TISDIC 2019

THE THIRD INTERNATIONAL CONFERENCE ON TRANSPORTATION INFRASTRUCTURE

AND SUSTAINABLE DEVELOPMENT - TISDIC 2019
THE THIRD INTERNATIONAL CONFERENCE ON TRANSPORTATION INFRASTRUCTURE
AND SUSTAINABLE DEVELOPMENT - TISDIC 2019

The third international

conference on transportation
infrastructure and sustainable
development - Tisdic 2019

CONSTRUCTION PUBLISHING HOUSE/ HANOI/ VIETNAM/ 2019

Publishing responsibility: Ngo Duc Vinh
Editors: Anand Puppala-Willam Young - Fabrizio Paolacci - Jie Han - Huynh Phuong Nam -Do Huu Dao
Editors Secrectary: Phan Hoang Nam – Nguyen Phuoc Quy Duy – Pham Ngoc Phuong
Readers: Nguyen Thi Binh, Tong Duc Phong
Typesetting: Typesetting department, Construction Publishing House.
Cover designer: Vu Binh Minh

Published by: The University of Danang – University of Science and Technology

and
Construction Publishing House, 37 Le Dai Hanh street, Hai Ba Trung district, Hanoi, Vietnam
Tel.: +84 4 39741416; Fax: +84 4 38215369; Email: [email protected]
Internet site: www.nxbxaydung.com.vn
All papers were reviewed by Scientific Committee
ISBN: 978-604-82-2893-4
No GPXB: 2706-2019/CXBIPH/01-224/XD 23th July, 2019
No QĐXB: 151-2019/QĐ-XBXD 30th July, 2019. Finished printing and depositing on August, 2019.
© 2019 The University of Danang – University of Science and Technology & Construction
Publishing House
Printed in Construction Publishing House, Add 10th Hoa Lu street, Hai Ba Trung district, Ha Noi.
The 3rd Int. Conf. on Transport Infrastructure & Sustainable Development (TISDIC 2019). Construction Publishing House. ISBN: 978-604-82-2893-4.

Contents

Preface XI
Organisation XIII

Keynote lectures
Keynote lecture 1: Non-Conventional stabilization to improve the stiffness of
expansive soil supporting pavement infrastructure in texas
Shi He, Surya S. C. Congress, Xinbao Yu, Sayantan Chakraborty, Anand J. Puppala 3
Keynote lecture 2: The role of the public in next generation asset management
William Young, Michael Shackleton, Alexa Delbosc, Sareh Balrololoom 11
Keynote lecture 3: Performance-based earthquake engineering analysis of
short-medium span steel-concrete composite bridges
Fabrizio Paolacci, Silvia Alessandri, Daniele Corritore, Phan Hoang Nam 25
Keynote lecture 4: Perspective on urban transportation strategy with BRT for
developing cities
Fumihiko Nakamura 38
Keynote lecture 5: Unified limit equilibrium design of geosynthetic-reinforced fill
walls and slopes
Jie Han 47

Session 1. Geotechnical Engineering and Underground Contrucsion

Invited lecture 1: Sustainable development of by-products based CLSM as
multi-purpose Geo-materials
Young-sang Kim, Do Tan Manh 57
Invited lecture 2: Problems and experiences of prediction of geo-risks in tunneling
in Vietnam
Nguyen Quang Phich, Nguyen Huy Vung, Trinh Van Thuong, Tran Nhat Minh 67
Invited lecture 3: Reliability of SPT numbers in predicting physical properties and
strength of cohesive soil
Truong Nhu Manh, Duong Hong Tham 75

V
The 3rd Int. Conf. on Transport Infrastructure & Sustainable Development (TISDIC 2019). Construction Publishing House. ISBN: 978-604-82-2893-4.

Invited lecture 4: Bidirectional load test for large bored piles of Cua Dai bridge
Do Huu Dao, Ha Hoang Viet Phuong, Nguyen Minh Hai, Anand J. Puppala,
Surya S. C. Congress, Young Sang Kim, Tran Quoc Thien 83
Bearing capacity analysis of bore pile in weathered rock in central Vietnam
Nguyen Chau Lan, Le Duc Tien, Bui Tien Thanh, Nguyen Ngoc Long 92
Finite difference method to determine lateral load bearing capacity of concrete pile
by taken into account soil and pile interaction
Dang Hong-Lam, Nguyen Duc-Hanh 98
Geocell wall - new solution for rock/ soil slope stabilization. application at a project
in the northern mountainous area in Vietnam
Le Thiet Trung, Pham Viet Anh 104
Determination of Unloading - reloading modulus and exponent parameters (m) for
hardening soil model from drained triaxial test of soft soil in Ho Chi Minh City
Ngo Duc Trung, Vo Phan, Tran Thi Thanh 112
Using discrete element method to simulate the settlement of soil saturated in
embankments due to static loading
Luong Nguyen Hoang Phuong, Tong Anh Tuan 119
Re-categorized soil layers and Performance Prediction of Deep Excavation in Kuala
Lumpur
Hai Khac Phan, Bin-Chen Benson Hsiung, Giang Hong Nguyen 123
Suffusion investigations under two flow directions
Tran Dinh Minh, Didier Marot, Fateh Bendahmane 134
Research of changing physical properties of embankment due to seepage effect by
the time
Nguyen Thanh Quang, Chau Truong Linh, Nguyen The Hung 144
Excavation behavior and the response of adjacent building at various stiffness using
Plaxis 2D and 3D
Huynh Quoc Thien, Tran Viet Thai, Nguyen Minh Tam, Lai Van Qui 156
Study of geopolymer synthesized from fly ash to improve soft ground
Tran Van Thu, Nguyen Van Chanh 170
A study on the main factors affecting the reinforcement corrosion in mechanically
stabilised earth walls and predict the service life of the wall
Chau Truong Linh, Nguyen Thu Ha, Vu Dinh Phung, Pham Van Ngoc 175
Cases study on the effect of down drag on bearing capacity of piles
Cao Van Hoa, Chau Truong Linh,… 183

VI
The 3rd Int. Conf. on Transport Infrastructure & Sustainable Development (TISDIC 2019). Construction Publishing House. ISBN: 978-604-82-2893-4.

A review on the application of deep cement mixing under lateral loads

Pham Van Ngoc, Erwin Oh, Dominic Ong, and Do Huu Dao 191
Study on load distribution between raft and piled in piled raft foundation in two
layers soil
Vo Van Dau, Vo Phan 200
Study on change of total stress applied on the pile head in the condition of
increasing water level
Vo Phan, Nguyen Tuan Phuong 207
Numerical simulation analysis of stress-deformation behavior of soil and
geosynthetic composite mass: a case study
Phan T.T Truc, M W. Gui, Pham Q. Thang 211
Analysis of theoretical equations for evaluation of bearing capacity of soft soils
Yongmin Kim, Yumi Chae, Jaehong Kim 220
Effect of montmorillonite content on the bearing capacity of soft ground improved
by cement deep mixing columns
Nguyen Ngoc Thang, Nguyen Minh Tam, Le Van Nam 227
Effects of mineral composition and physical - mechanical properties to calculate and
design ground improvement work in Quang Nam - Da Nang
Nguyen Thi Ngoc Yen, Bui Van Truong, Do Quang Thien, Nguyen Hoang Giang 234
Effect of unsaturation on the liquefaction of soil: case study of dense fine clean sand
Tran Khai Hoan, Saber Imanzadeh, Said Taibi, Hanène Souli,
Jean-Marie Fleureau, Mahdia Hattab, Luong Nguyen Hoang Phuong 244
Experimental evaluation of erosion resistance capacity of basic oxygen furnace
slag-treated clay
Nguyen Anh Dan, Gyeong-o Kang,Young-sang Kim, Takashi Tsuchida,
Young-hak Lee 252
Stabilization of weathered granite soil using ground bottom ash activated by various
industrial by-products
Tran Quoc Thien, Gyeong-O Kang, Do Huu Dao, Huynh Phuong Nam,
Young-Sang Kim 261
Bacterial enzyme induced calcite precipitation improving silty sand soil strength
Hoang Tung, James Alleman, Bora Cetin, Do Huyen 272
Study on the relationship between the plasticity index and the key parameters of the
vacuum consolidation process
Pham Quang Dong, Tran The Viet, Trinh Minh Thu 278

VII
The 3rd Int. Conf. on Transport Infrastructure & Sustainable Development (TISDIC 2019). Construction Publishing House. ISBN: 978-604-82-2893-4.

Determine fineness modulus of depositional coarse-grained soil in replacement for

alluvial sand in Thua Thien Hue coastal plain, Vietnam
Canh N.V., Do Quang Thien 285
Effect of sand on the engineering performance of cement asphalt mortar for fouled
ballast stabilization
Tri Ho Minh Le, Dae-Wook Park, Jung-Woo Seo, Koo-Sam Ji, Min-Ho Kang 293
Change detection of road surface from multi-temporal unmanned aerial vehicles
(UAV) images using machine learning
Nguyen Truong Linh, Dongyeob Han, Huh JungWon 302
Analysis of woven geotextiles for soft ground reinforcement with seaming strength
Han-Yong Jeon 307
Going underground - a solution for sustainable city in Viet Nam - aspect of
infrastructure
Vu Thi Thuy Giang, Do Nhu Trang 314

Session 2. Urban Transportation Engineering

Modal choice behavior of high school students in Danang city, Vietnam
Ong Truong Ich, Fumihiko Nakamura, Shinji Tanaka, Shino Miura,
And Ryo Ariyoshi 323
An improved version of cell-transmission model for air traffic flow
Tran Quang Khai, Kunihiko Hiraishi 335
How do lifestyle behaviours affect risky driving behaviors? a case study of
app-based motorcycle taxi riders in Danang, Vietnam
Vo Nhat Quang Dinh, Le Thanh Quang, Tran Thi Phuong Anh,
Nguyen-Phuoc Quy Duy 344

Development of safe bicycle infrastructure in Ukraine

Nataliia Kharytonova1,Valerii Vyrozhemskyi, Artem Bezuglyi 357
Pedestrian crossing behavior at signalized intersections in Ho Chi Minh city
Doan Duc, Nguyen Canh, Nguyen Son, Dang Thai, Truong Nhat 366
Observation of traffic conflicts using the traffic conflict technique: A case study in
Ho Chi Minh City
Doan Duc, Pham Kha 374
Development of trip generation model for Danang city by utilizing multiple linear
regression model
Cao Xuan My, Phan Cao Tho 379

VIII
The 3rd Int. Conf. on Transport Infrastructure & Sustainable Development (TISDIC 2019). Construction Publishing House. ISBN: 978-604-82-2893-4.

Advanced urban road intersection design: up-grade bike extra lane

Marwa S. El-bany 389
Development of bus models to evaluate road roughness and passenger ride comfort
Nguyen Van Teron, Patrick Swolana, Bernhard Lechner and YiikDiew Wong 397

Session 3. Structural and Bridge Engineering & Health Monitoring of

Transportation System
Invited lecture 1: Aerodynamic retrofitting of truss stiffening girder of long-span
suspension bridge
Hiroshi Katsuchi, Hitoshi Yamada, Shigeki Kusuhara and Taku Hanai 405
Invited lecture 2: Dynamic analysis of FGM conical shells surrounded by pasternak
elastic foundations
Nguyen Dong Anh, Le Quang Vinh, Nguyen Manh Cuong and Vu Quoc Hien 411
Advanced solutions to increase the efficiency in the use of pedestrian bridge
systems in Hanoi
Nguyen Minh Ngoc, Vu Anh 421
Flexural behaviour of precast concrete members connected by loop joints
Nguyen Quang Thinh, Hiroshi Mutsuyoshi, Takeshi Maki, Yosuke Ishihara 428
Seismic performance of an infilled moment-resisting steel frame during the 2016
central italy earthquake
Phan Hoang Nam, Fabrizio Paolacci, Hoang Phuong Hoa 437
GPU acceleration in OpenFOAM simulation of wind-over-top flow
Doan Sy Long, Hitoshi Yamada, Hiroshi Katsuchi 444
Dynamic responses of a functionally graded beam on nonlinear foundation under
moving harmonic loads by using different theories
Huynh Van Quang 454
Experimental study on ground ware propagations due to the impact of road vibration
rollers with a damping mass
Nguyen Lan, Hoang Phuong Hoa, Hoang Thai An 467
Damage detection in plate-like structures using improved modal strain energy
method
Le Cao Thanh, Ho Duc Duy 472
A study on behaviors of mass concrete due to heat of hydration and its pipe cooling
system
Vo Duy Hung, Nguyen Van My, Vo Van Viet 479

IX
The 3rd Int. Conf. on Transport Infrastructure & Sustainable Development (TISDIC 2019). Construction Publishing House. ISBN: 978-604-82-2893-4.

Session 4. Advanced Technologies and Materials in Transportation

System
Invited lecture 1: Weathering performance of new corrosion resistance steel at
coastal area
Kiyonobu Sugae, Hiroshi Iki, Hiroyuki Okada, Tomoya Tominaga,
Nguyen Thi Tuyet Trinh 489
Invited lecture 2: Investigating mechanical characteristic of ultra-high-performance
fiber-reinforced concrete at different strain rates
Ngo Tri Thuong, Tran Ngoc Thanh, Chau Truong Linh 497
Effect of fly ash and QuartzPowder on the properties of high-performance concrete
Nguyen Duc Vinh Quang, Aleksandrovaolga Vladimirovna,
Tkach Evgeniya Vladimirovna 503
The time-dependent effect over timeof water-to-cement ratio on the flow friction of
concrete pressure pipe
Nguyen Van-Vinh, Mai Chanh-Trung, Nguyen Hoang-Vinh, Nguyen The-Duong 516
Resistance of rubberized cement-based composites to acid attack
Phuong Ngoc Pham, Ahmed Toumi, Anaclet Turatsinze 522
Improving properties of OPC at high temperature by fly ash
Do Thi Phuong, Huynh Phuong Nam, Vu Minh Du 528
Estimation of distribution of chloride ions in cover concrete using electromagnetic
wave and X-ray fluorescence
Misaki Nakamura, Hitoshi Ito, and Toshiaki Mizobuchi 535
Recycled rock as coarse aggregate of shotcrete for tunnelling
Tran Thu-Hang, Nguyen Duc-Tran-Dong 543
Early predicting compressive strength of foamed concrete using an artificial
intelligence model
Ngo Ngoc-Tri, Nguyen Quang-Trung, Truong Ngoc-Son, Tran Thi-My-Dung 548
Author index 559

X
The 3rd Int. Conf. on Transport Infrastructure & Sustainable Development (TISDIC 2019). Construction Publishing House. ISBN: 978-604-82-2893-4.

Preface

The Third International Conference on Transportation Infrastructure and Sustainable Development

(TISDIC2019) was held in Da Nang city, Vietnam. The conference is to gather scientists, technology
experts, civil engineers, and related professionals who will provide a showcase for the latest
advancements and developments in construction and transportation infrastructures. The conference also
offers a forum to share and discuss the knowledge and the applications related to the strategies for
sustainable development of the transportation infrastructures
In the conference proceedings, all the papers have undergone the intensive review process which
performed by the scientific committee. Therefore, only accepted papers are included in the proceedings. The
proceedings comprised the selected papers from the subject areas of Urban Transportation Engineering;
Advanced Technologies and Materials in Transportation Construction; Geotechnical Engineering and
Underground Construction; Management & Operation of Transportation System; Health Monitoring of
Transportation System; Structural and Bridge Engineering; Structural Reliability and Risk Analysis.
We would like to acknowledge and give special appreciation to our keynote speakers for their valuable
contribution, our delegates for being with us and sharing their experiences, and our invitees for
participating in this conference. We would also like to extend my appreciation the scientific committee
for the devotion of their precious time and hard work to prepare for this conference. Finally, but not least,
we would like to thank all sponsors for their essential support for the conference.

Da Nang, August 2019

The Organizing Committee of TISDIC 2019
The 3rd Int. Conf. on Transport Infrastructure & Sustainable Development (TISDIC 2019). Construction Publishing House. ISBN: 978-604-82-2893-4.

ORGANIZATION

THE UNIVERSITY OF DANANG - UNIVERSITY OF SCIENCE

AND TECHNOLOGY

DANANG DEPARTMENT OF TRANSPORTATION

SUPPORT ORGANIZER
Faculty of Bridge and Road Engineering – Danang University of Science and Tecnology (DUST)
The University of Texas at Arlington (UTA)
The University of Kansas (UK)
Honam Regional Infrastructure Technology Management Center, Chonnam National University (CNU)
Vietnamese Social of Soil Mechanic and Geotechnical Engineering (VSSMGE)
Vietnam Bridge and Road Association (VBRA)

SPONSORS
Diamond : Truong Phuc Hai Construction Investment Limited Company
Gold : Hoa Binh Construction Group Joint Stock Company
Silver : ECC Construction Investment and Consulting Joint Stock Company
: Nippon Steel Corporation
: Tencate Geosynthetics Asia - Hung Viet Investment Construction J.S.C
: Beijing Samyon Instruments Co., Ltd
: Phan Vu Investment Joint Stock Company
: BK Engineering And Construction Company
Copper : Center of Research, Application And Consultancy of Foundation Engineering
: QES Construction Consultant Joint Stock Company
: Le Vu Joint Stock Company
: Dong Thang Trading Construction Limited Company
: Geotecnics Transfer and Application Joint Stock Company (GTA Vietnam)
: Tenox Asia Co. Ltd
: Center for Science Technology and Investement Consultancy
: Phu Sy Investment and building foundation Joint stock Company
: Bachy Soletanche Viet Nam
The 3rd Int. Conf. on Transport Infrastructure & Sustainable Development (TISDIC 2019). Construction Publishing House. ISBN: 978-604-82-2893-4.

Organization comittee
Chair : Assoc.Prof. Doan Quang Vinh Co-Chair: Dr. Dang Viet Dung
Vice chairs : Assoc.Prof. Le Thi Kim Oanh
: Dr. Huynh Phuong Nam
Member:
Mr. Le Van Trung
Assoc.Prof. Nguyen Dinh Lam
Assoc.Prof. Phan Cao Tho
Assoc.Prof. Hoang Phuong Hoa
Assoc.Prof. Chau Truong Linh
Dr. Doan Anh Tuan
Dr. Do Huu Dao
Dr. Cao Van Lam
Mr. Mai Trieu Quang
Mr. Nguyen Quang Trung
Mr. Nguyen Thanh Cuong

Scientific comittee
Co-Chairs
Prof. Anand Puppala Prof. Hitoshi Yamada
Prof. William Young Prof. Nguyen Dong Anh

Local committee members

Prof. Vu Dinh Phung Assoc.Prof. Pham Duy Hoa Assoc.Prof. Chau Truong Linh
Prof. Nguyen The Hung Assoc.Prof. Nguyen Quang Dao Assoc.Prof. Vo Ngoc Ha
Prof. Le Xuan Huynh Assoc.Prof. Nguyen Van Chanh Assoc.Prof. Le Thi Bich Thuy
Prof. Phan Quang Minh Assoc.Prof. Nguyen Xuan Hung Assoc.Prof. Tran Hoai Nam
Prof. Trinh Minh Thu Assoc.Prof. Nguyen Minh Tam Assoc.Prof. Nguyen Hong Hai
Prof. Do Nhu Trang Assoc.Prof. Nguyen Xuan Toan Dr. Phung Duc Long
Prof. Nguyen Quang Phich Assoc.Prof. Hoang Phuong Hoa Dr. Pham Van Long
Prof. Pham Duy Huu Assoc.Prof. Hoang Tung Dr. Le Thiet Trung
Assoc.Prof. Phan Cao Tho Assoc.Prof. Nguyen Duc Manh Dr. To Van Hung

International Committee members

Prof. Bengt H.Fellenius Prof. Balasubramaniam Prof. Marot Didier
Prof. Mounir Bouassida Prof. Watabe Yoichi Prof. Young-Sang Kim
Prof. James Alleman Prof. Hiroshi Katsuchi Prof. Ahmed Toumi
Prof. John W. Van de Lindt Prof. Yukihisa Kuriyama Prof. Yoon-Ho Cho
Prof. Fabrizio Paolacci Prof. Akira Hosoda Prof. Terri Norton
Prof. Masaharu Fukue Prof. Jean-Robert Courivaud Prof. Denzel M W Gui
Prof. Masaki Kitazume Prof. Erwin Oh Dr. Hitoshi Komatsu
Prof. Jie Han Prof. Anaclet Turatsinze Dr. NguyenMinh Hai
Prof. Zhusupbekov Askar Prof. Fumihiko Nakamura Dr. Do Quang Trung
Prof. Pham Huy Giao Prof. Chu Jian Dr. Phung Quoc Tri
The 3rd Int. Conf. on Transport Infrastructure & Sustainable Development (TISDIC 2019). Construction Publishing House. ISBN: 978-604-82-2893-4.

SECRETARIAT
Head Dr. Do Huu Dao
Members
Dr. Nguyen Duy Thao Dr. Tran Trung Viet Mr. Le Van Dung
Dr. Phan Hoang Nam Dr. Nguyen Phuoc Quy Duy Ms. Tran Thi Phuong Anh
Dr. Do Viet Hai Dr. Pham Ngoc Phuong Ms. Le Thi Ai Thi
Dr. Vo Duy Hung Mr. Le Thanh Hung Mr. Nguyen Tien Dung
The 3rd Int. Conf. on Transport Infrastructure & Sustainable Development (TISDIC 2019). Construction Publishing House. ISBN: 978-604-82-2893-4.

Keynote lecture 1: Non-conventional stabilization to

improve the stiffness of expansive soil supporting
pavement infrastructure in texas

Shi He1, Surya S. C. Congress1, Xinbao Yu1, Sayantan Chakraborty2, Anand J. Puppala1
1
Doctoral Student, 2Postdoctoral Fellow, 3Associate Professor, and 4Professor, Department of Civil
Engineering, The University of Texas at Arlington, Arlington, TX, U.S.A.

Corresponding Author: [email protected]

ABSTRACT: It is often difficult to find pavement subgrade soil with ideal characteristics. Stabilizers are
used to improve soil characteristics to match the pavement design requirements. Calcium-based
stabilizers such as lime, cement and fly ash are effective in enhancing the soil strength, stiffness and
workability in addition to controlling soil swell-shrinkage behaviour. However, manufacturing process of
these stabilizers emit a large amount of harmful gases that contribute to greenhouse effect. Furthermore,
in high-sulfate content soil, the calcium in traditional stabilizers reacts with alumina and sulphate to form
ettringite that contributes to swelling of the treated soils. Swell-shrink behaviours of expansive soil are
irreversible and contribute to the desiccation cracking that affect the performance of the pavement related
structures. Due to the disadvantages of conventional soil stabilizers, a new alternative liquid soil stabilizer
(LISS) is gradually gaining popularity within geotechnical fields. In this study, a comprehensive set of
laboratory tests including pH test, Atterberg limits test, standard compaction test, and resilient modulus
test were performed on soil samples treated with different dilution ratios of LISS. The obtained test
results indicated their effectiveness in treating expansive soils that are prevalent in the State of Texas.
LISS is an environmental-friendly soil stabilizer and results showed an increase the resilient modulus
after treatment. An universal model is used to estimate and compare the resilient modulus with the
laboratory measured values. More soils need to be tested to further understand the effectiveness on other
soil types.

Keywords: Liquid ionic soil stabilizer, resilient modulus, empirical constitutive model.

due to the volume change induced by seasonal

1. INTRODUCTION AND BACKGROUND fluctuation in moisture content (Petry and
Expansive soils are rich in arid and semi-arid Little, 2002; Puppala, 2016). Many previous
regions around the world including Australia, studies have addressed these issues and were
Canada, China, South Africa, India and the able to identify the potential factors responsible
United States (Sharma and Sivapullaiah, for this phenomenon. Montmorillonite mineral
2016).These soils exhibit excessive volume present in expansive soils is responsible for the
changes due to the ingress and egress of water erratic behavior of expansive soils subjected to
(Punthutaecha et al., 2006; Jones and Jefferson, alternate wetting and drying processes.
2012; Pranav, Rao, D and Rao, 2012). In the Researchers have come up with stabilization
State of Texas, many pavements laid over charts depending upon the amount of
expansive soils have exhibited severe distress montmorillonite (Pedarla, Chittoori and

3
Puppala, 2011). Few studies used sensors to water due to hydration reactions and crystal
measure volumetric moisture content and growth (Adams et al., 2008; Zhang et al., 2015;
matric suctions in the soils. They identified Puppala et al., 2019). Many previous studies
vegetation and climate present in the have reported adverse effects of traditional soil
surroundings influenced expansive soil stabilizers on the environment. Lime
movements and related pavement cracking production emits large amount of carbon
(Puppala et al., 2011). During rainy seasons, dioxide and rapidly increases the pH of the soil
expansive soils swell substantially due to the treated with it (Zhang et al., 2013). The
absorption of water, whereas, during the production of cement and fly ash consume
summer seasons, they lose moisture and shrink, plenty of energy and resources (Khadka,
often resulting in desiccation cracks. The Jayawickrama and Senadheera, 2018).
swell-shrink behaviors of expansive soil are Due to the reasons mentioned above, non-
irreversible. The recurring cracks on the traditional stabilizers, which do not have
ground surface, which do not seal perfectly calcium as one of the main constituents, are
after re-wetting, become a source of potential emerging in the market for treatment of sulfate-
problems (Jones and Jones, 1987; Jones and rich soils. The transportation cost of traditional
Jefferson, 2012; Puppala, Manosuthikij and soil stabilizer far outweigh the same for liquid
Chittoori, 2014). In the United State，around soil stabilizer (LISS) (Katz et al., 2001). These
60% of the homes constructed on the expansive non-traditional chemical stabilizers are
soils experienced minor damages while 10% available as concentrated liquids that can be
got seriously damaged (Holtz and Hart, 1978). diluted with water at the site, and then pressure
It is reported that the annual costs of repairing injected into the sublayers of soils or spread on
the homes and other infrastructures typically the surface of the problematic soil. Most non-
exceed $15 billion (Jones, 1992; Al-Rawas and traditional stabilizers are classified into three
Goosen, 2006; Shahbazi et al., 2017). groups: ionic, polymer, and enzyme (Rauch,
Katz and Liljestrand, 1993; Rauch et al., 2002).
Different treatment methods have been
Some of the products are used with secondary
developed to mitigate the harmful effects of
additives including surfactants, catalysts, and
expansive soils on the infrastructure durability.
ultraviolet inhibitors (Tingle et al., 2007).
Deep soil mixing have also been effective in
Despite of its low transportation cost and
mitigating the harmful effects of expansive soil
effectiveness in treating sulfate-rich soils,
(Madhyannapu and Puppala, 2014). Most
many engineers and practitioners do not
often, conventional soil stabilizers such as
consider to non-traditional stabilizers to treat
lime, Portland cement, and fly ash are utilized
problematic expansive soils. This can be
to improve workability, control swelling and
attributed to a few factors: insufficient
shrinkage behavior, and increase the strength
publications, lack of proprietary patent and
and stiffness properties of the foundation soil
misleading information related to the products
for light structures and subgrades of pavements
(Rauch, Katz and Liljestrand, 1993).
(Katz et al., 2001; Rauch et al., 2002; Zhang et
al., 2013; Puppala, Talluri and Chittoori, 2014). In this paper, LISS composed of sulfuric
Not all soils have been effectively treated by acid, phosphoric acid, citric acid and
conventional stabilizers. Sulfate-rich expansive surfactant is used as an expansive soil
soils treated by traditional soil stabilizers stabilizer. A comprehensive set of laboratory
containing calcium led to excessive volume tests such as Atterberg limits test, pH test,
change caused by ettringite formed due to the standard compaction test, and resilient
reaction of calcium with sulfate and alumina modulus test are conducted to evaluate the
present in the soil (Rauch, Katz and effectiveness of soil treatment using LISS.
Liljestrand, 1993; Rauch et al., 2002; Puppala These results have indicated an improvement
et al., 2018). Ettringite is prone to expand in the stiffness of the treated soils compared to
significantly when it comes in contact with those of untreated soils.

4
2. EXPERIMENTAL PROGRAM No. 40 sieve with various dosages of LISS
solutions. Then soil samples were kept in a
The test soil was sampled from Caddo Mills in moisture room to allow chemical reactions
Dallas region, Texas, USA. All the bulk soils between liquid chemical and soil specimens
were sampled at 0.9 m (3-ft) below the ground during the curing time. For the liquid limit
surface. A comprehensive set of laboratory tests (LL) test, distilled water was added until the
were conducted to evaluate the behaviour of the soil mixture reached the liquid state. After
soil treated with LISS. The dilution ratio adopted the LL test, the wet samples were air dried to
in the field is to mix 30 litres (8 gals) of the reach the plastic limit (PL) state. An E-180
chemical concentrate to 340 g of surfactant with PL rolling device as shown in Figure 2,
22,712 litres (6,000 gals) of water. In order to which was initially developed by Texas
simulate the field dilution ratio in the laboratory, it Department of Transportation (TxDOT), was
was carried out by mixing 5 ml liquid chemical used to measure the PL values. The major
concentrate and 0.057-g surfactant with 3.8 litres benefit of this device is that it can avoid
(1 gal) of water. Additionally, two other treatment human errors. The gap between the top roller
ratios were also selected to assess the effect of plate and the bottom roller plate is exactly
various LISS dosages shown in Table 1. 0.125 in. to meet the requirements of ASTM
D4318–17 (He, et al., 2018).
According to ASTM D698, standard
Proctor test was conducted to determine the
relationship between optimum moisture
content (OMC) and maximum dry density
(MDD) for treated specimens (Das, 2002).
Treated soil samples were prepared by
mixing 1.8 kg of pulverized dry soils with
different dilution ratios of LISS in separate
containers. These moist soil specimens were
kept in the moisture room overnight in order
to provide sufficient time for moisture
equilibration and chemical reaction between
the liquid chemicals and the soil before the
Figure 1. A deep injection rake for pressurized
injection of LISS (Source: Proctor test (He, Yu, Gautam and Hoyos,
http://www.prochemtex.com/Services.html) 2018). The OMC values obtained were used
in the liquid limit tests.
Table 1. Three LISS ratios for soil treatment Available literature and brochures of LISS
LISS 2nd manufacturing companies suggest that one of
1st Ratio 3rd Ratio the biggest advantages of LISS is its
Proportion Ratio
environmental-friendly nature. The pH test
Chemical results may provide some more understanding
Concentrate 5 5 10
on the long-term impact of the LISS on the
(ml) environment and its suitability to be used as a
Surfactant (g) 0.057 0.057 0.114 stabilizer. The pH tests were carried out
according to ASTM D4972. Approximately
Water (gallon) 1 2 1
300 grams of soil was completely submerged
Atterberg limit tests were carried out by into the LISS for curing periods of 1, 7, 14, 21
following the procedures outlined in ASTM and 28 days. The pH of the supernatant liquid
D4318-17 method. Each treated soil was monitored at the end of each curing period
specimen was prepared by mixing 100 g of to study the variation of pH due to progressive
pulverized dry soil samples passing through a chemical reactions.

5
 Resilient moduli tests were conducted the soil swell potential (Chen, 2012). Table 2
before and after the treatment of soil specimen shows the Atterberg limits of the Caddo Mills
using the equipment shown in Figure 3 and soil treated with various dilution ratios of
following the guidelines noted in AASHTO LISS. Although plasticity index (PI) reduced
T307. The test progress involved 15 stress after LISS treatment, the difference among PI
sequences with each of them having a unique values of untreated and treated is negligible.
combination of confining pressure and deviator To some extent, LISS is not very effective to
stress. During the test, each loading cycle decrease Atterberg limits, which is in line
comprised of 0.1 s of load pulse followed by with the findings of the previous studies
0.9 s of rest period. Although each sequence (Rauch, Katz and Liljestrand, 1993; Rauch et
includes 100 cycles, the mean value of resilient al., 2002)
modulus is calculated from the slope of
reloading curve for the last five cycles of each Table 2. Atterberg limits soil for Caddo Mills soil
loading sequence (Bhuvaneshwari, Robinson 1st 2nd 3rd
and Gandhi, 2018). Control
Ratio Ratio Ratio
Universal model that represents a range of
soils varying from cohesive to non-cohesive is Liquid limit (%) 76 79 72 75
introduced to analyze the relationship among Plastic limit (%) 18 29 15 22
resilient modulus, bulk stress, shear stress and Plastic index (%) 58 50 57 53
elastic moduli of soil (George, 2004). This model
assists the engineer to design the pavements in a 3.2 pH test
realistic manner (Han and Vanapalli, 2015). The primary purpose of the pH tests is to gain
some understanding of the extent of chemical
reaction between soil particles and the LISS
components. In this study, pH value of soil
treated by highest LISS concentration, i.e.
third dilution ratio, is compared with that of
the control sample. Figure 4a depicts the
Figure 2. The E-180 plastic limit rolling device
variation in pH of the treated soil in one
(He, Yu, Banerjee, et al., 2018)
month. Initially, the pH value of solution was
nearly 3.2 and increased rapidly during the
initial five hours of curing time, as shown in
Figure 4b. Later it gradually increased and
stabilized around 7.6 after a month of curing
time. A variation in pH with curing time
indicated progressive chemical reaction and
formation of new reaction products.
Furthermore, this test demonstrated that LISS
is environment-friendly material compared to
traditional stabilizers.
Figure 3. Resilient modulus test equipment (He,
Yu, Gautam, Puppala, et al., 2018) 3.3 Standard proctor compaction test
Figure 5 reveals the standard compaction curve
3. RESULTS AND DISCUSSION for Caddo Mills soil before and after treatment
3.1 Atterberg Limit Tests with LISS. For treated soil, the chemical
reaction between soil particles and LISS is
The liquid limit (LL) and plastic limit (PL) likely to occur before compaction, causing soil
tests reflect the effect of moisture content on particle flocculation. The MDD of soils
the mechanical properties of soil to estimate decreased after treatment.

6
 attributed to the need for water to react with
chemical components of LISS.

(a)

Figure 5. Standard compaction curve for Caddo

Mills soil

3.4 Resilient modulus test

Table 3 shows the resilient modulus of Caddo
Mills soil before and after treatment. Due to
stress softening effect of soil, the resilient
modulus (MR) value decreased with
increasing deviator stress (Rahman and
Tarefder, 2015). Also, MR value was found to
(b)
be influenced by OMC. Although LISS content
Figure 4. The pH test of sample treated by third ratio: in third ratio is more than first ratio, MR value
(a) pH variation over initial month of curing of first ratio is found to be larger than that
(b) pH variation in first 50 hours of curing. obtained corresponding to the third ratio. The
reduction of MR value can be attributed to the
Generally, for chemical treatment of soil
fact that OMC of soil treated by third ratio is
after compaction, the MDD of soils decreases
nearly 1.5% more than that of soil treated by
and the OMC increases as the specific gravity
first ratio, which can be observed in Figure 5.
of chemical is lower than the expansive soil
In sum, MR value of treated soil increased in
(Sweeney, Wong and Fredlund, 1988). The
comparison with untreated samples.
increase in OMC after treatment can be
Table 3. Resilient Modulus for Caddo Mills soil before and after treatment
Mr (Mpa) Mr (Mpa)
Confining Pressure Deviator Stress
Mr (MPa) Second Third Second
(kPa) (kPa) First Ratio First Ratio Third Ratio
Ratio Ratio Ratio
41.4 14 44.5 91.1 41.7 53.8 86.9 68 63.7
41.4 28 45.6 85.5 38.1 51.5 79.7 63.9 60.4
41.4 41 43.7 81.2 34.4 49 76.9 60.4 57.1
41.4 55 41.6 77.7 31.5 46.7 76 57.8 54.7
41.4 69 39.4 74.6 29.2 44.6 72.3 54.4 51.8
27.6 14 44.7 86.2 38.1 50.7 81.1 64.9 61.3
27.6 28 44 82.6 34.7 48.7 76.9 61.7 58.1
27.6 41 42.3 78.3 31.7 46.5 74.1 58.6 54.9
27.6 55 40.3 75.8 29.6 44.9 72 56.3 52.9
27.6 69 38.5 73.7 28.2 43.3 69.7 54 50.9
13.8 14 40.8 81.4 34.1 47.9 74.8 63.1 58.6
13.8 28 40.8 78.8 31.6 46.1 71.6 60 55.6
13.8 41 39.4 75.8 29.5 44.2 69.5 57.5 53.1
13.8 55 37.9 73.6 27.8 43 67.8 55.2 51.3
13.8 69 36.7 71.8 26.7 41.7 67.3 52.9 49.5

7
3.4 Universal Model for Predicting MR Results Soil Treated By First Ratio (28 Days)
90000

The universal model was developed on the

Universal Model MR (kPa)

85000

basis of bulk stress model created by Uzan in 80000

1985. In this model, the octahedral stress 75000

replaced deviator stress, which can better 70000

explain the stress state of the material (George, 65000

2004). Equation is listed below.

65900 70900 75900 80900 85900 90900
Laboratory MR (kPa)
k k3
      
2

M r  k1 Pa    oct   1
 Pa   Pa
Soil Treated By Second Ratio (7 Days)
  43000

41000

In this equation, k1, k2 and k3 are

Universal Model MR (kPa)

39000

coefficients relying on the material type and 37000

35000

physical properties. Pa is the atmospheric 33000

pressure,  is the bulk stress,  oct is the 31000

29000

octahedral shear stress, k1 is proportional to 27000

Young’s modulus and k2 indicates the soil

25000
26000 28000 30000 32000 34000 36000 38000 40000 42000 44000
Laboratory MR (kPa)

stiffness or hardening of material

Figure 6 reflects the regression analysis of Soil Treated By Second Ratio (28 Days)
resilient modulus of soils before and after 80000

chemical treatment. Table 4 summarizes all the

Universal Model MR (kPa)

75000

regression coefficients and coefficients of 70000

determination values. From the results, k1 and k2

65000

60000

are both positive whereas k3 is negative as the 55000

increasing shear stress softens the material. In 50000

addition, for most cases, k1 and k2 values of 52000 54000 56000 58000 60000 62000
Laboratory MR (kPa)
64000 66000 68000 70000

treated soils are greater than those of untreated

soil specimens because of higher strength and Soil Treated By Third Ratio (7 Days)

stiffness of soils after treatment. However, k3 56000

Unviersity Model ]MR (kPa)

54000
value before and after soil treatment are similar 52000

and this suggests that it is necessary to study the 50000

effects of LISS on shear strength of treated soil. 48000

46000

44000
Control Soil 42000
48500
40000
46500 40500 42500 44500 46500 48500 50500 52500 54500
Universal Model MR (kPa)

Laboratory MR (kPa)
44500

42500

40500
4. CONCLUSIONS
38500

36500
36500 37500 38500 39500 40500 41500 42500 43500 44500 45500 46500 The conclusions obtained from the laboratory
Laboratory MR (kPa)
test before and after treatment are listed as
Soil Treated By First Ratio (7 Days)
follows:
94000
1. When soil was mixed with LISS, the pH value
dropped to around 3.2. However, after one
Universal Model MR (kPa)

89000

84000 month, the pH value kept stable around 7.8,

79000 which showed that LISS was an environment-
74000 friendly alternative soil treatment.
69000
70000 75000 80000 85000 90000 95000
2. After treatment with LISS, MDD of soil
Laboratory MR (kPa) decreased while OMC of soil increased.

8
3. The resilient moduli of soil after treatment ACKNOWLEDGEMENTS:
increased compared with control sample.
This indicates that the LISS has increased Authors would like to acknowledge TX
moduli of the treated soil and can provide Prochemical for supporting the research
stronger support to pavements. support.

Soil Treated By Third Ratio(28 days) 5. REFERENCES

64500

Adams, A. G. et al. (2008) ‘Sulfate Induced Heave

Universal Model MR (kPa)

62500

60500 in Oklahoma Soils due to Lime Stabilization’,

58500
pp. 444–451.
56500
Al-Rawas, A. A. and Goosen, M. F. A. (2006)
54500

52500
Expansive soils: recent advances in
50500 characterization and treatment. Taylor & Francis.
48500 Bhuvaneshwari, S., Robinson, R. G. and Gandhi,
49000 51000 53000 55000 57000 59000 61000 63000 65000
Laboratory MR (kPa) S. R. (2018) ‘Resilient Modulus of Lime
Treated Expansive Soil’, Geotechnical and
Figure 5. Resilient modulus of Caddo Mills soil Geological Engineering. Springer, pp. 1–11.
before and after treatment. Chen, F. H. (2012) Foundations on expansive soils.
Elsevier.
Table 4. Regression coefficients of resilient Das, B. (2002) ‘Soil Mechanics Laboratory
modulus and coefficient of determination Manual’.
before and after treatment George, K. P. (2004) Prediction of resilient
k1 k2 k3 R2 modulus from soil index properties. University
of Mississippi.
Control 469.54 0.1278 -0.9711 0.925 Han, Z. and Vanapalli, S. K. (2015) ‘Model for
st
1 ratio predicting resilient modulus of unsaturated
913.86 0.099 -1.043 0.987 subgrade soil using soil-water characteristic
7 days
curve’, Canadian Geotechnical Journal. NRC
1st ratio
854.04 0.150 -1.032 0.970 Research Press, 52(10), pp. 1605–1619.
28 days He, S., Yu, X., Banerjee, A., et al. (2018)
2nd ‘Expansive Soil Treatment with Ionic Soil
Ratio 429.42 0.205 -2.020 0.989 Stabilizer’, Transportation Research Record:
7 days Journal of the Transportation Research Board.
He, S., Yu, X., Gautam, S. and Hoyos, L. R. (2018)
2nd
‘Influence of Ionic Soil Stabilizer (ISS) Dosage
Ratio 797.40 0.0711 -1.130 0.992
on the Stabilization Effectiveness of Expansive
28 days Soils’, in IFCEE 2018, pp. 103–112.
3rd He, S., Yu, X., Gautam, S., Puppala, A. J., et al.
Ratio 544.09 0.1286 -1.1194 0.994 (2018) ‘Resilient Modulus of Liquid
7 days Chemical-Treated Expansive Soils’, in
3rd GeoShanghai International Conference.
Ratio Springer, pp. 114–120.
651.14 0.0864 -1.1561 0.995
Holtz, W. G. and Hart, S. S. (1978) Home
28 days construction on shrinking and swelling soils.
Colorado Geological Survey.
Also, the predicted resilient modulus value Jones, D. E. and Jones, K. A. (1987) ‘Treating
from universal model is comparable with the expansive soils’, Civil Engineering—ASCE.
measured laboratory value. ASCE, 57(8), pp. 62–65.
In summary, this paper showed the effects Jones, L. (1992) ‘Expansive Soils Where Are They
of LISS treatment on Caddo Mills soil and Found ?’, i, pp. 314–325.
identified that the first dilution ratio is effective Jones, L. D. and Jefferson, I. F. (2012) Expansive
in treating the expansive sulphate-rich soils. soils. ICE Publishing.

9
Katz, L. et al. (2001) ‘Mechanisms of soil Puppala, A. J. et al. (2019) ‘Sulfate-Heaving
stabilization with liquid ionic stabilizer’, Studies on Chemically Treated Sulfate-Rich
Transportation Research Record: Journal of Geomaterials’, Journal of Materials in Civil
the Transportation Research Board. Engineering. American Society of Civil
Transportation Research Board of the National Engineers, 31(6), p. 4019076.
Academies, (1757), pp. 50–57. Puppala, A. J., Manosuthikij, T. and Chittoori, B.
Khadka, S. D., Jayawickrama, P. W. and C. S. (2014) ‘Swell and shrinkage strain
Senadheera, S. (2018) Strength and Shrink– prediction models for expansive clays’,
Engineering Geology. Elsevier, 168, pp. 1–8.
Swell Behavior of Highly Plastic Clay Treated
Puppala, A. J., Talluri, N. and Chittoori, B. C. S.
with Geopolymer.
(2014) ‘Calcium-based stabiliser treatment of
Madhyannapu, R. S. and Puppala, A. J. (2014)
sulfate-bearing soils’, Proceedings of the
‘Design and construction guidelines for deep Institution of Civil Engineers-Ground
soil mixing to stabilize expansive soils’, Journal Improvement. Thomas Telford Ltd, 167(3),
of Geotechnical and Geoenvironmental pp. 162–172.
Engineering. American Society of Civil Rahman, M. T. and Tarefder, R. A. (2015)
Engineers, 140(9), p. 4014051. ‘Assessment of Molding Moisture and Suction
Pedarla, A., Chittoori, S. and Puppala, A. J. (2011) on Resilient Modulus of Lime Stabilized
‘Influence of mineralogy and plasticity index Clayey Subgrade Soils’. ASTM International.
on the stabilization effectiveness of expansive Rauch, A. et al. (2002) ‘Measured effects of liquid
clays’, Transportation Research Record. soil stabilizers on engineering properties of
SAGE Publications Sage CA: Los Angeles, clay’, Transportation Research Record:
CA, 2212(1), pp. 91–99. Journal of the Transportation Research Board.
Petry, T. M. and Little, D. N. (2002) ‘Review of Transportation Research Board of the National
Stabilization of Clays and Expansive Soils in Academies, (1787), pp. 33–41.
Pavements and Lightly Loaded Structures— Rauch, A. F., Katz, L. E. and Liljestrand, H. M.
History, Practice, and Future’, Journal of Materials (1993) ‘AN analysis of the mechanisms and
in Civil Engineering, 14(6), pp. 447–460. doi: efficacy of three liquid chemical soil
10.1061/(ASCE)0899-1561(2002)14:6(447). stabilizers: VOLUME’, Work, 1.
Shahbazi, M. et al. (2017) ‘Optimization of carpet
Pranav, P. R. ., Rao, D, K. and Rao, G. V. V. R.
waste fibers and steel slag particles to
(2012) ‘A Laboratory Study on the Affect of
reinforce expansive soil using response
Rice Husk Ash & Lime on the Properties of surface methodology’, Applied Clay Science.
Marine Clay’, International Journal of Elsevier, 142, pp. 185–192.
Engineering and Innovative Technology, 2(1), Sharma, A. K. and Sivapullaiah, P. V (2016) ‘Ground
pp. 345–353. granulated blast furnace slag amended fly ash as
Punthutaecha, K. et al. (2006) ‘Volume change an expansive soil stabilizer’, Soils and
behaviors of expansive soils stabilized with Foundations. Elsevier, 56(2), pp. 205–212.
recycled ashes and fibers’, Journal of materials Sweeney, D. A., Wong, D. K. H. and Fredlund, D.
in Civil Engineering. American Society of Civil G. (1988) Effect of lime on a highly plastic
Engineers, 18(2), pp. 295–306. clay with special emphasis on aging.
Puppala, A. J. et al. (2011) ‘Threshold moisture Tingle, J. et al. (2007) ‘Stabilization mechanisms
content and matric suction potentials in of nontraditional additives’, Transportation
expansive clays prior to initiation of cracking in Research Record: Journal of the
pavements’, Canadian Geotechnical Journal. Transportation Research Board.
NRC Research Press, 48(4), pp. 519–531. Transportation Research Board of the National
Puppala, A. J. (2016) ‘Advances in ground Academies, (1989), pp. 59–67.
modification with chemical additives: From Zhang, M. et al. (2013) ‘Experimental feasibility
theory to practice’, Transportation study of geopolymer as the next-generation
soil stabilizer’, Construction and Building
Geotechnics. Elsevier, 9, pp. 123–138.
Materials. Elsevier, 47, pp. 1468–1478.
Puppala, A. J. et al. (2018) ‘Ettringite induced
Zhang, M. et al. (2015) ‘Calcium-free geopolymer
heaving in stabilized high sulfate soils’,
as a stabilizer for sulfate-rich soils’, Applied
Innovative Infrastructure Solutions, 3(1), p.
Clay Science. Elsevier, 108, pp. 199–207.
72. doi: 10.1007/s41062-018-0179-7.

10
The 3rd Int. Conf. on Transport Infrastructure & Sustainable Development (TISDIC 2019). Construction Publishing House. ISBN: 978-604-82-2893-4.

Keynote lecture 2: The role of the public in next generation

asset management

William Young1, Michael Shackleton2, Alexa Delbosc3, Sareh Balrololoom2

134
Department of Civil Engineering, Monash University, Melbourne Australia
2
Australian Road Research Board, Melbourne, Australia.

Corresponding Author: William Young. Address correspondence to: [email protected]

ABSTRACT: Transport infrastructure and sustainable development requires a co-ordinated, efficient and
well informed asset management system as its basis. This paper explores future developments in road
infrastructure and the role of the public’s view of these developments in next generation asset
management system. It initially reviews ideas on the future of roads. The review leads to a series of
qualitative questions of the public’s view of the future of roads. These were: Roads and their use will
remain largely the same; On major roads, cars and trucks will be separated from cyclists and pedestrians;
Parking on major roads will not be permitted; Local roads will give priority to pedestrians and cyclists; In
the future, roads will be much safer; The physical quality of roads and their surfaces will improve;
Vehicle travel will all be automated; Road travel will be more environmentally sustainable. People think
most of the factors investigated are changes that will, and should happen; and People will pay a road
charge for each trip. Paying a road toll stand out as the only one that people think will happen, but do not
want to happen. In terms of automated / driverless they're on the fence, they think it probably will happen
but they're in the middle of thinking it should/should not happen. In terms of what will happen in the
future the removal of parking on the major roads and separation of activities on major roads were overall
ranked the highest.
Keywords: Asset management, human view, future developments, roads.

1. INTRODUCTION This paper looks at the role of people’s view

of the future of roads in order to gain insights
Transport infrastructure and sustainable
development requires a co-ordinated, efficient into improved asset management systems. It
and well informed asset management system as explore what will happen and what should
its basis. Asset management refers to a happen as it points to the role of government.
systematic approach to the governance and The role of road assets in active transport
realization of value from the things that a modes (eg pedestrians, bicyclists) and their
group or entity is responsible for, over their impact on sustainable development is
whole life cycles. Asset Management can refer investigated along with driver behavior.
to shaping the future interfaces between the It initially reviews potential future directions
human, built, and natural environments raising questions on what changes will take
through collaborative and evidence-based place. It then uses a questionnaire to explore
decision processes. these directions for the general public. The

11
paper concludes that public input to the future environment, resilience and strategic asset,
of roads is an important ingredient in next addressing the theme that each new and
generation asset management systems. rehabilitated road will have a marginally smaller
impact on the environment in the future. RQ
2. A REVIEW OF THE FUTURE OF ROADS (2.1), asks if the impact of environmental change
on the sustainability and resilience of the road
Roads are presently the fundamental transport network is an area of considerable concern? RQ
infrastructure asset. They carry a lion’s share (2.2) looks at how do we ensure roads contribute
of today’s transport. They provide people with to a sustainable future? RQ (2.3), roads as a
day to day access to work, education, play and national strategic asset? RQ (2.4) explores are
retail activity. They facilitate the logistics improved methods of road construction and
chain through allowing movement of goods maintenance required to reduce the impact of
and resources from their extraction to roads on sustainability and the environment? RQ
manufacture and public outlets. They provide (2.5) investigate research required into
the framework for car, truck, bus, tram, light improving methods of recycling and reuse of
rail, motorbike, bicycle, and pedestrian resources to increase the sustainability of
mobility. Roads future is definite and there is a infrastructure? RQ (2.6) explores is research into
clear need for research to ensure this asset is alternatives to petroleum for road construction?
properly managed. Sub-Section 3 explores the relationship
This section outlines the results of a
between roads and society. Travel is often
literature review 1, 2, 3, 4, 5, 6, 7 the future of
considered as a derived demand with road
road assets. The future was generally seen as
providing free movement. Roads have been a
a time horizon of 30 years. It reports on 10
areas of interest and discusses each looking at community service obligation. The activity
potential research questions (RQ). The undertaken at the beginning and end of the trip
section explores two areas of asset has been the reason for travel. These activities
management: Usage and Enablers. are taken within and social or activity
framework. The questions raised are: RQ
2.1 Usage (3.1), should the unit of analysis be people and
good or the vehicles using the road system?
Sub-Section 1 introduces roads and their future RQ (3.2), is there a need for the integration of
addressing the theme that roads play a large part road systems into local and activity centers to
in the mobility of human communities. It improve urban livability and economic
addresses the Research Question (1.1), is there a competitiveness? RQ (3.3), is the integration
need to continually update our view on the role
of the mobility aspect of roads into the social
and future of roads? Road infrastructure, in the
and economic dimensions of urban area an area
future, will face increasing pressures and be
of considerable interest and in need of intense
impacted by a range of factors. Some of these
study? RQ (3.4) addresses how do we integrate
factors include: changing people’s expectations
and opportunities; people’s desire for mobility the operation of the road transport network and
that outstrip the ability to provide new their integration with the amenity and livability
infrastructure development; increased of cities? RQ (3.4), how important is resilience
population; an environment of rapidly changing in the people, movement, place interaction; RQ
technologies and information; changes in (3.5), should the designing roads provide for
weather patterns due to global factors; mobility/accessibility/amenity be complemented
constraints in capacity both land; and capital by the safety on these facilities? RQ (3.6), is
and an ever changing improvement in materials, the transfer of people from the road system to
construction and maintenance opportunities. their activity of choice in need of research? RQ
Sub-Section 2 introduces roads and (3.7), is the social value of roads in rural areas
sustainability. It focuses on the dimensions of of interest and in need of exploration?

12
2.2 Enablers investigation into a more detailed consideration
of safety in road design and the integration of
Sub-Section 4 raises the questions the
various transport modes required? RQ (6.3),
financial, business and economic models for
does the character of the road reserve and its
roads into the future is likely to change. It is
relationship to the safe system requires
likely that in the future roads will be funded
continuous assessment and improvement? RQ
and operated through a complex mix of local
(6.4), is there a need to balance the road design
private companies, government bodies and
for mobility and safety within the context of
international organizations. This will involve
the livability of cities and rural regions?
private ownership, government ownership,
Sub-Section 7 looks at the construction,
consortiums and partnerships: RQ (4.1), is the
rehabilitation and maintenance of the road
economic benefits of the road system being
cross-section, pavement and running surface
considered correctly? RQ (4.2), is it adequate
assets. This is an area of considerable
to look at the “Fitness for Purpose” of roads development in materials and smart
alone to determine road infrastructure and use infrastructure and next generation asset
priorities or should a total systems approach be management. Traditionally the road provide a
developed? RQ (4.3), is the regulation and reserve or way for people, goods, livestock,
licensing of the road system and the people vehicles, power-lines, water pipes, electric
who use it in need of exploration? RQ (4.4), is conduit, internet cables etc. to facilitate
the pricing of the road system and the people movement. Roads have been viewed as
who use it in need of exploration? RQ (4.5), simple technology which passively undertakes
how is the business model for road their role. Essentially they are viewed as a
infrastructure likely to change in the future? support infrastructure and a wearing surface.
Sub-Section 5 explores planning and policy The messages between roads, their users and
in the roads industry. The planning of roads is the broad environment use the road designer as
undertaken by government, transport and road an intermediary. This is changing. The
authorities. Because of the strategic and social research questions are: RQ (7.1), is the use of
importance of roads this is unlikely to change. the road cross-section in need of
However, the increasing importance of road reconsideration? RQ (7.2), is the materials
infrastructure and competition for funding used in pavement an area of considerable
within government is likely to result in the research opportunity? RQ (7.3), is research
increased roll of partnerships in the planning into the management and maintenance of the
and developments of roads. Some research road pavement in need of exploration? RQ
questions raised are: RQ (5.1), what are the (7.4), is research into pavements and the role of
major planning processes and policies information technology an area of interest? RQ
associated with environmental, economic and (7.5) what will be the loading on our
social trends? RQ (5.2), how equitable will pavements of the future? RQ (7.6), will
roads be in the future? improved materials be used to separate
Sub-Section 6 looks at design of vulnerable road users in high density and speed
infrastructure within the road reserve. Its main road environments.
theme for the future is that roads will be safer Sub-Section 8 focuses on physical road
in the future. Roads have been designed to infrastructure above and below the land
provide a high level of mobility. There needs surface, expanding previous discussion about
to be a balance between mobility and safety. roads within their cross-section. Traditionally
They also service a variety travel modes, the road provide a reserve or way for people,
consideration of the users of these modes is goods to facilitate movement. They are
required. The research questions raised in this generally tied to the land surface of the planet.
area are: RQ (6.1): are the guides to the design Bridges and tunnels provide the opportunity to
of roads in need of review? RQ (6.2) is the make more efficient use of the land surface.

13
Similarly Drone technology may in the future environments, and smart urban systems. RQ
provide an opportunity to move people and (9.7), is the use of new technology to provide
goods. The use of the air space above and data for the efficient planning, use and
below the land surface provides an opportunity operation of road system in need of research?
to increase the capacity of the road network. Sub Section 10 investigates information on
This section raises the research questions: RQ roads with particular concern about next
(8.1), how can the physical road space and generation data collection, editing, analysis,
areas above and below it be used to create storage and presentation. Reliable data is
efficiencies in the provision of physical required to make any decision on road asset
infrastructure? RQ (8.2), is there a role for the management. The methods of collecting
airspace above roads to be used for storing and analyzing this data are developing
infrastructure provision and vehicle rapidly. Road data collection in the future will
movement? RQ (8.3), is the materials used in be automated and increase in volume and
bridge and tunnel construction an area of accuracy. The research questions explored in
considerable research opportunity. RQ (8.4), this section are: RQ (10.1), what will be the
is the increases intelligence incorporated into approaches to collect, analyze, store and
bridge design an important area of research? present data in the future? RQ (10.2), how is
Sub-Section 9 explores developments in the enormous amount of data that will be
technology and intelligent road systems; the available in the future to be used? Sub Section
human technology interface. Information 10 has highlighted the need to look at new
technology is changing the driver, road, and generation data. The rapid development of
road environment interaction. This change and technology is providing data which can often
development is growing at a rapid rate and is be used by others. Road data collection in the
likely to have far reaching social, economic future will be automated and increase in
and environmental impacts. This section looks volume and accuracy.
at the changes taking place in information The literature review looked at the some
technology and their relationship to travel and potential developments in the asset
roads. That is: how can information management of roads. It pointed to a number
technology be used to develop efficient and of key questions. These can be summarised as:
sustainable road systems? Sub-Section 9 How will roads be used? What will the
relates to the following RQ’s: RQ (9.1) asks community want from roads in the future?
do traffic signal control systems have a role in Are the mobility, social and economic
future road and motorway systems? RQ (9.2), functions of roads sustainable and can they be
is the rapid change in automobile technological successfully integrated? Will future road
and its integration into the road system in need transport reduce social inequity or widen the
of further research? RQ (9.3), the various divide? Can we build resilience into roads to
modes of travel provide a richness in the use of deal with increasing congestion and extreme
roads. All of these modes are not developing events and challenges?
intelligence at the same rate. Is the intelligence What will the road network look like
of roads to cater for these users important and and how will it be funded? Can current
requiring investigation? RQ (9.4), will the models for financing and managing the road
roads of the future only develop in relation to network deliver the network of the future? In
land based activities? RQ (9.5), is the times of an uncertain connectivity future, how
increases intelligence incorporated into bridge should infrastructure design methods,
design an important area of research? RQ standards, design life cycle and materials
(9.6), is there a need for research at the overall change? Will the future network provide
environmental systems level to ensure the mobility and safety for all road users? What
system works and gets the best out of will big data and fast communications mean
developments in smart cities, smart to network management?

14
 What will be the road transport research covered all aspects of the future. Each of these
of the future? What is the best model for questions are discussed in turn.
research ecosystem for the roads industry in
Australia? What skills and disciplines should 3.2.1 What do you think are the biggest issues
we be encouraging in aspiring and next- relating to roads? (Q2)
generation connectivity experts? Does The responses were grouped into the:
Australia need to be part of the international Demand; Policy, planning and change drivers;
roads research community? Supply and design; and Operations; categories.
These were further subdivided into specific
3. A VIEW OF ROAD USERS issues. There were 656 issues raised by these
An important question related to the above is people. That is each person on average gave
what will happen vs what should happen, as it 1.38 issues. The particular issues were:
points to the role and stance of government, the Demand
management of assets and the priorities of the The main consideration here was the clear
community. This section addresses this issue view that there would be an increase in traffic
through presenting the results of a study of and that this would increase congestion. This
people’s view of the future of road assets. was particularly true in urban areas with the issue
3.1 The sample of low populations influencing the use of rural
roads. More specifically the responses were:
The survey was aiming at 300 responses. Increasing traffic and congestion 143/656
There was a large drop off in numbers from the 22%. Congestion was the major issue with 22%
panel - 770 people started the survey but there of the respondents highlighting it. This is
are only 280 (36%) usable responses. When expressed as congestion; traffic, traffic jams; and
this is combine with the social media pilot too busy roads, too many cars, too many cars
there was a total of 474 responses. The with one driver only in them, and overcrowding
demographics of the total sample was of roads is highlighted. The lack of planning
representative. The dominant state was NSW foresight to relieve congestion is expressed.
and the age profile was spread evenly over 18- There was little mention of the limitations on the
74 age groups. Twenty two percent were over infrastructure to cater for this congestion.
65 years of age. Latter questions in the survey Land use / transport interaction 15/656 2%.
showed a number of people with disabilities Many respondents pointed to the low
undertook the survey. There were slightly population in Australia and its large size. The
more females than males who completed the need to provide transport at a high quality in
survey. The majority of people were working
such a situation is challenging. This issue will
full time (38%) or retired (23%). The main
continue with particular emphasis on non-
income groups were the 0-$40,000 (38%) and
urban regions. Specific comments related to:
$40,001-$80,000 (30%). The majority of
too many people in the city should be moved to
people had access to one car and used it.
regional centres.
3.2 A General view Policy, planning and change drivers
This section focuses on questions where the The main issues with policy and government
respondents were offered opportunities to involvement related to planning and financing
respond generally about aspects of present and of roads. Tolls were seen as expensive. More
future road needs. The questions are very specifically the responses were:
useful since they elicited people’s view of the Government, planning 23/656 5%.
major issues related to the road asset. Their Reference to mismatch of speed signs and
main aim of the general qualitative questions is road design, lack of forward planning, and
to determine if the quantitative questions the economic impact of roads. The

15
government’s lack of commitment to road traffic lights and speed limits were a major
systems was mentioned. safety issue
Funding 14/656 2%. The funding of roads is More roads, capacity 30/656 5%. Four
seen as a considerable issue. This factor percent of the respondents mentioned the
includes: more funds for WA, cost cutting for inadequate capacity of the road system. Many
roads; too many tolls. The lack of funding arterial roads are not adequate for the volume,
particularly for rural roads. The application of not enough dual carriageway; and
expensive tolls and fuel tax are questioned. The infrastructure is not keeping up with traffic
need for more GST to build roads in Western growth. More underground freeways.
Australia was mentioned. The use of fuel tax Intersections and traffic signals 9/656 1%.
for road improvements was highlighted. The inconvenience of traffic signals was seen
Rural urban divide 13/656 2%. The divide as a concern. In particular poor timing.
between rural and urban areas was mentioned No change 7/656 1%. Several respondents
explicitly. The lack of rest areas along major felt that roads would not change and would be
roads was seen as a major safety issue. much as they are today. They felt the roads
Monotony, lack of spending on rural roads. were generally good.
Environmental and energy issues 10/656 Too many roads and road infrastructure
2%. Environmental were touched on in terms 6/656 1%. Some respondents thought there were
of road kill and reduction of fauna. Flood too many roads. This was expressed as too
damage was also mentioned. many roads, no highways everywhere, too many
Supply and design traffic signals, and you will never catch up.
The provision of roads, their design quality, Road works 4/656 1%. The large increase in
safety and the technology in building and the road works and the impact on capacity was
vehicles using them was a seen as a major mentioned on several occasions.
issue in the future. Parking and its impact on Parking 2/656 0%. Parking was mentioned
the capacity of roads was seen as an issue. by a number of respondents. It related to lack
Most of the responses appeared to relate to the of parking in local areas and in terms of its
higher level road system: distributors, arterials reduction in the capacity of many major roads.
and motorways. There was little reference to Operations
the local road system and its access The operations of the roads elicited a large
considerations. The responses were: response. These related to driver behaviour
The capacity of roads networks and the width composition, traffic control and enforcement.
of roads 61/656 9%. The improvement of the The responses were:
design and size of the infrastructure was a large The condition of roads 123/656 19%.
concern. The roads need refurbishment, need Another major issue was the maintenance of
wider lanes, poor design, not enough dual roads. This was directly related to: the need
carriageways, quality and standard improved, for refurbishment and replacement; potholes;
traffic signs, traffic signals, lack of design for overall poor condition; repairs. In general, it
the future, more dual carriageways, not wide was thought the quality of the asset and need
enough, speed limits, the roads are not too good, for refurbishment, renewal, lack of upkeep,
a lot of road edges are not too good, and quality lack of repair, potholes, maintenance was
of roads is terrible. Slippery road surfaces. prominent. Another important issue was to the
Safety 26/656 4%. Car accidents, death, and impact on general traffic associated with
driver training were also mentioned. Jay repairing the roads and new roads.
walking and other vulnerable road user Driver behaviour 93/656 14%. An
behaviour was highlighted. The need to fix interesting major issue was the behaviour of
dangerous intersections and roads was drivers: speeding, racism, self-absorbed
mentioned. The control of traffic through people; road rage; the use of drugs; alcohol;

16
lack of courtesy; stupid drivers; drivers using The demand (26%) and operations (40%)
cell phones while driving and bad drivers. groups attracts most attention emphasising the
Lack of respect for other road users by cars. road use focus on the road use more than the
Driver training was seen to be needed. quality of the asset. Local access and place
Composition 11/656 2%. The composition, issues did not attract a lot of attention
particularly the existence of trucks on rural indicating that roads were seen as a mobility
roads and highways was a concern. There was issue in this study.
a clear view that the number of trucks on the Asset management must take into account
road would increase and this had implications the user and their views of the asset. In terms of
for safety and congestion. Too many trucks asset management there were issues in each of
was mentioned. the four groups. In the demand area the
Public transport 11/656 2%. Public transport considerable concern about congestion and
was seen as an issue. More rail, bus and tram increasing traffic puts considerable demand of
facilities, Mention was made of freight travel the asset. It is clearly not satisfying the
on rail and the consequent reduction of truck customer. In terms of policy there is seen as a
traffic. Some suggestion of too much money is proof response from government and lack of
being spent on public transport was made. funding. The move towards privatisation of the
Bicyclist issues 6/656 1% Bicyclist were transport system and the introduction of tolls is
seen to be vulnerable at intersections. not totally supported by the community. There
Enforcement 3/656 0%. Police enforcement is voiced a need for more road infrastructure.
was mentioned but more in the light of police The supply and design of the system is not seen
hiding to catch drivers speeding than in terms to be coping with the increasing demand on it.
of presence. Safety, signage, wider roads and speed limits
Local issues 1/656 0%. Local issues were are a concern. In terms of operations the quality
also documented relating to safety (double of the asset is brought into question.
lines, school crossings), parking. Need for Maintenance for pot holes, surface conditions
double lanes in school zones are of concern. The use of the road by all
Walk (Pedestrian issues) 2/656 0%. Lack of modes gained some mention and improvements
a pedestrian network. in capacity in city areas with public transport
People with disabilities 1/656 0% There was gained the greatest interest. Clearly the
mention of people with disabilities and the increased use, quality design and deterioration
need for facilities for this group. of the asset are important issues.
The major response was one of the increased
traffic and consequent congestion (25%). The 3.2.2 How do you think the use of roads
condition of roads was also a concern (23%). will change in the next 30 years? (Q3)
Improved quality and design of the asset (10%)
This question was a major focus of the survey
and building more roads (4%) goes hand in
and was introduced to see if we missed out on
hand with these. These four responses account
any major issues in the quantifiable questions.
for 62% of total response. Driver behaviour,
The particular issues were:
speeding, arrogance, lack of courtesy was also
a major concern (15%) and safety (4%) Demand and usage
account for 19% of responses. Government Increased traffic, use and congestion 189/581
planning (5%) and funding (2%) also attracted 33%. By far the largest number of respondents
some attention. Speed and delay was not seen (33%) pointed to increasing traffic. Answers
directly as performance issue and was include: getting busier, more congestions, an
mentioned in the context of safety (4%) and increase in usage, more cars on road, more
driver behaviour (15%). This is an important issues, more and more cars, 30% increase in
change in view of road users. traffic, and more crowded, and poorer. Some

17
respondents indicated roads will be unusable. No Change 28/581 5%. A number of
This increase in traffic was interpreted as making respondents said there would be no change.
the road system worse with more issues and cars This was clearly stated with a small number of
using it. The increased congestion seems to be respondents seeing this as having a slightly
related to an increased number of cars on the negative dimension. No change unless
road. Cars were often seen as more attractive to infrastructure is overhauled.
road users than public transport and they had a Government, Policy and Planning 19/581
“right” to use them. However, a small number 3%. A number of respondents pointed to
pointed to public transport as a solution to the policy makers getting tougher and hence
congestion. People pointed to increased making the road system worse. There was
population as a reason for the increased mention of getting trucks of smaller roads and
congestions. This is also pointed out in the land onto freeways. The lack of funds in local
use / transport category. Overall, it appears that government to improve roads will result in a
a large number of respondent saw increased deterioration in this area. The relationship
between the number of users and the space
traffic and congestion as inevitable. A small
available was mentioned and hoped to improve
number of respondents were hoping that the
in the future. Mention was made of
number of cars on the road would decrease. This
controlling the time and use of the road system.
relates strongly to the congestion concern
Tolls, funding 11/581 2%. There was a
described above. number of respondent who indicated there
Land-use / transport interaction 11/581 2%. would be an increase in tolls and charges on
Some people mentioned the impact of road users like a fuel tax. For instance, more
increased sprawl on road travel. Increasing tolls, roads will get busier and more expensing
population was seen as a reason for increased to use, increased fuel cost will reduce travel.
traffic. Decreasing the population was also Increasing road prices, fund cutting. More roads
mentioned. Housing affordability in central will be toll. Some roads will be toll free. There
city driving decentralisation and longer trips. was also evidence that governments had a role
People work closer to home as congestion in funding infrastructure and that this may be
increases and public transport deteriorates. bias to particular constituencies. For instance,
Decentralisation will make more people drive need more GST in Western Australia, there is a
greater distances. Increased use less space. difference between urban and rural road funding
Decentralisation will reduce road volumes. and all funds will go to Sydney siders. A bias
Policy, planning and change drivers between different government parties was also
Technology 46/581 7%. In reaction to the mentioned. More expensive to use. Less
increasing demands on roads some respondents expenditure on maintenance and deterioration of
pointed to increasing technology. With more roads. The impact of costing funding on the
advanced cars being built. Some point to flying width of roads was also mentioned.
cars, fully automated vehicles improved quality Considerable change 2/581 0%. A number
and technologically developed roads. More of respondents indicated that everything would
driverless vehicles. More security cameras. change. Hence indicating that they felt to
Again this supported the inclusion of this future held many new transport developments.
question in the quantitative section. Roads will A common response to the future of roads was
be automated and more efficient vehicles, there would be a change. That is everything
(Electric vehicles, hydrogen, or fuel cells). will change.
Virtual workplaces will reduce travel. More Driver behaviour 11/581 1%. Driver
electric charge points. Smarter routes. More behaviour and attitude were mentioned by a
Uber like options. There was an indication that number of respondents. The improvement
there would be more surveillance pf vehicles would depend on the behaviour of Australian
on the road. and foreign drivers. Several focused on drug,

18
alcohol and cell phone use. Increasing use will increases and fuel prices increase to
make roads worse. Increase in road rage and unacceptable levels.
better road users was mentioned. Education Parking 1/181 0%. More parking in the CBD
programs related to behaviour and driving
Operations
ability was emphasised. Refresher courses
Public transport and demand changes
were also mentioned. This aspect was not
44/581 8%. There was a call for more efficient
looked at in the quantitative questions but
use of the road system and a reduction in roads
should be considered. Mention of elitism and
used by cars. An increased need and use of
driver attitude to other drivers.
public transport.
Environmental and energy 4/581 1%. The
The condition of roads and maintenance
removal of petrol driven cars was one future
19/581 3%. Roads will deteriorate resulting in
scenario. Cost of petrol will impact car use. lower speeds and more 4X4 vehicles. Roads
Hope petrol gone. Increased pollution. become smoother.
Supply and design Goods and truck movement 17/581 3%.
Infrastructure, design changes, efficiency and Increased truck volumes was mentioned. More
capacity 65/581 11%. In response to this increase trucks. Increase in accidents and truck crashes.
in congestion a number of respondents pointed to Mention was made of using rail to decrease
infrastructure solutions to the change. These freight movement by trucks on roads. Less
responses looked at increasing the number, trucks. Call for trucks to be automated. Keep
separation (divided roads), width, 4 lane roads trucks of local roads.
and connectivity of the road system. Some Safer 16/581 2%. There was a relatively
looked at looked at design changes to roads. small proportion of respondents who
More roads, new roads, more highways, wider mentioned safety. This was mainly aimed at
roads, separation of roads, and 4 lane roads being the roads being less safe and more crashes and
dominant. Several respondents pointed out that fatalities. Improved safety with more divided
relatively speaking Australia has good roads. roads being built. The use of divided roads
The inability of infrastructure development to was mentioned in other factors without being
keep up with the growth in traffic was linked to safety. They are trying to make roads
highlighted. There was some mention of safer for all people. Speed was mentioned in
different users, bicyclists, pedestrians and the regard to driver skill and the need to drive
separation of these users. It was hoped that roads slower. Increase in truck crashes. This
would improve, in particular the maintenance supported the inclusion of safety in the
and removal of potholes was mentioned. New quantitative questions. Change in vehicle fleet
materials would be used to improve roads. to keep people safe. More cars and roads will
Grade separation was mentioned. It will be a become more dangerous. More security
myriad of freeways and tunnels, there will be cameras, national road rules, more police
grade separated intersections, there will be more surveillance and presence
tunnels were mentioned. There was little or no Clearways, separation 10/581 2%.
mention of the local access or place Clearways and separation of different road
users attracted attention
considerations in the road hierarchy. Most
Bicycle 6/581 1%. Bicycle use was seen to
mention was made of the mobility questions.
increase and facilities for them develop
Build more capacity 24/581 4%. A number of
Cars 1/581 0%. Cars were mentioned as a
respondents felt that more roads will be required.
mode once.
Less use of the roads 7/581 1%. A number
of respondents felt the roads will be used less. Summary
Less use of roads due to congestion. Decrease The changes over the next 30 years were
use of roads as other modes of transport similar in character to the major issues
increase in usage. Less usage and congestion discussed earlier. The major changes were

19
seen as an increase in traffic and consequent the improvement of planning processes were
congestion (33%). More trucks accounts for seen as opportunities.
3%. Improved design, quality and capacity is Speed limit uniformity 10/714 1%
also a major change (11%) along with build Land use / transport interaction concerns
more roads (4%) and the condition of the 9/714 1%. Policy to control urban expansion
roads, maintenance and surface condition (3%). was mentioned.
These factors account for 62% of responses Environment and energy concerns 3/714
(54%). Increase use of public transport is seen 0%. Policy to control environmental and
as necessary (8%) and will result in capacity energy concerns was mentioned.
improvements. New technology in terms of Government, planning and policy 22/714
the vehicle, infrastructure and traffic control is 3%. The removal of tolls and the improvement
predicted to change (8%). No change accounts of planning processes were seen as opportunities.
for 5% and no response and unsure a further There was a call for more structured planning
10%. Government policy and planning (3%), processes and also
safety (3%), driver behaviour (2%). Supply, design and maintenance of roads
In terms of asset management the major Road expansion, capacity 156/714 33%.
concern was increased congestion on the roads The major point was the increase in capacity of
and the need for more and better infrastructure. the road system as the improvement of its
Roads a clearly seen for cars and most of the quality. More roads, dual carriageways, wider
future directions related to this aspect. roads, fly overs, separation of vehicle types.
Technology and public transport gain some Ensure roads are up to scratch, more bridges
comments for their improvement of the and tunnels, ensure roads constructed
operations of the road system. appropriately, wider roads and central barriers,
3.2.3 What changes would you make to roads in make road sturdier and less slippery when wet,
the future if you had unlimited resources? (Q4) double decker motorways. There was evidence
for a need for increased quality.
This question aimed to allow the respondents to Improved design and road capacity 110/714
think broadly about their priorities. There was 17%.
an average of 1.51 responses per respondent. The major point was to improve the design
Demand and usage and quality of roads. Build better roads,
Improved technology 24/714 4%. Technology separation of vehicle types, bicyclists,
did not receive a lot of attention. When it did it pedestrians. Ensure roads are up to scratch,
was rather extreme, flying cars, more driverless more lanes, better lighting, better connected,
vehicles. There was some consideration of ensure roads constructed appropriately, wider
new materials for roads and automatic roads and central barriers, make road sturdier
detection of errant behaviour. and less slippery when wet, double decker
Reduce traffic and congestion 16/714 2%. motorways. There was evidence for a need for
Removing congestion was suggested without increased quality.
specifying how to do it. The increasing of road The condition of roads, maintenance and
capacity outlined above was seen as one way road surface 74/714 16%. There was some
of doing this. interest in improving surface and maintenance
No change 2/714 0%. Some respondents of the road system. Better road surfaces, make
said there would be no change. pothole resistant, ensure roads are properly
surfaced, and better maintenance.
Policy, planning and change drivers
Road funding and tolling 26/714 4%. The Operations
funding responses related to more money for Improved public transport 45/714 7%.
particular states and the removal of tolls, no There was evidence in the responses that
tolls was the extreme. The removal of tolls and planning needs to take a new look and to focus

20
on alternatives like public transport. Issues lane configuration was also a major item
like limit cars, roads without trucks, fix (17%). The condition of roads attracted 16%
congestion, more roads, cheaper and available of responses. Together these three groups of
public transport. Ensure planners look ahead, responses took up 66% of the responses.
proper planning, Make a plan for the future, The asset management implications are
this does not appear to be happening. More clearly the need for more roads and
roads, expanded public transport, more public improvement of its condition. Improving the
transport, more motorways and public capacity of the roads by the improved
transport, increase spending on rail. provision of public transport attracted 7% of
Improve safety 35/714 5%. A small number the responses. A significant proportion looked
looked at increasing safety. Just improve at alternate modes of travel, particularly public
overall safety of roads, safer features. More transport to better utilise the road asset,
training and harsher penalties for bad improving the safety of the roads attracted 4%
behaviour, reduce speed limit, lock out systems of the responses. Improving the provision,
for drunks, opioids and raise the age for getting quality and efficiency of people movement
licence. Enforce maximum speed limits, through public transport was the major area
increase penalties for bad behaviour, Test people thought the road system could improve.
drivers every 10 years, Improving planning and policy procedures was
Rural urban divide 25/714 4%. Rural issues seen as an area of opportunity while there was
like divided highways, rest areas and a strong move towards the removal of tolls as a
deterioration of the roads attracted particular method of funding road improvements. The
interest. removal of trucks from some roads and the
Enforcement 27/714 4%. Enforcement movement of freight to rail transport was
was a particular issue with a call for more mentioned by some respondents. Technology
driver training and mandatory driver tests at improvements was seen as an integral part of
regular period. all of the initiatives. This question followed
Change driver behaviour 14/714 2%. the general theme of more and better
Improving driver behaviour was seen as infrastructure. The need for developing other
important by a number of respondents. Many of modes of transport, particularly public was
the technology improvements outlined above evident. A number of respondents had no idea.
were seen as ways of improving behaviour.
Bicycle facility improvements 7/714 1%. 3.2.4 Do you believe that the government
Improvements for bicycles, particularly will direct the agenda and spending on roads
separation of traffic flows attracted attention. in the future? Or will the agenda be directed
Clearway separation 17/714 2%. more by technology, private companies or
Separation of road and use of clearways was community demands? (Q7).
seen as an improvement. There is a general feeling in many of the
Rail freight 12/714 2%. Use of rail freight answers that government should direct the
rather than road freight was seen as an agenda but may not be able to do so. The
improvement. majority (49%) said that government would not
Other responses. Other responses included control the roads agenda and that technology,
parking issues, debris and cleanliness, remove communities and private companies would
trucks, car, bus, walk, improve carefully, have an influence. Of the remainder 31% said
lighting, renewable resources, and reduce roads governments would or should be in control.
attracted concerns. Six percent gave the most realistic answer that
Summary there is likely to be a sharing of responsibility
The major response was one of the increased and direction depending on the issue. For
capacity (33%) by provision and widening of instance, toll roads may be controlled by
roads. The improved design, widening and private companies since there is clear profit to

21
be made. However infrastructure for public
transport may be a private/government
partnership. Nine percent (9%) were not
certain who would be in control. The
remaining 5% gave no response.
3.3 Answers to guided questions
The questionnaire looked at a series of
questions developed from the literature review.
The questions were more guided in structure
and are listed in Table 1.

Table 1. Focus questions

Roads and their use will remain largely the same
as they are today.
On major roads like highways, cars and trucks
will be separated from cyclists and pedestrians.
Parking on major roads will not be permitted.
Local roads, and roads through shopping areas,
will give priority to pedestrians and cyclists Figure 1. Respondent view of what will happen
In the future, roads will be much safer for all users
In terms of what should happen the physical
The physical quality of roads and their surfaces quality of the roads, safety and separation of
will improve in the future
major activities were ranked highest (Figure 2).
Car, truck and bus travel will all be automated
(e.g., driverless) in the future
People will pay a toll or road charge for each trip,
with charges dependent on the time of day, route
and distance
Road travel will be more environmentally
sustainable in the future

The data from the numeric responses and the

qualitative answers of the data were cleaned
and skimmers are presented here. The
quantitative results show most things are
changes that people think will happen, and
should happen. Paying a road toll stands out as
the only one that people think will happen, but
on average do not want it to happen.
Automated/driverless they're on the fence, they
think it probably will happen but they're dead
on the middle of thinking it should/should not
happen. More specifically, in terms of what
will happen in the future the removal of
parking on the major roads and separation of
activities on major roads were overall ranked Figure 2. Respondent view of what they think
the highest (Figure 1). should happen in the future

22
 In terms of the less likely to happen, the 6. SUMMARY AND CONCLUSIONS
safety of the road system and no change were
ranked lowest. In terms of what should This paper explores the role of the public’s
happen no change and payment of tolls were view in next generation asset management
ranked lowest. systems. It first reviews the literature on the
future of roads and points to a number of areas
Figure 3 shows a comparison of these two where increased understanding is required.
responses. It can be seen that people think This review points to a number of areas where
they will be paying for the use of roads but the public’s view may be of interest. These
do not think this is what should happen. form the basis of a number of questions of the
Possible the clearest agreement is that change public’s view.
will not happen and it is not what people The study of the public’s view of the roads
want to happen. in the future points to increased traffic flow
The physical quality safety and separation of and potentially congestion. The general
the roads were seen as the things that should solution is increased capacity. This is mainly
happen. The removal of parking and separation seen to occur with improvement to the road
are seen as major things that will happen. system achieved by road design and capacity
In terms of asset management there was improvements. The impact of technology is
considerable support for improving the important in terms of vehicle use and
physical quality, separation of modes and infrastructure improvements. The overall
safety. In terms of what will happen the design of the system is seen to be important.
removal of parking and the separation of New initiatives in capacity provision,
modes were a major changes. Comparison of technology and alternate ways of doing
the open ended questions and the quantitative things are required. High capacity modes,
questions show some mismatch. The like public transport, are seen as one method
increase in traffic and the need for of improving capacity of the public and
infrastructure improvements were clearly goods movement.
focused on in the open ended questions. Also The main aim of the general qualitative
the behavior of drivers was not covered in the questions is to determine if the quantitative
structured questions. questions covered all aspects of the future and
if not what aspects were missing. Possibly the
most emphatic response was the issue of
increased traffic flow and congestion and the
view that this is going to increase. Other
concerns related to the structured questions and
reinforced their inclusion. One view that
appeared during the survey was a general
dissatisfaction with driver behavior.
The paper also focused on the responses to a
series of quantitative questions public’s view.
Most things are changes that people think will
happen, and should happen. Paying a road toll
stand out as the only one that people think will
happen, but on average do not want it to
happen. In terms of Automated / driverless
they're on the fence, they think it probably will
happen but they're dead on the middle of
Figure 3. Relationship between key variables thinking it should/should not happen. More

23
specifically, in terms of what will happen in 7. BIBLIOGRAPHY
the future the removal of parking on the major
Arup 2014, Future of highways, Arup, London, UK,
roads and separation of activities on major
<https://www.arup.com/publications/research/
roads were overall ranked the highest. In terms section/future-of-highways>. Viewed 19
of what should happen the physical quality of October 2017,
the roads, safety and separation of major BITRE 2016, Traffic on the national road network
activities were ranked highest. In terms of the 2013-2014, information sheet 80, Department
less likely to happen, the safety of the road of Infrastructure and Regional Development,
system and no change were ranked lowest. In Canberra, ACT.
terms of what should happen no change and Delbosc, A & Young, W (eds) 2017, Traffic
payment of tolls were ranked lowest. In a engineering and management, Institute of
Transport Studies, Monash University,
comparison of the two response it can be seen
Clayton, Vic.
that that people think they will be paying for Forman, RTT & Sperling, D 2011, ‘The future of
the use of roads but do not think this is what roads: no driving, no emissions, nature
should happen. reconnected’, Solutions, vol. 2, issue 5, pp. 10-23.
In terms of new generation asset Infrastructure Australia (2016) Australian
management to open ended questions focused Infrastructure Plan
on traffic congestion and design improvements. Lay, MG 1993, Ways of the world: a history of the
world’s roads and of the vehicles that used
Comparison of the open ended questions and
them, Primavera, Leichhardt, NSW.
the quantitative questions show some Transportation Research Board 2017, Critical
mismatch. The increase in traffic and the need transportation issues in the United States and
for infrastructure improvements were clearly world-wide, power point presentation, TRB,
focused on in the open ended questions. Also Washington, D.C.
the behavior of drivers was not covered in the Author Information:
structured questions. These aspects need to be Professor William Young, Australia, Monash
considered in future questions. University, [email protected]

24
The 3rd Int. Conf. on Transport Infrastructure & Sustainable Development (TISDIC 2019). Construction Publishing House. ISBN: 978-604-82-2893-4.

Keynote lecture 3: Performance-based earthquake

engineering analysis of short-medium span steel-concrete
composite bridges

Fabrizio Paolacci1,, Silvia Alessandri1, Daniele Corritore1, Phan Hoang Nam1,2

1
Department of Engineering, Roma Tre University, Rome, Italy
2
Faculty of Road and Bridge Engineering, The University of Danang - University of Science and
Technology, Da Nang, Vietnam

Corresponding author, E-mail: [email protected]

ABSTRACT: In this paper, the results of the performance-based earthquake engineering (PBEE)
analysis, carried out to assess the seismic behaviour of short-medium span steel-concrete composite I-
girder bridges, are presented and discussed. The selected case study is part of a group of bridges analysed
within the SEQBRI project, funded in 2012 by the European Union, which deals with a systematic
development of the PBEE analysis for short-medium span steel-concrete composite I-girder bridges.In
this respect,fragility and damage analyses of the selected bridge are performed using a proper component-
based numerical model along witha wide experimental campaign. These outcomes are then integrated into
the decision making analysis, where the selected decision variable is the repair cost ratio of the bridge.
The results show a good performance of short-medium span steel-concrete composite bridgesboth for
minor and major damage scenarios.
Keywords: Performance-based earthquake engineering, steel-concrete composite bridges, loss analysis,
concrete crossbeam.

1. INTRODUCTION depth of composite section; 2) no pre-stressing

Nowadays, short-medium span steel-concrete in concrete slab; 3) low dead weight and
composite I-girder bridges (SCC) made of hot limited foundations and settlements of
rolled steel beams are very popular, owing to supports; 4) simple erection methods because
their short construction time and reduced costs. of no steelwork on site; 5) short construction
Moreover, they are suitable for seismic areas on site and therefore, over passing of existent
due to their limited weight. With regard to railways or highways without any traffic
static loadings, these bridges can be easily restrictions. Favourable consequences of the
designed according to Eurocode 8 aforementioned properties arehighly resistance
(CEN19982005); however, the effectsof to earthquake, high durability, minimal overall
seismic loadings have not been yet adequately costs, high demolition and recycling capability,
investigated, and thus, they may exhibit etc. In addition, hot rolled (S355M/N-
damageeven in low-seismicity zones. S460M/N) steel produced according to the new
In general, steel-concrete composite bridges fine grain practice that is widely employed in
with hot rolled sections for small and medium the modern structural applicationscan be
spans (i.e., range between 25-40m) exhibit profitably used to increase the seismic
several advantages in terms of 1) small total performance of SCC bridges.

25
 The SEQBRI project, funded in 2012 by the concrete composite I-girder bridges. In this
European Union and conceived on these respect, fragility and damage analysis of the
premises, deals with a systematic development selected bridge are performed based on a
of the performance-based earthquake proper component-based numerical model and
engineering (PBEE) for short-medium span a wide experimental campaign. These outcomes
steel-concrete composite I-girder bridges that are then integrated into the decision making
includes seismic input randomness, demand analysis, where the selected decision variable is
and damage analysis as well as economic cost- the repair cost ratio of the bridge. The results
benefit estimations (Paolacci et al. 2017). In show a good performanceof shot medium span
particular, a new type of connection between steel-concrete composite bridges with respect
steel girder and piers that uses a transverse to other typologiesboth for minor and major
concrete crossbeam (CCB) is taken into damage scenarios
account. The steelwork parts of the composite
beams are prepared in the factory and equipped 2. DESCRIPTION OF CASE STUDY
with steel elements for connections (plates, The analysed case study is a straight 2-span
holes, welded studs, etc.). Then, the steel deck consisting of 4 main girders HE600B of
girders are transported on the construction site S460 steel grade (see Figure 1). The bridge is
to be placed on their supports. Without any 40m long and consists of 2 spans of 20m, while
propping, the steel span girders, which are the total width of the road cross-section is
simply supported at their ends on the head of 10.6m, with carriageway 6.5m wide and 2
the piers, operate like isostatic beams. The sidewalks 2.05m wide. The thickness of the
joint components (contact, studs, etc.) and the concrete slab is equal to 25cm. At the
reinforcement are installed. The slab is abutments, the steel girders are fixed to an end
concreted and the continuity of the composite reinforced CCB 0.6m wide. By this diaphragm,
beam is completed by concreting beam the deck is simply supported on normal
jointand the transverse beamover the pier. damping rubber bearings. At the intermediate
Possibly, separate supports can be considered pier, the steel girders are fixed to an
between the transversal beam and the pier to intermediate reinforced concrete cross-beam
limit the shear transfer through the joint. This 0.9m wide (see Figure 2). The pier (clear)
kind of joint has been studied in the past but height is 7m. A wall type pier of 0.6m thick
only for static load conditions. For example, and 7m wide is used. It is assumed that the
the Structure Laboratory of INSA Rennes was foundation soil is categorized as type B
involved in the project MIKTI (Lachal and according to EN1998. In this example, the soil-
Ariber 2002) to investigate innovative structure interaction (SSI) effect is neglected.
solutions for the design and the fabrication of Different types of CCB configurations have
beam-to-beam joints ensuring the continuity of been applied on several occasions (Hanswille
bridge composite beams (Somja et al. 2012). 2007). Many of them are variants of the three
Abbiati et al. (2018) tested this kind of bridges typical configurations, which are presented at
in cyclic conditions. the DIN Fachberichte 104 (2009). At each type
In this paper, the results of the performance- of the DIN proposal, the steel girder ends to a
based earthquake engineering (PBEE) analysis, head plate along with the whole height of the
carried out to assess the seismic performance cross section and the bottom flange of the girder
of short-medium span steel-concrete composite continues inside the CCB. The predominant
I-girder bridges, are presented and discussed. (usually tensile) forces at the top flange are
The selected case study is part of a group of transferred through the shear studs to the slab,
bridges analysed within the SEQBRI project, whereas the corresponding compression through
funded in 2012 by the European Union, which the contact between the two opposite bottom
deals with a systematic development of the steel flanges (types A and B) or through
PBEE analysis for short-medium span steel- concrete compression (type C). On the other

26
hand, if tensile forces are developed at the Thoughtrying to exploit the advantages and
bottom flange, these should be transferred avoid the disadvantages of the three typical
through the welded connection between the configurations, a new variant is formed and
flanges’ extensions (type A and B) or vertical investigated, especially forthe most
shear studs (type C for the intermediate CCBs criticalintermediate CCB. The new configuration
and all types for the edge CCBs). The shear of the intermediate CCB is closer to the DIN-FB-
forces are typically transferred through shear 104 variant B; however, the steel flanges are not
studs placed on the head plates, aligned parallel inserted into the CCB and the height of the head
to the bridge axis (types A and C). Alternatively, plate is limited (Figure 2). The idea is to transfer
at type B, the web is also inserted into the CCB the forces from the composite girder to the CCB
and transversally aligned shear studs transfer the through shear studs.
shear forces. The tensile force at the top flange is transferred
Last but not least, these configurations are gradually to the longitudinal reinforcement (or
proposed by the DIN FB in order to manage reversely a potential compression force to the
mainly vertical actions (i.e., dead and live concrete slab) through a group of vertical
loads), which produce negative moments and studs, which are placed on the top flange of
consequently tensile forces at the concrete slab the girder before the CCB. The top flange
and the top steel flange as well as compressive does not enter into the CCB. The shear force
forces at the bottom flange. For bridges is transferred to the CCB through a group of
subjected to seismic actions, significant tensile horizontal studs, transversally placed at the
forces might be exhibited also at the bottom sides of the girder’s web, which is inserted
flange of the steel girder, especially when the into the CCB. This group of studs is subjected
monolithic connection between CCB and pier only to pure shear, which is compatible to
is formed (see Eq. (1)). In such case, it is better their actual function. Holes for the placement
the use of variants A and B to be avoided. of reinforcement are foreseen at the web.
Concerning the magnitude of the tensile
stress, which might be exhibited at the bottom
flange of the steel girder, the following three
cases are distinguished:
- CCB1 - Bottom steel flange always in
compression,
- CCB2 - Bottom flange in compression or
in light tension,
- CCB3 - Bottom flange in compression or
in significant tension.
Using CCB1 and CCB2, the tensile force at
the bottom of the CCB is limited. This is the case
of short to medium span bridges, fully supported
on bearings. CCB3 could be met at bridges with
the monolithic connection between CCB and
Figure 1. Longitudinal section and cross section at pier. Significant tensile force at the bottom
span for selected case study flange is sustained by prestressed bolts or anchor
bars (see Figure 2). It is very beneficial that the
Considering that the CCB configuration is the duct of each anchor is filled with grease, which
most crucial detail, especially for composite offers the opportunity for inspection, re-tension
bridges in seismic prone areas and taking into or even replacement in case of an extreme
account the aforementioned discussion, the seismic event. In what follows, the vulnerability
widely used DIN FB type C is chosen to be of the selected bridge with CCB3 is analysed in
analytically as well as experimentally investigated. the framework of the PBEE.

27
 with  = 0.97,  = 1.3 mm-1, and  = 0.0045.
The effect of the shear studs along the deck is
modelled by one horizontal elastoplastic link,
whereas rigid links constraint the vertical
direction and the rotation between girders
and slab.
Studs welded in the transverse direction on
(a)
the web plate entering the CCB are simulated
with nonlinear links in the vertical and
horizontal direction. Each link simulates the
effect of 25mm shear studs.
The vertical head plates welded on the steel
CCB1 CCB2 CCB3 girders and directly in contact with the transverse
(b)
concrete beam are modelled by rigid links.
Figure 2. (a) Pier-deck joint with CCB, (b) Static Links standing for shear headed studs within
behaviour of pier-to-deck joints the joint are connected to these rigid links
according to the CCB configuration. In order to
2. NUMERICAL MODELING OF CCB- reproduce the constraint in compression due to
STEEL GIRDER CONNECTION the presence of the CCB, four compression gap
Using the FE code OpenSEES, a simplified elements have been modelled both at left and
nonlinear 2D numerical model of the bridge right sides of the CCB joints at two different
equipped with the CCB has been created, as levels, as shown in Figure 3.
illustrated in Figure 4. Only a single steel
girder with its tributary slab width equal to
beff = 2.65m wide is considered and thus only a
portion of pier is taken into account.
Nonlinearities have been explicitly considered
in the steel girders, concrete slab, pier, shear
headed studs of the CCB, shear headed studs
between girder and slab.
The steel girderand concrete slab have been
modelled using nonlinear beam elements with Figure 3. Numerical model of the pier-deck joint
fiber sections. The steelhas been modelled with CCB
according to Menegotto-Pinto model, whereas
the concrete has been modelled according to The compression gaps of the 2D numerical
Kent and Park model. models link the vertical end-plate to the axis of
The vertical shear studs connecting the steel CCB beam at mid-support, both in the left and
girderto the slab as well as the shear studs in the right side of the joint. Gap element is a
within the CCB are modelled using nonlinear link element whose axial stiffness is activated
links with an elastoplastic behaviour. Strength only when the absolute value of the negative
of the single shear stud Qu has been calculated relative deformation of its two end-nodes
according to CEN 1994-2 (2006), while the becomes greater than the initial gaps in the
related stiffness Ks has been evaluated by mean spring. A simplified evaluation of the stiffness
of the shear load-slip curvesdefined in Gattesco of the gap has been performed, assuming that
and Giuriani (1996): the compressive force coming from the girder
s would act uniformly on a surface equal to the
Q 
  1 e    S (1) area of the vertical head-plate. The two levels of
Qu pretension bars (DYWIDAG type) in the

28
bottom area of the CCB, have been modelled by been considered and the parameters of the
two elastic truss elements. The 2D FE model registration taken into account have been the
ofthe CCB is shown in the following figure. magnitude Mw, the distance from the fault
rupture DR and the local soil type. A range of
Magnitude Mw = 5.8-7.0 has been considered,
while to account for source mechanisms, far-
field earthquakes have been adopted and the
chosen distance DR has been comprised in a
range 15-40km.
Finally, among the 30 registrations chosen
by the PEER ground motion database, 14 have
been selected so that the mean spectrum best
approximate the target spectrum according to
the prescriptions of EN1998-2.
In order to better fit the target spectra, a
Figure 4. Experimental setup criterion of selection and modification of the
records has been considered. Fourteen input
This model has been validated through a
signals have been picked out so that the
wide experimental campaign. The setup of a
average spectra of their normalized time
test and a numerical/experimental comparison
histories to each PGA matching as best as
of the hysteretic force-displacement cycle is
possible the normalized reference spectrum.
illustrated in Figure 4 and 5. More details can
Moreover, a scaling factor has been applied to
be found in (Abbiati et al. 2018).
each input signal in order to minimize the
dispersion of the mean spectra compared to the
Eurocode spectrum. The scaling factor
modifies the original record only in amplitude.
Figure 6 shows the plots of the response
spectra of the selection.

Figure 5. Numerical and experimental hysteretic

response

3. SEISMIC INPUT SELECTION

A group of 14 accelerograms has been selected
from the PEER ground motion database. A Figure 6. Scaled response spectra of selection for
web application allows selecting natural time far-filed EC8 type 1
history records according to a specified range
for the characteristics of the recordings and the 4. PROBABILISTIC SEISMIC DEMAND
match to a defined response spectrum ANALYSIS OF CASE STUDIES
(http://peer.berkeley.edu/smcat/). The mean
squared error between the spectrum of each The selected EDPs for the chosen
registration and the target spectrum is representative case study are listed hereafter,
evaluated and the best 30 records were according to the classification of Mackie and
selected. The Eurocode 8 type 1 spectrum has Stojadinovic (2005) and Tubaldi et al. (2010).

29
 Global EDPs The multi-record IDA response of the pier
- Pier drift is the top displacement of the pier drift, obtained as the mean value of the multi-
normalized with respect to the total height. record IDA, is shown in Figure 7. The mean
value of the drift ratio for a PGA of 2.0g is
Intermediate EDPs
about 7%. The typical collapse ratio of 4%
- Bending moment-curvature (Mmax, max) of
(FEMA 356, 2000) is reached around 1.1g of
the pier at the bottom section.
PGA. For a PGA of 0.3g, which corresponds to
- Bending moment-rotation of the CCB
the PGA of response spectrum type 1 (soil B)
connection. These two measures have
chosen for the design of the bridge, the drift
been obtained for both the left and right ratio is about 1.5%. It is interesting to analyse
sides of the joint, where the centre of the moment and curvature at the base of the pier in
joint is identified with the middle axis of order to easily observe the nonlinear behaviour
the CCB element. The bending moment of of the section. Pier yields at its base for
the composite section just before the side earthquakes characterized by PGA of 0.3g
of interest has been defined as follows: while plastic hinge develops for stronger
Mmax = Mslab + Mgirder + Nz, where Mmax seismic actions. The ultimate value is obtained
is the bending moment of the slab, Mgirder between 1.2g and 1.4g.
is the bending moment of the girder, N is Maximum absolute values for bending
the axial force in the slab, and z the level moment in the composite beam and for rotation
arm (distance between girder and slab). of the CCB joint have been also recorded
- The rotation of one side of the joint is during dynamic analysis. In Figure 8, bending
given by the rotation of the vertical end moment-rotation relationship (mean value) of
plate with respect to the undeformed CCB at the left side, that is the maximum
x absolute bending moment versus the
configuration:   , where x is the
z corresponding rotation of the left side of the
relative displacement between the top joint, is shown. It can be seen that the onset of
node and the bottom node of the vertical nonlinear behaviour takes place "after" the
end plate and z is the height of the yielding of the pier, or rather for higher values
vertical end plate. of PGA (0.8-0.9g).
Representative force-slip curves, not shown
Local EDPs
here for brevity, have been derived using the
- Maximum force and relative displacement
mean values of the IDA curves for both the
of shear studs in the CCB in the horizontal
1st row and 4th row of shear studs. When the
direction.
lower group begins to yield, around 0.5g, the
Incremental dynamic analyses have been upper levels of connectors are subject to
performed(Vamvatsikos and Cornell 2002), smaller levels of stress. In particular, the
and multi-record IDA curves of the structural 1stgroup is still in the elastic range. The bigger
response quantities mentioned above have been deformations of the last line of studs are in
built. Engineering demand parameters are agreement with the fact that during the seismic
plotted versus the Intensity Measure (IM) action the neutral axis in the CCB is closer to
chosen as the peak ground accelerations (PGA the first rows of connectors.
of the selected group of input signals). Since In conclusion, for events with a PGA of 0.3g
the properties of the seismic response depend or under, the critical part is the pier only with
on the intensity of the ground shaking, a the formation of a plastic hinge at its base
comprehensive assessment requires numerous while the CCB does not seem to be affected by
nonlinear dynamic analyses at various levels of any relevant damage. The first level of damage
intensity in order to represent different possible within the intermediate CCB starts to arise
earthquake scenarios. The selected scan of from 0.5-0.6g, the value of PGA for which the
PGA is in the range (0.1g - 2.0g). lower group of studs within the CCB yield.

30
 deformations can result in spalling of cover
concrete, buckling of longitudinal reinforcement
or bar fracture. To quantitatively implement
PBEE for the considered reinforced wall-type
pier, the three damage conditions above
mentioned in this work have been considered as
damage measures. Berry and Eberhard (2003),
developed empirical equations to estimate
deformations at bar buckling and fracture and
concrete cover spalling based on theoretically
expected trends in drift ratios, plastic rotations.
The models were calibrated with existing
experimental results from the UW-PEER
Figure 7. IDA of the drift ratio of the pier reinforced concrete column performance
database, which includes the performance of
over 400 columns. Concrete cover spalling
represents the first flexural DS (DS), in which
there are marginal safety implications, there
may be a possible short-term loss of function,
and the cost to repair concrete spalling could be
significant. Buckling and fracture of the
longitudinal bars represent DSs in which safety
implications are significant, partial replacement
may be required, and a longer-term loss of
function may occur. Drift-ratio equations have
Figure 8. Bending moment-rotation (mean values)
of the CCB at the left side
been proposed by Berry and Eberhard (2003)
for damage estimation.
5. DAMAGE ANALYSIS AND FRAGILITY An alternative representation of the damage
CURVES condition in the pier is the plastic deformation
at the bottom section. The corresponding levels
In order to implement performance-based of damage, obtained in the pier, are for
earthquake engineering method for the concrete cover spalling 1.37%, for bar buckling
representative case of small-medium span 4.25% and for bar fracture 8.96%.
composite bridge and investigate its Figure 9a shows the fragility curves for cover
vulnerability, it has been necessary to accurately spalling, bar buckling and bar fracture in the
assess the probability of reaching particular reinforced concrete pier. As expected, the order
levels of damage. For this purpose, a correlation of the onset of damage from lowest to strongest
process between the most representative input conditions is: cover spalling, bar buckling
engineering demand parameters and damage and bar fracture. The probability of having
measures has been made. One of the most cover spalling is over 15% for 0.3g of PGA. It
popular methods for assessing the seismic grows then very fast, reaching 70% at the next
vulnerability of bridges is based on the step (0.4g). For 0.5 g, probability of cover
construction of fragility curves, which expresses spalling is already over 90%. The probability of
the probability of matching or exceeding a having bar buckling is under 1% until 0.8g. At
certain level of damage of a structure as a 1.2g, it is over 50% while at 1.6g it reaches
function of the intensity of the seismic action. 90%. The risk of bar fracture seems to arise
Piers are often the most vulnerable elements only for the strongest seismic inputs. The
in a bridge and column failure can have probability of fracture of the reinforcement is
catastrophic consequences. Excessive 0.5% for 1.6g and 15% for 2.0g.

31
 Fragility curves based on the plastic rotations Fragility curves for yielding and ultimate
of the bottom section of the pier were also built. curvatures of the pier at the bottom section, not
The general trend is similar to the aforementioned shown for brevity, have also been evaluated.
fragility curves based on the drift ratio. There is a probability of 5% for the section to
Probability of having cover spalling is 65% for a yield at 0.2g, while it is almost probable (80 %
PGA of 0.3g. Combining this result with the of P) at 0.3g. The probability of occurrence of
previous result, it seems very likely to have this the ultimate curvature grows in correspondence
phenomenon for the chosen case study, if a of higher ground motion levels: almost 10%
design earthquake will hit the structure. for 1.0 g of PGA, 30% for 1.2g, 60% for 1.4g
The risk of having bar buckling is still and 60% for 1.6g.
important, but lower than in the estimation based For the shear headed studs of the beam-to-
on drift ratio. The probability of having bar beam concrete connection, two levels of
buckling is almost 1% for 1.2g, 30% for 1.4g damage have been considered. In particular, the
and 50% for 1.6g. Bar fracture risk occurs once first level of damage is the yielding of the stud,
again in correspondence of high IM values, assumed as the last point of its elastic behaviour
although it is more probable than in the previous in the elastoplastic constitutive law chosen in
vulnerability estimation. In fact, probabilities the model. The second level of damage
associated with bar fracture are similar to bar corresponds to the complete rupture of the stud,
buckling ones but "shifted" of around 0.2g more. taken from the research work of Gattesco and
This is an interesting result which shows that the Giuriani (1996). The rupture of a shear headed
PBEE assessment may change when global stud, monotonically loaded, takes place for a
engineering demand parameters or intermediate horizontal slip of 9.7mm. The fragility curves
EDPs are considered. relative to the 1st row of shear studs are slightly
different (Figure 9b). In the first level of
transverse studs (upper level) the probability of
exceeding yielding slip is higher in the left side
than in the right one. Horizontal relative
displacement between the studs and the concrete
tends to yielding values with a significant non-
zero probability (2%) in correspondence of
PGA = 0.5g. The same probability grows from
2% up to 40% when PGA becomes 0.6g. In the
fourth level of transverse studs (lower level) the
(a) probability of exceeding yielding slip is slightly
higher in the right side than in the left one.
Moreover, significant values of probability
occur for lower PGAs than in the 1st line. When
PGA = 0.3g the probability of yielding damage
is almost 1%; when PGA = 0.4g the probability
of yielding damage is 30%; when PGA = 0.5g,
which corresponds to first levels of risk for 1st
row, the probability of 4th row is already equal
to 75%.

6. LOSS (COST-EFFECTIVE RISK) ANALYSIS

(b) Cost-effective risk evaluation is the very last
Figure 9. Fragility curves: (a) Drift of the pier and step of the global integrated probabilistic
(b) Horizontal relative displacement of performance-based evaluation procedure used
the studs for SEQBRI project, that explores all aspects

32
of the problem: a) global earthquake demand to (concrete cracking surface and/or volume,
local engineering demand parameters (EDP), percentage of broken or yielded reinforcement
b) associated damages, which are based on a rebars, percentage of broken studs or rotation
detailed non-linear analysis of each case study angle for example for what concerns CCB). At
including the experimental evaluation of the the end, the ultimate goal is to evaluate the
joint, c) the loss analysis. The result of this cumulative probability distribution function of
phase is a decision model, which is based on the decision variables DV (cost and/or repair
Decision Variables (DV) pertaining to the use
duration), as expressed by the PBEE framework
of a structure given a level of sustained
Integral defined by Eq. (1).
damage. Decision models are mathematical
relations between damage measures (DMs) and Most of those different successive analysis
decision variables (DVs). Commonly used parameters and intermediate outcomes, resulting
decision variables describe losses in terms of from the previous tasks, including numerical
repair cost, interruption in service (downtime non-linear analysis and experimental
or repair time), or loss of life: thus, decision evaluations. At each step, results are addressed
models are also called loss models. The in terms of probabilistic quantities such as mean
decision model may have several parts—those values, deviation or probability of exceedance
that relate DMs to repair quantities (Q) and (see previous paragraphs). Repair costs and
those that relate Q back to repair cost or repair duration databases are collected from different
time. A completely different set of decision former projects, US and French repair data
variables focuses on the remaining capacity of (Paolacci et al. 2017). Additional (but rather
the structure to function as intended. For
sparse) data are also collected from road/bridges
example, the return of a highway bridge to
maintenance services concerning European
differing degrees of functionality in a highway
recent “real field” seismic events such as Aquila
network is also an important loss criterion.
Functionality may be measured in terms of earthquake. Analysis preferentially focuses on
lateral load resistance in aftershocks, traffic direct costs related to structural damages, since
volume, lane and speed reductions, or access indirect costs appear to be very depending on
for emergency vehicles. the context (country, location, the proximity of
The main DV here adopted is the big cities, economical/ industrial activities, toll-
repair/construction cost ratio (direct cost). This roads) and are not directly related to SEQBRI
choice comes from the goal to provide an project bridge typology and specificity (SCC
immediate economic measure of earthquake bridges with CCB).
performance of the assessed structure. Indirect
6.1 Direct losses evaluation
costs like downtimes, even if mentioned, are
not explicitly considered in the analysis. Direct cost evaluation is essentially based on
Global earthquake demand is defined in terms repair quantities evaluated for different parts or
of peak ground acceleration (PGA), which elements of the structure for a given intensity
represents the most commonly used intensity measure or damage scenario and on associated
measure for structural seismic analysis and unit repair costs. For this matter, the bridge
design issues. Local engineering demand system is disaggregated into individual
parameters (EDP) consist in more representative components or ensembles, such as the columns
variables related to structural elements or CCB, denoted as performance groups (PGs)
deformation or load, such as pier drift, bending that are damaged, assessed, and repaired
moments, shear stress for examples. DSs are together using a specific combination of
expressed in terms of EDP limit values and different repair methods. Each of those PGs is
associated damage or repair amounts expressed also sub-divided into individual elements or
in terms of percentage of structural quantities sub-PGs.

33
6.2 Performance groups - PG3: CCBs (specific elements of the
analysed bridge typology): 1 PG per pier.
Different DSs are defined for each PG. A set of
- PG4: Deck (including concrete slab, steel
damage scenarios is defined for the analysed
girders and superstructures/equipment): 1
bridge that corresponds to most probable
PG per pier (associated tributary length,
events and relevant or representative
counted from mid-span to midspan),
combinations of DSs associated to the different
subdivided into concrete slab (including
PGs, derived from numerical analysis or
asphalt layers and sidewalk), edge
experimental tests.
(external) steel girders, central (internal)
Repair cost estimation (and loss modelling
steel girders and non-structural elements
in general) is almost always a structural system
(barrier rail).
problem, not simply determined at the
Figure 10 shows the PGs selected for the
component level. Therefore, it is often
case study.
necessary to combine (sum) numerous scalar
values in order to obtain, for example, a total
cost. In the SEQBRI project, the problem of
computing repair-related decision variables at
the structural system level accounting for the
correlations imposed by the selected repair
methods is treated using the notion of PGs.
APG is a collection of discrete DSs, associated
with different structural elements, correlated Figure 10. Performance groups
because they are affected by the selected repair
procedure. The tested case study is thus broken 6.3 Damage states and associated repair
down into PGs for each major bridge methods and quantities
component. Each PG represents a collection of Each DS corresponds to only one repair
structural components that act as a global-level method. Selection of repair methods is a crucial
indicator of structural performance and that step in computing the repair-related decision
contribute significantly to repair-level variable, repair cost ratios and repair time. A
decisions. PGs are not necessarily the same as selection of repair methods specific for the DSs
loadresisting structural components. For of the benchmark bridges has been stated. This
example, non-structural components may also selection is made based on the Caltrans database
be a PG, since they also suffer damage and (Mackie et al. 2008) of typical repair techniques
contribute to repair costs. The PGs generally since it has been defined to be generalizable and
correspond to things that are observable as a usable for performance-based evaluation of
unit. Grouping bridge components into PGs other structures. Each PG contains a number of
allows for more meaningful damage discrete DSs corresponding to repair quantities
assessment than grouping by component. The needed for restoring the bridge. The DSs are
definition of the PGs used for the SCC bridges numbered from DS0 to DS3 with higher
with CCB corresponds to the following numbers indicating more severe damage. The
structure disaggregation. Five PGs subdivided DS0 DS corresponds to the onset of damage
into sub-PGs have been defined for this study when repair costs begin to accumulate. For
(foundations are excluded): analysis, the repair cost of the bridge is treated
- PG1: Piers (bottom): 1 PG per pier, as 0 € below the DS0 level of damage. Damage
subdivided into an individual column for beyond DS0 is needed to be repaired and costs
multi-column piers. begin to accumulate. Slight damage less than
- PG2: Piers (top): 1 PG per pier, DS0 is assumed to be insignificant and not
subdivided into individual columns for needing repair. Damage computation requires
multi-column piers. the definition of maximum possible repair

34
quantities to define an upper limit to the Table 2. DSs of the major scenario
quantities and costs. The upper limit is called
Performance Damage Description
DS3 since it corresponds to the most severe
group state
possible DS for the elements in a PG. DS3
usually corresponds to complete failure and PG1 (Pier at DS2 Seal cracks and major
replacement of all elements in the entire PG, base) patching
even sometimes leading to the generalized PG2 DS1 Seal cracks and minor
collapse of the entire bridge structure. Details (Pier at top) removal and patching of
on repair methods here adopted can be found in concrete
(Paolacci et al. 2017). PG3 (CCB) DS2 Replacement of CCB
PG4 (Deck) DS2 Seal crack, clean deckand
6.4. Damage scenario and loss analysis apply methacrylate
To evaluate the expected repair cost of a
structure as consequence of a seismic event in
a given reference time period, one or more
damage scenario must be defined. Its
definition, which considers repairing items
more or less large, reflects a level of the
intensity measure (IM) of the ground motion at
the site. Two damage scenarios have been
defined and described in Table 1 and Table 2.
Once the scenario has been defined its
average cost can be calculated by summing the (a)
costs of the repairing items associated with each
PG. Figure 11a shows the repair-to-rebuild ratio
for both minor and major scenarios along with
the expected repair cost disaggregated for PGs
(Figure 11b). From the results, it clearly appears
that PG1 (base of piers) contributes most to
minor damage scenario expected to cost at high
value of the PGA, while peak contribution is
from PG3 at lower values of PGA for the minor
scenario and over the whole range of PGA for
the major damage scenario.

Table 1. DSs of the minor scenario

(b)
Performance Damage Description Figure 11. Expected repair cost conditioned to IM
group state for the minorand major damage
PG1 (Pier at DS1 Seal cracks and minor scenarios (a) and disaggregation of the
base) removal and patching of expected scenario (b)
concrete
From these curves, it appears that the
PG2 (Pier at DS0 --
top)
contribution for the minor damage scenario is
mainly from epoxy inject cracks and repairing
PG3 DS1 Seal cracks and minor minor spalls, independently of earthquake
removal and patching of
intensity. For major damage scenario, the
concrete
temporary support of the deck begins to rise
PG4 DS2 Seal crack, clean deckand rapidly as a contributing cost for higher
apply methacrylate intensities. In the analysis, it also can be noted that

35
one repair quantity can dominate because several It is evident that the application of this
PGs require that item in the associated repair complex framework can be totally justified
methods for that group. The method has been only when important damage conditions are
finally applied to the case study showing that: expected and for which the adoption of non-
- Significant probability of occurrence of linear models and probabilistic analyses appear
the minor damage scenario is due to justifiable. Given the fact that small-medium
0.3g < PGA < 0.6g. SCC bridges with CCB exhibit in mean
- Significant probability of occurrence of important damage conditions only for
the major damage scenario is due to PGA > 0.25g, it appears reasonable to employ
PGA > 0.7g. the PBEE methodology only for PGA values of
- An overall expected repair ratio between 0.35g or greater. These PGA values
5-8% is obtained for the minor damage characterize the most active prone-areas in
scenario and > 20% for PGA > 0.2g for Europe. For lower PGA values the design
major damage scenario. approaches to be adopted can rely on current
PG1, PG2 (pier) and PG3 (CCB) mainly design approaches prescribed by Eurocodes,
govern the decision-making analysis given that where partial safety factors cover aleatory
their associated costs results particularly uncertainties. In summary, no specific rules are
important. Therefore, any optimization action provided by Eurocodes to design CCBs, so, the
should act on these elements. design procedure proposed in the SEQBRI
project can be profitably used for this purpose.
7. CONCLUSIONS
8. ACKNOWLEDGEMENT
The SEQBRI project dealt with the full
application of the next-generation The Present work is supported by the funds of
performance-based earthquake engineering the European Project SEQBRI“Performance-
methodology (PBEE) employed for the Based Earthquake Engineering Analysis of
assessment of a new type of steel-concrete Short-Medium Span Steel-Concrete Composite
composite (SSC) bridges endowed with a Bridges”, Grant N° RFSR-CT-2012-00032.
special pier-to-deck connection.
The development of the PBEE framework 9. REFERENCES
for SCC bridges can be considered a novel Abbiati, G., Cazzador, E., Alessandri, S., Bursi,
contribution to the foundation of a new O.S., Paolacci, F., De Santis S.2018.
generation of earthquake engineering seismic Experimental characterization and component-
codes for bridges in Europe. The results are based modeling of deck-to-pier connections
novel. In particular, they entail: for composite bridges.Journal of
- A better understanding of the seismic Constructional Steel Research,50, 31-50.
behaviour of SCC bridges with CCB. Berry, M.P. and Eberhard, M.2003. Performance
- A reliable definition of limit states for this models for flexural damage in reinforced
new type of bridges. concrete columns.PEER Report, 2003/18.
- A proposal of a decision-making tool CEN1998-12005. Design of Structures for
based on next-generation PBEE Earthquake Resistance Part 1: General Rules,
methodology for SCC bridges with CCB Seismic Actions and Rules for Buildings.
in moderate/strong seismic prone areas. CEN 19942006. Eurocode 4: Design of composite
- A significant improvement of seismic steel and concrete structures.
design state-of-the-art of SCC bridges DIN FB 1042009. DIN-Fachbericht 104 -
with CCB, through the development of Verbundbrücken (Composite steel and
design guidelines for improving concrete bridges).
Eurocodes for the assessment of bridges in FEMA 3562000. Prestandard and Commentary for
seismic prone areas. the Seismic Rehabilitation of Buildings.

36
Gattesco, N. and Giuriani, E. 1996. Experimental Somja, H., Kaing, S. and Lachal A.2012. New
Study on Stud Shear Connectors Subjected to beam-to-beam joint with concrete embedding
Cyclic Loading. Journal of Constructional for composite bridges Experimental study and
Steel Research, 38(1), 1-21. finite element modelling.Journal of
Hanswille, G.2007. Composite bridges in Germany Constructional Steel Research, 77, 210-222
- State of the Art. Proceedings of 7th German Tubaldi, E., Barbato, M., and Dall’Asta, A.2010.
Japanese Bridge Symposium, Osaka. Transverse seismic response of continuous
Lachal, A. and Aribert, J.M.2002. National steel-concrete composite bridges exhibiting
program MIKTI, Conception de tabliers dual load path. Earthquake and Structures, 1,
performants dans le domaine des petites 21-41.
portées. Sujet 1-B: Techniques innovantes de Vamvatsikos, D. and Cornell, C.A. 2002.
raboutage, Cahier des charges. Incremental dynamic analysis. Earthquake
Mackie, K.R., Wong, J.M., Stojadinovic, B.2008. Engineering and Structural Dynamics, 31,
Integrated Probabilistic Performance-Based 491-514.
Evaluation of Benchmark Reinforced
Concrete Bridges. PEER report, 2007/09.
Paolacci, F. et al. 2017. Performance-based
earthquake engineering analysis of short-
medium span steel-concrete composite
bridges. Final Report, SEQBRI Project, Contr.
No: RFSR-CT-2012-00032, Research Fund
for Coal and Steel.

37
The 3rd Int. Conf. on Transport Infrastructure & Sustainable Development (TISDIC 2019). Construction Publishing House. ISBN: 978-604-82-2893-4.

Keynote lecture 4: Perspective on urban transportation

strategy with BRT for developing cities

Fumihiko Nakamura, Vice President Yokohama National University

Corresponding Author:
Address correspondence to First author: [email protected]

ABSTRACT: Bus Rapid Transit (BRT) system is one of the best alternatives for urban transportation
strategy especially in developing cities mainly because of its cost-efficiency. Starting from Curitiba,
Brazil in 1974, including Bogota, Colombia in 1999, many cities in Latin American Countries have
applied BRT system. In almost all cases, BRT systems are working as reliable service with higher-
capacity. In some special cases, BRT systems are well coordinated with urban transportation strategy
framework and/or with urban planning strategy framework, taking other transportation modes and land
use into account. On the other hand, there have been very few cases of BRT systems in Southeast Asian
Countries. Many cities have been considering BRT introduction, most of which, however, do not look to
have sufficient consideration either with urban transportation strategy or with urban planning strategy
such as Transit Oriented Development (TOD). The objective of the paper is to discuss the perspective of
BRT especially for developing cities. Firstly the paper reviews the history of BRT systems in the world
followed by the field survey results done in Curitiba,mainly from urban transportation strategy and urban
planning strategy points of view. Secondly, Multi-modal and Inter-modal aspects are well discussed,
where relationship with regular buses, balance between private car usage and parking policies are also
emphasized.Thirdly the framework of urban bus planning, management and operation is discussed with
roles of public sectors and private ones, based on the experiences in several advanced cities. Finally, in
terms of urban planning strategies, reviewing the original and applied concepts of TOD, the authors
discuss how TOD strategy could work with BRT systems. In conclusion, the authors address the
conditional possibilities and limitation of BRT systems especially in case of developing cities. More
specific implication is also mentioned in case of medium-sized cities in Southeast Asian countries.

1. INTRODUCTION higher-capacity. In some special cases, BRT

systems are well coordinated with urban
Bus Rapid Transit (BRT) system is one of the transportation strategy framework and/or with
best alternatives for urban transportation strategy urban planning strategy framework, taking other
especially in developing cities mainly because of transportation modes and land use into account.
its cost-efficiency. Starting from Curitiba, Brazil On the other hand, there have been very few
in 1974, including Bogota, Colombia in 1999, cases of BRT systems in Southeast Asian
many cities in Latin American Countries have Countries. Many cities have been considering
applied BRT system. In almost all cases, BRT BRT introduction, most of which, however, do
systems are working as reliable service with not look to have sufficient consideration either

38
with urban transportation strategy or with urban introduction and environmental impacts.
planning strategy such as Transit Oriented Thanks to the system, the usage of automobile
Development (TOD). The objective of the paper can be controlled and the system can be
is to discuss the perspective of BRT especially flexibly improved with the growth of city.
for developing cities. In Chapter 2, the history of Figure1 shows the year of starting operation
BRT systems in the world is introduced. The of major BRT cities.
chapter also discusses the cases in Curitiba, As in Figure 1, the earliest installation of
mainly from urban transportation strategy and BRT is in Curitiba, Brazil. In Curitiba Jaime
urban planning strategy points of view. In Lerner, the mayor of Curitiba took initiative in
Chapter 3, several related aspects are discussed, making the master plan which set bus system
including multi-modality, framework of as the important element in the city. Based on
planning, management and operation, the master plan, BRT in Curitiba finally started
implication with urban planning and design such its operation in 1974. After installation of BRT
as Transit Oriented Development. In chapter 4, in Curitiba, several Brazilian cities tried to start
as a chapter for conclusion, the authors address BRT systems one after another. Later on, some
the conditional possibilities and limitation of cities in Mexico, Colombia, Peru and Ecuador
BRT systems especially in case of developing was introduced BRTs. Especially BRT system
cities. More specific implication is also in Bogotá, the capital of Colombia, has been
mentioned in case of medium-sized cities in designed by themselves based on Curitiba's
Southeast Asian countries. BRT experience with some of their own ideas.
This Bogotá's bus system is named as
2. HISTORY OF BRT Transmilenio and it was started operation in
2.1 Definition of BRT 1999. Transmilenio has very high level of
performance with speed, capacity and
The word of "Bus Rapid Transit" first appeared presence. As Figure 2 shows, the number of
in the report published by USDOT (United BRT system is dramatically increased from
States Department of Transportation), 1975. 1999 to 2003 just after Bogotá’s start.
This book notes that BRT can be applied to Therefore Bogotá's Transmilenio can be
high dense cities center with minimum costs of regarded as a new milestone for BRT system.

Table 1. BRTs in Curitiba, Bogota and Jakarta

Curitiba, Brazil Bogotá, Colombia Jakarta, Indonesia

Features World's first BRT High performance Asian longest network

Bus with development Information technology

Population 1.85 million 7.36 million 9.59 million

Area 435 km2 435 km2 435 km2

GDP per capita US$ 10,816 US$ 5,967 US$ 3,005

The year of operation start 1974 1999 2004

Length of busways 72 kilometer 84 kilometer 123 kilometer

Number of vehicle 185 519 524

Average passenger per day 350,000 812,000 360,000

39
1970s 1980s 1990s 2000s 2010s
71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 00 01 02 03 04 05 06 07 08 09 10 11
C uritiba Bogotá Jakarta B angkok
Goiãnia Q uito São Paulo
P orto A legre G uayaqui
N agoya（K ey R oute） N agoya（Y utorito）
T aipei S eoul
K unm ing B eijingH angzhou
O ttaw a B risbane
A delaide S ydney
P ittsburgh Seattle O rland B oston Eugene
M iam i Los A ngels
P aris R ouen Lyon
C aen N antes
A m sterdam
Eindhoven

Figure 1. Open years of BRT operation

Figure 2. The number of new BRT cities in the year/Total number of BRT

Transjakarta in Jakarta city, Indonesia is one their own ways. Runcorn, one of the British
example of the followers of Transmilenio. A new towns, is famous for the busway system
non-profitable organization, called as ITDP set which is installed in accordance with the new
a branch office in Jakarta in order to install a town construction. Essen in Germany is the
new BRT system in Jakarta. ITDP Indonesia first city to have guided bus system, which can
takes initiative to learn more from be regarded as one of the BRT systems
Transmilenio. Transjakarta should be (Guided BRT). In recent years, more and more
appreciated as it has more than 120 km busway Indian and Chinese cities have been interested
in total, which was constructed within 5 years. in constructing BRT. According to the report
The scale and speed of the BRT infrastructure by EMBARQ, thirteen Chinese cities and four
construction is very remarkable and it has Indian cities have already installed BRT
given a big influence on other Asian cities. systems in 2012. African and Middle Eastern
Beside this history, buses in North America, cities such as Dar es Salaam and Amman also
including United States and Canada, Australia discuss to introduce exclusive lanes and pre
and European countries have been evolved in fare collection system.

40
 Japanese first BRT-like arterial bus started 2) Hierarchical network (called as RIT
to operate in Nagoya in 1982. In Seoul the bus (Integrated Route network of buses) with
system has been completely reorganized in different colored bus according with route
2004 and started to operate with center-located category regardless of operators. Figure 4
exclusive bus lanes which is also regarded as shows the concept of the hierarchical network
an example of BRT. Seoul municipality in which the color of lines is equivalent with
referred this reformation to Curitiba's bus the color of buses on the line. Red colored bus
system. There are also about ten Southeast operate on the main lines with busways in the
Asian cities, including Vientiane and Khon structural corridors. Orange colored bus
Kaen, trying to get new BRT systems. BRT operate on the feeder lines connecting with
system is expected to play more and more main lines at free-transferring interchanges.
important roles because construction of BRT is Green colored bus connects those interchanges
easier and cheaper than that of train. in ring lines. Silver colored bus operate along
As mentioned above, Curitiba is the root, red and green bus routes with the limited
Bogota is the turning point case and Jakarta is number of stops in order to enhance the
also one of the milestone in Southeast Asian operation speed to compete with private car
context. The authors summarizes the basic increase in early 1990’s.
information on BRT in those three cities.
2.3 Curitiba’s experience
Curitiba’s bus system has been well known for
its uniqueness in terms of the following points:
1) Center located busway with high rise
mixed use development along busways, which
create some sort of development axis or
structural corridor as shown in Figure 3.

Figure 4. Concept of the hierarchical bus network.

3) All the service planned by IPPUC, the

institute for urban planning, managed by URBS,
public company for urban infrastructure and
operated by private company.
4) Integrated bus interchange attached with
branches of city hall and some public service
facilities, and
5) Tube-style pre-boarding-fare-payment
bus station and articulate or biarticulate buses.
We should understand that dedicated center-
located busway is just one element of the
whole system. Most of the followers of
Curitiba did not considered these points but
just with center-located busways. Bogota and
Seoul are the well-known followers of
Figure 3. Busway in Curitiba Curitiba’s case. Bogota followed most of the

41
elements but no coordination with urban the lanes before and after bus boarding.
development and whole system management Continuous development along busways with
(Bogota is replacing existing private mini bus the ground and the second floors for business
operators gradually). Rather, Bogota focused and shops attract much activities with buses at
intensively on capacity, speed, and security the surface level of the streets. Curitiba
aided by information and communication respects freedom of pedestrian movement and
technology as well as modal shift from cars to relation between urban activities and buses.
buses, bikes and pedestrian with restriction of Difference between Curitiba’s busway and
car use by license plate control, local special Other BRT cases’ busway is illustrated in
tax for petroleum and yearly car-free day. Figure 5.
Seoul followed most of the elements but no
coordination with urban development. Seoul
busway is not busway but bus lanes without
curb segregation. Seoul has changed the whole
bus system with around 8,000 buses at once.
Jakarta followed Bogota’s experience
supported by ITDP but did not give sufficient
consideration on capacity and speed. It did not
invest on information and communication
technology. We should understand the
importance of framework setting and strategy
package setting.
Curitiba’s history could be learned more
from different aspects as shown below, which Figure 5. Difference of busway in Curitiba and
should be useful when we consider the other BRT cases
planning of BRT.
1) Curitiba wanted to introduce urban rail 5) Pre-boarding-fare-payment bus station
but gave up it because of funding problems. and biarticulate bus fleet was started in
Then Curitiba started to consider how they can Curitiba first in the world. The idea did not
manage the situation just by bus with full come from technology supply side but demand
accumulation of wisdom. Curitiba did not side. IPPUC team once deeply considered how
considered bus as just conventional mode but to carry 15,000 passengers per hour per
something new with some potential. direction just by bus with one minute interval.
2) Curitiba tried to avoid disordered The team realized that it is not enough just by
expansion of urbanization. Development axis large capacity buses. To make boarding time at
was originally designed as guideline for station as short as possible, prepayment station
urbanization. It should be flexible as it can system was developed. Needs-driven pressure
extend according with growth speed and other could make some sort of innovation revolution
exogenous factors. without considering so called common sense of
3) Busway with curbs was not strongly existing level of technology.
intended for capacity, speed and priority. The 6) As in many cities in Latin American
side lanes of busways are for local traffic region, lower income group residences were
whose volume is light and whose speed is located in suburban areas. In order to achieve
rather slow. From traffic engineering points of social inclusion policy for them, it was
view, busway is not necessary. The appealing important to provide connecting public
existence of bus infrastructure to citizen is transport system with a sort of new technology
more important in Curitiba. by which the residents should be proud. It
4) As traffic on side lanes of busways is could be strong evidence for them to be given
light and slow, bus passengers can easily cross big priority.

42
 7) The first pedestrianized street in Curitiba
is connected with one big tube-bus station. The
street was pedestrianized in 1972, when
Munich and Asahikawa also started
pedestrianized streets in their town centers. The
street is even now with full of citizens who are
much proud of it. Attractiveness should be
connected each other.
2.4 Current situation in Curitiba: Problems
and possible solutions
In spite of these Curitiba's efforts, our case
studies and interviews show that situation in Figure 6. Queue of waiting passengers at a tube stop
Curitiba has been changing completely since
2004. At some tube bus stops, as shown in
Figure 6, there are long queue during peak
hours which results from the lack of care about
design of tubes and usage. Moreover chronic
situation of management for operation,
operating companies have never tried to make
effort to serve better.
Brazilian federal government's policy of
promotion for purchasing and using cars
caused road congestion and less priority of
traffic signals dedicated for buses. Now traffic
jam in peak periods is very serious as shown
in Figure 7. Condominiums along the busways Figure 7. Traffic congestion in Curitiba in the
have become so popular that their prices got morning
raised more and many rich people who can
purchase cars have lived there. The issues Segregated busways and vehicles including
about utilization of platforms, management of articulate and biarticulate buses perform well
operation companies, training activities for even at present but the capacity of the stations
staffs, traffic management and regulation for are not well operated. During the peak hours
car usage are serious. As the political pressure some stations have queuing of waiting
from car users is very strong, bus priority passengers outside. It makes the passengers
signal system has been stopped and cycle time uncomfortable while their waiting for the bus.
at grade-intersection with busways are Each bi-articulated bus uses only one door (the
rearranged in order to give longer green time third door) for boarding at median stations
for crossing cars not for buses. Formerly, new though it has five doors. As a result, congestion
car parking construction had been prohibited occurs inside near the third door while other
in downtown up to 2004. This strong areas are not seriously congested and stopping
measures to control car demand to the city time at stations tend to become longer. This
center was abandoned in 2004. After that, a causes bus bunching and longer waiting time of
lot of parking garages have been constructed passengers, which declines reputation of buses
and now they are competing each other. by citizen. In addition to the problem of the
According to our survey, more than half of the stations, unsafe driving of buses is often
space is dedicated for garages in some blocks observed. This comes from poor management of
in the city center. drivers by operators, which should be strongly

43
controlled by URBS that is the public company 3) Security is the first priority for public
managing all Curitiba’s BRT systems. transport in order to attract existing car users.
As the corridors become occupied by rich Bogota’s BRT Transmilenio has shown
people, low income group has been obliged to reliable, speedy and secured public transport
live far side of the city or even outside of the system can attract middle-to high income white
city. As bus system is well integrated with collared business persons in the peak hours.
neighborhood cities, they can reach town center 4) Mobility must be designed carefully with
by integrated buses. Therefore, buses on the a wide range of variety, where people should
busway are well congested. However, as the bus be able to learn they can go safely and
network has flat fare system, which is welcome comfortably without private cars.
by poor sectors, cost efficiency becomes worse
as most of the passengers are long-distance 3. ASPECTS TO BE CONSIDERED FOR
riders. These negative points of BRT system, SUCCESSFUL INTRODUCTION OF BRT
attached with federal promotion of motorization,
3.1 Multi-modal and Inter-modal aspects
induced high amount of car use, which
politically made big pressure on elimination of The experiences of BRT in several cities tell us
strong traffic management and control. the importance of multi-modal and inter-modal
Furthermore, new residents in Curitiba who aspects. Here “multi-modal” aspect refers to
immigrated from other regions neither know choice environment of various modes. In other
the history of the city nor are proud of their words, urban transportation strategies should
city as human-oriented environmental capital provide the choice environment to all the
with very advanced way of city planning with travelers. Public transportation modes such as
lower cost and affluent of wisdoms respecting BRT should become one of the choices by all
education, historical memories and human the travelers. Public Transportation modes
dignity. They do not know the history of should be designed so that all the travelers
buses. Recently the policy of social welfare has recognize one possible alternative under their
been changed. Formerly all the citizen are travel conditions. Under this concept, the role of
requested to join city activities and work. regular buses and cars must be discussed clearly
Otherwise there had not been any subsidy. in detail. Integrated network of public
Nowadays, lower income group can get money transportation modes would enable efficient
without making any effort. Therefore, number mobility, consisting of BRT, regular buses and
of street beggars has been increasing as they do possibly paratransit modes. As car restriction on
not have to work. All of these factors have city centers also affects the multimodal aspect,
made security level worse and discouraged it would be important to set the role of cars in
people to use buses just because of poor the context of multimodal aspect. Connection
secured situation. between regular buses and BRT or that between
Those problems should be able to overcome if private cars and BRT could be discussed in the
we refer to Curitiba in the former days, Bogota next paragraph on intermodal aspect.
these days and Medellin with recent innovation. On the other hand, “Intermodal aspect”
The points to be learned are as follows: refers to the connectivity between modes,
1) Strategic location of low income group which include not only public transportation
residential district coordinated with their activity modes but walking, bicycle and so on as well.
patters and advanced public transport systems to Especially walking environment as a main
make them socially included and to provide feeder mode for public transportation systems
reliable and smooth access to the city center. should not be ignored. All the successful cities
2) History of the city should be learned by all regarding BRT in the world have quite a big
the citizen in Curitiba. Educational opportunity contribution to walking environment
should be provided even for senior people. improvement around BRT stations.

44
3.2 Separation of planning, management and operation
Table 2. Image of Separation of Planning, Management and Operation in case of Curitiba, Brazil
Local government with IPPUC Public Authority (URBS) Operator
planning Decide concept Decide everything
management Contract with operators.
Fare revenue collection and distribution
operation Monitor by CS do

In many cities, bus system has started as one of and infrastructure management. Operation
the service business. Therefore bus companies should be coordinated with traffic management.
plan the service, manage the company and
operate buses and drivers. Under this condition 3.3 TOD (Transit Oriented Develop)
without any control and regulation, companies BRT could provide most punctual and reliable
compete each other to get more passengers. In transport option to the citizen, which should be
most cases drivers also compete each other to emphasized though BRT itself will not mitigate
get more passengers as a part of or a full set of traffic congestion problem. If any city needs to
his/her salary is proportional to the number of mitigate traffic congestion problem, some sort
passengers carried. of modal shift should be considered. Modal shift
In order to overcome these problems, in developing metropolitan looks quite difficult.
learning from Curitiba’s case, the separation of Transportation textbooks tells us modal choice
planning, management and operation is highly mechanism, which could be applied anywhere.
recommended. There might be variety of the It should be reminded that this mechanism is
separation patterns. In case of Seoul city, based on the condition that all the traveler is
planning and management is done by the city, educated enough to make logical and rational
while the operation is done by the operators judgement, which is sometimes difficult in
based on the contracts with the city. In Curitiba, developing cities. Furthermore, security
planning is done by the City Planning Institute problems looks serious. Considering the issues
just under the mayor while the management is mentioned above, modal shift for BRT in
done by the public company for urban transport developing cities might be quite difficult.
infrastructure and system. Operation is done by
private operators based on the contracts with the
management public company as shown in
Table 2. In both cities, all the bus services are in
the same framework. Therefore, coordination
and integration of public transportation system
has been achieved at a very high level. On the
other hand, in case of Bogota, BRT system is in
the similar framework, while other regular bus
routes are out of it. Bogota city has a very clear
strategy that existing bus routes shall be
replaced one by one gradually, finally reaching
that all the routes would be in the same Figure 8. TOD case in Curitiba
framework. It might be one of the options.
Anyway, planning should be coordinated with In order to overcome this situation, TOD
city planning and urban design. Management concept should be powerful as it could change
should be coordinated with financial planning the life style of the citizen by enabling car-free

45
daily lives. One concern might be some REFERENCES
misunderstanding of TOD. Referring to the
Nakamura, F., Toyama, Y. and Tanaka, S. (2014).
original research documents on TOD, its A comprehensive review of BRT projects in
definition involves excellent walkable Latin America and its implication to Southeast
environment with various functions and activities Asian cities, an invited presentation at the 19th
which do not require the use of cars. Security for National Convention of Civil Engineering,
towns and BRTs are also needed. Once again in Kong Kaen University
Curitiba, as shown in Figure 8, walkable and Nakamura, F., Toyama, Y., Tanaka, S. and Wang, R.
dense development has been successfully applied (2013). BRT perspectives as a sustainable
aided by strategic building control. transport mode in developing cities, an invited
presentation at the 3rd International Symposium
4. CONCLUSION on Engineering, Energy and Environment,
organized by Thammasart University.
This paper reviewed the overview history of Toyama, Y. and Nakamura, F. (2013). A study on
BRT mainly in developing countries and BRT applicability on large cities in developing
focused mainly on Curitiba’s case as the first countries, WCTR (to be published into the
and excellent planning experience for buses WCTR journal).
Institute for Transportation Development Policy
and BRT. Learning from Curitiba’s history, the
(2007). Bus Rapid Transit Planning Guide
authors summarized the important aspects for Hensher, D. A., Thomas, F. G. (2008). Bus Rapid
BRT planning. Transit Systems: A Comparative Assessment.
The authors emphasize multi-modal and Transportation, vol.35, pp.501-518, Springer
intermodal aspects, the framework of Netherlands
planning, management and operation and Menckhoff, G. (2011). Some comments on
TOD concepts towards successful BRT proposed BRT scoring system,
systems in developing cities. http://www.itdp.org/documents/BRT_scoring_
BRT should show brand-new image, clean system_PT_summit_June_2011-final_-
and safe image and speedy and reliable image _Gerhard_Menckhoff.pdf, 2011
to the citizen especially potential users who are Federal Transit Administration, Office of
using cars and motorcycles at present. From Research, Demonstration and Innovation
(2009). Characteristics of BUS RAPID
the engineering points of view, capacity,
TRANSIT for Decision-Making, FTA
demand and traffic engineering should be Yabe, T. (2006). A Study on Planning Process for
important. Even though, the priority should be Bus Rapid Transit System, Yokohama
given to the clear and easy-understanding goal National University
as far as Curitiba’s experience is concerned. Dirgahayani, P.(2009). Managing Barriers towards
Simple but comprehensive proposal is required Intermodality Improvement based on Provider
in terms of connection with other public and User Perspectives to Promote Commute
transport, car restriction and urban activity Mode Shift to Bus Rapid Transit system, The
location setting. University of Tokyo
U.S. Department of Transportation (1975). Bus
ACKNOWLEDGEMENT Rapid Transit Options for Densely Developed
Areas, Office of the Secretary Federal
The authors thanks firstly to the experts at Highway Administration Urban Mass
IPPUC and URBS in Curitiba and Transportation Administration
Transmilenio in Bogota for their kind support Website of EMBARQ, http://www.embarq.org/
to our research interviews, secondly to Dr. Website of Transmilenio,
http://www.transmilenio.gov.co
Shinji Tanaka, Dr. Shino Miura for their
valuable advices to our research.

46
The 3rd Int. Conf. on Transport Infrastructure & Sustainable Development (TISDIC 2019). Construction Publishing House. ISBN: 978-604-82-2893-4.

Keynote lecture 5: Unified limit equilibrium design of

geosynthetic-reinforced fill walls and slopes

Jie Han
Department of Civil, Environmental, and Architectural Engineering, the University of Kansas,
Lawrence, Kansas, USA

Corresponding Author:
Address correspondence to First author: [email protected]

ABSTRACT: Geosynthetics have been successfully used to reinforce fill for walls and slopes for
sustainable development of infrastructure and natural protection. Geosynthetic-reinforced fill walls and
slopes have been commonly designed using the lateral earth pressure approach and the limit equilibrium
approach, respectively. In reality, failure of fill walls to slopes is progressive and can be analysed using
the same limit equilibrium approach. This paper describes the progressive failure of fill walls to slopes,
presents the evidences of fill walls and slopes having a similar failure mode, discuss the top-down limit
equilibrium analysis procedure, and demonstrates the use of the unified limit equilibrium method for
designing geosynthetic-reinforced fill walls and slopes even under special conditions.
Keywords: Geosynthetic, limit equilibrium, slope, stability, wall.

and Guo, 2017). For a geosynthetic-reinforced

1. INTRODUCTION fill wall, the current design method assumes a
linearly increasing lateral earth pressure based on
Fill is commonly used to construct fill walls and the Rankine or Coulomb theory (Figure 1a). Due
slopes through placement and compaction of fill to this lateral earth pressure distribution, the
in lifts. In the literature, walls and slopes are required tensile strength of the geosynthetic
artificially defined based on an inclination angle reinforcement increases with the depth (i.e., T1 <
of 70o (also called a slope angle) from the T2 < T3). For a geosynthetic-reinforced slope,
horizontal line. When the inclination angle is however, the required tensile strength of the
equal or larger than 70o, it is referred to as a wall. geosynthetic reinforcement does not change with
However, when the inclination angle is smaller the depth (i.e., T1 = T2 = T3) if there is no pullout
than 70o, it is referred to as a slope. issue. This comparison demonstrates that these
Unfortunately, the design methods for fill walls two design approaches for fill walls and slopes
and slopes are very different. Fill walls are often result in different magnitudes and distributions of
designed based on the lateral earth pressure tensile forces for geosynthetics at different
theories, e.g., Rankine or Coulomb’s lateral earth elevations, which require different reinforcement
pressure theories. However, fill slopes are often strengths and/or lengths. Such sudden and large
designed based on the limit equilibrium theories, differences happening at the boundary line
e.g., Bishop or Spencer’s slip surface between walls and slopes are not reasonable.
approaches. Geosynthetics (e.g., uniaxial Many studies have shown that geosynthetic-
geogrid and woven geotextile) have been reinforced fill walls and slopes fail in a similar
successfully used to reinforce fill walls and steep mode; therefore, it is not justified using very
slopes and create sustainable earthworks (Han different methods to design them.

47
 mode, and then focuses on the unified limit
equilibrium method for designing geosynthetic-
reinforced fill walls and slopes.

2. PROGRESSIVE MOVEMENT AND

TRANSITION FROM FILL WALLS TO
SLOPES
Figure 1. Reinforced fill wall versus slope Figure 2 shows the progressive movement and
failure of fill walls to slopes, which depend on
A unified method should be sought for
designing fill walls and slopes. Leshchinsky et lateral support and vertical load. When the
al. (1995) proposed this idea and a unified lateral load, Po, applied on the wall facing is
design approach to geosynthetic-reinforced equal to the overburden weight of the fill
slopes and segmental walls. In this approach, a multiplied by its coefficient of lateral earth
series of log spiral slip surfaces are analysed pressure at rest, the fill wall does not have any
through the toes of the walls and slopes to movement as shown in Figure 2(a). This state is
determine the distribution of mobilized friction so-called the Ko state. When the lateral load, P,
angles required to reach the limit equilibrium gradually decreases from Po, the fill wall starts to
state for each slip surface. Han and move laterally and potential slip surfaces develop
Leshchinsky (2006) developed a general and extend into the fill as shown in Figure 2(b).
analytical framework for design of flexible When the movement is limited, a bi-linear wedge
reinforced earth structures, in which planar with a top width of 0.3H (H is the height of the
failure surfaces were used to determine the fill wall) may be formed. This wedge is
distribution of required reinforcement tensile commonly used to design reinforced fill walls
resistance along reinforcement length using a with metallic reinforcement (i.e., inextensible
top-down approach. Leshchinsky et al. (2014) reinforcement) (Berg et al., 2009). If the lateral
improved the framework developed by Han load continues decreasing to the load, Pa
and Leshchinsky (2006) using log spiral failure (equivalent to the active lateral earth pressure),
surfaces instead of planar failure surfaces. the fill wall has an intermediate movement
Limit equilibrium of fill walls and slopes can (mostly happens during construction) and a
be analysed by an analytical method assuming Rankine active failure plane with an angle of
failure surfaces and ensuring force equilibrium =45o+/2 ( is the frictional angle of the fill) is
(e.g., the Bishop or Spencer method) or a formed as shown in Figure 2(c). This failure
continuum mechanics-based numerical method. plane is commonly used to design reinforced fill
A numerical method based on continuum walls with geosynthetic reinforcement (i.e.,
mechanics uses a strength reduction approach to extensible reinforcement). At this moment, the
obtain a limit equilibrium state. Both methods wedge is in a limit equilibrium state. However,
can calculate a minimum factor of safety (FoS) when the lateral load is lower than Pa, the wedge
and identify a critical failure surface for a fill loses its stability and slides down as shown in
wall or slope. Han and Leshchinsky (2004) Figure 2(d). Immediately after the wedge slides
found that the analytical limit equilibrium down, the lateral force becomes zero and the fill
method and the continuum mechanics-based wall becomes a steep fill slope as shown in
numerical method often result in similar factors Figure 2(e). This fill slope is not stable, another
of safety and critical slip surfaces. slip plane at an angle of  is formed, and the
This paper first discusses progressive wedge above the slip plane slides down. The
movement and failure of fill walls to slope to steep slope eventually becomes a flat slope at
demonstrate their relationship and transition, another limit equilibrium state as shown in
provides evidences of geosynthetic-reinforced Figure 2(f). This slope angle is also called the
fill walls and slopes having a similar failure repose angle.

48
 within the wall at the limit equilibrium state
identified by the numerical method is
approximately linear and close to the Rankine
active failure plane; however, the critical slip
surface identified by the Bishop method is
circular due to its assumption and only matches
the slip surface by the numerical method within
the lower portion. Figure 3(b) shows the
comparison of the critical slip surface within
the slope identified by the Bishop method and
the critical zone of the high shear strain rates
by the numerical method. This comparison
shows that the critical slip surface within the
slope is close to the critical zone.

Figure 2. Progressive failure of fill walls to slopes

The above progressive failure shows the

gradual transition from a fill wall to a fill slope
and can be prevented if geosynthetic
reinforcement is properly designed and installed.

3. FAILURE MODES OF FILL WALLS

AND SLOPES
Both analytical and numerical methods can Figure 3. Critical slip surfaces in wall and slope at
identify critical failure modes or surfaces the limit equilibrium state (modified
within fill walls and slopes. The analytical from Han and Leshchinsky, 2004)
method determines the critical failure surface
based on its lowest factor of safety among all Figure 4 shows that the critical slip surfaces
the factors of safety corresponding to a large within three-tiered walls at the limit
number of slip surfaces analysed. The equilibrium identified by the numerical method
numerical method often determines the and the Bishop method match well.
“potential” critical slip surface based on the The above discussions show that the limit
zone of high shear strain rates. Figure 3 shows equilibrium method can be used to analyse both
the comparison of critical slip surfaces geosynthetic-reinforced fill walls and slopes,
identified by the Bishop and numerical calculate identical factors of safety, and identify
methods for geosynthetic-reinforced fill walls similar slip surfaces as the numerical method
and slopes. Figure 3(a) also includes the based on the continuum mechanics. This
Rankine active failure plane within the wall. finding demonstrates that the limit equilibrium
Figure 3(a) shows that the critical slip surface method can be used as a unified method for

49
designing fill walls and slopes without the 4.2 Distribution of required tensile resistance
artificial separation of walls and slopes. along reinforcement
To ensure the fill wall with a height, H1, at the
limit equilibrium state with different failure
planes, the distribution of the required tensile
resistance for the reinforcement can be
determined using Equation (1) with different
inclination angles, I as shown in Figure 6.
Figure 6 shows no tensile resistance is needed
from the reinforcement at the failure plane
angle smaller than a certain value because the
fill itself is strong enough to maintain the
stability with this failure plane. As a result, no
further analysis is needed for the failure plane
Figure 4. Critical slip surfaces in tiered walls with an angle even smaller.
(modified from Leshchinsky and Han, 2004)

4. LIMIT EQUILIBRIUM APPROACH

4.1 Required tensile resistance of reinforcement
Han and Leshchinksy (2006) performed a force
equilibrium analysis to determine the required
tensile resistance, TRI, for a reinforcement to
maintain the stability of the wedge at the limit
equivalent state by assuming a planar slip Figure 6. Distribution of required tensile resistance
surface as shown in Figure 5: in Layer 1 to satisfy limit equilibrium for
 H12  sin  I  tan  cos  I  fill height, H1
TRI  (1)
2 tan  I  tan  sin  I  cos  I  When two reinforcement layers are
where  = unit weight of fill, H1 = height of involved, the calculated tensile resistance using
wedge,  = friction angle of fill, and 1 = Equation (1) with the wall height, H2, at the
inclination of failure plane. This above same slip surface should be divided by two
equation was obtained based on the limit equally for each reinforcement layer as shown
equilibrium condition (i.e., FoS = 1.0). When in Figure 7.
a fill wall is designed for a required FoS, the
friction angle in Equation (1) can be replaced
by a mobilized friction angle as follows:
   
m  tan 1  tan   (2)
  FoS  

Figure 7. Force equilibrium of a wedge with two

reinforcement layers

The distributions of the required tensile

Figure 5. Force equilibrium of a wedge with one resistance for both reinforcement layers can be
reinforcement layer determined as shown in Figure 8.

50
 also be long enough to ensure the rear has
sufficient tensile resistance to satisfy the limit
equilibrium condition.

Figure 8. Distributions of required tensile resistance

in Layers 1 and 2 to satisfy limit
equilibrium for fill height, H2

Figure 9(a) shows the required tensile (a) Overlapped

resistance in Layers 1 and 2 to satisfy limit
equilibrium for fill heights, H1 and H2,
respectively. To maintain the stability of the
fill wall at both heights, H1 and H2 at the same
time, Layer 1 in the front (i.e., close to the
facing) should carry extra load to satisfy limit
equilibrium for the first fill height, H1 as shown
in Figure 9(a). At the same time, Layer 1 in
the rear should carry extra load to satisfy limit
equilibrium for the second fill height, H2.
Since Layer 1 carries extra load in the front, (b) Adjusted
the demand for Layer 2 in the front is reduced Figure 9. Distributions of required tensile resistance
to satisfy the limit equilibrium. Figure 9(b) in Layers 1 and 2 to satisfy limit
shows the adjusted, required tensile resistance equilibrium for both fill heights, H1 and H2
in Layers 1 and 2 to maintain the stability of
the fill wall at both heights, H1 and H2 at the
same time. Since the analysis of equilibrium
started from the top reinforcement layer to the
bottom reinforcement layer, it is referred to as
the top-down limit equilibrium analysis
method. The procedure continues until all
layers are analyzed.
4.3 Required tensile strength for design
Figure 10. Required reinforcement tensile strength
The required tensile resistance using the above and length
procedure is the minimum requirement. For
practical applications, the required tensile Even though the above-discussed design
strength for each reinforcement should framework is based on a planar failure surface,
consider the tensile strength of reinforcement the design framework is also suitable for
material, the connection strength in the front, circular (Bishop), log spiral, bi-linear, or three-
and the pullout capacity in the rear as shown in part wedge (Spencer) methods. Figure 11
Figure 10. The connection strength is required shows the log spiral failure surface can also be
to ensure the front has sufficient tensile used for the design framework for
resistance to satisfy the limit equilibrium geosynthetic-reinforced walls and slopes
condition. The reinforcement length should (Leshchinsky et al., 2014).

51
 Figure 13 shows the outcome of a vertical
geosynthetic-reinforced fill wall with
secondary reinforcement. Clearly such
geosynthetic-reinforced fill walls cannot be
analysed using the lateral earth pressure
method but can be easily analysed by the limit
equilibrium method.

Figure 11. Log spiral failure surface (after Leshchinsky

et al., 2014)

This design framework has been included in

the US Federal Highway Administration
(FHWA) design document entitled “Limit
equilibrium design framework for MSE
structures with extensible reinforcement” by
Leshchinsky et al. (2016).
4.4 Limit equilibrium-based design software
This design framework has been included in
design software – ReSSA, which was
developed by ADAMA Engineering in the
USA. Figure 12 shows the outcome of a
vertical geosynthetic-reinforced fill wall using
Figure 13. Distributions of required tensile strengths
this design framework incorporated in the for primary and secondary reinforcement
ReSSA software. This figure clearly shows the layers in a geosynthetic-reinforced fill
distribution of required tensile strengths along wall (courtesy of Leshchinsky)
each reinforcement. For the upper three layers,
the maximum tensile strength is controlled by Figure 13 clearly shows that the distributions
the rear pullout capacity. The lower four layers of the required tensile strengths for primary and
have the reinforcement lengths more than what secondary reinforcement layers are different.
are needed. The required connection strengths The short secondary reinforcement layers are
for all reinforcement layers with the wall controlled by rear pullout capacities except for
facing are small. the lowest layer. The use of secondary
reinforcement layers reduces the required
maximum tensile strengths and connection
strengths from the primary reinforcement layers.
This result is in agreement with that in the field
study (Jiang et al., 2016).

3. CONCLUSIONS
This paper discussed the relationship between
geosynthetic-reinforced fill walls and slopes
and their gradual transition due to the reduction
of lateral support and failure of the unstable
wedge, pointed out the unreasonableness of the
Figure 12. Distributions of required tensile strengths artificial division between fill walls and slopes
in a geosynthetic-reinforced fill wall and the use of different theories for analysis,
(courtesy of Leshchinsky) presented the evidences that geosynthetic-

52
reinforced fill walls and slopes have a similar of Geotechnical and Geoenvironmental
failure mode at the limit equilibrium state, and Engineering, 130(12), 1225-1235.
described the unified limit equilibrium method Han, J. and Leshchinsky, D. (2004). “Limit
for designing geosynthetic-reinforced fill walls equilibrium and continuum mechanics-based
and slopes. This unified limit equilibrium numerical methods for analyzing stability of
MSE walls.” Proceedings of the 17th
method can address the tensile strength
Engineering Mechanics Conference, ASCE,
requirements for geosynthetic reinforcement University of Delaware, Newark, Delaware,
material, front connection, and rear pullout. USA, June 13-16.
This method can analyse not only a standard Han, J. and Leshchinsky, D. 2006. General
vertical geosynthetic-reinforced fill wall but analytical framework for design of flexible
also a complicated geosythetic-reinforced fill reinforced earth structures. ASCE Journal of
wall with secondary reinforcement. Geotechnical and Geoenvironmental
Engineering, 132(11), 1427-1435.
4. REFERENCES Han, J. and Leshchinsky, D. 2010. Analysis of
back-to-back mechanically stabilized earth
Berg, R.R., Christopher, B.R., and Samtani, N.C., walls. Geotextiles and Geomembranes, 28(3),
Design of Mechanically Stabilized Earth 262-267.
Walls and Reinforced Soil Slopes, Design & Jiang, Y., Han, J., Parsons, R.L., and Brennan, J.J.
Construction Guidelines, FHWA-NHI-00-043, (2016). Field instrumentation and evaluation
Federal Highway Administration, McLean, of modular-block MSE walls with secondary
VA, March 2009. geogrid layers. ASCE Journal of Geotechnical
Leshchinsky, D., Leshchinsky, O., Zelenko, B., and Geoenvironmental Engineering, 142(12),
and Horne, J. (2016). Limit Equilibrium 10.1061.
Design Framework for MSE Structures with Leshchinsky, D., Ling, H., and Hanks, G. 1995.
Extensible Reinforcement. FHWA-HIF-17- Unified design approach to geosynthetic
004, 120p. reinforced slopes and segmental walls.
Han, J. and Guo, J. 2017. “Geosynthetics used to Geosynthetics International, 2(5), 845-881.
stabilize vegetated surfaces for environmental Leshchinsky, D., Kang, B., Han, J., and Ling H.
sustainability in civil engineering.” Frontiers 2014. Framework for limit state design of
of Architecture and Civil Engineering in geosynthetic-reinforced walls and slopes.
China, 11(1), 56–65. Transportation Infrastructure Geotechnology,
Leshchinsky, D. and Han, J. (2004). “Geosynthetic 1(1), 129-164.
reinforced multitiered walls.” ASCE Journal

53
The 3rd Int. Conf. on Transport Infrastructure & Sustainable Development (TISDIC 2019). Construction Publishing House. ISBN: 978-604-82-2893-4.

Invited lecture 1: Sustainable development of by-products

based CLSM as multi-purpose Geo-materials

Young-sang Kim1* , Do Tan Manh 2

1
Department of Civil Engineering, Chonnam National University, Gwangju 61186, South Korea
2
Department of Civil Engineering, Hanoi University of Mining and Geology, Hanoi, Vietnam
*
Corresponding Author: [email protected]

ABSTRACT: This paper deals with developing the controlled low strength material (CLSM), which is a
self-compacting and self-leveling material, as multi-purpose geo-materials considered as profitable
materials such as backfill or structural fill materials, and thermal grout materials. Various CLSM
mixtures, produced by those wastes (e.g., pond ash, fly ash, excavated soil, red mud, gypsum) and binders
(i.e., both traditional binders and newly developed binders), were assessed by their engineering properties
(flowability, fresh density, bleeding, initial setting time, unconfined compressive strength, microstructural
analysis, and thermal conductivity) and environmental effects (leaching contaminants). As a result, the
developed CLSM mixtures performed well in the aspects of engineering properties and environmental
impacts, expectedly. In the aspect of geothermal systems, higher thermal conductivity and lower total cost
of coal ash-based CLSM than those of conventional grouts were observed by the laboratory tests. Along
with the acceptable results from the laboratory, good workability and sufficient heat exchange rate were
also obtained from the field test. All findings from this study revealled that the goals of sustainable
development of by-products based CLSM have been worked out and the developed CLSMs could be used
in the field as multi-purpose geo-materials.
Keywords: Controlled low strength material, back fill, geothermal systems, thermal grout, industrial by-
products.

1. INTRODUCTION plastic soil-cement. Moreover as reported in

ACI 229R (1999), there are various inherent
The American Concrete Institute (ACI) advantages of using CLSM instead of
defines controlled low strength material compacted fill in the applications mentioned.
(CLSM) as a self-leveling, self-compacting, These benefits include readily availability
and cementitious material primarily used to (i.e., using locally available materials, ready-
replace conventional backfill soil and mixed concrete suppliers can produce CLSM
structural fillings that result in unconfined to meet most project specifications), easy to
compressive strengths of 8.3 MPa or less (ACI place (i.e., because CLSM is self-leveling, it
229R 1999). CLSM is not considered as a needs little or no spreading or compacting),
type of low-strength concrete, but rather as a reduction in labor and equipment costs (owing
structural backfill. CLSM is known by many to self-leveling properties and no need for
different names such as flowable fill, compaction), versatile (i.e., CLSM mixtures
controlled-density fill, unshrinkable fill, can be adjusted to meet specific till
flowable mortar, soil–cement slurry, and requirements and mixes can be adjusted to

57
improve flowability), fast return to traffic (i.e., in cases of good flowability, self-compacting,
CLSMs can be placed quickly and support permanent fill, uniformity, stability and
traffic loads within several hours), reduction durability properties, the 28-day compressive
in excavation costs (i.e., CLSM eliminates strength ranging from 0.7 MPa to 8.3 MPa.
having to widen trenches to accommodate CLSM can be employed as an anticorrosion
compaction equipment), worker safety (i.e., backfilling (underground metal pipelines) if it
workers can place CLSM in a trench without satisfies a high degree of flowability, self-
entering the trench, reducing their exposure to compacting, self-leveling, early strength,
possible cave-ins), all-weather construction uniformity and durability properties, especially
(i.e., CLSM will typically displace any high electrical resistivity and a less than 2.0
standing water left in a trench from rain, MPa of 28-day compressive strength. More
reducing need for dewatering pumps), less recently, the application of CLSM for thermal
inspection (i.e., during placement, soil backfill backfilling (underground power cables,
must be tested after each lift for sufficient geothermal systems) has been commenced to
compaction CLSM self-compacts consistently investigate. The requirements for this
and does not need this extensive field testing), application can be listed here like a high degree
no storage (i.e., since ready-mixed concrete of flowability, self-compacting and self-
trucks deliver CLSM to job site in quantities leveling, early strength and durability
needed with no leftover till to haul away) properties, high density and low porosity, low
(Ling et al. 2018). If future excavation is permeability, the most importantly high
anticipated, the maximum long-term thermal conductivity (ACI 229R 1999). Based
on the abovementioned criteria, many studies
compressive strength should generally be less
have been performed with the objective of
than 2.1 MPa (ACI 229R 1999); therefore,
CLSM used in a wide range of applications
CLSM is advantageous owing to its relatively
such as a backfill (Brewer 1992; Do et al.
low strength. More importantly, the
2015; Do and Kim 2016; Kim et al. 2016; W.
potentiality of reusing by-products as
Sullivan 1997), utility and pipe bedding
components in CLSM is another advantage
(Adaska and C. Krell 1992), void fill (Gray et
that should also be considered because this al. 1998; Hook and Clem 1998), and bridge
makes a reduction in disposal demanding and approach applications (Snethen and Benson
finally leads to a contribution toward 1998). Other applications include using CLSM
sustainable development (Razak et al. 2009). for structural fill (ACI 229R 1999; Buss 1989;
Depend on criteria requirements, CLSM can Clem et al. 1995), soil stabilization (Do et al.
be used in various applications in the field. 2018; Do and Kim 2017; Do et al. 2017; Green
CLSM can be used as a general backfill such et al. 1998), and geothermal systems (Do et al.
as void fillings, filling abandoned underground 2019; Do et al. 2018; Do et al. 2017; Do et al.
structures if it has a high degree of flowability, 2017; Do et al. 2018; Kim et al. 2018).
durability properties, a less than 0.5 MPa of The goal of this study is focused on the
28-day strength. The application of CLSM for sustainable development of by-products
excavatable backfill (e.g., underground based CLSM as multi-purpose geo-materials
pipelines-water, sewer, and storm-drainage considered as profitable materials such as
pipelines, roadway trench, conduit bedding) backfill or structural fill materials, and
can be adopted with CLSM having high degree thermal grout materials. To accomplish these
of flowability, self-compacting and self- goals, the specific objectives were listed:
leveling, less subsidence and quick setting Development of CLSM using excavated soil
time, durability properties, easy for re- and coal ash; Development of a new
excavation and a less than 2.1 MPa of 28-day cementless binder for CLSM using entirely
compressive strength. The use of CLSM for by-products; Development and application
structural backfilling (e.g., bridge approach, of coal ash based CLSM for geothermal
foundation support, retaining walls) is feasible systems (Laboratory and field experiments).

58
2. SUSTAINABLE DEVELOPMENT OF ES could be a feasible material in improving
BY-PRODUCTS BASED CLSM AS the stability (bleeding level) of the proposed
MULTI-PURPOSE GEO-MATERIALS CLSM mixtures. The bleeding is found to be
relatively decreased with an increase in ES
2.1 Development of CLSM using excavated content (Figure 2).
soil and coal ash
The laboratory tests were conducted on CLSM
made with coal ash (e.g., pond ash and fly ash)
generated from coal combustion process of
thermal power plants in Jeolla Province (South
Korea) and excavated soil removed from
construction sites. A test matrix of mixture
proportions was developed and mixtures were
produced using a laboratory mixer. First,
control mixtures were produced with cement or
cementless binder (CB), Class F fly ash, pond Figure 2. Bleeding of various CLSM mixtures with
ash, and water. Subsequently, excavated soil respect to ES content (Kim et al. 2016)
(ES) was substituted for pond ash in amounts
of 10, 20, 30, and 40% replacement by weight. The initial setting times of control mixtures
The engineering properties of mixtures such as were shorter. However, the times were delayed
bleeding, flowability, fresh density, initial with the replacement of ES or cementless
setting time, and unconfiled compressive binder in CLSM mixtures but were still below
strength were tested in accordance with the maximum allowable limit required for
applicable ASTM Standards (2004): C 940, D general CLSM (Figure 3).
6103, D 6023, C 403, and D 4832, The compressive strengths of all proposed
respectively. mixtures measured conformed to the strength
Flowability of prepared mixtures requirement of re-excavation, less than 2.1
decreased with a corresponding increase in MPa (i.e., general fill) as reported in ACI
the ES content, and the desired flowability of 229R. In addition, the compressive strength
over 20 cm can be reached consistently at ES decreased with a corresponding increase in the
content of up to 30%, regardless of binders ES content, regardless of binders used. The
used. In addition, it is worth noting that the strength reduction of CLSM made with CB
flowability of CLSM with CB was slightly was remarkably more pronounced than that
higher compared to that of CLSM with with cement (Figure 3).
cement regardless of ES contents as shown in
Figure 1.

Figure 3. Unconfined compressive strength and

initial setting time of proposed CLSM
Figure 1. Flowability of various CLSM mixtures with mixtures with respect to ES content
respect to ES content (Kim et al. 2016) (Kim et al. 2016)

59
However, this proposed amount of excavated soil consistently at the G/Rm, Fa/RmG, and L/RmG
was only applicable in the local excavated soil used ratios of (0.40-1.33), (1.29-1.33), and (0.14-
in this study. It can be varied by different 0.71), respectively. The CLSM mixtures in
excavated soils (in other areas) that have different phase I (i.e., the control mixture determination)
from characteristics such as fine percent, fineness showed more stably when compared with those
modulus, liquid limit, plastic limit, and others. in phase II (i.e., effect of fly ash content), and
phase III (i.e., effect of lime content). In phase I,
2.2 Development of a new cementless binder
an increase in gypsum content (i.e., involving in
for CLSM using entirely by-products
the ratio of G/Rm) could lead to a relative
The new cementless binder (CB) derived decrease in bleeding of the fresh CLSM
from a cementitious mixture composed of fly mixtures. Gypsum might be, therefore, a
ash, red mud, lime, and gypsum was feasible material in the new cementless binder
developed by using by-products for the to improve the stability (bleeding level) of the
production of controlled low strength proposed CLSM mixtures.
material (CLSM). All engineering properties The initial setting time of the proposed
of CLSM made with the new proposed CLSM mixtures was a proportion of G/Rm,
cementless binder were investigated. A Fa/RmG, and L/RmG ratios. While an increase
comprehensive experimental program was in fly ash and lime content might accelerate the
planned to evaluate the engineering setting process, an increase in gypsum content
properties of CLSM mixtures made with the might lead to a relative decrease in setting time
new cementless binder. The experimental of the fresh CLSM mixtures. The maximum
program was divided into three phases. The allowable limitation of initial setting time as
mix proportions of various CLSM made with required for general CLSM can be achieved at
the new cementless binder (Fa-RmLG) were the G/Rm, Fa/RmG, and L/RmG ratios of
fabricated in the experimental program. In (0.40-6.00), (1.71-3.00), and (0.43-0.71),
phase I, five mixtures Fa-RmLG1, Fa- respectively.
RmLG2, Fa-RmLG3, Fa-RmLG4, and Fa-
RmLG5 were made by changing the ratios of
G/Rm from 0.40 to 0.75, 1.33, 2.50, and 6.00,
respectively, (i.e., other weight fractions
were fixed). From these mixtures, a control
mixture with the best acceptable engineering
properties was chosen for further
investigation (i.e., phase II and phase III).
Phase II and phase II were designed based the
control mixture by changing the ratios of
Fa/RmG and L/RmG from 2.14 (control Figure 4. Unconfined compressive strength and
mixture) to 1.29 (Fa1-RmLG), 1.71 (Fa2- initial setting time of proposed CLSM
RmLG), 2.57 (Fa3-RmLG), 3.00 (Fa4- mixtures on phase I (Do et al. 2019)
RmLG) and 0.43 (control mixture) to 0.14 The compressive strengths of all proposed
(Fa-RmL1G), 0.18 (Fa-RmL2G), 0.57 (Fa- CLSM mixtures, except for the first case in
RmL3G), 0.71 (Fa-RmL4G), respectively, to phase III (Fa-RmL1G), met the strength
investigate the effect of fly ash content requirement of even structural fill or pavement
(phase II) and lime content (phase III) on the bases reported in ACI 229R despite the fact
engineering properties of all CLSM mixtures. that Portland cement was absent in the
Flowability of the CLSM mixtures made proportions of CLSM mixtures. The ratios of
with the new cementless binder decreased with G/Rm, Fa/RmG, and L/RmG, used to control
a corresponding increase in the ratios of G/Rm, the cementless binder proportions, had a direct
Fa/RmG, and L/RmG, and the desired relationship with the compressive strengths of
flowability of over 20 cm can be reached the proposed CLSM mixtures.

60
Figure 5. Unconfined compressive strength and
initial setting time of proposed CLSM
mixtures on phase III (Do et al. 2019)
Figure 7. Relationship between 28-day strength
In addition, it is worth noting that red mud and and the molar ratio of
(CaO+Al2O3)/SiO2 (Do et al. 2019)
gypsum contents in the new cementless binder
positively affected to the early strengths (3-day 2.3 Development and application of coal ash
strength and 7-day strength) and 28-day strengths based CLSM for geothermal systems:
of the CLSM mixtures, respectively. A highly Laboratory and field experiments
alkaline environment released from the high red
mud content-mixtures might be a reason Conventionally, controlled low strength material
explaining for a development in early strengths (CLSM) is defined as a self-compacting and self-
whereas the high gypsum content-mixtures more leveling material. High thermal conductivity
releasing sulphate ions might be encouraged CLSM can be, however, used as the heat transfer
which react with alumina phase of fly ash to medium (thermal grout) in geothermal systems
enhance more formation of ettringite and due to its good flowability and low shrinkage.
eventually develop 28-day strength (Figure 6). Fesibilty of newly developed CLSM (with and
without cement) was investigated for being used
in geothermal systems. In the laboratory test
program, all properties of CLSM made with a
cementless binder (CB) were determined and
evaluated by comparing with those of CLSM
made with cement. In this work, four CLSM
groups were produced, tentatively named WB-C,
WB-CB, WS-C, and WS-CB, corresponding to
the ratios of water to binder W/B (i.e., changing
amount of binder content), water to solid W/S
(i.e., changing amount of water content), and two
binder types including cement and cementless
Figure 6. SEM image of Fa-RmLG5 in phase I (28 binder. Bleeding, fresh density, flowability, and
days) (Do et al. 2019) setting time, unconfined compressive strength
The relationship between 28-day strength tests were measured in the laboratory test
and the molar ratio of (CaO+Al2O3)/SiO2 was program. Thermal conductivity test was
also explored. As a result, the 28-day strengths conducted simultaneously with the 28-day
of all CLSM mixtures linearly increased upon unconfined compressive strength test. For a
increasing the (CaO+Al2O3)/SiO2 molar ratio, thermal conductivity test (i.e., conforming to
expectedly. This molar ratio could be also used ASTM D5334), the thermal needle probe used
to predict the 28-day strength of every CLSM consists of a heating element and a temperature
mixture made with the new cementless binder measuring element and is inserted into the
(Figure 7). specimen.

61
 As a result, all of the prepared CLSM mixtures In the aspect of economy, the proposed
performed similarly well and met the requirements CLSMs were more beneficial than conventional
reported in ACI 229R regarding all engineering grouts. Using CSLM thermal grout can reduce
properties (e.g., flowability, bleeding, setting time, 17-32 %, and 9-23% when compared to the neat
unconfined compressive strength) and the key cement grout and high solids bentonite,
parameter (thermal conductivity). It is concluded respectively. It is important to note that the
that the proposed CLSM mixtures in this study can proposed CLSMs were more economical than
be feasibly used as a thermal grout for borehole the conventional grouts since CLSMs had higher
heat exchangers. In addition, the use of cementless thermal conductivity and lower material costs
binder could lead to improving the flowability, (i.e., a very small amount of binders coupled
which is the key point on the performance of with a large number of by-products). But, more
CLSM in the aspect of workability of thermal importantly, the relative cost of CLSM mixtures
grout for borehole heat exchangers. Even though made with cementless binder was similar to that
the use of the cementless binder in CLSM mixtures of CLSM mixtures made with cement, regardless
decelerated the setting process, stability but still
of both ratios (W/B and W/S) as shown in Figure
met all requirements of CLSM despite the full
9. This finding would contribute to the
replacement of cement in the production of CLSM.
viewpoints of economics and sustainable
It is worth noting that thermal conductivities
development when the cementless binder can be
of CLSM mixtures made with cementless
binder were satisfactory when comparing to used for the production of CLSMs, replacing the
those of conventional grouts. This finding conventional grouts.
plays an important role if the proposed CLSMs
made with cementless binder are considered to
be used a thermal grout for borehole heat
exchangers (Figure 8).

(a)

Figure 9. Relative cost of the proposed CLSM

mixtures in comparison with congenital
grouts in the geothermal system (Do et
al. 2019)

In order to ensure the efficiency of the

proposed CLSM grout for borehole heat
(b)
exchangers in the field, a large-scale test is also
conducted. Figure 10 shows the location of the
built geothermal system in Yeosu Campus of
Chonnam National University, Jeolla Province,
South Korea. Two experimental houses
(control buildings) with the same dimension
(6.6m x 3.3m in length, 3m height) were made
for an evaluation of coal ash-based CLSM as a
Figure 8. Thermal conductivity of CLSM in thermal grout for borehole heat exchangers. In
comparison with conventional grouts this area, the initial temperature of the ground
respect to the ratios of W/B (a) and W/S is 15°C and the geothermal slope is 23°C. As a
(b) (Do et al. 2019) reference, a co-deposition temperature of

62
16.15°C at 50 m was used for the calculation. with coal ash-Figure 12a and bentonite-Figure
Thus, the average temperature was selected as 12b). The workability of the coal ash-based
15.57°C and used for the design. CLSM from the field tests was in the
acceptable requirements reported in ACI 229R,
(a) (1999). Finally, the ground heat exchanger is
connected to test buildings.
(a) (b)

Figure 11. Installing U-type (a) and Spiral-type

(b) into boreholes

(b)

Figure 10. Location of tested geothermal system Figure 12. Backfilling grouts into boreholes:
(a) 3D model of the geothermal system (b) CLSM made with coal ash (a) and
bentonite (b)
The geothermal system was composed of
five boreholes divided into two experimental The heat exchange rate of the proposed
groups (U-type and Spiral-type) as shown in CLSM from the large-scale test was compared
Figure 11. In each group, control grouts and found to be higher than that of
(bentonite and silicate sand) were also conventional grouts as shown in Figure 13.
employed to compare with the newly More importantly, the proposed grout (CLSM)
developed thermal grout (coal ash-based represented superiority to silicate sand on the
CLSM). The depth and diameter of the U-type performance in the field.
boreholes in the geothermal system were 40 m Even though silicate sand had much higher
and 0.15 m whereas those of the Spiral-type thermal conductivity than that of CLSM, its
boreholes were 25 m and 0.25 m, respectively. heat exchange rate was, however, lower than
The piping materials used in the geothermal the heat exchange rate of CLSM. This result
system installations were Polyethylene (PE) means thermal performance of the vertical
and Polybutene (PB) for U-type and Spiral- ground heat exchanger was influenced by the
type, respectively. The inner and outer airtightness of the contact between the heat
diameters of PE and PB tubes are 34.5 mm, exchanger and the backfill material rather than
42.2 mm and 16 mm, 20 mm, respectively. In the thermal conductivity of the backfill
this system, the pipes were used to convey the material itself. The good flowability of CLSM
heat transfer water to circulate into and out compared to that of silicate sand improved the
from and back to the heat pump equipment. airtightness, which affected the performance
After that, thermal grouts are backfilled into improvement of the heat exchanger. In
boreholes as shown in Figure 12 (CLSM made conclusion, good workability and sufficient

63
heat exchange rate obtained from the field test thermal grout materials. CLSM was synthesized
strongly reinforce the feasible use of the coal by coal ash and various industrial by-products in
ash-based CLSM as a thermal grout for Jeolla Province (South Korea) to assess not only
boreholes heat exchangers. the general properties (e.g., flowability, fresh
density, bleeding, initial setting time, unconfined
compressive strength, microstructural analysis)
but also environmental effects (leaching
contaminants). Based on the results of this
investigation, the following conclusions can be
drawn:
Beneficial utilization of excavated soil in
coal ash-based CLSM was first investigated in
this study for the possible applications of
trenching and backfilling. Pond ash was fully
recycled to replace natural sand in the
production of coal ash-based CLSM.
Excavated soil was then substituted for pond
Figure 13. Comparison of performance of thermal
grouts in the geothermal system ash in amounts of 10, 20, 30, and 40% by
weight to evaluate the effects of excavated soil
2.4 Contaminants of CLSM content on the characteristics of the pond ash
based CLSM. As a result, the partial
Heavy metals in CLSM also should be replacement of excavated soil to pond ash up to
considered as an aspect of the environmental 30% in the CLSM mixtures would decelerate
impacts since, as reported in ACI 229R, the setting process, reduce compressive
CLSMs are usually applied as backfill,
strength and flowability but still meet all
structural fill, or void fill materials that contact
requirements for CLSM despite the small
directly with the surface or groundwater. These
amount of cement used (4.78%). Especially, it
conditions could cause a potential risk of water
is worth noting that an increase in excavated
pollution through leaching of heavy metals into
soil content could lead to an improvement of
groundwater. All the prepared CLSM
specimens in this study were subjected to the stability as well as a slight reduction in the
contaminant test (TCLP). Overall, the environmental impacts due to the eco-friendly
concentrations of all elements including Ni, attributes of the excavated soil itself. This
Cu, As, Cd, Pb, Zn, Mn, and Cr were below the research suggested a possible 30% of
regulatory levels of Ni (5 mg/l), Cu (100 mg/l), excavated soil can be used in the coal ash-
As (5 mg/l), Cd (1 mg/l), Pb (5 mg/l), Zn (100 based CLSM.
mg/l), Mn (50 mg/l), Cr (5 mg/l), which are Considering a reduction of environmental
specified in Code of industrial wastes for pollution and sustainability contribution, a new
landfill disposal (2008) and Code of Federal cementless binder was developed using by-
Regulations of US government (1997). products in Jeolla Province, South Korea (e.g.,
Therefore, all hardened CLSM specimens in fly ash, red mud, and gypsum) for the
this study were classified as non-hazardous production of CLSM. To assess the technical
materials in the aspect of contaminants. feasibility of this cementless binder in CLSM,
effects of red mud, gypsum, fly ash, and lime
3. CONCLUSION contents on the properties of CLSM were
explored. The important points obtained from
This paper adressed the sustainable development the experimental results were the positive
of by-products based CLSM as multi-purpose effects of red mud and gypsum on the early
geo-materials considered as profitable materials strengths and 28-day strength of CLSM,
such as backfill or structural fill materials, and respectively. A highly alkaline environment,

64
released from red mud and lime, might be a 4. REFERENCES
reason explaining for a development in early
ACI 229R, C. (1999). "Controlled Low Strength
strengths (3-day and 7-day strengths). In
Materials (ACI 229R-99)." American
addition, gypsum might release sulphate ions
Concrete Institute, Farmington Hill, MI, USA.
which react with alumina phase of fly ash to Adaska, W., and C. Krell, W. (1992). Bibliography
form ettringites. on controlled low-strength materials (CLSM).
Based on the viewpoint reported in Brewer, W. E. (1992). "The Design and
American Concrete Institute 229R, controlled Construction of Small Span Bridges and
low strength material, owing to a very good Culvert Using Controlled Low Strength
workability and low shrinkage, was evaluated Material (CLSM)." FHWA/OH-93/014, Ohio
Department of Transportation, Columbus, OH,
as a novel back-fill material for thermal grout
129 p.
of borehole heat exchangers through its Buss, W. E. (1989). "Iowa Flowable Mortar Saves
engineering properties (flowability, bleeding, Bridges and Culverts."Transportation
fresh density, initial setting time, unconfined Research Record 1234, TRB, National
compressive strength, and microstructural Research Council, Washington, DC, 30–34.
analysis), environmental impacts, and more Clem, D. A., Hook, W. H., and Goldbaum, J.
importantly, thermal properties as well as total (1995). "Flowable Fill Makes Bridge
cost analysis. From the laboratory test results, Replacement Economical." Aberdeen’s
it is found that all of prepared CLSM mixtures Concrete Journal.
Do, T.-m., Kim, Y.-s., and Ryu, B.-c. (2015).
performed well in the aspects of the
"Improvement of engineering properties of
abovementioned characteristics. It is pond ash based CLSM with cementless binder
concluded that the proposed CLSM, owing to and artificial aggregates made of bauxite
a higher thermal conductivity and lower total residue." International Journal of Geo-
cost than conventional grouts, can be feasibly Engineering, 6(1), 8.
used as a thermal grout for borehole heat Do, T. M., Do, A. N., Kang, G.-O., and Kim, Y.-S.
exchangers. In order to ensure the efficiency (2019). "Utilization of marine dredged soil in
of the proposed CLSM grout for borehole heat controlled low-strength material used as a
exchangers in the field, a large-scale test was thermal grout in geothermal systems."
also conducted to verify the workability and Construction and Building Materials, 215,
613-622.
heat exchange rate of the proposed CLSM in
Do, T. M., Kang, G.-O., Go, G.-H., and Kim, Y.-S.
the field. As a result, the workability of the (2019). "Evaluation of Coal Ash Based CLSM
coal ash-based CLSM from the field tests are Made with Cementless Binder as a Thermal
in the acceptable requirements reported in Grout for Borehole Heat Exchangers." Journal
ACI 229R (1999). In addition, the heat of Materials in Civil Engineering, 31(6),
exchange rate of the proposed CLSM from the 04019072.
field test is found to be higher than that of Do, T. M., Kang, G.-O., and Kim, Y.-S. (2018).
conventional grouts. A good workability and "Thermal conductivity of controlled low
sufficient heat exchange rate obtained from strength material (CLSM) under various
degrees of saturation using a modified
the field test strongly reinforced the feasibility
pressure plate extractor apparatus – A case
of the coal ash-based CLSM as a thermal study for geothermal systems." Applied
grout for boreholes heat exchangers. Thermal Engineering, 143, 607-613.
Based on all findings with respects to Do, T. M., Kang, G.-O., and Kim, Y.-s. (2019).
engineering properties and environmental "Development of a new cementless binder for
effects, it can concluded that the goals of controlled low strength material (CLSM)
sustainable development of by-products based using entirely by-products." Construction and
Building Materials, 206, 576-589.
CLSM have been worked out and the
Do, T. M., Kang, G., Vu, N., and Kim, Y.-s.
developed CLSMs could be used in the field as (2018). "Development of a new cementless
multi-purpose geo-materials. binder for marine dredged soil stabilization:

65
 Strength behavior, hydraulic resistance Kim, Y.-s., Dang, M. Q., and Do, T. M. (2016).
capacity, microstructural analysis, and "Studies on compressive strength of sand
environmental impact." Construction and stabilized by alkali-activated ground bottom
Building Materials, 186, 263-275. ash and cured at the ambient conditions."
Do, T. M., and Kim, Y.-s. (2016). "Engineering International Journal of Geo-Engineering,
properties of controlled low strength material 7(1), 15.
(CLSM) incorporating red mud." International Kim, Y.-s., Do, T. M., Kim, H.-k., and Kang, G.
Journal of Geo-Engineering, 7(1), 7. (2016). "Utilization of excavated soil in coal
Do, T. M., and Kim, Y.-s. "Stabilization of Marine ash-based controlled low strength material
Dredged Sediments (MDS) by Industrial (CLSM)." Construction and Building
Byproducts in Honam Area." Proc., The 27th Materials, 124, 598-605.
International Ocean and Polar Engineering Kim, Y.-s., Do, T. M., Kim, M.-J., Kim, B.-J., and
Conference, International Society of Offshore Kim, H.-K. (2018). "Utilization of by-product
and Polar Engineers. in controlled low-strength material for
Do, T. M., Kim, Y.-s., and Dang, M. Q. (2017). geothermal systems: Engineering
"Influence of curing conditions on engineering performances, environmental impact, and cost
properties of controlled low strength material analysis." Journal of Cleaner Production, 172,
made with cementless binder." KSCE Journal 909-920.
of Civil Engineering, 21(5), 1774-1782. Ling, T.-C., Kaliyavaradhan, S. K., and Poon, C. S.
Do, T. M., Kim, Y.-s., and Dang, M. Q. (2018). "Global perspective on application of
"Stabilization of marine dredged sediment controlled low-strength material (CLSM) for
using lime-fly ash-red mud-gypsum binder." trench backfilling – An overview."
Do, T. M., Kim, Y.-s., Dang, M. Q., and Vu, N. T. Construction and Building Materials, 158,
T. "Thermal Conductivity of Controlled Low 535-548.
Strength Material (CLSM) Made with Razak, H. A., Naganathan, S., and Hamid, S. N. A.
Excavated Soil and Coal Ash." Proc., Congrès (2009). "Performance appraisal of industrial
International de Géotechnique–Ouvrages– waste incineration bottom ash as controlled
Structures, Springer, 808-815. low-strength material." Journal of Hazardous
Do, T. M., Kim, Y. S., Kang, G. O., Dang, M. Q., Materials, 172(2), 862-867.
and Tran, T. Q. "Thermal Conductivity of Snethen, D. R., and Benson, J. M. (1998).
Controlled Low Strength Material (CLSM) "Construction of CLSM Approach
Made Entirely from By-Products." Proc., Key Embankment to Minimize the Bump at the
Engineering Materials, Trans Tech Publ, 244- End of the Bridge” in The Design and
248. Application of Controlled Low-Strength
Gray, D. D., Reddy, T. P., Black, D. C., and Materials (Flowable Fill)." American Society
Ziemkiemcz, P. F. (1998). "Filling Abandoned for Testing and Materials (ASTM), 165–179.
Mines with Fluidized Bed Combustion Ash W. Sullivan, R. (1997). "Boston Harbor Tunnel
Grout” in the Design and Application of Project utilizes CLSM." ACI Concrete
Controlled Low-Strength Materials (Flowable International, 19, 40-43.
Fill)." American Society for Testing and
Materials (ASTM), 180–193. Author Information:
Green, B. H., Staheli, K., Bennett, D. , and 1. Full Name: Young-sang Kim
Walley, D. (1998). "Fly-Ash-Based CLSM 2. Prof/Dr/Mr/Ms: Professor/Ph.D.
Used for Critical Microtunneling 3. Country/University/Company: Republic of
Applications” in the Design and Application Korea/Chonnam National University
of Controlled Low-Strength Materials
4. Email: [email protected]
(Flowable Fill)." American Society for
Testing and Materials (ASTM), 151–164.
Hook, W., and Clem, D. A. (1998). "Innovative 1. Full Name: Tan Manh Do
Uses of CLSM in Colorado” in The Design 2. Prof/Dr/Mr/Ms: Lecturer/Ph.D.
and Application of Controlled Low-Strength 3. Country/University/Company: Vietnam/ Hanoi
Materials (Flowable Fill)." American Society University of Mining and Geology
for Testing and Materials (ASTM), 137–150. 4. Email: [email protected]

66
The 3rd Int. Conf. on Transport Infrastructure & Sustainable Development (TISDIC 2019). Construction Publishing House. ISBN: 978-604-82-2893-4.

Invited lecture 2: Problems and experiences of prediction

of geo-risks in tunneling in Vietnam

Nguyen Quang Phich, Nguyen Huy Vung, Trinh Van Thuong, Tran Nhat Minh
Faculty of Architecture and Civil Engineering, Binh Duong University, Thu Dau Mot, Vietnam
Xuan Sy TRAN, Quoc Vu TRAN
Faculty of Automobile Engineering Technology, Binh Duong University, Thu Dau Mot, Vietnam

Corresponding author, E-mail: [email protected]

ABSTRACT: The geological conditions by tunneling are generally complicated and in many cases there are
potentials of geological risks (geo-risks) to develop into accidents. Prediction of geological risks is therefore
one of the most important work in the design stage in order to have a safe construction of underground
structure. The paper presents some problems and experiences in prediction of geological risks in Tunneling in
Vietnam using numerical methods and proposals some hints for the future works. 
Keywords: Geological risks, prediction, problems and experiences, tunneling in Vietnam, UDEC, FLAC 2D.

after the excavation phase. (Nguyen et al. 2005,

1. INTRODUCTION Trinh 2005, Nguyen 2006, Nguyen 2007, Trinh,
Tunnel construction in civil engineering and in Broch and Lu, 2010, Vu et al. 2014). Therefore,
underground mining in Vietnam had many prediction and analysis of geo-risks are very
problems in excavation and operation, because important works just at the design stage in order
that is generally constructed in the geological to mitigate geo-risks and to determine
complicated rock mass, due to the presence of appropriate support system and reinforcement
discontinuities, such as joints, layers, folds, method, reasonably and quantitatively in
faults etc. In many cases there are tunnels consideration of stability of tunnel.
excavated in the vicinity of the existing To predict or identify and mitigate
structures and in weathered mix of soil and rock geohazards in tunneling in the design stage, a
in shallow depth. Besides it is difficult to predict number of software are used then since 2000 in
the geological condition and structure of the Vietnam to analyze the forms and types of
ground (rock mass and soil) correctly, the damages and risk connected to the design of
acquisition of physical parameters, as well as tunnels. (Nguyen 2007, Nguyen et al. 2012,
the prediction of the ground and system Nguyen et al. 2014, Nguyen et al 2019). If the
behaviors is extremely challenging. Until form and type of damage and risk depending
present, a variety of empirical, analytical and on the different designs then are known, it is
numerical methods are available for stability possible to choose appropriate design category
and serviceability analyses of underground and suitable support for the tunnels.
structures in rock masses, but the design This paper presents some simulation
problem is largely solved based on experience results and the experiences by using UDEC
and rule of thumb. Over several decades, some (Itasca 1996, Nguyen, Nguyen and Do 2007)
hydropower projects in Vietnam have suffered a and FLAC (Itasca 2005, Nguyen, Nguyen and
series of tunnel failures in different types from Do 2007) for illustration of geo-mechanical
rockfall, over collapse to cave-in during and processes in tunneling in hard rock masses.

67
2. GEOHAZARDS STUDY has a dip angle 0f -450 and constant spacing of
3 m. The joint set 2 has a dip angle of 800,
2.1 Influence of overburden on settlement and the constant joint spacing varies from 1
and cave-in m to 10 m in different models. The
Although all the water tunnels in Vietnam's underground opening has large horseshoe
hydropower projects are located in hard rock cross-section (straight wall, semicircular
masses, there are many sections of tunnels arches) with the wide and height of 10 m.
located not far from the boundary between hard The hard rock mass is below the ground
rock and covered soil. And it is just at these cover, which has an average density of 2.0 g /
locations that a lot of collapses occured and cm3 and a thickness of 50 m. Thus the value
some of them leaded to the formation of of vertical stress on the upper boundary of
sinkholes to the ground surface. To determine the model is 1MPa. The underground
the critical condition, several simulation tunnels are located under the upper boundary
models using UDEC were performed. 10 m (case 1) and 40 m (case 2). The rock
In the following show a case study for block (intact rock) and the joint sets J1 and
headrace tunnel in Huoi-Quang hydropower J2 have mechanical parameters compiled in
plant project. The considered rock mass has Table 1. The UDEC models for two cases are
two sets of persistent joints. The joint set 1 shown in Figure 1.

Table 1. Mechanical Parameters of Rock and Joint Sets

Intact rock Joint set 1 Joint set 2
3
Density  = 2600 (kg/m ) Normal stiffness kn1 = 6.0 (GPa/m) Normal stiffness kn2 = 3.8(GPa/m)
Bulk modulus K = 13.333 (GPa) Shear stiffness ks1 = 2.0 (GPa/m) Shear stiffness ks2 = 1.2 (GPa/m)
Shear modulus G = 8.000 (GPa) Tensile strength Tj1 = 0.01 (MPa) Tensile strength Tj2 = 0.01 (MPa)
Cohesion c = 1.0 (MPa) Cohesion cj1 = 0.05 (MPa) Cohesion cj2 = 0.05 (MPa)
Friction angle  = 45 (degrees) Friction angle j1 = 20.0 Friction angle j2 = 20.0 (degrees)
Tensile strength T = 1.06 (MPa) (degrees)

a) b)

Figure 1. UDEC models: a) model size

60x30m with hard rock
overburden of 10m; b) model
size 60x60m with hard rock
overburden of 40m

68
 The simulation results of varying the distance joints. The greater the distance between the joints,
between joints of joit set 2 are plotted in Figure 2 the more inhomogeneous is the stress distribution.
show that for near-surface excavation (case 1) Different spacing between the joints lead to
the rock blocks move along the surface of the different loosening appearances. It is not easy to
joint set 2, depending on the spacing of the joints explain these phenomena in general. However,
and on the interface between the joints and the except for the case b), the calculation results show
perimeter of the opening. It can be stated that in that the largest vertical displacement vector Umax
this case, the closer the distance between the gradually decreases as the distance between the
joints of the joint set 2, the faster the sliding joints of the joint set 2 (dJ2) becomes smaller, as
movements occur. And if the distance between shown in Table 2. Because of large block
the joints of the joint set 2 is large enough, movement, sinkhole could arise. In order to avoid
sliding along the joint surface could not occur. collapse to the surface, rock anchors with
Based on the results in Figure 2 it let also see shotcrete must be applied immediately after the
loose zones (bright areas) at the roof and the excavation. The simulation results for greater
bottom of the opening which can take little or no distance between the joints of joint set 2 shown
load depending on the distance between the joints that collapse and cave–in could not occur.
of the joint set 2. Of great interest in this example And in case of deep opening the
is the analysis of the distribution of stress and distribution of stresses and displacements
displacement in the rock mass. The rule of shown in Fig. 4 has the same manner like the
distribution of the major principal stress case of the shallow opening, but because of
component along two green lines through the great horizontal stress a cave-in is not occured.
centre of the underground tunnel shows very Rock bolting or shotcrete with steel mesh are
clearly the influence of the distances between the to used as support elements.
Table 2. Influence of joint Spacing of joint Set 2 on largest vertical Displacement
dJ2 (m) 3,0 2,5 2,0 1,5 1,0
Umax (m) 1,321 0,144 1,252 1,175 1,013
a b

c d

e f)

g h

Figure 2. Stresses and displacements distribution around the near surface opening by different joint
spacing of joint set 2: a) 1.5m; b) 2m; c) 2.5m; d) 3m; e) 6m; f) 7m; g) 8m and h) 10m

69
Figure 3. Stresses and displacements distribution around the deep opening by different joint spacings of
joint set 2: a) 3m; b) 1,5m

material and mechanical parameters of the

2.2. Influence of the presence of fault rocks, the fault and their joints are
Tunneling in mountainous coal mining region summarized in Table 3
Quang Ninh in sedimentary rock mass with
active tectonic requires excavation crossing
through increasingly difficult geological
mudstone
conditions. Fault is one of the most important
factors affecting on tunnel stability. The
tunnel and drift may have to be constructed
across a faulted zone as it is not always
possible to avoid crossing active faults. The
form of instabilities depends on different
factors.. The following presents a case study
simulating the geo-mechanical processes sandstone
around a coal mine drift that passes through a
fault of 1m wide in sedimentary rock mass
with two layers of mudstone and sandstone
(see Figure 4). Different models for
parameter study were developed in order to
investigate the influence of a fault in the coal Figure 4. UDEC model: drift through fault in
mining in sedimentary rock masses. The jointed rock mass with two layers.

Table 3. Mechanical Properties of Rocks and Fault

Rocks and fault Density  Bulk K Shear G Fiction  Cohesion C Temsion strength
g/cm3 GPa GPa độ MPa K MPa
Sand stone 2,65 14,2 7,0 29 2,6 4,6
Mudstone 2,50 10,0 5,2 26 2,2 3,6
Fault 1,40 4,3 2,3 35 0 o

Table 4. Mechanical Parameters of Fault

Normal stiffness kn Shear stiffness ks Cohesion C Friction  K
108Pa/m 108Pa/m Pa Degree Pa
Sandstone 9,0 3,0 0 20 0
Mudstone 7,0 2,2 0 18 0
Fault 3,0 1,2 0 12 0

70
 meaningful predictions with a single tool. In
this way it is possible to find out the form of
geohazards not only in the preparing stage but
also during the excavation when the geological
conditions are updated. And so geohazards
could be prevented or reduced.
2.2. Influence of the distribution of rock layers
In sedimentary rocks mass there are often
different rock layers with different distribution
order. The most analytical solutions assume
that the rock mass is homogeneous and
isotropic. There have also been some solutions
for homogeneous anisotropic masses which,
Figure 5. Propagation movement of rock mass however, have not been extensively applied to
surrounding the drift reality because of their complexity. The
following investigation case shows that using
The displacement vectors in surounding numerical methods such as FLAC 2D it is
rock mass after excavation are plotted in Figure possible to solve the boundary problems taking
5 for 2000 calculation steps. The simulation into account the existence of rock layers.
results give very interessting features about the Two models have been analyzed with
propagated movements of surounding rock different formations of layers in rock masses.
mass showed in Figure 5. Case 1: from roof to bottom (left to right) are
It could conclude that: layers of sandstone , claystone, coal,claystone
1) The displacement of the soft rock mass and siltdstone. Case 2: from roof to bottom (left
(mudstone) is much more greater than this one to right) are layers of sandstone, claystone, coal,
of moderately soft rock mass (sandstone); siltstone and sandstone. The tunnel has a semi-
2) The movement direction of surrounding circular arched that is dug in coal, located 12 m
rock mass depends on the relative stratification under the surface of hard rock layers. The study
of the rock layers and the fault. area is 30x30m in size, and is large enough for
3) Since the rock mass in the left side wall the size of underground works. The rock mass
of the drift is soft mudstone, rock sqeezing is consists different layers of coal claystone,
occured and strong floor heave is regonized. sandstone and siltstone with the mechanical
In this case the rock support must be parameters compiled in Table 5. The simulated
combined of active and passive support models are shown in Figure 6 and Figure 7.
elements in order to strengthen the rock mass, The simulation results provide complete
to reduce the deformation and to remain the informations about the rules of mechanical
perimeter of the drift. changes occur in the rock mass around the
So due to the coming together of numerical tunnel, including the rules of stress
modelling improvements such as numerical distribution, the formation of failure area, the
formulations for joint behaviour, and the displacement of the perimeter of the tunnel and
widespread availability of computing power, it the subsidence (settlement) of the upper hard
is now possible to perform practical analysis of rock boundary of the model.
discontinuous rock masses with the UDEC. Based on the obtained results, following
The simulation results show that a basically remarks can be drawn: when the rock mass has
advantage of UDEC is the flexible modeling of layered structure, all the rules of stress
broaden range of continuous and discontinuous distribution, displacement and the formation of
rock mass behaviours without any priority destructive areas depend on the distribution of
assumptions on failure mechanisms. As a layers. So one must be careful when applying
result, it is possible to examine different design the rules received by simple analytical
ideas, and obtain meaningful results or make solutions or experiences.

71
 The two simulated models showed that in layer. The distribution of the rock layers
the second model the displacements and the clearly affect the processes of the stress
failure zone have relatively higher values, distribution and movement in the rock mass;
although in the model 2 the hard sandstone is The destruction zone in the case 2 becomes a
located both in the upper and in the lower little wider.

Table 5. Mechanical Parameters of sedimentary hard rocks in the FLAC 2D-model.

Density  Cohesion C Friction angel  Bulk modulus K Shear modulus G
Rock Types
g/cm3 MPa Degree GPa GPa
Sandstone 2,61 1,00 40 11,60 8,70
Coal 1,30 0,01 35 2,60 1,30
Siltstone 2,50 1,00 25 10,00 7,00
Claystone 2,60 0,10 30 9,60 2,70

a) b)

c)

d) m

Line of upper boundary

Subsidence

m
Figure 6. a) FLAC model (model dimension 60x30m); b) Failure zone around the tunnel, c) displacement
of tunnel perimeter and d) settlement of the upper boundary for case 1 (sandstone-claystone-
coal-claystone-siltstons

72
 a) b)

c)

d)
m
Line of upper boundary
Subsidence

m
Figure 7. a) FLAC model (model dimension 60x30m); b) Failure zone around the tunnel, c) displacement
of tunnel perimeter and d) settlement of the upper boundary for case 2 (sandstone-claystone-
coal-siltstons-sandstone
3. DISCUSSION AND CONCLUSION From the planning and design practice
For prediction of risks or geohazards in shows that the simulation of the
tunnelling and underground works there are geomechanical process has to be carried out
needs of the development of an accurate very carefully in the design stage. The
geotechnical model and adequate knowledge of geomechanical changes are diverse and
spatial variation in rock mass geological strongly dependent on the presence of
condition. That are the prerequisites for the complementary geological conditions, in
planning and risk management. particular the uncertainty of these conditions.

73
 The application of the numerical methods is Nguyen Q.P. et al. 2012. On the Influence of Joints
absolutely necessary and one would have to onto Geohazards in Tunnelling. Proceedings
develop various different simulation models of The Nineteenth National Conference of
from the geological conditions and investigate Mechanics. Hanoi, 8-9 Dezember 2012. P.
them in order to obtain realistic results. 140-144. (in Vietnamese).
Nguyen Q.P. et al. 2014. Numerical simulation of
the influences of water inrush on underground
4. REFENCES coal mining stability in vietnam. Proceedings
Itasca consalting group 2005. FLAC- Fast of the conference “Mine Planning and
Langarian Analysis of Continua . User’s Equipment selection”. DOI: 101007/978-3-
Guide .3rd.edn (FLAC Version 5.0, April 319-02678-7-60.Springer International
2005). Publishing Switzerland 2014. 629-636.
Itasca Consulting Group 1996. UDEC-Universal Nguyen Q.P. et al. 2019. Influences of selected
Distinct Element Code, Version 4.0. User’s boundary conditions and size of the study area
Guide. Minneapolis, Minnesota, USA. on the forecast and analysis results of
Nguyen Q.P., Anh D.N,. Nguyen M.K. 2005. Risks geological hazards in underground
in tunneling with TBM and mitigation construction when using numerical methods.
measure. Scientific-technical Jurnal Mining Vietnam Jurnal of Science and Technology.
and Geology. Hanoi University of Mining and ISSN 1859-4794.Volum 61-Numbers 3-March
Geology (HUM).Nr.12 (10-2005). 60-64 2019. P.36-41. (vietnamese)
(vietnamese) Trinh Q.N. 2006. PhD Thesis: “Analysis of a
Nguyen Q.P. 2006. Risks in Underground cave-in problem in hydropower tunnel in
Construction- Causes and Mitigation Issue. Vietnam”. Norwegian University of Science
Scientific-technical Jurnal Mining and and Technology (NTNU). Trondheim,
Geology. Hanoi University of Mining and Norway.
Geology (HUM).Nr.16 (10-2006). 69-72 Trinh Q.N., Broch E. and Lu M. 2010. 2D versus
(vietnamese). 3D modelling for tunnelling at a weakness
Nguyen Q.P., Nguyen V.M., Do N.A. 2007. zone. Rock Engineering in Difficult Ground
Numerical Methods-FLAXIS 3D and UDEC. Condition-Soft Rocks and Karst-Vrkliam (ed)
Construction Publishing House, Hanoi (in Taylor&Francis Group, London, ISBN 978-0-
vietnamese) 415-80481-3.
Nguyen Q.P., 2007. Stability analysis and design Vu V.T. et al. (2014). Geohazards and prevention
of reinforcement support for underground mearsures from practical point of view.
opening in jointed rock mass with a small Proceedings of the Workshop Geohazards:
fault. Scientific-technical Jurnal Mining and Predition, Analysis and Prevention.. Hanoi 30.
Geology. Hanoi University of Mining and Mai 2014. P. 47-53. (in Vietnamese). P. 239-
Geology (HUM).Nr.16 (4-2007). 58-60 245. (in Vietnamese).
(vietnamese).

74
The 3rd Int. Conf. on Transport Infrastructure & Sustainable Development (TISDIC 2019). Construction Publishing House. ISBN: 978-604-82-2893-4.

Invited lecture 3: Reliability of SPT numbers in predicting

physical properties and strength of cohesive soil

Truong Nhu Manh 2, Duong Hong Tham1

1
Faculty of Civil Engineering, Saigon Technology University, Ho Chi Minh City, Vietnam
2
Master of Engineering

Corresponding author, E-mail: [email protected]

ABSTRACT: In-situ test has currently played an important role in geotechnical engineering and
subground technology. Beside lab tests on undisturbed samples of soil, many different kinds of site tests
was used and proved to be more efficient in design and construction such as PMT, CPT, SPT etc. Among
them, Standard Penetration Test (SPT data) is a very common and easy to carry out at site. Decades ago
SPT was proposed to be reliable to non-cohesive soil, and unreliable to fine soil because of unpredictable
and dispersed data, so that SPT was not applicable to cohesive soil. This paper aims to determine how
reliability of SPT data is in predicting physical and mechanical properties, i.e soil strength and modulus
of elasticity of fine grained soil. By analysing 40 boreholes out of 18 projects in Ho Chi Minh City, many
quantitatively significant correlation between SPT numbers and physical and mechanical properties of
soil was found.
Keywords: Standard penetration test; cohesive soil; statistical correlation; physical properties of soil;
bearing capacity of driven pile.

very fine sand or clays SPT data are so

1. INTRODUCTION scattered that people can say that its correlation
Standard Penetration Test has been currently between values in SPT and other relevant tests
an useful procedure to obtain data for analysing to be a crude approximation.
soil properties and for predicting bearing This paper assessed the level of reliability of
capacity of precast elements installed into soil SPT data in predicting physical properties and
such as piles or columns. This test has proved soil strength of fine soil.
to be reliable for non cohesive soil, especially
medium dense sand or coarse grained soil. In 2. REVIEWS ON STANDARD
some textbooks of foundation engineering, PENETRATION TEST
except some deposits containing boulders
which data from the test on those soils cannot 2.1 Uncertainties in getting proper N numbers
be used properly for investigation because of By dropping a 63,5kg weighing hammer onto a
the small size of device as compared to that of drill rod from a height exactly equals to
boulders, the test still be viewed as “rather 760mm, as per ASTM D-1586, SPT data are
unreliable” for clays and fine soil in general numbers N necessary to create a length of
(Peck, 1976). For cohesionless soils, the penetration of 300mm. To avoid errors, the rod
numbers of SPT stands for the dense states of is to be seated firmly first, so the blows for
the soils and for fine soils such as silty sand, or first 6 inches (150mm) of penetration are not

75
taken into account. So the next 150 mm is for ER
N 60  N (2)
collecting the main data. This penetration 60
might cause an increase of pore water pressure in which, N is number of blows.
for the next blows plunged into clay layers. During carrying out tests, numbers of blows
Energy or work done _ denoted as mgh, will be counted, and final SPT number will be
calculated about 473 Joule_ in penetrating the sum of n blows as algebraic average. However, as
tube 50/35mm diameter into soil is definitely for some researcher, multiple average gives more
loss due to several reasons, as follows: friction reasonable value of N than algebraic average
of hammer as gliding against the rod, string to 1
pull against pulleys, machinery friction of robe N   ( Ni ) n (3)
around cylinder, skin friction between
Based upon studies and researches carried
penetrometer and soil around it, etc. Therefore,
out on different aspects of geotechnical
a energy ratio (ER for short) is defined as
engineering, different values of ER were
following formula:
recommended as follows:
Emeasure Aoki and De’Alencar (1975) : ER =70%
ER  (1)
Etheoritical Shioi and Fukui(1985) ER=55%
A range of values about effectiveness Meyerhoff (1976) ER=55%
percentage of energy delivered to the test was Besides correction factor that will be
suggested. Kowas (1983) pointed out the mentioned later hereinafter, in Vietnamese
percentage is 55%; Seed (1985) suggested the condition about skills and experience of
60% percentage based on lots of technical technicians at site; a number 55% seems to be
reports in nationwide of America, meanwhile appropriate to use in pile bearing capacity.
Bowles suggested a number of 70% to be faily ER=55% will be chosen to determine N SPT
reasonable. It means that energy delivered to data from site investigation (Hoang T.Q and
work needs an adjustment in order to get an Tam, N.M, 2016)
effective energy. For instance, if 60% of Bearing capacity of piles can be calculated
energy ratio is suggested, it means by using SPT data with formula of Japanese
Institute of Architecture (1988)

Table 1. Formulas for pile bearing capacity using N SPT

Formula by Skin friction Point bearing Note

Aoki and De’Alencar Qs 

a.k
N s a, k = 14, 1(sand); Qp 
k
Nb
(1975) 3 .5 1.75
=60, 0.2 (clay) respectively

Shioi and Fukui

sand Q s  2 N s clay Q s  10 N s san Qp  (1  0.04 Db ) Nb clay
B
(1985)
Db
Q p  0.06 ) Nb
B

Meyerhoff (1976) Q p  0.4 N bC1C2

` Q s  n S N s ns=1
Non-
-displacement pile; =2
Displacement pile
Bazaara and Kurkur Q s  n S N s ns=2~4 Qb  nb N b nb=0.06~0.2
(1986) Nb average of N taken 1B above
and 3,75B below pile tip

76
Correction Factors Two stages of N numbers correction, i.e N
to N’ and then with silty and fine sandy soil, N’
Because of difference in efforts/energy to N”.
for various depths of penetration and
because of the difference in characteristics 2.2 Uncertainties in correction factors
of pore water pressure dissipation between Firstly, N data will be corrected to the depth of
cohesive soil as compared to that of by non testing. Correction factor depends on
cohesive soils, two correction factors were overburden pressure po. In terms of statistical
postulated to obtain the more precise number aspects, the pressure varies randomly with unit
of SPT data. weights of soil in layers of foundation.
Bazaraa (1967) proposed the following Therefore this factor also is a random
corrections to the actual count N, based on the parameter, with specified coefficient of
over burden pressure: variation COV(γ). However, unit weight has a
For po  75 kPa normal distribution and relatively small
4N coefficient of variation (less than 10% (Phoon
N' (4a) K.K, Kulhawy F.W.) Hence, correction to
(1  0.04 po )
depth of tests (N’) is more explanatory than
For po  75 kPa correction to silty and fine sand or cohesive
4N soil (N”).
N' (4b)
(3.25  0.01 po ) 2.3 Uncertainties when applying measured ,
corrected data into model
where N’ is corrected N value; N is After correction, factors are used in many
observed N-value; different formulas such as point bearing
po is over burden pressure, (kPa) = γD; capacity, skin friction and someway they are
D is depth of penetration (m); used to predict physical properties of soil. The
γ is unit weight of soil at the time of testing. question is that does SPT predict clearly
That is to correct N numbers with respect consistency of cohesive soil likethat with
to depth of testing; then if testing layer is of density of sandy soil ? The uncertainties here
fine or silt, it has to correct again because a are in the way of interpreting N data to get
temporary raise of pore water pressure can information of fine cohesive and silty soil.
resist partially the dynamic loads and the N Fortunately, literature reviews showed some
value could be higher than real value. measures to account for this question. It is to
Therefore a modification was applied to N’, use Square root of sum of square SRSS laws,
especially to observed N- value which bigger with symbolic as below (Phoon K.K, Kulhawy
than 15. F.W., 1999):
Formula for obtaining the final corrected
COV 2 (PD )  COV2 (PM )  COVm2 (PM )  COVM2 (Mt ) (6)
value if SPT is performed into saturated silt,
fine sand and silt below water table and bigger where PD is designed parameter, PM is measured
15 is: quantity (i.e from lab tests or in situ tests); index ω
1 indicates inherent natural attributes; and index m
N "  15  ( N ' 15) (5) indicates results from measurement; index M
2
indicates modeling . As such, PD stands for the
Pore water pressure has to be considered, most likely values in design.
that implies effective pressure must be taken
into account together with temporary pore 3. METHOD
water pressure increase. Hence, what
happens after the pressure dissipates ? or In order to assess the reliability of SPT
how about blow number of tests above the numbers in predicting the physical properties
water table ? and strength of cohesive soil, the procedure is:

77
 - Classify data into 4 groups: medium sand, Because SPT number cannot be only related
silty sand, clayey sand and sandy clay. to one specific property, multi variate
Soil data of 40 boreholes taken from 18 regression analysis is applied via Data Analysis
projects are tabulated as in table 2, in which tool of Excel software.
soil was again categorized into three groups:
non-cohesive soil (sand), fine soil (clay) and
cohesive soil (both clayey sand and sandy clay)
for clearly physical properties.
Because SPT data must be corrected by
transforming N into N’ for sand and N’ into N”
for clay, depth of sampling was taken into
account in all regression equations.

Table 2. Collected Data of 18 projects

Data Figure 1. Data Analysis tool in Excel.
Type Project Number
Status With level of confidence is 95%, corrected
Number of samples N is chosen to be dependent variable Y and
Sand 2,6,10,17, 18 20 Medium independent variables Xs.
(BH1) density
Clayey 1, 2, 3, 4, 6, 7, 185 Mainly
sand, 8, 9, 10, 11, plastic
sandy 12, 13, 16, 17,
clay 18 (BH1)
Clay 1,2,3,4,5,6,7 236 Semi
9,10,11,12, solid to
16,17,18 stiff, low
(BH2) plasticity
Figure 2. Regression window in Excel
- In order to find out the correlation between
N’ will be utilized for sand and N” for both
SPT numbers and soil physical properties such
sand, clayey sand and sandy clay.
as unit weight γ, moisture ω, void ratio e,
This paper will assess the reliability of SPT in
modulus of elasticity E, plasiticity index Ip and
predicting the physical-mechanical properties of
shear strength (cu, φ), data were tabulated as
cohesive soil. With soil strength parameters
described in table 3 for each soil group.
(cohesion and internal friction angle) and physical
Table 3. Data for correlation analysis
properties related to SPT number, regression
equation will be obtained and used in evaluating
Modulus of elasticity E
Overburden pressure po
Depth of sampling Z

Dried unit weight γd

bearing capacity of soil foundation or driven pile.

Plasticity index Ip
Initial voil ratio e
Water content ω

Corrected N’’
Corrected N’
Raw SPT N
Project No

Comparisons between three values :

(kN/m3)

(kPa)

(kPa)

a) of soil properties (physical parameters

and soil strength) in terms of SPT numbers
(this study).
of conventionally obtained lab tests.
2 2 0.38 13.02 1.07 0.24 35.30 22.61 2 4 10
2 4 0.39 13.03 1.06 0.25 70.80 21.49 2 3 9
of previous studies about correlations
2 6 0.29 14.55 0.84 0.17 88.44 24.69 5 5 10 between soil properties (physical, mechanical
2 6 0.33 14.19 0.90 0.19 78.63 23.26 5 5 10 properties and compressibility).
…
18 48 0.20 16.68 0.63 0.25 520.1 45.92 33 16 16 will be studied in order to choose the most
18 50 0.19 16.61 0.61 0.18 539.7 48.68 34 16 16 appropriate values to use in foundation
18 52 0.18 16.60 0.62 0.19 558.9 42.70 37 17 16
18 54 0.18 16.74 0.60 0.17 578.3 45.15 31 14 15 engineering.

78
4. RESULTS prescribed only explained nearly 46%
dependent variables N’. It meant that more
Some results were described in tables 4, 5and than 54% was due to others uncertainties in
equations of multi-variable regression are measurement, errors in lab tests etc.
shown below: These properties of sand were rather reliable
in predicting N’.
Table 4. Regression statistics and variance
Silty sand (observed 70 samples) showed no
analysis ANOVA
correlation at all.
Regression equation for clayey sand and
sandy clay
N” = 9.955 +0.036Z – 0.026 ω + 0.813γd +
0.098Ip + 0.009E (9)
R2 (R Squared adjusted) = 0.341 (slightly
smaller as that of Sand with R Squared
adjusted = 0.456) indicated that the
predictors or a few independent variables as
Results of multi variable regression for prescribed only explained nearly 34%
different kind of soil (Sand, Clay, Sandy Clay dependent variables N’. It meant that more
and Clayed Sand) are tabulated as in Table 5 than 66% was due to others uncertainties in
below: measurement, errors in lab tests etc. These
properties were relatively reliable in
Table 5. Significance of analysis of variance predicting N”.
(ANOVA) for sand, sandy clay/clayey sand, clay Regression equation for clay
N” = 78.629 -0.044Z – 82.65 ω -2 γd + 7.31Ip
Regression
ANOVA +0.048E (8)
Statistics
Type
R Adjusted Significance
square R2
F
F R2 (R Squared adjusted) = 0.177 indicated
that the predictors or a few independent
Sand (20
0.623 0.456 6.194 0.003 variables as prescribed only explained nearly
samples)
18% dependent variables N’. It meant that
Clayey more than 82% was due to others uncertainties
sand, in measurement, errors in lab tests etc.
sandy 0.363 0.341 16.905 2.07E-15 These properties were weakly reliable in
clay (185
samples)
predicting N”.
Based on soil data and corrected numbers of
Clay
(236 0.198 0.177 9.426 2.99E-9 SPT data, by conductiong multi-variable
samples) regression analysis, some results are:
- Sand : relatively usable
Regression equation for sandy soil Sandy Clay/ Clayey Sand: tentatively usable
Clay: tentatively usable with remarkable
N’= 224.9 – 0.13 Z + 0.022 ω – 7.89 γd –
139.67e + 0.02E (7) caution.
The most interesting issue was that the R2
where Z is the depth of sampling, other adjusted of both kinds of quite different
symbols are abovementioned in Table 3. characteristics, i.e sand is quite different with
In multi regression analysis, R2 adjusted clay (R2 adjusted = 0.456 for sand and = 0.341,
is used instead of (R Squared). R2 adjusted = for clayey sand and sandy clay, but for clay, R2
0.456 (or R = 67.5%) indicated that the (adjusted) =0.177. So SPT is reliable for sand
predictors or a few independent variables as and clayey sand, and weakly reliable for clay.

79
 This might be hardly applicable in For clay, there were uncertainties regarding
predicting the bearing capacity of pile installed pore water pressure raised and decreased
in cohesive soil as per The Institute of Japanese during process of penetrating the rod. There
Architecture 1988 that was well described in were different correlation between corrected
TCVN 10304:2014). factor and SPT without correction, as indicated
in Fig.3a, 3b, 3c
Correlation coefficient (R) stands for the
Correction method and formula is very
relationship between two variables, and R essential in predicting soil properties.
squared (adjusted) is defined the coefficient of
determination in multi-variable regression. 5.2 Model of analysis
Although the R squared was relatively small, Disregarding to scale of fluctuation (e.g distance
but strongly related to each other, expressed in within it, data has strongly correlated with each
very small value Significant F in the most right other), multi layer characteristics may result in
column of Table 6. weak correlation between N” and properties for
cohesive soil. As such, model required a more
Table 6. Regression statistics and variance effective tool of correction due to overburden
analysis ANOVA pressure, status of preconsolidation history
interaction between tip of penetrometer and its
shaft like 4D above pile tip and 1D lower pile tip.

 For sand moisture, dried density, initial

void ration and overburden pressure (depth
of samplings) affected most to corrected
SPT numbers instead of raw data without ,
correction.
 For clay, depth of sampling, plasticity index
did not affect SPT both N’ and N”.
Cohesion and modulus of deformation had
a slight effects on N”. This might be
unclear.
 For cohesive soil, depth of sampling
affected most significant the corrected SPT
numbers. Hence, correction was necessarily a)
obligatory.
 Corrected SPT numbers were applicable for
both sand (N’) and clay (N”), with different
coefficient of determination.

5. DISCUSSION
5.1 Correction Factor
A lot of formulas of correction factor CN for
SPT numbers for sand due to overburden
pressure. This study used Bazaraa’s formula
(1967). Small correlation coefficient may be
explained by not taking relative densitive into
account; dried unit weight was not perfectly a
criterion of density. b)

80
c)
Figure 3. Inconsistent trend of corrected SPT.
a) Borehole BH1 (all N”);b) Borehole
BH1 (some N’, other N”); c) Borehole
BH2, all corrected N” Figure 4. Correlation between SPT numbers and
There was an inconsistency in trends of shear strength parameters
corrected SPT as compared to those of raw
Regression equations for main parameters of
SPT before correction. Only two conditions of
soil strength and compressibility for different
correction, that is to say due to overburden
pressure N to N’ and then if fine sand or silty, soil are in Table 7
N’ to N”, it may lead to scattering data and and Table 7. Single variable regression equations
inconsistent trends as in Fig. 3 a, 3b and 3c.
for different kinds of soil (units in SI, i.e kPa
5.3 Reliability of lab test data and degree)
SPT data are often listed together with lab test Soil Confidence lever
data for regression purpose. Lab tests always type 85% 95%
contain errors in measurement in conventional E E=111.54+1.826N’ E=118.8+2.431N’
lab tests, undrained condition, and without any Sand C - -
necessary modification, especially disturbed φ φ =32.6°-0.01N’ φ =32.93°-.02°N’
soil samples like sand. E
Clayey E=57.67+2.38N’’ E=61.63+2.37N’’
For instance, in layers of clay, depth of
sand/ C C=0.134+0.001N’’ C=0.141+0.002N’’
sampling, plasticity index does not affect SPT Sandy
both N’ and N”, cohesion and modulus of Clay
φ φ =19.81+0.52N’’ φ =20.33+0.56N’’
deformation had a slight effects on N”. This E E=52.6+1.7N’’ E=56.7+1.939 N’’
might be inadequate when it does not include Clay C C=0.367+0.02N’’ C=0.407+0.022N’’
water content in regression equations. The fact
φ φ =15.89+0.36N’’ φ =16.5+0.39N’’
is that soil strength may slightly decrease due
to an increase of moisture. This leads to a Results obtained from abovementioned
smaller of N value. regression analysis can be compared to those of
previous works conducted by Mostafa Abdou
5.4 Single-variable regression analysis for
Abdel Naiem Mahmoud (2013) in which shear
shear strength parameters of cohesive soil
strength of silty clay with sand soil can be
At 95% confidence level: there were weak calculated as follows:
correlation between corrected SPT number N” φ (in degree) = 0.209N” + 19,68
and many different soil parameters altogether c (in kG force/cm2) = 0.014 N” – 0.18
but it was a significant correlation between N” where N” is the corrected SPT numbers,
(or N’) and E, C, φ. E in kPa.

81
6. CONCLUSIONS correction of the SPT data to use suitably in
practical design.
In general, SPT data had some slight
correlation with some properties of cohesive 7. REFERENCES
soil such as plasticity index, water content and
shear strength parameters. Some correlations Hoang T.Q and Nguyen Minh Tam. 2016.
are not clearly significant in multi variate Prediction on driven pile bearing capacity to
regression analysis, also in single variable SPT numbers and numerical method. ME
regression. The reasons for that are a) model thesis. HCM University of Technology
under consideration had to supplement some Vietnam National University.
Kudmetha, K.K., Dey, A. 2012. Uncertainty in
other variables; b) correction for data taken in
predicting bearing capacity of piles in sand
site and c) errors (uncertainties) in using SPT data. International Symposium on
measurement of lab test data must be quatified. engineering under Uncertainty,2012.01.06
In terms of soil properties (as independent Mostafa Abdoul Abdel Naiem Mahmoud. 2013.
variables), dependent variable SPT data for Reliability of using standard penetration test
sand were rather reliable to be a predictor, (SPT) in predicting properties of silty clay
except silty sand, relatively reliable for clayey with sand soil. International Journal of civil
sand and sandy clay; and weakly reliable for and structural engineering. Volume 3, No 3,
clay. Single variable regression may provide ISSN 0976 – 4399.
better correlation than that of multi-variable doi:10.6088/ijcser.201203013050 , 545-556
regression. Manh Nhu Truong. 2018. Assessment the
In order to use SPT data in computing reliability of SPT number in predicting
bearing capacity of precast pile installing into
bearing capacity of jacked-in/driven piles, it is
cohesive soil in District 1, HCMC. Master
necessary to supplement more general thesis, Open University in HCMC.
coefficients especially correction to Peck R.B, Hanson W.E and Thornburn T.H. 1976.
penetrometer, and precise data of lab tests for Foundation Engineering, 2nd edition, Mei Ya
comparison. This paper is still carrying on and Publication Inc., Taiwan pp.514.
obtaining some results for comparative Phoon K.K, Kulhawy F.W. 1999. Characterisation
analysis, currently aiming at providing relevant of geotechnical variability. Canadian
understandings about the importance of Geotechnical Journal No 36 (1999) 612-624

82
The 3rd Int. Conf. on Transport Infrastructure & Sustainable Development (TISDIC 2019). Construction Publishing House. ISBN: 978-604-82-2893-4.

Invited lecture 4: Bidirectional load test for large bored piles

of Cua Dai bridge

Do Huu Dao1, Hoang Ha Viet Phuong2, Nguyen Minh Hai3, Anand J. Puppala3,
Surya S. C. Congress3, Young Sang Kim4, Tran Quoc Thien4.
1
University of Science and Technology, the University of Danang, Danang, Vietnam
2
Quang Ngai Department of Transportation, Quang Ngai, Vietnam
3Department of Civil Engineering, The University of Texas at Arlington, Arlington, TX, U.S.A
4
Department of Civil Engineering, Chonnam National University, Gwangju, Republic of Korea

Corresponding Author: Do Huu Dao

Address correspondence to First author: [email protected]
ABSTRACT: This paper presents a case study of the loading test for Cua Dai bridge (Tra Khuc river,
Quang Ngai, Vietnam). The bridge, measured 1.876,8m in bridge length, and 2.500m in total length. With
two abutments and 37 piers supported by a number of 470 reinforced-concrete bored piles with the
diameters of 1200mm and 1500mm for approaching spans and main spans, respectively. The engineers
had to deal with a hard-working condition when testing the bearing capacity of the bored piles in a large
off-shore area, especially the safety and potential damage from the upcoming flood season in Vietnam.
Hence, the self-balanced Osterberg was employed to carry out the loading test for the D1200-mm piles
(for piers T6 and T30) and D1500-mm piles (for piers T19, T20, and T21) at the maximum testing load
13 MN and 18 MN, respectively. Finally, the test could effectively determine the sleeve resistance and tip
resistance of the reinforced-concrete bored piles, reaching up to 35MPa and 9 MN, respectively. As the
result, the Osterberg testing method did help reduce the constructing time of the understructure parts.
Keywords: Loading test, sleeve resistance, tip resistance, Osterberg test, load box.

1. INTRODUCTION The Osterberg method, based on the self-

Cua Dai bridge was constructed 1.500 m far balanced mechanism of Professor Osterberg
from the coastal line, crossing the river mouth (1989) (Fellenius, 2018), was created to solve
of Tra Khuc river. It includes 37 spans as the mentioned problems with higher
shown below: workability than the traditional methods in all
(39+440)+(640)+(540+39)+(75+4120+7 aspects. The loading method was performed in
accordance with ASTM D1143 (ASTM 1134,
5)+(39+440)+(540)+(340+39)m. With length
2007). After that, it was improved and
Lc = 1.876,8m, width of lanes 4x3,5=14,0m, and
developed by M. G. England, P. F. Cheesman
pedestrian lanes 21,5=3,0m. The designed (England and Chessman, 2011), and finally
bearing capacities are 6MN and 9MN for the standardized complying to ASTM D8169
D1200-mm and D1500-mm piles, respectively. (ASTM D8169/D8169M, 2018). The Osterberg
The construction work was in river mouth site method has been applied widely such as the
and water level changed continuously due to the case studies of John H. Schmertman, John A.
effect of tides, hence, making the working space Hayes (1997), Roberto Nova, Bruno Becci (2008),
and supplying road were very challenging to the Tan S.A and Fellenius B.H (2012), N.M. Hai – D.H.
engineers, especially the loading test. Thus, Dao (2013), Bengt H. Fellenius, & Ann, T.S.
applying a suitable method of loading test for the (2014) (Chang and Zhu, 2004; Nova and Becci,
reinforced-concrete bored piles is a very 2008; Tan and Fellenius, 2012; Nguyen and
important decision for whole of the project. Do, 2013; Fellenius and Ann, 2014).

83
 the river and sea, respectively, making a hard
condition for engineers to carry out the loading
test for the bored piles. The river mouth is
quite wide, it is about 1.350m with the
riverbed’s elevation ranging from -0.4m to -3.3
m, the listed piers T19, T20, and T21 are
located in the middle area of the riverbed. In
this case, a suitable solution to avoid the effect
Figure 1. Five main towers of Cua Dai Bridge are of the underflow and surface flow is using
building Ocell test for the loading test.
The self-balanced Osterberg was employed to In this project, the earthquake-resistant
carry out the loading test for the D1200-mm design was performed and applied, which is
piles (for piers T6 and T30) and D1500-mm designated at the magnitude of VII and
piles (for piers T19, T20, and T21) at the a = 0.0912 complying to TCVN 11823 (2017)
maximum testing load 13 MN and 18 MN, (TCVN 11823, 2017).
respectively. The reinforced-concrete bored
piles were equipped strain gages at 4 different 3. PILE CONSTRUCTION AND TEST
locations to measure the deformation along the PROGRAM
piles’ body and analyze the skin friction of the
piles’ surface, especially, the piles placed in the The bores were drilled by using a special
bedrock. A comprehensive case study is machine, which has a high capacity to use in the
presented in this paper to show the effectiveness various ground condition. When facing the
and convenience of the Osterberg testing bedrock, the driller can use a special tip with high
method applied for Cua Dai - one of the largest stiffness to bear large moment to penetrate the
bridges in the central area of Viet Nam. rock. Installing Ocell testing equipment was done
along with the manufacturing of steel cages.
2. SOIL AND HYDROGEOLOGY PROFILE The main equipment of the loading test
includes 03 hydraulic jack with a capacity of 3
The geological characteristics of the ground MN/jack for D1200 piles (maximum capacity
beneath the constructed piers T6, T19, T20, 3MN  03  2 = 18 MN) and 03 hydraulic jack
T21, and T30 are described as following: layer with a capacity of 5 MN/jack for D1500 piles
1 – relative dense fine sand with the thickness, (maximum capacity 5 MN  03  2 = 30 MN)
ranged from 5,1m to 10,8m; layer 2 – dense for a testing time. The maximum capacity of
medium sand – a product of weathered granite jacks must be higher than the values of 138,5%
soil, with a thickness, ranged from 10,5 m to and 167,7% of the applied load for the D1200
11,6 m, there was a relative dense silt layer (2a) and D1500 piles, respectively. The bottom and
surrounding the area beneath the pier T30; layer upper part of load box are made of steel with
3: weathered granite rock, TCR = 35%, RQĐ = 1250 mm in diameter and 50mm in thickness.
0, the thickness of layer 3 ranged from 1,9 m to Drilled holes on the surface with D250 mm in
2,2 m; layer 4: weathered granite rock, TCR = diameter were made for cleaning the piles’ tip
(58-70)%, RQĐ = (50-57)%, the thickness of and pouring concrete into the boreholes. The
layer 4 ranged from 8,5 m to 10,9 m. layer 5: load boxes were installed at the elevation 0,5m
granite bedrock. Based on the ground condition, from the bottom for 3 testing piles to measure the
the designed bored piles were embedded into tip resistance and the sleeve resistance. Wires
layer 4 – weathered granite rock. Figure 2 with 8mm in diameter were installed inside the
shows the geological section of the ground of steel pipe with 21mm in diameter linked the
the construction site. bottom and upper sides of the box with the top of
Hydro-geological characteristics: As bored piles for measuring the displacement of the
mentioned above, the bridge was constructed bored piles’ tip and body at the installed-Ocell
1.500 m far from the coastal line at the river points. Strain gages are vibrating sensors to
mouth. Hence, the understructure of the bridge measure the deformation along the piles’ body.
is strongly affected by the flow and tide of both The testing program is presented in Figure 3.

84
 Figure 2. Bore holes of test piles: T6, T19, T20, T21 and T30

Figure 3. Details of instrumented for test piles

25 couples of circular steel bars with 25mm Table 1. Test pile's parameter and time of
in diameter and 26 couples of circular steel experiment
bars with 32mm in diameter were used for the
D1200 mm and D1500 mm piles. At the load No Test pile Construction day Testing day
box, the circular bars were welded onto the 01 C1-T6 st
1 April, 2018 26th April, 2018
surface of the box. Others equipment were
installed into the piles including 03 pipe D56 02 C10-T30 19th April, 2018 15th April, 2018
mm and 01 pipe D114mm for measuring 03 C13-T19 26th June, 2018 1st August, 2018
ultrasound and checking the contact between
pile’s tip and poured concrete. Table 1 shows 04 C2-T20 14th May, 2018 16th June, 2018
the test pile's parameter and time of the 05 C4-T21 25th April, 2018 16h May, 2018
experiment.

85
 Installing procedure (ASTM D8169/D8169M,
2018)
1. Installing the bottom flange
2. Locating and installing 03 5-MN jacks
3. Installing the upper flange and weld to
make the load box
4. Installing the load box into the steel cage
5. Installing the D21 pipes and displacement
-measuring bars Talltel
6. Welding the D21 pipes onto the upper
flange
7. Welding the D21 pipes onto the bottom a) Positioning jacks on the steel plate
flange
8. Installing strain gages along the steel
cage’s body
9.Checking the workability of strain gages
10. Checking for the last time at all
components
Steel cage’s installation and concrete
casting
1. Drilling the bored holes with bentonite
2. Checking the depth of bored holes’
bottom b) the load cell box
3. Cleaning bored holes’ bottom
4. Lowering the first part of the steel cage
including load box into the bored holes
5. Lowering the second part of steel cage
6. Lowering the third part of steel cage
7. Checking the steel cage installation
8. Cleaning bored holes’ bottom again
9. Concrete work
10. Checking the quality and curing the
concrete up to 21 days.
Ocell testing procedures (ASTM D1143,
2007) c) Installing the load cell box in the steel cage

1. Making working space (6m x 6m)

2. Installing the loading frame
3. Installing 06 gages to measure the
deformation of the flanges and pile’s cap
4. Installing measuring apparatus of the
vibrating strain gages
5. Loading with an interval of 10% of
designed bearing capacity up to 200% of it.
Record the displacement and deformation
values.
6. Finish the test and analyze data. d) Lowering the steel cage to the drill hole
The testing process is presented briefly in
Figure 4. Figure 4. Images of manufacturing and lowering
the load cell box down the drill hole

86
 Pictures about loading process are shown in process with that of pile C13-T19, the total
Figure 5. displacement at the upper flange, the pile’s
tip, and pile’s cap was 3,095mm, 4,795mm,
and 1,145mm, respectively. The relationship
between loading and displacement,
displacement and elapsed time are shown in
Figure 7a and 7b, respectively.
With the maximum loading is up to two
times of the designed bearing capacity, the
displacement of pile’s body and pile’s tip were
in the acceptable values. Besides, the
settlement of the piles’ tip was smaller than 5
mm. It indicates that the bored piles were
working in a good condition that it could
a) Installation of measuring equipment mobilize its maximum bearing capacity to
serve the supper-structures.

5
Movement of
4 the Top plate

3
2 Movement of
pile head
1

MOVEMENT (mm)
0
-1 0 0.9 1.8 2.7 3.6 4.5 5.4 6.3 7.2 8.1 9
LOAD (MN)
-2
-3
Movement of
-4 bottom plate

-5
-6

b) Loading and recording data -7

Figure 5. Loading for the tested pile a) Load-movement curves of the pile

4. TEST RESULTS AND ANALYSIS

5
4
This paper presents and discusses the testing 3
results of 02 piles D1500 C13-T19 and C2- 2
Movement of top plate
MOVEMENT (MM)

T20. 1
0
Bearing capacity results and pile’s
-1
displacement -2
For case of pile C13-T19, with the -3
maximum load 18 MN (9 MN affected to -4
-5
pile’s body and 9 MN affected to pile’s tip), -6
Movement of bottom plate

the total displacement at the upper flange, -7

pile’s tip, and pile’s cap were 3,645mm, 0 50 100 150 200 250 300 350 400
6,415mm, and 1,385mm, respectively. The TIME (MIN)

relationship of loading and displacement,

displacement and elapsed time are shown in b) Movement-time curves of load cell box
Figure 6a and 6b, respectively. For the case
of pile C2 - T20, with the same loading Figure 6. Result of tested pile C13-T19

87
 5 Based on the collected deformation of the
4 Movement of
Top plate
piles, load at the depth where strain gages
3 installed is calculated as the following formula:
2 Movement
of pile head Pi   i .E. A (1)
1
Where: Pi – load at the depth i;  i –
MOVEMENT (mm) 0

-1
0 0.9 1.8 2.7 3.6 4.5 5.4 6.3 7.2 8.1 9 deformation when the loading is changed; E –
modulus of elasticity; A – a cross-section of piles.
LOAD (KN)
-2

-3
Movement of
bottom plae The load distribution along the pile’s body
-4 with respect to intervals of loading of pile C13-
-5 T19 and pile C2-T10 are presented in Figures
-6 10 and 11, respectively. The maximum load
was 9MN at the pile’s tip, then decreased along
a) Load-movement curves of the piles
with the depth due to the effect of skin friction,
4
it is zero at the ground surface.
3 The figures show that the loading distributions
2
Movement of top plate
along the plie’s body were different between the
1 loading box and ST4 location. This is because
MOVEMENT (MM)

0 those piles’ part were embedded into the rock

-1 layer, hence, the skin friction value of piles at
-2
this depth is much larger than that of piles
-3
located in the soil layer. The rock layer can bear
-4
-5
Movement of bottom plate and absorb the load effectively, thus, the load
-6 significantly decreased after passing layer 3 and
0 50 100 150 200 250 300 350 400 4. The load decreased around 63,1% and 71,94%
TIME (MIN) from the loading box to ST4 location for the case
b) Movement-time curves of load cell box
of pile C13-T9 and C2-T20, respectively.
70
Figure 7. Result of tested pile C2-T20 ST #4
ST #3
60 ST #2
ST #1
Shear resistance discussion 50
Strain ∆μe

Strain gages is a vibrating sensor in a form 40

of wires, which are symmetrically installed at 30

the designed depth (ST1 (-5m), ST2 (-9m), 20

ST3 (-14m) and ST4 (-19m)) in case of pile 10

C13-T19. Databox Geokon box will collect the 0

0 1800 3600 5400 7200 9000
displacement and deformation data, which are Load (kN)

shown in Figure 8. In case of pile C2-T20, the Figure 8. Load - strain curve of TP C13-T19
installed strain gages’ depth was as follow,
ST1 (-8m), ST2 (-12m), ST3 (-17m) and ST4 50
ST #4

(-21m), relationship of loading and piles’ 45

40
ST #3
ST #2

deformation is shown in Figure 9. Figures 8 35

ST #1
Biến dạng ∆μe

and 9 indicate that the deformation of tested 30

piles was similar and not critical. However. At 25

20
the depth of -19m for pile C13-T19 and -21m 15

for pile C2-T20 (the depth where ST4 10

5
installed), the deformation of piles was much 0

significantly increased. This reveals that the 0 1800 3600

Load (kN)
5400 7200 9000

sleeve resistance of the piles contacted rock

layer 3 is significant. Figure 9. Load - strain curve of TP C2-T20

88
 Load (kN) The mobilization of average shear resistance
0 1800 3600 5400 7200 9000 between bored piles C13-T19 and C2-T20 and
-1
soil with respect to depth is shown in Figures
12 and 13, respectively. Especially, it was
-3
241,7kPa for the case of pile C13-T19 (RQD =
-5 45%-60%, TCR = 50%-60%); and 229,1kPa
-7
for the case of pile C2-T20 (RQD = 40%-50%,
TCR = 40%-55%). These values are a good
-9
agreement with previous studies (Carrubba,
Depth (m)

-11 1997; Li et al,2011). The total displacement

and skin friction of soil layers are shown in
-13
Table 2.
-15
300
-17
250 18-23.0m

-19
200
-21

fs(kPa)
150
-23
100 14-18m

Figure 10. Load tranfer curve of TP C13-T19 50

9-14m

5-9m

3-8m
0
Load (kN)
0 1800 3600 5400 7200 9000
0 1800 3600 5400 7200 9000
Load (kN)
-0.5

Figure 12. Shear resistance of TP C13-T19

-2.5

-4.5 250
21-24.5m

-6.5 200

-8.5
150
fs(kPa)

-10.5
Depth (m)

100
17-21m
-12.5
12-17m
50
7-12m
-14.5
0-8m
0
-16.5 0 1800 3600 5400 7200 9000
Load (kN)
-18.5

-20.5 Figure 13. Shear resistance of TP C2-T20

-22.5
Load - movement curve equivalent:
-24.5 P-S method is performed by accumulating
the corresponding load with respect to the
Figure 11. Load tranfer curve of TP C2-T20
displacement of the pile’s tip. Extrapolation
Shear resistance between piles and soil method could be applied to estimate the
This can be determined as follow: further data by non-linear regression equation
Pi Pi based on the given data. Figure 14 and 15
1
fi = (2) present P-S relationship for pile C13-T19 and
U.l i
C2-T20, respectively. Figures show that the
Where: Pi and Pi-1 are load at the elevation i maximum bearing capacity of the bored piles
and i-1 (kN), U is pile’s perimeter (m), li is could be larger than 18 MN and expectedly
calculated segment of piles. up to 25 MN.

89
Table 2. Summary of depth - displacement and shear resistance of test piles.

TPC13-T19 TP C2-T20

Displacement Shear resistance Displacement Shear resistance

Depth (m) Depth (m)
(mm) (kPa) (mm) (kPa)

-3 -> -5.0 1.385 18.2 -3 -> -8.0 1.145 13.7

-5 -> -9.0 1.58 29.0 -8 -> -12.0 1.49 17.4

-9 -> -13.0 2.23 39.4 -12 -> -17.0 2.05 27.7

-14 -> -18.0 2.72 75.9 -17 -> -21.0 2.66 75

-18.0 -> -23.0 3.645 241.7 -21.0 -> -24.5 3.095 229.1

strongly affected by flows and tides. Osterberg

0 3.6 7.2
LOAD (MN)
10.8 14.4 18 21.6 25.2
method can be a good candidate to carry out
0 the loading test of bored piles instead of the
-2 traditional methods in most of working
-4
-6
conditions.
Head load test
MOVEMENT (MM)

-8 Osterberg method could classify the tip

-10 resistance and sleeve resistance based on its
-12
-14
Extraplation line
special working mechanism.
-16 Osterberg method did show the
-18 displacement along the bored piles and could
-20
-22
analyze the behavior of bored pile when
loaded. Therefore, this is a dominant advantage
Figure 14. Load - movement curve equivalent of which have not been done in case of the
TP C13-T19 traditional testing methods.
The skin friction between bored pile and
LOAD (MN)
rock layers was shown and analyzed. It was
0 2 4 6 8 10 12 14 16 18 20 22 24 26 28
241,7kPa in case of pile C13-T19 (where RQD
-1
-3
= 45%-60%, TCR = 50%-60%) and 229,1kPa
-5
Head load test
in case of pile C2-T20 (where RQD = 40%-
50%, TCR = 40%-55%).
MOVMENT (MM)

-7
-9
-11
Load - movement curve equivalent was
-13 Extrapolation line
plotted effectively. Besides, based on the
-15 extrapolation, the results could estimate the
-17
maximum bearing capacity of the bored piles
-19
-21
could even be larger than 18MN as designed.

Figure 15. Load - movement curve equivalent of ACKNOWLEDGMENTS

TP C2-T20
This work was supported by Projects
5. CONCLUSION management investment in the construction of
transport works Quang Ngai province, the
The self-balanced Osterberg test performed at Center of Research, Application and
Cua Dai bridge – Quang Ngai is the suitable Consultancy of Foundation Engineering -
method for the scope of large-diameter piles Danang University of Science and Technology.
bearing huge burdens in the area where is We also thank the cooperative research

90
program with Arlington Texas University static load testing. Proceedings of the 15th
(USA) and Honam Regional Infrastructure European Conference on Soil Mechanics and
Technology Management Center - Chonnam Geotechnical Engineering. 855-860.
National University (Korea) for cooperating Minh-Hai, Nguyen – Huu-Dao, Do (2013). Non-
and discussing this research results. Conventional Pile Loading Tests in Vietnam.
Proceedings of the 18th International
Conference on Soil Mechanics and
REFERENCE Geotechnical Engineering, ISBN 978-2-
ASTM D1143 (2007). Standard Test Method for 85978-477-5 (vol. 4). Pages: 2747-2750.
Piles Under Static Axial Compressive Load. Paolo Carrubba (1997). Skin friction on large-
American Society for Testing and Materials. diameter piles socketed into rock. Canadian
ASTM D8169/D8169M (2018). Standard Test Geotechnical Journal. 230.240.
Methods for Deep Foundations Under Bi- Roberto Nova, Bruno Becci (2008). Experimental
Directional Static Axial Compressive Load. and Numerical Assessment of Osterberg Load
American Society for Testing and Materials. Tests On Large Bored Piles in Sand. 11th
Bengt H. Fellenius (2018). Basics of Foundation International Deep Foundation Institute
Design. Electronic Edition. 2375 Rothesay Conference Proceeding – New York , 225-233.
Avenue Sidney, British Columbia, Canada, Tan S.A, Tan and Fellenius B.H (2012). Failure of
V8L 2B9. a Barrette as Revealed in a Bidirectional Test.
Bengt H. Fellenius, & Ann, T.S (2014). Analysis Geotechnical Publication, 307-321.
of Bidirectional-cell Tests for Icon TCVN 11823.2017. Highway Bridge Design
Condominiums, Singapore. Specification. Part 10 – Foundation.
Ming-Fang Chang, P.E., M. ASCE,1 and Hong W. W. Li, c. T. Wong, k. P. Yim, m. K. Leung
Zhu (2004). Construction Effect on Load (2011). Shaft Friction between Marble and
Transfer along Bored Piles. Journal of Concrete in Rock Socketted Large Diameter
Geotechnical and Geoenvironmental Bored Pile: A Review of State of Art.
Engineering © ASCE. 426-437. Procedia Engineering, Published by Elsevier
M. G. England, P. F. Cheesman (2011). Optimum Ltd. 1752-1758.
loading specifications for O-cell bi-directional

91
The 3rd Int. Conf. on Transport Infrastructure & Sustainable Development (TISDIC 2019). Construction Publishing House. ISBN: 978-604-82-2893-4.

Bearing capacity analysis of bore pile in weathered rock in

central Vietnam

Nguyen Chau Lan1, Le Duc Tien2, Bui Tien Thanh3, Nguyen Ngoc Long4
1,3,4
Civil engineering Department, University of Transport and Communications, Hanoi, Vietnam
2
PhD student, University of Transport and Communications, Hanoi, Vietnam


Corresponding author, E-mail: [email protected]

ABSTRACT: Weathered rock or IGM (Intermediate Geo Materials) is still a controversy in designing
bearing capacity of the bored pile. At present, Vietnamese standards separately defined the load capacity
in soils and rocks. That leads to both underestimation and overestimation in case IGM assumed to be soil
and rock respectively. In this paper, the calculation in a project in Central Vietnam (Quang Tri province)
was based on experienced equations. Furthermore, static load tests compression test and pull out the test
was conducted for bored piles with 0.8m in diameter which installed in the weathering rock. In these test,
instruments were installed including the strut meter, strain gage, and extensometers which were
distributed along bored piles to measure a side bearing capacity and tip capacity of the piles. In addition,
bored piles were modeled by a finite element method to compare with field tests.
Keywords: Weathered rock, IGM, bearing capacity, finite element program, static load test.

1. INTRODUCTION Intermediate Geological Material (IGM) into

design calculations. Normally, SPT-N value or
Currently in Vietnam, the design of drilled unconfined strength of IGM material can be
shaft is applied according to AASHTO LRFD used to calculate the bearing capacity of the
standard for bridge projects or Vietnam drilled shaft (Hayes and Simmonds 2002;
standards for high rise building Hwan 2015; Long 2016).
projects (AASHTO 2012). In these standards,
the estimation of the load capacity of the pile is Several researchers have proposed methods
divided into two cases: when the pile is placed to calculate the bearing capacity of the drilled
in the soil, the formulas related to the shaft into the IGM layer (Hayes and Simmonds
resistance of the ground is used, such as the 2002)(Carter 2014; Horvath, R. G., and
internal friction angle of sandy soil and Kenney 1979; Lianyang and H. 1998; Mayne,
unconfined strength of clay soil and when the P.W. and Harris 1993; Rowe and Armitage
pile is installed to the rock, it is possible to use 2010). However, there are still limited studies
the compressive the resistance of rock. consider the bearing capacity of the pile where
However, for the central region, such as Quang the pile tip installed in IGM layer in Vietnam.
Tri province, there is a limited experience for Thus, this paper presents the results of static
the case drilled shaft installed in a soft rock loading test and tension test for 800 mm in
layer with a strong weathered rock. This leads diameter of the drilled shaft in IGM layer. In
to difficulties for applying design and testing addition to, Finite Element Method (FEM) also
standards. In the case of weathered rock, conducted for comparison testing result and
previous studies introduced the concept of FEM results.

92
2. METHODS devices including pile head displacement,
loadcell, pile deformation measurement,
2.1 Static loading test and pull out test extensometers are shown in Figure 1.
2.1.1 Geological condition Load cell, displacement and data logger
Layer 1 to 3 is sandy clay and layer 4 is
claystone (strongly weathered rock), thickness Layer 1: sandy
Rebar meter: 4
and other properties and be found in Table 1
and Fig.1. Layer 2: very soft
Layer 3: Sandy clay mixed with gravel
2.1.2 Static loading test
Drilled shaft, D=0.8m; L=14.5m
Test pile of 800mm diameter has a penetration
depth of 14.5m from flatform level with 5.6 m Clay stone, strong weathered
of rock socket length into claystone of RQD Extensometer: 3 positions
from 20 to 30.
The load applied to the pile head
corresponds to each level of load Figure 1. Instruments along the drilled shaft pile
corresponding to the% of the design load. The
design load for the drilled shaft is 153 ton a. Displacement
according to AASHTO LRFD 2007 standard. Displacement measuring device (DT-100A;
The maximum load for static load test is 150% KYOWA, JAPAN) with an accuracy of
of the design value. The instruments for both 0.01mm and maximum displacement of 10cm,
static load test and pull out the test are installed is connected to the data logger system, and
along with the pile. For each pile, measuring fixed to the standard beam. There are 4
devices including pile head displacement, displacement measuring devices at the top of
loadcell, pile deformation measurement, the pile (Figure 2). Displacements were used
extensometers are shown in Figure 1. for measuring the displacement of pile head.

2. METHODS
2.1 Static loading test and pull out test
2.1.1 Geological condition Figure 2. Measuring the displacement of pile head

From layer 1 to layer 3, soil is is sandy clay. b. Extensometer

Layer 4 is claystone which is strong weartherd Extensometers are steel bar and were installed
rock. The thickness, mechanical properties and at top, center and tip of the pile. These
other properties are found in Table 1 and Fig.1. instruments were used for measuring
displacement of pile along the pile’s length.
2.1.2 Static loading test
The static loading test was conducted for a
drilled shaft pile. The diameter of pile and length
of pile is 800mm and 14.5m, respectively.
The load was applied to the pile head. The
design load for the drilled shaft is 153 ton
according to calculated value following
AASHTO LRFD 2007 standard . The
maximum load for static load test is 150% of
the design test. The instrument for both static
load test and pull out the test are installed
along with the pile. For each pile, measuring Figure 3. Extensometer equipment for measuring
displacement of pile

93
c. Strain gage Ec: elastic modulus of concrete;
Strain gages were arranged at 4 different c: deformation of concrete at each depth is
depths and 2 strain gages were attached to each determined by the sensor (strain gage).
depth. A total of 8 strain gages were used for The side resistance is calculated according
measuring the displacement of concrete. Strain to the equation below:
gage was attached on the main reinforced bar f = (Pi-Pi-1) / A = (Pi-Pi-1) /π.D.L (2)
(Figure 3). where Pi: force at ith elevation; Pi-1: force at
(i-1)th
elevation; A is the area; D is the pile
diameter; L is the distance between two strain
gage mounting positions.
2.2 Finite element method
FB-Pier is a nonlinear finite element analysis
program capable of analyzing multiple bridge
pier structures interconnected by bridge spans.
Figure 4. Stran gage for concrete displacement The program was developed by Bridge
Software Institute and university fo Florida.
2.1.3 Pullout test FB-pier was used for modeling compression
load test and pull out test of the drilled shaft
The schematic of pullout test for 800 mm in (Fig 5). In this model, soil -pile interaction can
diameter of the drilled shaft is shown in Figure be modeled as p-y curves and t-z
4. The design load for this drilled shaft was curve. Layers 1, layer 2 and layer 3 can be
about 173 ton, according to calculation from modeled as clay. The main parameters are unit
AASHTO LRFD standard [1] . weight, undrained strength Su (See Table 1).
Load cell
steel plate
For the fourth layer of soil is classified as
1200x1200x100
strongly weathered claystone, which is
modeled as Weak Rock (Reese) model with the
Hydraulic
1160

jact 500T 2 steel I900,

L = 10m
calculation method (Axial / Torsional) form of
900

Drilled Shaft IGM and Hyperbolic in FB-pier.

4110

Concrete Bar 16Ø25

1000

The parameters of this layer are following:

 Unit weigh=19 kN/m3
Measuring
1000

system

 RQD=20%
Drilled shaft
 Modulus = 398Mpa
4800 800 4800  Unconfined strength = 4.8 MPa
These parameter are inputed to FB-pier as
Figure 5. Pull out test
shown in Fig 6.
2.1.3 Analysis data from static test and pull
out test
The obtained data such as deformation of
concrete and displacemtnt of the pile at loading
stages can be used for analyzing side resistance
and tip resitance. The applied force at each
depth is calculated according to the formula:
P=c.Ec.Ac (1)
where
Ac: area of concrete, is the cross-section
area of the pile and the area of reinforcement; Figure 6. Model drilled shaft in FB-pier

94
Table 1. Soil parameters for FB-pier model
Unconfin
Thick Undrained Unit
Layer ed
ness strength Su weight
strength
(m) (kPa) (kN/m3) (kPa)
Sandy
1 2.2 24 18 -
clay
Soft clay
mixed
2 1.3 36 18 -
with
sand
Sandy
clay
3 8.1 180 18 -
mixed Figure 9. Side resistance vs. depth for pull out test
gravel
Clayston
e, strong
4 weathere 5.6 - 19 4800
d

Figure 7. Input parameters for claystone layer.

Figure 10. Load vs. settlement for compression test
3. RESULTS
3.1 Loading test result
According to measured value of strain gage
and displacement, side resistance and tip
resistance can be drawn with a depth of pile
(Fig. 7 and Fig.8).

Figure 11. Load vs. settlement for pull out test

Load and settlement curve for compression

and pull out test are shown in Fig. 9 and Fig
Figure 8. Side resistance vs. depth for
10. ,respectively.
compressition test

95
3.2. FEM method results Table 3. Comparision results for pull out test

Pull out test

LRFD Measured FEM result

method value
Tip
resistance 0.1728 0.1662 0.1511
qp (MPa)

Table 2 shows the comprarision result of

theory (LRFD), measument and FEM model. It
is clear that for IGM layer 4, the side resistance
of pile is consistance for 3 methods, whereas
the tip resistance for LRFD method is higher
Figure 12. Loading –settlement curve for compression
compare to other method.
test
Table 3 shows the tip resistance results for
pull out test. In this table, the value of LRFD
mehod also is higher than that of measured
value and FEM value.
Both compression and pull out the
test showed that the value of the settlement of
the pile into the weathered rock is quite small.
3.4 Side resistance correlation
According to AASHTO standard, side
resistance of pile in IGM layer can be
calculated as the following equation
qs=α. .qu (3)
where qu is unconfined compression strenth
Figure 12. Loading - settlement curve for Pull out test of IGM Several correlations have been
developed by various authors for estimating
From FEM results, load and settlement unit side resistance in IGM layer and
curve for compression and pull out test are also relationships to unconfined compressive
present in Fig. 11 and Fig 12. , respectively. strength qu as well as SPT N blow count. The
3.3 Comparison results methods adopted by AASHTO use qu as the
main design parameter as shown in equation 3.
Table 2. Comparision results for compression test Base on testing result from static load test
(Fig.6), qs can be corred to qu as equation below:)
Compression test
qS (correction) = 0.81. qu (4)
LRFD Measured FEM
method value result 4. CONCLUSIONS
Side
resistance 0.1728 0.1096 0.1118
From testing and FEM results, several
qs (MPa) conclusions can be draw the following:
- Side resistance are similar for all method
Tip
resistance 0.7094 0.48 0.5542 caculation. However, LRFD method provided
qp (MPa) the higher value of side and tip resitance.

96
 - A correction value for drilled shaft into the Hwan, Seung. 2015. “Evaluation of Ultimate
weathered rock layer (IGM) in Quang Tri Bearing Capacity of Drilled Shaft Socketed in
province is proposed. But more field tests Weathered Zone.”
shoud be conducted to provide a precise results Lianyang, Zhang, and Einstein Herbert H. 1998.
for practical design. “End Bearing Capacity of Drilled Shafts in
Rock.” Journal of Geotechnical and
Geoenvironmental Engineering 124(7): 574–
5. REFERENCES 84. https://doi.org/10.1061/(ASCE)1090-
AASHTO. 2012. Standard Specifications for 0241(1998)124:7(574).
Highway Bridges. Washington, D.C. Long, James H. 2016. “Static Pile Load Tests on
Carter, John P. 2014. “Analysis and Design of Driven Piles into Intermediate-Geo
Drilled Shaft Foundations Socketed into Materials.” (0092).
Rock : Final Report.” (May). Mayne, P.W. and Harris, D.E. 1993. Axial Load-
Hayes, Jack, and Tony Simmonds. 2002. Displacement Behavior of Drilled Shaft
“Interpreting Strain Measurements from Load Foundations in Piedmont Residuum. Research
Tests in Bored Piles.” Proc., 9th International Report No. E-20-X19 to FHWA by Georgia
Conference on Piling and Deep Foundations: Tech Research Corp, Atlanta.
663–69. Rowe, R. K., and H. H. Armitage. 2010. “A Design
Horvath, R. G., and Kenney, T. C. 1979. “Shaft Method for Drilled Piers in Soft Rock.”
Resistance of Rocksocketed Drilled Piers.” In Canadian Geotechnical Journal 24(1): 126–42.
Proc., Symp. on Deep Found., ASCE, New
York, , 182– 214.

97
The 3rd Int. Conf. on Transport Infrastructure & Sustainable Development (TISDIC 2019). Construction Publishing House. ISBN: 978-604-82-2893-4.

Finite difference method to determine lateral load bearing

capacity of concrete pile by taken into account soil and pile
interaction

Dang Hong-Lam 1,, Nguyen Duc-Hanh 1,

1
University of Transport and Communications, No.3 Cau Giay Street, Hanoi, Vietnam

Corresponding author, E-mail: [email protected]

ABSTRACT: The lateral load bearing capacity of piles is an attractive issue in pile foundation. Some
first studies were proposed by Broms (1964), Matlock (1970), Reese et al (1974) to determine the
ultimate lateral load bearing capacity and they were favourite for tentative estimation of the capacity of a
single pile with many prior assumptions. Recently, some numerical methods such as finite difference
method or finite element method were proposed and studies with the consideration of soil interaction
around pile. This interaction mainly governs the distribution of shear load and moment load of
longitudinal pile and it normally presented by p-y curves. Finite difference method is utilised in this paper
to determine the lateral load bearing capacity of a pile with some soil and pile interaction curve models.
The calculated results are then compared with the results from a widely used FB-Pier program. Finally,
comments and recommendations are given for lateral load bearing capacity determination.
Keywords: Lateral load bearing capacity of pile, finite difference method, FB-Pier, soil and pile
interaction.

the case for short piles, and (b) bending of the

1. INTRODUCTION pile, governed by the plastic yield resistance of
Several methods have been proposed to solve the pile section, which is applicable to long
the problem of a laterally loaded pile. The piles. The procedure was presented in the form
problem can be generally defined as computing of design charts and tables. Using his procedure,
pile deflection and bending moment as a a single pile which was either free-headed or
function of depth below the ground surface. In perfectly fixed against rotation could be
literature, soil models are mainly classified into analyzed (Broms, 1964a,b; Heyer and Reese,
three main types: Ultimate Load Analysis by 1979). In this model, lateral load on pile causes
Broms’s Method (Broms, 1964a,b; Heyer and pile's displacement, and arising reaction load in
Reese, 1979), Methods based on theory of lateral direction on pile meet limited value
elasticity by Winkler model (Poulos, 1971; correlative with active limited state and passive
Matlock and Reese, 1960; Davisson, 1970) limited state. Only earth passive pressure is
and Methods based on theory of non-linear examined and divided into two separate cases:
behaviour by P-y model. (Matlock, 1970; cohesive soil (φ=0) and cohesionless soil (c=0).
Reese et al, 1974; Pradhan, 2012). Passive pressure coefficient of soil is examined
For laterally loaded piles, Broms (1964a,b) by limit state theory of Rankine
developed a simplified solution based on the An elastic solution for the problem of a single
assumptions of (a) shear failure in soil, which is pile subjected to lateral loading was presented by

98
Poulos (1971). Poulos assumed the soil to be an curves provided in literature will be used and
elastic, homogeneous, isotropic half-space with a some techniques is presented to estimate the
constant Young's modulus and Poisson's ratio. lateral bearing capacity of concrete pile. Thus,
The pile was modelled as a thin, rectangular, the main purpose of this paper to help
vertical strip, with soil pressures constant across engineers to have a better understand p-y curves
the pile width. A general method for determining and to apply in practice.
moments and displacements of a vertical pile
embedded in a granular soil and subjected to 2. CALCULATION OF PILE LATERAL
lateral load and moment at the ground surface BEARING CAPACITY BY FINITE
was given by Matlock and Reese (1960). Under DIFFERENCE METHOD
simpler Winkler’s model, an elastic medium (soil
2.1 Finite difference technique
in this case) can be replaced by a series of
infinitely close independent elastic springs. From curve equation of pile's axis described via
Based on this assumption, differential fourth lever shown by formula (3),
k z = px (1) the pile was divided into n sections as Figure
where kz is modulus of subgrade reaction; p 1with same section length L. Replacing soil -
is pressure on soil; x is deflection pile interaction by disconnected links at (n+1)
The subgrade modulus for granular soils at a points numbered from 0 to n.
depth z is defined as Differential equations of pile axis for n+1
k z = nh .z (2) point are written as equation (3).
d4y
where nh is a parameter depending on the soil EJ .  b.c. y 
d .z 4 (3)
properties. In current practices, for cohesive EJ
soils, the value of kz is commonly estimated  yi 2  4. yi1  6. yi  4. yi 1  yi 2   b.c. yi  0
 L 
4

based on the undrained shear strength Su of where: E is Young modulus of pile material;
cohesive soils such as kz = 67Su (Davisson, J is inertia moment of pile cross section; i is
1970). For cohesionless soils, the value of nh is index from 1 to n-1; L is length of section
commonly estimated based on the friction angle
pile; b is width of pile; c is horizontal reaction
of cohesionless soils as proposed by ATC-32
stiffness of soil.
(1996). The Winkler model was indicated in the
with
foundation design standards (TCXD 205-1998,
c.L4 .b
TCVN 10304-2014) issued by Vietnam a 6
E.J .n 4
Ministry of Construction. These methods which
are based on the theory of elasticity are not therefore:
yi  2  4. yi 1  a. yi  4. yi 1  yi  2  0 (4)
generally applicable for design due to assigning
a constant value to represent the interaction Apply for sections from i = 1 to i = n-1 and
between pile and soil. achieve (n-1) equations contain (n+1) unknown
In fact, the relation deformation-load changes lateral displacement. Two equations are
non-linearly and reaches the limit value, pgh, supplemented from balancing statics conditions.
when the load increases. The nonlinear System of (n+1) equations contains
relationship of p-y can be performed by chart. equation(4) and 2 balancing statics equations
The numerical method is normally required to allow to define unique lateral displacement
analyse the pile soil interaction when using non- values of pile and correlative internal force
linear p-y curves. Although the model of p-y value of pile. Finite difference method can be
curves was introduced in the Bridge design applied for every displacement-reaction
standard issued by Ministry of Transportation in model. However, it becomes easier when that
22TCN 272-2005 and TCVN 11823-2017, only relation is linear because value a in that
some general guidelines were given. Hence, it is equations is unchanged. For non-linear
difficult to construct p-y curves and to control behavior, it is more complicated and will be
the results of lateral bearing capacity of pile for solved by iterative method by technique
engineering practices. In this paper the p-y mentioned in Section 2.2

99
 ground

Figure 2. Soil's relation p-y for clay

Tail of pile The ultimate soil resistance can be obtained

by using the equation of Matlock (Figure 7) for
clay
y (5)
p = 0,5. p u 3
Figure 1. Finite difference model for lateral load y 50
bearing pile The ultimate soil resistance can be obtained
by using the following equation:
2.2. Procedure to determine lateral bearing
capacity pu = Su .B.N p (6)
Np normalized ultimate soil resistance, the
The following steps will be used to analyse pile
general function, describing the variation of Np
horizontal bearing capacity by differential
at the z depth is given by
method.
Step 1: Calculate and build p-y curves  vo' z
Np = 3 J 9 (7)
Step 2: Establish system of equations Su B
Divide the pile into n section in which each Su is undrained shear strength at the depth z,
section has length of L. Then, set up system B is pile width
included n equations with n unknowns of
σ’vo is effective overburden stress,
horizontal displacements.
J is empirical coefficient. Matlock
Step 3: Solve the system of equations to get
recommended the use of 0.5 in design.
horizontal displacements corresponding to each
point. y50 = 2,5. 50 .B (16)
For the non-linear p-y curves, this step is where ε50 is strain at 50% of the maximum
solved by iterative method because the stress difference determined from triaxial
component of stiffness matrix in the step 2 is compression test. Table 1 is the recommendation
dependent on the horizontal displacement at of ε50 in case lack of triaxial compression test
the step 3.
Step 4: Find shear force and moment value Table 1. Recommendation of 50 parameter
along of the pile.
Soil Liquid High Low Light Very
plasticity plastic plastic Stiff Stiff
2.3. p-y curves to determine the lateral
bearing capacity of pile Su (kPa) 12 ÷ 24 24÷ 48 48÷ 95 95÷ 190 >190
ε50 0,02 0,01 0,007 0,005 0,004
Figure 2 describes some types of p-y relation
charts of common soil. These charts are built Reese et a1(1974) proposed criteria for
based on practical experiences or based on cohesion1ess soils for analyzing the behavior
results of field test such as pressuremeter test of piles under static and cyclic loading.
in borehole. Experimental p-y curves were obtained from

100
the results of tests on the fully instrumented
piles. Both theory and empiricism were then
employed to obtain mathematical expressions
that fit the experimentally derived p-y curves.
Figure 3 is the sketch of p-y curves for
cohesionless soil proposed by Reese et a1,
(1974). These curves are detailed in Reese et
a1, (1974); Matlock (1970); Pradhan (2012)
Figure 5. section of concrete pile

Figure 3. Soil's relation p-y for sand

3. CASE STUDY
A 3.0 m clay layer with undrained strength (Su)
of 25kPa, density of 16kN/m3 and deformation Figure 6. p-y curves of clay layer (from 0m to 3m
coefficient (ε50) of 0.003 underlain by a 16.0 m depth)
sand layer with friction angle of 30o, density of
16kN/m3 and coefficient (k) of 27155 kN/m3
(Figure 4). A drilled pile with the diameter (D) of
1 m using 20 steel bars with 20 mm in diameter
was installed into this soil profile (Figure 5).
The pile - soil interaction, p-y curves, is
modelled as Matlock (1970) in Figure 6 for
weak clay below water lever (the top layer of
Figure 4)and modelled as Reese (1974) in
Figure 7 for sand layer. The p-y curves show
that the horizontal reaction stiffness around soil
increases with the increasing of z depth.

Figure 7. p-y curves of sand layer (from 3m to

Figure 4. geological section 19m depth)

101
 Figure 8 shows the displacements; shear force Note: The commercial FB-Pier software was
and moments along pile corresponding the lateral built by group authors Marc Hoit, Cliff Hays,
load of 150 kN applied to the top of the drilled Mike McVay and Mark Williams, Bridge
pile. The maximum lateral displacement of this software institution (BSI) of Construction
pile is 5.05 mm from the finite difference method Faculty, Florida University, USA.
while its value is 5.68 mm from the commercial
FB-Pier software (Figure 8). Hence, the lateral Table 2. Comparison of results taken from
displacement is matched quite well between finite difference method and FB-Pier
finite difference method and the FB-Pier (only Finite
11.1 % in the maximum error). Although there is Item difference FB-Pier Difference
a good match for the displacements of two method
results, a large difference in moment and shear
Maximum
values were obtained. The minimum shear value lateral 5.05 mm 5.68 mm 11 %
is -67.02 kN at the depth of 7.0 m from the displacement
difference method while it is -100.00 kN at the
Minimum -67.02 -100.00
depth of 5.90 m (Figure 9) from FB-Pier. The 33 %
shear value kN kN
maximum moments are 318.32 kN.m from the
difference method and 416.00 kN.m from FB- Maximum
318.32 416.00
moment 23 %
Pier at the same position (Figure 10). All kN.m kN.m
value
comparisons are shown in the Table 2. All curves
from these two calculation method have a similar
Shear force (kN)
trend as the same soil - pile interaction model
was used for both cases. However, the extremal -300 -200 -100 0 100 200 300 400
values between two cases are slightly different 0
because the finite difference method was
implemented manually while finite element 2
method was done by the FB-Pier software.
Displacement (mm)
4
-10 0 10 20 30 40
0

6
2 FB-Pier software

4 8
Depth (m)

"Finite difference
method"
6
10
FB-Pier software
8
Depth (m)

12
10 "Finite difference
method"

12 14

14
16
16

18
18

Figure 8. Displacements Figure 9. Shear force

102
 Moment (kN.m) Mechanics and Foundation Division,
-100 100 300 500 700 American Society of Civil Engineers, Vol 89,
0 No. SM3 pp 123-157.
Broms B. B. 1964b. Lateral Resistance of Piles in
Cohesive Soils, Journal of the Soil Mechanics
2 and Foundation Division, American Society
of Civil Engineers, Vol 89, No. SM2, pp 27-
63.
4 Davisson M. T. 1970. Lateral load capacity of
piles, Highway Research Record, NO.333
6 (1970) pp. 104-112.
Heyer B. J. and Reese L. C. 1979. Analysis of
single piles under lateral loading, Research
8 Report 244-1: Development of Procedures for
Depth (m )

the Design of Drilled Foundations For Support

of Overhead Signs.
10
Matlock H. 1970. Correlations for Design of
FB-Pier software Laterally Loaded Piles in Soft Clay, Presented
12 at the Second Annual Offshore Technology
Conference, Houston, Texas, Vol 1, pp 577-
"Finite difference 588.
method"
14 Matlock H. and Reese L.C. 1960. Generalized
Solutions for Laterally Loaded Piles, JSMFD,
ASCE, Vol. 86, N. SMS, Part 1.
16
Poulos H. G. 1971. Behavior of Laterally Loaded
Piles: I-Single Piles, Journal of Soil
18 Mechanics and Foundation Division,
American Society of Civil Engineers, Vol. 97,
No. SM5, Proc. Paper 8092, pp 711-731.
Figure 10. Moments Pradhan D.L. 2012: Development of P-y Curves
for Monopiles in Clay using Finite Element
4. CONCLUSIONS Model Plaxis 3D Foundation, Master Thesis,
Norwegian University of Science and
This article presents a study of lateral bearing Technology, Department of Civil and
capacity of concrete rectangular pile by finite Transport Engineering .
difference method based on the theory of soil Reese L. C., Cox W. R. and Koop W. R. 1974.
pile interaction, p-y curves. The results were Analysis of Laterally Loaded Piles in Sand.
Proceeding of Fifth Annual Offshore
then compared with those from commercial
Technical Conference, Paper No. OTC 2080.
software FB-Pier. This article can help Houston Texas.
engineers to have a better understand the Tan W. L., Song S. T. and Hung W. S. 2012.
theory as well as practice of p-y theory to Effective Subgrade Coefficients for Seismic
determine lateral bearing capacity. Performance Assessment of Pile Foundations,
the 15th World conference on earthquake
5. REFERENCES engineering, Lisbon, Portugal.
TCVN 10304-2014 Pile Foundation – Design
22TCN 272-05. 2005. Bridge design standard. Standard. National University of Civil
Ministry of Transportation, Vietnam. Engineering, Vietnam.
ATC-32. 1996. Applied Technology Council ATC- TCVN 11823. 2017. Highway Bridge Design
32: Improved Seismic Design Criteria for Specification. Ministry of Transport, Vietnam.
California Bridges: Provisional TCXD 205. 1998. Foundation design standard,
Recommendations, Redwood City, California. Ministry of Construction, Vietnam.
Broms B. B. 1964a. Lateral Resistance of Piles in
Cohesionless Soils, Journal of the Soil

103
The 3rd Int. Conf. on Transport Infrastructure & Sustainable Development (TISDIC 2019). Construction Publishing House. ISBN: 978-604-82-2893-4.

Geocell wall - new solution for rock/ soil slope stabilization.

application at a project in the northern mountainous area
in Vietnam

Le Thiet Trung1,, Pham Viet Anh1,

1
Soil Mechanics Division, National University of Civil Engineering, Hanoi, Vietnam

Corresponding author, E-mail: [email protected]

ABSTRACT: The northern mountainous region is an important socio-economic development area of

Vietnam. With the natural conditions of a humid tropical region, monsoon … geological disasters have
been a threat to the life of the community, destroying the mountainous landscape.
Although the risk of geological hazards in mountainous areas is important, researches and applications of
slope stabilization solutions in Vietnam are very limited. The current solutions can be classified into
following types of solutions: i) slope surface repair; ii) construction of surface and underground water
drainage systems; iii) covering the open stone surface (spraying concrete covering layer, planting grass
...) to limit the rock weathering process; iv) installation of warning systems for landslide and erosion; v)
protection and development of vegetation systems (planting protection forests, grass planting ...).
In recent years, a new solution for protecting the slope is to build retaining wall in combination with
Geocell material that has been designed and successfully tested in some construction sites in the Northern
mountainous areas. With different advantages such as high safety factor, resistant material (up to 50 years
of durability), reduced construction time, low construction cost ... this solution can be widely applied in
the economic and social conditions of Vietnam.
This paper presents the design, construction process, layout of displacement monitoring system and results
of measurement of Geocell wall of a project in the Northern mountainous region. The monitoring results
show that the calculated results and displacement predictions are quite close to the measured reality.
Keywords: Slope stabilization; rock; soil; geocell material; retaining wall.

The center is invested in the form of socialization

1. PROJECT INTRODUCTION with total amount of over 1,000 billion dong.
The Obstetrics and Pediatrics Center, Phu Tho Phu Tho Provincial General Hospital acts as the
Provincial General Hospital, with a scale of 500 Project Owner, Hanoi Construction and
beds, started construction in October 2016 on an Architecture Design Joint Stock Company as the
area of 7 hectares, located on Nguyen Tat Thanh design, project management and construction
street, Nong Trang Ward, Viet Tri City, Phu Tho supervision consultant.
Province, adjacent to Hung Vuong University The Obstetrics and Pediatrics Center is
with the standard design of 4-star hotel hospital, designed with two (2) main blocks: Outpatient
ensuring Green - Clean - Beautiful elements, Clinic & Treatment and Administrative Block
harmonious architecture, friendly environment. with a construction area of 9,486m2 and a

104
height of 7 floors; Service Examination,
Inpatient Clinic Block with a construction area
of 1,540.6m2 and a height of 5 floors.
The area using retaining walls is a relatively
flat terrain, bounded by the lake to the
Northeast and residential houses from the East
to the Southeast (Figure 1).

Figure 2. Displacement of concrete retaining wall

In general, reinforced concrete retaining

wall is inclined with structural cracks, signs of
destroying current status, many unusual
locations, without keeping the working status
as original design. At the wall section of 7.2m
Figure 1. Project satellite image high near the lake, the largest horizontal
displacement (at the wall top) is 0.51m, the
The finishing ground elevation of the
smallest is 0.4m. At other wall panels, these
retaining wall changes from +19.28m to
values ranges from 0.06m to 0.294m.
+21.00m. The slope of each land lot is
determined according to the control limit to
ensure the most optimal drainage slope, the 2. EVALUATING THE CAUSES OF
natural ground elevation ranges from +12.00m to INSTABILITY OF REINFORCED
+18.00m. Elevation difference between inside CONCRETE RETAINING WALL
and outside project ranges from 2.5m to 7.5m. Overturning instability: In the original design,
The design consultant selects the solution of reinforced concrete retaining wall is semi-
reinforced concrete semi-gravity wall to retain gravity one, the wall stability is based on the
the finishing embankment. The retaining wall weight of wall and the embankment on the
surface is equal to the finishing ground bottom slab of the wall. However, with the size
elevation, the wall height varies from 2.5m to of the wall and the width of the bottom slab
7.2m depending on the elevation difference varying from 1.2m (wall section of 2.5m high)
between inside and outside project. The wall to 2.4m (wall section of 5m high) and 3.6m
section of 2.5m high uses strip footing on (wall section of 7.2m high), this weight is not
natural ground, with dimension of 0.3m thick enough to keep stable under horizontal
and 1.2m width. The wall section of 5m high pressure of the embankment.
uses strip footing on natural ground, with Sliding instability on the ground: As the
dimension of 0.3m thick and 2.4m width. The weight of wall and soil is small, thus friction of
wall section of 7.2m high uses pile foundation the foundation base - soil and pile resistance is
with bottom slab of 3.6m width. Every 5m not enough against the horizontal loads. The
length of wall, one arranges 4 piles with constitutive connection between piles and
30x30cm cross-section and 9m length. The walls is not guaranteed.
designed pile bearing capacity is Ptt = 40 ton. The drainage system behind the wall is
After the construction of wall and finishing ineffective, broken, clogged, increasing the
of soil embankment behind the wall, a number water pressure on the wall, reducing the shear
of wall sections inclines, deviates from the resistance of the soil and increasing the soil
original design position. The inclination pressure behind the wall.
difference at the subsidence position between Load bearing capacity of the foundation at
retaining walls is quite large (Figure 2). the wall base or of the pile is not sufficient.

105
3. GEOCELL WALL DESIGN 3.2 Geocell Wall Principles
The repair design proposals must ensure the For the embankment section where phenomenon
technical requirements, making use of the of wall instability occurs, proposed treatment
designed reinforced concrete retaining wall and solution includes:
saving costs. After many proposed options Use Geocell retaining wall behind the
such as expanding bottom slab of reinforced concrete wall. Geocell retaining wall is arranged
concrete wall, using soil anchor, using steel close to the old wall. Between the new wall and
sheet wall to form double wall, ... the Geocell the old wall, geotextile is used to reduce the soil
retaining wall has been selected because of lateral pressure on the old retaining wall. Back
many advantages in this project. fill soil is compacted to K=0.9 (Figure 3)
Arrange drainage D110 pipes along and
3.1 Geology of retaining wall area behind the geocell retaining wall and every 4m,
one perpendicular drainage pipe running out of
The stratum of the area to be treated includes the old retaining wall with drainage slopes of
the following layers: 1% approximately.
Layer 1: Back fill soil: less plastic, soft to
hard plastic clay. This layer is distributed across
the entire survey area. The thickness of layer is
from 3.70m to 7.20m. Standard penetration test
number (N30) varies from 3 to 7.
Layer 2: Mud soil - soft organic soil,
plastic state. This layer distributes areas near
the lake in the survey area, thickness of
2.20m. N30 = 4.
Layer 3: Low plastic clay, hard plastic state.
This layer distributes areas near lake in the
survey area, thickness varies from 1.50m to
3.70m. N30 = 10.
Layer 4a: Mud soil - organic, very plastic,
hard plastic state, thickness of layer is 3.10m. Figure 3. Typical design of Geocell retaining wall
N30 from 16 to 18.
Layer 4b: Mud soil - organic, very plastic, In case soft ground is found during
soft plastic state, thickness of layer is from construction, use such ground reinforcement
3.5m to 7.3m. N30 = 8. methods as partial sand cushion at the base of the
Layer 5: Very plastic, soft to hard plastic clay, wall or combined with bamboo poles (Figure 4)
thickness of layer is 10.30m. N30 from 6 to 12.
Layer 6a: Less plastic, soft to hard plastic
clay, thickness of layer ranges from 4.00m to
6.20m, N30 from 12 to 19.
Layer 6b: Less plastic, semi-hard to hard
clay. N30 from 18 to 46.
Layer 7: Sandstone, white gray, blue gray,
strong weathered. Top elevation of the layer
ranges from +7.90m to -5.20m. N30  100
hammers.
The soil investigation results show that the
geological condition is relatively complicated,
especially in the areas near the lake, the Figure 4. Reinforcement of soft soil with bamboo
geology is weak. poles.

106
3.3 Design of Geocell Wall In which:
Mg: Moments for protection against
Calculation theory of Geocell retaining wall is
overturning around the front edge.
relatively explicit, has been included in
Tt: Moments causing overturning around the
Vietnam Standard TCVN 10544: 2014 [1]. The
front edge.
design of Geocell retaining wall shall calculate
[K]: allowable safety factor of overturning
the following issues [1], [2]:
around the front edge, specified in TCVN
The overall calculation of soil slope with
9152:2012 [4].
retaining wall, in which the structure of
Geocell retaining wall is modeled by a soil
mass with improved physical and mechanical
properties [3], [2].
Sliding stability calculation at the bottom
surface of wall foundation (Figure 5), with
safety factor of flat sliding K is determined by
the formula [1]:
Rg
K  K 
Tt

In which: Figure 6. Overturning around the front edge

Rg : total force for protection against flat
sliding, Stability calculation of the load bearing
Tt : total force causing flat sliding at the capacity of the wall foundation (Figure 7)
bottom surface of wall foundation. with safety factor for the load bearing capacity
[K]: allowable safety factor of flat sliding at of the wall foundation determined by the
the bottom surface of wall foundation, formula [1]:
specified in TCVN 9152:2012 [4]. qu
K  K 
 max
In which:
qu: load bearing capacity of the wall
foundation.
σmax: maximum stress on the wall foundation.
[K]: allowable safety factor for the load
bearing capacity, specified in TCVN
9152:2012 [4].

Figure 5. Flat sliding to the bottom surface of wall

foundation

Stability calculation of overturning around

the front edge (Figure 6) safety factor of
overturning around the front edge (K) is
determined by the formula [1]:
M g
K  K  Figure 7. Bearing capacity instability of wall
M t foundation

107
 Stability calculation of sliding to the
bottom surface of the ith Geocell layer (Figure
8) with safety factor of sliding to the bottom
surface of the ith Geocell layer (Ki) determined
by the formula [1]:
Rgi
Ki   K 
Tti
In which:
Rgi : total forces for protection against flat
sliding to the bottom surface of the ith Geocell Figure 9: Overturning stability around the front
layer. edge in the ith Geocell layer
Tti : total forces causing flat sliding to the
bottom surface of the ith Geocell layer. The calculations are done for all sections.
[K]: allowable safety factor of flat sliding to Calculation results for the most dangerous
the bottom surface of wall foundation, section, in condition of water saturation, are
specified in TCVN 9152:2012 [4]. shown in the following table:

Table 1. Calculation results

Calculation cases Kcalcul [K]
Stability regarding flat sliding
to the bottom surface of 3,052 1,2
foundation
Overturning stability around
1,206 1,2
the front edge
Stability regarding load
bearing capacity of wall 1,5 1,2
Figure 8. Bottom flat sliding instability according
foundation
to the bottom of the ith Geocell layer
Stability regarding flat sliding
Calculation of overturning stability around to the bottom of the ith 1,28 1,2
the front edge in the ith Geocell layer (Figure 9) Geocell layer Fsmin
overturning safety factor around the front edge Overturning stability around
in the ith Geocell layer, Ki is determined by the the front edge in the ith 2,81 1,2
formula [1]: Geocell layer
M gi
Ki   K  Wall subsidence 5cm 10cm
M ti

In which: 4. CONSTRUCTION OF GEOCELL WALL

Mgi: total number of moments for protection Step 1: Excavation of soil to the base of
agaist overturning around the front edge in the existing reinforced concrete wall (Figure 10).
ith Geocell layer. Based on the approved design, Geocell flexible
Tti: total number of moments causing wall has a special structure, one part placed
overturning around the front edge in the ith directly on the foundation of reinforced
Geocell layer. concrete wall to keep this structure stable.
[K]: allowable safety factor of overturning Therefore, during construction, the contractor
around the front edge, specified in TCVN has excavated until the bottom level of the
9152:2012 [4]. foundation slab of reinforced concrete wall.

108
 Step 2: Reinforcement of soft ground in some Step 4: Construction of Geocell layers combined
areas (Figure 11). At the wall sections near the with sand filling and compacting to K90
lake, as the geology is complicated with soft soil (Figure 12). Geocell layers aim to increase
layers, before the foundation of Geocell wall is shear resistance for the largest shear force
constructed, reinforcement with bamboo poles is bearing area of Geocell wall [6].
carried out to increase the bearing capacity of the
foundation and reduce the deformation of the
foundation. According to the design, density of
bamboo poles is 30 poles/m2. During
construction, all bamboo poles have been pressed
to the designed density and are completely under
the water table, ensuring the longevity of
bamboo poles according to regulations.
Step 3: Construction of Geocell wall foundation.
Geocell wall foundation is constructed by
pouring concrete directly onto the 03 first
Geocell layers. Geocell layers are spread and Figure 12. Geocell combined with sand layer
linked in accordance with the guidelines in
TCVN 10544:2014 [1], [5]. According to the Step 5: Construction of Geocell layers of
design, the bottom of geocell wall foundation is flexible walls, combining with K90 soil cover.
inclined toward the embankment soil to increase The soil covering and compaction is carried
the slip resistance and overturning resistance of out for layers of 20cm thickness, exactly the
the wall around center of the turning point, i.e. same thickness as a Geocell layer. 16-ton roller
the outer edge of flexible wall. Construction is used but does not use vibration mode to
process is done in accordance with this design. ensure the compaction of soil but does not
cause any impact on the existing reinforced
concrete wall. The area between the inner edge
of the reinforced concrete wall and the outer
edge of the Geocell wall, with the top width of
approximately 1.5m, will be filled with soil
and compacted later (step 7). During
construction, testing to determine the density is
done for each soil layer (Figure 13) [5], [6].

Figure 10. Excavation until the base of existing

concrete wall

Figure 11. Construction of Geocell wall foundation Figure 13. Determination of the density of soil

109
 Step 6: Construction of drainage system. It is impossible to conduct monitoring of
The drainage system is arranged into 3 layers, geocell wall deformation and displacement,
including water collecting pipes and drainage because the wall structure is completely
pipes. Structure of typical cross-section of a located between the landfill and reinforced
drainage pipe consists of reverse filter layer, concrete wall.
made from stone 1-2, pipe coated with The purpose of building Geocell wall is to
geotextile + PVC pipe with diameter of D110; reduce or eliminate the pressure of the landfill
punch a hole of d30, density [5]. on the reinforced concrete wall. Therefore, we
During the construction process, from June evaluate the effectiveness of Geocell wall
to August 2018, there were many heavy rains. indirectly by measuring the displacement of
Drainage system has shown that drainage reinforced concrete wall.
efficiency is quite high (Figure 14). The principle for measuring displacement of
reinforced concrete wall is considering the wall
has no translational movement but rotary
displacement only.
Measurement is carried out from the
beginning of the earthwork process (step 1)
and the interval is 07 day
Measurement results show that reinforced
concrete walls have no displacement after the
construction of Geocell walls.

5. CONCLUSIONS
Through the design and construction of
Geocell flexible retaining wall at the Obstetrics
Figure 14. Drainage system works effectively in and Pediatrics Center - Phu Tho Provincial
rainy days (flowing out of reinforced General Hospital, we can see that there are
concrete wall) different effective solutions to keep the slope
stable, both excavation and embankment one.
Step 7: Fill the land between reinforced In case of embankment slope, Geocell
concrete wall and Geocell wall. In this area, the material-used retaining wall is a solution with
contractor does not use 16-ton roller to compact many advantages, including high stability,
the land but use 600kg roller (hand roller) to reasonable construction cost, fast construction
ensure safety for reinforced concrete wall. progress, simple construction equipment. The
Step 8: Surface finish and grass planting construction of Geocell wall allows the use of
(Figure 15). materials in place and local labors.
Besides, the Vietnamese Standard “TCVN
10544:2014 Neoweb cellular confinement in
infrastructure constructions - Requirement of
design, construction and acceptance” that has
been issued is also an advantage for the design
and construction of structure of this type.

6. REFERENCES
Standards TCVN10544:2014, Neoweb cellular
confinement in infrastructure constructions -
Requirement of design, construction and
Figure 15. Surface finish and grass planting acceptance, 2014.
A. J. Gilchrist, "Design and construction of Geocell
Retaining wall displacement monitoring Mattress as embankment foundation," 1988.

110
G. M. Latha, K. Rajagopal and a. N. R.
Krishnaswamy, "Design of geocell supported
embankments," in Proceedings of the 2nd
Asian Geosynthetics Conference., 2000.
Standards TCVN9152:2012, Hydraulic structures -
Designing Process for Retaining Walls, 2012.
BS8006 Section 7: Reinforced slopes, Standard
reinforced earth design techniques.
CIRIA C516 “Modular gravity retaining walls –
design guidance”, Standard gravity wall
design methods to check for overturning,
bearing, ect.

111
The 3rd Int. Conf. on Transport Infrastructure & Sustainable Development (TISDIC 2019). Construction Publishing House. ISBN: 978-604-82-2893-4.

Determination of unloading - Reloading modulus and

exponent parameters (m) for hardening soil model from
drained triaxial test of soft soil in Ho Chi Minh City

Ngo Duc Trung1*, Vo Phan2, Tran Thi Thanh3

1
Department of Civil Engineering, Van Hien University, Ho Chi Minh City, Vietnam
2
Department of Civil Engineering, Ho Chi Minh City University of Technology, Ho Chi Minh City, Vietnam
3
Department of Civil Engineering, Ho Chi Minh City University of Industry, Ho Chi Minh City, Vietnam

Corresponding author, E-mail: [email protected]

ABSTRACT: This study analyzes and determines the modulus of deformation and the dependence of the
modulus of deformation on the stress state of Ho Chi Minh City’s soft soil. Parameters were determined
as stiffness parameters: E50 and Eur in Hardening Soil model. Tri-axial compression testing was performed
for very soft clays at depths of 4-6 m and 12-14 m, soft clays in the range of 18-20 m and 24-26 m
according to drainage conditions with unloading and reloading. From test results, identifies the exponent
parameter (m) and the correlation coefficient Eur / E50 for soft soil in Ho Chi Minh City for deep
excavation calculation.
Keywords: stress paths, unloading and reloading, deep excavation, soil model, hardening soil.

1. INTRODUCTION soils, the chosen exponent is m = 1. According

to Janbu (1963), the value of m is about 0.5 for
In Ho Chi Minh City (HCMC) in recent years, sand and clay in Norway. Whereas von Soos
the calculation of geotechnical works often (1980) has a m value of 0.5 <m <1 [8]. Usmani
uses finite element method with constitutive (2007) [7] proposed that m = 0.67 in stress-
models. The biggest problem for design strain analysis of Delhi clay sand.
engineers is to properly analyze the behavior of Thus, the choice of the m-parameter makes it
the soil by selecting the right constitutive difficult for engineers to correlate the empirical
model and input parameters. expressions, since the amplitude is still relatively
The Hardening Soil (HS) model is based on wide and results in large discrepancies.
the Dun-can Chang model showing more This paper will identify the exponent
advances than the Mohr-Coulomb (MC) parameter (m) and the correlation coefficient
model. Similar to the MC model, stress states Eur / E50 for soft soil in Ho Chi Minh City on
of stress are expressed by the friction angle φ, the basis of drained triaxial compression test as
the cohesion force c, the dilatation angle ψ, but defined in the HS model.
the stiffness of the soil is expressed with
greater precision by using 3 different input 2. OVERVIEW OF SOFT SOIL IN HCMC
modulus variables: secant modulus E50ref ;
unloading-reloading modulus Eurref and tangent HCMC belongs to the Saigon River delta
oedometric modulus Eoedref
. The HS model also (Figure 1), the stratigraphic structure of this
explains the dependence of the stiffness on area belongs to the Fourth Century Tan Sinh
stress. The level of dependence of stress is period and the Tan Can Dai period accumulates
given by the exponent m. In order to simulate which form a total of 6 layers of natural soil
stress dependence according to the logarithmic (Figure 3). Layer 1 and layer 2 consist of slurry
law, Schanz et al. (1999) [5] investigated soft and thick soils with a depth of 20 ÷ 30m, high

112
organic content, high water content of 85-
104%, void ratio e = [1.5 ÷ 2.5] soft soil is very
compact, high liquid IL index, reaching 1.85.

Figure 3. Soil profile along the right bank of

Saigon Rive [3]

This study was conducted on two soft clay

of HCMC: very soft clays and soft clays with a
depth of 4 to 30 m below groundwater, which
often affects the stability and deformation of
Figure 1. Distribution of soft clay in SE region [3]
underground structures.

Soft soil of HCMC is located in: Binh Thanh 3. OVERVIEW OF THE HS MODEL
District, Can Gio District, District 6, District 7,
The HS model developed by Schanz et al.
District 4 and Binh Chanh District (Figure 2).
(1999) [1] is based on the classical elastic-
Soft soils are highly compressive, with very low
plastic theory to simulate the resilient and
load capacity. One or more of the physical
flexible behavior of the soil. Its elasticity uses
properties, durability and deformation of the
two stiffness modules, ie the secant modulus
soil are within the following range:
E50 and unloading-reloading modulus Eur.
- Void ratio e = [1.5÷2.5]
Plasticizers follow the nonlinear flow rule and
- Water content W  65 the directional re-orientation standard, to
- Water unit weight w = [14÷16] kN/m3 describe the relationship between stress and
- Undrained shear strength Su < 50 kPa strain of the soil in a hyperbolic curve.
- standard penetration test N30 < 4 Yield surface:
- cohesion intercept c < 10 kPa
qa (1   2 ) 2(1   2 )
- Settlement ratio a1-2 > 5 m2/kN f1    p (1)
- Deformation modulus E < 5000 kPa E50 qa  (1   2 ) Eur
qa (1   3 ) 2(1   3 )
f2    p (2)
E50 qa  (1   3 ) Eur
qa ( 2   3 ) 2( 2   3 )
f3    p (3)
E50 qa  ( 2   3 ) Eur
With qa, E50 and Eur are defined by formulas
(4), (5), (6) and the notation p for plastic stress.
m
 c cot g   '3 
E50  E50ref  ref 
(4)
 c cot g  p 
m
 c cot g   '3 
Eur  Eurref  ref 
(5)
 c cot g  p 
2sin  qf
q f  (c cot    3 ) , qa  (6)
1  sin  Rf
Figure 2. Distribution of soft ground in HCMC [3]

113
 defined by the smallest stress σ3 = pref as the
default value in Plaxis is pref = 100 (kN/m²).
However, defining the parameters Eurref , Eoed
ref
in
Plaxis generally chooses the default word for all
types of soil as formulas (8) often make
calculations difficult [6]:
ref
Eoed  E50ref ; Eurref  3E50ref (8)

4. DETERMINATION OF M AND EUR / E50

Figure 4. Definition of E50 and Eur for drained
triaxial test
OF THE HS MODEL FOR SOFT SOIL IN
HO CHI MINH CITY
4.1 Drained triaxial compression test
As defined in the HS model, parameters must
be determined from the triaxial compression
test. The triaxial compression test was carried
out in accordance with the drainage diagram
(CD) with cell pressure pref = σ'3 = 100kPa
which was unloaded and reloaded, resulting in
Figure 5. Definition of Eoed
ref
in oedometer test results a axial strain and deviator stress graph. From
this chart it is possible to determine the
In the un-loading and re-loading stress parameters φ ', c', E50ref , Eurref .To determine the
paths, the stress-strain relationships are still in parameter m depicts the dependence of the
the form of hyperbolic, and empirical studies stiffness on the stress for soft soil in HCMC,
[2] show that modulus E50 in the unloading and the author carried out experiments on 12 clay
reloading experiments is larger than in the samples at depths from 4 to 30m below the
conventional triaxial compression tests many groundwater table with drained triaxial tests
times and different from each soil type. In this have unloaded and reloaded at the pressure
study, we focused on the Eur / E50 ratio for the chamber level '3 are 50, 100, 200 and 400
soft clay in HCMC. kPa. The experimental sample area of Binh
Equation (4), (5) defines E50, Eur, and Eoed is Chanh district is relatively thick and
defined by the following equation: homogeneous. Results of the analysis of
m
ref  c cot g   '3  mechanical properties are given in Table 1.
Eoed  Eoed  ref 
(7)
 c cot g  p 
ref Table 1. Physical characteristics of soft soils in
Eoed is tangent oedometric modulus in
HCMC
oedemeter test at the vertical stress -σ1 '= pref.
The advantage of the HS model over the MC Very soft
Parameters Symbol Soft clays
model is not only the use of hyperbolic strain clays
curves instead of linear relations but also the Water content (%) Wn 70 ÷ 100 60 ÷ 70
control of the stiffness dependence on the stress Void ratio e 2.0 ÷ 3.0 1.3 ÷2.
load. When using the MC model, the user must Water unit
n 14.0 ÷ 16.0 16÷ 17
select a fixed Young modulus value while the weight (kN/m3)
real stiffness level depends on the pressure level. Dry unit weight
d 7.5 ÷ 8.5 8.5 ÷ 12
It is then necessary to estimate the pressure level (kN/m3)
in the soil and use that pressure level to obtain Liquid limit (%) WL 70 ÷80 45÷70
the appropriate stiffness value. With the HS Plastic limit (%) WP 30 ÷ 40 20 ÷ 30
model the difficult selection of input parameters Saturation (%) S 95 ÷98 99 ÷100
is no longer necessary. Instead, the modulus is

114
 The results of the experiment for the two
clay layers are shown in Figure 6 to Figure
13 belows:

Figure 10. The relationship (1 – q) of sample

No.7,8,9

Figure 6. The relationship (1 – q) of sample

No.1,2,3

Figure 11. The relationship (1 – v) of sample

No.7,8,9

Figure 7. The relationship (1 – v) of sample

No. 1, 2, 3

Figure 12. The relationship (1 – q) of sample

No.10,11,12

Figure 8. The relationship (1 – q) of sample

No. 4, 5, 6

Figure 13. The relationship (1 – v) of sample

No.10,11,12

Figure 9. The relationship (1 – v) of sample

From the stress-strain diagram (q, 1), we
No. 4, 5, 6 define c', ' and the parameters as Table 2.

115
Table 2. Shear srength parameters
Depth Sampl c’ φ’  1' f [k  3' f [k y
[m] e No. [kG/cm2] (0) G/cm2] G/cm2] p ref
Vert soft clays
4÷6 1 0.12 25.9 1.40 0.36 0.485
2 3.05 1.09 1.074
3 5.29 1.86 1.690
12 ÷ 14 4 0.09 26.3 2.16 0.86 0.885
5 5.01 1.90 1.758
6 9.48 3.72 3.300
Soft clays
18÷20 7 0.10 26.0 2.17 0.88 0.899
8 3.92 1.81 1.674
9 8.94 3.89 3.394
24÷26 10 0.11 26.3 2.37 0.98 0.982
11 4.02 1.91 1.747
12 9.64 3.99 3.442

4.2 To determine the power m from the

drained triaxial compression test Figure 14. The regression equation correlates
between E50 / E50ref and  y / p ref
The parameter m represents the dependence of
the stiffness on the stress state of the ground. In From there, the value of parameter m
this section, the author proceeds to define the determined from the triaxial compression test
exponent m from the modulus of the distortion in
through the secant modulus E50 is as follows:
the HS model according to expressions (4), (5).
Based on the definition of E50 in the HS - Very soft clay: m = [0.72 ÷ 0.93] (10)
0.8686
model, formula (4), we have: E50  y 
 0.9854  ref  , R2 = 0.9865 (11)
 y 
m
 c cot  '  
' m
E50ref p 
E50  E50ref  ref   E50ref  3

p   c cot  ' p 
ref Soft clay: m = [0.72 ÷ 0.84] (12)
(9) E50  y 
0.7585
 E50   1.0287  ref  , R2 = 0.9875 (13)
 m  log     ref  E50ref

y
  E50  p 
 p ref 
 
This value is consistent with the experimental
The power m can be determined as E50 as results of von Soos (1980) [8] that m is between
shown in Table 3.
0.5 m  1.0 with the lower catchment as sand
Table 3. Modulus E50 , E50ref and m from the triaxial and the upper margin is soft clay.
compression test From the stress-strain diagram obtained
from the experient, draw the tangents line Eur
Depth Sampl E50 E50ref E50 y m as defined by the modulus Eur of the HS model
[m] e No. [kG/cm2] [kG/cm2] E50ref p ref [-]
to determine the loading and re-loading Eur,
Very soft clays
4÷6 1 17.29 33.88 0.51 0.485 0.93
resulting in present in Table 4.
2 33.88 1.00 1.074 - Based on the definition of Eur in the HS
3 49.44 1.46 1.690 0.72 model, formula (5), we have:
12 ÷ 14 4 21.15 21.15 1.00 0.885 - m m
 y   c cot  '  3' 
5 33.72 1.59 1.758 0.83 Eur  Eurref  ref   Eurref  ref 
6 60.51 2.86 3.300 0.88 p   c cot  ' p 
Soft clays (14)
 Eur 
18 ÷ 20 7 20.72 20.72 1.00 0.899 -  m  log     ref 
8 29.12 1.41 1.674 0.76 
 p
y
ref


 Eur 
9 52.05 2.51 3.394 0.75
24 ÷ 26 10 22.93 22.93 1.00 0.982 - From equation (14), the parameter m is
11 30.66 1.48 1.747 0.72 determined according to the unloading module
12 64.69 2.81 3.442 0.84 Eur as shown in Table 4.

116
Table 4. Eur , Eurref and m from test results Eur  y 
0.789

 1.0389  ref  , R2 = 0.9956 (18)

Eurref p 
Depth Sampl Eur E ref
ur
Eur y m
[m] e No. [kG/cm2] [kG/cm2] Eurref p ref [-]
4.2 Determination of correlation coefficient
Very soft clay Eur / E50 for soft soil in HCMC
4÷6 1 68.98 127.27 0.54 0.485 0.85
2 127.27 1.00 1.074 - With the default set of parameters of the HS
3 205.90 1.62 1.690 0.92 model in Plaxis, Eurref / E50ref = 3 is often chosen
12 ÷ 14 4 106.09 106.09 1.00 0.885 - [8]. However, the actual ratio is very different
5 177.23 1.67 1.758 0.91 for each soil type. From the results of
6 284.03 2.68 3.300 0.82
experiments on soft soil in HCMC. The authors
Soft clay
propose this ratio as Table 5.
18 ÷ 20 7 99.39 99.39 1.00 0.899 -
8 149.33 1.50 1.674 0.79
Table 5. correlation coefficient Eur / E50 for
9 276.73 2.78 3.394 0.84
24 ÷ 26 10 108.31 108.31 1.00 0.982 -
soft soil in HCMC
11 173.42 1.60 1.747 0.84 Depth c E50 Eur Eur
[m] [kG/cm2]
12 298.81 2.76 3.442 0.82 [kG/cm2] [kG/cm2] E50
Very soft clays
4÷6 0.5 17.29 68.98 3.99
1.0 33.88 127.27 3.76
2.0 46.97 205.90 4.16
12 ÷ 14 1.0 21.15 106.09 5.02
2.0 33.72 177.23 5.26
4.0 60.51 284.03 4.69
Soft clays
18 ÷ 20 1.0 20.72 99.39 4.80
2.0 29.12 149.33 4.87
4.0 52.05 276.73 5.32
1.0 22.93 108.31 4.72
24 ÷ 26
2.0 34.35 173.42 5.05
4.0 64.69 298.81 4.62

Figure 15. The regression equation correlates

ref
between Eur / Eur and  y / p ref
From the relationship between Eur / Eurref and
 y / p ref (formula (14)), the regression line of
TRENDLINE as shown in Figure 15, we have
the following results:
Very soft clay: m = [0.82 ÷ 0.92] (15)
0.8325
Eur  y 
 1.0185  ref  , R2 = 0.9896 (16)
Eurref p  Figure 16. The regression equation correlates
Soft clay: m = [0.79 ÷ 0.84] (17) Eurref  E50ref of soft soil HCMC

117
 From above results, the mean value of the Very soft clays:
correlation coefficient Eurref / E50ref for soft soil is Eurref
 [3.99  5.26]
given HCMC is: E50ref
Very soft clays:
Soft clays:
Eurref
 [3.99  5.26] (19) Eurref
E50ref  [4.62  5.32]
E50ref
Eurref  4.5462E50ref  2.077 ,R2 = 0.9371 (20)
ref
E 6. REFERENCES
Soft clays: ur
ref
 [4.62  5.32] (21)
E
50 Brinkgreve R. B. J. (2005), Selection of Soil
E ref
 4.8383E ref
 2.5745 , R = 0.9766
2 Models and Parameters for Geotechnical
ur 50 (22)
Engineering Application, Journal of
This ratio differs considerably from the Geotechnical and Geoenvironmental
default value in Plaxis according to Vemeer [5] Engineering, ASCE.
for all soil types: Janbu N., (1963) Soil compressibility as
Eurref  3E50ref (23) determined by oedometer and triaxial tests. In:
Proceedings of European Conference on Soil
Mechanics and Foundation Engineering.
5. CONCLUSIONS Wiesbaden; 1963. p. 19e25.
(1) The stiffness of the soil depends on the Long P.V., D.T. Bergado, Nguyen L.V. and
Balasubramaniam A.S. (2013), “Design and
state of stress, the dependence of the hardness
Performance of Soft Ground Improvement
on the stress state of soft soil of HCMC is in Using PVD with and without Vacuum
the range: Consolidation”, Geotechnical Engineering
Determined from the drained triaxial Journal of the SEAGS & AGSSEA Vol. 44
compression test through E50: No.4 December 2013 ISSN 0046-5828.
Very soft clays: R. H. G. Parry, Ph.D., A.M.I.E.Aust., Triaxial
m = [0.72 ÷ 0.93] Compression and Extension.
Schanz T., Vermeer P. A., Bonnier P. G. and
Soft clays:
Brinkgreve R. B. J. (1999), Hardening Soil
m = [0.72 ÷ 0.8] Model: Formulation and Verification, Beyond
Determined from the drained triaxial 2000 in Computational Geotechnics, Balkema,
compression test through Eur: Rotterdam, pp. 281-290.
Very soft clays: K. Terzaghi, R. B. Peck & G. Mesri, Soil
Mechanics in engineering practice 3rd ed.,
m = [0.82 ÷ 0.92] John Willey, 1995.
Soft clays: Usmani A., Characterization of shear strength
m = [0.79 ÷ 0.8] behavior of Delhi silt and application to
boundary value problems, PhD Thesis.
Mean m value for soft soil of HCMC: Delhi: Indian Institute of Technology Delhi,
Very soft clays: 2007;
m ≈ 0.86 Von Soos P., Properties of soil and rock. In:
Soft clays: Grundbautaschenbuch, Part 4. 4th ed. Berlin:
Ernst and Sohn; 1980 (in German).
m ≈ 0.80
(2) Soil has a large modulus and non-linear
in the stress path of loading and unloading, the
actual stiffness of the soil is much higher than
the modulus of deformation obtained from
conventional test. With soft soil in HCMC ratio
Eurref / E50ref as follows:

118
The 3rd Int. Conf. on Transport Infrastructure & Sustainable Development (TISDIC 2019). Construction Publishing House. ISBN: 978-604-82-2893-4.

Using discrete element method to simulate the settlement

of soil saturated in embankments due to static loading

Lương Nguyen Hoang Phuong 1 , Tong Anh Tuan 2

1
Dong A University, 33 Xo viet nghe tinh Road, Danang, Vietnam
2
University of Transport and Communications, 3 Cau giay Road, Hanoi, Vietnam,
[email protected], [email protected]

ABSTRACT: This paper presents the numerical model of hydro–mechanical coupling DEM–PFV using
a combination of the discrete element method (DEM) for the solid phase and a pore–scale finite volume
(PFV) of the flow problem. Numerical results at the microscopic scale on settlement of an embankment
with soil saturated are presented denoting that the capable apply of the model DEM – PFV for predicting
the settlement of granular material used in road embankments subjected to static loading. Comparison
between the numerical result and the Terzaghi’s analytical solution of a soil consolidation problem shows
the validation of the model DEM – PFV.
Keywords: Numerical model DEM–PFV, settlement, soil saturated, consolidation.

1. INTRODUCTION flow modeling based on a pore space

Compared with the Finite Element Methods discretization by FEM method [1] or Lattice
(FEM), modeling the flow of pore fluid by Boltzmann scheme (LB) [8]; (ii) Continuum
Discrete Element Method (DEM) would allow a based models in which flow and solidfluid
better understanding of the complex aspects of interactions are defined with simplified semi
flow in porous media, and also the interaction of empirical models based on Darcy’s law [4][10];
flow in porous materials [1][3] with the and (iii) Porenetwork models developed to
mechanical behavior of granular materials at the predict the permeability of materials and
microscale [6][7]. DEM method has been multiphase flow [1][6].
developing in the last few decades [5] and is In the first section of this study, we give the
most commonly applied in the study of principles of the DEM and the governing equations
mechanical behavior of granular materials of the DEMPFV coupling [3][4][9][10][11], an
[2][4][9][10][11], however, this method does not opensourse software is developed in Lab 3SR1,
get used very often in Vietnam. written by C++ and Python, and based on Linux
Numerical modeling of hydromechanical system. Application of the hydro–mechanical
coupling of DEM method and flow in porous coupling DEM–PFV for simulation of settlement
media has attracted attention of the scientific of soil saturated in embankments due to static
communities in the field of geomaterials, soil loading, and numerical results are then compared
mechanics [2][4], and geotechnical engineering with solution of Terzaghi oedometric consolidation
[9][10][11]. Coupled hydromechanical models in the second section.
developed essentially differ in the modeling
techniques adopted for the fluid phase, and they 1
Laboratory of Soils, Solids, Structures and Risks,
can be sorted in three families: (i) Microscale Grenoble (France)

119
2. COUPLED DEM  PFV MODEL Where: mk mass of the particle k ,  k

n
volume of the particle k ,  s is a pressure
2.1 Solid phase (Discrete Element Model)
effect on the surface particle in the direction of

g
n
The DEM is essentially a free mesh Lagrangian the unit vector ,  s density of particles,

f
technique where each particle of the material is acceleration of gravity; jc ,k interactions force

F
a sphere identified by its own mass, radius and
moment of inertia. For each time step of the between particles j and k , c,k total force
computation, interaction forces between applied to the particle k , and nc the number of
particles and consequently the resulting forces interactions of the particle k .
acting on each of them, are deduced from With 6 xN degrees of freedom of
spheres’positions through the interaction law. translational and rotational motion, position
Newton’s second law is then integrated vector was written by a matrix:

X
J
T J
through an explicit secondorder finite ..
 1
( c
W) (2)

T
difference scheme to calculate the new position c
of spheres. Where moment of inertia, force
vector, W received by adding to mass matrix

M
components of rotational motion, and N
number of particles in the simulation.

X
Particles positions t t are updated at each
time step t by a discretization of acceleration
with a centered second order centered finite
difference scheme:
X t t  2 X t  X t t
 J 1 ( Tc  W ) (3)
t 2
Once the new position for each particle is
computed, the list of interactions is updated
and new interaction forces are computed at
each time step t :
 c . .

 Ft  t  Ft
c
 B ( X t t , X t t / 2 ) t

. X  Xt (4)
Figure 1. Elasticplastic contact model  X t t / 2  t t
 t
Linear elastic law provides the contact force DEM used the computation cycle in two
as a function of this relative displacement steps, computation of particles positions and
between two interacting grains. The normal computation of interaction forces.
f

force n and the shear force t are proportional

2.2 Flow phase (Pore-scale Finite Volume)
to the normal displacement U n , and tangential
. Pore-scale space discretization. Pore bodies
displacement U t between two particles, and their connections are defined locally through
depend on the normal stiffness k n and the a regular Delaunay triangulation. (see Fig. 2).
tangential stiffness k t through the particle
radius and the characteristic elastic modulus E
of the material (see Fig. 1).
X

x
ω

.. .. ..
Acceleration of movement k   k , k  of
a particle k obtained by applying the Newton’s
second law of motion:
Figure 2. Adjacent tetrahedra in the regular
n

g
F

.. nc
mk x k   s dS   s d   c ,k
j  mk  c ,k
 mk (1) triangulation and dual Voronoi
j 0
k k
network, in two dimensions [10]

120
 Continuity equation. The rate of volume {F f }  [S]{P} (11)
change of the pore equals to the sum of fluxes Finally, an explicit ordinary differential
q equation of the second order scheme is solved
exchanged ij between adjacents tetrahedral
through the facet S ij : by eq. 9.
. j4 j4
V if    (u  v )dS   qij
f c
(5) 3. SIMULATION OF SETTLEMENT OF
j  j1 S f j  j1 SOIL SATURATED IN EMBANKMENTS

u
v
ij

c
Where: (  ) is the velocity of the fluid DUE TO STATIC LOADING BY DEM –
PFV MODEL
relative to that of the solid.
From Stokes equation, Darcy’s law and Numerical setup: (i) By generating a dense
HagenPoiseuille relation, the relation between sphere packing in a cube; (ii) imposing
fluxes qij with the rate of volume change of a pressure at the top and bottom boundaries;
tetrahedral and the pressure field { pk } can be (iii) imposing hydraulic gradient and static
loading at the top and bottom boundaries;
written as [11]:
(iv) no-flux and non-slip conditions are
. j4
p  pj
V if   qij  kij i  Kij ( pi  p j ) (6) imposed on the lateral boundaries; (v)
j  j1 lij measure the settlement deformation. Fig. 3
Forces on solid particles. The total force introduces boundary conditions of pressure;
exerted by the fluid on a particle results from the walls are rigid and impermeable;
the pressure field and can be evaluated by Voronoi diagram; and display the pressure
summing three terms: (i) the buoyancy force field with N  5000 .
F

b,k
, (ii) the piezometric pressure p,k , and (iii)
F

the viscous shear stress  ,k :

(7)
Fk    gz ndS   p * ndS    ndS  F  F p , k  F , k
b,k

 k  k  k
n

With: the unit normal of the contour.

At each time step, the rate of volume change
of the pores is computed, the pressure field
obtained by eq. 6, and forces are computed for
the next step from eq. 7.
2.3 Coupled DEM-PFV model
Solution of the coupled DEM-PFV model is
based on two matrix relations linking the
generalized displacements to the pressure
field:
. Figure 3. Voronoi diagram and boundary
[G]{P}  [E]{X}  {Q q }  {Q p } (8) conditions of force and pressure [11]
..
[ M ]{X}  {F c }  {W}  {F f } (9) The main parameters of the numerical test:
Where: eq. 8 corresponds to the mass N  5000 , static loading   5000Pa and
conservation equation and eq. 9 describes the hydraulic conductivity k  3 107 m / s . The
movement of the solid phase. evolution of settlement in space and time
Pressure field will be obtained by a first obtained from the simulation is shown on
explicit order scheme from eq. 10: Fig 4, where: S settlement in time, S fin the
X

X
G
P

 (t  t )  (t ) 
[ ]{ (t )}  [ ]  { q (t )}  { p (t )} (10) final settlement, and Tv a non-dimensional
 t 
The solid  fluid interactions are computed time parameter. The numerical result was in
in equation (11): good agreement with Terzaghi’s analytical

121
solution, in terms of evolution of the 5. REFERENCES
settlement in time.
Bonilla R.R.O.: Numerical simulation of undrained
granular media. PhD Thesis. University of
Waterloo (2004).
Bryant S., Blunt M.: Prediction of relative
permeability in simple porous media. Physical
Review A 46(4), 20042011 (1992).
Catalano E., Chareyre B., Cortis A., Barthélemy
E.: A Pore-Scale Hydro-Mechanical Coupled
Model for Geomaterials Investigation of
internal erosion processes using a coupled
DEM-Fluid method. In: Particles 2011, II
International Conference on Particle-Based
Methods, Fundamentals and Applications.
Figure 4. Comparisons between the numerical
results of settlement and Terzaghi’s Barcelona, Spain, 2628 October (2011).
analytical solution. Chareyre B., Cortis A., Catalano E., Barthélemy
E.: Pore-Scale Modeling of Viscous Flow and
The evolution of pore pressure in space, Induced Forces in Dense Sphere Packings.
z / H and in time, Tv is shown on Fig. 5. The Transport in Porous Media 92(2), 473493
(2012).
pore pressures results are normalized P / Pmax to Cundall P.A., Strack O.D.L.: A discrete numerical
the external stress value Pmax . The evolution in model for granular assemblies. Geotechnique
space and time of the pore pressure is found to 29, 4765 (1979).
be in good agreement with the analytical DiRenzo A., DiMaio F.P.: Comparison of contact-
solution of Terzaghi (see Fig. 5). force models for the simulation of collisions in
DEM-based granular flow codes. Chemical
Engineering Science 59(3), 525541 (2004).
Jing L., Ma Y, Fang Z.: Modeling of fluid flow and
solid deformation for fractured rocks with
discontinuous deformation analysis (DDA)
method. International Journal of Rock
Mechanics and Mining Sciences 38(3), 343
355 (2001).
Ma J., Wu K., Jiang Z.: An implementation for lattice
Boltzmann simulation in low-porosity porous
media. Physical Review E 81(5) (2010).
Tong A.-T., Catalano E., and Chareyre B.: Pore-
Scale Flow Simulations: Model Predictions
Compared with Experiments on Bi-Dispersed
Granular Assemblies. Oil & Gas Science and
Figure 5. Comparisons between the pore pressure Technology – Revue d’IFP Energies
and Terzaghi’s analytical solution nouvelles, 67(5), 743–752 (2012).
Tong A.-T.: Modélisation micromécanique des
4. CONCLUSION couplages hydromécaniques et des mécanismes
d’érosion interne dans les ouvrages
The DEM-PFV model was applied to the hydrauliques, http://www.theses.fr (2014).
simulation of the settlement of soil saturated in Tong Anh Tuan and Bruno Chareyre.:
embankments due to static loading. The Micromechanical modelling of internal
evolution of settlement and pore pressure in erosion by suffusion using DEM-PFV coupled
space and time were analysed and discussed. model. In: Vietnam – Japan Workshop on
The numerical results obtained were in good Estuaries Coasts and Rivers. Hoi An,
agreement with Terzaghi’s analytical solution. Vietnam, 7-8 Sept (2015).

122
The 3rd Int. Conf. on Transport Infrastructure & Sustainable Development (TISDIC 2019). Construction Publishing House. ISBN: 978-604-82-2893-4.

Re-categorized soil layers and performance prediction of

deep excavation in Kuala Lumpur

Phan Hai Khac*, Bin-Chen Benson Hsiung, Nguyen Giang Hong

Department of Civil Engineering – National Kaohsiung University Science and Technology, Kaohsiung,
Taiwan-China

Corresponding author, E-mail: [email protected]

ABSTRACT: This paper evaluates the performance of a 13.2m deep excavation in the city centre, which
mainly rested on the weather residual soil of the Kenny Hill Formation and 3- dimensional simulation of
the excavation is carried out. Results from some current experiment tests as well as empirical approaches,
are adopted in order to define the values of soil stiffness. The ground profiles then are re-categorized into
three including of (1) Upper Kuala Lumpur sand (UKLS), which is the loose to medium dense sand; (2)
Lower Kuala Lumpur sand (LKLS), which is dense to very dense sand; and (3) Kuala Lumper sand/silt-
stone (KLSS), which is the sandstone and siltstone layer. The three-dimensional (3D) numerical
simulation using the commercial software PLAXIS 3D is adopted to predict the performance of deep
excavation stated above. By using the 3D studies, the optimization of the soil stiffness from in-situ test
data for UKLS is also explored through a comparison of field measurement data with back-analysis
results. At last, a comparatively better consistency was found between computed and measured data in
terms of wall deflections obtaining from the use of constitutive soil model of Hardening Soil model rather
than the Mohr-Coulomb model. In addition, alternative definitions of soil stiffness of various layers of
sands in Kuala Lumpur are also approved.
Keywords: Kenny Hill formation; deep excavation in sand; pressuremeter test (PMT); constitutive soil
models; soil stiffness.

1. INTRODUCTION a major of rising concern for these

developments is the ability of geotechnical
Applying reliable and available information engineers to understand the excavation
and modern technology into geotechnical performances and to eliminate the possible
design and analysis is eligible to predict impacts of the construction stage to their
ground and structure displacements induced by adjacent structures. For instance, the
deep excavations. Moreover, it aims to explore engineering geologic problems in this city have
and compare induced ground displacements been well documented by Tan, 2006. One of
once input soil parameters are obtained from ground formations in the research area is the
various approaches in this paper. particular weathered residual soils on top of
The rapid urbanization of Kuala Lumpur sedimentary rocks, named as Kenny Hill
City of Malaysia requires a lot of underground Formation. It is aware that undisturbed soil
constructions to solve for sustainable samples are difficult to be obtained. Thus, for
development under highly urbanized areas such having high quality of soil test data (Law et al.,
as car-park basements, underground terminal 2014), it is indicated that the soil modulus is
stations, and road & railway tunnels. However, unaccepted to be obtained through laboratory

123
test due to comparatively low quality of reinforcement concrete diaphragm wall. The
samples. According Hiung et al., (2017), bottom-up method was adopted for the
Young’s modulus is the most essential key construction, in which the sequences of
parameters to predict the excavation behaviors, construction as following: (i) installation of
adopting available information from back diaphragm wall, (ii) excavation and
analysis of soil stiffness often base on the case installation the strutting system, (iii)
of local histories; therefore, it is crucially excavation until the final depth, (iv)
important for engineering to make some construction of the underground structure, and
judgments in the prediction of excavation (v) backfilling and reinstatement respectively.
performance. Moreover, the correlation of soil
stiffness parameters against standard 2.2 Subsurface conditions
penetration test (SPT-N) value commonly used Fig.1 depicts the cross-section and subsurface of
to represent the actual soil condition. the excavation and supporting systems released
However, Law et al., (2014 and 2016) by Law et al. (2014, 2016). The ground profiles
pointed out that the average values of SPT-N mainly consist of sandy silt layers as upper recent
of an excavation located in Kuala Lumpur was alluvium deposits underlined by residual soils and
undertaken as a reference base for definition of weathered rocks of the Kenny Hill Formation.
soil stiffness for the simulation. However, the According to Lim, 1995, this formation is a
reliability of SPT-N is commonly doubted process of interbedded sedimentary rocks, which
based on previous experiences. Alternatively, it had undergone mild and regional metamorphic
is recommend using the other method to materials, such as siltstone and shales/mudstone
approach the soil stiffness from the field test overlain by stiff over-consolidated soils
instead in this paper. The weathered residual predominately of sandstone/siltstone to quartzite
soil of Kenny Hill Formation thus re-categorize and schist/phyllite respectively. The ground
into three groups of soils. Considering the condition at the site includes a 7.5m thick of silty
excavation performance, 3D finite element sand layer (recent alluvium deposits) underlain by
analysis using software PLAXIS 3D for residual soils of Kenny Hill Formation. The bulk
analyzing are delivered. The research selected density of the residual soil layer generally ranges
Hardening Soil (HS) model, which is from 19kN/m3 to 22kN/m3 with depth. The
considered more suitable rather than Mohr- groundwater table is observed at approximately
Coulomb (MC) model for a stress- relief case, 2.5m below the ground surface.
to further validate the key soil parameters for
excavation simulation.

2. PROJECT BACKGROUND
2.1 Brief project description
The case history of deep excavation in Kuala
Lumpur City, Malaysia has been selected for
the study and numerical simulation. The
project including three levels of the car-park
basement; the basement excavation is a
rectangular, approximately 43m in width and
78m in length. It was retained by using a 0.6m
thick and 20m deep for diaphragm wall. The
maximum excavation depth is up to 13.2m
and the pit is only supported by a single level
steel H-beam strut (400  400mm) at 4 m Figure 1. Excavation stages and soil profile in
below the ground surface, except cross-section

124
 As indicated previously, SPT-N value is Upper Kuala Lumpur sand (UKLS), which
often used as an index of subsurface is the loose to medium dense sand ranged from
information for the geotechnical design. The 0 to 13.5 mbgl (SPT-N < 30).
average SPT-N at layer 1 as upper recent Lower Kuala Lumpur sand (LKLS), which
alluvium is around 7.5 blows/300mm; is dense to very dense sand ranged from 13.5 to
however, the SPT-N increases from 20 to 100 25.5 mbgl (30 < SPT-N < 100).
blows/300mm at the deeper soil layers Kuala Lumpur sand/silt-stone (KLSS),
corresponding with layer 2 to layer 5 in this which is the sandstone and siltstone layer
beyond 25.5 mbgl (SPT-N > 100).
study. The highest SPT-N was recorded up to
There are three linear relationships have
150 blows/30mm at layer 6 in which believed
been separately established amongst these
that the presence of quartz veins or phyllite
group layers in terms of Young’s modulus.
fragments encountered in the boreholes. A
As for the first group layer, implementation
classification of the soil layers consist of three of Young’s modulus comparison between
groups can be found on the site on the base of UKLS with KHH sand aim to evaluate the
SPT-N (Skempton, 1986) which including of consistency of soil stiffness obtained from
loose silty sand at the first layer (SPT-N < 10), the PMT test. Hsiung et al,. (2016)
medium dense sandy silt at the second and mentioned that Young’s modulus of loose to
third layers (10 < SPT-N < 30), and hard medium dense sand can be generated by
material (SPT-N > 50). using the empirical Eq. 1. It is obvious that
the tendency of UKLS soil stiffness
3. RE-CATEGORIZED THE SOIL LAYERS interpreted from in-situ tests were pretty
smaller than that of in KHH sand.
3.1 Soil stiffness E  1250( z  9.7) (with z is the depth of soil layer)
Adopting from data of over 30 available (1)
boreholes pressuremeter tests, said action Regarding the second group layer, the
might improve the consistence of LKLS is considered to be dense to very dense
characteristics of current weathered residual sand material. Thus, it appears to like the
soils of the Kenny Hill Formation, especially intermediate and lowest unit of Berlin sand
for the soil stiffness used in design and analysis material with the void ratios e00.5-0.6 and
e00.3-0.4, respectively. It is noted that the
for deep excavation. These tests were also
Berlin sand unit weight commonly used
taken in central Kuala Lumpur city, which their
=20kN/m3 and the effective friction angle
locations are not far away from the current
roundly 33-35o (Schweiger, 2002;
excavation project, with various depths up to Nikolinakou et al., 2011). Ensure that few
30 m below the ground surface. In addition, the PMT tests have been conducted in this layer,
similar soil characteristic such as Kaohsiung a combination of Young’s modulus between
(KHH) sand (in Taiwan) and Berlin sand (in LKLS and Berlin sand is made to seek for the
Germany) are selected to compare with this most consistency of soil parameters.
Kuala Lumpur center city sand. These points Similarly, the last group layer is the KLSS
indicated that the ground profiles herein are layer with the SPT-N up to 150
simply redefined based on the soil stiffness of blows/300mm. Although there were three
each soil layer. available PMT tests checked in this group
Fig.2 shows the soil classification, there are layer, a linear interpolation of soil stiffness
three types of soil can be found in the site from test results was adopted to be used in
categorized as following group descriptions: analyses later.

125
Figure 2. Schematic outline for soil re-categorized of Kenny Hill Formation

3.2 Analysis methodology iii. For KLSS layer: EKLSS  1588740 z  37083380 (kPa)
(5)
Three scenarios of back calculation analysis The third scenario developed is based on the
(a.k.a inverse analysis) are made based on the assumption that the UKLS layer is similar to
in-situ and laboratory tests (SPT, PMT, CU) the one in KHH sand. In other words, the
and empirical equations from related soil relationship between Young's modulus and
characteristics (Kaohsiung sand and Berlin depth (Eq.1) is selected to estimate the
sand) to obtain reasonable stiffness properties excavation deformation. In the meanwhile, the
of residual soils for estimating the soil stiffness of LKLS and KLSS layers are
deformation prediction of deep excavation. still adopted from Eq. 4 and Eq.5.
Details of these scenarios are as following
below and summarized in Table 1. Table 1. Adopted soil stiffness in numerical
The first scenario would be made to back- back-analyses
calculate the excavation deformation on the
Young’s modulus, E (kPa)
basis of current SPT-N. The soil stiffness is No Scenarios
generated by using the Eq.2 recommended by UKLS LKLS KLSS
Hsiung et al., (2009) for estimation of 1 SC. 1 Eq. 2 Eq. 2 Eq. 2
Young’s modulus of sandy soil related to
SPT-N: 2 SC. 2 Eq. 3 Eq. 4 Eq. 5
E  2000N (kPa) (2) 3 SC. 3 Eq. 1 Eq. 4 Eq. 5
The second scenario established is based on
three linear interpolations of Young’s modulus 4. FINITE ELEMENT ANALYSIS
of each group soil layers as stated above which
are simply rewritten by 3 equations below: 4.1 Numerical model
i. For UKLS layer: EUKLS  1961.8z  1347.2 (kPa) (3) In this paper, PLAXIS 3D software (version
ii. For LKLS layer: ELKLS  10396 z  28900 (kPa) (4) 2017) is used to conduct the deformation

126
prediction of deep excavation. Fig. 3 computational time and cost without
represents a one-quarter of an excavation affecting the numerical results
model due to an advanced of symmetry
simulation. Hence, the total dimensions of
the Finite Element (FE) model were reduced
by half of each horizontal and vertical
direction, approximately 115 m and 135 m,
respectively. These distances were
approximately seven times the maximum
depth of excavation, He, militating against
the effects of boundary conditions (Finno et
al., 2007). The depth of model boundary is
chosen up to 40 m due to absence of data for Figure 3. The 3D finite element mesh and boundary
subsoil conditions. In accordance with the
symmetric model, the sizes of excavation
4.3 Soil and structure input parameters
were modeled by half of their primary and
complementary wall lengths equal to 37.5m The soil was assumed to behave like a
and 21.5m, respectively. It is also noticed perfectly plastic material, described by the
that the influences of corner effects become Mohr-Coulomb model. The model consists of
insignificant once the distance is 30 m from six input parameters such as the effective
the corner. In other words, the excavation is friction angle', cohesion c’, Poisson ratio
considerably under the plain strain ', dilation angle , coefficient lateral earth
conditions and the wall deflection remains pressure at rest K0, and Young’s modulus E’.
stable at distance of 30 m in primary wall The descriptions of soil properties for three
length (Ou, et al., 1996; Hsiung et al., 2016; group layers are listed in Table 2; Young’s
Hsiung et al., 2018). In addition, the FE moduli of soils are adopted corresponding to
model only comprised 82479 10-node each scenario in the table. The drained
tetrahedral elements with a total of 120801 behaviors are assigned for UKLS and LKLS;
notes. Again, using a one-quarter model however, the KLSS layer is assumed to
reduces the sizes of the 3D model, saving possess undrained behavior.

Table 2. Simplify soil properties of case KLCC

Depth f' c’ Young’s modulus, E (kN/m2) y'

Layers K0
(m) ( o) (kN/m2) Scenario 1 Scenario 2 Scenario 3 (o)

7.5 28 1 1.5E+04 6.0E+03 1.7E+04 0 0.531

UKLS 10.5 30 5 4.0E+04 1.6E+04 2.3E+04 0 0.500
13.5 31 5 6.0E+04 2.2E+04 2.7E+04 1 0.485
18 33 8 1.0E+05 1.9E+05 1.9E+05 3 0.455
LKLS
25.5 35 15 2.0E+05 2.6E+05 2.6E+05 5 0.426
KLSS 30 35 100 3.0E+05 7.0E+06 7.0E+06 5 0.426

127
 The structural elements were assumed as a installed into the diaphragm wall with its
linear elastic material. The temporary strut length is deeper than wall toe to 3m. Two
systems were modeled by the anchor inclinometers located at one side of the
element; meanwhile, the plate element used complementary wall, and four inclinometers
for the diaphragm wall. It is noted that the installed at each side of the primary wall.
diaphragm wall is assumed “wished-in-place” Since the one-quarter model has been made,
in which the installation effects of diaphragm only three available inclinometers are
wall was considered to be negligible in terms selected to make comparisons between
of changes in stresses and deformations. predictions with field data measurements in
Besides, the stiffness of wall is reduced by terms of wall deflections. IN-6, IN-7, and
approximately 30% and 40% of reduction in IN-8 are standing along the complementary
strut stiffness as suggested by Ou (2006). wall, location at a distance equal to one-four
Table 3 lists the input material properties of of wall length, and installation at the
diaphragm wall and the strut input parameters approximately center of long walls
are listed in Table 4. Further details of respectively.
structural information can be found in these
previous works of Law et al., 2016 and Ang
et al., 2018

Table 3. Diaphragm wall elastic properties

(Plate element)
Parameter Name Value Unit
Compressive fc’ 40 MPa
strength of
concrete
Thickness d 0.6 M
Young’s E 28  106 kPa
modulus
Young’s 70% 19.6  106 kPa
Figure 4. Plan view of inclinometers arrangement
modulus E
3
Unit weight w 9 kN/m 5.2 Results of analysis
Poisson’s ratio v 0.2
Fig. 5 shows that the comparisons of the wall
Table 4. Steel struts properties (Anchor element) deflections at first and final excavation (stage
of excavation I and II) among three scenarios
Level Strut Section EA 60%EA
size Area (kN) (kN)
at three cross-sections associated the location
(m2) of wall inclinometers such as IN-6, IN-7, and
H400 IN-8. The results of predictions were
Level
 0.0219 4.483106 2.6898106 compared to the field data measurements as
1
400 well, to further validate the suitability of each
scenario. Fig 5a also pointed out the wall
5. RESULTS OF ANALYSIS deflections were measured by the inclinometer
5.1 Seleted inclinometers namely IN-6, which installed nearby the
corner of the complementary wall. The
Fig.4 displays the locations of 10
inclinometers to observe the horizontal wall maximum of wall deflections was
movements. These instruments were approximately 20 mm at the ground surface

128
and 42mm at the middle of the wall within due to insignificant of corner effect.
stage I and stage II, respectively. Among the Similarly, the computed wall deflections
three scenarios generated a similar tendency were larger than that of the previous
of wall deflection. However, the second inclinometer, IN-7, among three scenarios.
scenario predicted higher maximum lateral Though the numerical prediction of the
wall movement values at the first stage up to second scenario was in excellent agreement
1.5 times than others, meanwhile, there was with the measured maximum wall deflection
little difference between the first and third in stage II, the significant overestimates were
scenario. Although the wall deflections at found in the top and toe of the wall even
stage II among three scenarios seem similarity roundly two-fold of increasing wall
of tendencies and values, there were the only deflection in stage I. The numerical analyses
close agreements with the field data predicted maximum wall deflection from the
measurement at nearby the top and toe of the first and third scenario up to 88 mm at the
wall, whereas the underestimates of the middle of the wall, in which 10mm is lower
maximum wall deflections were found less than that of reading data. Meanwhile, these
than 5%. predictions were pretty consistent with
Regarding the IN-7 shown in Fig.5b, the shallow depths; and hence, the overestimates
numerical analyses were overestimated the of wall deflections are most likely related to
wall deflections compared with the the soil stiffness of LKLS below. In this
measurement data in both excavation stages. case, the soil properties appeared to reflect
However, the field measurement data was higher strength in this group layer. Therefore,
the soil stiffness of LKLS should be taken
pretty closer to computed at the first and third
into account by selecting more relative
scenario, whereas the maximum wall
reference data from field and laboratory soil
deflections of the second scenario were
tests. Further analysis and study are thus
greater than others by approximately 2 times
necessary to obtain accurate predictions.
at ground surface and 1.4 times at the middle
Above all, scenario 1 and 3 appear to
of diaphragm wall corresponding with stage I
generate reasonable values of wall
and stage II. The unreported diagonal struts
deflections than scenario 2 in terms of
installed nearby the corner may probably
numerical analysis. The soil stiffnesses of
strengthen wall system stiffness, and hence,
each group soil listed in Table 1; therefore,
the wall deflections from reading data were
it can be adopted to design the Kenny Hill
smaller than that of predictions.
Formation. As a result, empirical equations
Unfortunately, the strutting system in this
can be obtained to predict excavation
model was re-arrangement as suggested by
performance. More importantly, the UKLS,
Ang et al., (2018) due to lack of their details
which imply for loose to medium dense
of coordinates and spacing of lateral sand, is better to obtain the similarity to
supporting system. Kaohsiung sand characteristic in terms of
As for the IN-8, this section was located at soil stiffnesses rather than current PMT test
the center of excavation (Fig. 5c). Therefore, or even SPT test. The LKLS and KLSS can
the measured maximum wall deflections (i.e. be obtained from the results of the PMT test,
46mm and 98mm) were greater than that and higher Young’s modulus should be
of finding at IN-7 (i.e.  30mm and  60mm) further studied.

129
 Lateral wall movement (mm) Lateral w all movement (mm)
0 40 80 120 Lateral w all movement (mm)
0 40 80 120 0 40 80 120
0 0 0

Strut level at 4m Strut level at 4m Strut level

at 4m
5 5 5
1st Exc. 1st Exc. 1st
to 4.5mbgl to 4.5 mbgl Exc.
to 4.5
Depth (mm)

Depth (mm)

Depth (mm)
mbgl
10 10 10

15 15 Final Exc. 15 Final Exc.

Final Exc. to13.2 mbgl
to13.2 mbgl to13.2 mbgl

Field data Simulations

20 Stage Stage 20 20
Scenarios
I II I II

25 25 25

(a) IN-6 (b) IN-7 (c) IN-8

Figure 5. Comparison of predicted and measured wall deflection at each stage of excavation among 3 scenarios

stiffness, E ref , triaxial unloading/reloading

6. DISCUSSIONS 50

stiffness, Eurref , and oedometer loading tangent

stiffness, Eoed , at the reference pressure pref that is
ref

The preceding numerical analyses using MC

model have shown that it is possible to obtain usually taken as 100 kPa (1 bar). It is not like the
reasonable estimates of wall deflections by using MC model in which the stiffness parameters are
the results of in situ tests as well as empirical constant, the HS model takes into account the
equations. The third scenario is obtained to stress dependency of soil stiffness. The HS
further validate the current input soil stiffnesses model requires 10 material parameters to
by using HS model (Schanz et al., 1999) for characterize the behavior of sand, these data are
UKLS and LKLS group layers. Meanwhile, the listed in Table 5. It is noticed that the almost soil
MC model still used for the last group KLSS due parameters are similar to the MC model
to having slightly effects on wall movements. In (exception of soil stiffnesses). E is approximately
equal to 1.5 times E50, and Eoed were set equal to 3
ref

the HS model, the soil stiffness is elaborately

calculated by using three additional soil times E ref and E ref respectively as suggested by
50 50

stiffnesses instead of Young’s modulus, E, in Schanz et al., (1999); Schweiger (2009); Hsiung
MC model such as triaxial loading secant et al, (2017); and Ang et al., (2018).
Table 5. Input parameters of UKLS and LKLS group layers for the HS model.
Layer Depth E50ref Eoedref Eurref c'  φ'  '  vur  m Rf Ko
7.5 2.0E+04 2.0E+04 6.1E+04 1 28 0 0.2 0.5 0.9 0.531
UKLS 10.5 2.1E+04 2.1E+04 6.4E+04 5 30 0 0.2 0.5 0.9 0.500
13.5 2.2E+04 2.2E+04 6.7E+04 5 31 1 0.2 0.5 0.9 0.485
18 1.5E+05 1.5E+05 4.4E+05 8 33 3 0.2 0.5 0.9 0.455
LKLS
25.5 1.7E+05 1.7E+05 5.2E+05 15 35 5 0.2 0.5 0.9 0.426
m
 c 'cos  '  'sin  ' 
Notice: The following relationship is used to calculate E50 (PLAXIS, 2017), E50  E50ref   , which ’3 is the
 c 'cos  ' p sin  ' 
ref

minor principal stress, a positive value for compression; m is the power determining the rate of variation of E50 with ’3

130
Figure 6. Predicted excavation performance for the middle of excavation section (IN-8)

Fig. 6 plots of the predictions obtained by changes from triangle to trapezoidal shape
both MC and HS model parameters at the corresponding from first excavation to final
middle of excavation (inclinometer IN-8 excavation. The loose sand in the UKLS group
section). The slightly different gap between layer may lead to producing larger horizontal
these models was found in terms of wall wall movements and ground surface settlement
deflections and ground surface settlements. once the excavation depth is reached to the
Regarding the wall deflections, although the final design. As the results, the maximum
tendency of wall movement is similar in stage values occurred at 3m and 6m away from the
I. Almost 6% higher maximum value of wall wall computed from the MC and HS model,
deflection obtained from MC model compared respectively. These distances are similar the
to the HS model. It extends that wall ranges of maximum ground surface settlement
deflections from the HS model were closer to for soft to medium clay suggested by Clough
reading data than the that of the MC model for and O’Rourke’s (1990) as 0   d/He  0.75
the shallow excavation. Similarly, the HS (with d is the distance away from the wall, and
model showed closer values of wall deflections He is the depth of excavation). However, the
at the top and middle of the wall versus transition zones, where settlement decreases
measured data than that of the MC model. from the largest to almost none, are pretty
However, only 3.5% of the differential value of larger than their suggested of d/He  2 (d/He =
maximum wall deflection has been observed 2.5 in MC and d/He = 3 in HS). For this case
between the two models. As for the ground study, the influence range of ground surface
surface settlement induced by excavation, the settlement is up to 3He.
MC model computed slightly higher maximum
settlement at the final excavation than HS 7. CONCLUSIONS
model (i.e. 92mm versus 88mm), while, similar
maximum values were found between two The performance of a case history deep
models at the first stage (i.e. 44mm) located excavation in residual soils of Kenny Hill
nearby the retaining wall due to the shallow Formation has been described in this paper.
excavation. Unfortunately, there is not The subsurface conditions were re-categorized
available data evaluates those results. on the basic of soil stiffness those calibrated
It is obviously seen that the envelopes of the from the local field soil tests, PMT, laboratory
ground surface settlement are significant tests, and empirical equations of similar soil

131
condition. The deep excavation was used 8. REFERENCES
finite–element simulation by using the MC
model to select the reasonable scenario into Ang J. S., Hsiung B. C. B. and Ching Hung 2018).
design and analyses. Furthermore, the HS Plane strain ratio and waling size evaluation of
deep excavation of Kuala Lumpur using 3d
model was delivered too in order of validation
finite element analysis. Proceeding 20th
the soil stiffness and further potential
SEAGC – 3rd AGSSEA Conference in
application in the practical deep excavation. conjunction with 22nd Annual Indonesian
Following below conclusions are drawn to National Conference on Geotechnical
summarize the major points in this study: Engineering. Jarkata – Indonesia, 6 –
Three major group soil layers of Kenny Hill November 2018. ISSN 978-602-17221-9. Pp.
Formation include of (1) Upper Kuala Lumpur 193-201
sand (UKLS), which is the loose to medium B.K. Tan 2006. Urban geology of Kuala Lumpur
dense sand; (2) Lower Kuala Lumpur sand and Ipoh, Malaysia, IAEG. Paper number 24,
(LKLS), which is the dense to very dense sand; The Geological Society of London, 2006.
and (3) Kuala Lumpur sand/silt-stone (KLSS), Clough, G.W and ORourke, T.D. 1990.
which is the sandstone and siltstone. Construction induced movements of in situ
The numerical simulations were pretty walls, Design and Performance of Earth
agreement with measured diaphragm wall Retaining Structures. ASCE Special
deflections by using the first scenario and third publication, No.25, pp 439-470.
scenario instead of using the second scenario. Hsiung, B. C. B. 2009. A Case Study on the
The paper recommends using the second Behavior of a Deep Excavation in Sand.
scenario to analysis due to inadequate Comput. Geotechn. Pp.665-675
confidences about SPT-N values on this site. Hsiung, B. C. B., Yang, K. H., Aila, W., Hung, C.,
The similar soil characteristic from 2016. Three-dimensional Effects of a Deep
Excavation on Wall Deflections in Loose to
Kaohsiung sand (loose to medium dense sand)
Medium Dense Sands. Comput. Geotechn.
is selected to use in UKLS group layer in this
Hsiung, B. C. B., Yang, K. H. 2017. Displacement
study in terms of soil stiffness. The Young’s of ground and diaphragm wall induced by
modulus herein calibrated by using the deep excavations in loose to medium dense
function with depth: E=1250(z+9.7). sand. Proceedings of the 19th International
Meanwhile, the soil stiffness of LKLS group Conference on Soil Mechanics and
layer is calibrated from the local field tests, Geotechnical Engineering, Seoul, 2017
PMT, and the soil stiffness of the dense sand in Hsiung B. C. B., Yang K. H., Aila W. and Ge L.
Berlin as E=10396(z+2.78) As for the KLSS 2018. Evaluation of the Wall Deflections of
group layer, the available soil stiffness from Deep Excavation in Central Jakarta using
PMT tests was obtained by using equation Three-dimensional Modeling. Tunnelling and
E=1588740(z-23.35) Underground Space Technology, Vol. 72,
The more complex soil model of HS gave pp.84-96.
better agreement with reading data rather than Law K. H., Ismail Z. and Roslan H. 2016. 3D
the MC model. The setup values of soil Finite Element Analysis of a Deep Excavation
stiffness are followed the empirical equations Considering the Effect of Anisotropic Wall
that can be applied for UKLS and LKLS to Stiffness. 19th Southeast Asian Geotechnical
predict the performance of excavation in Conference & 2nd AGSSEA Conference
Kenny Hill Formation. (19SEAGC & 2AGSSEA) Kuala Lumpur,
The trapezoidal envelope shape of ground pp659-664.
Law K. H., Siti Z. O., Roslan H and Zubaidah I.
surface settlement has been found in the final
2014. Determination of Soil Stiffness
excavation. The maximum ground surface
Parameters at a Deep Excavation Construction
settlement located at the distance away from Site in Kenny Hill Formation. Measurement,
the diaphragm wall, d, less than 0.75 times the Vol. 47, pp. 645-650.
excavation depth, He. It is also noticed that the Lim, T.H. 1995. Some engineering geology
maximum influence distance of ground surface characteristics of the Kenny Hill Formation,
settlement is up to 3 times He. Kuala Lumpur. Warta Geologi, 21(1), 9-11

132
Nikolinakou, M.A., Whittle, A.J., Savidis, S., Skempton, A. W. 1986. Standard Penetration Test
Schran, U., 2011. Prediction and interpretation Procedures. Geotechnique, vol. 36, no. 3,
of the performance of a deep excavation in pp.425-557.
Berlin sand. Journal of Geotechnical and Schweiger, H. F. 2002. Benchmarking in
Geoenvironmental Engineering 137, 1047- geotechnics. Part 1: Results for benchmarking;
1061. Part 2: Reference solution and parametric
Ou, C. Y., 2006. Deep Excavation: Theory and study. CGG_IR006_2002, Institute for Soil
Practice. Taylor & Francis, Netherlands Mechanics and Foundation Engineering,
Ou, C. Y., Chiou, D.C., and Wu, T. S. 1996. Three- Computational Geotechnics Group, Graz
dimensional Finite Element Analysis of Deep Univ. of Technology, Graz, Austria.
Excavations. Journal of Geotechnical and Schweiger HF. 2009. Influence of constitutive
Geoenvironment Engineering, pp337-345 model and EC7 design approach in FEM
Schanz, T., Vermeer, P. A., and Bonnier, P. G. analysis of deep excavations. PLAXIS,
1999. The hardening soil model: formulation ISSMGE International Seminar on Deep
and verification. Beyond 2000 in Excavations and Retaining Structures, vol.
Computational Geotechnics, Balkema, 3D. Budapest, pp. 99-114, 2016.
Rotterdam, pp281-296.

133
The 3rd Int. Conf. on Transport Infrastructure & Sustainable Development (TISDIC 2019). Construction Publishing House. ISBN: 978-604-82-2893-4.

Suffusion investigations under two flow directions

Tran Dinh Minh 1,2,, Didier MAROT 2, Fateh BENDAHMANE 2

1
The University of Danang- University of Science and Technology, Da Nang, Vietnam
2
University of Nantes, Institute GeM, UMR CNRS 6183, 58 rue Michel Ange, Saint-Nazaire, France

Corresponding author, E-mail: [email protected]

ABSTRACT: Internal erosion processes in earth structures and their foundations may increase the failure
risk of such structures. Suffusion, one of the main internal erosion processes, selectively erodes the fine
particles that move through the coarser particles. Thus to ensure the safety assessment of hydraulic earth
structures, it is important to characterize the suffusion susceptibility. In literature, most of the tests were
performed under vertical flow, whereas the flow in earth structures can be horizontal as well. To study the
potential influence of the flow direction, a new apparatus is developed. Two cohesionless soils were
selected with a slight variation of the grain size distributions. A series of suffusion tests was carried out to
verify the erosion susceptibility under vertical and horizontal flows. The post-suffusion gradations show
that under horizontal flow, the specimens become more heterogeneous. Moreover, both tested soils
appear more resistant under horizontal flow. These results demonstrate the significant influence of the
seepage flow direction and the interest of the new device.
Keywords: internal erosion, suffusion, experimental characterization, energy, flow direction.

medium can be the catalyst for slope instability

1. INTRODUCTION at the scale of hydraulic embankments (Fry et
Internal erosion is one of the main causes of al., 2012). Thus, to ensure the safety
instabilities within hydraulic earth structures assessment of hydraulic earth structures, the
such as dams, dikes or levees (Foster et al., suffusion susceptibility must be characterized.
2000). Fell and Fry (2013) proposed to According to Wan and Fell (2008),
distinguish four types of internal erosion: suffusion will occur in sand-gravel soils if the
concentrated leak erosion, backward erosion, three following criteria are satisfied: (i) the size
contact erosion, and suffusion. This paper of the fine grains must be smaller than the size
focuses on suffusion which is a selective of constrictions between coarser grains; (ii) the
erosion of fine particles under the effect of amount of the fine grains that fills the pores of
seepage flow within the matrix of coarser the soil skeleton formed by the coarser grains
particles. It is recognized that suffusion may must be small. If the volume of the fine grains
cause changes in porosity which can lead to is higher than the volume of the void, then
important modifications in the hydraulic and coarse grains will float in a matrix of fine
mechanical characteristics of the soil (Chang grains; (iii) the velocity of seepage flow
and Zhang, 2011; Ke and Takahashi, 2012; through the soil matrix must be high enough to
Marot et al., 2009; Moffat et al., 2011, among put out the detached fine grains through the
others). Finally, modifications of the porous pores of the coarser grains. In the same

134
manner, Garner and Fannin (2010) describe the be explained by the presence of preferential
main initiation conditions for suffusion with flows created by the erosion process leading to
the aid of a diagram comprising three a steady state. With the objective to consider
components: the material susceptibility, the the influence of the difference of head and that
critical stress condition, and the critical of the flow rate, Marot et al. (2011) proposed a
hydraulic load. new analysis based on the total flow power.
The two first criteria are mainly governed by The total flow power is the summation of the
the grain size distribution, the constriction size power transferred from the fluid to the solid
and the mechanical states. In this context, particles and the power dissipated by the
Kenney and Lau (1985), Wan and Fell (2008), viscous stresses in the bulk. According to
among others proposed different geometric Sibille et al. (2015), the transfer appears
criteria based on the size distribution of negligible in the case of suffusion, so the
particles to assess the potential of soil to authors suggest to characterize the fluid
suffusion. The finding of Moffat et al. (2011) loading from the total flow power which is
permits to identify the less conservative expressed by:
criterion between the Kenney and Lau’s P flow   γ w Δz  ΔP  Q (1)
criterion and Kézdi’s one. With the same where w is the specific weight of water,
objective, Marot et al. (2016) compared the z = zA - zB, zA and zB are altitudes of sections
aforementioned criteria with the criterion A and B respectively, P = PA - PB is the
proposed by Chang and Zhang (2013), in order pressure drop between the sections A and B
to identify three grain size distribution criteria respectively and Q is the volumetric water flow
that can be used for assessing the potential rate. z > 0 if the flow is in the downward
suffusion susceptibility. However, the direction, z < 0 if the flow is upward and the
detachment and transportation of particles
erosion power is equal to Q P if the flow is
induced by suffusion is also related to the
horizontal.
action of the fluid phase. Based on tests
performed under upward flow, Skempton and The expended energy Eflow is the time
Brogan (1994) proposed to characterize the integration of the instantaneous power
suffusion initiation by the critical hydraulic dissipated by the water seepage and it is
gradient. Reddi et al. (2000) showed that fine computed until reaching the steady state (i.e.
particles which do not pass through the hydraulic conductivity tends to stabilize
constrictions of coarse particles can be clogged and the erosion rate tends to decrease). For the
and this clogging induces a decrease of the same duration, the cumulative eroded dry mass
hydraulic conductivity. Rochim et al. (2017) is determined and the erosion resistance index
performed suffusion tests under downward is expressed by:
multistage hydraulic gradient and showed that  Cumulative eroded dry mass 
I α =  log   (2)
suffusion process can be described in three  E flow 
steps. First, the hydraulic conductivity
decreases simultaneously with the decrease of Depending on the values of Iα index, Marot
erosion rate, which suggests that some et al. (2016) proposed six categories of
detached particles can be filtered within the suffusion susceptibility, from highly erodible
soil itself. In a second stage, a rough increase to highly resistant (corresponding susceptibility
of the erosion rate occurs simultaneously with categories: highly erodible for Iα < 2; erodible
the increase of the hydraulic conductivity. for 2 ≤ Iα < 3; moderately erodible for
Thus, the predominant process during this 3 ≤ Iα < 4; moderately resistant for 4 ≤ Iα < 5;
second phase seems to be the detachment and resistant for 5 ≤ Iα < 6 and highly resistant for
transport of solid particles. Finally, the Iα ≥ 6).
hydraulic conductivity tends to stabilize while For characterizing the initiation and
the erosion rate decreases. This third phase can development of this complex internal erosion

135
process, experimental devices comprise a rigid
wall permeameter (Kenney and Lau, 1985;
Skempton and Brogan, 1994; Wan and Fell,
2008; Moffat and Fannin, 2006; Sail et al.,
2011) or a modified triaxial cell designed to
force fluid to percolate throughout the sample
(Bendahmane et al., 2008; Marot et al., 2009;
Chang and Zhang, 2011; Ke and Takahashi,
2014; Slangen and Fannin, 2017). These
laboratory devices permit to apply a vertical
flow on upward or downward direction.
However, the suffusion process is induced by Figure 1. Schematic diagram of the new apparatus
seepage flow and the hydraulic conductivity of
compacted soil is not an intrinsic parameter The soil sample is contained in an acrylic
because it depends on the soil fabric (Smith cylinder that is 80mm in inner diameter and
and Konrad, 2011) and on the flow direction, 500mm in height. The height of specimens can
which can be in site completely different from vary from 300mm to 400mm. The cell is fixed
vertical. by two vertical beams with two clamps to
In this context, the main objective of this avoid any disturbance and to provide support
research is to develop a multidirectional flow for the inlet and outlet ports, which are used in
apparatus which can be used to study soil the case of a horizontal flow along the
suffusion susceptibility under vertical or specimen diameter. The cell top end is closed
horizontal flow. by a cap equipped with an inlet port, an
opening for the passage of the axial load ram
2. APPARATUS, TESTED SOILS AND and a vent. The piston contains a 15mm thick
METHODOLOGY layer of glass beads to diffuse the fluid
uniformly on the specimen top surface. The
2.1 Experimental setup specimen is positioned on a 10mm opening
The general configuration of the testing size grid and a mesh with a selected pore
apparatus is shown in Fig. 1. This device was opening size can come in between. The same
designed to permit the injection of water in type of mesh is introduced in the horizontal
vertical downward direction or in horizontal outlet. The vertical funnel-shaped draining
direction. For applying horizontal flow, the system and the horizontal outlet are both
cylindrical cell can be placed horizontally and connected to an effluent tank. This effluent
the seepage flow percolates along the specimen tank is equipped with three inlet valves and an
length. But thanks to an inlet and an outlet overflow outlet to control the downstream
ports, a horizontal flow can also be injected hydraulic head. Two inlets permit to collect the
along the specimen diameter. The device effluent for horizontal and vertical flows. The
allows the fine particles detached by the third inlet is connected to the upstream
seepage flow to go out of the cell and to be reservoir in order to saturate the specimen by
captured in an effluent tank. The loading an upward flow. Within the effluent tank, a
system is used to apply axial stress on the rotating system contains eight beakers, each
specimen through a pneumatic piston which with a linen bag for the sampling of eroded
can move up or down. A data acquisition particles carried with the effluent. The
system driven by a dedicated software permits upstream and downstream reservoirs have an
to follow the time evolutions of hydraulic overflow in order to maintain constant the
conductivity, hydraulic gradient, effluent mass, water level.
the gradient of pressure and flow rate At the outlet of the water collection system,
throughout performing the tests. a balance can be used to measure very low

136
flow rate. Otherwise, two flow meters with
complementary ranges permit to measure the
flow rate continually. Under vertical or
horizontal flow, the applied difference of
pressure is measured by the same differential
pressure sensor which is connected with four
pressure ports: at the horizontal inlet and
outlet, the base pedestal and the top cap. The
data acquisition records automatically the time
evolutions of: the hydraulic conductivity, the
hydraulic gradient, the effluent mass, the Figure 2. Grain size distribution of tested soils
gradient of pressure and the flow rate Table 1. Characteristics of tested gradations
throughout performing the tests. and potential of suffusion susceptibility
2.2 Tested soils and potential susceptibility to Soil P (%) Gr (H/F)min KL WF CZ
suffusion B 1.6 2.5 0.035 U / S
Two non-plastic or low plasticity soils were R2 1.2 WG 0.195 U S U
selected with a slight variation of grain size
distributions and different grain shapes to Note: P = percentage of particle smaller than
0.063mm; Gr = dmax/dmin (dmax and dmin: maximal
investigate suffusion susceptibility. A laser
and minimal particle sizes characterizing the gap in
diffraction particle-size analyzer was used to the grading curve); WG: widely graded soil; F and
measure with demineralized water and without H are the mass percentages of the grains with a
deflocculation agent, the grain size distribution size, lower than a given particle diameter d and
of tested soils (see Fig. 2). Both soils were between d and 4d respectively; (H/F)min is the
created by mixing different materials. The gap- minimum value of Kenney and Lau’s ratio; KL:
graded soil B is composed by the mixture of Kenney and Lau’s criterion; WF: Wan and Fell’s
the sand S1 (25%) and the gravel G3 (75%) criterion; CZ: Chang and Zhang’s criterion;
marketed by Sablière Palvadeau (France). The /= method not relevant for considered soil.
well-graded soil R2 is created by the mixture Finally, no clear classification can be drawn for
of 28% sand S1 and 72% gravel from a French soils B and R2 thanks to the three used criteria.
dike DR. Therefore, the erodibility characterization needs
According to the grain size based criterion suffusion tests.
proposed by Kenney and Lau (1985), both
2.3 Test procedure
soils B and R2 are indeed internally unstable
(see Table 1). For the gap-graded soil B, the The experimental methodology can be
percentage P of particle smaller than described in three steps: specimen preparation,
0.063mm is less than 10%, and the gap ratio specimen saturation and suffusion test.
Gr is smaller than 3. Thus the Chang and Tested soils were moistly compacted in the
Zhang’s (2013) method assesses soil B as acrylic cylinder to prevent soil segregation and
internally stable. As the percentage of fine P per layers of 5 cm height in order to limit the
is less than 5% and (H/F)min < 1 for the well- heterogeneity. Each layer was produced by using
graded soil R2, Chang and Zhang’s (2013) a static compaction technique to limit the soil
method classified this soil as internally anisotropy until the fixed density was reached
unstable. According to Wan and Fell (2008), and the final specimen height was 350 mm. Then
the method that they proposed seems not to the cylinder cell is connected to the funnel-
be applicable for gap-graded soils and soils shaped draining system and the loading system at
with a mass of fine fraction lower than 15%. its bottom and top, respectively. All specimens
Then this method is only relevant for soil R2 were subjected to an air pressure of 200 kPa
which is classified as internally stable. through the piston and the pore opening size of

137
the downstream mesh is 1.2 mm in order to flow direction. Table 2 details the initial dry
allow the erosion of all fine particles. unit weight after saturation , the applied
The saturation step begins with the injection hydraulic gradient i, the seepage flow length L
of carbon dioxide (from the bottom) for a and the value of the erosion resistance index I.
duration of ten minutes to improve the
dissolution of gases into water and afterward, the Table 2. Properties of tested specimens
specimen is saturated with tap water in the SoilTested  i L Iα
upward direction by gradually increasing the Specimen (kN/m3) (-) (cm) (-)
level of the upstream tank. The water level was B B-V 17.4 0.05-0.88 35.0 3.8
increased until it reached the top of the specimen.
B-Vrep 0.06-1.30 35.0 3.7
The inflow rate was sufficiently small to prevent
B-H 0.05-0.86 35.0 4.1
soil segregation in the specimen and to achieve a
high degree of saturation. The whole saturation B-Hrep 0.22-3.29 8.1 4.0
phase requires approximately 24h. At the end of R2 R2-V 18.6 0.06-3.43 35.0 4.3
the saturation step, pressure ports are in turn R2-H 0.09-3.25 35.0 4.4
linked with the differential pressure sensor to
measure pressure drop during suffusion test. 3.1 Hydraulic conductivity and erosion rate
for soil B
For the suffusion test step, the upstream
reservoir is fixed and the downstream reservoir For each flow direction, two specimens of soil
is moved down, step by step to apply the multi- B were tested: specimens B-V and B-Vrep
stage hydraulic gradient. Fig. 3 shows the under vertical flow and B-H and B-Hrep under
multi-stage hydraulic gradient applied on horizontal flow. Fig. 4a and Fig. 4b show the
specimens of soil B under vertical flow. time evolution of the hydraulic conductivity
and the erosion rate, respectively.

Figure 3. Multi-stage hydraulic gradient for test of

soil B Figure 4a. Time evolution of hydraulic conductivity
of soil B
A beaker in the rotating sampling system is
selected for the duration of each stage of the
hydraulic gradient. At the end of each test, the
eroded fine particles caught by the linen bags
are dried in an oven for 24 hours and their
mass is determined.

3. TEST RESULTS AND ANALYSIS

Six specimens were tested. The name of each
test starts with the soil letter and the second
letter, V or H clarifies the vertical or horizontal Figure 4b. Time evolution of erosion rate of soil B

138
 It can be noted that under vertical flow, the The hydraulic conductivity under horizontal
initial hydraulic conductivity of test B-V is flow is higher than which under vertical flow
similar to which of test B-Vrep, roughly 2.5 10- (see Fig 5a). The hydraulic conductivity of R2
3
m/s. This result shows that the repeatability of is quite less than that of soil B and tends to
specimen preparation is fairly good. continuously decrease during the suffusion
The hydraulic conductivity of test B-V process under both flow directions. The erosion
increases gradually in early stages then rises rate fluctuates during the testing process but it
steeply in later stages when the hydraulic is always smaller than 9 10-4 kg s-1 m-2 (see
gradient is increased (at t = 0, 30, 60, 90, 120, Fig. 5b). The conjunction of hydraulic
150, 180 and 210 min, see Fig. 4a). The conductivity decreasing and small magnitude
variation of erosion rate of test B-V exhibits a of erosion rate suggests that the predominant
continued increase from 6 10-4 kg.s-1.m-2 to 6 process induced by the seepage flow is the
10-3 kg.s-1.m-2 (see Fig. 4b), but slightly filtration process.
decreases at the final stage. The hydraulic
conductivity of test B-Vrep increases slowly
and fluctuates in the range of 2.4-3.2 10-3 m/s.
The predominant response to these stages
seems to be the filtration process which
produces a general decrease of hydraulic
conductivity induced by the clogging of fines
particles in the skeleton of the coarse particles.
This is clearly shown by the erosion rate during
these stages which is lower than which of B-V.
At the beginning of test B-H, the hydraulic
conductivity is highest, around 5 10-3 m/s, it
slightly increases to 6.3 10-3 m/s and finally
stays constant until the end of test. The
hydraulic conductivity of test B-Hrep steadily Figure 5a. Time evolution of hydraulic
increases from the initial stage about 3.1 10-3 conductivity of soil R2
m/s to the middle stage about 9.1 10-3 m/s,
corresponding to a strong increase of erosion
rate. Finally, the hydraulic conductivity
gradually decreases and reaches 5.8 10-3 m/s.
The comparison of the results under both flow
directions shows that the hydraulic conductivity
during the suffusion process is higher under
horizontal flow than under vertical flow.
For vertical flow, once the erosion rate is
higher than 4 10-3 kg s-1 m-2, then the hydraulic
conductivity begins to sharply rise until the final
stage. Whereas under horizontal flow, the
hydraulic conductivity is higher than 5 10-3 m/s
even if the erosion rate is smaller than 4 10-3 kg Figure 5b. Time evolution of erosion rate of soil R2
s-1 m-2.
3.3 Post-suffusion gradations of both soils
3.2 Hydraulic conductivity and erosion rate With the objective to explain the discrepancy
for soil R2 of the evolution of hydraulic conductivity of
For the soil R2, two specimens were carried specimens under vertical and horizontal flows,
out: R2-V under vertical flow and R2-H under the grain size distribution was measured after
horizontal flow. suffusion tests. Specimen B-V was divided into

139
three parts: top layer, middle layer, and bottom Initial and post-suffusion grain size
layer (top, middle and bottom are distinguished distributions of tests R2-V and R2-H are
according to the elevation). Fig 6a shows that the plotted in Fig. 7a and 7b, respectively. For both
loss of fine particles is the largest in the top layer, tests, specimens were divided into top, middle
then in the middle layer and finally in the bottom and bottom layers. The results show quite no
layer with a fine percentage of 19.2%, 21.5%, variation of grain size distribution under
and 22.2%, respectively. vertical flow and a small loss of fine fraction
For horizontal flow, the specimen B-H was under horizontal flow. These results may
divided into four parts: upstream top layer, explain the small variations of hydraulic
upstream bottom layer, downstream top layer, conductivity and erosion rate. In addition,
and downstream bottom layer. Fig. 6b shows that under horizontal flow, the loss of fine fraction
no distinction can be made between upstream is larger in the top layer than other layers
and downstream parts. However, the percentage leading to a preferential flow in the top layer as
of fine particles decreases in top layer and for soil B. Moreover, the slight variations
increases in bottom layer with a final value of confirm that this soil appears more resistant
17.4% and 27.4%, respectively. In consequence, face to suffusion process than soil B.
under horizontal flow the specimen becomes
more heterogeneous than under vertical flow.
Moreover this result suggests that under
horizontal flow, a preferential flow may appear
in the top layer which can explain the higher
value of the hydraulic conductivity measured
during tests B-H and B-Hrep.

Figure 7a. Grain size distribution of soil R2 under

vertical flow

Figure 6a. Grain size distribution of soil B under

vertical flow

Figure 7b. Grain size distribution of soil R2 under

horizontal flow

3.4 Suffusion susceptibility of both soils

The cumulative loss of dry mass versus the
cumulative expended energy of soil B are
plotted in Fig. 8. At the steady state (i.e. when
Figure 6b. Grain size distribution of soil B under the hydraulic conductivity tends to stabilize
horizontal flow

140
and the erosion rate tends to decrease) the horizontal flow than under vertical flow. For
erosion resistance index (Iα) is computed. soil B (i.e. the most erodible soil), the effect of
Under vertical flow, Iα is equal to 3.7 and 3.8 flow direction on the resistant erosion index is
for tests B-V and B-V rep, respectively. The more pronounced. It should be due to the
corresponding classification is moderately significant influence of gravity when the
erodible. Whereas under horizontal flow, the constrictions are quite large. These results
values of Iα are 4.1 and 4.0 for tests B-H and demonstrate the significant influence of the
B-Hrep, respectively. So the suffusion seepage flow direction.
classification is moderately resistant.
4. CONCLUSION
In literature, most of suffusion tests was
performed under vertical flow. In field
conditions, the seepage flow direction can be
vertical or horizontal. Therefore, a new device
was developed in order to investigate the
suffusion susceptibility under vertical or
horizontal flow and to study the seepage flow
direction influence. The tested soils are a well-
graded soil (named R2) and a gap-graded soil
(B), but a slight variation of grain size
Figure 8. Cumulative loss mass versus cumulative distributions makes them different. No clear
expended energy under both flow classification can be drawn for these soils,
directions, soil B thanks to the three used criteria which are only
based on the gradation.
Six specimens were prepared by static
compaction to limit the soil anisotropy and
suffusion tests were carried out under vertical or
horizontal flow. The comparison of post-
suffusion gradations with the initial ones permits
to note that under horizontal flow, the specimens
become more heterogeneous with a higher loss
of fine particles in the top layer. The erosion
resistance index (Iα) is based on the dry mass of
eroded fines and expended energy. This index I
is computed at the steady state (i.e. when the
Figure 9. Cumulative loss mass versus cumulative hydraulic conductivity tends to stabilize and the
expended energy under both flow
erosion rate tends to decrease). Thanks to the
directions, soil R2
value of I, the suffusion susceptibility
Fig. 9 shows the cumulative loss of dry mass classification can be determined. The results
versus the cumulative expended energy for show that soil B is considered as moderately
tests R2-V and R2-H. At the steady state, Iα is erodible under vertical flow, and moderately
4.3 and 4.4 for tests R2-V and R2-H, resistant under horizontal flow. Soil R2 is
respectively. Thus the corresponding suffusion classified as moderately resistant under both flow
classification is moderately resistant. In directions. In greater detail, both tested soils
consequence, under both flow directions, the appear more resistant under horizontal flow.
soil R2 is more resistant than soil B. Moreover, Thus, these results demonstrate the
for both tested soils, the resistance face to significant influence of the seepage flow
suffusion process appears higher under direction on the suffusion susceptibility. In this

141
context, the new device gives the opportunity Kenney, T. C., & Lau, D. 1985. Internal stability of
to select the direction of the injected flow. granular filters. Canadian Geotechnical
Further studies are required to investigate Journal, 22, 215-225.
the influence of other factors, such as the flow Marot, D., Bendahmane, F., Rosquoet. & Alexis,
direction according to the interface between A. 2009. Internal flow effects on isotropic
layers of compacted soils. confined sand-clay mixtures. Soil & Sediment
Contamination, pp. 294-306.
5. ACKNOWLEDGMENTS Marot, D., Regazzoni, P. L., & Wahl, T. 2011.
The authors thank the company IMSRN, the Energy based method for providing soil
Ministry of Education and Training of surface erodibility rankings. Journal of
Vietnam, the University of Danang, Vietnam, Geotechnical and Geoenvironmental
for providing financial support for this work. Engineering (ASCE), 48, 1772-1787.
Marot, D., Rochim, A., Nguyen, H. H.,
6. REFERENCES Bendahmane, F., & Sibille, L. 2016. Assessing
the susceptibility of gap graded soils to
Bendahmane, F., Marot, D., & Alexis, A. 2008.
internal erosion characterization: proposition
Experimental parametric study of suffusion
of a new experimental methodology. Nat
and backward erosion. Journal of
Geotechnical and Geoenvironmental Hazards, 1-24.
Engineering, ASCE, 134(1), 57-67. Moffat, R., & Fannin, J. 2006. A large
Chang, D. S., & Zhang, L. M. 2011. A stress- permeameter for study of internal stability in
controlled erosion apparatus for studying cohesionless soils. Geotechnical Testing
internal erosion in soils. Geotechnical Testing Journal, 23(1), 116-122.
Journal, 34(6), 579-589. Moffat, R., Fannin, R.J., & Garner, S.J., 2011.
Chang, D.S., & Zhang, L.M. 2013. Extended Spatial and temporal progression of internal
internal stability criteria for soils under erosion in cohesionless soil. Canadian
seepage. Soils and Foundations, 53(4), 569- Geotechnical Journal, 48, 399–412.
583. Reddi, L. N., Lee, I., & Bonala, M. S. 2000.
Fell, R., & Fry, J. J. 2013. Erosion in
Comparision of internal and surface erosion
geomechanics applied to dams and levees.
Bonelli S. Editor. ISTE-Wiley, pp.1-99. using flow pump test on a sandkaolinite
Foster, M., Fell, R., & Spanagle, M. 2000. The mixture. Geotechnical Testing Journal, 23(1),
statistics of embankment dam failures and 116-122.
accidents. Canadien Geotechnical Journal, Rochim, A., Marot, D., Sibille, L., & Le, V.T
(37), 1000-1024. 2017. Effect of hydraulic loading history on
Fry, J. J., Vogel, A., Royet, P., & Courivaud, J.R. suffusion susceptibility of cohesionless soils.
2012. Dam failures by erosion: Lessons from Journal of Geotechnical and
ERINOH data bases. Proc., 6th Int. Conf. on Geoenvironmental Engingneeing (ASCE),
Scour and Erosion, Paris, pp. 273-280. 04017025: 1-10
Garner, S.J., & Fannin, R.J. 2010. Understanding Sail, Y., Marot, D., Sibille, L., & Alexis, A. 2011.
internal erosion: a decade of research
Suffusion tests on cohesionless granular
following a sinkhole event. Int J Hydropower
matter. European Journal of Environmental
Dams 17:93-98.
Ke, L., & Takahashi, A. 2012. Strength reduction and Civil Engineering, 15(5), 799-817.
of cohesionless soil due to internal erosion Sibille, L., Lominé, F., Poullain, P., Sail, Y., &
induced by one -dimensional upward seepage Marot, D. 2015. Internal erosion in granular
flow. Soils and Foundations, 52(4), 698-711. media: direct numerical simulations and
Ke, L., & Takahashi, A. 2014. Experimental energy interpretation. Hydrological Processes,
investigations on suffusion characteristics and 29(9), 2149-2163.
its mechanical consequences on saturated Skempton, A. W., & Brogan, J. M. 1994.
cohesionless soil. Soils and Foundations, Experiments on piping in sandy gravels.
54(4): 713-730. Géotechnique, 44(3), 440-460.

142
Slangen, P. & Fannin, R.J. 2017. The role of
particle type on suffusion and suffusion.
Géotechnique Letters, 7(1), pp. 6-10.
Smith, M., & Konrad, J.M. 2011. Assessing
hydraulic conductivities of a compacted dam
core using geostatistical analysis of
construction control data. Canadian
Geotechnical Journal, 48, 1314–1327.
Wan, C. F., & Fell, R. 2008. Assessing the
potential of internal instability and suffusion
in embankment dams and their foundations.
Journal of Geotechnical and
Geoenvironmental Engineering (ASCE),
134(3), 401-407.

143
The 3rd Int. Conf. on Transport Infrastructure & Sustainable Development (TISDIC 2019). Construction Publishing House. ISBN: 978-604-82-2893-4.

Research of changing physical properties of embankment due to

seepage effect by the time

Nguyen Thanh Quang 1,, Chau Truong Linh 2, Nguyen The Hung 3
1,2,3
University of Science and Technology, Danang of University

Corresponding Author: Nguyen Thanh Quang,
E-mail: [email protected]

ABSTRACT: Currently, empirical method is commonly used to study the influence of seepage on the dam
incidents. For the study of seepage influence, the method of physical modeling is applied by studying each
phenomenon due to the seepage influence on the scale model of the dam. From the observation,
measurement, collection and analysis of data, we can make assessments of the influence of seepage on the
dam stability. In particular, the determination of physical properties of embankment during the operation of
reservoirs is very important and necessary which not only shows the seepage effect on changing physical
properties of embankment but also is the foundation for examining and finishing the mathematical
equations so that the performance is close to the experimental data. From the objective of researching the
seepage effect, this article mentioned the building research equipment of changing physical properties of
embankment due to seepage effect by the time. Through this equipment, combining with other experiments,
we can determine the changes of physical properties of embankment due to seepage effect by the time
including cohesion (C); friction angle (φ); density (γ) and permeability coefficient (Kt).
Keywords: Seepage, dam incidents, earth dam, physical properties, embankment.

accounted for 35.4% and other failure accounted

1. INTRODUCTION for about 19.7%.
In recent years, there have been many dam Currently, empirical method is commonly
failure and incidents in the world causing serious used to study the influence of seepage on the
catastrophes and in Vietnam, these incidents also dam incidents. For the study of seepage
often occur (www.vncold.vn). The effect of dam influence, the method of physical modeling is
failure causes heavy losses in people, property, applied by studying each phenomenon due to the
infrastructure and disruption of economic and seepage influence on the scale model of the dam.
social activities. In Vietnam, as of 2018, the While designing earth dams, it is necessary to
country has invested in building 6,648 reservoirs ensure the stable permeability condition so that
with a total capacity of about 13.5 billion m3, the permeability cannot exceed the permissible
distributed in 45 provinces/ cities of the country, flow and velocity, causing underground erosion
including 702 large reservoirs and 5,946 small and washing away the materials which takes
reservoirs. According to the report “Current dam bad effect on the sustainability and longevity of
safety in Vietnam” of the Ministry of Agriculture the project (Vietnamese standard TCVN 8216,
and Rural Development, by 2018 there were 2009). However, when seepage flow exceeds
1,200/6,648 broken dams and reservoirs the permissible flow and velocity, it will result
(www.vncold.vn). In particular, the rate of dam in the risk of dam failure. Therefore,
failure due to seepage accounted for about permeability is one of the important factor when
44.9%, broken roof protection structure designing earth dams.

144
 Currently, empirical method is commonly The cell accommodates a specimen of
used to study the influence of seepage on the 0.035m-in-diameter and 0.070 m-in-length.
dam incidents. For the study of seepage
influence, the method of physical modeling is
applied by studying each phenomenon due to the
seepage influence on the scale model of the dam.
From the observation, measurement, collection
and analysis of data, we can make assessments of
the influence of seepage on the dam stability.
From the objective of researching the seepage
effect, this article mentioned the building
research equipment of changing physical
properties of embankment due to seepage effect
by the time is shown Fig.1 below: Figure 3. Installing specimen into the model

Figure 1. Research equipment of changing physical

Figure 4. Flow pump and pressure tank and automatic
properties of embankment due to seepage
pump control system
effect by the time
Research equipment of changing physical The increment of inflow rate system consists
properties of embankment due to seepage of a 15-35-atm flow pump (Fig 4) and a
effect by the time is improved based on the pressure tank to put into the specimen through
structure of triaxial testing. The device consists an automatic pump control system. (Fig 4).
of an automatic inflow pump control system, a The adjusted pressure is 0 ~ 30 atm depending
triaxial cell system similar to the structure of on the calculation in the experimental
compression cell and the outflow control condition.
system. The diagram illustrating the structure The seepage flow moves through the
of each cell is shown in Fig 2. specimen by two pores in two metal sample
caps (Fig 5). The flow tubes are designated as
0.002 m-in-diameter. During the experiment,
the seepage pressure is always in accordance
with the calculated value in the experimental
investigations.

Figure 5. Two metal caps, porous stone/disc

Figure 2. Cell structure in research equipment of
To ensure that no local erosion occurs when
changing physical properties of
embankment due to seepage effect by the seepage flows into the specimen, the test
the time specimen is installed with more porous stone

145
above the specimen. Porous stone is 0.035m- water filter system) goes into the water cell,
in-diameter corresponding to upper cap one ensuring that the system is closed.
and 0.005m-in-thickness (Fig 5). The seepage The determination of physical properties of
outflow control system is installed similarly to embankment during the reservoir operation is
the inflow. The flow tubes are designated as very important and necessary which not only
0.002 m-in-diameter. shows the seepage effect on changing physical
The research equipment of changing properties of embankment but also is the
physical properties of embankment due to foundation for examining and finishing the
seepage effect by the time can be controlled in mathematical equations so that the
either mechanical or automatic modes through performance is close to the experimental data.
control cabinet installed beside the cell (Fig 6). Through this equipment, combining with
In order to ensure safety during equipment other experiments, we can determine the
operation, the control cabinet installed with changes of physical properties of embankment
thermal relay which can automatically interrupt due to seepage effect by the time including
and stop the operation of the device when there cohesion (C); friction angle (φ); density (γ) and
are any problems. permeability coefficient (Kt).

2. EXPERIMENTAL MATERIALS
2.1 Materials
We can make some general comments that the
embankment is mainly cohesive soil with a
wide range of cohesive. Based on the factors
including: physical properties, classification of
Figure 6. Control cabinet
embankment types and experiment time, the
To determine the cohesive soils after the soil type selected for research is aluvi (aQ). In
seepage effect by the time, the specimen is natural condition, the soil has γc = 1.4÷1.6
installed into Geocomp triaxial testing T/m3, W = 20÷25%. When saturated, there are
equipment (Fig. 7). parameters such as φ = 16°÷20°, C = 0.10÷0.4
To ensure that it does not affect the kG/cm2, K = 10-4÷10-5 cm/s. This soil type has
specimen, a triaxial cell cap is designated a clay content of 15÷35% and can be used to
similar to Geocomp triaxial cell cap. build dams or dam cores. Similar to this soil
type, the topic mentioned the main dam of Ta
Trach reservoir, Thua Thien Hue province.
The soil is exploited in the 5th expansion zone
of the VD2 (Drawing Ref. N°123D-DC-VL-
08÷N°123D-DC-VL-10) in the downstream of
the VD2 – Ta Trach reservoir in Thua Thien
Hue Province (Geological report of Ta Trach
reservoir, 2010).

Figure 7. Geocomp triaxial testing equipment in

the lab of Danang University of Science
& Technology

To ensure that no air bubbles appear during

the equipment operation, in each cell and
pressure tank, we have made air degassing Figure 8. The location of taking specimens (the 5th
valve. During operation, all the water supply to expansion zone of the VD2) - Ta Trach
the pump and the water output (through the reservoir, Thua Thien Hue Province

146
2.3 Test materials 2.2.3 Determination of shear strength
2.3.1 Determination of grain composition The experiment to determine the shear strength
of experimental materials was performed on
The grain composition of the soil is a triaxial testing equipment under (Vietnamese
percentage (%) under the weight of the standard TCVN 8868, 2011).
particle-size groups existing in the soil. Test
results are determined under (Vietnamese
standard TCVN 4198, 2014) and (ASTM D422
– 63, 2007).

Figure 11. Install triaxial cell into the equipment

The test results of shear strength of materials

under CU and CD diagrams:

Figure 9. Result of grain composition of research

materials

2.2.2 Determination of maximum dry density

and optimum moisture content
In order to determine the soil weight in the
specimen, it is necessary to determine two
parameters including maximum dry density
and optimum moisture content by standard Figure 12. Triaxial compression test results under
CU diagram
compaction test under (Vietnamese standard
TCVN 4201, 2012).

Figure 10. Standard laboratory compaction tests Figure 13. Triaxial compression test results under
for materials CD diagram

147
2.3 Test procedures This time of saturation process is shown as
Figure 15, 16, and 17.
2.3.1 Test Specimen preparation Back Pore pwp
Cell Value Time
pressure pressure difference
The test specimens are compacted with a dry pressure
ub u u
B t
unit weight relevant to 97% of the largest dry (kG/cm2) (-) (h)
(kG/cm2) (kG/cm2) (kG/cm2)
unit weight and corresponding moisture after 0 0 0 0
being compacted. 0.5 - 0.16 0.16 32.0 18
The specimen is compacted in 3 soil layers 0.5 0.4 0.34 36
with 23mm-in-thickness for each one to ensure 1.0 - 0.70 0.36 72.0 54
consistency in tightness in the whole specimen 1.0 0.9 0.75 72
(Fig. 14). The height and diameter are 70mm 1.5 - 1.23 0.48 96.0 90
and 35mm respectively. The compaction 1.5 1.4 1.28 93
weight is 0.715 (kG), the weight of the 1.9 - 1.68 96
compaction + conductor is 1.303 (kG), the
height of the fallen compaction is 30cm.

Table 1. Soil volume put into the specimen

Maximum dry volume 1.68 (g/cm3)
Maximum moisture 16.23 (%)
Tightness of the specimen 0.97
Knife volume 72.066 (cm3)
Soil volume put into the
136.50 (g)
specimen
Figure 15. The saturation process of specimen 1

Back Pore pwp

Cell Value Time
pressure pressure difference
pressure B t
ub u u
(kG/cm2) (-) (h)
(kG/cm2) (kG/cm2) (kG/cm2)
0 0 0 0
0.5 - 0.19 0.19 38.0 18
0.5 0.4 0.33 36
Figure 14. Soil sample after being designated 1.0 - 0.71 0.38 76.0 54
1.0 0.9 0.77 72
2.3.2 Specimen saturation process
1.5 - 1.25 0.48 96.0 90
The purpose of specimen saturation is to make 1.5 1.4 1.31 93
all pores filled with water. To do that, we often 2.4 - 2.21 96
raise the pore water pressure in the specimen to
the level of filling the entire air in the pores. To
implement this process, the equipment uses
water pressure (seepage pressure) to influence on
the specimen, at the same time increasing the cell
pressure to retain a small positive effective stress.
All specimens used in this experiment were
first fully saturated, creating the consistency of
the initial moisture or saturation. Soil specimens
are considered to be fully saturated when pore
water pressure coefficient B reaches a value
greater than or equal to 0.95 (Head, 1986). Figure 16. The saturation process of specimen 2

148
 Back Pore pwp
Value Time
pores which decreases the pressure. The trend
Cell pressure
pressure
pressure difference
B t of reducing B is similar to the research
ub u u
(kG/cm2)
(kG/cm2)
(-) (h) results of Evans and Fang (1988). However,
(kG/cm2) (kG/cm2)
during specimen saturation, B is always over
0 0 0 0
0.95 and the specimen is considered saturated
0.5 - 0.18 0.18 36.0 18
during the experiment.
0.5 0.4 0.27 36
1.0 - 0.69 0.42 84.0 54 2.4 The process of changing seepage flow by
1.0 0.9 0.77 72 the time
1.5 - 1.26 0.49 98.0 90
1,5 1.4 1.32 93 After the saturation finishes, the specimen are
3.4 - 3.22 96 consolidated under cell pressure and seepage
pressure, in other words, the soil specimens are
isotropically consolidated by effective
pressure. The value of effective pressure is
calculated according to the effective stress
principle applied to the main dam of Ta Trach
reservoir, Thua Thien Hue Province.

Figure 19. Cross-section of main dam, Ta Trach

Figure 17. The saturation process of specimen 3 reservoir, Thua Thien Hue Province

Based on the period and timing of the a) We can determine cell pressure in the
saturation process on a specimen set, it is formula:
 z   w .z
possible to show the relationship between
seepage pressure, chamber pressure by the time In which:
(applicable to all specimens during the - Natural density:
experiment ) in two following stages:  w  0,97.1,68(1  0,1623)  1,89 (kG / cm2 )
- z = 17m
The cell pressures determined for each cell:
0.5 σz; 1.0 σz; 2.0 σz with the actual condition
are 1.6 atm - 3.2 atm - 6.4 atm respectively.
b) We can determine seepage pressure in
the formula:
 t   nc .z
The seepage pressures determined for each
cell: 0.5 σt; 1.0 σt; 2.0 σt with the actual
condition are 0.85 atm - 1.7 atm - 3.4 atm
respectively.
Figure 18. The saturation process of specimens The seepage pressure for each specimen is
calculated corresponding to the required
The saturation degree tends to decrease experimental result. Corresponding to each
slightly during the first time of stage 2. This requirement, the performance of changing the
change occurs because the bubbles are made seepage pressure to each specimen is shown in
in the specimen when the water displaces the the following figure:

149
  Cohensive specimen considering the seepage
effect according to actual condition (case 1)

Figure 20. The saturation process and seepage

effect of specimen 1 (case 1) Figure 24. The saturation process and seepage
effect of specimen 2 (case 2)

Figure 21. The saturation process and seepage

effect of specimen 2 (case 1)

Figure 25. The saturation process and seepage

effect of specimen 3 (case 2)

The time to determine the changes of physical

properties at T = 5, 15, 30, 45, 60, 90 days

3. TEST RESULTS
Figure 22. The saturation process and seepage
effect of specimen 3 (case 1) 3.1 Density changes by the time
Through the specimen test on the research
 Cohensive specimen considering the
equipment of changing physical properties of
seepage effect under the increase seepage
embankment due to the seepage effect by the
effect (case 2)
time corresponding to both actual condition
and the condition of increasing seepage
pressure, results of density changing by the
time is shown in the following table 2:

Table 2. The result to determine the density

changing by the time on the specimen (Unit: kG/m3)
Time
0 5 15 30 45 60 90
(day)

Figure 23. The saturation process and seepage Case 1 18.90 18.80 18.80 - - -
18.90
effect of specimen 1 (case 2) Case 2 18.70 18.20 17.50 17.50 17.40 17.40

150
 The result of changing embankment density the density change of embankment. In
is shown on Fig. 26. general, during the experiment, the specimen
tends to compress in the accumulated
erosion process of fine grains. This shows
that when fine-grained erosion occurs, at the
same time the rearrangement of soil particles
in the specimen occurs. This experimental
result is similar to that of Moffat et al.
(2011) described the volume deformation
process corresponding to the suffusion. This
process also results in the distribution of
particle size in the specimen as researched
Figure 26. The chart shows the results of the by some authors such as Kenney and Lau
density change by the time (1985), Chang and Zhang (2011). However,
due to limitations in experimental condition,
Some comments for the test results: the change of grain composition curve of the
- The soil density changes under the first post-experimental specimen was not
regression line relationship with time t mentioned.
corresponding to actual condition:
3.2 Shear strength changes by the time
  0.0038t  18.898 with R2 = 0.7619
- The soil density changes under the first Through the specimen test on the research
regression relationship with time t in two equipment of changing physical properties
stages corresponding to the condition of of embankment due to seepage effect by the
increasing seepage pressure. time corresponding to both actual condition
Stage 1: and condition of increasing seepage
pressure, the result of shear strength
  0.0471t  18.914 with R2 = 0.9994
changing by the time is shown in following
Stage 2: Table 3, Table 4:
  0.0019t  17.557 with R2 = 0.7143
Table 3. The result to determine the cohesion
The change of embankment density can be
(C) changing by the time on the specimen
explained as follows:
(Unit: kG/cm2)
The change in soil density is proportional to
the amount of eroded soil accumulating by the Time
0 5 15 30 45 60 90
time. Corresponding to the condition of actual (day)

seepage pressure, the effective time of seepage Case

21.75 21.30 20.69 - - -
1
flow is not long enough resulting in a small 22.16
amount of accumulated eroded soil and few Case
21.50 20.00 19.50 19.40 19.15 19.10
2
change in density. However, corresponding to
the condition of increasing seepage pressure,
the time is longer, the seepage effect on the Table 4. The result to determine the friction
specimen is large which causes an increase in angle changing by the time on the specimen
soil erosion. At a later time (t = 60 days), the (Unit: degree)
amount of soil erosion decreases and almost Time (day) 0 5 15 30 45 60 90
disappears, the soil density is almost constant
during this stage. Case 1 14,975 14,580 14,177 - - -
15,210
It can be seen that the continuous fine- Case 2 14,830 14,320 13,900 13,780 13,710 13,650
grained erosion on the specimens can lead to
the position rearrangement of the particles The changing result of cohesion C of
which results in the volume deformation and embankment is shown on Figure 27.

151
 seepage pressure, the change of permeability
coefficient by the time is shown in Table 5.

Table 5. The result to determine the

permeability coefficient changes by the time on
the equipment (Unit: cm/s)
Time (day) 0 5 15 30 45 60 90
Case 1 -6
1.10-6 3.5.10-6 5.10-6 - - -
1.10
Case 2 1.5.10-6 4.10-6 4.5.10-6 5.10-6 5.5.10-6 6.10-6

Figure 27. The chart shows the results of the Comment on test results: By the time t, the
cohesion change by the time permeability coefficient has increased. The
increase of permeability coefficient results
The soil cohesion changes under the tertiary
from the specimen consolidation process by the
regression relation with time t:
C  1E  07t 3  0.0002t 2  0.0493t  21.989 ,
time t.
R2=0,9901
4. DISCUSSIONS
The changing result of embankment density
is shown on Figure 28. Based on the graph of parameters C, φ, γ, Kt by
the time, we can make a conclusion that by the
time of reservoir exploitation due to seepage
effect, the physical properties of embankment
has changed. To make comment on this
change, I make some following points:
- During the period of dam construction and
operation in the first period, due to seepage
effect, the moisture of tembankment has
changed. This affects the physical properties of
embankment. However, this change is still
Figure 28. The chart shows the results of the quite slight, because the time of seepage effect
friction angle change by the time is not long.
- During the reservoir operation in
The friction angle of soil changes under the accordance with the operation procedure, due
tertiary regression relation with time t: to seepage effect, the physical properties of
  3E  06t 3  0.0006t 2  0.0497t  15.209 , embankment continue to change and tend to
R2=0.9321 change less after a long time (from 30 years or
Comment on test results: By the time t, the more). Based on the factors affecting the dam
shear strength of soil due to seepage effect stability due to seepage effect, the main reason
slowly reduces. The decrease of shear strength of this change may be the underground erosion.
shows that the soil permeability coefficient is This change can be explained in two stages:
proportional to the accumulated amount of - Stage 1: When the seepage takes effect,
eroded fine grains by the time. deformation deformation has not yet occured
because the fine grains are eroded, the total
3.2 The permeability coefficient changes by volume of specimen remains unchanged and
the time the volume of eroded fine grains will be
Through the experiment for the specimen on replaced by water in the same volume
the research equipment of changing physical (saturated specimen).
properties of embankment due to seepage - Stage 2: The erosion washed away a
effect by the time corresponding to the actual large amount of fine grains, forming a new
condition and the condition of increasing structure. This will result in the distribution

152
of the soil specimen to another stable form. The permissible safety factor in this case is
At this stage, the void coefficient of the [K] = 1.15 (Vietnamese standard TCVN 8216,
specimen also changes. 2009) (Table 7).
c) Case 3: Check and calculate the dam
5. CALCULATION APPLICATION stability and permeability in the above two
cases by the time of reservoir exploitation
The construction project of Ta Trach reservoir (calculated for 100 years).
is built in Duong Hoa Commune, Huong Thuy Parameters of input embankment material is
District, Thua Thien Hue Province, 18 km taken from regression line relation by the time
away from the southwestern of Hue City. This in Section 3.3, including:
project was built on Ta Trach River (a main - Density:
branch of the upstream of Huong River).    0.0038t  18.898 , R2 = 0,7619
5.1 Some general parameters (Technical - Cohesion C:
design report of Ta Trach reservoir, 2007): C  1E  07t 3  0.0002t 2  0.0493t  21.989 , R2= 0,9901

- Elevation of breakwater top: ▼+56.0m - Friction angel φ:

- Length of dam top: 1187m   3E  06t 3  0.0006t 2  0.0497t  15.209 , R2 = 0,9321
- Width of dam surface: 10m + The permeability coefficient taken under
- Form and structure: Multi-block mixed the design:
rock dam (including waterproofing block in the Kt = 1.10-5
middle, imbed blocks at upstream and 5.2 Case 1: Inspection of downstream stability
downstream of the dam)
According to (Vietnamese standard TCVN
8216, 2009) (Table 8) on the calculation of dam
stability, the article content focuses on the
calculation of dam stability in the following cases:
a) Case 1: After the water level is full and
stability seepage - testing downstream (basic
combination). With condition in the upstream is
retention level, the downstream is the average
water level of the supplying period. That is:
- Water level in upstream: ▼+45m
(retention level)
- Water level in downstream: ▼+2.5m Figure 29. The position of sliding surface of dam
The permissible safety factor in this case is body after the seepage is stable
[K] = 1.35 (Vietnamese standard TCVN 8216,
2009) (Table 7).
b) Case 2: After withdrawing water -
checking the upstream slope stability (basic
combination). With condition in the upstream
is (retention level) drawing to the water level
that is safe for the dam in case of risk
incidents, and downstream is water level
corresponding to the max Qrelease when Figure 30. The safety factor after the seepage is
removing water from the lake. stable (K = 1.498)
- Upstream water level is ▼+45m (retention level)
drawing down to the water level to ensure the According to the calculation in the case, when
safety of the dam in case of risk incidents is 25m the water is full and the seepage is stable, the
(the water level before flood). safety factor result is K = 1.498 > [K] = 1.35.
- Downstream water level: ▼+13.1m Therefore, the downstream is stable in the case.

153
 Comment: From the calculation results 5.4.1 Checking downstream stability considering
based on the basic calculation above, it shows the change of physical properties due to seepage
that the downstream ensures stability in effect
permeability and slip.
5.3 Case 2: Inspection of upstream stability

Figure 33. The position of sliding surface of dam

body in case 1 considering the change of
Figure 31. The position of dam body after the physical properties
water level is lower from 45m to 25m

Figure 32. The safety factor after after the water Figure 34. The safety factor after in case 1
level is lower from 45m to 25m considering the change of physical
(K = 1.279) properties (K = 1.328)

According to the calculation in the case, when In this case, the safety factor is K = 1.328 < [K] =
the water is full and the seepage is stable, the 1.35. Therefore, the downstream is unstable at time
T = 100 years considering the change of physical
safety factor result is K = 1.279 > [K] = 1.15.
properties due to seepage effect by the time.
Therefore, the upstream is stable in the case.
Comment: From the calculation results 5.4.2 Checking upstream stability considering
based on the basic calculation above, it shows the entire embankment
that the upstream ensures stability in
permeability and slip.
5.4 Calculation results of case 3 when
considering the change of physical properties
due to seepage effect
When checking and calculating the dam stability
and permeability in case 1 and 2 by the time of
reservoir exploitation (for 100 years), I use the
input parameters C, φ, γ and Kt based on the
regression line relation of these parameters with
time (Fig. 26, Fig. 27, Fig. 28). Parameters were
Figure 35. The position of sliding surface of dam
taken at time t = 100 years and considered for the
body in case 2 considering the change
entire embankment. of physical properties

154
 K = 1.328 and upstream is K = 1.124. According
to the permissible safety factor, the downstream
and upstream are unstable at this time.
Calculation results show that the use of
stability results of PLAXIS software and the
chart to determine the change of physical
properties of embankment due to seepage
effect by the time are the real bases to make
Figure 36. The stability factor of dam body in
calculation for early warning of the dam
case 2 considering the change of
instability in Vietnam.
physical properties (K = 1.124)
7. REFERENCES
In this case, the safety factor is K = 1.124 <
ASTM D422 - 63, Standard Test Method for
[K] = 1.15. Therefore, the upstream is unstable
Particle-Size Analysis of Soils.
at time T = 100 years considering the change
Chang, D.S., Zhang, L.M., A stress-controlled
of physical properties due to seepage effect by
erosion apparatus for studying internal erosion
the time.
in soils, Geotech. Test. J, vol. 34, no. 6, pp.
579-589, 2011.
6. CONCLUSION Evans, J.C., Fang, H.Y, Triaxial permeability and
From the statistical results of some actual strength testing of contaminated soils,
construction works in Vietnam and other Advanced Triaxial Testing of Soils and Rock,
published documents, based on the scope of ASTM STP 977. American, Society for
adhesion change (one of the main factors Testing and Materials, Philadelphia, , vol. In:
affecting the change of physical properties of Donaghe, R.T., Chaney, R.C., Silver, M.L.
embankment due to seepage effect), I have (Ed.), pp. 387-404, 1988.
chosen ancient alluvium corresponding to Geological report of Ta Trach reservoir – Thua
embankment of the main dam of Ta Trach Thien Hue Province, Shop Drawing, 2010.
reservoir as research soil. Head, K.H., Manual of Soil Laboratory Testing,
From the test results on the research John Wiley and Sons, 1986.
equipment, I found that during the experiment, Kenney, T.C., Lau, D., Internal stability of granular
the parameters of physical properties of filters, Can. Geotech. J, vol. 22, no. 2, p. 215–
embankment include density, shear strength and 225, 1985.
permeability coefficient also changed. The Moffat, R., Fannin, R.J., Garner, S.J., Spatial and
result of this change is shown through many temporal progression of internal erosion in
different regression lines from figure 26 to 28 cohesionless soil, Can. Geotech. J, vol. 48, no.
by me. The regression line relation by the time 3, p. 399–412, 2011
can be used to determine the physical properties Technical design report of Ta Trach reservoir, 2007.
of embankment corresponding to actual Vietnamese standard TCVN 4198:2014,
operation time of the dam. Construction soil - Method of particle
Based on the calculation results on the composition analysis in the laboratory.
numerical model by Plaxis software shows: Vietnamese standard TCVN 4201:2012,
- Stability result in design stage of Construction soil - Method of determining the
downstream roof is K = 1.498 > [K] = 1.35 standard tightness in the laboratory.
and upstream is K = 1.279 > [K] = 1.15. Vietnamese standard TCVN 8216:2009, Design of
Therefore, according to the design document, compacted earth dams.
the dam ensures the stability according to Vietnamese standard TCVN 8868:2011, test to
Vietnamese standard TCVN 8216, 2009. determine incohesive shear strength – no
- With the project level, the service life is 100 water drainage and cohesive – drainage of
years. Based on the test results of changing cohesive soil on the triaxial testing equipment.
physical properties of embankment due to Vietnam National Committee on Large Dams and
seepage effect, corresponding to the time of 100 Water Resources Development, [Online],
years, the safety factor of downstream is www.vncold.vn

155
The 3rd Int. Conf. on Transport Infrastructure & Sustainable Development (TISDIC 2019). Construction Publishing House. ISBN: 978-604-82-2893-4.

Excavation behavior and the response of adjacent building

at various stiffness using Plaxis 2D and 3D

Huynh Quoc Thien 1, Tran Viet Thai 2, Nguyen Minh Tam 3, Lai Van Qui 4,
1,3,4
Faculty of Civil Engineering, Ho Chi Minh City University of Technology, Ho Chi Minh City, Viet Nam
2
Hoa Binh Construction Group Joint Stock Company, Ho Chi Minh City, Viet Nam

Corresponding author: Van Qui Lai*

Address correspondence to First author: [email protected]

ABSTRACT: The behavior of deep excavation and response of adjacent building for a deep excavation
case in urban areas are presented. The commercial finite element software, namely Plaxis 2D and 3D, are
used to perform the horizontal displacement of diaphragm wall, the settlement of adjacent building and
the damage level of adjacent building during excavation sequences. The damage criteria based on angular
distortion β and lateral strain εL of adjacent building are employed to assess the building’s damage
potential. The comparison among the results of modeling and field measurements are implemented. The
good agreement among those results is recorded. Besides, the damage levels of adjacent building are also
investigated by varying the stiffness of the building structure. Some remarkable notations for designing
deep excavation in the urban areas to ensure safety throughout the construction process are also
recommended.
Keywords: Excavation behavior, the response of adjacent building, damage potential, angular distortion β,
lateral strain εL.

seriously, various neighboring buildings on

1. INTRODUCTION shallow foundations were largely damaged.
The urbanization rate has been increasing in There had several main reasons for bad
recent years, the construction of deep designing of deep excavation. First of all, soil
basements in the narrow areas with adjacent investigations are not fully conducted.
buildings is inevitable. The design and Almost projects, neither Consolidated
construction of deep excavations in urban Drained (CD) triaxial test nor Oedometer
areas have been become challenges that (OED) test are carried out. This leads to some
requires ensuring the stability of the noticeable mistakes in excavation behavior
excavation as well as minimizing impact on analysis methods. The result is a number of
the adjacent buildings. If the excavation was publicized cases of failures because of the
not well analyzed in advance, there would
lack of soil information. Secondly, practical
have some bad affects to surrounding areas or
engineers lack basic knowledge and
adjacent structures. In particular, a special
concern is structure’s adjacent buildings on necessary experience in assessing the
shallow foundations or melaleuca piles, response of adjacent building due to deep
which is highly sensitive to movements excavation. Specially, in case of low-rise
caused by ground deformation. Due to this framed buildings supported by spread
reason, the visible cracks on adjacent footings are easily damaged. In specifically,
building would be occurred. In fact, the horizontal displacement of diaphragm

156
wall is the main reason affecting potential 2. DAMAGED CRITERIA FOR ESTIMATING
damage of adjacent buildings through BUILDING DAMAGE POTENTIAL
causing angular distortion β and lateral strain
A number of evaluation criteria have been
εL of those ones. A practical example about proposed for estimating the potential damage
damaging adjacent building due to deep of adjacent buildings or performing the damage
excavation in soft soil of Ho Chi Minh City levels of adjacent buildings next to a deep
was shown Figure 1. excavation. However, almost that criteria use
three parameters including the angular
distortion β, lateral extension strain εL of
adjacent building and a principal strain εp
formed by both the angular distortion and the
lateral strain. In that, the angular distortion β is
a forced deformation caused by vertical ground
displacement, which is determined by average
settlement slope minus the tilt of a structural
bay. It damages neighbouring buildings with
diagonal cracks. The lateral extension strain εL
Figure 1. Damaging adjacent building due to deep is deformation due to lateral ground
excavation displacement, which is determined by the
extension of the base divided by the base
Therefore, this study focuses on the length. It causes vertical cracks on adjacent
horizontal displacement of diaphragm wall and buildings. And the principal strain εp is
the responses of low-rise framed building assumed as the maximum strain on the
supported by spread footings. A case study in building structure. The details of this
Ho Chi Minh city is presented. The Plaxis 2D determination can be seen in Figure 2. In usual,
and 3D model using Hardening Soil (HS) the maximum principal strain εp is compared
model were employed for predicting with the critical strains for each different
diaphragm wall deflection and adjacent damage category to estimate potential damage
building movement. As the reference, some levels. In special cases, for buildings with short
empirical formulas were proposed to determine length are mainly damaged by angular
soil parameters for the Plaxis model. The distortion, while lateral strain is negligible
comparison between predicted wall influence on those ones.
displacement and measured field data was
conducted. The good agreements among the
(b)
deformation results were recorded. Besides, to (a)
assess the damage potential of adjacent
building, some criterion values were employed.
The comparison among calculated values,
measured values and criterion values were (c)
implemented. Furthermore, the effects of the
stiffness of building structure on the potential
damage of the building were also investigated.
In final, several recommendations for
designing deep excavation in the urban areas
were presented. Note: θ = direction of crack formation and the
angle of the plane on which εp acts, measured from
The paper includes main four parts:
vertical plane.
Introduction, Damaged criteria for estimating
building damage potential, A case study in Ho Figure 2. Define angular distortion, lateral strain
Chi Minh and conclusions. and maximum principal strain.

157
 Burland et al (1977) firstly proposed the lack of serviceability can have commercial and
damage levels for adjacent buildings by economic impacts on the building and the
collecting, analyzing observed damage from activities it hosts.
measured field data, the observed and  Structural damages are related to
calculated crack width, as shown in Table 1. In cracking or excessive deformations of the
that, they defined six levels of damage, bearing structures and can lead to the partial
numbered 0 to 5 with increasing severity, by or total collapse of the building. Structural
basing on the criteria of the degree of severity, damages can sometimes remain partially
the width of crack, limiting tensile strain (or the hidden beneath the finishes. However,
maximum principal strain εp). For practical whitewash and plaster are good indicators of
example, with measured width of cracks less the cracking propagation.
than 5mm or calculated result of εp in the range
of 0.05 – 0.075%, the damage of adjacent Crack width
Negligible

Lateral strain εL [x10-3]

building is in level two. Besides that, Burland < 0.1mm

Aesthetic
Very Slight
et al (1977) were also proposed simple chart, < 1mm
Slight
as shown in Figure 2c, to assess building 1-5mm

Structural Functional
damage potential. Moderate
5-15mm
In the next generations, Boscardin and Severe
Cording (1989) developed the work of 15-25mm
Very Severe
Angular distortion β [x10-3]
Burland et al (1977) by using the suggestion >25mm

of MacDonald (1956) and Bjerrum (1963).

Figure 3. Damage criterion based on state of strain
They proposed an updated practical chart for
at a point
estimating building damage potential, as
shown in Figure 3. In that, there had three In the recently study, a new term, called
categories of damage classified by visible damage potential index DPI first introduced
damage repairs and six degree of severity by Son and Cording (2005) was also
classified by crack width. The three applicable for assessing the damage levels of
categories of damage can be explained as adjacent building. The DPI is a modification
below. of the maximum principal tensile strain εp
 Aesthetic damages are related to slight
and can be calculated by Equation 1. And the
cracking of the structures, affecting mainly are
levels of building damage can be assessed
easy to repair and generally, redecoration is
throughout Table 2.
sufficient to cover the light cracks.
DPI   p / 1 / 200   100 (1)
 Functional damages are related to the loss
of functionality or serviceability of parts of the One advantage of the DPI concept is that it
building (e.g. doors and windows may be stuck provides a relative measure of the maximum
and pipelines can be damaged) or of sensitive principal strain. The levels of building
devices located inside the building (such as damage could be assessed in terms of a
precision instruments that are sensitive to relative DPI values. The use of the DPI index
differential movements); the structural integrity allows to implement damage evaluation
of the building is not affected, however, the criteria easily.

158
Table 1. Damage classification established by Burland et al. (1977)
Crack Limiting
Category Degree of
Description of typical damage width tensile
of damage severity
[mm] strain (%)
0 aesthetic Negligible Hairline cracks <0.1 0–0.05
1 aesthetic Very slight Fine cracks which are easily treated using <1.0 0.05–0.075
normal decoration. Damage generally
restricted to internal wall finishes. Close
inspection may reveal some cracks in
external brickwork or masonry.

2 aesthetic Slight Cracks easily filled. Redecoration <5.0 0.075–0.15

probably required. Recurrent cracks can be
masked by suitable linings. Cracks can be
visible externally and some repointing may
be required to ensure weather - tightness.
Doors and windows may stick slightly.
3 aesthetic/ Moderate The cracks require some opening up and can 5–15 0.15– 0.3
functional be patched by a mason. Repointing of (many
external brickwork and possibly a small cracks with
amount of brickwork to be replaced. Doors width >3
and windows sticking. Service pipes may mm)
fracture. weather - tightness often impaired.
4 functional/ Severe Extensive repair work involving breaking- 15–25 (but >0.3
serviceability out and replacing sections of walls, depend on
especially over doors and windows. the number
Windows and door frames distorted, floor of cracks)
sloping noticeably. Walls leaning or bulging
noticeably, some loss of bearing in beams.
Service pipes disrupted.
5 structural Very severe Structural damage which requires a major >25 (but
repair job involving partial or complete depend
rebuilding. Beams lose bearing, walls lean on the
badly and require shoring. Windows broken number of
with distortion. Danger of instability cracks)

Table 2. Levels of building damage according

to DPI 3. A CASE STUDY IN HO CHI MINH CITY
Level of building damage DPI (damage 3.1 Project description
caused by excavation potential index)
0 Negligible to very slight 0-15 The project, namely Madison, was located in
15 Thi Sach street, District 1, Ho Chi Minh
1 Very slight to Slight 15-25
City. This project consists of 17 stories and 3
2 Slight to moderate 25-35 basement levels located on an area of
3 Moderate 35-60 2360m2. The deep excavation was carried out
according to the Semi Top-down method and
4 Severe 60-85
was retained by a diaphragm wall having
5 Very Severe >85 800mm thick and 37m in depth.

159
 consisting of 5 phases and the levels of
excavation. Besides, the Table 3 was also
prepared to give more information of
construction sequences. To observation the
behavior of diaphragm wall during construction
sequences, the inclinometer equipment was
installed as Figure 6.
The adjacent building located next to deep
Figure 4. Project location in District 1, HCMC excavation was low-rise building and founded
on shallow foundations. The framed structures
Figure 5 below shows the cross section, of the building were carefully surveyed before
geological condition, the construction sequences excavation.
Surcharge load
20 kN/m 2
Construction
0m L1 Slab +0.40mGL sequences

Filling soil 175mm B30

-2.0m 15m
1. Very Soft -3.2m B1 Slab -3.20mGL
silty clay
175mm B30 -3.80mGL
-4.5m (SPT=0)
2. Stiff clay
-5.6m (SPT=6)
B2 Slab -6.50mGL
200mm B30 -7.30mGL

H400 -10.30mGL

B3 Slab -10.80mGL
600mm B30 -11.80mGL

3. Medium
dense sand -15.50mGL
(SPT=15)
12m
-19.70mGL

Note: ground level (GL)

D-Wall -26.00mGL
400 B35

-34.5m

4. Very stiff D-Wall -37.00mGL

clean clay 800 B35
(SPT=35)
-50.0m
5. Very
dense sand
-90.0m (SPT=37)

Figure 5. Construction section of Madison project

160
Table 3. Construction sequences of Madison
project

Finishing
Construction sequences
date

1st excavation to -3.8mGL, 31/03/2017

1 and install B1 slab and L1
slab. (Cycle 24)
Figure 7. The structural plan of adjacent buildings
2nd excavation to -7.3mGL, 26/04/2017
2
and install B2 slab. (Cycle 38) Lateral displacement (mm)
0 20 40 60
3rd excavation to -11.8mGL, 25/05/2017
0

3 install H400 steel struts at -

10.3mGL (Cycle 57) -5

4th excavation to the bottom 27/06/2017 -10

4 levels of foundation (-
15.55mGL) (Cycle 82)
-15

Depth z(m)
-20

-25

-30
31/03/2017_ID02

26/04/2017_ID02
-35
25/05/2017_ID02

-40 27/06/2017_ID02

Figure 6. Arrangement of inclinometer points

The plan of adjacent building was shown

in Fig 7. To survey the damage levels of
adjacent building due to the deep excavation,
some settlement points in adjacent buildings,
Figure 8. Monitoring data of lateral wall displacement
were also installed and monitored during and settlement of adjacent building at
construction sequences. Besides, cracks on various stages
the structures of adjacent building were also
observed carefully and continuously, as 3.2 Analyze excavation behavior and response
shown in Fig 17. The observation data of of the adjacent building
horizontal displacement of diaphragm wall
and the settlement of adjacent buildings were Back analysis by 2 and 3 - dimensional finite
showed in Fig 8. element (FE) were chosen for analyzing the

161
excavation behavior with the aims to properties is comparatively difficult, so the
determine the suitable stiffness parameters of SPT (N) value, is commonly used for
soil model and compare the response of engineering practice. Shear strength
parameters are determined from direct shear
adjacent buildings by Plaxis 2D model with 3D
test or according to the correlation with the N
model. Hardening Soil model (HS) in Plaxis
index. By author's studies from back analysis
program was used to simulate the behaviour of of horizontal displacement of diaphragm wall,
the soil in all the layers. HS is based on the secant modulus could be estimated from
isotropic hardening and it had the following the equation E50ref=2000Nf(z). Where
basic characteristics: stress dependant stiffness f(z)=1+(-0.0286z-0.1429)0.5 and z depth.
according to the power law, plastic straining Besides, drain analysis, which no excess pore
both due to primary deviatoric loading (shear pressure is generated, was used for analyzing
sandy soil behavior.
hardening) and primary compression
(compression hardening, cap yield), elastic For clay, the total stress undrained analysis
un/reloading, dilatancy effect and failure with undrained internal friction angle θu=0
and undrained shear strength Su were used
according to the Morh-Coulomb criterion.
cu=Su. The Su value can be defined from vane
Almost soil parameters in HS are commonly
shear test or direct shear test. The secant
defined from the CD test and OED test.
modulus for clay was determined from back
However, these soil tests weren’t carried out in
analysis results: E50ref equal to 300Su in case
this case studies. So, soil parameters were
of soft soil and 500Su for stiff clay as shown
determined by Direct Shear (DS) test ,
in Table 4. Those obtained results are
Standard Penetration Test (SPT), the Vane
completely reasonable with previous studies
Shear Test (VST) and empirical formulars
such as: Lim et al (2010), Dao et al (2013),
from back analysis as described in Table 4.
Yong et al (2016), Hsiung et al (2016) and
For sand, it is understood that to get good
Tran et al (2018).
quality sampling for determining the soil

Table 4. Input parameter for soil layer

Depth N ’ c’ Su E50ref Eoedref Eurref

Soil layer m
(m) (value) (deg) (kPa) (kPa) (kPa) (kPa) (kPa)

1. Soft silt clay 1-4 0-1 - - 17-23 300Su 1

2. Stiff clay 4-6 6-9 - - 34-47 500Su 0.8

3. 1-3.6 Medium
6-35 8-25 26-28 5-8 - 2000Nf(z) 0.5
dense sand E50ref 3E50ref

170-
4. Very stiff clay 35-50 30-60 - - 500Su 0.7
270

5. Very dense sand 50-90 25-52 31-33 5-10 - 2000Nf(z) 0.5

Note: In back analysis, the thick sand layer 3 with a thickness of 30m was divided into 6 fine layers with a
thickness of 5m, namely from layer 3.1 to layer 3.6, obeying the change of N (SPT) values according to
depth z.

162
 For input parameters of structures, the In terms of two - dimensional finite element
diaphragm wall was simulated by plate (FE) using Plaxis 2D, the adjacent building was
element. The slabs and the steel struts were simulated such as a flat framed structure
simulated by plate element and node to node including floor, beams and columns, foots.
anchor element in Plaxis 3D 2017 respectively, These structures were demonstrated by plate
and those ones were simulated by elements of elements with flexural stiffness EI (kN/m2/m)
fixed-end anchor in Plaxis 2D version 8.6. The and axial stiffness EA (kN/m). The actual
linear elastic model was adopted to model the dimensions of adjacent building structures
diaphragm wall, slabs and steel struts. In two- included floors (100mm), beams (200x300mm),
dimensional simulation, etabs model was columns (200200mm), shallow foundations
employed to determine the axial stiffness AE of (1200800mm) and length of spacing
opening slabs based on the elastic relationship (Lspacing=4.0m). 2D framed stiffness was
between unit force (P) and corresponding computed for a unit 1m, that equals to total
displacement (Δy) as Equation 2. stiffness of the floor and the beams in a spacing
PL (2) divided by this spacing length as Equation 3.
EA 
y Dead load and live load were general assigned
to weight of the plate element (w). The input
parameters of flat framed structure for Plaxis
2D model was presented in Table 5 follow.
EI floor  EI beams EAfloor  EAbeams (3)
EI1m  ; EA1m 
Lspacing Lspacing

Table 6. Input parameters for 2D frame structure

EA EI w ν
Parameters
(kN/m) (kNm2/m) (kN/m/m)
Figure 9. Opening layout of L1 slab
Floor+beam 3,105,000 5,287 6.00 0.15
Column 270,000 900 0.25 0.15
Foundation 6,500,000 350,000 5.60 0.15

In terms of three - dimensional finite element

(FE) using Plaxis 3D, space framed structure of
adjacent building including floor, beams and
columns, foots were properly modeled
dimensional reality. Plate and beam elements in
Figure 10. Etabs model for determining EA of L1 slab
Plaxis 3D were used to model for floor, foots and
Table 5. Input parameters of structures for beams, columns respectively. Dead load was
Plaxis 2D model automatically calculated through element size
and specific weight of the materials. Live load
EA EI w ν Lspacing
Parameters was evenly distributed on the floor.
(kN/m) (kNm2/m) (kN/m/m) (m)
80m
D wall 24x106 1.28x10 6.4 0.15 -
6

L1 slab 27x103 - - - 1.0 80

40m
m
3
B1 slab 70x10 - - - 1.0
B2 slab 150x103 - - - 1.0
200m
H400 165x103 - - - 7.5
Note: L=1m for length of fixed-end anchor in Plaxis Figure 11. 2-dimensional analysis model by
2D model Plaxis 2D

163
 3rd excavation to -11.8mGL 4th excavation to-15.50mGL
220m
200m

80m

 
Figure 12. 3-dimensional analysis model by
Plaxis 3D

For the purpose of comparison between the

results of back-analysis and measurement.    
Monitoring data of ID02 measuring point,
which located the nearest adjacent building and
the center of the width of diaphragm wall, was
selected. The main purpose is to avoid three-
dimensional effects of the excavation on the
wall deformation. The results of the analysis by
Plaxis 2D model and 3D model were compared
with the field measurement shown as Figure 13 Figure 13. Comparison of predicted value and
and Figure 14. field measurements at various stages.
These predicted results perfectly show Plaxis 2D model Plaxis 3D model
good agreement between predicted values 1st excavation to - 3.8mGL
and field measures in each stage of
excavation. Plaxis 3D and 2D simulations
give quite similar results in horizontal
displacement of diaphragm wall, the
settlement and tilt of adjacent building. These
things show that the selected soil parameters
were perfectly accurate for excavation
behavior analysis.
1st excavation to - 3.8mGL 2nd excavation to -7.3mGL
Tilt i=0.477mm/m Tilt i=0.283mm/m
4th excavation to -15.50mGL

Tilt i =-0.87mm/m Tilt i =-0.842mm/m

Note: (+) tilt toward the excavation, (-) tilt outward
the excavation

164
 Table 7. Calculated ground slope, building tilt,
lateral strain εL angular distortion β and
damage potential index DPI at various stages
1st excavation to - 3.8mGL
Span Bay 1 Bay 2 Bay 3
Ground slope 0.57 0.89 1.05
Tilt 0.48 0.48 0.48
Figure 14. The change in the tilt of adjacent building angular distortion β 0.09 0.41 0.57
according to construction sequences. lateral strain εL 0.00 0.00 0.00
DPI index 1.00 4.10 5.70
3.3 Assessing the damage levels of adjacent
nd
building 2 excavation to -7.3mGL
Span Bay 1 Bay 2 Bay 3
To assessing the damage levels of adjacent
Ground slope -0.31 -0.43 -0.61
building, the the chart of Boscardin and
Tilt 0.23 0.23 0.23
Cording (1989) and the DPI value of Son and
angular distortion β 0.54 0.66 0.84
Cording (2005), mentioned in part 2, were
used. From Plaxis models, the parameter lateral strain εL 0.00 0.00 0.00
values of angular distortion β and lateral strain DPI index 5.4 6.60 8.40
rd
εL of building structure were calculated by 3 excavation to -11.8mGL
displacement of ground surface and building Span Bay 1 Bay 2 Bay 3
structure at each step of excavation model, as Ground slope -1.33 -0.56 -0.67
shown in Table 7 and Figure 15. Tilt -0.21 -0.21 -0.21
angular distortion β 1.12 0.35 0.46
lateral strain εL 0.00 0.00 0.00
DPI index 11.20 3.50 4.60
th
4 excavation to -15.50mGL
Span Bay 1 Bay 2 Bay 3
Ground slope -2.87 -0.20 -0.43
Tilt -0.87 -0.87 -0.87
angular distortion β 2.00 0.67 0.44
lateral strain εL 0.00 0.00 0.00
DPI index 20.00 6.70 4.40
Note: x10-3
By plotting the calculated values of β and
εL to the chart of Boscardin and Cording
Figure 15. Movement of the ground surface and (1989) as Figure 16, it can be seen that the
adjacent building structures at the 4th damage level varies from very slight to slight
excavation step. during excavation sequences. Furthermore,
the values of DPI index denfined by Son and
y A  yB  xc  x A    xD  xB 
Slope = ; Tilt = Cording (2005) were also dertermined and
LAB 2H
shown in Table 7. The max DPI values in
β = Slope - Tilt ; εL =  xC  xD    xA  xB  excavation stages varied from 5.7 to 20. It
LAB  LCD means that damage level varied negligible to

165
slight. comparisons between calculated According to Burland (1977), visible cracks
values based on two criteria and measured with width smaller than 5mm are categorised
data were also presented in Table 8. The as an aesthetic damage with degree of severity
predicted results perfectly show good from very light to light. The calculated values
agreement with field measures in each stage are completely accurate with the results of
of excavation. measuring adjacent building damage at the
field as shown in Table 8.

Table 8. Comparisons between calculated

values based on criteria of building damage
assessment and measured data
Damage Damage
assessed by assessed by Son
and Cording Building damage
construction Boscardin
(2005) measured at the
sequences and
field
Cording DPI Damage
(1989). max level
1st excavation
Figure 16. Assessing building damage potential to - 3.8mGL
Negligible 5.7 Negligible No damage
according to propose Boscardin and
Negligible
Cording (1989). 2nd excavation
Negligible 8.4 to very No damage
to -7.3mGL
slight
From the observation records, in the 3rd and 4th The first
stages of the construction process, excavation to - cracks in
Negligible
11.8mGL and -15.50mGL, the adjacent building 3rd excavation Very external
11.2 to very
to -11.8mGL slight brickwork,
was still in good condition of normal slight
masonry and
serviceability. Main structures of the building ceiling
such as slabs, beams, columns didn’t appear The cracks
cracks. Doors, windows and stairs remained widen in
intact. Appearing only several fine cracks in brickwork,
very masonry and
external brickwork and masonry and plaster 4th excavation
Slight 20.0 slight to ceiling with
to -15.50mGL
ceiling with width of 0.2-0.7mm and length of 60- slight width of 0.2-
85cm, as shown in Figure 17. These cracks were 0.7mm and
length of 60-
not significantly affected on the structural 85cm
integrity of the building and easily treated by
using normal decoration to cover the light cracks. 3.4 Analyzing the damage potential of
It was aesthetic damages with very slight degree. adjacent building at various stiffness
As the analysis in section 3.3, the damage
potential of adjacent building depends on the
stiffness of building. However, in the reality,
the stiffness of adjacent building is not easily
accurate determined because it depends on its
new or old characteristics and existing cracks.
Besides, brickwork walls also affect on total
stiffness of building which is often not
considered when calculating the stiffness of the
frame structures. Therefore, analyzing the
response of adjacent building at various
stiffness is very necessary in predicting
Figure 17. Cracking measurement at the field. damaged potential. In this part, the stiffness of

166
frame structure analyzed in section 3.3, was
considered to increase or decrease by 1.5 to 4
times compared with the initial calculated
stiffness. This change was aimed at surveying
the influence of stiffness on the specific
parameters describing the damage level of
adjacent building.
As the results, the Figure 18 shows that the
tilt of adjacent building does not seem to
change significantly when EI considerably Figure 20. The change in the DPI index according
changes to 4 times in all excavation stages. to the stiffness of adjacent building at 4th
Therefore, it could be stated that the structural excavation to -15.5mGL
stiffness does not significantly affect the tilt of
the adjacent building. Lateral strain εL and angular distortion β,
which are two main parameters to assess
damage potential of adjacent building, were
also investigated. In this case study, lateral
strain εL obtained from the models is very
small, approximately equal to 0. Therefore,
angular distortion β (β=slope-tilt) is a directly
evaluating parameter for damage levels of the
adjacent building through DPI index, as
Figure 18. The changes in the tilt of adjacent Equation 1. As mentioned above, when EI
building decreases, the tilt does not change considerably
and the slope angle goes up due to the increase
Besides that, the differential settlement of differential settlement. So that there is an
among isolated footings of building frames increase in the DPI index of building, as shown
increased significantly when EI decreased in Figure 20. It means that damage level of the
more than 2 times compared to the initial EI. adjacent building increases when EI decreases.
That thing is described through the increasing On the contrary, when the EI stiffness
of slope angle among each bay of building, as increases the DPI index is unchanged. It could
shown in Figure 19. On the contrary, when EI be explained that in case of the building
goes down more than 2 times compared to the structure have enough EI stiffness to share the
initial EI, the differential settlement seems to load evenly among columns and resist ground
be unchanged because of the stable deformation, the tilt and differential settlement
maintenance of the slope angle. This general of building are maintained steady.
response of adjacent building occurs at all
stages of excavation. 4. CONCLUSIONS
From this paper, several conclusions were
summarized as follows:
Using Hardening Soil model in Plaxis
software to calculate for deep excavation had a
sufficient level of reliability. The results of
those analyses show good agreement between
predicted and monitored values of lateral wall
displacement and the settlement of adjacent
building in both Plaxis 2D and 3D model.
Figure 19. The change of slope according to the Soil stiffness parameters E50 can be
stiffness determined by correlation with SPT value (N)

167
for cohesionless soil and undrained shear Dao, T.V.T. Le, T.N. & Nguyen. C.O. 2013. A
strength (Su) for cohesive soil. These values case study of braced excavation using steel
are 2000Nxf(z) in which f(z)=1+(-0.0286z- sheet pile wall in thi vai soft soil clay, Viet
0.1429)0.5 and z depth for cohesionless soil, Nam. JS-Okinawa, Japan.
300Su for soft soil and 500Su for stiff clay. Fok, P. Neo, B. H. Veeresh, C. Wen, D. & Goh, K.
The methods of assessing building damage H. 2012. Limiting values of retaining wall
displacement and impact to the adjacent
potential proposed by Boscardin and Cording
structures. The IES Journal Part A: Civil &
(1989) and formulated damage potential index
Structural Engineering, 5(3), 134-139.
(DPI) by Son and Cording (2005) are Hsiung, B. C. B. Yang, K. H. Aila, W. & Hung, C.
completely accurate with recorded and 2016. Three-dimensional effects of a deep
measured data at the field. These methods excavation on wall deflections in loose to
could be applied in the design stages as well as medium dense sands. Computers and
construction stages of deep excavation to Geotechnics, 80, 138-151.
predict the degree of adjacent building damage Huynh, Q. T. & Nguyen, M. T. 2018. Estimating
in advance. the value of secant module E50 of thick sand
The stiffness of the adjacent building has layer in district 1, Ho Chi Minh city by finite
negligible effect on its tilt change. When the EI element method. Vietnam Journal of
stiffness increases or decreases in the range Construction, Vol. 57, No.609, 136-142.
from 1.5 to 4 times the tilt of adjacent building Lim, A. Ou, C. Y. & Hsieh, P. G. 2010. Evaluation
seems to be unchanged. The differential of clay constitutive models for analysis of
settlement between the shallow footings rises deep excavation under undrained
conditions. Journal of GeoEngineering, 5(1),
significantly when the EI stiffness reduces.
9-20.
The damage levels of adjacent building are Lin, H. D. Mendy, S. Dang, P. H. Hsieh, Y. M. &
assessed through the DPI index (damage Chen, C. C. 2016. Responses of adjacent
potential index), which increases considerably ground and building induced by excavation
when the EI stiffness decreases from 1.5 to 4 using 3D decoupled simulation. Japanese
times. In contrary, when the stiffness goes Geotechnical Society Special Publication,
down from 1.5 to 4 times, these values do not 2(40), 1437-1440.
change appreciably. Sabzi, Z. & Fakher, A. 2015. The performance of
Monitoring system should be sufficiently buildings adjacent to excavation supported by
carried out to ensure safety throughout the inclined struts. International Journal of Civil
construction process and assess the effects on Engineering, 13(1), 1-13.
adjacent buildings. The main purpose is to Son, M. & Cording, E. J. 2005. Estimation of
strictly control excavation risks and the building damage due to excavation-induced
damage of adjacent buildings. ground movements. Journal of Geotechnical
and Geoenvironmental Engineering, 131(2),
162-177.
5. REFERENCES
Son, M. & Cording, E. J. 2008. Numerical model
Cording, E. J. Long, J. L. Son, M. Laefer, D. & tests of building response to excavation-induced
Ghahreman, B. 2010. Assessment of ground movements. Canadian geotechnical
excavation-induced building damage. In 2010 Journal, 45(11), 1611-1621.
Earth Retention Conference-Earth Retention Son, M. 2015. Response analysis of nearby
Conference American Society of Civil structures to tunneling-induced ground
movements in sandy soils. Tunnelling and
Engineers,101-120.
Underground Space Technology, 48, 156-169.
Căpraru, C. & Chirică, A. 2013. Some Remarks on
Son, M. 2016. Response analysis of nearby
the Influence of Deep Excavations on
structures to tunneling-induced ground
Neighbouring Buildings. 2013 - Seventh
movements in clay soils. Tunnelling and
International Conference on Case Histories in Underground Space Technology, 56, 90-104.
Geotechnical Engineering, Missouri Schuster, M. Kung, G. T. C. Juang, C. H.&
University of Science Technology. Hashash, Y. M. 2009. Simplified model for

168
 evaluating damage potential of buildings
adjacent to a braced excavation. Journal of
Geotechnical and Geoenvironmental
Engineering, 135(12), 1823-1835.
Tran, V. T. & Huynh, Q. V. 2018. Back analysis
on deep excavation in the new alluvial
sediment layer in Ho Chi Minh City.
Numerical Analysis in Geotechnics –
NAG2018, Ho Chi Minh City, Vietnam.
Yong, C. C. & Oh, E. 2016. Modelling ground
response for deep excavation in soft
ground. International Journal, 11(26), 2633-
2642.
Zhang, X. Yang, J. Zhang, Y. & Gao, Y. 2018.
Cause investigation of damages in existing
building adjacent to foundation pit in
construction. Engineering Failure Analysis,
83, 117-124.

169
The 3rd Int. Conf. on Transport Infrastructure & Sustainable Development (TISDIC 2019). Construction Publishing House. ISBN: 978-604-82-2893-4.

Study of geopolymer synthesized from fly ash to improve

soft ground

Tran Van Thu1, Nguyen Van Chanh2

1
Department of Transportation Engineering, Ho Chi Minh City University of Transport, Ho Chi Minh City, Vietnam
2
Department of Civil Engineering, Ho Chi Minh City University of Technology, Ho Chi Minh City, Vietnam

Corresponding Author:
Address correspondence to First author: [email protected]

ABSTRACT: Geopolymer is a material that can replace Portland cement. This material is the product of
the reaction process between silicon-rich materials and aluminum with alkaline activated solution. This is
an environmentally friendly material due to the utilization of industrial waste, such as fly ash, while
reducing the greenhouse effects because of replacing portland cement. This study summarizes the results
of research on geopolymer application in the field of improvement soft ground in the world and Vietnam
in recent years. This paper also proposes the research direction and ability to apply this technology in the
field of improvement soft ground soil in Vietnam.
Keywords: Geopolymer, fly ash, geopolymer concrete, improvement soft ground, sustainable
development.

participate in geopolymer reactions [6, 7]. In order

1. INTRODUCTION to take advantage of the huge source of fly ash
from Vietnam, contributing to reducing
Geopolymer technology has been studied and
environmental pollution, ensuring sustainable
developed in recent decades [1]. However, the
development in the field of infrastructure
application of this technology to soft ground to
construction, it is necessary to conduct research
reinforce the foundation is a new issue and there
on characteristics and processes. synthesis of
are still many issues to be studied to clarify
geopolymer from soft soil sources and fly ash of
especially in specific conditions in Vietnam. The
Vietnam, factors affecting the mechanical
term "Geopolymer" was built in the 1970s by
properties of this geopolymer.
French scientist Prof. Joseph Davidovits [1] and
he began to build the principle for making
2. MATERIALS AND METHODS
geopolymer. By relying on the reactivity of
aluminosilicate and Alkaline alkaline solutions to 2.1 Materials
create high-intensity products. The material
system for making geopolymer materials consists 2.1.1 Fly ash
of two main components, initial materials and Fly ash is generated during the combustion of
alkali activators. Aluminosilicate materials to finely ground coal in thermal power plants.
supply Si and Al for geopolymerization process During combustion, the minerals in coal (clay,
occur (usually fly ash, metacaolanh, silicafume feldspar, quartz and shale) are melted,
...[4]. The most commonly used alkali activators suspended and blown out of the combustion
are NaOH, KOH and liquid sodium Silicate chamber along with the exhaust gas. When
solutions to create an alkaline environment and cooled, it will solidify into spherical particles

170
called fly ash. The source of fly ash used in while the amount of SO3 is significant in the
this study comes from Vinh Tan thermal power soil , from 0.07 to 2.84%. The above results are
plant. The composition of these fly ash is relatively consistent with the analysis results of
determined by X-ray fluorescence spectrometer mineral composition of the soil. It can be seen
(XRF) method, the result is as follows: that this is a source of materials mainly
composed of silicon and aluminum as a possible
Table 1. Results of chemical composition analysis source for synthesizing geopolymer.
of fly ash (%)
2.1.3 Alkaline activator solution
Fly ash SiO2 Al2O3 Fe2O3 K2O MgO TiO2 CaO SO3 Na2OLOI
Vinh
55.71 27.51 7.02 5.31 1.17 0.99 0.93 0.44 0.25 8.32
Alkali activators used to synthesize
Tan 2 geopolymer such as NaOH, KOH and liquid
glass Sodium Silicate are all basic chemicals
2.1.2 Clay commonly used in industry. The alkali
activated solution used in this study has a
Clay source in Vietnam is quite abundant,
concentration of 8M; 10M and 12M. NaOH
stretching from North to South. Chemically,
solution is mixed well with sodium silicate
clay is a term used to describe a group of
liquid glass solution in a ratio of 1: 1.
hydrated phyllosilicate minerals, usually with a
diameter of less than 2 μm. 2.2 Methods
Clay consists of phyllosilicate minerals rich
in oxides and hydrocarbons of silicon and Input sources used to synthesize geopolymer
aluminum as well as a large amount of water include: clay mud taken in Ho Chi Minh City;
involved in the structure and change with each fly ash obtained from Vinh Tan 2 thermal
type of clay. Clay is generally created due to power plant, Binh Thuan province; Sodium
the chemical weathering of silicate rocks under hydroxide (dry NaOH) and liquid glass
the action of carbonic acid, but some types of solution. Fly ash and clay mud are dried,
clay are formed by hydrothermal activities. The crushed and then dried in a proportion of fly
composition of some types of clay in Ho Chi ash / mud ratio of 0: 100; 20:80 and 40:60.
Minh City and some provinces in the Mekong These samples are then mixed with Alkali
Delta are as follows: additives against a total volume of 0.25 solids;
0.40 and 0.50. Alkali solution selected includes
Table 2. Results of chemical composition 8M concentration NaOH solution; 10M and
analysis of clay (%) 12M, combined with Na2SiO3 solution at a
ratio of 1: 1. After being shaped and cured at
Clay
source
SiO2 Al2O3 Fe2O3 K2O MgO TiO2 CaO SO3 Na2OMKN ambient temperature (around 300C), the
Ho
samples were taken to determine the
Chi 52.00 20.05 6.73 1.77 1.32 1.16 0.06 0.42 0.58 13.99 compressive strength (after 7 days, 14 days and
Minh 28 days after maintenance). Which samples
have high compressive strength, fly ash
The results of mineral composition analysis composition, clay mud and activated solution
showed that: in clay mud soil, clay minerals are concentration are appropriate.
mainly kaolinite and illite with their content
similar, followed by chlorit and finally 3. RESULTS AND DISCUSSION
montmorillonit. In clay mud soil, highly
dispersed minerals such as montmorillonite and 3.1 Factors affecting compressive strength
illite occupy from over 19 to 25%, kaolinite
minerals account for 16-18%. The content of the 3.1.1 Effect of alkali activated solution
main oxides in the soil is SiO2 from 52.00 to concentration on compressive strength
71.02%, Al2O3 accounts for 12.50 to 20.05%, Experimental results are shown in Figure 1,
the content of Fe2O3 accounts for 3.76 to 7.01%, Figure 2 and Figure 3.

171
 When fly ash is not used, the maximum
compressive strength achieved is the lowest (qu
= 33.8 daN/cm2), corresponding to a
concentration of NaOH solution is 8 mol/l;
3.1.2 Effect of fly ash content and curing
time on compressive strength
Experimental results are shown in Figure 4,
Figure 5 and Figure 6.
Figure 1. Effect of alkali activated solution
concentration on compressive strength
when fly ash content is 0%

Figure 4. Effect of fly ash content and curing time on

compressive strength when concentration of
alkaline activated solution is 8 mol/l
Figure 2. Effect of alkali activated solution
concentration on compressive strength
when fly ash content is 20%

Figure 5. Effect of fly ash content and curing time

on compressive strength when
concentration of alkaline activated
solution is 10 mol/l
Figure 3. Effect of alkali activated solution
concentration on compressive strength
when fly ash content is 40%

The result of the obtained test shows that:

The maximum compressive strength
achieved with the fly ash content is 40%
(qu = 48.1 daN/cm2), the ratio of activating
solution accounts for 40% of solids and
NaOH solution concentration. is 10 mol/l;
When the fly ash content is 20% of the
solids content, the use of NaOH solution is 8 Figure 6. Effect of fly ash content and curing time on
mol/l; 10 mol/l or 12 mol /l result in compressive strength when concentration of
compressive strength of almost equivalent; alkaline activated solution is 12 mol/l

172
 The result of the obtained test shows that: 3.2.2 Effect of fly ash content and curing
The compressive strength gradually increases time on elastic modulus
with the amount of fly ash used and reaches a
Experimental results are shown in Figure 10.
relatively stable value when the fly ash content
is used about 40%. The compressive strength
gradually increases over time and reaches a
relatively stable value after 14 to 28 days.
3.2 Factors affecting elastic modulus
3.2.1 Effect of alkali activated solution
concentration on elastic modulus
Experimental results are shown in Figure 7,
Figure 8 and Figure 9.

Figure 10. Effect of fly ash content and curing

time on elastic modulus when solution
concentration is 10mol/l

Most of the experimental samples, the value

of elastic modulus E ≥500 daN/cm2, after 7
days, satisfy the requirements of making the
bottom of the road foundation according to the
standard [8] . After a period of 14 days to 28
Figure 7. Effect of alkali activated solution
days, the value of the elastic modulus of the
concentration on elastic modulus when
fly ash content is 0% geopolymer material is large enough to make
up most parts of the highway. Thus, the use of
geopolymer synthesizes clay and fly ash to
make roads of great potential;
Experimental results of elastic modulus are
quite similar to those achieved with compressive
strength. The largest repression module is
achieved (E = 3209 daN/cm2 = 320 Mpa), with
the use of fly ash content of 40%, the ratio of
activating fluid accounts for 40% of solids and
Figure 8. Effect of alkali activated solution concentration of NaOH solution is 10 mol/l;
concentration on elastic modulus when Modulus of elasticity increases with
fly ash content is 20% increasing fly ash content, when fly ash content
reaches about 40% value of elastic modulus
increases quite slowly.

4. CONCLUSIONS
Based on the experimental results, it can be
concluded that: The optimum fly ash content is
about 40%, the ratio of activating solution
accounts for 40% of solids and the
Figure 9. Effect of alkali activated solution concentration of NaOH solution is 10 mol/l;
concentration on elastic modulus when Geopolymer synthesized from clay and fly
fly ash content is 40% ash with the value of elastic modulus E ≥500

173
daN/cm2, after 7 days of curing, satisfying the Jian He.2012. Synthesis and Characterization of
requirements of making highway bed bottom Geopolymers for Infrastructural Applications.
according to TCVN 4054: 2005, after 28 days PhD Thesis, Nottingham University, UK.
The date the maximum elastic modulus value Madawala, Eric and Ajit .2014. Reuse options for
is obtained E = 3209 daN/cm2. Thus, the use of coal fired power plant bottom and fly ash.
Reviews in Environmental Science and
geopolymer synthesized from clay and fly ash
Bio/Technology.
to make highway and improve soft ground has Ma, Y., Hu, J. and Ye, G.2012. The Effect of
great potential. Activating Solution on the Mechanical
The chemical composition of clayey mud in Strength, Reaction Rate, Mineralogy, and
Ho Chi Minh City and the Mekong Delta microstructure of Alkali- Activated Fly Ash. J
region is quite similar, so the above experiment Mater Sci, 47, pp 4568-4578.
results can be referenced for the whole Mekong Palomo, A., Alonso, A., Fernandez-Jimenez, A.,
Delta region. Sobrados, I. and Sanz, J.2004. Alkaline
Activation of Fly Ash: NMR Study of The
5. REFERENCES Reaction Products. J. Am. Ceram. Soc., 87 (6),
pp1141-1145.
Davidovits, J.2008. Geopolymer Chemistry and N.V. Chanh, B.D. Trung and D.V. Tuan.2008.
Application. 2nd edn, Institut Geopolymere, Recent research geopolymer concrete. The 3rd
Saint- Quentin, France. ACF International Conference - ACF / VCA.
Hardjito, D.2005. Studies of Fly Ash-based TCVN 4054: 2005. Highway − Specifications for
geopolymer Concrete. PhD Thesis, Curtin design.
University of Technology, Perth, Australia.

174
The 3rd Int. Conf. on Transport Infrastructure & Sustainable Development (TISDIC 2019). Construction Publishing House. ISBN: 978-604-82-2893-4.

A study on the main factors affecting the reinforcement

corrosion in mechanically stabilised earth walls and
predict the service life of the wall

Chau Truong Linh 1,*, Nguyen Thu Ha 2, Vu Dinh Phung 3, Pham Van Ngoc 4
1,2,4
Faculty of Bridge and Road Engineering, University of Science and Technology, Da Nang City, Viet Nam
3
Faculty of Transportation Engineering, Faculty of Civil Engineering, Thuyloi University, Viet Nam
*
Corresponding author, E_mail: [email protected]

ABSTRACT: In this study, the Principal Component Analysis method is applied to determine the
main factors affecting reinforcement corrosion in mechanically stabilised earth (MSE) walls. A
response surface model is used to build the correlation function between corrosion and main factors,
which then applied to design a program to predict the service life of the MSE wall used steel
reinforcement (MSE-T). Studies use databases from the French Road and Highways Research
Institute, and National School of Bridges and Roads by measuring mechanical-physical-chemical
parameters of 770 test specimens extracted from 125 walls on the highway system throughout France.
The results of MSE-T program are used to simulate the scenarios of steel corrosion of MSE walls by
FLAC software, which could show the stress-strain displacement of the walls at the extracted time.
The research results could provide a warning of degradation of the MSE wall in the end stage of
working period to enhance safety in the operating process and propose the solutions to improve the
longevity of the MSE walls.
Keywords: MSE wall, MSE-T program, numerical modelling, corrosion scenario, principal component
analysis (PCA), response surface methodology.

1. INTRODUCTION factors like the physical-chemical properties

of soil, the condition of environments, and
The MSE wall has been applied popularly in the quality of reinforcement. Many MSE
construction, especially in reinforcing the walls collapsed during the operation could
vertical slope and in constructing high affect the transportations and stabilisation of
embankment. The MSE wall shows many the embankment (LCPC 2003, Chau et al.
outstanding advantages such as long 2016). Therefore, predicting the service life
durability, low price, being suitable for of this type of wall accurately will provide
infrastructures subjected to dynamic loads, the warnings, maintain and repair it to
and reducing differential settlement for the prevent risky accidents (Chau et al. 2014, Le
infrastructures on soft soil (Duong 2012, et al. 2016). The authors have designed an
LCPC 2003, Chau et al. 2016, Les 1979). In MSE-T program to calculate the service life
the MSE wall, the service life of the wall of reinforced MSE wall, and analysis results
depends on the strength of reinforcement from this program will be used to evaluate
during the serving time. However, the level the stress-deformation of the wall by FLAC
of reinforcement corrosion depends on many 2D software.

175
2. THE MAIN FACTORS AFFECTING reinforcement corrosion are selected (F1 and F2
THE REINFORCEMENT CORROSION are combination CPA 1 and combination CPA 2,
IN MECHANICALLY STABILISED respectively). The result is shown in Figure 1.
EARTH WALLS
2.1 The database for research
In order to build the program, the authors used
the empirical data of France Institute of Roads
and Highways (SETRA), France National Bridge
and Road University. During period 2004-2007,
the institute conducted many experimental tests
to survey the status of reinforced soil, which was
applied for retaining walls and abutments. The
steel-reinforced MSE wall was applied for over
125 retaining walls on the highways system Figure 1. Analysis of the parameters affect the
throughout France. They collected 770 samples reinforcement corrosion on the
of reinforcement and soil from those correlation unit circle
infrastructures to calculate and analyse the
Table 1. Matrix correlation between calculated
properties of reinforcement and soil (Chau et al.
parameters
2016). The collected data are essential for
evaluating the level of reinforcement corrosion Factors Effect factors
accurately and predicting the service life of the Average width (mm) 0.129
reinforced MSE wall.
Average thickness E0 (mm) -0.268
2.2 Analysis of the factors affecting the Tensile strength (MPa) 0.052
reinforcement corrosion by using Principal Tensile force of reinforcement
Component Analysis -0.119
(N)
The purpose of Principal Component Analysis Humidity W -0.134
(PCA) is to perform a linear mapping of data to Impedance R (W.cm) -0.188
a lower-dimensional space in such a way that pH -0.337
the variance of the data in the low-dimensional
Ion [Cl-] mg/kg -0.027
representation is maximised. In practice, the 2-
covariance matrix of the data is constructed, and Ion [SO4 ] en mg/kg 0.537
the eigenvectors on this matrix are computed. Zinc thickness Ezn (μm) 0.615
The eigenvectors that correspond to the largest Elongation (%) -0.068
eigenvalues can now be used to reconstruct a Corrosion 1.0
large fraction of the variance of the original
data. Moreover, the first few eigenvectors can Figure 1 and Table 1 show that the
often be interpreted in terms of the large-scale parameters affecting the degree of
physical behaviour of the system. The original reinforcement corrosion have a correlation with
space has been reduced to space spanned by a corrosion, including impedance (R), humidity
few eigenvectors (Jolliffe 2002). (W), pH, sulfate concentration (SO42-) average
It is the fact that not all 770 samples contain thickness E0, Zinc thickness Ezn. In retaining
adequate information. Therefore, only 66 wall located in the coastal area, Cl-
samples with fully necessary data were selected; concentration have a clear influence on the
the PCA was used to evaluate the covariance level of reinforcement corrosion. Therefore,
matrix, which shows the relationship among the seven influence factors effect on the
main components. Then the covariance reinforcement corrosion are choice, included R,
coefficients which affect the rate of W, pH, ion Cl-, ion SO42-, Ezn, E0.

176
2.3 Modeling Response surface to estimate Where α0 is linear constants, Xi is variables;
the rate of reinforcement corrosion i is elected factors affecting the rate of
corrosion (i from 1 to 7). Each parameter Xi is
Response surface is a collection of mathematics
assigned by 66 values from the results of
and statistics to build the experimental models,
experiments; Yi is Experimental corrosion; Ỹi
thoroughly designed in experiments. The object
is Modelling corrosion (Table 3).
is to optimise a behaviour (output variable) that
are affected by many independent variables Table 3. Input parameters
(input variables) (Faravelli 1989). In this method,
the error is assumed random. RSTOOL in Sample for experiment (1÷ 66)
MATLAB is applied to find out the regression
Corrosive value (experimental) Yi
function of 4 models, as shown in Table 2.
Impedance R X1
Table 2. Regression functions in Matlab
Humidity W X2
Models Regression functions
pH X3
Linear
Y   0   i X i Ion Cl-
X4
response i 1

surface Ion SO42- X5

Pure quadratic
Y   0   i X i   i X i 2 Medium thickness E0 X6
response i 1 i i

surface Galvanised thickness Ezn X7

Interaction Corrosive value (simulation) Ỹi
Y   0   i X i    i X i X i 1
response i 1 i i

surface The model is analysed by the full quadratic

Full quadratic response surface gives results as shown in
Y   0   i X i    i X i X i 1   i X i 2 Figure 2 and Figure 3, and the equation of
response i 1 i 1 i 1

surface regression is as follow:

Figure 2. Analysis by the full quadratic model

Figure 3. Comparison between experiment and model

177
Y 1  54, 2555  0, 0160  R  5, 9328  W+9, 5666  pH  0, 0582  Cl 3. THE PROGRAM TO PREDICT SERVICE
0, 5331  SO4  19, 2802  E0  0, 7659  E ZN  (5,12 E  05)  R  W LIFE OF REINFORCED MSE: MSE-T
+(3, 95 E  05)  R  pH  (1, 04 E  05)  R  Cl  (1, 62 E  05)  R  SO4
0, 002617  R  E0  (4, 99 E  06)  R  E ZN  0, 012071  W  pH 3.1 Flow charts, algorithm and theme program
0, 000766  W  Cl  0, 001348  W  SO4  1, 2568437  W  E0
0, 0096743  W  E ZN  0, 0130224  pH  Cl  0, 0412501  pH  SO4 Using Microsoft Visual Studio and using
0, 6355959  pH  E0  0, 0270347  pH  E ZN  0, 0007228  Cl  SO4 language C# to build the program. Microsoft
0, 0231137  Cl  E0  0, 0003445  Cl  E ZN  0, 0468423  SO4  E0 Visual Studio is an integrated development
0, 0002234  SO4  E ZN  0,1286062  E0  E ZN  (1, 09 E  07)  R 2 environment (IDE) from Microsoft. It is used to
0, 039367  W 2  0, 596285  pH 2  (6, 26 E  05)  Cl 2 develop computer programs, as well as websites,
0, 000148  SO4  0, 346901  E0 2  0, 003017  E zn 2 web apps, web services and mobile apps. C# is a
general-purpose, object-oriented programming
(1) language, simple and modern. Diagram of
According to the correlated model and calculation is shown in Figure 4.
correlated coefficient R2, the model of full
quadratic has strongest reliability (R2 = 0.981)
as shown in Table 4. Thus the model of full
quadratic is used for calculating to determine
the level of reinforcement corrosion. Where: Ӯi
is the average of the Yi

Table 4. Correlated coefficient R2 between

experiment and model
Model R2  1 
 (Y  Y )
i i
2

 (Y  Y )
i i
2

Linear 0.741
Pure quadratic 0.875
Interaction 0.957
Full quadratic 0.981
Figure 4. General diagram of calculation

Figure 5. Input screen

178
3.2 Input data, calculation, results parameters is according to (Chau and Pham
2014, Le and Chau 2016). Selecting two
General and detailed information: define the typical scenarios of reinforcement corrosion:
name of work, content, date of input, export continuous corrosion due to chemical
data and output parameters. Input parameters properties of soil (scene 1) and local
include the physical-mechanical of the soil corrosion due to cracking of the wall (scene
behind the wall and the parameters of the 2) (Table 5 and 6).
calculated properties of the reinforcement, as
shown in Figure 5. Table 5. Wall and reinforcement parameters
Selecting the type of calculation: preliminary
calculation and detailed calculation are used for Wall height H (m) 10.5
unbuilt works and the works already under
construction, respectively. Selecting the type of Number of layers 14.0
corrosion: define the point of time starting
Length of reinforcement (m) 8.0
operation and starting corrosion. Level of
increasing (A) is selected, depending on the Reinforcement diameter CT5 (mm) 10.0
natural condition as well as the properties of
reinforcement that choose A following the type Galvanized reinforcement 2.0
of corrosion (Figure 6a and 6b). (bar/connect)
After analysing, software results includes Ground width not reinforced (m) 7.0
A0: the degree of reinforcement corrosion after
the first year since the infrastructures start the The soil under the wall (m) 5.0
operation, A1: the degree of reinforcement
corrosion in the first year since the work is Table 6. Initial parameters for calculating
started corrosion. After calculating the service corrosion
life of the wall, the results should export to a
word and excel format containing all the input Scene 1 Scene 2
parameters and results.
R (W.cm) 2913 1350
3.3 Example of the application of MSE-T
program to calculate the service time of W (%) 8.6 13.9
reinforced MSE wall pH (%) 10.6 6.9
Application of Trinh Tung bridge at the Ha Cl- (mg/kg) 82 219
Noi - Lao Cai expressway. The physical-
chemical properties of soil as for input SO42- (mg/kg) 37 94

Figure 6a. Preminaliry calculation - Prediction of the service life of the wall in scene 1

179
Figure 6b. Detail calculation - Prediction of the service life of the wall in scene 1

4. THE COMBINATION OF MSE-T AND Table 7. Results from software MSE-T and
FLAC PROGRAM TO ESTIMATE simulation parameters in FLAC 2D
STABILITY OF MSE WALL OVERTIME
Scene 1 Preminaliry Detailed
In order to observe the change of reinforcement calculation
over time and the change of stress-deformation,
deformation between wall and reinforcement, the A (μm) 0 -
authors connect the MSE-T and FLAC 2D to
perform the calculation (Table 7). n 0.65 0.65
The service life of the wall is calculated since A0 (μm) 20.02 -
the structure of the wall is operated until the
remains of strength value of reinforcement are A1 (μm) 20.02 29.05
less than half of its original value. At this point,
Service life (years) 118.1 58.94
the wall is still able to work, and the reinforced
steel continues to be corroded. According to BS
Adding the code into FLAC, the software
8006 (1995), the average thickness loss (on a
will automatically analyse the change of stress,
reinforcement surface) at the time point t (year)
deformation of the wall elements over time
due to corrosion is Δe = A.tn with the
until the work is collapsed.
galvanization: n = 0.65.

Figure 7. Model of the MSE Figure 8. Model of MSE on FLAC

180
 a) In Operation b) Collapse in scene 1 (148 years) c) Collapse in scene 2 (67,7 years)

     

Figure 9. Increasement of shear deformation

     
Figure 10. Horizontal deformation of the wall 

Figure 11. Tensile force in the reinforcement layers in operation

Figure 12. Remaining tensile force in the reinforcement layers in varying times

Figure 13. Deformation of the wall shell over time

181
 The code added into the software is written parameters, pictures of stress-deformation,
by C++ and based on the relationship between transposition of shell wall and tension chart
the reduction of cross-side (thickness and along on reinforcement.
width of reinforcement), elongation and the The study results could provide a warning of
remaining tension of the reinforcement. This MSE-T damage at the end of stable operation
relationship is built in reinforcement in order to improve safety during exploration
experiment from Chau et al. (2014). The result and to propose solutions to increase the service
of calculation by using FLAC after integrated life of the wall.
from MSE-T is shown in Figure 7-13.
In all reinforcement layers, the second layer 6. REFERENCES
of scene 1 (from the foot of the wall to the top)
is broken first because of the influence of the Duong Ngoc Hai 2012. Design and construction of
embankment. In scene 2, is the 11th reinforced earth walls, Vietnam Construction
reinforcement, this area is most corroded and Publishing House Limited Company (in
affected by cavitation. Vietnamese).
It could be seen that the axial force in LCPC, ed. 2003. Guide technique -
reinforcement locates at near the wall shell (1m Recommandations pour l’inspection détaillée,
from the edge of the wall) and reduce steadily le suivi et le diagnostic des ouvrages de
along to the ends of reinforcement. For scene 1, soutènement en remblai renforcé par des
from 0 to the year 120th, the change of tension éléments métalliques, 102.
strength value in reinforcement is negligible. Chau Truong Linh, Corfdir A., Bourgeois E. 2016.
Since the 148th year onwards, it is distributed Corrosion des armatures sur le comportement
significantly the tension over the area of des murs en terre armée - Effect of
corroded reinforcement due to the impact of reinforcement corrosion on the behaviour of
physical-chemical properties of the soil. For earth walls reinforced by steel elements,
scene 2, in the early to the 60th year, the change Éditions Universitaires Européennes.
of traction value in the reinforcement was Standard BS 8006:1995 - Standard practice of soil
negligible. Since the 68th year onwards it is and other fill materials. Vietnam Construction
distributed the tension over the area of corroded Publishing House Limited Company (in
due to the impact of aggression. Vietnamese).
Standard 22TCN 272:2005 - Ministry of transport
5. CONCLUSION in Vietnam (in Vietnamese).
Les ouvrages en terre armée, recommendation et
By analysing of the soil and reinforcement data
règles de l’art. 1979. Guides techniques.
from 770 samples using Principal Component
Analysis, the main factors affecting the degree of SETRA, ed. Ministère des Transports.
reinforcement corrosion are determined, including Chau Truong Linh, Pham Van Lim 2014. Study on
R, W, pH, ion Cl-, ion SO42-, Ezn, E0.
behavior and life of reinforced earth walls
Using Response Surface model to build a through reinforcement corrosion scenarios.
relative function between the degree of Science Journal of Transportation, Vietnam,
reinforcement corrosion with seven main 12, 36-39 (in Vietnamese).
factors affecting. Le Hong Long, Chau Truong Linh 2016. Studying
The authors have created the program to the effects of different types of natural
predict the service life of reinforced MSE; the aggregates on soil interaction - the core in the
results can be used to evaluate existing works earth retaining wall with self-fabrication core
if having actual data on physical-chemical has considered the life due to corrosion.
properties of soil and reinforcement. Science Journal of Transportation, Vietnam,
The program has been applied to calculate 7, 69-72 (in Vietnamese).
the MSE wall for the structure in Ha Noi - Lao Jolliffe I.T. 2002. Principal Component Analysis,
Cai Expressway with two different corrosion Second Edition. Springer Ed.
scenarios. Faravelli L. 1989. Response surface approach for
Using the results from MSE-T program to reliability analysis. J. Eng. Mech. 115(12),
add into FLAC 2D software to export the 2763- 2781.

182
The 3rd Int. Conf. on Transport Infrastructure & Sustainable Development (TISDIC 2019). Construction Publishing House. ISBN: 978-604-82-2893-4.

Cases study on the effect of down drag on bearing

capacity of piles

Cao Van Hoa1,, Chau Truong Linh2

1
Department of Civil Construction, Hochiminh University of Architecture, Hochiminh City, Vietnam
2
Department of Civil Construction, Danang University, Danang City, Vietnam


Corresponding author, E-mail: [email protected], [email protected]

ABSTRACT: Piles behaviour is characteristic by load bearing capacity and displacement. Pile
displacement in normal practice is an allowable value, with assumption that pile load is small enough to
keep bearing soil around the piles to behave as elastic material. Many researchers recognized that
negative friction together with the external load may become very large that it exceeds allowable bearing
capacity. Negative friction is developed due to consolidation settlement of soil surrounding piles, caused
by lowering ground water level, load from superstructure and/or surcharge load. This study tries to
explain the behaviours of pile: both load bearing capacity and pile - soil settlement. The study shows that
when force equality plane (FEP) and settlement equality plane (SEP) coincide, pile capacity is balanced
with external load plus down drag and settlement of soil balances with pile allowable displacement.
Keywords: Negative friction, down-drag, consolidation, pile group, settlement, case study.

1. INTRODUCTION (3) Due to surcharge-loaded soils: This

situation occurs when a surcharge load is
Friction occurs when there is a relative sliding placed on the ground surface around a
between pile and its surrounding soil. Negative previously installed pile in which the subsoil
friction may occurs: was in equilibrium or when groundwater level
(1) Due to remolding of soil when piles are is lowering. The surcharge load will cause
driven in clay: An estimate of its magnitude is settlement which is prevented by these piles,
given by Johnson and Kavanagh (1968). The thereby mobilizing negative skin friction.
value of negative friction in this case is Depending on the magnitude of the surcharge
normally smaller than any other cases; load, and the nature of the compressible soil,
(2) Due to soils undergoing consolidation: this situation is likely to cause the maximum
Compressible soils that are undergoing active amount of down-drag force on the piles.
consolidation when piles are drived will (Fellenius, 1972);
produce down-drag. This consolidation Negative friction creates down-drag force in
settlement causes the usual mechanism of piles as additional load, and this force is
dissipation of excess pore-water stress but it is proportional to the pile’s length. Johanessen and
prevented from occurring adjacent to the pile Bjerrum (1965), Bjerrum et al. (1969), Bozozuk
because of the adhesion and friction of soil to (1972) noted that the value of down-drag force
the pile (Endo et al., 1969). Subsoil is also exceeded the allowable load, the one that is
consolidated due to stress from superstructure normally used in estimation of pile capacity.
distributed into soil through raft and piles based Fellenius & Broms (1969) and Fellenius (1969)
on Bousinesque and/or Mindlin solutions. published their measured result, showing that

183
down-drag foce can spontaneously increased due It can be recognized that negative friction
to soil re-consolidation after driving piles. does not cause failure of piles geotechnically or
Koerner (1971) did many experiments in physically, the main concerns for design are:
order to analyze the behaviour of negative (1) The additional down-drag force
friction. The results showed that under small developed from negative friction, and hence
load, negative friction only appeared on the top the structural pile integrity;
half of pile, and the larger the load, the further (2) The additional displacement of the pile
down the negative friction influenced on it. head.
Endo et al. (1969) executed tests on longer
piles, thus he found the location of Force 2. ASSUMPTIONS OF PILE BEHAVIOUR
equality plane (Neutral plane).
The magnitude of displacement required in 2.1 Condition #1: Force equality
order to develop negative friction, is mentioned Assuming that piles are installed in soil
by many researchers. Worker and Darvall undergoing consolidation. The condition of
(1973) believed that with 35 mm settlement at force equality required the load distributed
surface ground by surcharge load from 3 m along a pile to be equal to the pile bearing
hight embankment around a pile is enough to capacity. The horizontal plane at the point
mobilize negative friction down to the depth of where the load equal the bearing capacity is
18 m. Bjerin (1977) confirmed that negative called “Force equality plane – FEP”.
friction appeared at the maximum depth of 25
m when the ground surface 0.12 m away from 2.1.1 Bearing capacity of pile
pile was settled 5 mm, and at the distance of 5
Bearing capacity of pile includes toe resistance
m it settled 8 mm. Bozozuk (1981) found out
and shaft resistance. Fellenius (2011) believed
that a reversal direction of shear forces down to
that the toe resistance mainly developed from
a depth of 20 m occurred when loading a pile
the behavior of compresstion, and it has no
and it generates a relative movement of about 5
ultimate value. When the load is increased, the
mm at the pile head.
volume of soil is compressed to the maximum,
The reported observations indicate that no
but it never reachs failure or peak value. Thus
"slip" between pile and soil takes place and
the toe resistance is only an assumption value. It
that insignificant movement is all that is
can be assumed that behavior of toe resistance is
needed to generate shear stress or to reverse the
similiar as behaviour of compressing force at
direction of shear along the pile-soil interface.
the plate loading test. Site experiments of
Poulos (1997) highlighted some
Ismael (1985), Bergdahl et al. (1984) or
misconceptions about negative friction. It is the
Osterberg test of Fellenius (2011) showed that
concept that negative friction reduces the
on load - displacement curve, there is no failure
ultimate bearing capacity of piles, thus some
point. Behavior of toe resistance can be
method of designs attempt to compute the
simulated by hyperbol model, and it can be
reduced ultimate bearing capacity in which the
determined by many methods. For this study,
downdrag load is substracted from the original
toe resistance will be determine by method β:
load capacity. This concept is generally
rt  N t  ' z  D (1)
incorrect because pile is just only failed
geotechnically. When failed, pile displacement Rt  At rt  At N t  ' z  D (2)
become larger than soil settlement, whereas where
negative friction requires soil settlement larger Nt Toe “bearing capacity” coefficient
than pile displacement. There are cases that D Depth to pile toe
negative friction can reduces the ultimate 'z=D Effective overburden stress at the
bearing capacity. However, there seems to be pile toe
no evidence to demonstrate that these At Toe area (normally, the cross sectional
phenomena has been experienced. area of the pile)

184
 The shaft resistance developed from friction Figure 1a shows curves of load and bearing
is characteristic by shearing force, therefore it capacity along a pile (Fellenius, 2011). It
has ultimate value (Fellenius, 2011). That shaft shows that the load acting on pile increased
resistance is a real value. The external load is from top to toe because of the influence of
distributed along pile length due to shaft down-drag. And the pile bearing capacity
resistance. The remaining load at pile toe shall decreases from the top to toe because of the
be borne by toe resistance. For this study, shaft influence of positive friction.
resistance (both negative and positive) will be The condition of force equality is when the
estimated by: load curve (formulae 5) meets bearing capacity
rs  c '   ' z (3) curve (formulae 6). In figured 1a, the plane
R s   As rs dz   As (c'   ' z )dz (4) passing the point where Pz curve intersect Rz
where curve, is called “Force Equality Plane - FEP”.
c‘ Effective cohesion intercept It can be noted that:
 Bjerrum-Burland coefficient - With a specific value of toe resistance, the
'z Effective overburden stress FEP’s location will be at the same place, even
As Circumferential area of the pile at Depth z using different calculation methods or different
(surface area over a unit length of the pile) friction parameters, because the curves of
The distributed bearing capacity can be negative and positive friction are symmetrical;
determined by: - When lengthen the pile, the value of toe
resistance increase lightly while shaft
Rz  Rult   As . . z' dz
(6) resistance is greatly increased; thus FEP get
 Rt  Rs   As . . z' dz lower toward its toe, and vice versa. By
Whereas, Rt is toe resistance in which soil is selection pile length, one can adjust FEP
being behave elastically, Rs is total ultimate location.
shaft resistance. - To locate the FEP correctly, it is very
important to determine the pile toe resistance
2.1.2 Distributed load along pile length as accurate as possible.
Distributed load includes external load acting 2.2 Condition #2: Settlement equality
on pile head and down-drag load. It can be
determined by: At location where subsoil settlement is equal to
Pz  Po   As . . z' dz (5) pile displacement, where there’s no slip
between pile and subsoil, the friction is equal
Whereas Po is external load, include loads
to zero. Above this location, soil is settled
from superstructure.
more than pile displacement, thus it develops
2.1.3 Force equality plane negative friction. Below this location, pile
displacement is greater than soil settlement, it
developes positive friction. Plane that intersect
this location is called “Settlement Equality
Plane – SEP”.
2.2.1 Subsoil settlement
Subsoil settlement is originated from the soil
consolidattion. Estimation of soil settlement
need to consider not only the direct and/or
indirect load impacting on subsoil, but also the
effects of lowering groundwater level and/or
a) b)
urbanization load.
Figure 1. Force and displacement equality planes Subsoil in urban area is always under stress
by Fellenius of new constructions and other human

185
activities. Erkens et al. (2015) warned that settlement superposition method is to
HCM city is among top 5 cities in the world determine soil settlement, disregarding strength
that are settling every year. During the last 25 of raft. It may be estimated by:
year, HCM city is settled almost half a meter, e1i  e2i
and its main cause is the overuse of S   in1 Si   in1  hi (10)
1  e1i
groundwater for everyday living. Le Van
Whereas:
Trung (2017) recorded that ground at HCM’s
S The final soil settlement at ground.
suburban districts get settled from 5 to 10 mm
Si The final settlement of soil layer i
every year, with the main cause is overuse of
n Number of soil layers
groundwater. Specifically, it is recorded that
e1i Void ratio of soil layer i at the load
Nguyen Huu Canh St. is 1.2m lower since it
level equal to overbuden stress
was built (2002). Many researchers in the
e2i Void ratio of soil layer i at load level
world noted that with only very few milimeter
equal to total value of overbuden stress and
of ground settlement, negative friction can
surcharge stress.
reach down to depth of several tens meters.
hi Thickness of layer i.
Therefore, negative friction is a real in pile
For deep foundation, the soil settlement can
construction practice in Ho Chi Minh city.
be estimate based on Mindlin solution as
Calculation of subsoil settlement is based on
similar assumption in PRAB, writen by
calculation overbuden stress and stress from
Matsumoto & Kityodom (2002, 2003).
external load.
n
Stress from shallow foundation and wi    ij Pj (9)
surcharge load are distributed in subsoil based j 1

on Bousinesq solution: Where:

 z  k. p (7) wi = Settlement of node i
Where: aij = Soil flexibility coefficient denoting
z l settlement at node i due to a unit load acting at
K A factor depend on and node j
b b
z Depth from loading area Pj = Force acting at node j
b Raft width Subsoil settlement at any point in semi-
l Raft length space under foundation is a superposition of
Additional stress due to lowering water the settlements caused by external loads (from
level: shallow foundation, pile foundation,
embankment surcharge and water lowering).
 z  H i ( i   ib )   fill h fill   w H (8)
Where: 2.2.2 Pile displacement
H i Thickness of soil layer in draining area
Pile head displacement, sh is consisted of pile
i Natural unit weight toe displacement, sb and elastic deformation of
 ib Bouyant unit weight pile, se:
 fill Natural unit weight of upper soil layer sh = sb +se (11)
(filling) Allowable pile head displacement,
h fill Thickness of upper soil layer (filling) according to the British, French, Japanese
standards and by many other researchers, is set
w Water unit weght from 2.5% to 10% of pile diameter. Chinese
H Depth from ground water level to top standard sets displacement at 40 - 80mm, or:
level of pressured aquifer se =2PL/3EA+20 (mm) (12)
Subsoil under shallow foundation and soil Poulos & Davis (1980) proposed:
surcharge is settled because of the stress db
originated from external load, and it can be sh  (13)
30 F
determined by various ways. The sub-layer

186
 Elastic deformation of pile can be The building used open-end steel piles of 500
determined by: mm in diameter and 35m in length. It was
L  i .Li
believed that these piles have more than enough
se   0 (14) bearing capacity. But two years after completion,
E
its displacement is recorded to be 150mm, and
2.2.3 Settlement Equality Plane 200 mm two years after that. In the end, the
“Settlement Equality Plane – SEP” is a vital building was demolished because of its excessive
condition to determine the area where pile is displacement.
under positive friction or negative friction. In figure 2, Fellenius (2011) based on Inoue
Figure 1b shows curves of pile displacement, et al. (1977)’s paper, displayed a construction
subsoil settlement and SEP. site and geotechnical profile. It can be seen
It can be noted that: that there are two layers of silty clay of about
- If pile displacement is larger or equal to 10 m thick at depths of 10 and 25m, what are
soil settlement, no negative friction developes laid between fine sand layers. Because there is
at pile shaft; no available data of consolidation test of these
- If pile displacement is smaller than soil layers recorded, therefore this article will use
settlement, negative friction developes along the data of similar soil in Vietnam, specifically
the whole pile shaft. at Phú Mỹ 1 Thermal Plant Project, Bà Rịa –
- The intersection where pile displacement Vũng Tàu province, to demonstrate the case of
curve meet soil settlement curve, devides the damage due to down-drag.
pile into upper half where negative friction The building got excessive displacement
developed and lower half of where positive because it built on subsoil undergoing
friction developed. consolidation, mainly due to 3.5m thick sand fill.

2.2 Pile design

When designing a pile, during the matching
of SEP and FEP, these three cases may
occured: < 102 m >

Pile is considered to be capable to bear

external load if the location of FEP is below SAND FILL

the location of SEP, because its bearing FINE SAND

capacity is greater than the load. SILTY CLAY: w = w L = 40% - 60%; τu = 40

KP
- Pile is not capable to bear load if FEP is FINE SAND

above SEP, meaning that its bearing capacity is SILTY CLAY: w = wL = 40% - 60%; τu = 80

smaller than external load.

SILT &
SAND
FINE SAND
- If these two planes coincide, pile bearing
capacity is equal to external load, in this case
these planes are called Neutrality Plane as per Figure 2. Construction site and geotechnical profile
many researchers. Negative and positive
friction can be determined when the neutral Figure 3a shows curves of pile load and
plane is in this balancing status. capacity against pile length. Figure 3b displays
pile displacement and soil settlement curves. It
3. VERIFICATION AND CASE STUDY can be recognized that SEP is at 26m depth
(Fig. 3b), whereas FEP is at 22.5m depth (fig.
3.1 Speculative example of down drag 3a). Therefore, even pile bearing capacity is
2.6 times greater than pile load, the load at
This example simulates Inoue et al. (1977)’s case depth of SEP is larger than pile bearing
of a three-stories building in Japan, with a foot capacity what cause building excessive
print of 15  100m, which displaced excessively. displacement.

187
 The Bitexco tower was built in District 1,
HCM city (Fig. 4), with 68-stories. Some of
statistic geotechnical data of soil in District
1, HCM city are listed on Table 1. The
building has 4m thick large-size raft, placing
on a 31 m thick sand layers and a layer of
medium-size sand down to 100 m depth, in
between is a 15m thick layer of hard yellow -
brownish clay (Châu Ngọc Ẩn, 2000; Nguyễn
Figure 3. The estimated result of pile bearing K.H. and Phienwej, 2016). The tower has a
capacity, pile displacement and soil total of 109 piles of 1.5 m diameter, consisted
settlement of 69 piles of 75m-long at raft perimeter and
40 piles of 85m-long at the raft center. The
It can be noted that:
building foundation was simulated and
- The above building is designed as
analyzed by Cao Văn Hóa (2015, 2016) by
conventional pile group method, whereas piles
using PRAB program.
are designed with relatively large safety
The results are displayed in Figure 5. It
coefficient. However it was built on very soft
shows that FEP is at 90m depth (Fig. 5a),
soil layers what hasn’t been finished
whereas the SEP is at 50m, 38m, or 20m in
consolidation yet, the subsoil is under a large
stress of 3.5m of sand fill, as well as urban according to 3 different asumptions, in which
live load. Thus the negative friction develops (Fig 5b):
to the depth of 26m. (1) External load is the full building load
- Negative friction caused by subsoil considering as uniform and acting on
consolidated settlement results the building restangular rigid base, which is distributed in
excessive displacement. In order to prevent subsoil by Boussinesq solution - SEP 3;
building from displacement, it is considered (2) External load is a portion of building
to lengthen piles, in this example more than load caried by raft plus urban live load (in this
3.5m. study we assume it is 10 KN) – SEP 2, or;
(3) External load is only the load carried by
3.2 Verification of pile design for Bitexco Tower raft - SEP 1.

a) b)
Figure 5. The estimated results of pile bearing
capacity, pile displacement and soil
Figure 4. Plan of Bitexco Tower Pile raft settlement

188
Table 1. Statistic soil parameters for District 1, HCMC
Compression tests CI CII
Thick γ1 γ2 φI φII
Soil layers ness e e E e
(m) g/cm3 g/cm3 1KG/cm 2KG/cm 3KG/cm 4KG/cm deg. deg. KG/c KG/c
2 2 2 2

m2 m2
1. Sand fill 3,2 1,81 1,83 0,701 0,645 0,616 0,608 8 10 0,12 0,15
2. Clay 1,6 1,96 1,98 0,613 0,593 0,581 0,578 14 16 0,3 0,32
3.Bluish gray
clay, 2,8 1,92 1,94 0,647 0,608 0,586 0,578 13 15 0,21 0,24
hard
4. Fine sand,
33,4 1,91 1,94 0,672 0,649 0,638 0,630 27 29 - -
medium dense
5. Yellowish
brown clay, 15,5 2,00 2,05 0,526 0,498 0,484 0,479 16 18 0,45 0,5
hard
6. Medium-
size sand, 4,4 2,01 2,06 0,517 0,503 0,496 0,490 31 33 - -
medium dense

It can be noted that: is given. Then, shaft resistance is the only factor
- The Bitexco Tower is designed by using that affect the location of FEP. Considering that,
conventional pile group method. Because piles the curves of negative friction and positive
are installed in stable and consolidated subsoil friction are symetrically; Thus, location of force
of a 300-year-old city, and it is not affected by equality plane is unchanged for a specific pile
the lowering of groundwater. From Fig. 5a the length. Location of FEP determined by this way
blue curve illustrates the load distrubution is conservative, that ensure bearing capacity
along pile calculated by PRAB, one can larger then total load (include external and down-
recognize that after matching of SEP and FEP drag). FEP is normally in lower half of pile.
the load distribution matched well with result - Soil settlement and pile displacement
from PRAB. Thus, the pile capacity is more behaviour are characteristic by SEP. Soil
larger than pile load and the pile is influenced settlement is to estimate as a superposition of
by negative friction insignificantly. the settlements caused by external loads
- If there was no other requirement, the piles (include from shallow foundation, pile
in this building can be shortened greatly. The foundation, embankment surcharge and water
determination of final piles’ length is based on lowering). SEP is passing the point where soil
the matching of FEP and SEP. settlement is equal to pile allowable
displacement, and this point devide a pile into
two halves. Adjusting of pile displacement or
4. CONCLUSIONS
pile length - but keep unchanged pile
Negative friction always exists and it displacement, one can match SEP with FEP.
influences on pile behaviour. Behaviour of soil - Design of piles in consideration of
- pile in consideration of down-drag is related negative friction is to match FEP and SEP. To
on settlement equality plane (SEP) and force do so, the pile can be lengthened or shortened.
equality plane (FEP): Pile is considered capable to carry designed
- Load and bearing capacity behaviour are load if FEP is located below SEP.
characteristic by FEP. The toe resistance should - The magnitude of friction, included both
select conservative because displacement of pile negative and positive friction, can be
caused by pile toe load, and the load at pile head determined when FEP and SEP are coincided.

189
ACKNOWLEDGMENT Poulos, H.G. 1997. Piles Subjective to Negative
Friction: A procedure for design. Jurnal of
In this paper, the authors used PRAB Southeast Asian Geotechnical society, volume
program, a present to Construction 28.
Department of University Architecture of Poulos, H.G. 2001. METHODS OF ANALYSIS
HCM city by Prof. T. Matsumoto. The OF PILED RAFT FOUNDATIONS. A Report
authors of this paper sincerely are thankful Prepared on Behalf of Technical Committee
for Prof Matsumoto’s team. TC18 on Piled Foundations, July.
Thiem Quoc Tuan, Huynh Ngoc Sang & Tran Le
The Dien. 2007. Tìm hiểu cơ chế gây lún mặt
5. REFERENCES
đất do khai thác nước dưới đất ở khu công
Cao Van Hoa, 2015. An analysis on raft deformation nghiệp Hiệp Bình Phước, Quận Thủ Đức,
in piled raft foundation on HoChiMinh city sand Thành phố Hồ Chí Minh, Tạp chí phát triển
by PRAB. Vietnam IBST Journal of Science and khoa học công nghệ, tập 10, số 6.
Construction Technology, No. 2/2015, ISSN Tomlinson, M.J. 1994. Pile design and construction
1859-1566, pp 52-60. practice. E & FN SPON, an imprint of
Fellenius, B. H. 2004. UNIFIED DESIGN OF Chapman & Hall, London.
PILED FOUNDATIONS WITH EMPHASIS
ON SETLEMENT ANALYSIS. Geo-Institute
Geo-TRANS Conference, Los Angeles, July
27-30.
Fellenius, B.H. 1984. Negative skin friction and
settlement of piles. Second International
Seminar, Pile Foundations, Nanyang
Technological Institute, Singapore, 28-30 Nov.
Inoue, Y., Tamaoki, K. & Ogai, T. 1977.
Settlement of building due to pile downdrag.
Proc. 9th ICSMFE, Tokyo, July 10-15, Vol. 1,
pp. 561– 564.
Kitiyodom, P. & Matsumoto, T. 2002. A simplified
analysis method for piled raft and pile group
foundations with batter piles. International
Journal for Numerical and Analytical
Methods in Geomechanics, vol. 1369, pp.
1349–1369.
Kitiyodom, P. and Matsumoto, T. 2003. A
simplified analysis method for piled raft
foundations in non-homogeneous soils.
International Journal for Numerical and
Analytical Methods in Geomechanics, vol. 27,
Issue 2, pp.85-109.
Koerner, R.M. & Mukhopadhyay, C. 1971. Behavior
of negative skin friction on model piles in
medium plasticity silt. Drexel University.
Lee,C.J., Bolton, M.D. & Al-Tabaa, A. 2001.
Recent finding on negative skin friction in
piles and pile group in consolidating ground.
5th International Conference on Deep
Foundation Practice incoporating Pile talk,
Singapore, 4-6/Appril.
Okabe, T. 1977. Large negative skin friction and
friction-free methods. Proc. 9th ICSMFE,
Tokyo, Vol.1, pp. 679 – 683.

190
The 3rd Int. Conf. on Transport Infrastructure & Sustainable Development (TISDIC 2019). Construction Publishing House. ISBN: 978-604-82-2893-4.

A review on the application of deep cement mixing under

lateral loads

Pham Van-Ngoc 1,3*, Erwin Oh2, Dominic Ong2, and Do Huu-Dao 1

1
Lecturer, University of Science and Technology, The University of Danang, Vietnam
2
Lecturer, School of Engineering and Built Environment, Griffith University, Australia
3
PhD candidate at School of Engineering and Built Environment, Griffith University, Australia

*Corresponding author, E-mail: [email protected].

ABSTRACT: Deep cement mixing (DCM) is one of the geotechnical solutions to protect riverbanks
or to prevent erosion along beaches. The horizontal bearing capacity of the DCM is insufficient to
resist high lateral loads which may lead to large horizontal displacement or even failures of the
structure. Therefore, the lateral resistance of the DCM material is the main challenge which needs to
be studied in depth. This paper is going to summarise previous studies on the lateral response of
DCM to improve embankment and to strengthen retaining walls. In addition, lateral resistance
improvement methods, such as using steel beams, concrete piles, T-shape columns, fibres, anchors,
and PVDs, are presented based on previous research. This review is significant for the design of
DCM structures under high horizontal loads.
Keywords: Deep cement mixing, lateral resistance, embankment, retaining walls, improvement methods.

ground improvement and foundations. DMM

1. INTRODUCTION includes dry and wet mixing (jet grouting),
Deep mixing method (DMM) is an in-situ which requires more cement content than the
soil treatment method in which native soils or dry mixing method (Rollins et al., 2010);
fills are blended with cementitious or other however, it provides higher average strength
stabilising agents (binder) (Bruce, 2000). (Rollins et al., 2010).
Materials such as cement, lime, slag and fly The advantages of DMM include
ash, are commonly used as stabilising agents economics, flexibility, savings of materials and
(Bruce, 2000). This method was developed in energy by exploiting the properties of the soil
1967 to increase the stability of foundations at the site, producing only small vibrations and
(in Japan), reduce settlement and increase the low construction noise with low costs and
stability of embankments (in Sweden) rapid installation (Bruce, 2000; Chen et al.,
(Bruce, 2000; Indraratna et al., 2015; 2013; Holm, 2003; Indraratna et al., 2015).
Kitazume & Terashi, 2013; Moseley & However, it has a limitation in terms of column
Kirsch, 2004). Since the 1980s, DMM has length, final bearing capacity and lateral
been applied widely in Japan, Europe, China, resistance (Dehghanbanadaki et al., 2013).
Singapore, Hong Kong and America with DMM can be applied as the foundation of
more than 100 million m3 of soil stabilised in various structures such as earth retaining wall
over one thousand projects (Bruce, 2000; structures, foundations, soil reinforcement,
Holm, 2003; Kitazume & Terashi, 2013). In land levees and floodwalls stabilization,
Vietnam, DMM has been used recently in remediation, barrier against liquefaction,

191
embankments, excavation walls, reinforcement
piles, breakwaters and high-rise buildings
(Bruce, 2000; Denies & Van-Lysebetten, 2012;
Porbaha, 1998; Terashi, 2005). For some
structures, DCM needs to resist the lateral
loads from earth pressure, water and tide
pressure, landslides, or earthquake loads. By
applying DCM, the lateral resistance of the
embankment structures could be increased
from 60% to 260% (Rollins et al., 2010).
However, that horizontal bearing capacity of
DCM is insufficient to resist high lateral loads
which may lead to large horizontal
displacement or even failure of the structure. Figure 1. Landslide at the riverbank in Vietnam
To support the embankment or retaining
structures, the lateral resistance of DCM could 2. RECENT STUDIES
be enhanced by using steel beams, concrete 2.1 Embankments
piles, T-shape columns, PVDs or fibres (Chang
et al., 2009). This paper presents previous A popular application of DCM materials is to
studies on the lateral response of deep cement support embankments, as embankment
mixing material. construction on soft soil could be subject to
Recently, global warming and river/beach tensile failure, shear failure, bending moment
erosion have become significant concerns in or side slope failure (Filz & Navin, 2006). A
many countries. In Vietnam, such erosion typical section of embankment reinforced by a
situation has occurred seriously in the centre DCM column is shown in Figure
area of the country such as at Hoi An beach,
My Khe beach and also along the riverbanks in
the North and South areas as presented in
Figure 1 (DDMFC, 2004). Such erosion had
led to the disappearance of 2500 km of beach
and riverbanks in Vietnam (Ho et al., 2016), as
well as the collapse of infrastructure and
thousands of hectares of land (DDMFC, 2004).
The situation is even more of a threat because
these areas are where more than 50% of the
total population resides (Woodroffe, 2002), 2.
and consequently, the majority of buildings
Figure 2. Embankment supported by deep cement
and transport infrastructure are concentrated in
mixing column (Filz & Navin, 2006)
these areas. Therefore, protecting riverbanks
and beaches are a priority policy in many In addition, in some cases, the embankments
provinces in Vietnam. and pavements need to be widened to increase
In summary, the lateral response of DCM is transportation capacity which then could lead
a critical aspect of the geotechnical application. to a differential settlement or slope failure due
Therefore, its lateral response ability when to soft ground. Han et al. (2007) showed that
applied in Vietnam needs to be reviewed the soil-cement column (SCC) reduce not only
carefully. It then could be applied widely in total settlement and differential settlement of
practice to prevent soil erosion or improve the widening embankment, but also reduce lateral
bearing capacity of embankments. displacement of the system (Figure 3).

192
Figure 3. Example of widened embankment supported
by the soil-cement column (Han et al., 2007)

Furthermore, physical models, centrifugal

Figure 4. The layout of a composite foundation
tests and numerical models have also been with SCC and PVDs (Ye et al., 2015)
applied to analyse the effect of DCM
columns to stabilise embankments. One of To improve the ratio of area replacement, a
the technical aspects of an embankment type of SCC called a T-shape deep mixing
supported by a deep mixing column is lateral (TDM) column has been applied in the field, in
resistance. For DCM materials, compressive which the portion near the surface is opened its
strength is higher than tensile strength, which diameter, and the deeper part of the DCM
means that the system could collapse in the column shows a smaller diameter (Yi et al.,
case of high lateral loads affecting the DCM 2012) (Figure 5). Using TDM could decrease
column. There are several methods to 67% total settlement of ground, reduce 6.5% of
improve the lateral response of the DCM cement and 19% of construction time, as well
column, such as using prefabricated vertical as decrease 31% lateral movement (Yi et al.,
drains (PVD), T-shape columns, steel beams, 2015; Yi et al., 2012; Yi et al., 2018). In terms
and concrete piles. of lateral displacement of the ground, the
Case histories indicate that settlement of maximum lateral movement value of the TDM
surrounding untreated soil is always higher column is smaller and shallower than
than soil treated by DCM under embankment conventional SCC (Yi et al., 2012). Moreover,
loading conditions (Bergado and Lorenzo the lateral displacement ratio at the
2002). The difference in settlement between embankment centerline of the TDM column is
treated and untreated soils can be as high as 8 smaller and its enhancement of the stability of
to 20% of the average settlement at the ground the embankment is better than usual SCC (Liu
surface (Bergado et al. 2002). et al., 2011). These researchers have conducted
In 2015, Ye et al. (2015) presented a laboratory tests, numerical models by FLAC
centrifugal model result on an embankment on 3D, and field tests to make the conclusion that
a composite foundation combined with SCC the TDM column provides a viable, economic
and prefabricated vertical drains (PVDs) as solution and bring more benefits than
presented in Figure 4. As a result, the lateral conventional SCC (Liu et al., 2011).
displacement decreases when the length of
SCC is increased, which then leads to improve
the stability of the embankment. Besides, SCC
transfers the loads to the lower layer and PVDs
can accelerate the consolidation of the soil
layer below the SCC system (Ye et al., 2015).
Moreover, the composite foundation increases
the shear strength of ground and consolidation
rate, as well as decreases pore water pressure
and settlement over time (Bergado et al., 2018; Figure 5. An illustration of the TDM column
Ye et al., 2012). supporting the embankment (Yi et al., 2012)

193
 Another solution to improve the lateral So, it could be said that SDCM has a bearing
resistance of the DCM system is to use capacity and lateral resistance much higher
concrete piles or steel beams to insert into the than conventional DCM and concrete pile, and
centre of fresh soil mixing material. It is it is suitable for reinforcing the embankment
known as stiffened deep cement mixing on soft clay (Li et al., 2018).
(SDCM) or deep cement mixing bored pile In the research of Wonglert and Jongpradist
(DCM-BP) (Li et al., 2018; Voottipruex et al., (2015), they present the principle of designing
2011a). SDCM is a composite structure of process, choosing the length and diameter of the
concrete pile or steel beam and DCM column, core pile, and simulation method by PLAXIS
which then works together by transferring the 3D. In which, the clay layer was modelled as the
loads to the surrounding soil. Figure 6 shows Hardening soil model, soft clay and medium
the example of the SDCM using square stiffness clay were models as the Soft-soil model,
prestressed concrete pile. The composite and the DCM material and sand layer were
structure can resist the shear forces and assumed to behave as elastic-perfectly plastic
material as the Mohr-Coulomb model. They
bending moments from the embankment loads.
found that by increasing the core length, the
Hence, it prevents pile failure, soil failure, and
bearing capacity of the composite foundation is
increases the stability and bearing capacity of
increased, and the total settlement of the ground
the embankment (Bergado et al., 2018).
is decreased (Wonglert & Jongpradist, 2015).
However, the diameter of concrete core pile
significantly affects the bending capacity of the
SDCM pile and lateral movement of the
embankment (Bergado et al., 2018; Voottipruex
et al., 2011b; Wonglert & Jongpradist, 2015). In
addition, they found that increasing the length of
the concrete pile does not increase the lateral
load capacity of the SDCM (Voottipruex et al.,
2011b). Furthermore, the SDCM piles increase
the skin friction of the concrete piles and long-
term strength of the structure, which then
enhances the surrounding lateral resistance and
the performance of the core concrete piles (Li et
Figure 6. Details of the SDCM pile (Voottipruex al., 2018).
et al., 2011b) In terms of durability of SDCM, Denies et
al. (2015a) shown that the degradation of DCM
In 2013, Raongjant and Jing (2013)
material with time depends on wet-dry and
conducted a research on SDCM under lateral
freeze-thaw cycles, carbonation, diffusion of
cyclic loading and they found that the SDCM
calcium, sulphate and chlorides attack. Hence,
with the H-steel beam core increases lateral
using concrete core piles in the coastal areas
resistance capacity 3-4 times. Similar results
could avoid the impact of corrosion into fresh
were shown in the research of Voottipruex et
soil mix material (Denies et al., 2015a).
al. (2011b), Jamsawang et al. (2008), Wonglert
and Jongpradist (2015). They conducted a 2.2 Retaining walls, levees and flood walls
simulation model by PLAXIS 3D software and
compared simulated results to observe data to For the retaining wall (excavation) applications,
make the conclusions that the SDCM piles the DCM wall has been applied popularly as it is
reinforced by concrete core pile increase the more advantageous in an urban area with high
bearing capacity by 2.2 times and lateral ground-water level, low vibration and noise
capacity by 15 times higher than the typical (Rutherford et al., 2007). Some journal articles
DCM piles. Besides, SDCM resists lateral present the advantages and limitations of
loads 5.4 times higher than concrete pile only. excavation supports and the guideline for

194
designing and constructing the DCM walls the effects of lateral earth pressure, erosion from
(Denies et al., 2012; Rutherford et al., 2007). river flow or sulphate attack.
Furthermore, DCM has also been applied to
reinforce an earth levee and floodwall to protect
rice fields and infrastructure (Kitazume et al.,
2015b) as well as prevent erosion along
riverbanks and beaches.
For these structures, lateral earth pressure
and external loads may create collapse or
damage of structure as the DCM materials only
have low bending stiffness. Therefore, some
solutions have been suggested such as using
steel beams (Denies & Huybrechts, 2017;
Denies et al., 2015b; Raongjant & Jing, 2013;
Rutherford et al., 2007), concrete piles
(Voottipruex et al., 2011b; Yin & Fang, 2010),
fibre steel or polymer (Sukontasukkul &
Jamsawang, 2012), to increase bending
stiffness in order to resist shear and bending
moment. Besides, physical models, large-scale
models and numerical models have also been
applied to investigate the principle working Figure 7. Example of DCM reinforce the levee
process of these composite foundations. (Filz et al., 2012)
DCM has been applied in some levee
reinforcement projects as shown in Figure 7. In
America, DCM was used to create a stable
foundation, increase settlement and lateral load
resistance (Schmutzler et al., 2013). In Vietnam,
Kitazume et al. (2015a) had applied the soil-
cement mixing walls to reinforce earth levees in
Dong Thap province. It helped to prevent the
levee’s piping, sliding and seepage. DCM has also
been applied to build an earth levee using a small Figure 8. The reinforced earth levee using DCM
lightweight equipment system (Tran-Nguyen et (Tran-Nguyen et al., 2018)
al., 2016). Recently, Tran-Nguyen et al. (2018)
presented their research project on using DCM to Steel and polypropylene fibres have also
reinforce the levee in Vietnam. By installing one been used to improve the flexural performance
and two rows DCM at the core of the earth levee of the soil-cement column to resist horizontal
with the cement contents from 250-300 kg/m3 loads (Sukontasukkul & Jamsawang, 2012). In
and observing the lateral displacement (Figure 8), the research, they had added the fibres with a
they concluded that the lateral displacement of the different volume fraction of 0.5, 0.75 and 1%.
system in the dry season was more significant; As a result, the fibres had incorporated into the
however, the levees were still stable. Also, in the mix, created a bridge to restrain cracks in the
flood season, the DCM row can cut off seepage material and increased flexural strength.
and keep the levee stable. The same design Besides, they also found that polypropylene
process is shown in the research of Filz et al. performs better than steel fibres.
(2012). In these studies, the authors determine the The deep cement mixing column reinforced
strength and uniformness of the reinforced earth by steel beam (I or H shape) and concrete pile
levee (DCM material), but they do not consider are popular in excavation application. Much

195
research has been performed on that method then they found that the bending resistance
(Denies & Huybrechts, 2017; Meepon et al., increases up to 126% in comparison to the use
2016; Peixoto et al., 2012a, 2012b; Rutherford et of a steel beam only. Therefore, it could save
al., 2007; Wonglert & Jongpradist, 2015). Steel 38% of steel used in the structure.
beams and concrete piles have been inserted into The concrete pile is also a choice for core
fresh DCM material to resist the bending reinforcement of DCM. Research showed that
moments and shear forces from earth, water the bearing capacity of SDCM depends on core
pressure and other loads (Denies & Huybrechts, dimension and core volume ratio (Wonglert &
2017; Rutherford et al., 2007). Furthermore, the Jongpradist, 2015). With a length of the concrete
H-shape steel beam has been applied by core around 0.85 lengths of DCM column, the
Rutherford et al. (2007) to restrain bending and bearing capacity of the SDCM column could be
shear failure and keep the excavation stable. The 15 times higher than the conventional DCM
bending stiffness of DCM material is around 2- column. Voottipruex et al. (2011b) simulated
4% of steel beam (Rutherford et al., 2007), so SDCM by PLAXIS 3D to observe its failure
when working together, the lateral forces mode and concluded that using concrete pile is
transmitted to steel beams follow an arching better than an I or H steel beam.
effect principle as presented in Figure 9. They Another design of retaining structure in
also provide the design process with a choice excavations is to combine reinforced concrete
regarding the dimension and distance between piles with the DCM columns, in which the
the steel beams. DCM columns are arranged in a continuous
In practice, an IPE 450 steel beam was used arch and concrete piles are placed at the two
in the projects at Cannes, France (Peixoto et al., toes of the arch as shown in Figure 10 (Shao et
2012a, 2012b). The steel beam in combination al., 2005). The lateral earth pressure will
with soil cement material can resist the bending transmit to the toe of the SCC arch and the
moment by earth pressure, hydrostatic pressure concrete piles are used to resist those forces.
and overloads on the surface.
By laboratory investigation and full-scale
test of the DCM and SDCM walls, Meepon et
al. (2016) observed the behaviour of five types
of walls. They found that one row of DCM
column reinforced by H-shape steel beam at
each column could resist bending moment due
to the horizontal force with a linear relationship
with horizontal displacement. So, it could
replace two or three rows of DCM column Figure 10. The compound DCM column and
without reinforcement. concrete pile (Shao et al., 2005)

3. CONCLUSION
This paper reviews the lateral response of the
DCM column applied for embankment
reinforcement and retaining walls under
horizontal load and bending moment. It is clear
that the DCM column improves the stability of
Figure 9. Arch effect of DCM reinforced by the the embankment and retaining walls and
steel beam (Rutherford et al., 2007) decreases the lateral displacement of the ground.
However, under high lateral loads, the structures
In 2017, Denies and Huybrechts (2017) need to be strengthened by combination with the
conducted a large scale test to evaluate the concrete piles, steel beams, PVDs, steel or
combination of DCM material and steel beam; polypropylene fibres. In addition, using the H-

196
shape DCM column or combining the concrete Nicolas Denies & Gust Van-Lysebetten. 2012.
pile and the DCM column arranged in the General Report–Session 4–SOIL MIXING 2–
continuous arch can increase bending stiffness DEEP MIXING. Proceedings of the the
International Symposium of ISSMGE-TC211.
and decrease the lateral displacement of the
Recent research, advances & execution
structures. Moreover, the DCM may be used to aspects of ground improvement works,
reinforce levees or floodwalls. Within the ground Brussels, Belgium. (www. bbri. be/go/tc211)
condition at Danang area, Vietnam, the DCM Nicolas Denies, André Vervoort & Noël
reinforced by concrete pile could be applied to Huybrechts. 2012. Soil mix walls as retaining
build a wall for preventing erosion and slope structures, Belgian practice. Proceedings of
collapse of riverbanks and beaches. the TC211 International Symposium on
Ground Improvement IS-GI.
George Filz, Tiffany Adams, Michael Navin & AE
4. REFERENCES Templeton. 2012. Design of deep mixing for
DT Bergado, PV Long, S Chaiyaput & AS support of levees and floodwalls. In Grouting
Balasubramaniam. 2018. Prefabricated and Deep Mixing 2012, 89-133.
George M Filz & Michael P Navin. (2006).
Vertical Drain (PVD) and Deep Cement
Stability of column-supported embankments.
Mixing (DCM)/Stiffened DCM (SDCM)
Retrieved from Virginia Center for
techniques for soft ground improvement. Transportation Innovation and Research.
Proceedings of the IOP Conference Series: Jie Han, Sadik Oztoprak, Robert L Parsons & Jie
Earth and Environmental Science. Huang. 2007. Numerical analysis of
Donald A Bruce. 2000. An introduction to the deep foundation columns to support widening of
soil mixing methods as used in geotechnical embankments. Computers and Geotechnics,
applications. 34(6), 435-448.
Der‐Wen Chang, Bor‐Shiun Lin & Shih‐Hao Viet-Cuong; Ho, Ngoc-Quynh; Nguyen & Huy-
Cheng. 2009. Lateral load distributions on Hieu Bui. 2016. Evaluation of the status of
grouped piles from dynamic pile‐to‐pile stability and efficiency solutions, river
interaction factors. International Journal for protection, seaweed protection technology in
Numerical and Analytical Methods in Vietnam. Journal of Science and technology in
Geomechanics, 33(2), 173-191. Vietnam, 34, 1-12.
Jin Jian Chen, Lianyang Zhang, Jun Feng Zhang, Goran Holm. 2003. State of practice in dry deep
Yan Fei Zhu & Jian Hua Wang. 2013. Field mixing methods. Proceedings of the Grouting
Tests, Modification, and Application of Deep and Ground Treatment.
Soil Mixing Method in Soft Clay. Journal of Buddhima Indraratna, Jian Chu & Cholachat
Geotechnical and Geoenvironmental Rujikiatkamjorn. 2015. Ground Improvement
engineering, 24-34.
Case Histories: Chemical, Electrokinetic,
DDMFC. 2004. Bank Erosion in Mekong Delta
Thermal and bioengineering methods:
and along Red River in Vietnam. Retrieved
Elsevier.
from Holland.
P Jamsawang, D Bergado, A Bandari & P
Ali Dehghanbanadaki, Kamarudin Ahmad, Nazri Ali,
Mahdy Khari, Payman Alimohammadi & Nima Voottipruex. 2008. Investigation and
Latifi. 2013. Stabilization of soft soils with deep simulation of behavior of stiffened deep
mixed soil columns. EJGE, 18, 295-306. cement mixing (SDCM) piles. International
Nicolas Denies, N Huybrechts, F De Cock, B Journal of Geotechnical Engineering, 2(3),
Lameire, J Maertens, A Vervoort & A 229-246.
Guimond-Barrett. 2015a. Thoughts on the Masaki Kitazume, Tanaka Hitoshi & Toru
durability of the soil mix material. Kobayshi. 2015a. Effects of deep mixing
Nicolas Denies & Noël Huybrechts. 2017. Deep method to reinforce earth levees in the
mixing method for the construction of earth Mekong Delta, Vietnam.
and water retaining walls. RILEM Technical Masaki Kitazume, Hitoshi Tanaka, Bao K Le,
Letters, 2, 1-9. Long P Le, Chinh TM Ho, Tien B Nguyen,
Nicolas Denies, Noel Huybrechts, Jan Maertens & Chau D Truong & Phuong A Mai. 2015b.
André Vervoort. 2015b. Design of soil mix Laboratory investigation of soilcrete created
walls for earth and water retaining structures. from Mekong Delta’s soils mixed with
Proceedings of the Deep Mixing. cement.

197
Masaki Kitazume & Masaaki Terashi. 2013. The 2007. Design process of deep soil mixed walls
deep mixing method: CRC Press. for excavation support. ISSMGE International
Hongjiang Li, Songyu Liu & Liyuan Tong. 2018. Journal of Geoengineering Case Histories,
Field investigation of the performance of 1(2), 56-72.
composite foundations reinforced by DCM- Wesley Schmutzler, Filippo Maria Leoni & Kyle
bored piles under lateral loads. Construction Sansing. 2013. Levee Foundation Remediation
and Building Materials, 170, 690-697. Using the Deep Mix Method.
Song-Yu Liu, Yan-Jun Du, Yao-Lin Yi & Anand J Yong Shao, Emir Jose Macari & Weiming Cai.
Puppala. 2011. Field investigations on 2005. Compound deep soil mixing columns
performance of T-shaped deep mixed soil cement for retaining structures in excavations. Journal
column–supported embankments over soft of Geotechnical and Geoenvironmental
ground. Journal of Geotechnical and engineering, 131(11), 1370-1377.
Geoenvironmental engineering, 138(6), 718-727. Piti Sukontasukkul & Pitthaya Jamsawang. 2012.
I Meepon, P Voottipruex & P Jamsawang. 2016. Use of steel and polypropylene fibers to
Behaviors of soil cement columns and stiffened improve flexural performance of deep soil–
soil cement column wall in shallow excavation. cement column. Construction and Building
Lowland technology international: the official Materials, 29, 201-205.
journal of the International Association of Terashi. 2005. Keynote lecture: design of deep
Lowland Technology, 18(3), 197-208. mixing in infrastructure applications.
Michael P Moseley & Klaus Kirsch. 2004. Ground Proceedings of the Deep Mixing'05:
International conference on deep mixing best
improvement: CRC Press.
practice and recent advances.
Artur Peixoto, Tecnologia de Fundações Geo-
Hoang-Hung Tran-Nguyen, Bao Khac Le & Long
Rumo, Estela Sousa & Pedro Gomes. 2012a.
Phi Le. 2016. Behaviors of field soilcrete
Earth Retaining Structure using Cutter Soil
created by a light weight equipment system to
Mixing technology for the “Villa Paradisio” reinforce an earth levee in Vietnam.
Project at Cannes, France. Proceedings of the Malaysian Journal of Civil Engineering, 28(1).
International symposium of ISSMGE-TC211. Hoang-Hung Tran-Nguyen, Long Phi Le, Bao
Recent research, advances & execution Khac Le & Hien Minh Thi Ly. 2018. Field
aspects of ground improvement works. 31 trials on the soil cement mixing technology to
May-1 June 2012, Brussels, Belgium. reinforce earth levees in the Mekong Delta,
Artur Peixoto, Tecnologia de Fundações Geo-Rumo, Vietnam. Asean Engineering Journal,
Estela Sousa & Pedro Gomes. 2012b. Solution 13(6.530), 18.480.
of earth retaining structure using Cutter Soil Voottipruex, DT Bergado, T Suksawat, P
Mixing technology:“Parking Saint Nicolas” Jamsawang & W Cheang. 2011a. Behavior
Project at Cannes, France. Proceedings of the and simulation of deep cement mixing (DCM)
International symposium of ISSMGE-TC211. and stiffened deep cement mixing (SDCM)
Recent research, advances & execution aspects piles under full scale loading. Soils and
of ground improvement works. 31 May-1 June foundations, 51(2), 307-320.
2012, Brussels, Belgium. Voottipruex, Taweephong Suksawat, DT Bergado
A Porbaha. 1998. State of the art in deep mixing & Pitthaya Jamsawang. 2011b. Numerical
technology: part I. Basic concepts and simulations and parametric study of SDCM
overview. Proceedings of the ICE-Ground and DCM piles under full scale axial and
Improvement, 2(2), 81-92. lateral loads. Computers and Geotechnics,
Werasak Raongjant & Meng Jing. 2013. Field 38(3), 318-329.
testing of stiffened deep cement mixing piles Anucha Wonglert & Pornkasem Jongpradist. 2015.
under lateral cyclic loading. Earthquake Impact of reinforced core on performance and
Engineering and Engineering Vibration, 12(2), failure behavior of stiffened deep cement mixing
261-265. piles. Computers and Geotechnics, 69, 93-104.
Kyle M Rollins, Mark Herbst, Matthew Adsero & Colin D Woodroffe. 2002. Coasts: form, process
Dan Brown. 2010. Jet grouting and soil and evolution: Cambridge University Press.
mixing for increased lateral pile group Guan Bao Ye, Zhen Zhang, Hao Feng Xing, Mao
resistance. In GeoFlorida 2010: Advances in Song Huang & Chao Xu. 2012. Consolidation
Analysis, Modeling & Design, 1563-1572. of a composite foundation with soil–cement
Cassandra J Rutherford, Giovanna Biscontin, columns and prefabricated vertical drains.
Demetrious Koutsoftas & Jean-Louis Briaud.

198
 Bulletin of Engineering Geology and the
Environment, 71(1), 87-98.
Guanbao Ye, Qingwen Zhang, Zhen Zhang &
Hongtao Chang. 2015. Centrifugal modeling
of a composite foundation combined with
soil–cement columns and prefabricated
vertical drains. Soils and foundations, 55(5),
1259-1269.
Y Yi, S Liu & AJ Puppala. 2015. Laboratory
modelling of T-shaped soil–cement column
for soft ground treatment under embankment.
Geotechnique, 66(1), 85-89.
Yaolin Yi, Songyu Liu, Yanjun Du, Zhiduo Zhu &
Guangyin Du. 2012. The T-shaped deep
mixed column application in soft ground
improvement. In Grouting and Deep Mixing
2012 (pp. 389-399).
Yoalin Yi, Pengpeng Ni & Songyu Liu. 2018.
Numerical simulation of T-shaped soil-cement
column supported embankment over soft
ground. Proceedings of the China-Europe
conference on geotechnical engineering, SSGG.
Jian-Hua Yin & Zhen Fang. 2010. Physical
modeling of a footing on soft soil ground with
deep cement mixed soil columns under
vertical loading. Marine Georesources and
Geotechnology, 28(2), 173-188.

199
The 3rd Int. Conf. on Transport Infrastructure & Sustainable Development (TISDIC 2019). Construction Publishing House. ISBN: 978-604-82-2893-4.

Study on load distribution between raft and piled in piled

raft foundation in two layers soil

Vo Van Dau1,, Vo Phan2

1
Faculty of Civil Engineering, Ho Chi Minh City University of Technology, Ho Chi Minh city, Vietnam
1
Department of Transportation Engineering, College of Engineering Technology, Cantho University, Vietnam
2
Faculty of Civil Engineering, Ho Chi Minh City University of Technology, Vietnam

Corresponding author, E-mail: [email protected]

ABSTRACT: The paper focuses on load distribution of the raft and piles in pile-raft foundation system
based on the number of piles in raft foundations according to the Poulos - Davis - Randolph method
(PDR) and Finite Element Method (FEM). Plaxis 3D Foundation was employed as a tool for Finite
Element Method in this study. Methods of analyzing and calculating the load sharing between the raft and
axial loads based on the soil parameters from site investigation. The results determined the load-shared
between the raft and the piles in pile-raft foundation in cases of surveying on 16, 20 and 24 piles,
respectively, under 2 layers of the soil.
Keywords: Piled rafts, load distribution, load sharing.

1. INTRODUCTION when the pile raft design will lead to inaccurate

description of how to distribute the load on the
Today, a majority of the population increase in
raft and the piles of the raft foundation.
huge cities, strong economic growth, The
According to PDR (Poulos, Davis and
number of people moving from rural to urban
Randoph) provide a method for calculating the
areas, continues to attract much interest,
load distribution of rafts and piles under the
Population pressure requires cities to have
effect of building loads. The research results
housing and infrastructure development.
show that the load between raft and pile in pile
Therefore, commercial centers or high-end
group of raft foundation with survey cases: 16
apartments are mainly multi-storey houses and
piles, 20 piles and 24 piles in 2 layer soil
basements. So, piles raft are applied as an
conditions.
optimal solution in terms of structure as well as
construction costs. 2. PDR METHOD FOR CALCULATION
In pile raft design, designers often conceive OF SETTLEMENT
of the entire load of the work received by the
piles, the contribution of the raft is often 2.1 Vertical load capacity
ignored, even if the bottom of the raft is in For the raft
contact with the ground, this is the design The ultimate bearing capacity is calculated
perspective very prudent, because the raft is by equation (1). Table 1 shows the
actually transmits a part of the load to the interpretation of parameters in equation (1).
q  cNc Fcs F Fci  qN q Fqs Fqi  0.5 BN F s F F i
ground. Therefore, if the raft work is omitted ult cd d
(1)

200
Table 1. Shape, depth, inclination factors S
Note: typical pile spacing 3   3.5
recommended for use D
S center-to-center pile spacing
Factor Relationship Source
BNq B B
D pile diameter
Fcs 1 , Fqs 1 tan, Fs 1 0.4
LNc L L De beer For the foundation
Shape
pr  q
Qult  n.Qult
ult
Where L=length of the (1970) pile (6)
foundation (L>B) where: n- total number piles
Condition (a): D  1 f

B 2.2 Load – settlement behavior

Df
Fcd 1 0.4 ,
B Calculation for raft stiffness (elastic or initial):
Fqd 1 2 tan(1sin) 2
Df
, Fd  1  .a.Es (7)
B K ri 
I
Depth Condition (b): Df
1
B where: a= foundation radius a  ( BxL) / 
 Df 
Fcd 1 0.4tan1 , Es=Young’s modulus soil
B
Ip= influenced factor for vertical displacement
 Df 
Fqd 1 2tan(1sin)2 tan1 ,  = Poisson’s ratio;
B
a
Fd 1 h: thickness of soil layer . ,   Ip
h
Meyerho
2 2
 0   0 
 90 
f (1963);
Fci  Fqi  1  0  , Fi  1  0 
  
Inclinat 0 Hanna
Where  =inclination of
ion and
the load on the foundation
Meyerho
with respect to the vertical
f (1981)
These shape factors are empirical relations
based on extensive laboratory tests

The total bearing capacity of the raft is

calculated by equation (2). In that equation, A
is the area of the raft.
ult
Qraf  qult xA (2)
For the piles Figure 1. Influence factors for vertical displacement
pile  Q S  Q P  W pile
Q ult (3) of rigid circle (Poulos and Davis, 1974) [2]
n n
QS   Qsi   f si Asi (4)
i 1 i 1 Calculation for single pile: (undrained case)
QP  qE AE (5)  4 2 tanh  l l 

 (1  )   l r0  (8)
Where: K1  Gr r0 
1 4 tanh l l 
 1 
Qult ultimate load   (1  ) l r0 
QS shaft resistance Eu d r
QP end bearing G ; r0  ,  0 (under-reamed piles)
2(v ' 1) 2 rb
WPile weight of pile
Gl
fSi ultimate skin friction, layer i, = (end-bearing piles)
 K o vo' tan   (1  sin  ) vo' tan  Gb
 vo' vertical effective stress (at mind layer)
Stiffness of group piles
kpi=k1 n p (9)
 friction angle
Asi perimeter area of pile, layer i Calculation stiffness of the pile - raft
qE ultimate end bearing pressure foundation:
AE cross section area of pile Kpri=X kpi (10)

201
 1  0.6( K ri / K pi ) and: f T e (Plasticity) (18)
X (11) D   0
1  0.64( K ri / K pi )  
0.2 g is the plastic potential function which is

1  0.8( K r / K p )
 Kr / K p  (12)
introduced to fix the problem of theory of
1 associated plasticity in estimating dilatancy.
p  (13)
(1   ) Non-associated plasticity is denoted as g ≠ f.
Therefore, the relationship between effective
3. FINITE ELEMENT ANALYSES stress rates and strain rates can be expressed as

    D e 

De
g f T e 
D   (19)
3.1 Model formulations  d     
In which α = 0 (elasticity) and α = 1 (plasticity)
Linear Elastic Model f T e g
d D (20)
This is the simplest model used for materials,    
which is based on the Hooke’s law for isotropic For multi surface yield contour, the above
linear elastic behavior. The relationship between equations should be extended as:
effective stress and strain is expressed in term of g1 g g (21)
 p  1  2 2  3 3  ...
the rate as below:      
   D e (14) Where li (i = 1, 2, 3,…) can be defined from
e
Where D is the elastic material stiffness the yield functions fi (i = 1, 2, 3,…), respectively.
matrix. Effective Young’s modulus E and The yield condition used in Mohr-Coulomb
effective Poisson’s ratio  are used in this model, model is an extension of Coulomb’s friction
which are attached in De matrix. Linear elastic law to general states of stress. In principle
model is inappropriate to model behavior of the stress space, this condition consists of six yield
soils which have highly non-linear behavior. functions as below:
1 1 (22)
This model is suitable to simulate behavior of f1a   2   3    2   3  sin   c cos   0
2 2
structures (e.g. piles, raft or walls) where the 1 1
strength properties of materials are very high f1b   3 2   3   2  sin   c cos   0
      (23)
2 2
compared with those of soil. In Plaxis, this model 1 1
f 2a   3  1    3  1  sin   c cos   0 (24)
is usually used together with Non-porous type of 2 2
material behavior to exclude pore pressures from 1 1
f 2b   1   3    1   3  sin   c cos   0 (25)
these structural elements. 2 2
Mohr - Coulomb Model 1 1
f3a   1   2    1   2  sin   c cos   0 (26)
The Mohr-Coulomb model is an elastic 2 2
perfectly plastic model which is a constitutive 1 1
f3b   2   1    2   1  sin   c cos   0 (27)
model with a fixed yield surface and the 2 2
behavior of points within the yield surface is Where , c are the friction angle and
purely elastic. Based on the basic principal of cohesion of the soil respectively. The condition
elastoplasticity, equation (3.1) can be written as: fi = 0 for all yield functions together give a
   D e     p  (15) hexagonal cone as shown in Fig.2
Where  p is the plastic strain rate component
which is defined by:
g
 p   (16)
 
Where l is the plastic multiplier which is
defined from the yield function, f, as below:
  0 for: f  0
or:  f D e  0 (Elasticity)
T
(17) Figure 2. The Mohr-Coulomb yield surface in
  principal stress space for c = 0
  0 for: f  0 (Brinkgreve et al. (2007)).

202
 The plastic potential functions of Mohr- For analyses of model piled rafts, a
Coulomb model are defined as below: comparative study will be performed for 3
1 1 cases of model of piled raft on two sands. The
g1a   2   3    2   3  sin (28)
2 2 models of 4×4 piled raft, 4×5 piled raft and
1 1
g1b   3   2    3   2  sin (29) 4×6 piled raft are considered in the study.
2 2 For piled raft in Mekong Delta sand,
1 1
g 2 a   3   1    3   1  sin (30) considering a 19m × 25m rectangular raft with
2 2 a thickness of 1 m, it will be supported by 16
1 1
g2b   1   3    1   3  sin (31) piles, 20 piles and 24 piles, respectively. The
2 2 depth of raft is 1 m, and the water table at the
1 1 ground surface. The length of piles is 25m, and
g 3a   1   2    1   2  sin (32)
2 2 the piled spacing is 3m. Based on the works of
1 1 Poulos (1991)
g3b   2   1    2  1  sin (33)
2 2
Where  is the dilatancy angle of the soil. 4. PRACTICE CALCULATION
Hence, there are five parameters including c, 
and  for plasticity and E and  for elasticity 4.1 PDR method for piled raft foundation
are required for Mohr-Coulomb model. Vertical capacity of raft in sand is calculated as
q ult  cN c Fcs Fcd Fci  qN q Fqs Fqi  0.5BN  Fs Fd Fi
3.2 Model of piled raft and sand
Where
The behavior of structures (piles and raft) is
modeled by Linear Elastic Model with Young  Nc  48.58
modulus and passion ratio as shown in Table 
 I
sand  35.5   Nq  35.75
0

2.1. The properties of the structural materials 

used in this study are the same for all cases.  N  30.2
Mohr-Coulomb Model is used to simulate the BN q
behavior of the soils. The material type used Fcs  1   1.399
LN c
is drained.
B
Fqs  1  tan   1.29
Table 2.1 Properties of structural materials L
used in the analyses B
Fs  1  0.4  0.783
L
Description Young’s Poisson Unit
modulus ratio weight Df 1
  0.052  1
E (kN/m2)  (kN/m3) B 19
6 Df
Pile 20×10 0.2 24 1
Fcd  1  0.4  1  0.4   1.02
Raft 20×10 6
0.2 24 B 19
D
Fqd  1  2 tan  (1  sin  )2 f  1.01
B
Table 3 Properties of sand used in the analyses F d  1
Description unit Sand 1 Sand 2  0 
2

3 Fci  Fqi   1  0   1
Unit weight kN/m 17 19  90 
Friction Degree 35.5 38  0  
2
00 
2

angle F i   1  0    1  0 
1
    35.5 
Poisson’s - 0.3 0.28
ratio qult  3543kN / m 2
ult
Young kN/m2 53.000 135.000 Qraf  qult  A  2356.1MN
modulus
q   v1   ' xh  17kN / m2

203
 Vertical capacity of pile:  l 
Sand 1:   ln  2.5  (1   )   4.47
 r0 
Qs20m  f s20m xAs  364.8kN
l  2 /  (l / ro )  0.92
Sand 2:
Qs26m  f s26m xAs  673kN e2 l 1 e2 x 0.92  1
tanh(l )    0.73
e2 l  1 e2 x 0.92  1
Qs20m  Qs26m  1037.8kN  1.038MN Stiffness of piled group:
Qp  qp  Ap  (cNc  vpNq  dN ) Ae K pi  K1 n p  710.581 24  3481MN / m

d2 Load carried by piled group (%):

where Ap   0.79m2 Pp 1 K pi   rp K r
4    0.64
 Qp  7676kN Pp  Pr Pr K pi  (1  2 rp ) K r
1
Pp
Weight of pile
WP   c xVP  0.491MN where Vp : volume of pile where:
Total load of piles Qult pile  Qs  Qp  WP  8.223MN
 rp 
0. 2
K r / K pi   0.26
1  0.8( K r / K pi )
(pile compression)
pile  Qs  WP  1.038  0.491  1.529MN (pile
Qult Load carried by raft (%)
tension)
1 p 
Pr

1

1   rp K r  0.36
Load capacity of 24 piles Pp  Pr Pp K p  (1  2 rp ) K r
1
24
Q pile  24  8.223  197 .352 MN Pr

Load distributions in piled raft: Table 4. Load carried by piled group (%)- PDR
Stiffness for Raft:
 .a.Es  14.594  53 Number of piles
Kri    2209MN / m
I 1.1 % 16 20 24
a  ( BxL ) /   14 .594 m Load carried by piled 60 62 64
a 14 .594 Load carried by raft 40 38 36
  0 .36 , u  0 .3  I   1 .1
h 40
Stiffness for single pile 4.2 FEM for piled raft foundation
 4 2 tanh l l 
 (1   )     Properties of Soil and Raft
 l r0 
K 1  G r r0  
 1 1 4 tanh l l  Table 5 Parameters of model soils for FE
  (1   ) l r0  analysis
 710 .58 MN / m
Eu 53 Description unit Sand 1 Sand 2
G   20.38MPa
2(  1) 2(0.3  1) Material Model MC MC
d 1 Unit weight kN/m3 17 19
r0    0.5 ; Friction angle Degree 35.5 38
2 2
r0 c' kN/m2 1.0 1.0
 1
rb Poisson’s ratio - 0.3 0.28
Gl Permeability, k m/day 34.50 34.50
  0.386 ;
Gb Young modulus kN/m2 53.000 135.000
G avg Interfaces, 0.7 0.7
 1; Rinter (-)
Gl
EP Dilatancy angle, Degree 0.0 0.0
  1324.8 ψ'
Gl

204
 MC: Mohr-Coulomb 5.2 FE analysis of piled raft
Two layers of soil were simulated in
this analysis. Load carried by piled group ‐ PDR   

Bè Cọc

Table 6. Material properties of raft and pile 100

90 79
80 72

(%)
70
Description unit Raft pile 60
58

50 42

Type of Linear, Linear, 40

30
28
21
behaviour iso- iso- 20
10

tropic tropic 0
16 20 24
NUMBER OF PILE

Unit weight kN/m3 24 24 Figure 5. Load distributions in piled raft (%) - FEM
Poisson’s ratio - 0.17 0.3
2 7
5.3 Load distributions in piled group (%) –
Young modulus kN/m 2.7 x 10 2.7 x 107 PDR and FEM
Load carried by piled group (%)
90
79
80 72
70 62 64
60 58
60
50

%
40 PDR
30 FEM
20
10
0
16 20 24
NUMBER OF PILE

Figure 3. Simulation of raft foundation on sands

Figure 6. Load carried by piled group (%)
Table 7. Load carried by piled group (%) - FEM
Load applied to raft piles by PDR and FE
Number of piles method is divided to both raft and piles. The
% 16 20 24 PDR method and the FE method show that
Load carried by piled 58 72 79 piles load increases when the number of piles
increases.
Load carried by raft 42 28 21
6. CONCLUSIONS
5. RESULTS
Load on a pile raft foundation are distributed to
5.1 PDR analysis of piled raft both raft and piles.
The number of piles increases to 4 piles, the
Load carried by piled group ‐ PDR   
load on piles increases by 10%.
Bè Cọc
When reducing the number of piles, the load
100
90 on raft and piles along the bearing, load
80
capacity of rafts and piles is maximized.
(%)

70 62 64
60
60
50 40 38
40
30
36
7. REFERENCES
20
10
0
Brinkgreve, R. B. J. et al. (2007). Plaxis user’s
16 20 24 manual, Version 2.0. Rotterdam: Balkema.
NUMBER OF PILE
Poulos, H. G. and Davis, E. H. (1974). Elastic
solutions for soil and rock mechanics. New
Figure 4. Load distributions in piled raft(%) - PDR York: John Wiley

205
Poulos, H.G, (2000), Practical design procedures
for piled raft foundations. Design applications
of raft foundation. Hemsley J. A., editor.
London. Thomas Telford.
Poulos, H. G. (2001a). Methods of analysis of piled
raft foundations. A report prepared on behalf
of technical committee TC18 on piled
foundations. ISSMGE.
Poulos, H. G. (2001b). Piled raft foundations:
Design and applications. Geotechnique, 51
(2), 95–113
Poulos, H.G and Davis E.H (1980), Piled
foundation analysis and design, the University
of Sydney.
Randolph, M. F. (1994). Design methods for pile
groups and piled rafts. State of the Art Rep.,
Proc., 13th ICSMFE5: 61-82

206
The 3rd Int. Conf. on Transport Infrastructure & Sustainable Development (TISDIC 2019). Construction Publishing House. ISBN: 978-604-82-2893-4.

Study on change of total stress applied on the pile head in

the condition of increasing water level

Vo Phan, Nguyen Tuan Phuong

Faculty of Civil Engineering, Ho Chi Minh City University of Technology, Viet Nam, E-mail: [email protected].
Faculty of Civil Engineering, MienTay Construction University, Viet Nam, E-mail:
[email protected].

ABSTRACT: Soft soil treatment by combining reinforced concrete piles with geotextile is one of the
most popular solutions in the soft soil area of the southern provinces. The advantage of this solution
was a significant reduction in settlement due to pile head load, the quality of the constructions was
controlled conveniently and a little affection of the physic or chemical environment on the pile
manufacturing materials (compared to the cement soil piles), significantly reduced replenishment
materials (compared to the pre-loading methods) and some other advantages. However, the higher
effectiveness of solution is, the higher stress concentration ratio on the top piles is. In the condition of
increasing water level will change stress concentration ratio on the top piles, which affect the
effectiveness of solution. The content of the article studies on change of total stress applied on the
pile head in the condition of increasing water level.
Keywords: Load transfer, the stress concentration ratio on the top piles, arching effect.

1. INTRODUCTION solving the stress equation based on the

"Geosynthetic reinforced pile-supported trapdoor model and drawing the vertical stress
embankment on the soft soils". The "pile line in the sand in the case of the arching effect
elements" (reinforced concrete piles, stone and no arching effect were based on the
piles, wooden piles, sand piles, deep soil relationship between the load factor (P/γH) and
mixing columns...) are placed in a regular the ration (Z/B) showed in Figure 1.
pattern through the soft soil down to a lower
load bearing stratum, "Pile elements" are
typically arranged in triangular or rectangular
patterns in practice. The weight of the
embankment can be transmitted directly to the
pile heads by the arching effect or indirectly
through the membrane effect of the geotextile
layer. Pile elements would bear and transmit
load under the hard soil through the friction of
the pile with the surrounding soft soil.
Terzaghi (1943) showed the results of
studying the effect of the arching effect by Figure 1. The effect of arching effect to static
stress line

207
 No Sign of Location of earth pressure cell
earth
pressure
cell
1 Ps9 earth pressure cell was placed
between pile and geotextile layer
2 Ps1 earth pressure cell was placed
between two piles and on the
geotextile layer
3 Ps4 earth pressure cell was placed
Figure 2. The effect of arching effect to the on the pile head measuring
vertical stress path according to horizontal vertical stress
Terzaghi method
4 Ps8 earth pressure cell was located
at distance of pile ¼ distance of
2. TEST MODEL pile and geotextile layer
Test model was built 16 reinforced concrete 5 Ps10 earth pressure cell was placed
piles B.20 (M.250), pile length L = 14m between pile and geotextile layer
including 02 modules of each module of 7m. 6 Ps14 earth pressure cell was placed in
Geosynthetic reinforcement used Woven the sand layer 0.4m away from
geotextile has high tensile strength of 100 the pile head vertically
kN/m, maximum geogrid-strain 10%. Sand
7 Ps11 earth pressure cell was placed in
covered on the top of pile is big grain sand γs = the sand layer 0.8m away from
19 kN/m3, embankment height h=4m on the the pile head vertically
concrete plate thick 0,2m.
8 Ps6 earth pressure cell was placed in
the sand layer 1.2m away from
the pile head vertically
9 Ps2 earth pressure cell was placed in
the sand layer 1.6m away from
the pile head vertically

3. TEST RESULTS
3.1. Vertical stress was earth pressure cell Ps9
and Ps1 with load distribution 48kN/m2, the
underground water level was below pile head

Figure 4. Vertical stress was earth pressure cell

Ps9 and Ps1when the load distribution
was 48kN/m2, the underground water
Figure 3. The plan, cross section of test model level was below pile head

208
 Figure 4. The stress value was earth pressure significantly and the total vertical stress in the
cell Ps9 and Ps1 different from each other. The pile head reduced showed in Figure 6.
effect of geotextile reduced to the load Vertical effective stress distribution in the
distribution on soft soil and the load tend to soft soil is calculated by analytical method
transmit on pile heads combined with measurement data from the test
model, showed in figure 6.
3.2 Vertical stress was earth pressure cell Ps9
and Ps1 with load distribution 48kN/m2, the
stage of river level increase 63cm.

Figure 6. Vertical stress was calculated by

Figure 5. Vertical stress was earth pressure cell analytical method and measurement
Ps9 and Ps1 when the load distribution results from the test model
was 48kN/m2, the stage of river level
increase 63cm. The stress value of earth pressure cell Pzs
was located at different depths in the sand layer
Figure 5. Due to the river level increase such as depth of 1.6m; 1.2m; 0.8m; 0.4m and
63cm which greatly affects load capacity of at the top of pile was 0.0m. The increase and
soft soil, the tendency was soften the soil decrease stress in embankment showed in
below, the capacity of soil support decreases. Figure 6.
At the moment, most of the load distribution
will be focused on the bearing pile so the 4. CONCLUSION
vertical stress is at earth pressure cell Ps9
increases. 1. When the effects of prolonged rainstorms
caused local flooding, the total stress applied
3.3 Stress distribution in embankment was on the pile head increased, while the total
determined by analytical method and test method stress applied on the soft soil negligible
Vertical stress was caused by the weight of the change.
embankment. developed linear lines according 2. The stress concentration factor (n) is
to the depth. When the underground water level smaller in the dry season and increases in the
increases, the vertical total stress of submerged rainy season When the effects of prolonged
soil changed because of natural buoyancy. rainstorms caused local flooding.
When arching appeared, Vertical stress was
caused by the weight of the embankment 5. REFERENCES
changed significantly because of the effect of Terzaghi, K. (1943), Theoretical Soil Mechanics,
arching, so that a part of the total vertical stress Jonh Wily anh Sons, New York, 1943, pp 71-76
tends to distribution in the center of pile head, Aubeny, C.P. / Li, Y. / Briaud, J.L. (2002):
which made the total vertical stress distributed Geosynthetics reinforced pile supported
on reduced soil. When lowering the embankments: numerical simulation and
groundwater level, the total vertical stress design needs; Geosynthetics- 7th ICG-
distributed in the soft soil increased Delmas, Gourc& Girard (eds), pp. 365-368.

209
B. Le Hello, B. Chevalier, G. Combe, P. Villard,
Coupling finite elements and discrete elements
methods, application to reinforced
embankment by piles and geosynthetics.
BS 8006 (1995): British Standard, code of practice
of strengthened/ reinforced soils and other
fills, chapter 9.
Bergado, D. T., Anderson, L. R, Miura, N. and
Balasubramaniam, A. S. (1996). Soft Ground
Improvement in Lowland and Other
Environments, ASCE.
Chau Ngoc An (2012) Foundation, Ho Chi Minh
City National University Publishing House, pp
453 ÷ 446
Collin, J.G. / Watson, C.H. / Han, G. (2005):
Column-Supported Embankments solves time
constraint for new road construction;
Proceedings of the Geo-Frontiers Congress,
Austin, Texas, pp. 1-9.
D.T. Bergado, J.C. Bottles, New technical
measures to improve soft soil in construction,
Education Publishing House 1994, pp 58 ÷ 60
EBGEO (2007): Empfehlung for den Enwurf und
die Berechnung von Erdkurpern mit
Bewehrung aus Geokunststoffen; 2007.
Goh, A.T.C. / The, C.I. / Wong, K.S. (1997):
Analysis of piles subjected to embankment
induced lateral soil movements; Journal of
Geotechnical and Geoenviromental
Engineering, Vol. 123, No. 9, pp. 792-801
Gourge Samir Fahmi Farag, Leateral Spreading in
basal reinforced embankments supported by
pile – like element, University Kassel,
Germany 2008, pp 125
Hans-Georg Kempfert Berhane Gebreselassie,
Excavations and Foundations in Soft Soils,
University Kassel, Germany.

210
The 3rd Int. Conf. on Transport Infrastructure & Sustainable Development (TISDIC 2019). Construction Publishing House. ISBN: 978-604-82-2893-4.

Numerical simulation analysis of stress-deformation behavior

of soil and geosynthetic composite mass: a case study

Phan T.T Truc.,1,2 M W. Gui2, Pham Q.Thang 3

1
Dept. of Civil Engineering, Mientrung University of Civil Engineering, Tuyhoa City, Phuyen, Vietnam
2
Dept. of Civil Engineering, National Taipei University of Technology, No 1, Taipei 10608, Taiwan
3
Dept. of Civil Engineering, University of Texas Rio Grande Valley, Edinburg, TX, 78539, United States

ABSTRACT: This paper aimed to verify numerical modeling of compaction load to stress-deformation
behavior of soil and geosynthetic composite (SGC) mass under working stress conditions. Data from a
full-scale instrumented of SGC mass by Pham (2009) was used for a numerical analysis. Some
parameters of behavior of SGC such as compaction-induced stresses, stress and strain in
reinforcements, and angle of dilation of a soil geosynthetic composite will be investigated in this study.
The FEM modeling is built to verify the results from the field-scale experiments and show clearly show
the effect and call attention to the importance of the compaction conditions to stress-deformation
behavior of SGC mass.
Keywords: soil geosynthetic composite, compaction-induced stresses, angle of dilation.

1. INTRODUCTION These reinforced soil walls are designed by

using methods such as AASHTO Specifications
Modern techniques for mechanically stabilizing (2014), FHWA Guidelines (Berg et al. 2009),
soil were introduced in the 1960s. First used in and NCMA method (2009). They are
France, a method known as “reinforced earth” collectively referred to as Mechanically
used embedded narrow metal straps to reinforce Stabilized Earth (MSE) walls. MSE walls with
soil. In 1972, this technique was adopted in the geosynthetics as reinforcement have sometimes
United States by the California Division of been referred to as Geosynthetic Mechanically
Highways for construction of retaining walls. Stabilized Earth (GMSE). Reinforcement
Many other soil reinforcement methods were spacing for GMSE walls is typically between
researched and implemented following the first 0.3 and 0.9m. The relatively wide spacing stems
applications in the United States. Geosynthetics from a fundamental design concept of a tieback
are modern materials used to improve soil system.
conditions by providing tensile resistance and The significant beneficial effects of placing
stability. In the case of soil reinforcement, a sheet reinforcement on small spacing, however,
primary application of geosynthetics involves
have gained increasing attention in recent
reinforcing steep slopes. With the use of
years. The significant benefits of close
geosynthetics, the construction of reinforced
steep slopes is often more affordable and reinforcement spacing were first realized in
technically feasible when compared to actual construction and later validated by many
traditional construction techniques. field-scale experiments (Adams et al. 2007,
Furthermore, geosynthetic reinforcement is a Pham 2009, Wu and Pham 2013). These
cost-effective solution for stabilizing recurring experiments have confirmed that reinforcement
slope (Kim et al. 2019). spacing plays a far greater role than

211
reinforcement strength in the capacity of a
Compaction load


reinforced soil system. It has also been
established that the behavior of reinforced soil Position 1 Position 2 Position 3

with closely spaced geosynthetic reinforcement

z

can be accurately characterized as a SGC (Wu

and Pham 2013). 
In this study, the stress-deformation
behavior of soil geosynthetic composite mass (a)
are investigated for some issues such as h
compaction-induced stresses, stress and strain K2 ,c and K3,c line

in reinforcements and angle of dilation of a soil Ki,c

line

geosynthetic composite based on field-scale F

n
experiments by Pham (2009). G
2 D

'h,c,r
E
1
2. OVERVIEW BEHAVIOUR OF A SGC MASS C
B

A
There are many fundamental engineering
differences between MSE and SGC including v' 
'
v 'v,max v

recognition of compaction-induced stresses (b)

(CIS), geotextile-soil interaction, reinforcement
Figure 1. (a) Positions of compaction load, (b) Stress
spacing versus aggregate particle size, stresses
paths at depth z along section I-I as
and strains in the reinforcement. The focus of compaction load moves toward the section.
this paper can be listed as follows.

2.1 Compaction-induced stresses (CIS) 2.2 Stress and strain in reinforcements

Duncan and Seed (1986), Ehrlich and Many studies have acknowledged that the
Mitchell (1994), Pham (2009), and Wu and working stresses within the reinforcement of
Pham (2013) elaborated on the concept of constructed walls are apparently lower than
increase residual lateral stresses within
predicted using current design methods Allen et
composite structures caused by the
compaction processes (Figure 1). The basic al. (2001), Wu (2001), Holtz and Lee (2002),
concept is that elevated lateral stresses Pham (2009), Wu and Pham (2013), and Wu et.
develop within the soil due to compaction. al (2018). According to Pham (2009), based on
The vertical and lateral stresses are a
well-instrumented, large-scale, confined
function of the compaction equipment.
When the compactor is removed, the vertical compression testing of SGC composites
stresses return to normal which is equivalent constructed using woven geotextiles, the research
to the overburden load. However, a portion provides substantial proof that the relationship
of the lateral stresses remains locked-in to between reinforcement strength and spacing is
the soil as a result of the reinforcement
not linear. These works have demonstrated that
layers. These locked-in stresses increase the
lateral resistance within the soil mass by spacing plays a far more important role in the
providing an apparent increase in confining engineering properties of a composite soil mass
pressure. Conversely, the level of locked-in than the reinforcement strength. This engineering
stresses would be expected to decrease
principle is fundamental to the design of other
rapidly as the spacing between the
reinforcement increased. composite materials.

212
3. FINITE ELEMENT NUMERICAL
SIMULATION OF (SGC) TEST

3.1 Field scale of (SGC) test

Pham (2009) conducted a series of tests on
field scale SGC mass. The setup of the SGC
test is shown in Figure 2. The soil-
geosynthetic composite specimens in the tests
were 2.0m high and 1.4m wide and in a (a)
plane-strain condition. The soil was diabase
crushed gravel having a maximum particle
size of 33mm. Triaxial tests (specimen
diameter = 150mm and height = 300mm)
reveal that the soil has an internal friction
angle,  = 50o, and cohesion, c = 70 kPa, in
the stress range of interest for the SGC tests.
The soil was reinforced with a woven
geotextile of wide-width strengths of 70 and
140kN/m at the vertical spacing of 0.2 and
0.4m. The geotextile with T = 140kN/m was
obtained by gluing two sheets of the same
geotextile with T = 70kN/m. Uniaxial tension (b)
tests were conducted to verify the tensile Figure 2. (a) The Soil-Geosynthetic Composite (SGC)
strengths. The conditions of the SGC tests are Test Setup, (b) Failure Planes of the
shown in Table 1. Composite Mass after Testing in Test 2

Table 1. Test conditions of one unreinforced test (Test 1) and four SGC tests (Tests 2, 3, 4, and 5)

Wide-Width
Geosynthetic Confining Reinforcement,
Test Designation Strength
Reinforcement Pressure Spacing, S
Reinforcement

Test 1 None 34 kPa None No reinforcement

Test 2 Geotex 4x4 34 kPa T = 140 kN/m Sv = 0.2 m

Double-Sheet
Test 3 34 kPa T = 140 kN/m Sv = 0.4 m
Geotex 4x4

Double-Sheet
Test 4 34 kPa T = 70 kN/m Sv = 0.4 m
Geotex 4x4

0
Test 5 Geotex 4x4 T = 70 kN/m Sv = 0.2 m
(unconfirned)

213
3.2 Numerical model with the assumption of fully bonded soil-
Finite element (FE) method of analysis was reinforcement interface was valid. The
employed to simulate the field-scale SGC hardening soil model was selected for
experiments described behaviour of SGC mass simulation of the behavior of the fill material in
(Test 2). The analysis was conducted by using the SGC tests. It is a second-order hyperbolic
the finite element Plaxis 2D (Plaxis B. V 2002). elasto-plastic hardening model involving
(Figure 3). friction hardening to express the plastic shear
In this analysis, geosynthetic reinforcement strain in deviatoric loading and cap hardening
was simulated by a linear elastic model that to express the plastic volumetric strain in
input parameters are shown in Table 2. The primary compression. The soil parameters of
values of interface propertie between soil-
the hardening soil model determined from
reinforcement (Ri = 0.9) is valuable in this
research. It should be noted that Ri = 0.9 is drained triaxial. The input parameters of FEM
used to compare with the case (Pham 2009) modeling are shown in Table 2.
v

Reinforcement

c c

H
Sv = 0.2 m

W = 0.7H

(a)

(b)
Figure 3. (a) Typical Geometric and Loading Conditions of a SGC mass, (b) FEM modeling in Plaxis 2D

214
Table 2. The input parameters of FEM modeling
Material Value

Soil properties (Diabase)

Model Hardening Soil
Peak plane strain friction angle, ϕ (°) 50
Cohesion, c (kN/m2) 70
a
Dilation angle , Ψ (°) 19
Unit weight, γ (kN/m3) 16.8
Eref50 a, (kN/m3) 62374
Eur50 = 3* Eref50 , (kN/m3) 187122
Stress dependence exponent, m 0.5
Failure ratio, R 0.9
Poisson's ratio, υ 0.2
Pref (kN/m2) 100
Reinforcement (Single- sheet Geotex 4x4)
Elastic axial stiffness (kN/m) 1000
Reinforcement spacing (m) 0.2
Modular block properties
Model Model Linear elastic
Stiffness modulus ((kN/m2) 3*106
Unit weight, γ (kN/m3) 12.5
Poisson's ratio, υ 0
Block-Block interfaceb
Model Mohr-Coulomb
Stiffness modulus ((kN/m2) 3*106
Unit weight, γ (kN/m3) 0
Poisson's ratio, υ 0.45
Angle of internal friction , ϕ (°) 33
Cohesion, c (kN/m2) 2
Soil -Block interfaceb
Model Mohr-Coulomb
Unit weight, γ (kN/m3) 0
Poisson's ratio, υ 0.45
Angle of internal friction , ϕ (°) 33.33
Cohesion, c (kN/m2) 46.67
Soil -Soil interfaceb
Model Mohr-Coulomb
Unit weight, γ (kN/m3) 0
Poisson's ratio, υ 0.45
Angle of internal friction , ϕ (°) 45
Cohesion, c (kN/m2) 63

a
Based on drained triaxial testing.
b
Defined interface elements with zero thickness in Plaxis Software.

215
4. RESULTS AND DICUSSION values. Figure 5 shows the lateral displacements
on the open faces of the specimen at applied
4.1 Compaction-induced stresses (CIS) pressures of 400 kPa, 1,000 kPa, 2,000 kPa and
The global stress-strain relationships, as 2,500 kPa. It is seen that the simulated
obtained from FE analysis and the SGC Test 1 displacements are in good agreement with the
and Test 2 are shown in Figure 4. The Figure 4 measured values.
show that the results with consideration of CIS These results are consistent with those of
gives slightly better simulation of stress-strain Pham (2009) and this result may be explained
curve. These results are in agreement with Pham by the fact that because of the conditions of
(2009) findings which showed that the good compaction in SGC mass that the
compaction energy used in the SGC tests was values of interface property between soil-
very low. As a result, the magnitude of CIS was reinforcement affects not to be significant to
very small, and the effect of CIS on the global the stress-deformation behavior of this SGC
stress-strain relationship was not significant. mass.
A comparison of the distribution of strains
4.2 Stress and strength in reinforcemen in the reinforcement in SGC Test 2 between
Figure 5 shows the lateral displacements on the the FE analyses and measured data are shown
open faces of the specimen at applied pressures in Figure 6. It is seen that the simulated strains
of 400 kPa, 1,000 kPa, 2,000 kPa and 2,500 are in good agreement with the measured
kPa. It is seen that the simulated displacements values in both two cases at reinforcement layer
are in good agreement with the measured 1.6 m and 0.8 m from the Base.

Figure 4. Comparison of Global Stress-Strain Relationship of SGC Test 1 and Test 2

Figure 5. Comparison of Lateral Displacement at Open Face of SGC Test 2

216
 (a)

(b)
Figure 6. Comparison of Reinforcement Strains of SGC Test 2: (a) At reinforcement Layer 1.6 m from
the base; (b) At reinforcement Layer 0.8 m from the base.

Test 2 as obtained from FE analysis and the

4.3 The angle of dilation of a soil geosynthetic
measured data. SGC Test 2 was conducted under
composite
a confining pressure of 34 kPa and to provide
Figure 7 shows a comparison of the global stress- additional data under different confining pressures
strain and volume change relationships of SGC of 100 kPa and 200 kPa by FEM. It is seen that

217
the global stress-strain and volume change were approximately -8o, -12o, -13o under
relationships under the confining pressure of 34 confining pressures of 34 kPa, 100 kPa and
kPa as obtained from FE analysis are in good 200 kPa, respectively. This suggests that the
agreement with the measured data. These results presence of geosynthetic reinforcement has a
are consistent with those of Pham (2009). tendency to “suppress” dilation the
One of the issues that emerges from these surrounding soil. A soil having less tendency
findings is that the reinforcing mechanism of to dilate will become stronger. The dilation
GRS mass can be viewed in terms of the behavior offers a new explanation of the
angle of dilation. The angle of dilation of a reinforcing mechanism, and the angle of
geosynthetic-reinforced soil mass is smaller dilation provides a quantitative measure of
than the angle of dilation of an unreinforced the degree of reinforcing effect of a GRS
soil mass. From Figure 7 the angles of mass. These results are in line with those of
dilation of the soil geosynthetic composites previous studies, Pham (2009)

Figure 7. FE Analyses of SGC Test 2 under different confining pressures

5. CONCLUSIONS the reinforcing mechanism, and the angle of

dilation provides a quantitative measure of the
The results from FE analyses and the tests are degree of reinforcing effect of a SGC mass.
also in good agreement in failure value of
deviatoric Stress, horizontal displacement, strain 6. REFERENCES
of reinforcement and previous studies Pham
(2009). Hardening Soil model have good results AASHTO. 2014. American Association of State
Highway and Transportation Officials: LRFD
for GRS simulation by FEM modelling
Bridge Design Specifications, 7th Edition.
Interface property element between soil- Adams, M. T., Schlatter, W. and Stabile, T. 2007.
reinforcement affects not to be significant to the Geosynthetic-Reinforced Soil Integrated
stress-deformation behavior of this SGC mass Abutments at the Bowman Road Bridge in
in good compaction condition. The presence of Defiance County, Ohio. Geotechnical Special
geosynthetic reinforcement has a tendency to Publication No. 172: Proc. Geo-Denver 2007,
suppress dilation the surrounding soil, and ASCE, Denver, Colorado.
reduce the angle of dilation of the soil mass. The Allen, T. M. & Bathurst, R. J. 2001. Application of
dilation behavior offers a new explanation of K0-stiffness Method to Reinforced Soil Wall

218
 Limit State Design. Final Research Report to Retaining Walls. Report No. WA-RD 532.1,
Washington State Department of Washington State Department of
Transportation, Seattle, Washington State. Transportation, Washington.
Berg, R. R. Christopher, B. R. & Samtani, N. C. NCMA. 2009. Design Manual for Segmental
2009. Design of Mechanically Stabilized Earth Retaining Walls, 3rd Edition, TR 127B,
Walls and Reinforced Soil Slopes, Publication National Concrete Masonry Association,
No. FHWA-NHI-10-024, Volume 1, Federal Herndon, VA, pp.302.
Highway Administration, McLean. Pp.332. Pham, T. Q. 2009. Investigating Composite
Duncan, J. M. & Seed, R. B. 1986. Compaction- behavior of Geosynthetic-Reinforced Soil
induced earth pressure under Ko conditions, J (GRS) Mass. Ph.D. Dissertation, University of
of the Geotechnical Engineering Colorado Denver.
Division, ASCE, 112(1), 1–22. Plaxis B.V. 2002. Plaxis 2D - Version 8 Manual.
Elton, D. J. & Patawaran, M.A.B. 2005. Balkema, Rotterdam.
Mechanically Stabilized Earth (MSE) Wu, J. T. H. 2001. Revising the AASHTO
Reinforcement Tensile Strength from Tests of Guidelines for Design and Construction of
Geotextile Reinforced Soil. Technical Report, GRS Walls. Report CDOT-DTD-R-2001-6,
Alabama Highway Research Center, Auburn Colorado Department of Transportation,
University. University of Colorado Denver, pp.148.
Ehrlich, M. & Mitchell, J. K. 1994. Working stress Wu, J. T. H. & Pham, T. Q. 2013. Load carrying
design method for reinforced soil walls. J. capacity and required reinforcement strength
Geotech. Eng., ASCE, 120(4), 625-645. of closely spaced soil-geosynthetic
Kim, Y. J., Kotwal, A. R., Cho, B. Y. Wilde, J. & composites. J. Geotech. Geoenviron., ASCE.
You, B. H. 2019. Geosynthetic Reinforced 139(9), 1468–1476.
Steep Slopes: Current Technology in the Wu, J. T. H., Tung, C. Y., Adams, M. T. & Nicks,
United States, Applied Science, 9(10): 2008. J. E. 2018. Analysis of Stress-Deformation
Holtz, R. D. 2010. Reinforced Soil Technology: Behavior of Soil-Geosynthetic Composites in
from Experimental to the Familiar. Terzaghi Plane Strain Condition, Transportation
Lecture, Geo-Florida, Palm Beach. Infrastructure Geotechnology, 5(3), 210-230.
Holtz, R. D. & Lee, W. F. 2002. Internal Stability
Analyses of Geosynthetic Reinforced

219
The 3rd Int. Conf. on Transport Infrastructure & Sustainable Development (TISDIC 2019). Construction Publishing House. ISBN: 978-604-82-2893-4.

Analysis of theoretical equations for evaluation of bearing

capacity of soft soils

Yongmin Kim1, Yumi Chae1, Jaehong Kim1,

1
Department of Civil and Environmental Engineering, Dongshin University, Naju, South Korea

Corresponding author, E-mail: [email protected]

ABSTRACT: The practical use of geotextile mat has been increased to improve the bearing capacity of
the soft ground in the west and south coast in Korea. In the beginning stage of construction in soft ground,
appropriate tensile strength of geotextile mat should be needed for trafficability of construction equipment
after covering up the soil on the mat. The theoretical bearing capacity of the soft ground has been
described by Yamanouchi (1985). However, many researchers have modified the equation to consider
bearing capacity with respect to strength and thickness of sand mat on soft ground. This study suggests a
modified equation to evaluate appropriately the bearing capacity of the soft ground with the incorporation
of the seam tensile strength of geotextile and soil strength of the sand mat layer.
Keywords: Geotextile, soft ground, tensile strength, yamanouchi equation, bearing capacity.

1. INTRODUCTION In order to improve and develop these soft

grounds, the utilization of various geotextiles
The soft ground, which is widely distributed on has been increasing in the field, South Korea.
the west coast and the southern coast, has
Unstable due to fracture of geotextile mat
many dredged landfills and dumping to reclaim
occurs by weakening of the fabric and
agricultural land and industrial complexes.
Since the permeability coefficient of soft soil is strength of the seam. Although research on the
too small, the consolidation is slow and the material of geotextile mats is still actively
water content is very large, so the bearing performed in the world, there is a lack of
capacity of soft ground is not only weak, but research to improve the tensile strength of
also differential settlement occurs in the long geotextile seam, ultraviolet damage to
term, causing problems such as collapse and geotextile mat, and inadequate application of
sinking of construction equipment in the field. the stitching fiber. When the geotextile design
of the road is not in the maximum principal
stress direction, seam tensile strength is
considered to be more than 50% of the fabric
strength, but only 50% of it is applied as the
seam tensile strength when geotextile mats are
laid on a wide soft ground.
Globally, geomembrane and geogrid have
been increasingly used in a field of
construction work, and the need for a seam of
Figure 1. Belt conveyor equipment that covers soft geosynthetic composites is still increasing in
soil with sandy soils (KAITA, 2018) the field (An et al., 2009; Yuu et al., 2007)

220
 An et al. (2009) investigated the orientation distributed contact width should increase to
of seam tensile strength of the geogrids reduce the distributed contact pressure and
through laboratory tests and finite element increase the thickness of sand mat in order to
analysis. The comparison of the magnitude of increase the safety factor of the bearing capacity.
the tensile strength along the direction and the Ju et al. (2007) studied a modified bearing
analysis of the maximum resistance values by capacity equation of Yamanouchi (1984)
measuring the seam tensile strength of the formula to explain the contradictions by
geogrid was presented. laboratory tests (model tests). This
Jeon et al. (2008) examined the behavior of modification takes into account the magnitude
the seam area of a geotextile tube through of the load acting of sand mat on soft soils and
theoretical studies and experiments to the seam tensile strength of geotextiles.
determine the effective sealing structure and In this study, the modified equation of Ju et al.
reasonable design strength. When the stitching (2007) will be analyzed to adequately re-evaluate
fiber of the geotextile was similar, the seam the bearing capacity of the soft ground. The main
type of geotextiles had a significant effect on consideration is in estimating the bearing
the seam tensile strength and it was also capacity of the sand mat above the soft ground
confirmed that the on-site condition of the containing seam tensile strength of geotextile.
stitching fiber affects the tensile strength.
Yuu et al. (2007) determined that the long- 2. REVIEW OF CONVENTIONAL EQUATION
term design allowable strength of a geotextile
tube on the maximum strength of the weak part 2.1 Yamanouchi (1985)
of the structure, and that the design strength of Yamanouchi (1985) formula is a method of
the tube is based on the seam tensile strength comparing the magnitude of the load acting on
rather than the tensile strength of the the soft soil by the distributed load and the
geotextile. In addition, they proposed how equipment stresses according to the soil
seam tensile strength of the tube is an thickness and the bearing capacity considering
important factor in coastal land harbor the seam tensile strength of the geotextiles.
construction by the geotextile tube.

Figure 3. Schematic design of Yamanouchi method

Figure 2. Overturn of construction equipment due to to calculate bearing capacity with
uplift and settlement in soft ground geotextile tensile strength (Yamanouchi,
(KAITA, 2018) 1985)

In order to improve the soft ground, the The angles and parameters of Yamanouchi
trafficability of construction equipment should (1985) equation are difficult to define
ensure the bearing capacity of the ground. The numerically. The load distribution angle ()
semi-empirical method of Yamanouchi (1985) occurs in the ground where the load is
equation and Meyerhof (1974) equation of dispersed, and the Yamanouchi (1985) is
bearing capacity formula have been applied to assumed to be 2:1 (vertical : horizontal). As
calculate the strength of soft ground. The shown in Fig. 3, it can be expressed in Eq. (1)
constitutive theory of Yamanouchi (1985) according to the tensile force ( ) due to the
formula contains logical contradictions that behavior of the geotextile mat laid on the soft

221
soil and the flexure caused by the distributed to determine the load distribution angle 1.
contact pressure on the surface. It is the Therefore, the equation of load distribution
allowable bearing capacity (qa) formula of soft angle assumes a relationship as shown in Eq.
soil using safety factor (Fs) ‘4’. (2) (Ju et al., 2007).
1  H  Df    2sin  1    f ( B, d ) (2)
qa  . 1   . 5.3.cu  Ta .   
Fs  b   b r As shown in Eq. (2), if the dispersion angle
 is considered as a function of the load width
4 S R(1  cos  ) 
 a   Df  (1) ‘B’ and the thickness ‘d’ of the sand mat layer,
b  H  Df  ‘2dtan1’ can be substituted for the sand mat
layer thickness ‘d’ in the equation proposed by
where, Fs is factor of safety, cu is undrained Yamanouchi, therefore, Ju et al. (2007)
shear strength of dredged landfill layer, H is suggested a modified bearing capacity equation
soil layer thickness, b is width of distributed as Eq. (3),
contact pressure, Ta is allowable tensile
1  2d tan 1  D f    2sin  1 
strength of geotextile considering damage qa  . 1   . 5.3.cu  Ta .   
during construction,  is load distributed angle Fs  b   b r
between tensile strength (T) of geotextile mat 4S a R 1  cos  2  
and horizontal plane (≈15˚).    Df  (3)
b  2d tan 1  D f 
In Yamanouchi’s (1985) proposal, if the
width of contact pressure is increased for the 2.2 Meyerhof (1974)
purpose of decreasing the distributed contact
pressure, the bearing capacity is reduced and Unlike the Yamanouchi empirical formula, the
there is a logical contradiction that the Meyerhof method is able to calculate the
thickness of the sand mat layer should be tensile strength of geotextile in determining
increased to raise the safety factor of the bearing capacity by considering the
bearing capacity. inhomogeneous layer in the ground. The
Ju et al. (2007) proposed a modified equation is as Eq. (4),
equation of bearing capacity that explains the
contradictions of Yamanouchi equation. They
pointed out the Yamanouchi equation that the
load is dispersed and disappeared as the ground
deeper into the ground even if the contact
pressure acting on the ground increases. They
performed laboratory tests in order to prove
that the larger the thickness of the sand mat
layer is, the larger the load dispersion effect is. Figure 4. Scheme of Meyerhof method to consider
In a simple design for the improvement of soft the inhomogeneous layer in soft soil
soil, the load distribution angle 1 from the (Meyerhof, 1974)
ground is 26.57° which is applied the ratio of
depth ‘2’ and dispersion ‘1’. However, the load Meyerhof (1974) proposed a formula to
distribution angle can be influenced by the evaluate the tensile strength of geotextiles by
modifying a conventional bearing capacity
width of the load and the thickness of the sand
equation shown in Eq. (4).
mat layer.
In particular, in various conditions of the  b 2  b 
soft ground-geotextile-sand mat layer system, 1  0.2. L  .5.3.cu   .H . 1  L  . 
1      
geotextiles are used to exert tensile force qa 
Fs  2.D f  K s .tan  2.Ta .sin  
when subjected to a load, so it is always a  1     Df  
 H  b b 
problem to use the concept of 2:1 constantly

222
 where, Fs is the factor of safety, cu is
undrain shear strength, H is the thickness of
sand mat layer, b is the width of contact
pressure on the ground, Ta is allowable tensile
strength of geotextile considering damage
during construction,  is the angle between the
crushed geotextile mat and the horizontal
plane (≈15˚),  is friction angle of sand mat
layer, Ks is punching shear resistance factor.
The punching shear resistance factor (Ks) used Figure 5. Punching shear failure (Kim, 2010)
in the formula are shown in Table 1. Table 1
shows the relation between the friction angle
of soils and punching shear factors depending
on soil strength.

Table 1. Punching shear factors with the soil

friction angle (Meyerhof, 1974)
Soil friction angle Punching shear factor
(°) (Ks)
20 1.89 Figure 6. General shear failure (Kim, 2010)

25 2.22 In the case of loading on dense or compacted

30 3.06 sand overlying soft clay, contact pressure or
foundations are constructed on compacted sand
35 4.45 layers that overlie soft clay soils. If the
40 6.95 thickness of the sand layer under the foundation
is relatively small (Fig 5), the failure surface
45 11.12 may extend into the soft clay layer. This is
50 19.15 punching shear failure shown in Fig 5.
However, if the sand mat layer under the
3. BEARING CAPACITY OF COMPOSITE loading is large, the failure surface will lie
SOILS entirely in the sand mat layer, as shown in Fig.
6 (general shear failure).
One of the key requirements for applying If it is assumed that the upper layer of stiff
shallow foundations or contact pressure in soil and the lower layer of soft soil. When the
layered soils considering shear failure is that contact pressure is distributed in the ground
the soil layer where the foundation or contact surface, the shear failure type of the sand layer
pressure is located must have sufficient is generally divided into punching shear failure
strength. The above-mentioned method of and general shear failure. In this case of
estimating the bearing capacity has been punching shear failure, it may be mentioned
described as a homogeneous soil condition that the fundamental width (B) is greater than
under this assumption, but in the field the depth (H) from the base floor to the lower
heterogeneous ground conditions such as a layer, and the friction angle of the undrained
composite ground or various layered ground condition of the clay layer can be considered
can be frequently observed. zero and then the bearing capacity of the clay
Fig 5 and 6 show an example describing the layer (qb) and the sand layer (qt) can be
bearing capacity calculation formula composed obtained as Eq. (5) and (6).
of heterogeneous layers that can be easily seen  B
qb  1  0.2  5.14 c 2   1 ( D f  H) (5)
during soft ground improvement.  L

223
 1
qt   1 D f N q (1) Fqs (1)   1 BN (1) F s (1) (6)
2
The bearing capacity formula of Meyerhof
(1974) is derived as Eq. (7). The increased
thickness of the sand layer is a factor that
reduces its bearing capacity in this formula.
Therefore, a revised proposal is needed to
calculate the appropriate bearing capacity.
1  B
qa  1  0.2  5.14 c2
Fs L
 B   2 D f  K s tan 1
 1H 2  1   1     1D f
 L  H  B
1 Figure 7. Allowable bearing capacity by the thickness
  1D f N q1 Fqs1   1BN 1 F s1 (7) of sand mat layer
2

4. PROPOSAL OF MODIFIED THEORETICAL Meyerhof (1974)’s formula takes into

EQUATION OF BEARING CAPACITY account the bearing capacity for
inhomogeneous layer in the upper part, but
The equation of Yamanouchi (1985) and Ju et with the exclusion of the thickness of sand mat
al. (2007) did not take into account the layer, there is little change in the increase in
thickness and strength of the sand layer, but the bearing capacity. A proposed equation that
only the strength of the clay layer and the was modified by considering the strength of
tensile strength of geotextile were considered. two layers could be found to have somewhat
In this study, the formula of Meyerhof (1974) smaller bearing capacity than the results of the
and the advantages of Yamanouchi (1985) Yamanouchi (1985) method. The dispersion
were mixed together and proposed as Eq. (8). angle and the pressure width transmitted to the
1  H 2Ta sin  K s tan   soft ground caused by the load dissipation of
qa  . 1   .  5.3cu    (8)
Fs  b B b  contact pressure on the surface ground should
For field application of these formulas, the be applied to the seam tensile strength of
soil strength parameters are the factor of safety geotextile mat. But the Yamanouchi (1985)
4, cu undrained shear strength 25 kN/m2, H formula is applying the opposite. And, the
thickness of sand mat layer, b width of contact thickness and strength of the sand mat layer are
pressure on the ground 1m, Ta allowable tensile having a lot of impact on the increase in the
strength of geotextile considering damage bearing capacity.
during construction 150 kN/m2,  angle The strength of the soft ground containing
between the crushed geotextile mat and the geotextile varies depending on the magnitude
horizontal plane (≈15˚),  friction angle of the of the seam tensile strength. Since the seam
sand mat layer 35°, Ks punching shear tensile strength of geotextile mat is recognized
resistance factor 4.45. only by 50% of the geotextile strength, the
The input data for the field application of change in bearing capacity as a result of the
the formulas were determined based on the increase in seam tensile strength is shown in
construction equipment used for installing Fig. 8. Thus, the thickness and strength of the
geotextile mat for reinforcement of soft sand mat layer are having a lot of impact on the
ground. When geotextile is placed on a increase in the bearing capacity.
saturated clay layer and deposited into a sand The allowable bearing capacity is improving
mat layer, the allowable bearing capacity up to 133 kN/m2 from 121 kN/m2 with the
calculated by four formulas for contact thickness of the sand mat layer increasing from
pressure on the ground surface compared as 10 cm to 2 m. Table 2 shows the increment
shown in Fig. 7. shown in Fig. 8.

224
 multiplied the ratio of the thickness of sand
mat layer to bearing capacity instead of the
punching shear resistance factor, so the
appropriate bearing capacity of the sand mat
layer can be calculated. It is necessary to
construct economical soft soil reinforcement
by supplementing the results of Meyerhof
(1974) formula which is evaluated too small.

6. ACKNOWLEDGEMENTS
This research was supported by a
grant(18RDRP-B076564-05) from Regional
Figure 8. Allowable bearing capacity by the Development Research Program funded by
thickness of sand mat layer Ministry of Land, Infrastructure, and Transport
of Korean government.
Table 2. Variation of allowable bearing
capacity (qa) according to the seam tensile 7. REFERENCES
strength (Ta) of geotextile with a thickness of
sand mat layer An, B.W, Lim, D.Y, Lee, K.Y, Chung, C.G, and
Jeon, H.Y(2009). “Interpretation on Geogrids
Allowable bearing capacity, (kN/m2) Junction and Tensile Behaviors by Finite
H=0.5 H=1.0 H=1.5 H=2.0 Element Analysis”, Textile Science and
m m m m Engineering, 46(4), 239-244.
Ju, J.W, Seo, K.W, and Jang, M.H(2007). “Bearing
50 65.99 92.39 118.79 145.18 capacity of soft ground reinforced by
60 68.15 95.41 122.67 149.92 geosynthetics and sand mat,” Korean
70 70.31 98.43 126.56 154.67 Geosynthetics Society, 2007(1), 1-7.
Jeon, H.Y, Yuu, J.J, and Chang, Y.C(2008),
80 72.46 101.45 130.43 159.42 “Seaming Behaviors Interpretation of
Geotextile Tube,” Fall Geotechnical
5. CONCLUSIONS Engineering Conference, 2008, 1593-1603.
Kim, S.I(2010). “Foundation Engineering,” Kyobo
In this study, the modified equation, which book center, Seoul, pp.111-113.
calculates the bearing capacity of soft ground KAITA (2018). “Development of Sealing
reinforced by geotextile, was evaluated by Technology and Efficient Installation Method
applying the seam tensile strength and for Increasing Joint Strength of Geotextile in
thickness of sand mat layer appropriately. The Soft Ground” Korea Agency for Infrastructure
strength and thickness of the sand mat layer in Technology Advancement, R&D Report,
the field are the main factors for improving the 18RDRP-B076564-05, South Korea.
bearing capacity because it increases directly Leshchinsky, D. (2004), “Design software for soft
soils,” Geotechnical Fabrics Report, March,
the trafficability on soft ground. Therefore, it is
22(2), 18-23.
necessary to correct conventional equations
Meyerhof, G.G. (1974), “Ultimate Bearing
underestimated in the calculation of bearing Capacity of Footings on Sand Layer Overlying
capacity and apply them to the modification Clay,” Canadian Geotechnical Journal, 11(2),
formula by properly assessing the unstable 224-229.
factors in construction on soft ground. Yamanouchil, T. (1985), “Recent development in
The construction method of reinforcing soft the used synthetic geofabrics and geogrids”,
ground by utilizing geotextile is influenced by Symposium on Recent Developments in
many surrounding environments for calculating Ground improvement techniques, Bangkok,
bearing capacity. Yamanouchi (1985) theory 205-224.

225
Yuu, J.J, Oh, Y.I, and Jeon, H.Y(2007), “A
Comprehensive Consideration on the Seam
Strength of Geotextile Tubes: (1) failure
behaviors and the required strength,” Korean
Geosynthetics Society, 2007(11), 241-246.
Yuu, J.J, Oh, Y.I, and Jeon, H.Y(2008), “A
Comprehensivie Consideration on the Seam
Strength of Geotextile Tubes: (2) Basics of the
Stitching Seam,” Korean Geosynthetics
Society, 2008(4), 215-210.

226
The 3rd Int. Conf. on Transport Infrastructure & Sustainable Development (TISDIC 2019). Construction Publishing House. ISBN: 978-604-82-2893-4.

Effect of montmorillonite content on the bearing capacity

of soft ground improved by cement deep mixing columns

Nguyen Ngoc Thang 1, Nguyen Minh Tam 2, Le Van Nam 2

1
Civil Engineering Faculty, Tien Giang University, Tien Giang, Vietnam
2
Civil Engineering Faculty, Ho Chi Minh City University of Technology, Ho Chi Minh city, Vietnam

Corresponding Author: NGUYEN Ngoc Thang

Address correspondence to First author: [email protected]

ABSTRACT: The Mekong Delta is the largest delta of Vietnam, covered Holocence sediment due to
sedimentation of the Mekong river system. The minerals of soft clay in this area usually include:
Montmorillonite (MMT), Illite, Chlorite, and Kaolinite. MMT in clay affect to the soil swelling
properties. The developed swelling pressure will destroy the lightly construction and road surface. This
research study focused on the influence of MMT content to the unconfined compressive strength of Deep
Cement Mixing (DCM) by using the unconfined compressive strength of DCM, X-ray CT method and
laboratory scale model. All samples were stabilized in fresh water at 28 days curing before performing the
test. The effect of MMT content on the unconfined compression property of DCM specimens has been
performed using the standard samples. In addition, the 1-g laboratory model has been used to evaluate the
influence of MMT content in the clays on the compressive strength of soft ground supported by DCM
columns. The results proved that the compressive strength of soft ground improved by DCM columns
when the MMT content increased.
Keywords: Montmorillonite, DCM, soft soil, mekong delta.

1. INTRODUCTION Chanh - Ho Chi Minh from 11.3% to 13.3%, in

An Giang from 5% to 10% (James L. Post,
The Mekong Delta is located in the southern of
1971) and in Mekong Delta from 0% to 8%
Vietnam. As a delta area, most of the soil
(Nguyen Huu Chiem, 1993).
layers are categorized as the soft soil. The
properties of the soil are high water content, In this study, the unconfined compression test
low stiffness and less frictional angle and less is used to investigate the optimum cement content
cohesion as well. of soil cement specimens. The series of laboratory
During recently years, stabilization of soft model is also used to simulate the influence of
ground by DCM method has been applied MMT content to the bearing the capacity of
popularly in this area. There are many factors stabilized soil by soil cement columns.
affect to mechanical properties of stabilized
soil. The mineralogical properties of soil 2. MATERIALS
strongly affect the strength of treated soil
2.1 Soil
(Terashi, M., 1997). The minerals of soil
usually include: MMT, Illite, Chlorite, and The soil is one of the typical soft soils in the
Kaolinite, Quartz… MMT mineral affect to the Mekong Delta. The properties of soil are
soil swelling properties. According to Phan Thi obtained by the laboratory tests, is given in
San Ha et al. (2007), MMT content in Binh Table 1.

227
Table 1. The properties of the soil in the 3. METHOD TEST
Mekong Delta
3.1 Unconfined compression test
ID Parameters Value
1 Water content, W(%) 45.14 Generally, the unconfined compressive
2 Wet unit weight, w(kN/m3) 17.43 strength of soil cement is between 50kPa and
3 Dry unit weight, w (kN/m3) 12.01 4,000kPa, which is 50 to 100 times greater than
4 Initial void ratio, eo 1.29 that of pre-improved soil but still much lower
5 Liquid Limit, LL(%) 47.59 than that of concrete.
6 Plastic Limit, PL(%) 25.20 The focus of the material design is that the
7 Plasticity Index, IP(%) 22.39 quality of the product must be achieved to
8 Cohesive, c(kPa) 6.70 satisfy the minimum strength and other design
9 Friction angle, (o) 3o53’ requirements. Although the soil cement mixing
10 MMT content (%) 6.00 contractor often determines the mix design, it
is important for the design engineer to
2.2 Bentonite understand the factors contributing to the
Bentonite is an absorbent aluminum strength of the soil cement.
phyllosilicate which in general terms are impure In this study, the soil cement specimen is
clay consisting mostly of MMT. Bentonite made with the diameter of 50mm and the height
expands when wet – sodium bentonite can absorb of 100mm, which different conditions: Cement
the several hundred present of its dry weight in content is 5%, 10%, 15%, 25%; Bentonite
water. It is commonly used in drilling, to make content is 0%, 5%, 10%, 15% (MMT content is
slurry wall, and to form impermeable barriers. 6%, 9%, 12%, 15% respectively); water content
The properties of bentonite are given in Table 2. is 80%; curing condition is freshwater.
The influence of cement content on the
Table 2. Bentonite properties unconfined compressive strength (UCS) was
investigated by a series of unconfined
ID Parameters Value compression tests have been performed in
1 Specific Gravity Gs(%) 2.73 laboratory at 28 days of curing. In total, 48
2 Liquid Limit, LL(%) 405.10 unconfined compression tests were performed.
3 Plastic Limit, PL(%) 51.50 The results are shown in Figure 1 and Figure 2.
The UCS of soil cement specimens decreases
4 Plasticity Index, IP(%) 353.30
almost linearly with the increasing of MMT
5 MMT content (%) 65.00 contents, as shown in Figure 1 and Figure 2.
Based on the Figure 1, the optimum cement
2.3 Cement content can be determined when the soft soil has
Portland cement PCB40 is used for this study. different MMT contents.
The properties of cement are given in Table 3.

Table 3. Cement properties

ID Parameters Value
1 Specific surface (cm2/g)  3300
2 Time of setting
Initial set (min)  45
Final set (h)  10
3 Chemical (%)
MgO  5.0
Figure 1. Influence of cement content on UCS at
SO3  3.5 28 days

228
 with its width of X-ray attenuation is
calculated. The coefficient of absorption is
used in order to calculate the following so
called “CT-value”:
  w
CT  value  ( t )k (1)
w
Where µt is coefficient of absorption at the
scanning point, µw is coefficient of water, and
k is a constant called Hounsfield unit. Hera,
this constant is fixed at a value of 1000. Thus
the CT-value of air should be -1000 because
the coefficient of absorption for air is zero.
Figure 2. Influence of MMT content on UCS at Likewise, this value for water is 0 from the
28 days definition of Eq. (1). It is well known that the
CT-value has a linear relationship with
3.2 X-ray CT method material density.
The CT-value of soil cement specimen is not
To clearly understand the affection of MMT, a so much different by changing MMT content,
series of X-ray CT is scanned for soil cement but large change when increasing the total
specimens at initial the unconfined water-to-cement (Figure 5 and Figure 7). The
compression test for 28 days curing, which CT-value of soil cement specimen changed the
specimens are made from kaolinite clay, mean value of 842, 853, 844, 836 when MMT
bentonite content is 0%, 5%, 10%, 15% (MMT content change 0%, 3.3%, 6.5%, 9.8%
content is 0%, 3.3%, 6.5%, 9.8% respectively), respectively with wT/c=3; Otherwise with
cement content is 20%, total water-to-cement, wT/c=5, this value change 653, 686, 675, 668 .
wT/c is 3 and 5. The position of X-ray CT is The result of X-ray CT method showed the
shown in Figure 3. density of soil cement decrease when the soil
has much MMT mineral, this reason affect to
unconfined compressive strength of soil
cement specimens.

Figure 3. Influence of MMT content on UCS at

28 days

In this system of X-ray CT scanner used in

this research, the collimated X-ray penetrates
from all around the circumference of the
specimen by rotating and traversing the
specimen table. The detected analogy data are
changed to digital values and then, a
coefficient of X-ray absorption in each spatial
unite called a “voxel” which is unit volume Figure 4. Density of specimens by X-ray CT
whose size is 0.293mm  0.293mm square, image, wT/c=3

229
 in each sample. This result shows a linear
relationship among all the results, so the
change of the density of the specimen can be
discussed with the CT-value.

Figure 5. CT-value of specimens, wT/c=3

Figure 8. CT-value-density relationship for DCM

The failure mode of DCM specimen is

brittle failure. The maximum strain is 1.5%.
The X-ray CT images clearly see the cracks in
sides DCM specimens are shown in Figure 9.

Figure 6. Density of specimens by X-ray CT

image, wT/c=5

Figure 9. The cracks in sides DCM specimens,

wT/c=5

3.3 Laboratory model test

The main reason for model test is the difficulty to
perform extensive quantitative field tests of
reasonable costs. With laboratory model tests the
most important factor affecting the behaviour of
Figure 7. CT-value of specimens, wT/c=5 the columns can be determined. In order to confirm
the validity of the proposed homogenization
Figure 8 shows the same relationship for method, 1g-model test of the improved ground
DCM specimens, in which all the samples were with soil cement columns are conducted. The
scanned at 20 different height and the average model is designed based on the models of Ailin
CT-value was calculated using these CT-value Nur I.o, Hafez M.A., Norbaya S. (2011), K.Omine,

230
H.Ochiai (1999), Kitazume, M., Okano, K., and The water content of the consolidated soil
Miyajima, S. (1999), Mirja Kosche (2004), was approximately from 49% to 50% and the
S.Larsson (1999). The test box is shown wet unit weight was from 17.2 to 17.4kN/m3.
schematically in Figure 10 . After the consolidation was completed, the
surface of soft soil was flattened through a
trimmer and some of the soil was removed to
bring its height 200mm. The soil was ready for
constructing soil cement columns inside
(Figure 15a).
The soil cement columns were prepared
using a technique similar to the ones used to
construct bored piles. Before drilling, the guide
plate that was used to assure the accuracy of
geometry was placed carefully on top of the
consolidated soil (Figure 15b). The auger of
the drill is continuous with the diameter of
20mm and the length of 300mm (Figure 15c).
The hole is kept by plastic tube (Figure 15d).
The columns material was injected into pre-
bored using a injection pump (Figure 15e). The
Figure 10. Schematic of laboratory model test holes were filled up through inserting the
300mm long hose with 15mm outer diameter
In model ground preparation, dry clay
to the bottom of each hole (Figure 15f).
power with 0%, 5%, 10%, 15% bentonite in
weights was mixed with freshwater to slurry
a water content of approximately 80%
(Figure 11). This slurry was poured in the
test box with 300 mm in height (Figure 13).
They had a draining sand layer in the bottom
(Figure 12). Pressure was applied at the top
to consolidate the slurry. The consolidation
pressure was increased in step 0kPa, 2.5kPa, (a) (b)
5.0kPa, 10.0kPa (Figure 14).

(c) (d)
Figure 11. Slurry Figure 12. Sand layer

(e) (f)
Figure 13. Soil layer Figure 14. Consolidation
Figure 15. The process of making soil cement columns

231
 Continue done with 31 other columns. Finally, MMT content 6%. Otherwise, only increase
the stabilized soil in laboratory model was kept 1.25 times with MMT 15%. This result is
in room temperature for curing 28 days. shown that: The MMT mineral in soil is the
The system test was performed using large affection to improved soil by Deep
loading mechanism that was specifically Mixing Method.
designed for this research. It is composed of an
oil pressure supplier, a pressure transducer, a
loading piston and a controller (Figure 16). The
bearing capacity of stabilized soil is the
maximum axial load attained per unit area or
the load per unit area at 5% axial strain,
whichever occurs first during performance of a
test (Figure 17).

Figure 18. The bearing capacity of stabilized and

unstabilized soft soil

4. CONCLUSION
Based on the study described above, the
conclusions of this study are listed as follow:
The unconfined compressive strength of soil
Figure 16. Laboratory model
cement specimens decreases almost linearly
with the increasing of MMT contents. Based
on the Figure 1, the optimum cement content
can be determined when the soft soil has
different MMT contents.
The CT-value of soil cement specimen is
decreased when MMT content increase, it
means the density of soil cement decrease
when the soil has much MMT mineral. The
unconfined compressive strength decreases by
decreasing of CT-value. The X-ray CT images
clearly see the cracks in sides DCM specimens,
the failure mode of DCM specimen is brittle
Figure 17. The assembled system of test failure.
The bearing capacity reduces when the
The results of research is showed that with
MMT content increased.
the same cement content, the bearing capacity
reduces when the MMT content increased
(Figure 18). When MMT content increases 9% 5. REFERENCES
(from 6% to 15% then the bearing capacity of Ailin Nur I.o, Hafez M.A., Norbaya S. 2011. Study
stabilized soil will decrease 2.6 times (from of Bearing Capacity of Lime-Cemant
66kPa to 25kPa). The bearing capacity of Columns with Pulverized Fuel Ash for Soil
stabilized soil increases 3.3 times to Stabilization Using Laboratory Model. EJGE,
unstabilized soil, from 20kPa to 66kPa with 16, 1595-1605.

232
James L. Post, Richard L. Sloane. 1971. Clays and Phan Thị San Ha, Le Minh Son, Nguyen Hoang
Clay Minerrals, 19, 21-29. An. 2007. Classification of clay mineral by
K.Omine, H.Ochiai. 1999. Homogenization Methylene Blue Absorption test (MBA).
method for numerical analysis of improved Science & Technology Development, 10, 07-
ground with cement-treated soil columns. 2007 (in Vietnamese), 63-74.
International Conference on Dry Mix S. Larsson. 1999. Shear box apparatus for
Methods: Dry Mix Methods for Deep Soil modelling chemical stabilised soil –
Stabilization, Rotterdam Balkema, 161-168. Introductory tests. International Conference
Masaki Kitazume, Minoru Yamamoto, and on Dry Mix Methods: Dry Mix Methods for
Yasushi Ukdaka. 1999. Vertical bearing Deep Soil Stabilization, Rotterdam Balkema,
capacity of column type DMM ground with 115-121.
low improvement ratio. International Terashi, M. 1997. Deep Mixing Methods – Brief
Conference on Dry Mix Methods: Dry Mix state – of-the-art, 14th International
Methods for Deep Soil Stabilization, Conference on Soil Mechanics and
Rotterdam Balkema, 245-250. Foundation Engineering.
Mirja Kosche. 2004. A laboratory model study on Author Information:
the transition zone and the boundary layer 1. Full Name: Ngoc Thang NGUYEN
around lime-cement columns in kaolin clay, 2. Prof/Dr/Mr/Ms: Dr
Division of Soil and Rock Mechanics, Royal 3. Country/University/Company: Vietnam /Tien
Institute of Technology, Stockholm, MSc Giang University /Civil Engineering Faculty.
Thesis, 62p. 4. Email: [email protected]
Nguyen Huu Chiem. 1993. Geo-Pedological Study
of the Mekong Delta, Southeast Asian Studies,
31, 2, 158-186.

233
The 3rd Int. Conf. on Transport Infrastructure & Sustainable Development (TISDIC 2019). Construction Publishing House. ISBN: 978-604-82-2893-4.

Effects of mineral composition and physical - mechanical

properties to calculate and design ground improvement
work in Quang Nam - Da Nang

Nguyen Thi Ngoc Yen1, Bui Van Truong2, Do Quang Thien3,

Nguyen Hoang Giang4
1
Faculty of Bridge and Road - DUT, Da Nang, Viet Nam, [email protected]
2
Thuy Loi University, Ha Noi, Viet Nam.
3
Hue Science University, Hue, Viet Nam.
4
Duy Tan University, Da Nang, Viet Nam.
*Email: [email protected]

ABSTRACT: The results on studying show: Mineral composition of soft soils include Quartz, Illite,
Kaolinite in majority; the content of organic of soils are low (<10%), tenderly deacrease with depth. The
main chemical composition of soils are SiO2; Al2O3; sand and silt particles occupy the majority.
Consolidation ratio Cv = 1,70 - 2,14 m2/year (silty mud clay) and Cv = 1,31 - 1,34 m2/year (mud clay).
Horizontal consolidation ratio, which represent the heterogeneous characteristics of soft soils:
m = Ch(tp)/Cv = 1,32 - 4,19; m = Ch(ap)/Cv = 1,82 - 1,86 (Silty mud clay); m = Ch(tp)/Cv = 1,09 - 4,14;
m = Ch(ap)/Cv = 1,52 - 1,91 (mud clay). Small deformation modulus (E0,5-1 ≤ 5000 kPa) and low shear
strength. Application calculation results show that: Mineral composition and physical-mechanical properties
of soft soils play an important role in deciding the ground improvement method and its effectiveness.
Keywords: Consolidation coefficient, soft clayey soils, ground treatment, soil improvement,
ratio m = Ch/Cv, Quang Nam – Da Nang.

1. INTRODUCTION detail through many works of many different

authors: Nguyen Viet Tinh (2001); Nguyen
In the calculation, design works, special works Manh Thuy (2005); Nguyen Thi Nu (2014), this
on soft ground. In addition to the normal is an important and valuable document used in
physical - mechanical properties of soft soil, the calculating, designing and choose solution for
mineral composition and cohesive properties, the ground treatment. However, in the central region
shear strength of soft soil plays an important role. in general and the coastal plain of Quang Nam -
On that basis, can choose solution for ground Da Nang in particular, this issue has not been
treatment, forecast on consolidation time, studied in depth by the authors. Therefore,
settlement, stable embankment height. Currently, through the research results, the authors analyze
in the world, the above problem is concerned by the effects of the mineral composition and the
many authors to research in depth with many physical - mechanical properties (P-MP) of soft
devices and on different models [Terzaghi. K soil in the coastal plain of Quang Nam -
(1924, 1940, 1941), Taylor.D.W (1940, 1942), Da Nang to choose the treatment solution as well
Bjerrum. L (1967), Casagrande. A (1938), it has as the effectiveness of the treatment solution
been recognized and applied in construction Quang Nam - Da Nang coastal plain multi-
standards (ASTM 2435, JGS 2000, BS 1377,…). origin soft soils have unstable distribution,
In Vietnam, especially the Northern Delta and alternative thickness and almost not expose to
Southern Delta, This issue has been studied in surface. Based on additional engineering

234
geological investigation [7] showed that there (Mekong Delta from 0,64 to 10,14% and
are 8 soft soils formations which had the Southern area of Ho Chi Minh City from 50 to
different composition, origin, time of forming. 60% [Nguyen Viet Tinh (2001), Nguyen Thi Nu
They are ambQ23, mbQ22, ambQ22, mQ22no, (2014)]) and show in figure 2.
abmQ21, amQ13(2) mQ13(2)đn, mlQ13(1)ht but However, the presence of organic humus
only four of them (ambQ23, mbQ22, ambQ22 and glue particles on the surface of soil particles,
abmQ21 sediments) have the larger distribution have increased the thickness of the water film
and relate to ground and foundation designing surrounded the clay particles. Therefore, the
work. So that, the author deeply studied for 4 water content is higher meanwhile the effective
formations. void ratio reduces, it means the permeability is
lower, preventing the compaction of soil and
2. STUDY RESULTS AND DISCUSSIONS influence to soil treatment, improvement by
vertical drainage.
2.1 Mineral composition (MC)
Mineral composition plays a very important
role, decides the thickness and viscosity of

Content of organic, %
pellicular water surrounding particle of soil,
hence it affects to mechanical properties,
especially shear strength and consolidation
properties of soft soils.
Content of mineral, %

Figure 2. Content of organic in soft soils of

QN-DN coastal plain and other areas

2.3 Chemical composition of soft soils

The main chemical composition in soft soils are
SiO2, Al2O3 having high ratio (Organic mud clay
Mineral composition of soft soils in the Northern Delta and Ho Chi Minh City:
Figure 1. Content of mineral in soft soils of SiO2 = 45,84 - 58,32%; Al2O3 = 18,36 - 24,59%;
QN-DN coastal plain Fe2O3 = 2,89 - 6,73%; FeO = 2,07 - 3,25%;
MKN = 11,56 - 13,99% [Nguyen Viet Tinh
The results on analysing of MC in figure 1 (2001), Nguyen Thi Nu (2014)] and show in
show: Main minerals included Quartz, Illite, figure 3. At normal condition, the dissolution of
Kaolinite in Quang Nam-Da Nang coastal plain SiO2 and Al2O3 is very poor, when the alkaline
have higher content than MeKong Delta (the environment has a high pH degree, those oxide
Northern Delta: Kaolinit 15%, Illite 15%, are easy to be dissolved lead to the destroy
Montmorilonit 5%, Quartz 38 - 40% [6]). The secondary minerals in soils, esspecially clay
popular of Illite as well as the appearance of minerals (Illite, Kaolinite and Montmorilonite),
Montmorilonite impact to plasticity, swellness thus the deformation – consolidation and shear
and deformation properties of soft soils, strength characteristics of soils will be improved,
consequently influence to ground improvement and enhance the soil improvement effect in the
ability. case of using vertical drainage solution. Morever,
SiO2 and Al2O3 in soluble form is a factor that
2.2 Content of organic
create a part of hardened bonding materials,
Content of organic in soft soils are low (<10%), which increases the strength of soil – cement
deacreasing trend with depth and less than soft mixtures once soil improvement by inorganic
soils of the Northern Delta and Southern Delta method is applied.

235
 consolidation coefficient greater than mud
clay. On the other hand, when treating the
ground with inorganic binders (cemeny - soil,
lime/cement - soil) then the increase in
content of clay particle reduces the ability to
mix inorganic binders into the soil, reducing
the strength of cement - soil because the clay
particles are small in size, large contact
surface area will require the function greater
Figure 3. Content of mineral in soft soils of QN-DN amount of cement to bond between cement
and soil.
2.4 Particle size of soft soils
2.5 Mechanical properties
In general, particle content > 2 mm in mud
clay, mud sandy clay accounts for a very low Material composition is a factor that plays an
rate (figure 4, 5). important role in determining the cohesion and
shear resistance of soft soil, including vertical
consolidation coefficient (Cv) and horizontal
Percentage of cumulative, %

consolidation coefficient (Ch). The higher the

content of clay minerals, organic matter and
the content of clay particles in the soil, will
reduce the drainage capacity of the soil,
slowing the consolidation process of the soil.
Vertical consolidation characteristic of soft
soils (Cv)
Grain size, mm The results on analysing of Vertical
Figure 4. Content of particles of mud sandy clay of consolidation characteristic in figure 6a and
QN-DN coastal plain and Mekong Delta figure 6b show: Soft soils having different
composition, state, properties, age or/and
origin bring the change of vertical
Percentage of cumulative, %

consolidation coefficient. In which, mud

sandy clay ambQ22 gave the highest vertical
consolidation coefficient (Cv = 2,14 m2/year)
while mud clay mbQ22 gave the lowest one
(Cv = 1,31 m2/year).
Deformation - consolidation parameters of
soft soils are one of important properties used
Grain size, mm to choose, calculate and design ground
improvement treatment method. Deformation-
Figure 5. Content of particles of mud clay of QN-
DN coastal plain and other areas consolidation characteristics closely depend
on MC of soft soil. Percentage of sand and silt
The dominance of the content of clay and occupy in majority; content of Quartz,
silt particles in mud clay soil will reduce its Kaolinite, Illite of soft soils in researched area
permeability, extend the settlement time of are more than Northern Delta and Southern
embankment. The reduction of clay and silt Delta ones; therefore, consolidation
particles, and increase of sand grains makes coefficient are higher while compression
mud sandy clay with permeability and coefficient are lower.

236
 3.75
3.50
The back analysis result after in-situ
3.25 settlement monitoring data showed that soft
3.00
2.75 clayey soils (mud sandy clay, mud clay) in the

Cv, m2/year
2.50 researched area has the horizontal consolidation
2.25
2.00 coefficient Ch(ap) = 1,99- 3,08 m2/year. This
1.75 value matches other studies in Viet Nam and
1.50
1.25 some countries in the world: soft soils in Cuu
1.00
0 50 100 150 200 250 300 350 400
Long river plain Ch(ap) = 1,27 – 3,03 m2/year
P, kPa [Nguyen Thi Nu (2014)]; soft clayey soils of Ba
Ria – Vung Tau Ch(ap) = 2,06 m2/year [Nguyen
Figure 6a. Vertical consolidation coefficient of
Duy Quang (2010)].
mud sandy clay
Besides, the ratio m of studied soils change as
follow: m = Ch(tp) /Cv = 1,95 – 2,34 m2/year; m =
Ch(ap)/Cv = 1,52 – 1,91 m2/year. These results are
also suitable with other achieved researches in
Viet Nam as well as all over the world: Soft clay
in Japan m=1 (Tan, Chew, 1996 [Tan.S.A and
Chew.S.H (1996)]; Soft clay in Singpore,
Sweden m=1,5 -3 [Tan.S.A and Chew.S.H
(1996)]; Soft clay in Ha Noi m=1,06 - 4,01
(Nguyen Viet Tinh, 2001 [Nguyen Viet Tinh
(2001)]); Soft clay in Cuu Long river plain
m=1,35-5,61 [Nguyen Thi Nu (2014)],...
Figure 6b. Vertical consolidation coefficient of
mud clay

Horizontal consolidation characteristic of

soft soils (Ch)
Horizontal consolidation coefficient used to
calculate and design ground improvement
solution by vertical drain method in order to
increase consolidation process of soft soils
after treatment. The results on analysing of
horizontal consolidation characteristic in figure
7a, 7b, 7c show: Figure 7a. Horizontal consolidation coefficient
Horizontal consolidation coefficient and the through other compression pressures
ratio m between in-situ settlement monitoring
result (Ch(ap)) and in-door test (Ch(ap)) are In general, our study on soft clayey soils in
different. For mud sandy clay (ambQ23, Quang Nam - Da Nang coastal plain is
mbQ22), Ch(ap) = 2,51 – 3,08 m2/year and Ch(tp) compatible with variuos researches not only in
= 2,25 – 2,47 m2/year; while for mud clay Viet Nam but also over the world. Thus, it is
(ambQ23, mbQ22), Ch(ap) = 1,99 – 2,73 m2/year suggested to use as reference documents for
and Ch(tp) = 2,14 – 2,40 m2/year. These results designing work of soil improvement.
may be affected by taking, preservating, The horizontal consolidation coefficient of
preparing, or testing samples processes that soil depends on clay mineral content, organic
could make the natural structure or physical matter, clay particle content. The higher the
properties of soil samples had changed. content of clay minerals, organic matter and clay
However, the distinct value between Ch(ap) and content in soft soil, the slower the consolidation
Ch(tp) is not much, so that our study is highly process of the soil, ie the cohesion coefficient of
dependable. the soil will be low.

237
 undrained cohesive force
Unconsolidation –
Soft soil

Figure 7b. Ratio m = Ch(tp)/Cv through other Figure 8. Unconsolidation - undrained cohesive
compression pressures force of soft soils in QN-DN coastal
plain and other areas
Consolidation coefficient

undrained cohesive force

Consolidation –
Soft soils
Note: MSC – Mud sandy clay ; MC– Mud clay Soft soil

Figure 9. Consolidation - undrained cohesive force

Figure 7c. Vertical consolidation coefficient (Cv) of soft soils in QN-DN coastal plain and
and horizontal one (Ch(tp)) and after Mekong Delta
Asaoka’s one (Ch(ap))
Once soils get the consolidation ratio U =
2.6 Shear strength of soft soils 95%, their shear strength increase considerably
Shear strength of soft soils when (figure 9): Mud sandy clay ambQ23 increase
Unconsolidation - undrained cohesive force 46,99%; mud sandy clay mbQ22 28,84%; mud
shown in figure 8 are very low: Mud sandy sandy clay ambQ22 65,04% and mud sandy clay
clay: cuu= 5,7-7,9 kPa,  uu =0035’-0048’; Mud abmQ21 74,56% alternately. Mud clay amQ13(2)
increase 38,85% and mud clay mbQ22 59,53%.
clay: cuu=7,1-7,2 kPa,  uu =0031’-0035’;
The increase of shear strength (ccu,  cu ) make
The increase shear strength of soil increases
constructive properties of soft soils raise
the bearing capacity of the ground, while
effective when treating ground. Morever,
improving the efficiency of soft ground
effective shear strength of soils reach high
improvement. Besides, the shear strength of
values: c’=5,1-8,2 kPa;   17 0 57'230 28' (for mud
soft soil also depends significantly on the
material composition of the soil, the main sandy clay) and c’=6,4-7,3 kPa;   16 0 52'19 0 02'
existence of clay minerals, the presence of (for mud clay).
organic matter and the content of clay particles
2.7 Recommendation to use physical-
will reduce the unit cohesive force of soil.
mechanical properties (P-MP) of soft soils
However, the dominance of sand and silt
served to road design work
particles will increase the internal friction
angle of the soil. When the content of clay Based on above results, the author suggests to
particles and plastic index increases, the use P-MP of soft soils in Quang Nam-Da Nang
effective friction angle of the soil decreases but coastal plain in calculate, design ground
the strength of the effective unit cohesive force improvement work by vertical drain method
of the soil increases and vice versa. displayed in table 1.

238
Table 1. Reccommendation to use physical-mechanical properties use to calculate and design
ground improvement work
Parameters Mud sandy clay Mud clay
(ambQ23) (mbQ2 ) 2
(ambQ2 ) 2
(abmQ2 )1 2
(mbQ2 ) (amQ13(2))
38,9-45,8/ 40,0-50,0/ 43,0-53,1/ 46,3-50,1/ 55,3-59,6/ 50,1-52,1/
W, %
43,3 43,1 46,7 48,4 57,5 51,4
1,15-1,29 1,11-1,37/ 1,25-1,44/ 1,29-1,48/ 1,14-1,64/ 1,44-1,52/
Physical properties

e0
1,25 1,22 1,32 1,40 1,53 1,48
35,3-41,8/ 36,1-41,7/ 38,4-44,2/ 38,1-40,2/ 47,3-50,4/ 42,0-45,1/
LL,%
39,6 38,5 40,9 39,2 48,7 43,3
10,5-16,9/ 11,9-16,9/ 11,0-19,3/ 13,4-16,2/ 20,1-26,0/ 18,2-19,2/
PI,%
13,6 15,1 15,2 14,8 22,6 18,6
1,14-1,43/ 1,18-1,50/ 1,17-1,65/ 1,52-1,77/ 1,30-1,51/ 1,38-1,51/
IL
1,28 1,31 1,39 1,63 1,39 1,44
Cv, 1,63-2,78/ 1,38-1,91/ 1,62-2,66/ 1,61 - 2,15/ 1,08 - 1,46/ 1,36 - 1,51/
m2/year
Consolidation

1,89 1,70 2,14 1,84 1,31 1,43

coefficient

Ch, 1,80-3,88/ 1,96-5,42/ 1,61-3,75/ 1,89-4,84/

m2/year - -
2,25 2,47 2,14 2,40
Ch/ Cv 1,25-4,19/ 1,32-3,51/ 1,12-3,55/ 1,09-4,14/
- -
2,17 1,95 2,22 2,34
4,9 - 14,1/ 6,9 - 16,4/ 6,5 - 12,4/ 9, 1 - 11,9/ 12,4 - 15,4/ 10,9 - 13,2/
a1-2, kPa-1
10,5 11,7 13,8 10,7 13,8 12,3
Deformation parameter and

e0 1,15-1,29 1,11-1,37/ 1,25-1,44/ 1,29-1,48/ 1,14-1,64/ 1,44-1,52/

pressure history

1,25 1,22 1,32 1,40 1,53 1,48

Pc, kPa 42,7 - 83,3/ 56,6 - 67,3/ 47,9 - 55,9/ 56,9 - 79,4/ 52,4 - 66,1/ 42,9 - 59,8/
56,8 61,5 52,6 62,3 57,9 45,5
0,459-0,653/ 0,362 - 0,556/ 0,511 - 0,731/ 0,583 - 0,711/ 0,517 - 0,889 0,485-0,641/
Cc
0,540 0,441 0,610 0,660 0,688 0,512
0,063-0,082/ 0,049 - 0,067/ 0,068 - 0,096/ 0,061-0,092/ 0,077 - 0,104 0,071-0,098/
Cr
0,073 0,057 0,082 0,081 0,09 0,088
 uu 0028’-1019’/ 0030-0056’/ 0033’- 0050’/ 0031’-0038’/ 0029’-0055’/ 0028 - 0032’/
strength, UU

degree 0048’ 0040’ 0039’ 0035’ 0035’ 0031’

Shear

6,9 - 9,0/ 5,8-8,2/ 5,4-6,0/ 6,8 - 7,9/ 7,0 - 7,3/

Cu,kPa 7,0- 8,7/7,9
7,9 7,2 5,7 7,2 7,1
VST Su, kPa 8,9-9,6 10,2 9,9-10,4 9,4-10,7 9,7-10,1 9,7-10,3
12010’-
11003’-13004’ 11037’-13014’ 11014’-12045’ 11009’-11036’ 10018’-11054’
 cu degree 12048’
12016' 12031' 11035' 11028' 11006'
12011'
Shear strength, CU

11,1-12,2 9,6-10,8 10,3-13,4 8,8-11,1 10,5-12,6 10,5-11,3

Cu, kPa
11,7 10,1 11,9 9,9 11,6 9,9
16043’-
18052’-22005’ 18055’-23028’ 19018’-20046’ 16022’-17057’ 16024’-19002’
 degree
'
17006’
20033' 19019' 20002' 17010' 17043'
16048'
7,1-8,1 6,7-8,2 7,2-7,8 5,1-7,3 6,4-7,3 6,6-6,8
C’ kPa
7,67 6,8 7,5 6,2 6,9 6,6

239
3. APPLY TO CALCULATE A REAL 1: from Km0+000 to Km8+000 and package 2:
BUILDING from Km8+000 to Km 16+880).
- Stability and settlement of road ground
Based on the resuts on MC and P-MP of soft untreated
soils, the author select the parameters of soilt to On all specific calculation profile,
calculate and design the ground improvement consolidation settlement and time to settle are
method for 2 buildings. Those are Nguyen Tat greater than specification on building’s
Thanh road project (Km0+00 to Km5+987,5) settlement (figure 9 and table 2).
and Da Nang - Quang Ngai expressway (pakage

Figure 9. Settlement analysis before treatment by Plaxis 8.5 at MC1

Table 2. Synthesize of analysed results on settlement of ground not treated

Calculated by Calculated by Shear

Thickness Excel Plaxis
HTK, HTT, stability
N0 Project Profile of soft
m m coefficient
soil (m)
Sc,(m) t,(month) Sc,(m) t,(month) Plaxis 8.5

1 MC1 17,20 2,95 4,19 1,12 130,7 1,12 1,20 1,32

Nguyen
2 Tat Thanh MC2 14,00 3,05 4,16 1,01 108,5 1,57 1,03 1,12
road
3 MC3 10,10 2,88 3,83 0,86 225,9 1,65 0,88 1,57

4 MC4 9,50 6,00 7,05 0,95 324,3 1,67 1,02 1,65

5 DN - QN MC5 5,30 6,80 7,38 0,53 110,6 1,61 0,62 1,67

expressway
6 (pakage 2) MC6 9,80 7,90 9,08 1,07 272,8 1,42 1,12 1,61

7 MC7 10,84 7,50 9,08 1,44 360,7 0,94 1,52 1,42

8 DN - QN MC8 21,50 4,50 5,61 1,01 1142,5 1,48 1,04 0,94

expressway
1,48
9 (pakage 1) MC9 21,00 4,70 5,64 0,85 1716,6 0,88

240
 - Stability and settlement of road ground Table 4. Calculating results on treatment by
treated sand well
The authors selects the wick drain (figure
Settlement
10) and sand well method (figure 11). Stability
(cm)
Calculating results on treatment by wick drain correspond to coeffici
method displayed in table 3 show that they Project Profile U=90% ent
meet the requirement of residual settlement and Plaxis
allowable time for ground treatment. Excel Plaxis 8.5

Table 3. Calculating results on treatment by

MC1 96,0 106 1,62
wick drain method Nguyen Tat
Thanh road MC2 87,0 93,1 1,72
Settlement MC3 74,0 76,2 1,85
(cm) Stability
correspond to coeffici MC4 82,0 84,1 2,07
Project Profile U=90% ent DN - QN MC5 46,0 49,4 1,91
Plaxis expressway
8.5 (pakage 2) MC6 93,0 97,3 1,87
Excel Plaxis
MC7 129 135 1,93
MC1 101 104 1,57 DN - QN MC8 88,0 91,3 1,44
Nguyen Tat expressway
MC2 90,0 97,1 1,65 MC9 79,0 82,4 1,92
Thanh road (pakage 1)
MC3 - - -
MC4 85,0 87,3 1,94
DN - QN MC5 48,0 52,2 1,87
expressway
(pakage 2) MC6 96,0 101 1,81
MC7 129 140 1,85
DN - QN MC8 91,0 95,3 1,41
expressway
(pakage 1) MC9 83,0 84,2 1,84

Figure 11. Calculating model of treatment by sand

well method in Plaxis 8.5 at MC1

Compare with monitoring results

The results of designing calculation for ground
treatment as follows (table 5): STXL= 120 cm,
SPVD = 104 cm; TPVD= 283 days; SSD = 106 cm;
TPVD= 304 days. So, the calculated results on
researched P-MP determined in this study are
Figure 10. Calculating model of treatment by wick more appropriate than real monitoring data
drain method in Plaxis 8.5 at MC1 (SQT = 115,9 cm; TQT= 342 days). The quite
large difference between calculation of designing
Calculating results on treatment by sand document (STXL= 76,4 cm; TXL = 264 days) and
well displayed in table 4 show that they meet the realiaty may be caused by the usage, selection
the requirement of residual settlement and P-MP of soft soils, especially consoliadtion
allowable time for ground treatment. characteristics, shear strength of soil.

241
Table 5. Comparison of calculating result by different method

Settlement before Designing treatment

Real treatment result
treatment,cm solution
Calculation PVD SD
Profile
Designing SPVD, TPVD SSD, TSD, Treatment THSTK TQT SQT,
docs Plaxis cm , cm method , day day cm
Excel day
8.5
day
MC1 76,4 112 120 104 283 106 304 PVD-1,3T 264 342 115,9
MC2 87,3 101 103 97,1 289 93,1 308 PVD-1,3T 276 339 106
MC3 50,0 86 088 - - 76,2 276 PVD-1,3T 246 306 -
MC4 113,4 95 102 87,3 478 84,1 427 SD-1,9V 462 705 102
MC5 61,43 53 062 52,2 431 49,4 440 SD-1,9V 450 437 59,1
MC6 64,89 107 112 101 471 97,3 449 SD-2,0V 430 605 111,2
MC7 70,50 144 152 140 462 135 441 SD-1,9V 435 702 152,1
MC8 115,7 101 104 95,3 469 91,3 459 SD-2,0V 424 727 102,8
MC9 99,0 085 088 84,2 436 82,4 433 SD-1,9V 463 727 91,1

4. CONCLUSIONS 5. REFERENCES
Mineral composition of soils include Quartz, Tan.S.A and Chew.S.H, 1996, Comparison of the
Illite, Kaolinite in majority; the content of hyperbolic and Asaoka observational method
organic of soils are low (<10%), tenderly monitoring consolidation with vertical drains,
deacrease with depth. The main chemical Soils and Foundations, Vol.36, No.3, pp. 31-42.
composition of soils are SiO2; Al2O3; sand and Nguyen Duy Quang, P.H.Giao, T.Seah, 2010,
Settlement calculation and back – analysis of
silt particles occupy the majority. They are
soil properties for a test embankment on a soft
important role in deciding the ground clay ground improved by PVD and vacuum –
improvement method and its effectiveness, as assisted at a site Vung Tau, Vietnam, Viet
follows: Nam Geotechnical journal, No.14/2010.
The existence of mineral Monmorilonite Report on settlement monitoring of project “Da
with high content reduces the effectiveness of Nang – Quang Ngai Express way, Package A1
treatment solution. (Km0+000÷Km8+800); Package A2
Organic compounds reduce permeability, (Km8+000÷Km16+800), Package A4
prevent soil compaction and affect the ability (Km21+500÷Km32+600). Da Nang priority
to improve soil when using vertical drainage infrastructure investment project: Lengthen
solutions. Nguyen Tri Phuong Street (Km 2+715,04
Sand and silt particles occupy the majority ÷Km 5+987,50) and approach road to Khue
Dong bridge.
will reduce permeability, extend the settlement
Nguyen Manh Thuy, 2002, Selection suitable
time of embankment. technical solution for soft soil improvement in
Consolidation parameters and shear Southern of Ho Chi Minh city, geological
strength of soft soils depend on the change doctor of philosophy thesis.
of mineral composition. The soft soils have Nguyen Viet Tinh, 2001, Engineering geological
consolidation coefficient and shear strength features of lower – middle Holocene sediment
are low, so the treatment time is long, originated lacustrine – swampy of Hai Hung
reducing the effectiveness of the treatment formation (lbQ21-2hh). Assessment the
solution. application ability and prediction the change
under engineering work and urban

242
 development, example for Ha Noi area,
geological doctor of philosophy thesis. Mining
– Geology University.
Nguyen Thi Nu, 2014, Research on engineering
geological features of soft clayey soil amQ22-3
located in Cuu Long river plain for road
ground improvement, geological doctor of
philosophy thesis. Mining – Geology
University.
Nguyen Thi Ngoc Yen, 2018, Research of physic-
mechanical properties of soft soil in Quang
Nam – D Nang coast plain for designing road
engineering, Technical Doctoral thesis, Thuy
Loi University, Ha Noi, Viet Nam.

243
The 3rd Int. Conf. on Transport Infrastructure & Sustainable Development (TISDIC 2019). Construction Publishing House. ISBN: 978-604-82-2893-4.

Effect of unsaturation on the liquefaction of Soil: Case

study of dense fine clean sand

Tran Khai Hoan 1,, Saber IMANZADEH1,2, Said TAIBI1, Hanène SOULI3,
Jean-Marie FLEUREAU4, Mahdia HATTAB5, Luong Nguyen Hoang Phuong 6
1
Laboratory UMR CNRS 6294-University of Le Havre Normandy, Le Havre, France
2
Normandie Univ., INSA Rouen Normandie, Laboratoire de Mécanique de Normandie, 76801 Saint-
Etienne du Rouvray, France
3
Laboratory UMR CNRS 5513, Saint Etienne, France
4
Laboratory MSS-Mat, UMR CNRS 8579, Centrale Supelec, Saclay, Paris, France
5
University of Lorraine, Nancy, France
6
Faculty of Civil engineering, DongA University, DaNang, VietNam


Corresponding author, E-mail: [email protected]; [email protected].

ABSTRACT: This paper aims to study the behavior of dense fine clean sand subjected to dynamic loading in
the tri-axial cell. RF Hostun sand, a material commonly used in geotechnical laboratory tests in France, was
used in these experiments. The wet tamping method was used to prepare the samples. All the samples were
firstly saturated by circulating de-aired water from the bottom to the top to have different degrees of
saturation. At the end of this step, Skempton parameter B was measured. Thereafter, the samples were
consolidated isotropically and then subjected to cyclic deviator stress to study liquefaction. If after a given
number of cycles, the sample is not liquefied, the stress amplitude will be increased. The liquefaction
susceptibility of the sand is evaluated based on the value and the number of the cyclic loading causing
liquefaction. All tests show that, under cyclic loading, the samples were liquefied; however, the level of
deviator stress causing liquefaction depends on the degree of saturation. The samples with the same degree of
saturation were liquefied with the same level of deviator stress; nevertheless, there is a great difference in the
number of the load cycle. This consequence suggests that besides the void ratio and the saturation degree, the
different soil fabrics (caused by sample preparation technique: wet tamping) can effect on the sand
liquefaction susceptibility and as well the triggering time of liquefaction in situ or in the laboratory.
Keywords: RF Hostun sand, skempton parameter B, liquefaction, unsaturation, cyclic loading.

1. INTRODUCTION liquefaction, cyclic liquefaction, and cyclic

mobility. The two first types relate to the loose
Soil liquefaction is a phenomenon known for a soils and the last one relates to the liquefaction of
long time and has caused a lot of damages over dense soils. The most accepted criterion for
the world. The damage of liquefaction in the liquefaction in laboratory tests is that the sample
Indonesia earthquake in 2018 (Sassa et al. 2018) is liquefied if one of the following conditions
shows that until now, it is still a great danger for appears: pore water pressure increases to cell
human safety. In common usage, liquefaction is pressure leading to the loss of effective confining
usually related to saturated incoherent soils due stress; axial strain in one cycle reaches 5%.
to the increase of pore water results in the Many studies have focused on the
decrease of effective confining stress and in the liquefaction of saturated soils and the behavior
end, the soil behaves like a liquid. Liquefaction of saturated soils has been clearly understood
can be classified into three types: flow (Castro G 1969; Seed H B, Idriss I M 1982);

244
however, recent studies have shown that This material has following parameters:
liquefaction can be observed not only on specific gravity 2.65g/cm3, maximum grain
saturated soil but also on unsaturated sandy size 0.6mm, minimum grain size 0.12mm,
soils. Tsukamoto et al. (2014) showed that the friction angle 40o. Other parameters are shown
air bubbles can be found at 5m below the in table 1. Where D10, D50, D60 are the particle
groundwater table, it means that most of the size distributions of RF Hostun sand, e is the
building structures are based on unsaturated initial void ratio of the sample.
soil layers. The theoretical study of Martin et
al. (1978) predicted the effect of saturation
degree on the cyclic liquefaction resistance of
sands. Whereby, the unsaturated sands can be
liquefied and a small change of saturation can
causes a significant change in the cyclic stress
ratio causing liquefaction. This finding has
been confirmed by the laboratory results
(Chaney 1978; Yoshimi et al. 1989; Fourie et
al. 2001; Tran et al. 2018). Mulilis et al. (1978)
investigated the liquefaction of Monterrey sand
Figure 1. RFHostun sand
with the effect of Skempton’s coefficient B and
they found that the resistance of sand does not Table 1. The parameters of RF Hostun sand.
change when the value of B is between 0.91
and 0.97. Della et al. (2011) demonstrated that D50 D10 D60 emax emin
the dilatancy and the contractancy of soils (m) (m) (m)
change when the Skempton’s pore pressure 300 200 400 1.041 0.648
coefficient B decreases; however the influence
is not the same for all soils. It depends on the
soil type, initial density and confining pressure.
Although laboratory studies on the effect of
saturation degree have achieved some progress,
it is necessary to have more results to make
clear the combined effects with other factors
such as initial relative density, effective
confining stress, and loading path. This paper
presents three tests (Table 2) to study the
behavior of RF Hostun sand, the fine clean
sand usually used in geotechnical laboratory Figure 2. Comparison of grain size distribution of
tests in France. The first two experiments were RF Hostun sands to other liquefiable
carried out on fully saturated samples while the sands (Iwasaki 1986)
last experiment was performed on unsaturated
samples. Through these tests, the influence of The apparatus used for dynamic tri-axial
saturation degree and different soils fabric tests was used in this study (Figure 3). The
caused by sample preparation are studied. Dynamic (Cyclic) Triaxial Testing System
5Hz/5kN provided by VJ tech includes a
2. MATERIAL AND METHOD dynamic controller to generate and control
dynamic parameters, i.e. force, displacement
The material is fine quartz sand (RF Hostun) and pore water pressure. The cell pressure and
from Sika Co. The sand with its microscope back pressure are controlled by a Pneumatic
particle shape (Nadia 2001) is shown in figure Automatic Pressure Control (APC) device and
1 and the grain size distribution of the material a Hydraulic APC device, respectively. The
is presented in figure 2.

245
pressure imposed by the Pneumatic APC is 3.2 Sample saturation.
transmitted to the cell through an Air-Water
With the saturated tests, firstly, a cell pressure
Interface. This device’s duty is to make the
of 35kPa had been applied, and then, the
pressure more responsive and mellifluous, It
sample was circulated by de-aired water. The
includes a rubber ball embed in a cell. The
back pressure of 15kPa was applied to the base
pressure generator controls the air pressure
of the sample; de-aired water surged from
inside the rubber ball instead of cell pressure
bottom to top and went out of the sample. With
directly. The specimens used are 70 mm in
saturated tests, this process finished when there
diameter and 140 mm in high. Back pressure is
were no air bubbles in the outlet pipe. After
applied on both top and bottom of the sample
that, the cell pressure and back pressure were
while the pore water pressure is measured only
increased slowly to 770 and 750kPa
at the bottom of the sample. The displacement
respectively to dissolve the remaining air in the
can be controlled by the dynamic controller
sample (figure 4). In this process, the effective
(this device makes and controls dynamic load)
stress was kept equal to 20kPa. With the
or the axial displacement transducer.
unsaturated tests, the sample was circulated by
de-aired water until achieving the saturation
degree desired.

Figure 4. Cell and back pressure enhancement to

dissolve the remaining air bubbles inside
Figure 3. Triaxial dynamic test apparatus the sample

3. TEST PATHS Skempton’s coefficient B was used to check

the saturation of the sample. To do this
3.1 Sample preparation by moist tamping experiment, first, all the back pressure valves
are locked. Then, the cell pressure is increased.
The moist tamping method was used to prepare
This increase in pressure results in an increase
the sample with low densities. Firstly, the sand
in the pore water pressure. The ratio between
was dried and then a fixed quantity of water
the measured pore pressure increase and the
was added to have a mixture with the initial
imposed cell pressure increase is defined as
water content of 8%. This water content results
Skempton’s coefficient B:
in capillary forces between sand particles and u w
allows reaching low-density not achievable by B (1)
 3
other methods. The capillary forces are also
useful in keeping the sample stable after Where 3 and uw are the imposed
removing the split mold. The sample size is 70 increment of confining stress and the resulting
mm in diameter and 140mm high and initial measured increment of pore water pressure,
void ratio of 0.73. respectively.

246
 Because water is incompressible compared
to the soils skeleton, the sample is considered
to be fully saturated if B = 1. The appearance
of air bubbles with their compressibility
makes smaller B value. In practice, a B
between 0.95 and 0.97 is regarded as a signal
to indicate full saturation of sandy soils due
to some compliance in the test apparatus and
sample compressibility (Chaney 1978,
Jefferies 2016). With RF Hostun sand, the B Figure 5. Cyclic deviator stress with CSR = 0.15
= 0.97 is achieved in our laboratory when the
sample is fully saturated. This value was also The amplitude of the deviator stress was
used by Nadia (2011)and Arab (2016)for RF chosen initially to be 30 kPa corresponding to
Hostun sand as an indication of full cyclic stress ratio of 0.15 (load case 1). Cyclic
saturation state. stress ratio (CSR) is defined by the following
equation:
3.3 Isotropic consolidation c
qmax
To make the sample consolidate, firstly, all CSR  (2)
2 3' c
back pressure valves were closed and then
c
the cell pressure was increased to have the Where qmax / 2 is the maximum cyclic shear
difference of 100 kPa between the cell c
stress, qmax is the amplitude of deviator
pressure and the back pressure. This results
dynamic stress.  3c' is the effective
in the rising of pore water pressure. It takes
some minutes to have stable effective consolidation stress.
confining stress. With saturated tests, because After 100 cycles of the first cyclic load case,
B ≈1 the effective confining stress before and if the sample is still not liquefied the amplitude
c
after cell pressure increment was nearly equal of cyclic deviator stress qmax will be increased
while there was a large difference for to have CSR of 0.2 (load case 2), 0.25 (load
unsaturated experiments depending on the case 3), and 0.3 (load case 4). In case the
value of B. After having the stability of pore sample shows the tendency of liquefaction, the
water pressure, the back pressure valves were test will continue with unchanged CSR, so the
opened. The water went out of the sample number of cycles of the final load case may be
through both ends due to the imbalance greater than 100 (test 1).
between the back pressure and the climbed Table 2. Test information.
up pore water pressure. The cell pressure
gradually decreased until by the back Test B Dr Maximal CSR
pressure, the effective stress increased Test 1 0.97 79% 0.25
gradually and reached 100 kPa at the end of Test 2 0.97 79% 0.25
the process. The sample volume change
Test 3 0.7 79% 0.33
during consolidation was measured by
Hydraulic APC device in figure 2. Dr: relative density, B: Skempton’s coefficient B,
CSR: Cyclic stress ratio,
3.4 Dynamic deviator loading
4. RESUTLS
The saw teeth form load with a frequency of
0.1 Hz and 100 cycles were applied to the top 4.1 The results of saturated dynamic tests
of the sample. With this frequency, it took ten
Two saturated dynamic tests with the same
seconds to finish a cycle of load (figure 5).
initial conditions were carried out (test 1 and

247
test 2). After 100 cycles of the first and second Figure 7a shows the cyclic loading process
load case with deviator stress of 30 kPa and 40 of test 2 for load case 3 with CSR = 0.25. In
kPa, all samples were not liquefied or showed the first five cycles, the deviator stress
the signals of liquefaction. The CSR was fluctuates between 50kPa and -42kPa,
increased to 0.25 corresponding to deviator thereafter, the deviator stress amplitude starts
stress of 50 kPa. With this CSR the samples of decreasing. The test stops after 22 cycles
test 1 and 2 were liquefied. There are two when the maximum and minimum value of
criteria for liquefaction as mentioned above: the deviator stress amplitude are 30 kPa and
one is that the effective confining stress drops –30 kPa respectively. Similar to the first test,
to zero and the other is the axial strain  a  5 % at the end of the process, the amplitude of the
Figure 6a shows generally the variation of deviator stress is not zero.
deviator stress of test 1 when CSR = 0.25. At
the beginning of the process, the deviator stress
fluctuates between 50kPa and – 50kPa. After
60 cycles (corresponding to the time of 600
seconds) the deviator stress amplitude starts
reducing due to the building up of pore water
pressure resulting in the loss of soil resistance.
The test finished after 112 cycles. At the end of
the process, the amplitude of the deviator stress
is not zero.
Figure 7a. Deviator stress and pore water pressure
of test 2 when the CSR = 0.25

The pore water pressure caused by cyclic

loading in test 2 fluctuates (figure 7b). Its
amplitude reaches 800kPa, the value of cell
pressure, after 22 cycles.

Figure 6a. The deviator stress variation of test 1

when the CSR = 0.25
The development of pore water pressure due
to cyclic deviator stress was shown in figure 6b.
The pore water pressure increases from 720 kPa
to 800kPa. The sample is liquefied and the test
is stopped after 112 cycles when the maximal
pore water pressure equals to the cell pressure.
Figure 7b. The pore water pressure variation of
test 2 when the CSR = 0.25

The liquefaction of test 2 is shown more

clearly in the figure 8 by the results of the last
ten cycles. It is clear that the pore water
pressure builds up to cell pressure (800kPa) in
the 22nd cycle (The time is 220 seconds
corresponds to 22 cycles); however the pore
water is only equal to cell pressure at some
Figure 6b. The Pore water pressure variation of
test 1 when the CSR = 0.25
points and it is possible to conclude that the

248
liquefaction is cyclic mobility. At the end of Figure 10 shows the effective confining
the process, the deviator stress is stress change during cyclic loading of the last
approximately 35kPa significantly lower than ten cycles of test 2 (with CSR = 0.25). The
the initial value; nevertheless, the soil effective confining stress varies and is zero
resistance does not fall to zero suddenly. after 22 cycles when deviator stress is 35 kPa.

Figure 8. Deviator stress, pore water pressure and Figure 10. Effective confining stress and deviator
cell pressure of the last ten cycles of test 2. stress versus time of the last ten cycles
(CSR = 0.25) - saturated dynamic test 2
One of the standards usually used to verify
the liquefaction is excess pore water pressure The relationship between axial strain and
ratio ru . This ratio is calculated by normalizing deviator stress is also surveyed and put on the
the excess pore water pressure generated view of figure 11. The axial strain accumulates
during un-drained cyclic loading for the after each cycle and reaches -1.7% maximum.
effective consolidation stress as the following It means that the specimen has been stretched
equation: during cyclic loading.
u
ru  '
(3)
3c
Where u is the excess pore water pressure
and  3c' is the effective consolidation stress.
The maximum possible value for ru is 1.0 (or
100%), which occurs when the pore water
pressure equals to cell pressure or the effective
confining stress equals to zero and the sample
is considered to be liquefied. Figure 9 exhibits
clearly the variation of ru of test 2. This ratio
fluctuates due to the change of deviator stress Figure 11. Axial strain versus deviator stress of the
last ten cycles of saturated dynamic test 2
and finally equals to 1 after 22 cycles when the
deviator stress is almost 35 kPa. Figure 12 shows exactly what happens with
cyclic mobility phenomenon. The figure is
butterfly form. The q-p’ curve approaches the
failure criteria lines and fluctuates. where q is
deviator stress and p’ is effective mean stress.
The slope of the failure criteria line can be
calculated as equations (3) and (4). The lines
achieved from the test corresponds to effective
friction angle crit  37o. The deviator stress
Figure 9. Excess pore water pressure ratio and effective mean stress increases each time
variation due to the cyclic deviator stress the curve passes the phase transformation lines.

249
 6.sin crit The relationship between deviator stress and
M (Compression) (3)
3  sincrit effective mean stress is shown in figure 14. It
6.sincrit is noted that the slopes of the failure criteria
M (Extension) (4) lines in this figure are equal to the slopes of
3  sin crit
failure criteria lines in figure 12 (the failure
Where M is the slopes of the failure criteria
criteria lines for the saturated tests with
lines and crit is effective friction angle.
effective friction angle crit  37o ). The
Failure criteria lines deviator stress varies between 60 kPa and -50
kPa while the effective mean stress reduces
from almost 60 kPa to zero. When the curve
contacts the failure criteria lines, it changes
with the same slopes with these lines.

Phase transformation lines

Failure criteria lines

Figure 12. Deviator stress versus effective mean

stress of the last ten cycles (CSR = 0.25) of
saturated dynamic test 2.

4.2 The results of unsaturated dynamic test

The unsaturated test follows exactly the Figure 14. Deviator stress versus effective mean
procedure of saturated test. The difference is stress of the last ten cycles (CSR = 0.3)
only the degree of saturation. The sample with of unsaturated dynamic test 3.
B = 0.7 shows that it was not fully saturated.
Similar to the saturated tests, these pictures 5. CONCLUSION
below presents the results. In figure 13, the
This paper presents a laboratory study of
deviator stress of the first seven cycles varies
liquefaction on RF Hostun sand in both
between 60 kPa and – 50 kPa corresponding to
CSR = 0.3. This value is higher than the saturated and unsaturated state. Skempton’s
coefficient B = 0.97 was used to evaluate the
maximum CSR of the two above saturated
saturation degree of the sample. The saturated
tests. From 8th cycle, the amplitude of the
deviator stress starts declining. The sample tests have Skempton’s coefficient B = 0.97
while the unsaturated test has B of 0.7. The
liquefied after 23 cycles and the slope of the
investigation shows that the sample with a high
failure criteria lines is the same with this of
saturated tests (figure 14). relative density of 0.79 and Skempton’s
coefficient B of 0.97 can be liquefied under
cyclic loading.
Comparing the results of saturated tests and
unsaturated test, the conclusion is the decrease
of saturation results in the increase the
maximal cyclic stress ratio. The saturated
samples are liquefied by cyclic loading with
CSR of 0.25 while the unsaturated sample is
liquefied under cyclic loading with CSR = 0.3.
The CSR increases 0.05 (from 0.25 to 0.3)
Figure 13. Cyclic deviator stress of test 3 with when Skempton’s coefficient B decreases 0.27
CSR = 0.3 (from 0.97 to 0.7).

250
 The results of saturated tests show that, the International Consortium on Landslides,
although all samples have the same initial Doi 10.1007/s10346-018-1114-x.
conditions and the maximal CSR, there is still Seed, H.B., Idriss, I.M. 1982. Ground motion and
a big difference in the number of cycles soil liquefaction during earthquake, Berkeley,
causing liquefaction. This result leads to a University of California.
Tsukamoto, Y., Kawabe, S., Matsumoto, J.,
conclusion that, with the sands in the dense
Hagiwara, S. 2014. Cyclic resistance of two
state, besides the void ratio and the saturation unsaturated silty sands against soil
degree, the different soil fabrics (caused by liquefaction. Soils and Foundations.
sample preparation technique: wet tamping) Tran, K.H., Imanzadeh, S., Taibi, S., Souli, H.,
can play an important role when evaluating the Fleufeau, J.M., Bouchemelia, S., Pantet, A.
liquefaction susceptibility. 2018. Cyclic behavior of unsaturated Hostun
sand. 4th International Conference
6. REFERENCE Unsaturated Soils & Sustainable Construction
UNSAT, Oran - Algeria.
Arab, A., Sadek, M., Benkhatir, M. 2016. Tran, K.H., Imanzadeh, S., Taibi, S., Souli, H.,
Saturation Effect on Behavior of Sandy Soil Fleufeau, J.M., Pantet, A. 2018. Some
Under Monotonic and Cyclic Loading: A aspects of the cyclic behavior
Laboratory Investigation. Geotechnical and of quasi-saturated sand. 36èmes Rencontres
Geological Engineering. Universitaires de Génie Civil de l'AUGC.
Benahmed, N. 2001. Comportement mécanique France.
d’un sable sous cisaillement monotone et Yoshimi, Y., Tanaka, K., and Tokimatsu, K. 1989.
cyclique : application aux phénomènes de Liquefaction resistance of partially saturated
liquéfaction et de mobilité cyclique. Thèses sand. Soils Found, 29(3), 157–162.
doctoral.
Castro, G. 1969, Liquefaction of sands. Harvard
University, Cambridge.
Chaney, R.C. 1978. Saturation effects on the cyclic
strength of sands. Proc., ASCE Special Conf.
on Earthquake Engineering and Soil
Dynamics, ASCE, New York, 342–359
Della, N., Arab, A., Belkhatir, M. 2011. Static
liquefaction of sandy soil: an experimental
investigation into the effects of saturation and
initial state. ActaMech 218(1–2):175–186
Fourie, A., Hofmann, B., Mikula, R., Lord, E., and
Robertson, P. 2001. Partially saturated tailings
sand below the phreatic surface.
Geotechnique, 51(7), 577–585.
Iwasaki T. 1986. Soil liquefaction studies in Japan
(a state of art). Soil Dynamics and Earthquake
Engineering 5(1): 2-68. doi: 10.1016/0267-
7261(86)90024-2.
Jefferies, M., Been, K. 2016. Soil liquefaction – a
critical state approach. CRC press book. Taylo
and Francis Group, 472.
Mullilis, J.P., Townsend, F.C., Horz, R.C. 1978.
Triaxial testing techniques and sand
liquefaction. ASTM STP 654 DynGeotech
Test, 265–279
Sassa, S., Takagawa, T. 2018. Liquefied gravity
flow-induced tsunami: first evidence and
comparison from the 2018 Indonesia Sulawesi
earthquake and tsunami disasters, Journal of

251
The 3rd Int. Conf. on Transport Infrastructure & Sustainable Development (TISDIC 2019). Construction Publishing House. ISBN: 978-604-82-2893-4.

Experimental evaluation of erosion resistance capacity of

basic oxygen furnace slag-treated clay

Nguyen Anh Dan 1, Gyeong-o Kang2, Young-sang Kim3, Takashi Tsuchida4,

Young-hak Lee5
1
Department of Civil Engineering, Chonnam National University, Gwangju, 77 Yongbong-ro, Buk-gu,
Gwangju 61186, South Korea
2
Honam Regional Infrastructure Technology Management Center, Chonnam National University, 50
Daehak-ro, Yeosu, 59626, Jeollanamdo, South Korea
3
Department of Civil Engineering, Chonnam National University, Gwangju, 77 Yongbong-ro, Buk-gu, Gwangju
61186, South Korea
4
Department of Civil and Environmental Engineering, Graduate School of Engineering, Hiroshima University,
1-4-1, Kagamiyama, Higashi-Hiroshima, Hiroshima 739-8527, Japan
5
Institute of Agricultural Science, Chungnam National University, 99 Daehak-ro, Yuseong-gu, Daejeon 34134,
South Korea

Corresponding author: Gyeong-o Kang, Email: [email protected]

Address correspondence to First author: [email protected]

ABSTRACT: The main purpose of this study is to evaluate the variations in erosion resistance
capacity of basic oxygen furnace (BOF) slag-treated marine dredged clay under the several current
velocities with changing of water flow direction. The samples were artificially prepared with
different slag contents at various slag sizes. Erosion function apparatus (EFA) tests were carried out
to assess the erosion resistance properties of the BOF slag-treated clay. The results indicated that the
erosion rates and shear stresses () show the highest correlation with the undrained shear strength (Su)
of the sample. It was revealed that there is optimum slag content (30%) that occurs the lowest erosion
rate. In addition, the erosion resistance capacity of soil increased with an increase in slag particle
size. Finally, the resistance mechanism against the water flow of mixtures depended on curing time
and water flow direction.
Keywords: Basic Oxygen Furnace (BOF) slag-treated clay, erosion resistance capacity, erosion rate,
critical shear stress, erosion function apparatus (EFA).

1. INTRODUCTION Various methods have been developed for this

purpose, and an overview of these methods is
Nowadays, land reclamations are being presented by Chu et al., (2009).
performed in some coastal nations such as In this research, the BOF slag was used to
Korea, Vietnam, China, Singapore (He et al., treat clay. The BOF slag is a by-product
2016). This issue gives great challenging due generated during the iron and steelmaking
to lack of common fill material like sand or processes and currently used in various
rock. Thus, using marine dredge clay for filling fields. The advantages of this material are
is an effective solution. Clay can be obtained cheap and easy to obtain. By mixing the clay
from dredging projects, but has poor and the BOF slag at appropriate ratio can
engineering properties. Hence, the treatment of improve significantly the engineering
marine dredged clay is an important problem. properties of the clay.

252
 The marine and coastal works are always under unidirectional flow but also
affected by erosion. In general, erosion is both multidirectional flow by rotating a tube that
the hydraulic and geotechnical process, which contains soil specimen. As described in Fig.1b,
is related to the movement of soil owing to the to rotate the sample tube, a motor system is set
water flow (Harris & Whitehouse, 2017). The up beside the sample, and the other one is used
water flow in the erosion process creates the to controlling the level of a specimen. These
shear stress impacting on the ground. This improvements of the device allow assessing the
stress causes the soil erosion and moves soil effect of flow with different incidence angles.
particles to other locations. Moreover, the
water flow direction in the sea periodically
changes, because the tidal movements and each
type of tide display different characteristics.
For purpose using dredge clay as the
material of embankment or filling of coastal
works, when study treatment of clay, it is
necessary to evaluate not only the strength of (b) Sample
the material but also erosion resistance (a) Front view rotating part
capacity. Regarding erosion experiment, Figure 1. The EFA in Chonnam National University
Briaud et al., (1999) were proposed the process
and theory of EFA test. Besides, some Fig. 2 presents the principle diagram of the
experiments were also carried out to predict improved EFA used in this study. The test
erosion rate in cohesive soils around bridge method and sequence of this apparatus are
pies using EFA tests by Briaud et al., (1999, similar to those of the initial EFA, except for
2008). changing of current direction is considered
This paper aims to examine the erosion during the test. To determine the erosion
resistance capacity of clayed soil treated by resistance capacity of the material, normally
BOF slag under the various conditions. The the test is performed with an array of flow
specimens were prepared and tested by EFA to velocities caused by water flowing in a
consider the effect of the amount of slag, the rectangular flume (pipe). The soil sample is
effect of the slag particle size, the effect of the made by a thin wall tube with 76 mm outside
curing time and the flow direction. In addition, diameter and put at the bottom of the flume. A
the strength development of samples was also piston is located under the specimen to push
measured by laboratory vane shear tests and the soil in the tube until protrudes 1 mm into
unconfined compression tests. From the results the pipe (Fig. 3a). The standard EFA test
of strength tests, by the comparison with the monitors the erosion of 1 mm soil in 1 hour. In
erosion experimental data, can give the the case of 1 mm soil is eroded when the
conclusions about the effects of the factors experimental time is less than 1 hour, continue
above to erosion resistance capacity of BOF to push the sample and repeat the experiment.
slag-treated marine dredged clay.

2. EXPERIMENTAL PROGRAM
2.1 Erosion function apparatus (EFA)
The EFA was firstly developed by Professor
Briaud at Texas A&M University. Fig. 1 shows
the EFA located at Chonnam National
University. In comparison with the previous
apparatus, this EFA has improved, so it can
consider erosion behavior of soil not only Figure 2. Principle diagram of EFA test

253
 The marine dredged clay was filtered through
a 2-mm sieve to remove coarse particles or other
impurities. BOF slag was air-dried at a room
temperature of 20 ± 3°C for one day to fulfill the
saturated-surface-dry density condition. Artificial
seawater with 3.5% salinity was prepared as
(a) Before erosion (b) After erosion
additional water used in the desired water
Figure 3. Photographs captured before and after content of experiment samples to make the
erosion identical condition as marine clay in the coastal
area. The slurry after adding BOF slag and
2.2 Sample preparation additional water to marine dredged clay was
The initial water content was determined by fully mixed for 5 minutes using a hand mixer.
normalizing it by the liquid limit and its Properties of clay and BOF slag are shown in
conditions of samples were set as 1.5 times the Table 1. Erosion test conditions are described in
liquid limit. The BOF slag content, BOFVol (%), Table 2, and Fig. 4. shows the sample
defined as the ratio of the volume of BOF slag to preparation process.
the volumes of soils and water, as below: Table 1. Properties of clay and BOF slag
VBOF slag
BOFVol (%)   100 (1) Clay BOF slag
Vtotal
Liquid Specific Water Specific
where Vtotal is the total volume of the limit gravity content gravity
mixture and VBOFslag is the volume of BOF slag. 119.9 % 2.65 3.04% 3.15

Table 2. Erosion test conditions

Particle size
Target Slag content (%) Water content (%) Curing time (day) Flow direction
(mm)
10
20
Effect of the amount
30 1.5LL 1 Unidirectional <5
of slag
40
50
Effect of the particle <5
30 1.5LL 1 Unidirectional
size of slag <2
Effect of the curing 1 Unidirectional
time and flow 30 1.5LL 3 & <5
direction 7 Bidirectional

(a) Material preparation (b) Material mixing

(c) Sample making (d) Sample preservation

Figure 4. Sample preparation

254
2.3 Procedure and conditions of EFA test and it can be re-interpreted into the relationship
between shear stress, t, and the erosion rate, as
This study focuses on assessing the influence
shown in Fig. 5b. Herein, the critical velocity,
of directional changes in the tidal current on
Vc, is defined as the flow velocity directly before
the erosion resistance of the soil such as the
bidirectional flow without considering the the erosion rate reaches a value of 1 mm/h, and
magnitude of the tidal current, water depth of the t imposed on the surface of the sample by the
the tidal flow, etc. water at the Vc is defined as the critical shear
The procedure and conditions of the stress, tc (Briaud et al., 1999). tc, where the
improved EFA test were used in this study in erosion is initiated, has been used as one
accordance with the original EFA proposed by representative parameter to quantify soil erosion.
Briaud et al., 1999, except for the considering 1
of multidirectional current, as follows:  f V 2 (1)
8
1. Assemble the soil sample tube (76 mm
outside diameter) at the bottom of the rectangular where f = friction factor,  = unit mass of
flume and fill the pipe up with water. The sample water (1000 kg/m3), and V = mean flow
is kept saturate for approximately 1h. velocity in the pipe (m/s).
2. Set the initial flow velocity, it depends on
the type of material to decide the opening
velocity. Normally, using the beginning
velocity of 0.3 m/s for clayed soil and 0.1 m/s
for sandy soil.
3. Push the soil sample up until grow up 1
mm in the flume and control the first flow
velocity through the rectangular pipe as the
values mentioned above using the pumps. (a) Velocity-erosion rate (b) Shear stress-erosion rate
4. Observe the erosion of the 1 mm soil for 1 Figure 5. Presentation of EFA test results (Briaud
h or push the soil up into the pipe and repeat et al., 1999)
the experiment, if 1 mm erosion occurs.
For the multidirectional current test, The friction factor, f, obtained from the
assuming that 1h to perform the EFA Moody chart (Fig. 6). It is a function of the
experiment corresponds to 1 day, the soil Reynolds number of the pipe, Re (ρVD/), and
sample is rotated three times every 15 min for relative roughness, /D, i.e., D = diameter of
1h, while setting the incidence angles of the
the pipe (m),  = dynamic viscosity coefficient
current (i.e., 0o, 90o, 135o, and 180o), to
simulate the ebb and flood tides in 1 day. of the water (10-6 m2/s, 20oC), and  =
5. Record the total amount of erosion for 1 h roughness, ρ = the density of the fluid.
(1 mm × rising frequency), and then calculate
the erosion rate (mm/h).
6. Increase the current velocity and repeat
the steps above. The gap and range of the
velocity depending on the material and target
of the test (0.3 – 5 m/s).
2.4 Analysis of EFA test results
The EFA test results are presented in the form of
the curve indicating the change of the erosion
rate when the velocity increase, as shown in Fig.
5a. The shear stress, t, impact on the soil, which
is generated by the water flowing in the Figure 6. Moody chart (Represented by Munson et
rectangular flume can be estimated from Eq. (2), al., 2013)

255
 Briaud (2008) proposed a range of erosion In these experiments, the EFA tests were
rates for various soil types obtained using the carried out with various flow velocities, and
EFA for 15 years. In the United States, many the initial velocity which was set for each
experiments using the EFA have been carried sample is 0.3 m/s. After that, the flow velocity
out in each state and the results are collected was increased step by step to 0.5m/s, 0.7m/s,
and presented for the purpose of design as 1m/s, 1.5m/s, 2m/s, 3m/s and 4m/s.
shown in Fig. 7. As shown in the figures, the Fig. 8 shows the change of erosion rate
limit shear stress that causes erosion of the according to the flow velocities. The
soil is very different and depending on the relationship between the shear stress and the
type of soil. erosion rate are displayed in Fig. 9. The values
of shear stress were calculated using equation
(2). It is clear that the erosion rate does not
occur at velocities less than 1 m/s for all
specimens. After that, when velocity exceeds 1
m/s, there is a significant decrease of erosion
rate with an increase of the fraction of BOF
slag from 10% to 30%, and the smallest
erosion rate of material achieve when using
30% of BOF slag. However, when the
proportion of BOF slag climbs to over 30%,
the erosion rate increases again. These results
(a) Erosion rate versus velocity are matched with the values of Su, which are
illustrated in Fig. 10. It can be seen that the Su
at 1day ranges from 1.09 kPa at 10% of slag to
a peak of 13.46 kPa at 30%. Similarly, these
values for 7 days are 2.22 kPa and 127.40 kPa
at 10% and 30%, respectively. Furthermore,
although the percentage of slag increases by up
to 50%, Su decreases up to 8.02 kPa at 1 day
and 52.50 kPa at 7 days. The possible reason
why strengths at 40% and 50% of BOF slag are
smaller than those of 30% BOF slag is that
(b) Erosion rate versus shear stress even though 40% and 50% of BOF slags
Figure 7. Erosion categories for soils based on contain a larger amount of slags, 30% BOF
velocity and shear stress (Briaud, 2008) slag has higher fine particle content of BOF
slag then others, resulting in high chemical
3. RESULTS AND DISCUSSIONs activity.
3.1 Effect of the amount of slag
To access the effect of the amount of BOF slag
on the erosion resistance capacity of the
mixture, several EFA tests were performed to
determine the Vc and the c under the
unidirectional flow condition for five sample
cases of mixtures, including 10%, 20%, 30%,
40% and 50% at 1day curing. Besides, Su were
also determined by the laboratory vane shear
tests at 1 day and unconfined compression tests Figure 8. Relation of erosion rate and flow
after 7 days. velocity with various BOF slag contents

256
 3.2 Effect of the particle size of slag
These EFA experiments used unidirectional
flow and specimens with slag particle size
smaller 2mm and 5mm in order to compare the
differences in erosion resistance behavior of
samples. The proportion of BOF slag in the
mixture was 30% chosen from the results of
the tests that evaluated the influence of the
amount of slag.
Figure 9. Relation of shear stress and erosion rate The relation between shear stress and
with various BOF slag contents velocity and the relationship between shear
stress and erosion rate are illustrated in
Figure 11 and Figure 12, respectively. Overall,
the sample used small slag particle size is
eroded higher than that of large particle size.
At the velocities under 2m/s, the water flow
causes a low erosion for both types of samples.
However, when the water flow velocity
increase from 2m/s to 4m/s, the erosion rates
rise sharply until 72mm/h for the mixture used
slag particle less than 5mm and 120mm/h for
the material with slag particle under 2mm.

Figure 10. Change of Su with various BOF slag

contents

The determined Vc and c of samples, which

have different the amount of BOF slag, are
summarized in Table 3. The Vc and c grow
significantly from 1.07m/s and 2.84N/m2 at
10% of slag to 2.04m/s and 9.23N/m2 at 30%,
respectively. However, when the percentage of
BOF slag increases to 40% and 50%, there is a
decrease in critical velocity and critical shear Figure 11. Relation of erosion rate and flow
stress due to small strengths. It was found that velocity according to particle size
the erosion resistance characteristics of BOF
slag-treated clay have an agreement with the
behaviors of Su.

Table 3. Critical velocity and critical shear

stress with various BOF slag contents
Amount of slag in
10 20 30 40 50
the mixture (%)
Critical velocity,
1.07 1.09 2.04 1.66 1.27
Vc (m/s)
Critical shear Figure 12. Relation of shear stress and erosion rate
2.84 2.94 9.23 6.32 3.87
stress, τc (N/m2) according to particle size

257
 Figure 13 compares the Su of mixtures with (0o180o) at a certain velocity during the same
different BOF slag particle size at 1 day and 7 sustainment time that was allocated for the test.
days. In general, the Su of the sample with the For each specimen, at each velocity level, the
particle of slag below 5 mm constantly higher amount of erosion was observed for one full
than those of sample used particle of slag under hour. This means that a soil sample was rotated
2mm. At 1 day, the strength of the mixture has a in the order of 0o 180o  0o 180o every 15
slag particle size less than 2mm is 2.49 kPa, min for 1 h while shutting off the water flow
whereas the value for under 5mm is 13.46 kPa. upon each rotation, considering the twice-a-day
Similarly, the results of tests at 7 days are 46.43 tidal cycle.
kPa and 127.40 kPa, respectively. This means Figure 14. and Figure 15. show the
that, when the slag particle size increases, Su of comparisons of the variations in the relationship
the mixture grows, so the erosion resistance between the erosion rate and the velocity/ the
capacity of this material is also higher. shear stress of unidirectional and bidirectional
flow at 1 day, 3 days, and 7 days. The results
indicated that the erosion rate for all samples
increases remarkably after a specific velocity,
except for samples at 7 days, and that the specific
velocity of each sample varied depending on
curing time of the sample. In general, the erosion
rate decreases with an increase in curing time.
The results from strength tests show a tight
Figure 13. Su depending on slag particle size correlation with data from EFA tests above. The
development of Su of the mixture with curing
Vc and c of specimens, which were used time is presented in Fig. 16. At 1 day, Su of BOF
different particle sizes of BOF slag, are showed slag-treated clay is 13.46kPa. This can lead to the
in Table 4. As shown in Table 4, there is an possibility of erosion for all samples is the
increase of Vc and c corresponding to a rise of highest and similar between the unidirectional
BOF slag particle size. When slag particle size and bidirectional flow. At 3 days curing time, Su
under 2mm is used, the measured Vc and c are achieves 24.17kPa. There are clear differences
1.77m/s and 7.11N/m2, respectively, whereas between two curves, and the mixture is scoured
for the sample with a particle of slag less than faster under impacting of bidirectional flow, in
5mm, Vc and c reach to 2.04m/s and 9.23 N/m2. comparison to unidirectional flow. After 7 days,
Su of the mixture rises to 127.4 kPa. As a result,
Table 4. Critical velocity and critical shear the erosion resistance capacity reaches the
stress with different particle size highest value. Erosion could not be observed for
both types of flow directions, even though
Slag particle size < 2mm < 5mm
velocity goes up to 5m/s.
Critical velocity, Vc (m/s) 1.77 2.04
Critical shear stress, τc (N/m2) 7.11 9.23

3.3 Effect of the curing time and flow direction

The EFA tests were performed to consider curing
time and evaluate the influences of flow
direction. The samples with 30% BOF slag were
tested under unidirectional flow and bidirectional
flow at 1 day, 3 days, and 7 days curing time.
Regarding the tests under bidirectional flow, the Figure 14. Relation of erosion rate and flow velocity
erosion resistance capacity was measured by considering curing time and water flow
rotating each artificial soil sample three times direction

258
Figure 15. Relation of shear stress and erosion rate
considering curing time and water flow Figure 18. Erosion classification of mixture under
direction bidirectional flow (Briaud, 2008)

Table 5 compares Vc and c of the samples

with different curing time under unidirectional
and bidirectional flow. As an overall trend, Vc
and c occurred under unidirectional flow are
slower than those of bidirectional flow.
Besides, critical velocities and critical shear
stress also grow according to the time.
Figure 16. Development of undrained shear strength Table 5. Critical velocity and critical shear
with curing time
stress with different curing time under
Figure 17 and Figure 18 show the erosion unidirectional and bidirectional flow
properties of the material by using the erodibility Curing
classification chart in the correlation between the 1 day 3 days
time
velocity and erosion rate. Under both Directional Uni- Bi- Uni- Bi-
unidirectional and bidirectional flow, the BOF flow direction direction direction direction
slag-treated clay samples are found in the zone of
Critical
medium erodibility for all velocity level at 1 day. velocity, 2.10 2.03 2.71 2.66
After 3 days, when velocity higher than 2.5m/s,
Vc (m/s)
they are also put in medium erodibility, but at
velocities lower than 2.5m/s the samples are Critical
classified in low erodibility zone. Until 7 days, shear
9.74 9.15 15.62 15.09
the mixture becomes material having good stress, τc
(N/m2)
erosion resistance capacity, and they are placed
in low erodibility zone.
4. CONCLUSIONS
The following conclusions are drawn as follow:
- The erosion resistance capacity and
undrained shear strength of BOF slag-treated
clay depend on the proportion of slag in
mixture, and the highest values are found when
the percentage of slag is 30%. In addition, at
the same condition and mixed ratio, the
mixture with larger slag particle size provides
higher Su and also erosion resistance capacity,
Figure 17. Erosion classification of mixtures under in comparison with other samples having
unidirectional flow (Briaud, 2008) smaller particle size.

259
 - The erosion resistance capacity of BOF 6. REFERENCES
slag-treated clay changes according to the time
Briaud, J. L. 2008. Case Histories in Soil and Rock
and the flow direction. At 3 days, the Erosion : Woodrow Wilson Bridge , Brazos River
bidirectional flow causes the erosion faster Meander , Normandy Cliffs , and New Orleans
than unidirectional flow, and material is Levees. Geotechnical and Geoenviromental
classified in medium erodibility zone of Engineering, 134(10), 1425–1447.
erosion classification chart. After 7 days, the Briaud, J. L., Ting, F., Chen, H. ., Gudavalli, R.,
Kwak, K., Philogene, B., … Li, Y. 1999.
erosion almost does not occur, even though the
SRICOS: Prediction of Scour rate at bridge
samples are affected by the any type of flow. piers. In Report No TX-00/2937-1. Texas, USA.
Therefore, the mixture is placed in a low Chu, J., Varaksin, S., Klotz, U., & Mengé, P. 2009.
erodibility area. Construction processes State-of-the-art report.
- The erosion rate of all the samples Proceedings of the 17th International
increases rapidly over a certain velocity. Vc and Conference on Soil Mechanics and
τc of mixture reach the highest when 30% of Geotechnical Engineering: The Academia and
Practice of Geotechnical Engineering, 4,
BOF slag was used. Moreover, critical velocity 3006–3135. Alexandria, Egypt.
and critical shear stress also grow with the Harris, J., & Whitehouse, R. 2017. Scour
increase of BOF slag particle size and curing Development around Large-Diameter
time. In addition, critical velocity and critical Monopiles in Cohesive Soils: Evidence from
shear stress are generated by unidirectional the Field. Journal of Waterway, Port, Coastal,
flow higher than those of bidirectional flow. and Ocean Engineering, 143(5).
He, J., Jian, C., Tan, S. K., & Vu, T. T. 2016.
Sedimentation behavior of flocculant-treated
5. ACKNOWLEDGEMENT soil slurry. Marine Georesources and
Geotechnology, 35(5), 593–602.
This research was a part of the project titled
Munson, B. R., Okiishi, T. H., Huebsch, W. W., &
“Development of Design Technology for Safe Rothmayer, A. P. 2013. Fundamentals of
Harbor from Disasters”, funded by the Fluid Mechanics (7th ed.). USA: John Wiley
Ministry of Ocean and Fisheries, Korea. & Sons, Inc.

260
The 3rd Int. Conf. on Transport Infrastructure & Sustainable Development (TISDIC 2019). Construction Publishing House. ISBN: 978-604-82-2893-4.

Stabilization of weathered granite soil using ground

bottom ash activated by various industrial by-products

Tran Quoc Thien 1, Gyeong-o KANG2, Do Huu Dao 3, Huynh Phuong Nam 3,
Young-sang KIM*
1
Department of Civil and Environmental Engineering, Chonnam National University, Yeosu, 59626, 50
Daehak-ro, Yeosu, Jeollanamdo, Republic of Korea
2
Honam Regional Infrastructure Technology Management Center, Chonnam National University, 50
Daehak-ro, Yeosu, 59626, Jeollanamdo, Republic of Korea
3
Faculty of Road and Bridge Engineering, The University of Da Nang, University of Science and
Technology, 54 Nguyen Luong Bang, Lien Chieu, Da Nang, 550000, Viet Nam

Department of Civil Engineering, Chonnam National University, Gwangju, 77 Yongbong-ro, Buk-gu,
Gwangju, 61186, Republic of Korea.
Corresponding Author: Professor Young-sang KIM, [email protected]
Address correspondence to First author: [email protected]

ABSTRACT: The main objective of this research is attempting to reusing bottom ash (BA) in a goal
weathered granite soil (WGS) stabilization. The new cementless binder, tentatively named GBA(G/R)
was generated from the cementitious mixtures composed of ground bottom ash (GBA) activated by
phosphogypsum (G), lime (L), and red mud (R). For the experimental program, various mixtures of the
WGS stabilization were prepared with different proportions of binders GBA(G/R). The engineering
characteristics such as the unconfined compressive strength (UCS), secant modulus of elasticity (E50),
California bearing ratio (CBR), splitting tensile strength (SPS), corrosivity potential, and heavy metal
leaching test were determined as the main indicators for improvement in WGS with respect to the basic
environmental requirements. The expected improvements in the engineering properties of the stabilized
WGS mixtures were compared with those of the unstabilized WGS. The addition of the binder led to a
significant improvement in the 28 day UCS, which reached more than 5 MPa with the used G/R ratio in
the stabilized mixture equal to 1. In addition, scanning electron microscopy (SEM), X-Ray diffraction
(XRD), and energy-dispersive X-ray spectrometry (EDS) were used to investigate the presence of gels as
a possible reason for the improved WGS engineering properties. The results showed that the proposed
binder GBA(G/R) is feasible for the stabilization of weathered granite soil.
Keywords: By-products, new green binders, bottom ash, ground bottom ash, weathered granite soil.

1. INTRODUCTION area is hilly terrain, most operating and under-

construction infrastructures are located in hilly
In recent decades, rapid growths in population areas. Hence, residual granitic soil (WGS)-the
and urbanization have been observed in South most abundant soil type on the Korean
Korea. Consequently, the demands for land peninsula-has been exploited as subgrade
for the development of infrastructures and layer for construction fields such as mat
industrial facilities have been increasing foundations, embankments for abutments,
constantly. Because approximately 70% of the retaining walls, and road constructions (Kim

261
2007; Yun 2015). One of the main reasons environment for the stabilized soil mixtures
why WGS is not widely applied are its various by mixing the red mud with other mixture
engineering characteristics that depend on the components under dry conditions (Tchadjié et
weathering degree, composition minerals, and al. 2016; Djobo et al. 2016; Ranjbar and
compaction energy of the embankment. Kuenzel 2017; Ghadir and Ranjbar 2018;
Besides, most WGS types classify as clayey Hashemi et al. 2018, 2019).
sand and silty sand and their mechanical More than 6 million tons coal ash are
behaviors exhibit intermediate properties generated annually in South Korea. The fly
between sand and clay. Thus, the material is ash (FA) content was approximately 85–90%,
vulnerable when the soil is immersed in water whereas 10%–15% was bottom ash (BA),
(Kim 2007; Yun 2015). The local soils have which is less effectively reused than FA
often been stabilized by mixing in various (Park et al. 2009). Besides, 30 million tons of
materials to enhance their engineering phosphogypsum were accumulated and
properties. Cement is one of the traditional 200,000 tons of red mud were generated each
materials used for soil stabilization. South year (Mun et al. 2007; Do and Kim 2016).
Korea is an important cement producer and For each ton of phosphoric acid,
consumer and among the world’s top 10 CO2 approximately 5 tons of phosphogypsum are
emitters with an annual emission of 4 million generated (Verbeek and Du Plessis 2005;
tons of CO2 (Yoon et al. 2017). According to Deǧirmenci 2008; Chen et al. 2018).
calculations, when 1 ton cement is created, 1 Mitigating the ecosystem impacts such as the
ton CO2 is discharged into the air leaching of heavy metals into the soil and
environment (Davidovits 2002; Kim et al. ground water via the appropriate disposal of
2019). Thus, the Korean government is facing these industrial wastes is currently a high
severe air pollution and the greenhouse effect. concern in Korea.
To reduce the CO2 emissions by 37% from Recently, several soil improvement studies
the business-as-usual levels, it is essential to for indigenous soils have been conducted with
reduce cement-producing activities by finding recycled materials and various industrial by-
alternative cementing agents (Yoon et al. products as cement replacement. Several
2017). To mitigate the environmental effects, studies have investigated the reuse of bottom
cement is entirely or partially replaced with ash to examine its effect on the engineering
pozzolanic materials such as coal ash, blast- properties of soil (Nalbantoǧlu 2004; Güllü
furnace slag, volcanic ash, palm oil fuel ash, 2014; López López et al. 2015; Sharma and
rice husk ask, red mud (R), and Sivapullaiah 2016; Hastuty et al. 2017; Kim et
phosphogypsum (G). (Ouhadi and Yong 2008; al. 2018; Hashemi et al. 2018, 2019)
Sivapullaiah and Jha 2014; Modarres and BA has very poor self-hardening abilities
Nosoudy 2015; Tchadjié et al. 2016; Djobo et without chemical activators such as lime or
al. 2016; Pourakbar and Huat 2017; Ghadir cement (Sharma and Sivapullaiah 2016).
and Ranjbar 2018; Hashemi et al. 2018, 2019). Therefore, in this study, red mud, lime, and
These materials have been widely applied for phosphogypsum were used as activators for the
soil stabilization and are based on the working binders. They can accelerate the reaction
mechanisms of geopolymers composed of between the binder components in a gel
tetrahedral, linked silicates and aluminates. formation.
The aluminosilicate reactive parts in the The study might potentially increase the
pozzolanic materials are usually activated usage of BA, red mud, lime, and
with strong alkali solutions such as Na2O3 phosphogypsum as effective soil stabilizers
and NaOH owing to their strong base and replacement of Portland cement. This
potentials. Instead of directly using alkali would positively impact the environmental
solutions, red mud-a strong alkali by-product- conservation in terms of engineering and
was used in this study to generate the alkali economic factors.

262
2. EXPERIMENTAL PROGRAM Table 1. Physical Properties of WGS and new
binder’s components
2.1 Materials
Parameters WGS GBA G L R
The WGS originated from a site in Yeosu city,
Jeolla Province, South Korea. Fig. 1 presents Specific gravity, Gs 2.62 2.36 2.33 3.20 3.15
the grain size distribution of the WGS obtained
via sieve analyses and hydrometer tests. The Maximum dry 16.6
physical properties and X-ray diffraction density (KN/m3)
(XRD) results of the WGS are presented in Optimum Moisture 16.80
Table 1 and Table 2, respectively. Content (%)
Liquid limit, LL (%) 27.80
Plastic limit, PL (%) 23.06
Plasticity index, 4.74
PI (%)
Undissolved mass 51.72 45.08 81.62
(%)
Fine particles 15.05 17.51 53.10 67.20 100
< 75μm (%)
Soil classification Clayey
sand
Figure 1. Particle size distribution of WGS and the Colour Brown Grey White White Red
new binder components
Bottom ash (BA) was used as a main binder
component to create green binders. The BA
originated from a thermal power plant in the
Honam area, South Korea. Table 2 present its
chemical compositions. FA is a well-known
material in the construction field owing to its
pozzolanic reactivity. Bottom ash (BA) has a
similar chemical composition. However, its
particles are porous, angular, heterogeneous,
a) Raw bottom ash. b) Ground bottom ash. and much larger than FA particles.
Furthermore, it has many pores and craters on
the surface. For its effective use as main
binder component, it was ground with a ball-
milling machine (UBM-10L, RAMT) for 3 h
at 120rpm. After grinding, the specific surface
of the GBA was 0.20m2/g, which might be
sufficient for a significant strength
(Sathonsaowaphak et al. 2009). The particle
sizes, scanning electron microscopy (SEM)
images, and grain size distribution of the GBA
c) Particles of bottom ash and ground bottom ash
before and after grinding are presented in
Figure 2. Scanning electron microscope of raw and Figs. 2 and 3, respectively. The XRF an XRD
ground bottom ash. analyses of the GBA are shown in Table 2 and
Fig.5a, respectively.

263
Table 2. Chemical composition of WGS, ground The red mud (R) was obtained from the
bottom ash (GBA), cement (C), phosphogypsum alumina refining of bauxite via the Bayer
(G), lime (L), and red mud (R) by XRF analysis process in the Honam area. Owing to their high
pH values, the red mud and lime were applied
Chemical WGS GBA C G L R as alkaline activators. The phosphogypsum (G)
composition
used as sulfate activator was generated in the
(%)
manufacture of phosphoric acid via the wet
CaO 1.26 1.80 64.35 51.43 80.01 7.10 process. Among the activator components,
SiO2 55.17 62.5317.16 1.65 1.87 15.12 lime was considered as the major gel formation
agent. It provides not only an alkaline
Al2O3 30.73 20.91 3.57 0.32 0.93 19.87
environment but also Ca2+ ions for the calcium
SO3 - - 2.2 39.08 0.22 0.72 silicate hydrate (CSH), CASH, and CaCO3 gel
MgO 0.14 0.69 2.66 - 3.69 0.37 formation. The mechanisms of the gel
formation were studied and partially revealed
Fe2O3 6.66 8.7 4.03 0.30 0.75 22.21
in this study. The grain size distributions and
TiO2 0.95 1.28 0.29 - - 5.24 chemical compositions of phosphogypsum (G),
K2O 3.18 1.44 1.11 - 0.31 0.11 lime (L), and red mud (R) obtained from sieve
tests, hydrometer tests, and via a helium–neon
Na2O 0.44 0.39 - 0.20 0.11 14.92
laser optical system (ISO 13320) are presented
(LOI) 1.15 1.88 4.63 6.66 13.43 15.85 in Fig. 1 and Table 2, respectively. The XRD
spectra of the WGS, cement, and binder
components such as GBA, red mud, lime, and
phosphogypsum are shown in Fig. 5a.
In addition, the reactivity index was
investigated to determine the reactive Al and Si
contents in each binder component. The
reactive fraction was determined by stirring 1 g
of binder components in 100 mL NaOH 8
mol.L-1 solution for one day at 65°C. The
insoluble particles were filtered out with a
vacuum pump with a microfiber filter paper.
Next, they were washed with distilled water to
Figure 5a. XRD analysis of WGS, GBA, C, R, L, obtain a neutral pH, dried for 1 h at 250°C, and
and G (G: Phosphogypsum; Gi:Gibbsite; finally weighed. The reactivity index is
Q:Quartz; M:Mullite; H:Hematite; calculated as the ratio of the weight of
He:Hedenbergite; K:Kaolinite; C3S:Alite; insoluble particles to 1 g. The reactivity
C2S:Belite; C3A:Tricalcium aluminate; indexes of the binder components are shown in
C4AF:Celite; C2F:Brownmillerite).
Table 1.
2.2 Experiment
A series of laboratory tests was conducted to
determine the unconfined compressive strength
(UCS), California bearing ratio (CBR), and
splitting tensile strength (SPS) to observe the
Figure 5b. XRD analysis of WGS and stabilized strength behavior of the stabilized mixtures
mixtures (Q:Quartz; K:Kaolinite; G: with an optimal water content during the
Phosphogypsum; CSH:Calcium silicate curing time. Further, SEM, XRD, and EDS
hydrate; CH:calcium hydroxide; analyses were performed to determine the
E:Ettringite; Cc:Calcium carbonate ). microstructure characteristics of the stabilized

264
WGS mixtures. Finally, the corrosivity SEM and EDS analyses, tiny samples were cut
potentials and heavy metal leaching test were from the air-dried stabilized-soil specimens, and
measured with various combinations of WGS the debris on the surface was removed. Further,
and new green binders to assess the feasibility the samples were freeze-dried to remove
of the binders with respect to environmental moisture. The XRD samples were pulverized
protection. The mixture proportions are shown and sieved through a 50 µm sieve before mixing
in Table 3.
them with a methanol solution. Next, they were
Table 3. Mix design and type of tests dried for 20 min before the tests (Modarres and
performed in the experimental program Nosoudy 2015).
The corrosivity describes whether a material
Mix name Description might be dangerous to the environment or
WGS Raw soil human health owing to its harmful metal
mobilization when discharged into a water
WGS-GBA WGS, 20% GBA source or the ground. A material is corrosive
WGS-C WGS, 20% cement when a typical sample of the material is aqueous
and has a pH value equal to or below 2 or equal
GBA WGS, 20% GBA + 1.5% G + 6% L
(G/R=0.2) + 7.5% R
to or above 12.5 ( Tikalsky et al. 2004; Razak et
al. 2009; Kim et al. 2018). In this study, the
GBA WGS, 20% GBA + 3.0% G + 6% L corrosivities of the stabilized WGS mixtures
(G/R=0.5) + 6.0% R were measured by measuring the pH values of
GBA WGS, 20% GBA + 4.5% G + 6% L the respective leachates. The leachates were
(G/R=1.0) + 4.5% R collected by placing a cylindrical sample in an
GBA WGS, 20% GBA + 6.0% G + 6% L
amount of water conforming to a water-to-solid
(G/R=2.0) + 3.0% R ratio of 8 (L/S = 8) at 21°C. Moreover, the
concentrations of the contaminants in the
GBA WGS, 20% GBA + 7.5% G + 6% L leaching water of the stabilized mixtures were
(G/R=5.0) + 1.5% R
determined via the concentrations of heavy
metals leaching from the stabilized mixtures.
The bearing capacities of road materials are Because the stabilized WGS would be in direct
commonly evaluated via the CBR value. The contact with the surface or groundwater, the
CBR tests were carried out on specimens amount of heavy metals is an important
(prepared in a cylindrical mold of 175mm height characteristic that must be controlled. The
and 150 mm diameter) with the typical mixtures obtained leachate was investigated with
including raw WGS, WGS stabilized via cement, inductively coupled plasma atomic emission
GBA, and GBA with activator. After 28 days of spectrometry (ICPMS) (Kim et al. 2018).
curing, the specimens were immersed in water
for four days before testing. The SPS was used to 2.3 RESULTS AND DISCUSSION
evaluate the tensile strength, which is an
important engineering property for the design of 2.3.1 Strength characteristics
roads. It reveals the substantial resistance of a The compaction curve of the natural WGS soil
material to tensile stress. was determined according to ASTM D698.
To simplify the macro-behavior investigation Therefore, the maximal dry density (γdmax =
and comparison of the mixtures after 28 curing 1.66 g/cm3) and optimal water content (OMC
days, SEM, XRD, and EDS were employed. = 16.8%) could be determined. Fig. 6 shows
The Hitachi S-3000N, D8 ADVANCE, and a the UCS results of the compacted WGS for 3,
low-vacuum (LV) SEM were used. For the 7, and 28 curing days.

265
 Figure 6 presents the relationships between long-term (28 days) strength behavior by
the average UCS and phosphogypsum-to-red generating ettringite during the curing time.
mud ratio for various WGS mixtures These results are in good agreement with
stabilized with binder GBA(G/R). The 28 day those of previous studies (Do et al. 2019;
UCS varied between 3917.9–5477.4 kPa, the Kalkan 2006; Tran et al.2019).
secant modulus of elasticity varied between
343.2–517.9 MPa (Fig. 7). The 28 day UCS
values of the stabilized mixtures were higher
than those of the unstabilized ones. Further,
they were comparable to the 28 day UCS
values of the WGS samples stabilized with
cement. Please note that all WGS mixtures
stabilized with GBA(G/R) satisfied the 7 day
strength requirements for subgrade and
subbase layers in road construction (1.4 MPa)
according to the ACI committee 230 Report
(ACI 2009).
After 28 curing days, the UCS and secant
modulus increased with increasing Figure 6. UCS of the stabilized WGS mixtures
phosphogypsum-to-red mud ratio (G/R) from using mainly ground bottom ash
0.2 to 1.0. The maximum was reached for (GBA) activated by activators.
mixtures with GBA(G/R = 1.0) binder.
Afterward, the increasing G/R ratio (from 1 to
5) caused the mixture stiffnesses to decrease.
Further, the UCS and secant modulus reached
maxima for the mixtures with phosphogypsum-
to-lime ratios (G/L) and red mud-to-lime ratios
(R/L) of 0.75.
The red mud in the mixtures not only
provides OH- for the formation of CSH gel
but also releases calcium hydroxide. Another
product originating from lime possibly reacts
with the remaining alumina component in the
red mud and generates calcium aluminates Figure 7. Secant modulus of elasticity of the
(CA) (Ribeiro et al. 2013). Hence, beyond the stabilized WGS mixtures using main
optimal G/R ratio, a low amount of red mud binder GBA activated by the activators.
and a higher phosphogypsum level in the
mixture could lead to a lower strength Fig. 8 shows the SPS values of the raw
compared with that of the optimal mixture WGS and various stabilized WGS mixtures
(GBA(G/R = 1.0)). Thus, the strength after curing times of 3, 7, and 28 days. The 3,
behavior of the mixture is a function of the 7, and 28 day SPS values of the raw WGS
G/R ratio. ranged between 18.7-56.9 kPa, and those of the
Please note that the red mud components in WGS stabilized with GBA without activators
the new green binders strongly enhanced the increased insignificantly. The WGS mixtures
early strength behaviors of the stabilized WGS treated with GBA activated via G, L, and R
mixtures (e.g., GBA(G/R = 0.2), GBA(G/R = exhibited significantly improved SPS values.
0.5),GBA(G/R = 1.0)), as shown in Fig. 6. The SPS values increased to 1194.0 kPa for the
The phosphogypsum component in the new case with optimal phosphogypsum-to-red mud
green binders remarkably contributed to the ratio.

266
 and GBA(G/R = 5.0)) were chosen for the
SEM tests.
Figures 10 a & b present the SEM results of
the WGS particles and WGS-GBA mixtures,
respectively. It can be seen that the gels formed
on the GBA surface appeared unclear and
insignificant. This might be one of the reasons
for the insignificant improvements in the WGS
stabilized with unreacted GBA without the
activators.

Figure 8. SPS of raw WGS, cement treated WGS,

and the stabilized WGS with and without
the activators.

Fig. 9 shows the CBR values of the raw

WGS and various stabilized WGS mixtures
after 28 curing days. The raw WGS exhibited a a) b)
very low bearing capacity (2.01%), which was
Figure 10. SEM images for WGS particles and
insufficient for the CBR requirements for
stabilized mixture using GBA.
subgrade layers. Further, the CBR values of the
mixtures WGS-GBA stabilized with GBA,
without activators. Interestingly, Fig. 9
presents significant CBR improvements for the
WGS mixtures stabilized with GBA with
activators (130.5%). This improvement is
comparable to the case of WGS treated with
cement (WGS-C).
a) b)
Figure 11. SEM analysis for stabilized mixtures
using binders GBA(G/R=1) and
GBA(G/R=5).

The material was used owing to its high

alumina and silica contents for pozzolanic
reactions when activated with activators such
as phosphogypsum, lime, and red mud. The
SEM images of the mixtures stabilized with
GBA with activators (GBA(G/R = 1) and
GBA(G/R = 5)) are shown in Figs. 11a and b,
Figure 9. CBR of raw WGS, cement treated WGS, respectively. Regarding the GBA(G/R = 1)
and the stabilized WGS with and without mixture, most GBA particles seemed to be
the activators. hydrolyzed in the alkaline environment
composed of red mud and lime. It is believed
2.3.2 Microstructure analysis that two main gels were generated inside the
The SEM technique is one of the best methods GBA(G/R = 1) mixture: CSH gel surrounding
to characterize the morphologies the GBA particles and needle-shaped ettringite
(microstructures) of mixture components. Four gel surrounding the GBA and soil particles.
samples (WGS, WGS-GBA, GBA(G/R = 1.0), Figure 11b presents the SEM image of the

267
GBA(G/R = 5) mixture with visible ettringite oxygen (O). Hence, it was considered to be
gels. However, very little CSH gel surrounds ettringite gel. Besides, the signal of calcite (Cc)
the GBA particles. This might be caused by the was found along with the main CSH and
low amount of red mud and high amount of ettringite gels in the EDS analyses. These
phosphogypsum, which reduces the alkaline results confirm the presence of CSH, ettringite,
environment for the CSH formation. Hence, and calcite gels in the WGS mixtures stabilized
these factors might have led to a reduction in with the new green binders.
the strength and stiffness. These results are in
good agreement with those of a previous study
(Abdullah et al. 2017).
Besides, XRD analyses were conducted for
two samples (GBA(G/R = 1), GBA(G/R = 5)).
According to Fig. 5b, the pure WGS contained
minerals such as vermiculite (V), kaolinite
(K), quartz (Q), illite (I), and albite (A).
Figure 5b exhibits new peaks corresponding to
CSH, CH, ettringite (E), calcite (Cc), and
phosphogypsum (G). Calcium silicate hydrate a) EDS analysis of the gel supposedly named CSH.
(CSH), CH, ettringite (E), and calcite (Cc) are
well known as hardening agents that help to
improve the soil strength (Subbarao 2001;
Modarres and Nosoudy 2015). However,
GBA(G/R = 5) in Fig. 11 contained not only
the hardening agents (particularly, ettringite)
but also an abundant amount of
phosphogypsum. Owing to its low pH value,
phosphogypsum might reduce the pH value of
the mixtures, thereby retarding the gel Figure 12. EDS analyses indicated the main gels
formation process. Furthermore, hardening of the stabilized WGS mixture named
agents such as CSH, ettringite (E), and calcite GBA(G/R=1.0)
(Cc) might be surrounded by crystalline
phosphogypsum instead of interlocking with
each other. This leads to a reduced 2.3.3 Environmental effects
interlocking area and stiffness of the stabilized The pH values of the leachate samples
mixtures. This assumption agrees with those collected from the stabilized WGS mixtures are
of previous studies (Mehta 1983; Santhanam listed in Table 4. The pH values after 7, 14, 21,
et al. 2003). and 28 days varied from 10.40 to 11.62 for the
To conduct a comprehensive microstructure GBA(G/R) mixture. According to Table 4, the
analysis, the mixture GBA(G/R = 1.0) was pH values of the WGS mixtures stabilized with
investigated in an EDS analysis as an example. GBA without activators (WGS-GBA) did not
Figure 12 presents the EDS results: the approximate the alkaline environment
chemical elements of two different gels- necessary for the hydration of those binders to
supposedly CSH and ettringite-in the soil form cross-linking gels (pH values between
matrix shown in Fig. 11a. The dominant 5.75 and 8.53), which resulted in the
chemical elements in area 1 (Fig. 11a) were significant improvements of the WGS mixture
calcium (Ca), silicate (Si), and oxygen (O). strengths. Furthermore, these pH values were
Therefore, it was considered to be CSH gel. below and above the threshold interval of 2.5
The typical chemical elements of area 2 were to 12.5, respectively. Thus, the stabilized
calcium (Ca), sulfate (S), aluminum (Al), and mixtures were not corrosive.

268
Table 4. pH values of stabilized WGS mixtures 3. CONCLUSIONS
pH measured values The effects of the new green binder GBA(G/R)
Mixtures and activators (phosphogypsum, lime, and red
7 14 days 21 28 days
days days mud) on stabilized WGS mixtures were
investigated. The following main conclusions
WGS 5.87 5.87 5.87 5.87 can be drawn:
WGS-GBA 5.75 6.01 6.15 6.13 1. The WGS mixtures including the green
GBA(G/R=0.2) 10.99 11.52 11.06 10.62 binder GBA(G/R) exhibited significant
increases in strength compared with those of
GBA(G/R=0.5) 10.86 11.62 11.35 10.97 the unstabilized ones. The strengths of the
GBA(G/R=1.0) 10.81 11.6 11.08 10.66 mixtures with green binders were
GBA(G/R=2.0) 10.71 11.59 11.17 10.85 comparable with the strength of WGS soil
stabilized with cement.
GBA(G/R=5.0) 10.73 10.93 10.67 10.40 2. The 28 day UCS of the mixtures stabilized
EPA SW-846 2000 2.5<measured pH value<12.5 with the new green binder GBA(G/R)
reached their maximum for G/R = 1. The
Figure 13 lists the trace element strength behavior was a function of the G/R
concentrations in the leachates originating from ratio.
the stabilized WGS specimens. In general, Fe 3. The red mud amount in the new green binder
exhibited the highest concentration, followed by GBA(G/R) enhanced the early strength
Cu, As, Cr, Zn, Ni, Mn, Co, Pb, and Cd. The behavior of the stabilized WGS mixtures.
obtained concentrations were compared with the Further, phosphogypsum strongly enhanced
maximal contamination levels for drinking the long-term (28 day) strength behavior.
water in the “Treatment Standard for Hazardous 4. To effectively mobilize the pozzolanic
Wastes” in the Code of Federal Regulations of activity of the green binders, the main
the US government (Pollution Control & component GBA had to be activated with
Ecology Commission 2015) and were below the activators (G, L, and R). They contributed to
acceptable levels. Thus, the heavy metal significant improvements in the engineering
concentrations in the WGS mixtures stabilized properties of the WGS.
with the new binders might not exceed the 5. The mixtures stabilized with the new green
critical limits. Consequently, the investigated binders were not corrosive according to their
stabilized soils had no detrimental impact on the pH values. Further, the hardened WGS
environment, and the hardened WGS specimens mixtures stabilized with the green binders
were classified as nonhazardous materials. classified as nonhazardous materials.

4. ACKNOWLEDGMENT
This work was supported by the National
Research Foundation of Korea (NRF) grant
funded by the Korea government (MSIT)
(NRF-2018R1C1B6008095)

5. REFERENCES
Abdullah, A., Mohammed, H., and Ahmed, K.
(2017). “Improvement of geotechnical
properties of sabkha soil utilizing cement kiln
dust.” Journal of Rock Mechanics and
Figure 13. Element concentrations of leachates
Geotechnical Engineering, Elsevier B.V., 9(4),
from stabilized soils (unit: ppb) 749–760.

269
ACI. (2009). 230.1R-09 Report on Soil Cement. Hastuty, I. P., Roesyanto, and Jeriko, B. S. (2017).
American Concrete Institute. “A Study of the Effectiveness of the Use of
Chen, X., Gao, J., Liu, C., and Zhao, Y. (2018). Cement and Bottom ash Towards the Stability
“Effect of neutralization on the setting and of Clay in Terms of UCT Value.” Procedia
hardening characters of hemihydrate Engineering, 171, 484-491.
phosphogypsum plaster.” Construction and Kalkan, E. (2006). “Utilization of red mud as a
Building Materials, 190, 53-64. stabilization material for the preparation of
Davidovits, J. (2002). “Environmentally Driven clay liners.” Engineering Geology, 87(3-4),
Geopolymer Cement Applications.” 220-229.
Geopolymer 2002 Conference, (6), 1-9. Kim, B.-W. (2007). “Pullout Resistance of Geogrid
Deǧirmenci, N. (2008). “Utilization of Reinforced Granite Soil.” Journal of
phosphogypsum as raw and calcined material Experimental Psychology: General.
in manufacturing of building products.” Kim, Y. S., Dang, M. Q., Do, T. M., and Lee, J. K.
Construction and Building Materials, 22(8), (2018). “Soil stabilization by ground bottom
1857-1862. ash and red mud.” Geomechanics and
Engineering, 1, 105-112.
Djobo, J.-N. Y., Tchakouté, H.-K., Ranjbar, N.,
López López, E., Vega-Zamanillo, Á., Calzada
Elimbi, A., Tchadjié, L.-N., Njopwouo, D.,
Pérez, M. A., and Hernández-Sanz, A. (2015).
and Biernacki, J. (2016). “Gel Composition
“Bearing capacity of bottom ash and its
and Strength Properties of Alkali-Activated mixture with soils.” Soils and Foundations,
Oyster Shell-Volcanic Ash: Effect of 55(3), 529-535.
Synthesis Conditions.” Journal of the Mehta, P. K. (1983). “Mechanism of sulfate attack
American Ceramic Society, 99(9), 3159-3166. on portland cement concrete - Another look.”
Do, M. T., Kang, G. O., and Kim, Y. S. (2019). Cement and Concrete Research, 13(3), 401-
“Development of a new cementless binder for 406.
controlled low strength material (CLSM) Modarres, A., and Nosoudy, Y.-M. (2015). “Clay
using entirely by-products.” Construction and stabilization using coal waste and lime -
Building Materials, 206, 576-589. Technical and environmental impacts.”
Do, T. M., and Kim, Y. S. (2016). “Engineering Applied Clay Science, Elsevier B.V., 116-117,
properties of controlled low strength material 281-288.
(CLSM) incorporating red mud.” International Mun, K.-J., Hyoung, W.-K., Lee, C.-W., So, S.-Y.,
Journal of Geo-Engineering, 7(1), 7. and Soh, Y.-S. (2007). “Basic properties of
Ghadir, P., and Ranjbar, N. (2018). “Clayey soil non-sintering cement using phosphogypsum
stabilization using geopolymer and Portland and waste lime as activator.” Construction and
cement.” Construction and Building Materials, Building Materials, 21(6), 1342-1350.
188, 361-371. Nalbantoǧlu, Z. (2004). “Effectiveness of class C
fly ash as an expansive soil stabilizer.”
Güllü, H. (2014). “Factorial experimental approach
Construction and Building Materials, 18(6),
for effective dosage rate of stabilizer:
377-381.
Application for fine-grained soil treated with Ouhadi, V.-R., and Yong, R.-N. (2008). “Ettringite
bottom ash.” Soils and Foundations, 54(3), formation and behaviour in clayey soils.”
462-477. Applied Clay Science, 42(1-2), 258-265.
Hashemi, S.-S. G., Mahmud, H.-B., Djobo, J.-N. Park, S.-B., Jang, Y.-I., Lee, J., and Lee, B.-J.
Y., Tan, C.-G., Ang, B.-C., and Ranjbar, N. (2009). “An experimental study on the hazard
(2018). “Microstructural characterization and assessment and mechanical properties of
mechanical properties of bottom ash mortar.” porous concrete utilizing coal bottom ash
Journal of Cleaner Production, 170, 797-804. coarse aggregate in Korea.” Journal of
Hashemi, S.-S. G., Mahmud, H.-B., Ghuan, T.-C., Hazardous Materials, 166(1), 348-355.
Chin, A.-B., Kuenzel, C., and Ranjbar, N. Pollution Control & Ecology Commission. (2015).
(2019). “Safe disposal of coal bottom ash by Pollution Control and Ecology Commision
solidification and stabilization techniques.” Regulation No . 23.
Construction and Building Materials, 197, Pourakbar, S., and Huat, B.-K. (2017). “A review
705-715. of alternatives traditional cementitious binders

270
 for engineering improvement of soils.” Verbeek, C. J. R., and Du Plessis, B. J. G. W.
International Journal of Geotechnical (2005). “Density and flexural strength of
Engineering, 11(2), 206-216. phosphogypsum-polymer composites.”
Ranjbar, N., and Kuenzel, C. (2017). “Influence of Construction and Building Materials, 19(4),
preheating of fly ash precursors to produce 265-274.
geopolymers.” Journal of the American Yoon, Y.-I., Kim, Y.-E., Nam, S.-C., Park, S.-Y.,
Ceramic Society, 100(7), 3165-3174. Chun, I.-S., Lee, S.-D., and Kim, H.-S. (2017).
Razak, H. A., Naganathan, S., and Hamid, S. N. A. “Economic Analysis of CCU in the Korean
(2009). “Performance appraisal of industrial Cement Industry: CO2 Capture Using
waste incineration bottom ash as controlled KIERSOL & PCC Conversion.” Energy
low-strength material.” Journal of Hazardous Procedia, Elsevier B.V., 114(November
Materials, 172(2-3), 862-867. 2016), 6240-6245.
Ribeiro, D. V., Silva, A. S., Labrincha, J. A., and Yun, J.-W. (2015). “Behavior of Cement-Based
Morelli, M. R. (2013). “Rheological properties Fracture Grouting considering Characteristics
and hydration behavior of portland cement of Weathered Granite Soil.” Doctor
mortars containing calcined red mud.” dissertation, Korea University.
Canadian Journal of Civil Engineering, 40(6),
557-566.
Santhanam, M., Cohen, M. D., and Olek, J. (2003). Author Information:
“Effects of gypsum formation on the Tran Quoc Thien, Graduate student, Department of
performance of cement mortars during Civil and Environmental Engineering,
external sulfate attack.” Cement and Concrete Chonnam National University, Republic of
Research, 33(3), 325-332. Korea, [email protected]
Sathonsaowaphak, A., Chindaprasirt, P., and
Pimraksa, K. (2009). “Workability and
strength of lignite bottom ash geopolymer
mortar.” Journal of Hazardous Materials,
168(1), 44-50.
Sharma, A. K., and Sivapullaiah, P. V. (2016).
“Ground granulated blast furnace slag
amended fly ash as an expansive soil
stabilizer.” Soils and Foundations, Elsevier,
56(2), 205-212.
Sivapullaiah, P.-V., and Jha, A.-K. (2014).
“Gypsum Induced Strength Behaviour of Fly
Ash-Lime Stabilized Expansive Soil.”
Geotechnical and Geological Engineering,
32(5), 1261-1273.
Subbarao, G. and. (2001). “Micostructural
Development in Fly Ash Modified with Lime
and Gypsum.” 18(2), 377-382.
Tchadjié, L.-N., Djobo, J. N. Y., Ranjbar, N.,
Tchakouté, H. K., Kenne, B. B. D., Elimbi, A.,
and Njopwouo, D. (2016). “Potential of using
granite waste as raw material for geopolymer
synthesis.” Ceramics International, 42(2),
3046-3055.
Tikalsky, P. J., Bahia, H. U., Deng, A., and Snyder,
T. (2004). Excess foundry sand
characterization and experimental
investigation in controlled low-strength
material and hot-mixing asphalt.

271
The 3rd Int. Conf. on Transport Infrastructure & Sustainable Development (TISDIC 2019). Construction Publishing House. ISBN: 978-604-82-2893-4.

Bacterial enzyme induced calcite precipitation improving

silty sand soil strength

Hoang Tung 1,, James Alleman2, Bora Cetin3, Do Huyen 4

1
Faculty of Bridge and Road Construction Engineering, The University of Danang - University of
Science and Technology, Vietnam
2,3
Department of Civil, Construction, and Environmental Engineering, Iowa State University, Ames,
Iowa, USA, 50014
4
Faculty of Project Management, The University of Danang - University of Science and Technology, Vietnam

Corresponding author, E-mail: [email protected]

ABSTRACT: Using ordinary Portland cement for ground improvement is popular in construction field.
However, this method is high energy consumption and non-eco-friendly materials. The potential method
of microbial induced calcite precipitation (MICP) has been used to bio-stabilize sand for more than a
decade. Recently, a pure agricultural enzyme is used to similarly induce carbonate precipitation which
would again increase soil strength. However, both of these soil improvement methods still have inherent
disadvantages. For example, the MICP is limited in fine grained soils and the EICP research depends on
the commercial urease sources. This study investigates a new technique for bio-derived soil stabilization
based on a bacterial enzyme induced calcite precipitation (BEICP). The bacterial urease was collected by
using a simple sonication technique to break down viable cells of Sporosarcina pasteurii bacteria. The
sonication method is able to generate extracted enzyme far higher levels of urease activity, at levels
roughly 1.5 to 2 times higher compared to urease activity of the whole cells urease producing bacteria.
The BEICP method can form calcite crystals as bridges between fine (silt) and coarse (sand) grained soil.
The overall unconfined compressive strength of silty sand columns was higher than 1000 kPa. The
research also examines the micro-properties of BEICP-treated soil via scanning electron microscope
(SEM) and Energy Dispersive X-ray Spectroscopy (EDS) testing methods. The result indicates that
BEICP is a new approach of bio-derived method for soil improvement.
Keywords: Biocementation, biogeotechnics, enzyme, bacteria, soil stabilization.

1. INTRODUCTION process in geotechnical engineering fields

(from 2007 to 2018). The general field of
Natural soil properties vary significantly over microbial geotechnology has progressively
time and region. This variability and shifted into two different fields of
heterogeneity of the natural soils makes it application, including: (1) bio-clogging and
difficult to develop engineering soil (2) bio-cementation (Ivanov and Chu 2008).
improvement mechanisms which are able to The main aim of these applications is to
universally remediate all geotechnical soils enhance soil shear strength and to reduce the
(DeJong et al. 2011). Engineers have worked permeability of natural soils in order to
with a variety of methods to stabilize weak prepare these materials for construction
soils for different intended engineering and/or environmental remediation. However,
purposes. Recently, many studies have been research conducted on micro-organism
conducted to develop the suitable process, i.e., which is called ‘microbial
methodology for using micro-organism induced calcium carbonate precipitation’

272
(MICP), over the last decade has shown that 2. MATERIALS AND METHODOLOGY
the MICP method still has some limitations.
One distinct issue is that MICP treatment 2.1 Biomass culturing, enzyme extraction and
employs urease microbes who have an chemical solution
inherent constraint on the ability to This project’s commercial bacterial isolate,
physically migrate through soils unless their Sporosarcina pasteurii ATCC-11859, was
pore space voids are larger than that of obtained from the American Type Culture
medium to fine sands (Kavazanjian and Collection (Manassas, Virginia). The pre-
Hamdan 2015). In turn, MICP is probably not
sterilized culturing broth used an ‘ammonium-
suited for soil systems bearing finer-grained
trypticase soy broth’ (NH4 -TSB) growth
materials which would then impede bacterial
media which included: tryptic soy broth at 20
migration. A newer ureolysis method using
nano-scale and water-soluble urease enzyme g/L, ammonium sulfate at 10 g/L, Tris buffer at
source also induces carbonate precipitation 0.13 mol/L, and an overall solution pH of 9.0.
via ureolysis. The enzyme induced carbonate After incubation, this stock microbial culture
precipitation (EICP) method has been was then stored at 4 °C prior for subsequent
investigated for increasing strength, use. Urease enzyme lysed directly from live
permeability reduction, and mitigation of bacteria of Sporosarcina pasteurii was used in
fugitive dust by Nemati and Voordouw the present study. The procedures of extraction
(2003), Nemati et al. (2005), Neupane et al. and optimization method using sonication
(2013); Hamdan et al. (2013); Park et al. technique have been described previously
(2014); Kavazanjian and Hamdan (2015); (Hoang et al. 2018). The sonication method
Hamdan and Kavazanjian (2016). Most of typically produced an approximate urease
this current research has used an agricultural activity rate of 25.4 mM urea per min. The
urease which had been purchased from a concentrations of chemical solution used in the
commercial chemical company. Extraction of current study were 0.3 M at the same molar
urease from plants requires time (plant concentrations for soluble CaCl2 and urea.
rearing) and space, and it is produced in
small amounts. 2.2 Testing soil
One significant goal of this study, This study used Ottawa 20/30 silica standard
therefore, was to develop a better method to sand for a coarse-grained soil and loess soil
produce the urease enzyme, where the passing sieve No. 200 (0.075mm) for the fine-
approach is not only simpler but also able to grained soil. The Ottawa 20/30 silica sand as
generate higher urease activities than the described in ASTM C778 was used in the
ones observed within MICP method. This experiments. The sand properties have a
new method is called bacterial-enzyme particle diameter at 10% finer by mass (D10) of
induced calcite precipitation (BEICP). It was 0.58 mm, a particle diameter at 50% finer by
consequently used to treat a group of test mass (D50) of 0.72 mm, a coefficient of
columns with different percentages of fine- uniformity (Cu) of 1.17, a coefficient of
grained soils by weight. A particular curvature (Cc) of 1.02, specific gravity (Gs) of
advantage of using this new method is that 2.65, and emax and emin of 0.74 and 0.51,
we are now able to generate far higher levels respectively. The Iowa natural loess soil was
of urease enzyme activity, at levels roughly 2 used in this study as a fine-grained soil. The
times higher than those obtained with the loess soil included 0.7% of sand, 86.5% of silt,
conventional MICP method. Yet, another and 12.8% of clay. The loess soil has 34.5% of
clear advantage for the BEICP method is that liquid limit and 27.3% of plastic limit. The
this process can form calcite crystals as loess soil was sieved through U.S. Sieve
bridges between fine- (silt) and coarse- No.200 to collect only silt and clay particles
(sand) grains soil which increases the overall (particle size less than 0.075mm). Three
strength of silty-sand columns. different soil blends were evaluated, including:

273
1) 100 % sand and 0 % fines, 2) 90 % sand and cells or extracted enzyme bond on soil
10 % fines, and 3) 80 % sand and 20 % fines. particle surfaces; (2) biological liquid was
These three options are subsequently referred drained off the soil column; (3) a mixed
to as: 1) 100-0, 2) 90-10, and 3) 80-20. The chemical solution of urea and calcium
grain size distribution curves of tested soil chloride (0.3 M by 1:1 ratio) was used and
were shown in Figure 1. circulated for 9-12 h; (4) the soil column was
fluxed by deionized water for 2 h to remove
soluble byproducts and then the bottom cap
was removed to drain off all liquid for
approximately 10 h. After one cycle (urease
enzyme and urea/calcium chloride solution),
fresh biological solution and chemical were
used to recirculate in the soil column. Such
an enzyme solution and urea/calcium
chloride solution were repeated one cycle a
day for 8, 12, and 16 days.

Figure 1. Grain size distribution curves of tested

soils (after Hoang et al. 2018)

2.3 Soil column preparation and treatment

procedures
Soil stabilization columns were prepared for
three different soil mixtures with both sand-
only and silty-sand soils. A moist-tamping
method was then used to pack the specimens,
where this pre-moistened soil was gently
tamped in PVC columns with 50 mm
diameter and 100mm height dimensions. To
Figure 2. Soil column circulating-percolation treatment
achieve a similar specific density in each
layer, pre-determined amounts of soil were
added in ten successive layers of equal 3. RESULTS AND DISCUSSIONS
thickness (i.e., at 10 mm per layer) within 3.1 Sand and silty sand soil column cemented
each column. These column tamping steps by BEICP
were carefully conducted to achieve the
similar void ratio (e~0.60) levels within all The main purpose of this work is to form calcite
column samples. crystal bonding which can cement sand silt
The sandy soil or silty sand soil columns mixture soil. A new bio-cementation process
were treated using the following procedure employed bacterial urease accompanied with a
(Figure 2): (1) biological solution (extracted solution containing urea and calcium chloride,
urease for BEICP method or bacterial cells which was to success forming intact silty sand
for MICP method) was pumped into the column specimens. The successful capability for
sample from the top and gravity drained out BEICP processing to stabilize a full-depth
from the bottom. A peristaltic pump with sample column of silty-sand soils, though, is
silicone tubing was used to recirculate visually confirmed by the accompanying
biological liquid for 3 h with the rate of 0.8- photograph given in Figure 3 (e.g., after an
1.2 ml/minute which could keep bacterial eight-cycle treatment cycle).

274
 The relationship between fine content and
amount of calcium carbonate precipitation is
shown in Figure 4. The increase of
percentage of fine-grained in column sample
gives rise to CaCO3 content at the same
treatment cycles. However, the increasing
rate is slightly slow from 12 to 16 cycles of
treatment. Therefore, the strength of BEICP
samples of silty sand soil does not only
depend on carbonate content but also relate to
fine content.

Figure 3. Original BEICP-treated 80-20 sample

after eight treatment cycles (CaCO3
content = 4.74% (w/w), and UCS =
214 kPa)

3.1.1 Unconfining compressive strength

(UCS) of BEICP-treated soil
The UCS data in Figure 4 indicated that
increasing the strength of BEICP samples is a
result of rising the calcium carbonate content
correlated to increasing of numbers treatment
cycles. For example, samples of 90% of sand Figure 4. UCS versus CaCO3 precipitation of
and 10% of fine increase the CaCO3 content BEICP samples at different number of
treatment cycles of BEICP-treated soil
from 4.48% after 8th cycle to 14.45% at 16th
cycle correspond to 412 and 1088 kPa of UCS 3.1.2 Hydraulic conductivity of BEICP-
values respectively. The highest strength is treated soil
1539 kPa for the 100% sand cemented sample
with 7.89% of calcium carbonate content at Figure 5 shown the effect of fine content and
16th cycle while the cemented sample of 80% percentage of calcium carbonate
sand and 20% fine at 8th cycle provides the precipitation on the permeability of the
lowest strength of 214 kPa at 4.47% of the BEICP-treated samples. As an overall trend,
CaCO3 content. it is clear that the increase in the percentage
Kavazanjian and Hamdan (2015) also of both fine-grained soil and CaCO3
observed an increased level of CaCO3 precipitation correlates to the reduction of
formation with EICP processing after having permeability of cemented samples. After the
added bentonite to sand. Interestingly, 16th cycle of treatment, the permeability of
though, their EICP tests were not able to 80% sand and 20% fine BEICP-treated
form an intact, stabilized column sample. sample reduces from 2.56 x 10-4 m/s to 4.98
Oliveira et al. (2016) similarly report x 10-7 m/s at 14.16% (w/w) of CaCO3
elevated CaCO3 precipitation levels (reaching content. There is a clear possibility that the
17%) being developed during EICP tests on a increase of the amount of fine content took
heavily compacted sand-silt-clay soil mixture the place of void spaces of soil matrix which
(i.e., at respective 73.4%, 23.8%, and 2.8% caused the falling of hydraulic conductivity.
levels), and here again their finished product In addition, repeated of treatment cycles of
UC strength (i.e., ~250 kPa) was only BEICP-treatment accumulated calcite
slightly higher than their untreated, non- crystals results in the increase of the pores
stabilized material (i.e., ~175 kPa). clogging (Chu et al. 2012).

275
 strength of sand column. In addition, the
Calcite clusters fiiled the pore space in sand
matrix which is the result in reduction of
permeability. The area 1 in Figure 7 (a) was
selected to examine the spectrum of Si and Ca
minerals. Figure 7(d) presented the peak of Si
and Ca which confirmed the formation of
Calcite in sand matrix after BEICP treatment.

Figure 5. Permeability versus CaCO3 precipitation

of BEICP-treated samples at different
number of treatment cycles

3.1.3 Microscale investigation of biological

precipitation
To examine the physical characteristics of
bacterial enzyme bio-cemented soil including
the mineral phases and the pore structure, the
SEM, was performed on intact cementation
specimens. Figure 6 was a series of SEM
images displayed from a single enzymatic
cemented sand silt mixture specimen at various Figure 6. SEM images of BEICP-treated silty-sand
(80% sand plus 20% fine) soil: (a) sand
magnifications. Figure 6(a) showed the mixture
grains are separated by mixture of fine-
of fine-grains and calcium carbonate grained soil and calcite crystals
precipitation fill the void spaces of the soil precipitated; (b) (in color) calcium
sample. The sand grains were mainly separated mineral in red and aluminum mineral in
by fine-grained soil which was bound by green; (c) fine grains are bound by calcite
calcite crystals (red color), as shown at Figure crystals; (d) a single fine grain is attached
6(b). Figure 6(c) zoomed closer (500x) to the by calcite crystals
location where fine particles linked together
via the calcite crystal precipitated. The calcite
crystals also form in pore spaces between fine
grains. The Figure 6(d) showed a visible
evidence of calcite crystals bond the fine
particle and sand particle together. Therefore,
the mixture of fine grains and calcite crystals
precipitated formed a cemented-bridge to link
sand grains together which could increase
strength of silty sand soil sample.
To confirm the formation of Calcite mineral
in sand matrix after BEICP process, the EDS
test was performed with the BEICP-treated Figure 7. SEM and EDS results of BEICP-treated
sand after 16 cycles of treatment (Figure 7). sand (100% sand plus 0% fine) after 16
The EDS images in Figure 7 (b, c) revealed the cycles of treatment: (a) SEM at 50x (b) Si
Ca minerals (CaCO3) stuck around the Si mineral overlay, (c) Ca mineral overlay,
minerals (sand). The CaCO3 precipitated at the (d) EDS spectrum of biocemented sample
contact points of sand particles increased the at area 1, containing detectable Si, Ca

276
4. CONCLUSION Hoang, T., Alleman, J., Cetin, B., Ikuma, K., and
Choi, S.-G. (2018). “Sand and silty-sand soil
This paper reports a new method of bio- stabilization using bacterial enzyme induced
mediated soil stabilization, as well as a calcite precipitation (BEICP).” Canadian
corresponding assessment of bio-stabilization Geotechnical Journal, 1–66.
soil properties at both a micro-structural-scale Ivanov, V., and Chu, J. (2008). “Applications of
and at a macro-soil-scale under various microorganisms to geotechnical engineering
conditions. for bioclogging and biocementation of soil in
- The BEICP method proved to be capable situ.” Reviews in Environmental Science and
Biotechnology, 7(2), 139–153.
of improving soil strength and reducing
Kavazanjian, E., and Hamdan, N. (2015). “Enzyme
permeability when applied to non-plastic sands induced carbonate precipitation (EICP)
as well as two levels of low plasticity 90-10 columns for ground improvement.” Proc.,
and 80-20 (i. e., respective percentile fractions ASCE IFCEE 2015, M. Iskander, M. T.
for coarse-fine grain) soil materials. The UCS Suleiman, J. B. Anderson, and Debra F.
of biocemented soil increased when the Laefer, eds., ASCE, San Antonio, Texas,
calcium carbonate content increased. However, 2252–2261.
the amount of fine content reduced the strength Nemati, M., Greene, E. A., and Voordouw, G.
of biostabilized soil. (2005). “Permeability profile modification
- The microstructural analysis indicated that using bacterially formed calcium carbonate:
the BEICP method can form calcium carbonate Comparison with enzymic option.” Process
at the contact point of sand and fine particles. Biochemistry, 40(2), 925–933.
Nemati, M., and Voordouw, G. (2003).
These Calcite clusters bind soil particle
“Modification of porous media permeability,
together which increases the UCS of soil. In using calcium carbonate produced
addition, the permeability of soil columns was enzymatically in situ.” Enzyme and Microbial
reduced due to the filling CaCO3 in pore space Technology, 33(5), 635–642.
of soil matrix. Neupane, D., Yasuhara, H., Kinoshita, N., and
Unno, T. (2013). “Applicability of enzymatic
5. REFERENCES calcium carbonate precipitation as a soil-
strengthening technique.” ASCE Journal of
Chu, J., Stabnikov, V., and Ivanov, V. (2012). Geotechnical and Geoenvironmental
“Microbially induced calcium carbonate Engineering, 139(December), 2201–2211.
precipitation on surface or in the bulk of Oliveira, P. J. V., Freitas, L. D., and Carmona, J. P.
soil.” Geomicrobiology Journal, 29(March), S. F. (2016). “Effect of soil type on the
544–549. enzymatic calcium carbonate precipitation
DeJong, J. T., Soga, K., Banwart, S. a, Whalley, process used for soil improvement.” Journal of
W. R., Ginn, T. R., Nelson, D. C., Mortensen, Materials in Civil Engineering, 29(4), 1–7.
B. M., Martinez, B. C., and Barkouki, T. Park, S.-S., Choi, S.-G., and Nam, I.-H. (2014).
(2011). “Soil engineering in vivo: harnessing “Effect of microbially induced calcite
natural biogeochemical systems for precipitation on strength of cemented sand.”
sustainable, multi-functional engineering Journal of Materials in Civil Engineering,
solutions.” Journal of the Royal Society, 26(8), 47–56.
Interface / the Royal Society, 8(54), 1–15.
Hamdan, N., and Kavazanjian, E. (2016).
“Enzyme-induced carbonate mineral
precipitation for fugitive dust control.”
Géotechnique, 66(7), 546–555.
Hamdan, N., Kavazanjian, J. E., and S., O.
(2013). “Carbonate cementation via plant
derived urease.” Proceedings of the 18th
International Conference on Soil Mechanics
and Geotechnical Engineering, Paris 2013,
2489–2492.

277
The 3rd Int. Conf. on Transport Infrastructure & Sustainable Development (TISDIC 2019). Construction Publishing House. ISBN: 978-604-82-2893-4.

Study on the relationship between the plasticity index and

the key parameters of the vacuum consolidation process

Pham Quang Dong1,, Tran The Viet2, Trinh Minh Thu3

1
Central Regional College of Technology – Economics and Water Resources, 14 Nguyen Tat Thanh,
Hoi An, Quang Nam
2
Department of Geotechnical Engineering, Thuyloi University, 175 Tay Son, Dong Da, Ha Noi
3
Department of Geotechnical Engineering, Thuyloi University, 175 Tay Son, Dong Da, Ha Noi

Corresponding author, E-mail: [email protected]

ABSTRACT: This study addressed the relationship between the plasticity index (PI) and the key
parameters of the vacuum consolidation process including the consolidation time (t), the degree of
consolidation (U), and the thickness of the soft soil layer (H) when improving soft ground by vacuum
consolidation method. For this purpose, the coupled analysis technique which combines SEEP/W and
SIGMA/W modules in Geostudio software is applied for the simulation. Three different foundations
composing of different soft soils in Dinh Vu- Hai Phong, Duyen Hai - Tra Vinh, Nhon Trach - Dong Nai
were considered. The suitability of the predicted numerical models was calibrated using experimental
results of laboratory and field tests. The results of this study, therefore, help technical workers to quickly
give suitable predictions about the process of consolidation when applying vacuum consolidation method
in soft soil treatment.
Keywords: Vacuum consolidation, plastic Index, degree of Consolidation, soft soil, coupled analysis.

1. INTRODUCTION International airport. Since then it has become

well known and broadly accepted for
As a developing country with rapid growth, the improving very soft soil deposits, a more cost-
demand for infrastructure development on soft effective solution for projects worldwide
and compressible clayey soil in Vietnam (Griffin & O'Kelly, 2013).
increases significantly with the rise in the The effectiveness of the prefabricated
population, especially in urban areas. Such vertical drains (PVD) combined with vacuum
critical geotechnical condition causes many preloading has been discussed by many
challenges to the constructions of researchers. Comparing to other ground
infrastructures such as low bearing capacity, improvement approaches for a large area,
excessive settlement and slope failure (Xu et vacuum preloading is one of the most cost-
al., 2018). Preloading coupled with vertical effective methods in foundation engineering
drains has been a widely used method for soft practice (Indraratna et al., 2016; Kumar et al.,
ground improvement in engineering practice. 2015; Xu et al., 2018). According to Indraratna
The combination of vacuum preloading and et al. (2005), the behavior of soft clay
vertical loading has been proved to be an foundations stabilized by vertical drains is now
efficient solution for the treatment of soft soil able to give predicted results with satisfactory
foundation (Chai et al., 2005). The technique accuracy due to considerable progress that has
was originally suggested by Kjellman (1952) to been made in the past decade through rigorous
improve the soft foundation of the Philadelphia numerical analysis.

278
 Theoretically, the surcharge preloading has namely: Dinh Vu – Hai Phong City; Duyen Hai
the role to exert extra stress in the treated thermal power plant - Tra Vinh Province, and
foundation. This stress helps to compact the Nhon Trach thermal power plant – Dong Nai
voids induced by water being drained out by Province were considered. The soil is classified
electroosmosis and vacuum (Liu et al., 2014). following the unified soil classification system.
Therefore, comparing to the conventional All the necessary properties of the soft soils for
methods, the vacuum preloading method has the consolidation modeling of the selected sites
several advantages including 1) no fill material are determined by laboratory tests using
is required, 2) the construction periods are ASTM. Their average values are presented in
generally shorter and 3) the construction does Tables from Table. 1 to Table. 3.
not require heavy machinery (Chai et al., Table 1. Properties of soft soil in Dinh Vu –
2005). The presence of the vertical drains helps Hai Phong
to reduce the water drainage path and speed up
the dissipation of excess pore pressure (Chai et Soil parameters Unit Soil layer
al., 2006; Kumar et al., 2015). However, the Unit weight kN/m3 17.000
method is not effective when applying on soft
clay with very low hydraulic conductivity (Liu Permeability m/s 2.9 × 10-8
et al., 2014), as it is very hard to drain the Poisson’s ratio - 0.260
water out of fine soils by surcharge preloading Elastic modulus kPa 0.310
or vacuum preloading. Plasticity index % 24.640
Therefore, it can be stated that the system of Soil type - SC-CH
PVD subjected to vacuum preloading is a
beneficial technique, it helps to accelerate Table 2. Properties of soft soil in Duyen Hai –
radial consolidation and reduce a large rate of Tra Vinh
surcharge load (Indraratna et al., 2005). This
Soil parameters Unit Soil layer
study addresses the relationship between the 3
plasticity index and the key parameters of the Unit weight kN/m 16.40
vacuum consolidation process including the Permeability m/s 2.9×10-8
consolidation time (t), the degree of Poisson’s ratio - 0.27
consolidation (U), and the thickness of the soft
Elastic modulus kPa 3380.0
soil foundation (H) when improving soft
Plasticity index % 18.40
ground by vacuum consolidation method. For
this purpose, the coupled analysis technique Soil type - CH
which combining SEEP/W
Table 3. Properties of soft soil in Nhon Trach –
(Geoslope_International_Ltd, 2019a) and
Dong Nai
SIGMA/W (Geoslope_International_Ltd,
2019b) modules in Geostudio software were Soil parameters Unit Soil layer
utilized. The suitability of these numerical kN/m 3
14.17
Unit weight
models was calibrated using experimental
results of laboratory and field tests. Permeability m/s 2.94×10-8
Poisson’s ratio - 0.27
2. SITES FOR THE INVESTIGATION Elastic modulus kPa 230
In this study, three different types of soft soils Plasticity index % 33.8
in different project sites around Vietnam Soil type - MH

279
3. BOUNDARY CONDITIONS Table 4. Degree of consolidation of selected
soft soils
The study utilized the coupled analysis
technique combining SEEP/W Soft layer Time to get % degree of
Site PI (%) thickness consolidation (day)
(Geoslope_International_Ltd, 2019a) and (m)
SIGMA/W (Geoslope_International_Ltd, 80% 85% 90% 95%

2019b) in Geostudio to model the Vacuum 10 13,78 19,11 26,76 37,07

preloading combined with vertical drain Duyen 15 18,40 25,86 35,52 47,29
Hai
problem. To establish relationships between Tra
18,40 20 20,13 30,12 44,25 60,89
key parameters of the vacuum consolidation Vinh 25 23,33 36,20 52,78 77,72
process when considering selected soft soils, 30 25,55 39,01 59,62 90,70
we applied boundary conditions that reflect 10 18,61 25,99 35,46 46,82
the in situ. To assess the influence of the Dinh 15 23,69 32,71 44,38 58,46
treated soil depth, numerical models with five Vu
24,64 20 27,25 40,65 58,29 81,45
different scenarios for the soft soil thickness Hai
Phong 25 28,14 43,75 65,59 93,73
including 10m; 15m; 20 m; 25m; and 30m
30 31,82 50,10 75,28 109,17
were considered for the simulation and
analysis. The common vertical drain with a 10 28,33 39,00 51,01 62,69

cross section of 100mm × 4mm, a length Nhon 15 36,06 47,35 60,77 77,99
Trach
equal to the depth of the soft soil layer Dong
33,80 20 40,16 55,43 78,11 106,27
(varying from 10 to 30m as indicated) were Nai 25 46,18 65,25 90,39 128,32
installed in a triangular pattern at an effective 30 48,72 72,10 103,85 151,06
spacing of 1.0m.
As can be seen in Table 4, for all types of
The loading process was divided into two
studied soil, higher values of the soft soil
separated stages. In the first stage, a 0.5m thickness result in a longer time to reach a
sand blanket (γsand = 16kN/m3) was placed specific degree of consolidation. In addition, to
on the ground surface to create the platform reach the same degree of consolidation, soft soils
for installing the horizontal perforated pipes. with larger values of the plasticity index require
At the same time, an average vacuum more time. The corresponding relationship
pressure of – 55.0 kPa was applied and kept between the degree of consolidation and the time
constant for a period of 10 days. In the corresponding to various scenarios of the
second stage (after 10 days), we increased thickness of the foundation are illustrated from
the surcharge represented by the sand layer Figures 1 to Figure 5.
to 1.5 m and the vacuum pressure was also
rose to 89 kPa.

4. RESULTS AND ANALYSIS

4.1 Relationship between the degree of

consolidation, the thickness of the soft soil
layer and the time
The simulated degree of consolidation of
selected soft soil foundations corresponded to
Figure 1. Relationship between the degree of
different thickness of the treated soft soil was consolidation and time when the soft
presented in Table 4. soil layer is 10 m

280
Figure 2. Relationship between the degree of Figure 3. Relationship between the degree of
consolidation and time when the soft soil consolidation and time when the soft soil
layer is 15m layer is 20m

Figure 4. Relationship between the degree of Figure 5. Relationship between the degree of
consolidation and time when the soft consolidation and time when the soft
soil layer is 25m soil layer is 30m

4.2 Establish the relationship between the ground is defined to be from 10m to 30m. It is
plasticity index and the key parameters of the possible to consider the consolidation time (t)
vacuum consolidation process is a function of the plasticity index (PI), the
degree of consolidation (U%) and the
The speed of the consolidation of the thickness of the treated soft foundation (H)
foundation depends considerably on its which is a constant.
permeability and the thickness of the drainage + Case 2: to establish the above relationship
path (Devi et al., 2015; Shang et al., 1998; when the degree of consolidation is defined. In
Solanki & Desai, 2008; Sridharan & Honne, this case, we considered the consolidation time
2004). However, it is often more complicated to (t) as a function of the plasticity index (PI), the
define the relationship between the degree and thickness of the treated soft soil layer (H), and
time of consolidation using traditional methods. the degree of consolidation (U%) which is a
Therefore, in this study, we used the plasticity constant in this case.
index, a common soil parameter which is easily For the first situation, using the plasticity
defined using simple laboratory tests. To make it index given in Table 1 combining with
easy in making predictions of the change of numerical predictions, the results in establishing
parameters of the consolidation process when relationships between parameters of the vacuum
applying vacuum preloading method for soft soil consolidation in soft ground treatment
treatment, we considered the following cases: correspond to various ground thickness (H = 10
+ Case 1: To establish the above m to 30 m) are presented in Figures from Figure
relationship when the thickness of the soft 6 to Figure 10.

281
Figure 6. Relationship between the consolidation
time, the plasticity index, and the degree
of consolidation when the thickness of Figure 9. Relationship between the consolidation
the soft ground foundation is 10m. time, the plasticity index, and the
degree of consolidation when the
thickness of the soft ground foundation
is 25m.

Figure 7. Relationship between the consolidation

time, the plasticity index, and the degree
of consolidation when the thickness of
the soft ground foundation is 15m.

Figure 10. Relationship between the consolidation

time, the plasticity index, and the
degree of consolidation when the
thickness of the soft ground foundation
is 30m.

For the second case, using the plasticity

index presented in Table 1 combining with
the predicted results by numerical models, it
is possible to establish the relationships
between key parameters of the vacuum
consolidation in soft ground treatment
Figure 8. Relationship between the consolidation correspond to various degree of consolidation
time, the plasticity index, and the degree values of U = 80%, 85%, 90%, 95%. The
of consolidation when the thickness of results are shown in Figures from Figure 11
the soft ground foundation is 20m. to Figure 14.

282
 when the thickness of the soft ground varies
from 10 m to 30 m. It can be seen that, for all
cases, it is possible to link the consolidation
time and the plasticity index by a linear
relationship. Higher values of plasticity index
results in a longer time of consolidation and the
larger the thickness of the treated soft
foundation, the clearer the above relationship.

Figure 11. Relationship between the consolidation

time, the plasticity index and the
thickness of the soft layer when the
degree of consolidation U = 80%

Figure 13. Relationship between the consolidation

time, the plasticity index and the
thickness of the soft layer when the
degree of consolidation U = 90%

Figure 12. Relationship between the consolidation

time, the plasticity index and the
thickness of the soft layer when the
degree of consolidation U = 85%

Figures from Figure 6 to Figure 14 show the

relationship between the time of consolidation,
the plasticity index, the thickness of the treated
Figure 14. Relationship between the consolidation
soft soil layer, and the degree of consolidation time, the plasticity index and the
of three types of soft soils when improving by thickness of the soft layer when the
vacuum preloading method. These degree of consolidation U = 95%
relationships are represented by equations
denoting as t80, t85, t90, and t95. It should be 5. CONCLUSION
noted that t80, t85, t90, and t95 are the time (t) to
get the corresponding degree of consolidation From the achieved results when establishing the
of 80%, 85%, 90%, 95%. On the other hand, relationships between key parameters of the
t10, t15, t20, t25, and t30 are the consolidation time vacuum consolidation using numerical method,

283
it can be concluded that the established with Index Properties Of Different Soils in
equations of t80, t85, t90, t95 and t10, t15, t20, t25, t30 Manipur Valley. International Journal of
Engineering Research and Development, 11(5),
show the relationship between the consolidation 57-63.
time, plasticity index, degree of consolidation Geoslope_International_Ltd. (2019a). Seep/W user's
and the thickness of the treated soft ground of guide for finite element analyses. Calgary, Alberta,
different kinds of soft soils when the degree of Canada: Geoslope International Ltd.
consolidation and the thickness of the treated Geoslope_International_Ltd. (2019b). Sigma/W user's
soft ground are defined. Based on equations guide for slope stability analyses. Calgary,
Alberta, Canada: Geoslope International Ltd.
defining t80, t85, t90, t95 and t10, t15, t20, t25, t30, it is Griffin, H., & O'Kelly, B. (2013). Ground improvement
possible to predict the consolidation time and by vacuum consolidation – a review. Ground
the degree of consolidation for soft clays using Improvement, 167(GI4), 274–290.
the value of the plasticity index (PI) when doi:http://dx.doi.org/10.1680/grim.13.00012
improving the soft foundation applying vacuum Indraratna, B., Kan, M. E., Potts, D., Rujikiatkamjorn, C.,
& Sloan, S. W. (2016). Analytical solution and
consolidation method. Therefore, constructing numerical simulation of vacuum consolidation by
such relationships may help engineers to have vertical drains beneath circular embankments.
initial predictions about the consolidation Computers and Geotechnics, 80, 83-96.
processes in soft ground treatment applying doi:http://dx.doi.org/10.1016/j.compgeo.2016.06.008
vacuum consolidation. Indraratna, B., Sathananthan, I., Rujikiatkamjorn, C., &
Balasubramaniam, A. S. (2005). Analytical and
Numerical Modeling of Soft Soil Stabilized by
6. RECOMMENDATIONS Prefabricated Vertical Drains Incorporating
Vacuum Preloading. International Journal of
In this study, we only established relationships Geomechanics, 5(2), 114-124.
between mentioned parameters corresponding doi:10.1061/(ASCE)1532-3641(2005)5:2(114)
to various ground thickness (H = from 10 m to Kjellman, W. (1952). Consolidation of clay soil by means
30 m) and degree of consolidation U = 80% to of mamospheric pressure. Paper presented at the
95%. In this case, the effective distance Proc. Conf. on Soil Stabilization, Boston, USA.
Kumar, S. G., Sridhar, G., Radhakrishnan, R.,
between drains is 1m and the common Robinson, R. G., & Rajagopal, K. (2015). A case
dimensions of vertical drains are 4×100mm. study of vacuum consolidation of soft clay deposit.
The surcharge value which has been commonly Indian Geotech J, 45(1), 51-61.
used for soft soil improvement in Vietnam was Liu, H., Cui, Y., Shen, Y., & Ding, X. (2014). A New
utilized. Therefore, for the future works, it is Method of Combination of Electroosmosis, Vacuum
necessary to establish the relationship between and Surcharge Preloading for Soft Ground
Improvement. China Ocean Eng, 28(4), 511-528.
the plasticity index and the key parameters of doi:10.1007/s13344-014-0042-3, ISSN 0890-5487
the vacuum consolidation process for more Shang, J. Q., Tang, M., & Miao, Z. (1998). Vacuum
loading scenarios with a varying value of preloading consolidation of reclaimed land: a case
surcharge as well as the time to maintain it. study. Can. Geotech. J., 35, 740–749.
Moreover, the effect of the effective distance Solanki, C. H., & Desai, M. D. (2008). Role of
Atterberg limits on time rate settlement of alluvial
between drains should also be considered. deposits. Journal of Engineering and Technology,
Sardar Patel University, 21, 12-15.
7. REFERENCES Sridharan, A., & Honne, N. (2004). Coefficient of
Consolidation and its Correlation with Index
Chai, J. C., Carter, J. P., & Hayashi, S. (2005). Ground Properties of Remolded Soils. Geotechnical
Deformation Induced by Vacuum Consolidation. Testing Journal, 27(5), 1-6.
Journal of Geotechnical and Geoenvironmental doi:10.1520/GTJ10784
engineering, 131(12), 1552. Xu, Y., Ye, G., & Zhang, Z. (2018). Consolidation
doi:10.1061/(ASCE)1090-0241(2005)131:12(1552) Analysis of Soft Soil by Vacuum Preloading
Chai, J. C., Carter, J. P., & Hayashi, S. (2006). Vacuum Considering Groundwater Table Change. Paper
consolidation and its combination with presented at the GeoShanghai 2018 International
embankment loading. Can. Geotech. J, 43, 895- Conference: Ground Improvement and
996. doi:10.1139/T06-056 Geosynthetics, Shanghai.
Devi, S., P, Devi, K. P., Prasad, D. S. V., & Raju, G. V.
R. (2015). Study on Consolidation and Correlation

284
The 3rd Int. Conf. on Transport Infrastructure & Sustainable Development (TISDIC 2019). Construction Publishing House. ISBN: 978-604-82-2893-4.

Determine fineness modulus of depositional coarse-grained

soil in replacement for alluvial sand in Thua Thien Hue
coastal plain, Vietnam

Canh N.V., Do Quang Thien.

Hue Science University, Hue, Viet Nam.

*Corresponding author: [email protected]

ABSTRACT: The main aim of paper to propose new approach method for fineness module (Ms)
determination of granular soil (sand) for using as natural building materials, which base on soil
particle composition in Huong River and follow Vietnam engineering standard (TCVN) 7570:2006.
Simultaneously, apply the proposed method to determine fineness module of sand in soil particle
composition report of 725 samples collected from engineering geological investigation in Thua Thien
Hue coastal plain. The results present fineness module is calculated by 2 methods without significant
differences and value of Ms which follows TCVN standard 4198:2014 has more reliable. The
granular soil in research area has grain composition, impurities and fineness module (Ms = 0.99-2.56)
meet technical requirements of fine aggregates (fine sand and coarse sand) for concrete and mortal.
Key words: Non-cohesion materials, granular soils, fineness module, coastal plain.

1. INTRODUCTION 7570:2006; TCVN 7570:2006; TCVN 7572:2-

2006). However, almost databases are so
Our research illustrates that there are many localized and limited, that lead to the
documents on geological investigation for difficulties in research and determination the
construction in Thua Thien Hue territory, fineness modulus of sand and gravel for
especially the results of granular soil analysis concrete and mortar. As for several countries
(sand, gravel) with diversity, abundance, these around the world, the grain size distribution
soils are experimented with grain size determination is completely different from
distribution follows sieve analysis from Vietnam. Mostly, the standard for soil
Vietnam engineering standard (TCVN) 4198: classification in every countries are not similar
2014 (Ho Vuong Binh 1992 and et al; TCVN even some of them that cannot classified soil,
4198:2014). These are valuable data but they gravel and rubbles then they combine those
are not exploited and utilized for researching, coarse grain soils into a group called sandy soil
exploring engineering material mines which with gradation from small to coarse (Table 1)
distribute inside Quaternary sediments. (H. Zhang 2011; S. K. Duggal 2015; ASTM
Meanwhile, the determination of grain size C33; ASTM C136) in contrast with Vietnam
distribution for gravel and sand in the serve of engineering standard (TCVN) 7570:2006, it is
calculating the design and production of able to classify clearly the gradation of sand,
concrete and mortar in our country comply gravel - rubble. Nowadays, there are still no
follow TCVN 7570: 2006 and TCXD 127: any research projects concern to this topic in
1985 (Phung Van Lu and et al. 2002; TCVN Vietnam and all the world.

285
 Therefore, in order to apply completely Table 1. The grain size of sand, gravel, rubble
the most of the results of granular soil follow authors and several standards
particle analysis in the geological Conceptual
investigation for constructions (Ho Vuong Vietnam
Binh 1992 and et al), we strongly engineering TCVN ASTM BS JIS
standard (Vietnam) (USA) (UK) (Japan)
recommend a new approachable method in (TCVN
the classification of grain soil gradation to 4198:2014)
calculate fineness module of granular soil in
0,075
propose for engineering material in coastal
0,10 0,14 0,15 0,15 0,15
plain of Thua Thien Hue provinces with the
0,25 Sand 0,315Sand 0,30 0,212 0,30
desire to contribute to the improvement of
0,5 0,63 0,60 0,30 0,60
the theoretical foundation in the study of
1 1,25 1,18 0,425 1,18
grading for engineering materials such as
2 2,50 2,36 0,60 2,36
sand, gravel in nature.
5 5 4,75 0,85 4,75
2. RESEARCH METHOD 10Gravel 10 Gravel 9,50 1,18 9,50
20Rubble 20 Rubble 12,5 1,70 16
The current situation of sand and gravel 40 40 19 2,36 19
classifications in the fineness module present 70 70 25,5 3,35 26,5
that in several countries the classified 100 100 37,5 5 31,5
standard cannot distinguish between sand 50 6,30 37,5
and gravel - rubble that only group them in a 63 10 53
group called sand - soil with gradation from 75 14 63
fine to coarse (Phung Van Lu and et al. 100 20 75
2002) (Table 1). But for Vietnam standard 28 106
7570:2006, it classifies specifically the 37
composition of sand, gravel - rubble (TCVN 50
7570:2006). 63
From Table 1, it is obvious to see that the 75
selected grain size for fineness module of
Determine the percentage of residual
granular soil in each national standard is totally
particle size per sieve (sieve analysis)
different from grain size diameter applied in according to formula (1):
geological engineering investigation. Therefore, mi
we will not utilize analysis data of soil grain ai  .100% (1)
m
size composition of sand, gravel and rubble in Where: mi - The passing particle weight of
the geological investigation. The authors grain size on a sieve with a diameter (i); m -
propose to choose the grain size according to The total weight of the sample (1000 gram).
the experimental result of grain size analysis of Determine the percentage of accumulated
granular soil in the geological investigation to granular soil on sieve Ai follows (2):
calculate the fineness module of sand, gravel As for sand: Ai=a2 + a1 + ……..+ ai (2a)
and rubble as below steps (Ho Vuong Binh As for gravel and rubble:
1992 and et al). Ai=a100 + a70 + a40……..+ ai (2b)

286
 Accumulated percentage Ai is the total Author’s method (TCVN 4198:2014)
percentage of retention granular soil on the A2  A1  A0,5  A0,32  A0,1
Ms (2014) 
sieve size i and the other larger sieves than it. 100
Determine Fineness module (Ms) of 11,36  20, 7  58, 7  82,34  88,96

granular soil follow the formula (3): 100
 2, 62
As for sand:
A2  A1  ..  A0,10 Difference of fineness module of sand ∆Ms
Ms  (3a)
100 (%) determined follow 2 methods
As for gravel and rubble: Ms 
Ms(2014)  Ms(2006)
100%
A100  A70  A40  ..  A5 Ms(2006)
Ms  (3b) 2, 62  2, 41
100  100%  8, 71%
2, 41
Where: A2, A1,… A0,10 - The percentage of
retained sand on each sieve 2, 1, 0.5, 0.25 and Given as calculated results, it illustrates
0.1 in respectively; A100, A70, A40..... A5 - the that Ms determined under standard TCVN
accumulated percentage of gravel and rubble 7570:2006 is lower than Ms of standard
on sieves with size 100, 70, 40, 20, 10, 5 mm TCVN 4198:2014 (0.21%) with 8.71%
respectively. difference. This is because TCVN 7570:2006
applied too large sieve size that leads to the
3. RESULTS AND DISCUSSION unreasonable method in determining retention
percentage of sand in each sieve get a lower
3.1 Fineness module of sand follow TCVN
7570:2006 and new proposed method of value. Meanwhile, TCVN 4198:2014 by
authors (TCVN 4798:2014) through using a smaller size of sieve that
brings more favorable than the percentage of
As given in table 1 shows that due to non-
remaining sand in each sieve as higher as
different grain size limit between 2 methods,
more accurate for Ms value.
the authors only conduct comparing fineness
module of sand follows 2 mentioned 3.2 Fineness modulus of granular soil Thua
methods. To determine and compare the Thien Hue follow proposed method
fineness module of sand by calculating by 2
The granular soil formations used for natural
conceptual methods, we sample and analyses
engineering material in Thua Thien Hue
the grain composition of Huong river (aQ23)
coastal province includes sand, gravel,
(Table 2). Continuously, the data for grain
rubbles, rock mass and silt or clay. That all
size analysis are quantified follow 2 different
key compositions of Quaternary sediment
kinds of sieve, first step we ignore the
with Holocene - Pleistocene period; Alluvial
percent of grain size ≥ 5 mm. Then the
(a), alluvial - pluvial (ap), marine (m) and
fineness module of sand is identified follow
marine - eolian (mv) origins: a,apQ23, mvQ23,
each method as below:
mQ22, a,apQ21, a,apQ13(2), mQ13(2), a,apQ13(1),
The standard TCVN 7570:2006
a,apQ12. The distributed area of these
A2,5  A1, 25  A0, 63  A0,315  A0,14
Ms (2006)  formations is presented in cross-section on
100
7,35  14, 77  53,53  78, 69  86, 78 the map and Quaternary geological cross

100 section through routes in figure 1 and figure
 2, 41 2. To be specific:

287
Table 2. Results of grain size analysis of sand in Huong river (aQ23 )
Grain percentage Grain percentage
Proposed sieve Size of Sieve follow
Order (%) (%) follow
(TCVN4198:2014) TCVN7570:2006
(TCVN4198:2014) TCVN7570:2006
1 <0,10 8,89 <0,14 11,02
2 0,10 6,62 0,14 8,03
3 0,25 23,64 0,32 25,16
4 0,50 38 0,63 39,06
5 1 9,34 1,25 7,22
6 2 11,36 2,50 7,25
7 5 2,15 5 2,26
8 10 - 10 -
9 20 - 20 -
10 40 - 40 -
11 70 - 70 -
12 - - 100 -

Marine - eolian (mvQ23): located on Early Holocene marine sediment (mQ22):

sandbars with sand dunes get from 5 to 10 lithological composition proves quite uniform
meters to 30 - 33 meters high along study area as well gradated sand with a range from coarse
coastline (from Vinh Thai to Lang Co sub- sand (rarely) to typical gray-white silty sand.
district and usually being covered upon The gray-white sand distributed widely in
mentioned marine sand sediment (mQ22). coastal plain of the central part of Vietnam
Joined with this formation popularly is grey- especially in Quang Binh, Quang Tri, Thua
yellow medium-fine sand. In addition, from Thien Hue, Da Nang, Quang Nam and from the
grain size analysis of more than 100 grey- grain size analysed results of more than 600
yellow sand samples with different sandy grain samples determined by sieve analysis as
size as follow: Gravel = 1%, sand = 90%, silt = following: Gravel = 2%, sand = 87%, silt =
8%, clay = 1%. Grain density: Md = 0.22mm, 9%, clay = 2%. Grain density: Md = 0.23mm,
So = 1.49, Sk = 1.02. So = 1.64, Sk = 1.06.
Aluvia - pluvia sediments (a, apQ23): located
Aluvia - pluvia sediments of the later
on mainly at the river bottom, streams in
Holocene period (a, apQ21): this formation
Truong Son mountain especially rivers as Bo,
seems covered on by sediments with having
Huong, Truoi (Figure 3b). The key lithological
age Q22 and Q23 at 38.5 meters deep (at
composition is like boulders, cobbles, sand,
borehole LK306) to 10,2 meters (at borehole
rare silt, clay which structure the river delta,
CCT1-3) from the surface. Sand, coubles,
underwater sandbar, floating bar. Boulders and
sand in the river are usually quartz, silica rock, boulders in the Holocene period are created
less quartzit, granites, silty sand. From grain from quartz, silica, granite, sandy quartzite,
size analysis of nearly 30 samples of coarse basalt (igneous rock). Percentage of grain size
grain soil shows that the percentage of particle distribution includes: coubles - boulders =
size as coubles = 9%, gravel = 22%, sand = 10%, gravel = 19%, sand = 67%, silt = 3%,
62%, silty = 5%, clay = 2%. Grain density: Md clay = 1%. Grain density: Md = 1.89mm, So =
= 1.15 mm, So = 2.41, Sk = 1.34. 3.07, Sk = 1.28.

288
 Marine sediment mQ13(2): conceptual K74) to 30.5 meters at borehole (CCT5) from
sediment exposes various locations and this the surface. This sediment has lithological
sediment are also conducted study on the composition include sand, clay, boulders,
mapping of the geological map with average gravels (at boreholes CCT5, CSH1, LK309,
scale and large scale; with sediment entitle is LKHU8, LK314…). Through 140 samples of
Da Nang formation (Cat Nguyen Hung and et grain size analysis as boulders = 30%, gravel –
al. 1996; Nguyen Van Trang and et al. 1996) or sand = 68%, silt - clay = 2%. Grain density: Md
Phu Xuan formation (Do Van Long and et at. = 2.29 mm, So = 2.75, Sk = 1.61.
2000; Pham Huy Thong and et at. 1997). The sediments which have origin from
Where, Author Nguyen Xuan Duong classified middle Pleistocene alluvia - pluvial (a, apQ12)
conceptual formations into marine - alluvia distribute at several deep boreholes at the
multiple origins and belong to Da Nang depth 100.8 (LK309), 80.1 (LK3QT) to 42
formation (amQ13(2) đn) (Figure 3a). As for meters deep (LK429) from the surface. The
after 180 yellow sandy soil sample analysis, we
main lithological composition includes
have percanetage particle size composition as
boulders, gravels = 32%, sand = 65%, silt -
gravel = 2%, sand = 76%, silty = 19%, clay =
clay = 3%. Grain density: Md = 1.18mm,
3%. Grain density: Md = 0.02mm, So = 1.75, Sk
So = 2.13, Sk = 1.89.
= 1.08.
Based on the experiment data of 1.633
Aluvia - pluvia sediments (a, apQ13(2)) is
samples, we eliminated inaccuracy data from
covered from 50 meters deep (LK432) to 6
meters (CSH1) from the surface and only 159 samples (by 9.74%) and determine total
expose under the shapes of narrow limbs at values of grain composition follow laboratory
the depth 10 - 12 meters of the western site of data of 1.474 samples (by 90.26%).
the coastal plain. The grain size analyzed From calculated results on Table 3 show that
results of more than 150 samples indicate that the fineness modulus varies from 1.21 to 5.97
boulder = 17%, gravel = 20%, sand = 60%, which meet the engineering requirements of
silt - clay = 3%. Grain density: Md = 2.46mm, fine aggregates for concrete and mortal
So = 2.73, Sk = 1.72. material conducting follow Vietnam
Upper Pleistocene alluvial - pluvial engineering standard (Phung Van Lu and et al.
sediment at the down section (a, apQ13(1)) 2002; TCVN 7570:2006; TCVN 7570:2006;
distributes at 71 meters deep at borehole (LK1- TCXD 127:1985).

Figure 1a. Geological map of Thua Thien Hue coastal plain (Section 1)

289
Figure 1b. Geological map of Thua Thien Hue coastal plain (Section 2)

Figure 2a. Geological cross-section of route IV - VI’ in Phong Dien

Figure 2b. Geological cross-section of route V - V’ in Hue

Figure 3a. Bai Tram sand mine (Da Nang Figure 3b. Alluvial sand yard in Loc An commune,
formation, mQ13dn) in Loc Tien Phu Loc district
commune, Phu Loc district

290
Table 3. Fineness modulus of granular soil follow grain size analysis results of engineering
geological investigation in Thua Thien Hue coastal plain
Percentage of grain size passing on sieve, %

Geological age
Origin and
Fineness

Samples

40-20mm

20-10mm

1-0,5mm
Modulus,

10-5mm

<0,1mm
0,5-0,25

0,25-0,1
Soil type

>40mm

5-2mm

2-1mm

mm

mm
MS
a,apQ23

Sand contains
57 7,15 4,71 2,32 3,03 8,14 15,08 17,69 23,23 11,43 7,22 2.12
gravels
Medium grey
22 - - - - 1,19 4,63 19,31 34,5 22,57 17,8 1.74
dense sand
mvQ23

Small sand with

91 - - - - 0,48 2,64 11,86 26,79 36,31 21,92 1.38
medium dense
Coarse grey sand
13 - - 0,79 4,54 8,23 11,94 34,01 22,10 10,59 7,80 2,46 with medium
dense
Grey medium
mQ22

252 - - - - 2,99 7,29 19,16 34,73 22,69 13,14 1,94 sand with
medium dense
355 - - - - 0,50 4,23 12,27 26,75 37,68 18,57 1,47 Fine foam sand
41 - - - - 0,41 2,98 6,86 25,97 34,75 29,03 1,21 Foam silty sand
Gravels contain
9 1,52 38,42 18,24 6,38 7,51 9,15 6,74 5,51 3,97 2,56 1,09
dense sand
a,apQ21

Coarse sand,
75 - 0,41 1,14 2,61 8,35 11,04 28,07 26,37 14,87 7,14 2,38
medium dense
3 - - - - 0,62 2,15 4,57 26,29 50,16 16,21 1,28 Fine foam sand
Gravels contain
15 17,32 26,47 19,9 13,31 9,87 10,63 2,50 - - - 0,99
dense sand
Q13(2)
a,ap

Sand contain
157 0,05 1,79 5,56 8,74 14,26 18,01 20,93 15,73 10,15 4,78 2,48
dense gravels
Yellow medium
116 - - - - 2,10 8,24 16,49 31,01 25,63 16,53 1,81
dense sand
Yellow fine
mQ13(2)

57 - - - - 0,65 4,63 11,92 32,85 28,83 21,13 1,52

dense sand
Yellow silty sand
30 - - - - 0,22 2,61 10,27 26,46 30,65 2,79 2,56 with medium
dense
Gravels contain
14 2,19 30,33 19,73 9,75 9,12 10,39 8,66 3,5 3,66 2,67 1,24 sand with high
dense
a,apQ13(1)

Sand contains
83 - 1,89 4,25 7,84 11,33 16,26 23,89 20,21 10,53 3,8 2,44 gravel with high
dense
Medium sand
41 - 1,73 3,17 4,88 4,96 9,61 18,86 34,32 19,79 2,68 2,08
high dense
Sand contain
7 - 5,39 10,12 11,92 15,64 21,46 15,43 10,35 7,68 2,01 2,39 gravels, high
a,apQ12

dense
Coarse sand, high
36 - - 3,57 4,4 8,76 16,37 26,32 18,39 15,97 6,22 2,41
dense

291
 project of Hue urban for engineering
4. CONCLUSION
geological mapping of Hue city, scale
As for above scientific contents, we emphasis 1:1200000; Engineering geological report of
various conclusions as below: Cacut bridge, head office of Chan May - Lang
Co bay management, Kenh bridge, Long
- The fineness modulus of granular soil is
bridge, Cua Viet bridge, Cho Dinh bridge, Pho
not only identified follow current Vietnam Trach bridge...and other constructions.
engineering standard TCVN 7570:2006, also TCVN 7570:2006 - Aggregates for concrete and
calculated follow our proposal method (TCVN mortar - Technical requirements.
4198:2014) with errors less than 10%. Due to TCVN 4198:2014 for engineering soil - The
using large size of sieve, the Ms value methods of particle content determination in
determined to follow TCVN 7570:2006 gets the laboratory.
inaccuracy and lower than one follow TCVN ASTM C136 Test Method for Sieve Analysis of
4198:2014. Fine and Coarse Aggregates.
- Granular sediments are suitable for S. K. Duggal 2015. Building Materials. PHI
utilising as natural engineering material in Learning Private Limited, Delhi, ISBN: 978-
research area include several multi-origins as: 81-203-5091-5.
a,apQ23, mvQ23, mQ22, a,apQ21, a,apQ13(2),
mQ13(2), a,apQ13(1), a,apQ12. Mostly, sand have
particle and impurities content and fineness
modulus (Ms = 0,99-2,56) meet the technical
demands of fine aggregate for engineering
concrete and mortal material.
- It is so necessary to continue research,
testing of this method on determining fineness
module of granular soil in comparison to US
standard (ASTM), British standard (BS),
Japanese standard (JS).

5. REFERENCES
Cat Nguyen Hung and et al. 1996. Report of
geological mapping and mining exploration
with scale 1:50000 in Da Nang - Hoi An city.
Storage at Geological and Mineral
Department of Vietnam, Hanoi.
Do Van Long and et at. 2000. The characteristics
of Quaternary sediments in Binh Tri Thien
plain”, Geological map(98), pages.53-66.
Phung Van Lu, Pham Duy Huu, Phan Khac Tri,
2002. Construction material, Construction
Publishing House.
Pham Huy Thong and et al. 1997. Geological and
Mineral report in Hue city with scale 1:50000.
Storage at Geological and Mineral
Department of Vietnam, Hanoi.
Nguyen Van Trang and et al. 1996. Geology and
Mineral in Huong Hoa - Hue - Da Nang, scale
1:200000. Geological and Mineral
Department of Vietnam, Hanoi.
Ho Vuong Binh 1992 and et al. Engineering
geological documents in Thua Thien Hue: The

292
The 3rd Int. Conf. on Transport Infrastructure & Sustainable Development (TISDIC 2019). Construction Publishing House. ISBN: 978-604-82-2893-4.

Effect of sand on the engineering performance of cement

asphalt mortar for fouled ballast stabilization

Le Tri Ho Minh 1*, Dae-Wook PARK1, Jung-Woo SEO1, Koo-Sam JI2, Min-ho KANG
1
Dept. of Civil Engineering, Kunsan National University, 558 Daehak Ro, Kunsan, Jeonbuk, Republic of
Korea.
2
Samhyun company, 36-12, Gamillam-ro, Hanam-si, Gyeonggi-do, Republic of Korea.

Corresponding author, E-mail: [email protected]

ABSTRACT: In the stabilization strategies for fouled railway ballast, cement asphalt mortar (CAM) has
been developed as a promising method which combines the advantageous from both cement and asphalt
material with fast construction time. In this research, the effect of sand on the engineering properties of
CAM are evaluated. To cope with this objective, flowability and mixing stability tests are conducted to
identify the workability of materials at fresh stage. Then, 2hrs and 28 days unconfined compressive
strength test are employed to determine the strength development of CAM mixture. The microstructure of
hardened samples is analyzed by applying scanning electron microscope test (SEM). The test results
indicated that adding sand to CAM mixture not only enhances the rheology of fresh mixture but also
contributes to the strong structure bonding between cement hydration product and asphalt membrane.
Keywords: Cement asphalt mortar; fouled ballast stabilization; sand; mixing stability; compressive strength.

1. INTRODUCTION Especially, the use of cement asphalt mortar

(CAM) as an interlayer between track slab and
In recent years, the ballasted track structure is concrete roadbed has been applied in many
widely used in many countries due to its low developed countries such as Japan, and
construction cost and high performance in Germany (D’Angelo et al. 2016). Ouyang et at.
rutting resistance (Selig et al. 1994). However, (2014) states that the combined merits from
reports concern about the great deterioration of both cement and asphalt in this technique will
the ballast structure due to the continuous loads provide the riding smoothness for high-speed
from high-speed trains and environmental train and damping ability for stress relaxation in
impacts (D’Angelo et al. 2016). This problem the railway structure. However, these methods
will lead to an increase in maintenance and must be applied at the initial construction stage
rehabilitation costs in long-time service life with high-cost requirement. This drawback
(Pires and Dumont 2015). Hence, many hinders these solutions to be applied in large
strategies have been proposed to resolve this scale or in maintenance projects. Hence, the
issue. Brown et al. (2007) present the use of need to find a feasible stabilizing solution is of
geogrid in reducing the vertical and lateral importance (D’Angelo et al. 2016).
deformation of ballast structure. Research on As found in D’ Angelo research (D’Angelo
polyurethane coating ballast from Dersch
et al. 2016), the author states that asphalt
(2010) shows promising results in settlement
resistance with simple application. The stress- emulsion can help stabilize ballast particles by
bearing ability of fouled ballast structure can decreasing the permanent deformation of
also be enhanced by using the geo-pavement fouled ballast structure. Therefore, based on
technique from Kennedy (2009) study. this concept, this study aims to utilize CAM in

293
stabilizing fouled ballast. The contribution of types of emulsifier are used in this research
CAM as a bonding material between fouled including cationic, anionic, and nonionic
ballast particles is expected to promote strong emulsifier. Then, from each type of
rutting resistance and enhance the elasticity emulsifier, five different sand ratios are
behavior for the railway structure. In this applied to develop CAM mixtures: 0%, 25%,
method, high flowability CAM mixture is 50%, 75%, 100% by weight of cement. The
poured directly on top of the fouled ballast. rheology characteristic of varied mix
The self-leveling of material allows it to flow conditions is figured out by conducting the
through the whole ballast structure. With the mixing stability test and flowability,
fast setting time, the maintenance project can meanwhiles, the strength development of
be done at short time for traffic reopening CAM mixture are investigated by performing
which adapts the strict requirement from the unconfined compressive strength (UCS)
railway maintenance projects. The components test at an early age (2hrs) and 28 days.
of CAM can also be easily provided by local Finally, the microstructure of different
company such as asphalt emulsion (AE), conditions is analyzed from the SEM test.
cement (C), water (W), and superplasticizer
(SP). Therefore, it is practical to apply this 2. MATERIALS AND TEST METHODS
technique as a sustainable solution for ballast
2.1 Materials
maintenance.
Prior works on the CAM stabilized ballast The chemical composition and physical
shows that high AE/C ratio leads to the properties of Type II Portland cement (C)
increase in ductility behavior of CAM used in this study are shown in Table 1, and
mixture which improves the stress dissipation Table 2, respectively. As stated before, the
for the whole railway structure (Ouyang et al. asphalt emulsion (AE) modified with three
2014, Ouyang et al. 2018). The damping types of emulsifier are used in this research
ability from viscoelastic characteristic of including cationic (+), anionic (-), and
asphalt membrane can also be promoted by nonionic. The properties of asphalt emulsion
using the technique. However, high AE/C provide by Korean chemical company are
ratio may lead to a remarkable drop in presented in Table 3. The construction sand
strength. Hence, the proper use of asphalt (S) is employed in this test with the specific
emulsion should be greatly considered gravity of 2.65 g/cm3. The gradation
(Ouyang et al. 2014, Ouyang et al. 2018). distribution of sand is presented in Fig. 1.
Besides, the early strength gain in mixture Based on the suggestion from prior works
can be enhanced by using quick hardening (Le et al. 2019), the proper use of quick-
admixture with small rheology effect (Le et hardening admixture (QA) will increase the
al. 2019). However, there are lack of research strength gain at an early age. Hence, this
focus on the effect of sand or emulsifier type admixture is included in the mix design with
on the engineering properties of CAM a ratio of 14% by weight of cement. It is well
mixture. It is skeptical that these components known that adding superplasticizer (SP) will
may impose a strong influence on the improve the workability of the cement
characteristic of CAM to some extent. mortar and lower the use of mixing water. In
Therefore, the main aim of this study is to this study, the SP will be used at 2% by
evaluate the effect of different natural sand weight of water to improve the rheology of
ratio and emulsifier types on the mechanical CAM mixture. In the production of AE, the
properties of CAM at both fresh and emulsifier and admixture are added to
hardened stage. To achieve this target, three prolong the stability of asphalt droplets in

294
water without coalescence phenomenon. Table 1. Chemical Composition of Cement by
These components may generate the Percent.
excessive formation of air bubble which may SiO2 Al2O3 Fe2O3 CaO MgO SO3 Na2O3 f-Cao
lead to the deterioration in strength of
21.56 4.15 2.64 61.18 2.35 3.01 0.65 0.81
cement asphalt mortar. Therefore, defoaming
agent (D) is added to the asphalt emulsion at
Table 2. Physical Properties of Cement.
a ratio of 0.1% by weight of cement. Tap
water (W) is employed to mix all Fineness Specific Initial Final 28 day
components. The mix design used in this (residue surface setting setting compressive
study is exhibited in Table 4. on (m2/kg) (min) (min) strength
80 μm (MPa)
sieve)
(%)
0.5 330 145 205 55

Table 3. Properties of Anionic Asphalt

Emulsion
Density Storage Residue Residue Penetration Solid
(g/cm3) stability after on depth at content
1 day distillation 1.18 25° (0.1 (%)
(%) (%) mm mm)
sieve
(%)

Figure 1. Ballast and sand particle size distribution. 1.02 0.3 50 0.01 70 50

Table 4. Mixture proportion of CAM without sand (based on cement weight).

Mix (C) (QA) (AE) (S) (W) (SP) (D)
A0 100% 14% 100% (A) 0% 30% 2% 0.1%
A25 100% 14% 100% (A) 25% 30% 2% 0.1%
A50 100% 14% 100% (A) 50% 30% 2% 0.1%
A75 100% 14% 100% (A) 75% 30% 2% 0.1%
A100 100% 14% 100% (A) 100% 30% 2% 0.1%
C0 100% 14% 100% (C) 0% 30% 2% 0.1%
C25 100% 14% 100% (C) 25% 30% 2% 0.1%
C50 100% 14% 100% (C) 50% 30% 2% 0.1%
C75 100% 14% 100% (C) 75% 30% 2% 0.1%
C100 100% 14% 100% (C) 100% 30% 2% 0.1%
N0 100% 14% 100% (N) 0% 30% 2% 0.1%
N25 100% 14% 100% (N) 25% 30% 2% 0.1%
N50 100% 14% 100% (N) 50% 30% 2% 0.1%
N75 100% 14% 100% (N) 75% 30% 2% 0.1%
N100 100% 14% 100% (N) 100% 30% 2% 0.1%
*A: anionic asphalt emulsion; *C: cationic asphalt emulsion; *N: nonionic asphalt emulsion.

295
2.2 Mixing method Where,
 Pr: the mixing stability status (%),
Based on preliminary research (Le et al. 2019),
it was found that the mixing method plays an  m: weight of residue (oven dried)
important role in the rheology of cement  m1: weight of asphalt emulsion (50g),
asphalt mortar. The mixing effectiveness  m2: weight of cement (50g)
depends mainly on the initial mixing water
with the dry components. The prior works a b
(Ouyang et al. 2014, Le et al. 2019) suggest
that dry components (C, QT, S, D) should be
mixed with water first for 2 mins at a rotation
rate of 120rpm, then, asphalt emulsion is added
to this combination and the whole mixture is
mixed for an additional 1 min at a rate of
60rpm. The aim of this mixing method is to c d
ensure the stable condition of asphalt droplet in
asphalt emulsion. The water will help lower the
adsorption energy of cement particle to the N50
asphalt droplets in the asphalt emulsion and
thereby, prolong the demulsification process of
asphalt emulsion. If this process is triggered at Figure 2. (a)The mixing stability test, (b)the
an early time, the coalescence of asphalt flowability test, (c)sample after UCS
droplet can happen abruptly in the fresh stage test, (d)the SEM test
and lead to the formation of big and
inhomogeneous CAM particle. 2.4 Flowability test
2.3 Mixing stability test One of the main advantages of CAM which
drawn attention from recent research is the
The mixing stability is the main factor high flowability behavior. This ability will
affecting the rheology of asphalt emulsion. allow CAM to be pumped in the actual
Preliminary research (Ouyang et al. 2014, Le et construction site with fast and simple support.
al. 2019) suggests that mixture with good In this research, the CAM mixture is designed
mixing stability will receive homogenous to thoroughly flow through the ballast layer
particle size distribution and ideal flowability. and create a good connection between ballast
The demulsification rate of asphalt emulsion particles. This flowability test (KS F 2432,
can also be determined in this test since the 2004) includes the use of 400ml flow cone
residue content will represent the rate of with both close ends, the aluminum plate, and a
stability breakage of asphalt emulsion. Regards recording watch. After the mixing process,
to the test method (KS M 2203, 2018), 50g of 400ml of fresh material from each condition is
cement, 50g of asphalt emulsion, and 150g of poured to the flow cone lifted 150mm from the
deionized water are first mixed for surface of the aluminum plate. The time for the
approximately 3 mins at a mixing rate of whole paste to flow out of the steel cone is
120rpm. Then, the paste is mixed at a shearing considered as the flow time (Fig 2b). three
rate of 60rpm for additional 1 min and it was replicates are used to calculate the average
filtered through an 1.18 mm sieve size (Fig. flow time of each condition. In addition, to
2a). The remained solid residue on the sieve is define the optimum CAM volume for ballast
then oven dried and weight. Based on equation covering, the trial tests are conducted on the
1, the mixture stability (%) can be calculated: mixture with optimum flowability. The test
m includes the crystal-clear plastic mold with a
Pr  100 (1)
m1  m2 size of 150mm in diameter and 500mm in

296
height. The particle size distribution of ballast SEM S-4700 under high vacuum condition
layer is shown in Fig. 1. This aggregate is (Fig. 2d). As stated before, three replicates of
added to the mold and compacted by using each condition will be scanned, and the proper
steel rod and plastic hammer. After thorough image will be selected to analyze.
mixing, the fresh CAM is poured from the top
of the ballast layer. The optimum CAM 3. TEST RESULTS AND DISCUSSIONS
volume investigation is mainly conducted by
visual inspection which has a requirement of 3.1 The mixing stability test results
ballast covered by CAM from top to bottom Figure 3 illustrates the mixing stability test
without stacking phenomenon. results of CAM mixture with varied emulsifier
2.5 Unconfined compressive strength test types and different sand ratios. Regards to the
effect of emulsifier type on the establishment of
Due to the very short construction time from solid residue on the 1.18mm sieve size, it was
railway maintenance project, the strength at an noticed that cationic CAM mixture suffered
early age of CAM mixture is designed to adapt from remarkable residue content formation
this requirement. Hence, in this study, the UCS compared to the remain emulsifier CAM types.
of samples are not only measured at 28 days The residue percentage retained on the filter
strength but also at early age of around 2hrs. ranges from 14% to 9% when the sand ratios are
The UCS test is conducted in accordance with increased from 0% to 100% by weight of
the KS F 2351 (2010). When the mixture is
cement. This may be attributed to the cationic
thoroughly mixed, the fresh paste is cast into
emulsifier have a opposite charge type with the
laminate molds with a size of 5cm in diameter
cement (anionic charge type). It leads to the fast
and 10cm in height. Then, the specimens are
connection between the two components and
then transferred to the curing room with 98%
relative humidity and 23oC (KS F 2351 2010). accelerates the demulsification process of
After 2hrs, 3 samples from each condition are asphalt emulsion. Therefore, more asphalt
stripped out from the pre-cut molds and droplets coalesce on the cement particle surface
applied the UCS test (Fig. 2c). The remained in cationic CAM mixture at an early age which
samples are cured for 1 day more before leads to the formation of excessive residue on
unmolding process. They are left to be cured in the filtering apparatus. Meanwhiles, anionic and
the same environment until the testing day (28 nonionic CAM mixture show ideal mixing
days). The average results from 3 replicates stability results with less sign of big particle
will represent 1 condition. The universal formation and the residue content varying from
testing machine is employed in this test with 4% to 0.5%. It can be explained by the same
the loading rates of 0.5mm/min and charge type from anionic AE and neutral charge
0.1mm/min for 2hrs and 28 days strength of nonionic AE which lead to the stability
samples, respectively. Based on the suggestion between cement particle and asphalt droplets.
from prior research (Le et al. 2019), this The asphalt emulsion can be dispersed freely on
loading rate will ensure the accurate test results the cement particle at an early age without
generation. stability breakage and thereby, improving the
mixing stability.
2.6 SEM analysis
In addition to the impact of sand to this
The SEM analysis test is conducted with the rheology property of CAM mixtures, the test
aim of investigating the microstructure results reveal that adding sand will noticeably
organization of CAM mixture. It is expected enhance the mixing stability of the fresh
that CAM mixture with sand will exhibit a mixture. The effectiveness is prominent when
dense structure which promotes a strong the S/C ratios are increased to higher than
connection with cement hydration product. The 50%. As shown in Fig 3, the mixing stability of
SEM test is conducted by using Hitachi FE- CAM mixtures are recorded in anionic and

297
nonionic condition with the residue content of shown in mixing stability test. The
about 0.5%. Based on the laboratory test flowability of mixtures with sand is much
suggestion, this value is very low which may better than that without sand, especially
not impose negative impact to the mixing mixture with A/C ratio of 50%. The flow
stability of CAM mixture. In these mixtures, time in cationic mixture reduces from 30 to
sand may play a role as a neutral interlayer 22 secs when S/C ratio is increased from 0%
which protects the asphalt droplets from the
to 50%, then, it goes from 22 to 25 secs when
adsorption energy of cement particles and
S/C is designed at 100%. It should be noted
thereby, sustaining the stability of AE.
that adding too much sand may lead to the
However, when the S/C ratios are increased to
75% or 100%, the efficiency provided by sand high friction between CAM particle and
is not obvious. thereby, prolonging the flow time.
Based on the trial test for figuring out the
optimum CAM volume for ballast covering,
when the CAM content are below 5%, the
bonding material is unable to cover the whole
ballast aggregate. The visual inspection
suggests that the use of CAM of 7-10% by
volume of air void in ballast help provide the
proper flow of fresh CAM throughout the
whole ballast structure (Fig 5). It can flow
from top to the bottom of the crystal mold
without excessive stacking phenomenon at the
Figure 3. Mixing stability test result
bottom. Hence, this optimum content is
3.2 The flowability test results promoted to apply in further research.

The flowability test results of CAM mixtures

are presented in Fig 4. The test result
confirmed the findings from the mixing
stability test with poorest flow results found
in cationic mixture. It was investigated that
cationic asphalt emulsion may lead to the
formation of bigger CAM particles in the
mixture and lead to the agglomeration
phenomenon at the nozzle of the flow cone.
This issue will increase the flow time of fresh Figure 4. Flowability test result
mixtures substantially. However, the problem
can be resolved by using nonionic or anionic
asphalt emulsion. These conditions show
similar flow time with the average value of
around 18 secs. Due to the same and neutral
charge types of emulsifier to the charge type
of cement, the stability of asphalt emulsion
can be protected during the mixing stage. The
flowability results of mixtures with different Figure 5. Optimum CAM volume for stabilizing
A/C ratios also follow the same trend as ballast

298
3.3 The unconfined compressive strength test
results
Figure 6 represents the UCS test results of
CAM mixture modified with different
emulsifier types and S/C ratios. Although
mixtures with different emulsifier types show
different rheology results, they share the
relatively same strength gain at an early age.
For examples, anionic mixtures with S/C
ratio of 0.5 obtain the UCS value of about
0.35 MPa.
However, after 28 days the UCS of (a)
mixtures with nonionic emulsifier achieves
the highest strength, followed by anionic
mixtures and cationic mixtures. Due to the
inhomogeneous and poor particle size
distribution in mixtures with cationic asphalt
emulsion, cationic mixture may obtain poor
structure development and suffer from the
strength loss.
Regards to the effect of the strength
development of CAM mixture, it was
indicated from the test results that adding
(b)
sand not only provide good rheology but also
contribute to the strength gain of CAM Figure 6. UCS test results of mixture after 2hrs (a)
and 28 days (b).
mixtures. The test results reveal that the
higher the S/C ratio, the stronger the strength 3.4 The SEM analysis
develops in CAM mixture. It can be
The SEM results confirmed the findings
explained by the appropriate distribution of
from the UCS tests that mix A0 with cationic
CAM particle in the mixture with sand which
and 0% sand obtain very poor combination
leads to the strong bonding between asphalt
between cement hydration product and asphalt
membrane, cement hydration product, and
membrane. Meanwhile, mix N50 with nonionic
sand. Also, the test results agree with the emulsion and 50% sand receive a homogenous
findings from the mixing stability test that distribution of asphalt layer among cement
the use of 50% sand by weight of cement particle with obvious ettringite structure
may be the best cost-effectiveness solution development (Fig. 7). The dense structure
since adding more sand to a S/C ratio of 75% shows the good merging between the asphalt
or 100% only lead to small strength membrane and cement hydration product. This
improvement. Therefore, based on the above strong bonding may lead to the good stress
findings, mixture with nonionic asphalt bearing ability of CAM structure. Interestingly,
emulsion and S/C ratio of 50% is promoted to the SEM results also indicate that the ettringite
be used in further research to achieve can develop through the thick asphalt
promising results. membrane covering (N50).

299
 remarkably by using anionic or nonionic
asphalt emulsion. This may be due to the
similar or neutral charge type of to the charge
type of cement. The charging concept will
sustain the stability of asphalt droplet in the
fresh mixture and thereby, prolonging the
demulsification process of asphalt emulsion.
As found in mixture with cationic emulsifier,
the fast breakage of asphalt droplet stability
will lead to the early coalescence of asphalt
membrane on cement particle which causes
(A0) high volume formation of big CAM particle
at an early age.
The differences in strength development
between three emulsifier types is not obvious
at an early age. However, after 28 days
nonionic CAM mixtures outperformed the
remain conditions since this mixture obtain the
good connection between asphalt droplets and
cement particle.
The incorporation of sand in CAM mixture
not only improve the mixing stability of
mixture at fresh stage but also contribute to
the strength attainment. The rheology tests
results reveal that sand may act as a neutral
(N50) barrier that protects the asphalt droplets from
Figure 7. Mixing stability test results of mixture the adsorption energy of cement particle.
A0 (0% sand + cationic AE) and Without sand support, the demulsification of
mixture N50 (50% sand + nonionic AE) AE can be accelerated abruptly during the
mixing period.
4. CONCLUSIONS The UCS test results indicate that adding
sand will increase the strength gain of all
The progressive impacts from heavy train conditions. The strength of CAM mixture
loads and nearby environment lead to the increases noticeably when S/C ratio is
degradation of the ballast layer with fouled increased from 0% to 50%. However, the
ballast formation. This problem will lead to effectiveness is neglected when this ratio is
the permanent settlement of the whole designed up to 100%. Hence, among five sand
structure and require substantial maintenance conditions, the S/C ratio of 50% is suggested to
and rehabilitation projects. Hence, the main be applied in future CAM study to receive the
aim of this study is to systematically develop optimum mechanical performance.
a new fouled ballast stabilizing method using
CAM. There are many factors affects the ACKNOWLEDGMENT
mechanical properties of CAM, this study
mainly studies the effect of emulsifier and This research was supported by a grant from
sand on the characteristics of CAM before the Infrastructure and Transportation
applying to the actual ballast stabilizing Technology Promotion Research Program
project. The following results can be drawn funded by the Ministry of Land, Infrastructure,
from this research: The mixing stability and and Transport of Korean Government
flowability of CAM mixture can be enhanced (18RDRP-B076564-05)

300
5. REFERENCES Mortar. Transet. MATEC Web of Conferences
271, 03002.
Brown, S. F., Kwan J. & Thom N. H. 2007. Ouyang, J., Tan, Y., & Li, H. 2014. Factors
Identifying the key parameters that influence influencing rheological properties of fresh
geogrid reinforcement of railway ballast, cement asphalt emulsion paste. Construction
Geotext. Geomembr. 25 (6) 326-335. and Building Materials 68, 611-617.
Dersch, M. S., Tutumluer, E., Peeler, C. T. & Bower, Ouyang, J., Hu, L., Li, H. & Han, B. 2018. Effect
D. K. 2010. Polyurethane coating of railroad of cement on the demulsifying behavior of
ballast aggregate for improved performance, over-stabilized asphalt emulsion during
2010 Joint Rail Conference, 1, 337-342. mixing. Construction and Building Materials
D’Angelo. G., Thom, N. & Presti, D. L. 2016. 177, 252-260.
Bitumen stabilized ballast: A potential Ouyang, J., Zhao, J., & Tan, Y. 2018. Modeling
solution for railway track-bed. Construction Mechanical Properties of Cement Asphalt
and Building Materials 125 118-126. Emulsion Mortar with Different Asphalt to
Kennedy, J. H., Woodward, P. K., Medero, G.M. Cement Ratios and Temperatures. Journal of
& McKinney, J. 2009. Full-scale cyclic Materials in Civil Engineering 30.
geopavement & railway accelerated fatigue Pires, J. & Dumont, A.G. Railway ballast
testing, International Railway Engineering degradation, 15th Swiss Transport Research
Conference. Conference, 2015.
Le, T. H. M., Park, D. W. & Seo, J. W. 2019. Selig E. T., Waters J. M. Track. Geotechnology
Evaluation on the mechanical properties of and Substructure Management, T. Telford,
cement asphalt mortar with quick hardening 1994.
admixture for railway maintenance. Construction KS F 2351. 2010. Testing Method for Compressive
and Building Materials, 206, 375-384. Strength of Asphalt Mixture.
Le, T. H. M., Park, D. W. & Seo, J. W. 2019. KS F 2432. 2004. Testing Method for Consistency
Effect of Water and Emulsifier on the of Mortar Grouting.
Mechanical Properties of Cement Asphalt KS M 2203. 2018. Emulsified Asphalt.

301
The 3rd Int. Conf. on Transport Infrastructure & Sustainable Development (TISDIC 2019). Construction Publishing House. ISBN: 978-604-82-2893-4.

Change detection of road surface from multi-temporal

unmanned aerial vehicles (UAV) images using machine
learning

Truong Linh Nguyen1, Dongyeob Han2, Jungwon Huh2

1,2
Department of Marine and Civil Engineering, Chonnam National University, Yeosu, Korea

Corresponding author: Dongyeob Han

E-mail: [email protected]

ABSTRACT: In computer vision and image processing, change detection is one of the most commonly
encountered tasks. Using Unmanned Aerial Vehicles (UAV) with high-performance vision sensors for the
road surfaces survey has received considerable attention due to its reliability and safety. Due to the
combined effect of aging and deterioration of the road surface, some damage would always appear on the
road surface, the number of roads that need to be surveyed increases, and they require much cost.
Therefore, road surface inspection based on UAV with vision sensors is one of the promising strategies to
maintain the road. This paper proposes a method for detecting changes in road surface using a pair of
UAV images in a different time. A novel fully Convolutional Siamese Network is introduced to measure
changes by extracting the features and calculating the distance between them by Euclidean distance
function. The Contrastive Loss function will be used to enlarge the distance between the changed feature
pairs and to reduce the distance between the unchanged feature pairs. The experimental results using the
dataset show the effectiveness of our approach.
Keywords: change detection, Convolutional Siamese Network, machine learning, Unmanned Aerial
Vehicle (UAV), image processing.

1. INTRODUCTION unmanned aerial vehicle, and artificial neural

network, the surveying task has become easier.
The road is one of the most important
In Kim et al. (2018), a crack identification
infrastructures, play a crucial role in supporting
method was introduced to survey an aging
the development of an economy and human
concrete bridge. First, generate the point cloud-
society (Pan et al. 2018). The quality of roads
based background model of the bridge in a
will decrease during the use stage because of
the aging and deterioration of the road surface; flight. Then, inspection images from a high-
some damage would always appear on the road resolution camera attached on UAV were
surface. Potholes and cracks are the two most captured and stored to scan structural elements.
common categories of road surface damages. Finally, they conducted prerequisite, deep
Therefore, it is imperative to maintain roads in learning processing for both image
good condition. Road condition inspection is classification and localization; and crack size
the first work that the experts should do before estimation to quantify the cracks. Reference
road maintenance. Usually, the inspector needs markers were attached to the structures in
to go outside to survey, but it spends so much advance to quantify the cracks on captured
cost. Currently, with the support of computer, images. The convolutional neural networks

302
method was used to reduce the computational network (Bromley et al. 2018, Koch et al. 2015)
cost in crack detection tasks. Finally, the length to generate feature pairs. Then a simple
and thickness of the crack is calculated using predefined distance metric (Euclidean distance
various image processing techniques and function – l2) was used to measure the
validated through the field test. dissimilarity of the feature pairs. The contrastive
In Guo et al. (2018), author proposed novel loss function is applied to bring together
deep metric learning-based scene change unchanged pairs and separate change pairs.
detection, the method can measure changes
region using the learned implicit metric, which
casts the change detection task to an implicit
metric learning problem. Then, to overcome
noisy changes caused by large camera
viewpoint differences, a Threshold Contrastive
Loss (TCL) was developed, which presents a
significant improvement. Finally, authors
integrate the distance metric into the baseline
based on the Fully Convolution Neural
architecture, giving rise to better performance. Figure 1. Flow chart of Change detection task
In Pan et al. (2018), the authors proposed
one new application procedure that applied 2.1 Convolutional Siamese Metric Network
SVM (Support Vector Machine), ANN
(Artificial Neural Network), and RF (Random Siamese networks are neural networks
Forest) classifiers to detect the asphalt road containing two or more identical sub-network
surface cracks and potholes from the UAV components (Bromley et al. 2018). The network
multispectral pavement images. The has the same configuration with the same
contribution of this study is to investigate the parameters and weights. In conventional CNN,
feasibility of the proposed model and offer a there exist three types of layers: convolutional,
new tool for monitoring asphalt road pavement pooling, and fully connected layers.
condition to improve the efficiency of road
maintenance practice.
In this paper, a novel change detection
method will be introduced. Firstly, the features
extracted from the Convolutional Siamese
Network. Then, the distance between features
will be obtained by the Euclidean distance
function. Finally, the contrastive loss function
is used to enlarge the distance between the
changed feature pairs and to reduce the
distance between the unchanged feature pairs. Figure 2. Siamese Neural Network Architecture
The rest of this paper is organized as follows.
The Methodology is described in Section 2. The convolutional layers can extract the
Section 3 is the experiment results. Finally, the hierarchical features from the input image. The
conclusion of this paper is drawn in Section 4. pooling layers’ functionalities consist of
receptive field enlargement and dimensionality
2. METHODOLOGY reduction, which means to reduce the size of
the output feature maps. The fully connected
Figure 1 describes the processing steps of the layers are used as a classifier, which outputs
change detection task. A pair of images was the probabilities predicting the input image to
input through a full Convolutional Siamese each class.

303
2.2 Contrastive Loss Function (CLF) image background image with no special
object was selected for the reference image at
The contrastive loss was used to enlarge the
time t0, and others are an image at time t1.
distance of changed pairs and reduce the
distance of unchanged pairs simultaneously.
The contrastive loss was formulated as
follows:
 D  fi, fj  y (i, j )  1
CLF   (1)
 max(0, m  D  fi, fj  y (i, j )  0
where fi fj are feature vectors extracted from
the feature pair at the same position, D(fi , fj )
measures the distance between fi and fj using
Euclidean distance function. y(i,j) = 1 means
there is no change at this location. In this case,
the loss function tries to minimize the distance
between fi and fj. Whereas y(i,j) = 0 means
there is a change in this spatial position, Figure 4. Drone Phantom 4 RTK
encouraging the distance to be larger than the
Table 1. Phantom 4 RTK Specifications (DJI
margin, denoted m. In our paper m was set 2
Corporation)
(Guo et al. 2018)
Phantom 4 RTK Parameters
3. EXPERIMENT AND RESULTS Maximum altitude 6000m
3.1 Dataset and devices Maximum speed 58km/h
Maximum flight time 30mins
The testing dataset was obtained by Drone
Phantom 4 RTK (Table 1, Table 2) at GNSS Module GPS, GLONASS,
Deokyang Bridge (figure 3) in a different BeiDou, Galileo
time. The first time (t0) which the image was Stabilization system 3 axes (tilt, roll, yo)
captured is 2019-01-11 and the second time is Maximum controllable 90°/s
2019-04-17. angular velocity
FOV 70° (horizontal) ± 10°
(vertical)

Table 2. Camera Attached on Drone

Specifications (DJI)
Cameras
Sensors 1 "CMOS; Effective
Pixels: 20M
Lens FOV 84 °; 8.8 mm / 24
mm
Figure 3. Deokyang Bridge Area Maximum image size 4864×3648
(4:3); 5472×3648 (3:2)
To train the proposed network, this paper Photo, Video Format JPEG/MOV
used CDnet dataset (Goyette et al. 2012, Wang
Gimbal Stabilization: 3-axis
et al. 2014). The CDnet dataset consists of 31
Pitch: –90° to +30°
videos (91595 image pairs) depicting indoor
and outdoor scenes with pedestrians, boats, and Max Controllable
Angular Speed: 90°/s
trucks that captured in a different time. The

304
 The whole process of training, testing, and
checking results is done in python
programming language, pytorch platform
(Pake et al. 2017), and Linux 18.04 operating
system. The training hardware was two graph
cards (NVidia Titan XP).
3.2 Results
Some images are used to test the
effectiveness of the proposed method. As can
be seen in figure 5 – Test 1; figure 5(a) is the
image at time t0 which captured on 2019-01-
11 and figure 5(b) is the time t1 which
captured on 2019-04-17. In image t0, at this
time, the road surface does not have any Figure 6. Change detection task – Test 2
considerable damage, but in the image t1,
some large pothole on the road surface shows
up. The main task of the proposed method is
to detect the pothole and show the difference
between two image t0, t1. The results in figure
5(c) are good.
Similar to Test 1, the changed areas on Test
2 are also detected; the first pothole is located
in the upper right corner, the second pothole is
located the lower left corner and, the third
pothole is located the lower right corner.
There is one pothole was detected in Test 3
and two potholes were detected at upper right
corner in Test 4. As can be seen in figure
5,6,7,8, most potholes are detectable. Figure 7. Change detection task – Test 3

Figure 5. Change detection task – Test 1 Figure 8. Change detection task – Test 4

305
 Figure 5(d), 6(d), 7(d), 8(d) are the image Bridge Inspection Using an Unmanned Aerial
which generated by blending between the Vehicle. Sensors, 18, 1881, 1 -14.
results image and image at time t1 to see the Koch, G., Zemel, R., Salakhutdinov, R. 2015.
effectiveness of the method, As can be seen, Siamese Neural Networks for One-shot Image
the detected area (color areas) and real damage Recognition. International Conference on
Machine Learning. Lille, France, 1 – 8.
area are total matches. Pan. Y., Zhang. X., Cervone, G., Yang, L. 2018.
Table 3. Change detection rate of method Detection of Asphalt Pavement Potholes and
Cracks Based on the Unmanned Aerial
Test 1 Test 2 Test 3 Test 4 Vehicle Multispectral Imagery. IEEE Journal
Detect Rate 86% 79% 72% 91% of Selected Topics in Applied Earth
Observations and Remote Sensing, 11, 10,
As we can be seen at table 3, Table 3 shows 3701 – 3712.
the percentage of change positions that the Paszke, A., Gross, Chintala, S., Chanan, G., Yang,
method can detect for each of the two input E., Devito, Z., Lin, Z., Desmaison, A., Antiga,
images. In this paper, 4 images pairs are used for L., Lerer, A. 2017. Automatic differentiation
testing, the method can detect up to 86 percent in in pytorch. 31st Conference on Neural
Test 1, 79 percent in Test 2, 72 percent in Test 3 Information Processing Systems. Long Beach,
and over 90 percent in Test 4. This table shows CA, USA. 1 – 4
the effectiveness of the method. Wang, Y., Jodoin, P.M., Porikli, F., Benezeth, Y.,
The reason that the method cannot detect the Konrad, J., Ishwar, P. 2014. Cdnet 2014: an
whole change area is because of the changed expanded change detection benchmark dataset.
area is not a large area and there has not been any IEEE conference on computer vision and
significant change. To overcome this problem, in pattern recognition workshops. 393 – 400.
next research the edge detection will be
integrated to detect more change position.
6. ACKNOWLEDGEMENTS
4. CONCLUSIONS This research was supported by a grant (19RDRP –
In this paper, a change detection method based on B076564 – 06) from Regional Development
the Convolutional Siamese network has been Research Program funded by Ministry of
introduced for UAV image of the road surface. Land, Infrastructure and Transport of Korean
The features of two UAV image in a different government.
time was extracted. Then the distance of the Author Information:
features was generated to detect changes between
image pairs. Experimental results demonstrate the Author 1:
effectiveness of the method. In our future work, 1. Full Name: Mr. Truong Linh Nguyen
we will combine the edge detection and the
Siamese network to get better results. 2. Country/University/Company: Department of
Marine and Civil Engineering, Chonnam
5. REFERENCES National University, Yeosu, Korea
3. Email: [email protected]
Bromley, J., Guyon, I., Lecun, Y., Sackinger, E.,
Shah, R. 1994. Signature verification using a” Author 2:
Siamese” time delay neural network. 1. Prof. Dongyeob Han
Advances in neural information processing 2. Country/University/Company: Department of
system, 737 – 744.
DJI Corporation. 2019. https://www.dji.com/phantom- Marine and Civil Engineering, Chonnam
4-rtk National University, Yeosu, Korea
Guo. E., Fu, X., Zhu, J., Deng, M., Liu, Y., Zhu, Q, 3. Email: [email protected]
Li. H. 2018. Learning to Measure Changes:
Fully Convolutional Siamese Metric Networks Author 3:
for Scene Change Detection. IEEE Transactions 1. Prof. Jungwon Huh
on Multimedia, XX – XX, 1 – 10. 2. Country/University/Company: Department of
Goyette, N., Jodoin, P.M., Porikli, F., Konrad, J.,
Ishwar, P. 2012. Changedetection.net: A new Marine and Civil Engineering, Chonnam
change detection benchmark dataset. CVPR National University, Yeosu, Korea
Workshops. 1 – 8 3. Email: [email protected]
Kim, I.H., Jeon, H.M., Beak, S.C., Hong, W.H.,
Jung, Y.J. 2018. Application of Crack
Identification Techniques for an Aging Concrete

306
The 3rd Int. Conf. on Transport Infrastructure & Sustainable Development (TISDIC 2019). Construction Publishing House. ISBN: 978-604-82-2893-4.

Analysis of woven geotextiles for soft ground

reinforcement with seaming strength

Han-Yong Jeon
Department of Chemical Engineering, Inha University, Incheon, South Korea

Corresponding author, E-mail: [email protected]

ABSTRACT: Seaming property of woven geotextiles for soft ground reinforcement is influenced by process
parameters from manufacture to construction. 8 Geotextiles (4 woven and 4 nonwoven types), 4 geogrids and
2 geocomposites of [nonwoven/fibers/nonwoven] were used and the different seaming methods were applied
to compare the seam properties of 3 geosynthetics and transmissivity of geocomposites. Tensile strength
retentions of these geosynthetics were evaluated as the degree of damage by chemical degradation. Woven
geotextiles showed the higher seam strength in the order (SSd-1<SSd-2) > (SSa-1<SSa-2) > geospacer
without regard to the design strength. For nonwoven geotextiles, the order of seam strength is
geospacer>(SSa-1<SSa-2). Geogrids showed the higher seam strength but reduction factors were increased in
the order of band > geospacer without regard to the geogrid’s compositions. Finally, geocomposites showed
the higher seam strength in the order of geospacer > (SSa-1<SSa-2) but showed the transmissivity in the order
of geospacer > (SSa-1>SSa-2) without regard to the kinds of filled fibers and weight of geocomposite.
Keywords: Seaming property, geosynthetics, geotextiles, geogrids, geocomposites, reduction factor.

1. INTRODUCTION where the sewing threads reverse direction for

their return stroke. For geogrids, the connection
Geosynthetics have their seaming methods such techniques require a high level of dimensional
as mechanical or chemical seaming techniques stability within the geogrid structure to withstand
due to their own structures, physical, chemical the stress concentrations imposed by connection
and thermal properties (FHWA 1989, GRI 2007, techniques (Koerner 2012, Leu 2001). Filtration
IGS 2010). Furthermore, the seam strength of or drainage performance of geocomposites could
geosynthetics must be influenced by seaming be also influenced by the seam strength with
methods and for reinforcement/protection different seaming methods. Different seaming
applications with geosynthetics, the seaming is methods were applied to compare the seam
typically the weak connection to applied load properties of 3 geosynthetics and transmissivities
transfer in the reinforcement/protection system of geocomposites.
(Holtz 1995, Jewell 1996). Besides this, the
seaming of geosynthetics should be evaluated for
2. GEOSYNTHETICS AND THEIR SEAM
durability limitations such as installation
damage, creep deformation, chemical and TYPES
biological degradations etc. Seaming of 2.1 Geosynthetics and Their Functions
geotextiles by traditional sewing methods
necessarily causes needle damage to fibers and Geosynthetic is defined as a planar product,
yarns as well as creating stress concentrations manufactured from polymeric material used with

307
soil, rock, earth, or other geotechnical engineering or with soil. The application areas are
related material as an integral part of a man-made numerous and growing steadily and they
project, structure, or system; ASTM D4354, encompass the entire range of functions listed
D4759, D4873, D5617, D5818 and their functions for geosynthetics: separation, reinforcement,
are as separation, reinforcement, filtration, filtration, drainage, and containment.
drainage, containment (barrier) etc. Generally,
seven specific types of geosynthetics to be applied
to geotechnical engineering field are as geotextiles,
geogrids, geonets, geomembranes, geosynthetic
clay liners, geofoam, and geocomposites.
Among these geosynthetics, geotextiles,
geogrids and geocomposites are used in this (a)
study for soil reinforcement.
The following is brief explanation of these 3
geosynthetics and Figure 1 shows the
photographs of typical examples (Koerner 2012).
Geotextiles are permeable fabrics capable of
multiple functions and form one of the two
largest groups of geosynthetics and consist of (b) (c)
synthetic fibers. These synthetic fibers are Figure 1. Photographs of typical examples of (a)
made into flexible, porous fabrics by standard Geotextiles, (b) Geogrids, (c)
weaving machinery or are matted together in a Geocomposites
random nonwoven manner. Some are also
knitted. The major point is that geotextiles are 2.2 Seam types of geosynthetics
porous to liquid flow across their manufactured Most generally used for field seaming of woven
plane and also within their thickness, but to a geotextiles are called superimposed seam (SS)
widely varying degree. There are at least one and they are formed by placing one geotextile on
hundred specific application areas for top of another and, with a minimum of handling,
geotextiles that have been developed; however, joining these geotextile sections with one or
the fabric always performs at least one of four more rows of stitching (Figure 2).
discrete functions: separation, reinforcement, SSa seam (flat seam) is the weakest of all
filtration, and/or drainage. seam types used for geotextile field installations,
Geogrids are interconnected sets of ribs is made by mating two parallel sections of
used in reinforcement and represent a rapidly geotextile and sewing them together with one or
growing segment within the geosynthetics. more rows of stitches (SSa-1: one row of
Rather than being a woven, nonwoven, or stitches, SSa-2: two rows of stiches, SSa-3: three
knitted textile fabric, geogrids are plastics rows of stiches). There is a tendency when using
formed into a very open grid-like configuration this seam type to sew close to the selvage; this
- i.e., they have large apertures between practice will significantly reduce the ability to
individual ribs in the machine and cross evenly disperse stress between the geotextile and
machine directions. There are many application sewing thread stitches.
areas; however, they function almost SSd seam (butterfly seam) can be considered
exclusively as reinforcement materials. when the geotextile being used is loosely
Geocomposites consist of a combination of woven or has a tendency to slip at the edges.
geotextiles, geogrids, geonets, and/or Slippage or fraying is more likely to occur in
geomembranes in a factory-fabricated unit. loosely woven geotextiles that have fewer
Also, any one of these four materials can be machine and cross machine direction yarns
combined with another synthetic material spaced further apart. Two parallel sections of
(e.g., deformed plastic sheets or steel cables) geotextile are mated and each is turned back

308
toward itself to make a thickness of 4 piles and 3. EXPERIMENTAL
then sewn with either one or more rows of
stitches (SSd-1: one row of stitches, SSd-2: 3.1 Preparation of Samples and Seaming
two rows of stiches). This seam type 8 geotextiles ( 4 woven and 4 nonwoven
effectively prevents the selvage from being types), 4 geogrids and 2 geocomposites of
used as stitching surface. [nonwoven/fibers/nonwoven] structure were
Geospacer is a kind of connector to sew used as geosynthetic materials. Geocomposites
between 2 geosynthetics and Figure 3 shows the of 3-layer structure which have the excellent
seamed geosynthetics, e.g., nonwoven geotextile, drainage function under confined loading
geogrid and geocomposite by the geospacer. condition were manufactured by needle
In this Figure, it was seen that this seaming punching method. Three different punching
method has no problem to be used in the sewen patterns were applied to manufacture these
area instead of the overlapping the geosynthetics, geotextiles through the order of [up
especially the thick geotextiles which have the punching→up punching→down punching]
weight more than 1,000g/m2. process.
3.2 Assessments of Performance of Geosynthetics
In this study, seaming methods that applied
to geosynthetics were as following: (1)
geotextiles - sewing methods: flat seam
(; SSa-1, -2), butterfly seam ( SSd-1, -2),
geospacer, (2) geogrids - band and geospacer,
(3) geocomposites - flat seam (SSa-1, -2) and
geospacer. Table 1 shows the specifications
and seaming methods of these geosynthetics
and Figure 1 shows the various seam types
and geospacer. Seam properties of geotextiles
and geocomposites were examined by ASTM
D 4884(test method for seam strength of
sewn geotextiles) and those of geogrids were
investigated through ASTM D 4884.
Chemical resistance of geosynthetics was
tested in the condition of pH 3 and 8, 25°C
and 50°C during 360 days. These solutions
were adopted to compare the chemical
Figure 2. Schematic diagrams of various seam resistance by considering the waste leachates.
types and special geospacer The degree of chemical resistance of
geosynthetics was estimated by comparing
the tensile strength retention ratio after 360
days with ASTM D 5322. Especially for
geogrids, the reduction factor to affect the
long-term performance was determined by
ASTM D 5262 and GRI GG 3(b).
Transmissivities of geocomposites were
tested by ASTM D 4716. 2 types of geonet
composites - GNC-1, -2 - having the same
(a) NWGT (b) GG (c) GC thickness as smart geotextiles were used as
Figure 3. Photographs of seamed geosynthetics by comparison materials for drainage function
geospacer

309
Table 1. Specifications and seaming methods of geosynthetics.
Design Seaming
Geosynthetics Fiber Weight (g/m2)
Strength (kN/m) Methods
WGT-1 50
Woven WGT-2 100
Polyester
Geotextiles (WGT) WGT-3 150
Flat Seam,
WGT-4 200
Butterfly Seam,
NWGT-1 712
Nonwoven Geospacer
NWGT-2 Poly- 1,024
Geotextiles
NWGT-3 propylene 1,532
(NWGT)
NWGT-4 2,156
GG-1 80
High
GG-2 150
Geogrid (GG) tenacity Geospacer
GG-3 200
Polyester
GG-4 250
Geocomposite (GC) *Poly-
GC-1 1,306 Flat Seam,
[Nonwoven propylene
GC-2 2,084 Geospacer
/Fibers /Nonwoven] Waste Fiber

4. SEAM PROPERTIES optimum for the woven geotextiles. For

nonwoven geotextiles, the geospacer seam
4.1 Seam Strength shows the excellent seam strength than the flat
Figure 4 shows the seam strength of the and butterfly seams with the increase of
seamed geosynthetics with the seam type. For weight. It is thought that geospacer makes
woven geotextiles, the butterfly seam (SSd-1 more compact seam than the flat and butterfly
and SSd-2) shows the excellent seam strength seams for thicker nonwoven geotextiles. For
than the flat seam (SSa-1 and SSa-2) but the geogrids and geocomposites, the geospacer
geospacer seam shows the lower seam strength seam shows the excellent seam strength than
than the flat and butterfly seams. From this, it the band seam without regard to the design
was seen that the geospacer seam is not strength and geogrid’s composition.

(a) WGT (b) NWGT

310
 (c) GG (d) GC
Figure 4. Seam strength of geosynthetics with seam type

4.2 Chemical Resistance woven geotextiles were decreased with high

The degree of chemical resistance of pH value and temperature. This is the typical
geosynthetics in the acidic and alkaline phenomena of polyester fibers which were
solutions was represented in Figure 5 and 6, unstable against the high pH value and
respectively. The tensile strength retentions of temperature.

(a) WGT (b) NWGT

(c) GG (d) GC
Figure 5. Tensile strength retention of geosynthetics with seam type at pH 3, 25C

311
 (a) WGT (b) NWGT

(c) GG (d) GC
Figure 6. Tensile strength retention of geosynthetics with seam type at pH 8, 25C

RFcd = reduction factor for chemical

4.3 Long-term design strength of geogrid by degradation
reduction factor RFbd = reduction factor for biological
From GRI Standard Test Method GG4, the degradation
allowable strength of geogrid could be written
Figure 7 and 8 shows the long-term design
in the following equation to be taken into
strength of geogrids by the reduction factor of
consideration of the ultimate strength,
the above mentioned conditions. We used the
reduction factors for application of geogrids.
default values of each reduction factor to be
The long-term design strength of geogrid
written in the reference.
should be used in the following equation. The
reference document can be found in the GRI
Standard Test Method GG4(b) “Determination
of the Long-term Design Strength of Flexible
Geogrids”.
1
Tdesign  Tultimate [ ]
RFid  RFcr  RFcd  RFbd

where, Tdesign = long-term design strength

Tultimate = ultimate strength
RFid = reduction factor for installation
damage
RFcr = reduction factor for creep
deformation Figure 5. Reduction factor of geogrids

312
 Geocomposites showed the higher seam
strength in the order of geospacer > flat seam
( SSa-1<SSa-2) but showed the transmissivity
in the order of geospacer>flat seam (SSa-
1>SSa-2) without regard to the kinds of filled
fibers and weight of geocomposite.

ACKNOWLEDGMENTS
This research was supported by a
grant(18RDRPB076564-05) from Regional
Development Research Program funded by
Figure 7. Long-term design strength of geogrids Ministry of Land, Infrastructure and Transport
of Korean government.
4.4 Transmissivities of Geocomposites
Figure 8 shows the transmissivities of 6. REFERENCES
geocomposites by seam type. Geospacer seam
FHWA, 1989. Geotextile Design & Construction
showed higher transmissivity than the flat seam
Guidelines, U.S. Dept. of Transportation
(SSa-1>SSa-2) without regard to the kinds of Federal Highway Administration, Publication
filled fibers and weight of geocomposite. For No. FHWA HI-90-001, 24-46.
the similar specification, geocomposites GRI, 2007. GRI Standard Test Methods on
showed the better transmissivities than the Geosynthetics, Drexel University,
typical drainage materials, geonet composites.. Philadelphia, PA, USA
Holtzs, R. D. et al. 1995. Geosynthetic Design and
Construction Guidelines", U.S. Dept. of
Transportation Federal Highway
Administration, Publication No. FHWA HI-
95-038, 27-105.
IGS Educational Resource, 2010. Geosynthetics
Functions, International Geosynthetics Society
Jewell, R. A. 1996. Soil Reinforcement with
Geotextiles, CIRLA Special Publication,
Thomas Telford, Westminster, Chapter 5.
Koerner, R. M., 2012. Designing with
Geosynthetics", 6th Ed., Xlibris Co., USA.
Leu, W., Luane T., 2001. Applications of
Figure 8. Transmissivity of geocomposites Geotextiles, Geogrids, and Geocells in
Northern Minnesota”, Geosynthetics
5. CONCLUSION Conference 2001, 809-821, Portland, Oregon,
USA
Woven geotextiles showed the higher seam
strength in the order of butterfly seam (; SSd-
1<SSd-2)>flat seam(SSa-1<SSa-2)> geospacer
without regard to the design strength. For
nonwoven geotextiles of 1,000g/m2~, the order
of seam strength is geospacer > flat seam
(SSa-1<SSa-2). Geogrids showed the higher
seam strength in the order of band>geospacer
but creep deformation were increased in the
order of band > geospacer without regard to the
design strength and geogrid’s compositions.

313
The 3rd Int. Conf. on Transport Infrastructure & Sustainable Development (TISDIC 2019). Construction Publishing House. ISBN: 978-604-82-2893-4.

Going underground - a solution for sustainable city in Viet

Nam - aspect of infrastructure

Vu Thi Thuy Giang1, Do Nhu Trang2

1
Civil engineering Faculty, Hanoi Architecture University, Ha Noi, Viet Nam
2
Civil engineering Faculty, University of transport technology, Ha Noi, Viet Nam

Corresponding Author: Vu Thi Thuy Giang

Address correspondence to First author: http://hau.edu.vn

ABSTRACT: Vietnam is facing many challenges coupled with its rapid urbanization. Big cities such as
Ha Noi and Ho Chi Minh city are presented with unprecedented strain on their urban infrastructure.
Traffic congestion, lack of parking areas, asynchronous and overlapping utilities are among the serious
problems facing these cities. This paper provides some recommendations for sustainable urbanization in
Vietnam by going underground, based on foreign experiences and Viet Nam’s situation.
Keywords: urbanization, sustainable urban, underground, integrated.

1. INTRODUCTION Group published ITA Report N°11

“Underground Solutions for Urban
Economic growth is one of the main factors Problems” which addresses this topic [4].
that contribute to urbanization [2]. The 20th The main goal of this report is to create
century witnessed a booming of skyscraper awareness of the underground options in city
construction. Our globe is experiencing planning.
unprecedented growth in urban population. The The uses for underground structures have
corresponding sprawl of existing urban areas developed from primarily shelter to space for
and the appearance of new cities and infrastructure and a wide range of functional
urbanizing regions are occurring at rapid rates. facilities. The uses may be categorized into
The ever larger and denser concentration of several primary uses: (i) infrastructure for
population is causing serious concerns. transit and utilities, (ii) storage, and (iii)
Approximately 70% of the world’s inhabitants protection of the environment. Increasingly,
by 2050 (forecast to be 10 billion at that point) the public, especially in larger cities, demands
will live in urban areas [2], [4]. a higher quality of living environment with
In 1991, International Tunnelling respect to [4]:
Association (ITA) provided special studies - Reliable and safe transport of people and
and publications to address the use and goods,
construction of underground space by Tokyo - Water distribution and sewerage systems,
Declaration in 1991: “The 21st century is the - Sustainability of the environment and
century of underground space development containing sprawl,
and utilization” [4]. Working Group N°20 - More green spaces and recreational areas,
was formed to provide an overview of the - Reduced use of fuel, and fuel emissions,
typical challenges of urban city planning and - Noise control,
the solutions which are offered by the use of - Aesthetics,
underground space. In 2012, the Working - Efficient use of real-estate.

314
 Thousands of years ago, underground reliable and resilient, and to make
structures had provided shelter and storages for underground space desirable, so we need to
our human ancestors. Today, they also help in understand the key to sustainability and
the use of the limited and valuable space in service for these systems is trustworthiness.
urban areas by replacing traditionally surface As one of the fastest growing economies
facilities with tunnels, caverns and other in ASEAN, Vietnam has a seemingly endless
underground spaces. The main uses of the wish list for infrastructure. Currently, just 20
underground are [6]: per cent of the country’s national roads are
- Transport use (with emphasis on paved, and a recently approved plan to build a
infrastructure): provides for transportation 1,372 km north-south highway by 2030 is
tunnels and underground utilities. estimated to cost USD 14 billion. The rising
- Production use:extraction of underground population in major cities in recent years has
resources. strained and exceeded capacity of the existing
- Urban structure use: structure foundations connectivity networks and utilities systems.
and underground facilities such as shopping With 50% of Vietnam’s population expected
malls, recreational facilities, carpark and to be living in cities, Hanoi and Ho Chi Minh
working space are building rapid transit systems exceeding
- Storage use: space for materials best placed USD 22 billion in the hope of reducing private
underground such as hazardous materials and vehicle ownership and improving air quality
energy storage. [1], [3].
The urban “problems” are well documented Rapid urbanization and economic growth
and those that may be solved through the use have led to a rise in travel demand and the
of the underground space include [7]: use of personal vehicles. With high density of
- Crowding and lack of space (for work and population and urban structure, big cities in
recreation), Viet Nam, especially Ha Noi and Ho Chi
- Traffic congestion, Minh City, are facing an unbalanced
- Aging infrastructure and distribution of development of the economy and the
resources, infrastructure. This creates challenges in
- Environmental conditions such as noise urban transport such as traffic jams, high
and air pollution, accident rate, environmental pollution, and
- Aesthetic quality and image of urban serious lack of parking areas. Extensive use
environment quality, of personal vehicles in the cities has further
- Safety, security, and protection against contributed to these issues.
natural disasters, Traffic congestion has cost Ha Noi between
- Flooding, sewage conveyance and US$1 and 1.2 billion per year and has made air
treatment, pollution five times worse (the Institute for
- Synergy effects of the above. Transportation and Development Policy -
MOT). Meanwhile, Ho Chi Minh City is
2. CHALLENGES IN URBAN confronted with growing traffic congestion.
INFRASTRUCTURE DEVELOPMENT OF For example, according to the World Bank’s
VIET NAM studies [14], transit time from Ha Noi to Hai
Phong averages 4.5 hours because of road
2.1 Traffic congestion and lack of parking areas congestion compared to 2 hours without
The underground must add to the desirability congestion.
and vitality of the urban experience - and our The Municipal Transport Department of Ho
underground design must uplift and inspire Chi Minh City speculated that the road surface
those who use and visit the urban needs to be doubled by 2020. Several elevated
environment. Therefore, it is important to highways and metro lines are currently being
make underground infrastructure systems planned and constructed.

315
 oversight caused illegal construction, inter–
influence between sectors: tele-communication
cable, energy power cable, fresh water pipe,
sewer culvert, etc..

Figure 1. Congestion traffic

(www.baoxaydung.com.vn)

To reduce the use of personal vehicle, big

cities in Viet Nam are also planning to develop
and improve bus systems and metro rail
transit. Ha Noi has ambitiously planned several Figure 2. Construction cable underground on Le
modern metro rail networks such as the pilot Van Luong road using trench method
metro Nhon – Ha Noi station, Cat Linh – Ha disturbs the road surface and pavement.
Dong, Tran Hung Dao – Thang Long South, (From: Hồng Kỳ/TTXVN)
etc.All of them are still being constructed.
Urban railways and bus rapid transit (BRT)
systems are being developed and the pilot line
Yen Nghia – Cat Linh in Ha Noi is expected to
promote public transportation in big cities.
To solve the problem of parking area
shortage, the Government has agreed on the
underground car-parking solution in Ha Noi
(Thong Nhat Park, Nhan Chinh Park and Quan
Ngua Stadium) and in Ho Chi Minh city
(August revolutionary square, and Co Tan
Park). However, these projects have been only Figure 3. Construction sewer culvert in Pham Van
on paper and no exact date for execution has Dong road (Ha Noi) (From: Danviet)
been confirmed.
2.3 Water flooding
2.2 Over lapping of utility network
In Viet Nam, floods occur in the rainy season
In the first decades of 19th century, Viet Nam (June to November) influenced by the tropical
had several utility tunnels used to provide fresh monsoons, and also during the typhoon season
water or carry sewage in Ha Noi and Ho Chi (September to October) in the northern and
Minh city. Built by the French colony, some of central areas where typhoons land. Storm
them are still in use today. rainfalls in the upper basin of the Mekong
For the decades between 1950s and early cause flood disaster over wide areas in the
1990s, North of Viet Nam had stepped back in Mekong delta.
developing infrastructure. It was the same with Flood disasters in urban areas like Ha Noi
South of Viet Nam after 1975. and Ho Chi Minh City tend to be amplified
In early 2000s, the economic boom induced because of their poor drainage systems [0].
out-of-control emigration and created pressure The occurrence of heavy floods in Ha noi
in urban cities. The lack of infrastructure as an effect of climate change interrupts
policy as well as inadequate planning and normal lives of its citizens. In Ho Chi Minh

316
City, rising sea level causes ever more
serious flooding in lower areas. Although
decision-makers have employed some
solutions such as pumping, improving canals
and culvert, they proved to have only
temporary effects.
Recent flood disasters are [0]:
- The flood on September 30, 2011
Figure 4. Historical flooding in Ha Noi in 2008
triggered by the storm rainfalls of Typhoon (https://nld.com.vn )
Nesat caused damage in the central areas, and
flooding from the Mekong River caused
damage in Ho Chi Minh and its surrounding
area.
- The flood of November 2008 triggered by
the storm rainfalls of three days from
November 1 caused the worst flood damage in
the last 20 years in the capital of Hanoi and its
surrounding area, as well as 17 provinces in the
central region. The economic losses were Figure 5. Flooding tide in Ho Chi Minh City
estimated at about USD 400 million. (https://www.tienphong.)

Figure 6. Current and Planned Railways, Metro Tracks, Airports, and Ports Affected by projected in Ho
Chi Minh City Extreme Floods by 2050 without Proposed Flood Controls (Source: ADB) [0].

317
3.1 Foreign experiences - Japan built the Metropolitan area outer
underground discharge Channel, also known as
A modern city can only be sustainable if it has the G-cans project, in Tokyo to prevent over
the ability to withstand natural disasters or other flooding.
important risks induced by climate change - Malaysia: built in 2007, the Stormwater
causing raising sea level and heavy rain. Management And Road Tunnel (SMART
To this end, the development of the urban Tunnel) is a storm drainage and road structure
underground space in order to deal with with 9.7 km long, it is used to reduce the
vulnerability and resilience issues is extremely occurrence of floods and traffic jams in Kuala
important to ensure a sustainable urban city. Lumpur.
Many cities face water flooding risks and
tunnels are being built for flood management 3.2 Recommendations for urban cities in
purposes, as it is considered the most effective Viet Nam
solution. However, during emergencies it may To meet the demand of economic development,
be complicated or even impossible to evacuate in 2009, National Assembly accepted the Urban
in time a large city even if an early warning planning law No. 30/2009/QH12, which focused
system is available. on urban underground construction and space
A finalist in the eVolo 2010 Skyscraper management sector. In addition, some standards
Competition, Bunker Arquitectura gained are issued such as: Viet Nam Construction Code
international recognition for Earth scraper, a QCXDVN 01: 2008/BXD on Construction
55-story inverted underground skyscraper. The planning; National technical regulation QCVN
visionary design proposed a mixed-use 07: 2016/ BXD on technical infrastructure
underground development for installation in works, QCVN 13: 2018/BXD on Garage, QCVN
Mexico City’s main plaza. The design concept, 08: 2018/BXD on Subway works; Vietnam
which addressed multiple urban constraints, construction standard TCXDVN 104: 2007 on
included 10 floors dedicated to a Pre- urban roads - design requirements, etc.
Columbian museum; 10 floors of retail space However, cities must also strive to become
and housing; and 35 floors of office space. smart,efficient and sustainable places. In order
Descending 1,000 feet below the surface, the to achieves these goals, cities should consider:
inverted pyramid also featured a central void - Improve infrastructures and public transport
for natural light and ventilation, a glass ceiling flow.
and a variety of green spaces [6]. - Availability of dynamic systems to guide the
Some examples: traffic towards its final destination - the parking.
- The city of Oslo in Norway has been using - Adopt a more effective and ecological
underground gymnasiums, swimming pools, parking solution that allows drivers to find the
bowling halls etc. for many years. These free parking space quickly, safely and easily.
facilities can be converted to civil defences - Integrate infrastructure sectors.
shelters at short notice.

Figure 7. Underground swimming pool, Helsinki,

Finland [ITA] Figure 8. Inter-depencies of infrastructure sectors [11]

318
 When examining the more general cases of
multiple infrastructures connected as a “system
of systems”, interdependencies of infrastructure
sectors should be considered [11].
Figure 8 illustrates the interdependent
relationships among several components of
infrastructure. These complex relationships are
characterized by multiple connections among
infrastructure, feedback and feed-forward Figure 10. A multi-utility tunnel [0]
paths, and intricate, branching topologies. The
connections create an intricate web that, 4. CONCLUSION
depending on the characteristics of its linkages, Due to the advancement of underground
can transmit shocks throughout broad swaths construction methods, changes in technology
of an economy and across multiple and design are likely to emerge in parallel with
infrastructures. a better understanding of the impacts on
Urban development planning, on the one ground performance and related parameters.
hand, coordinates between spatial levels and That may require greater integration of systems
between sectoral plans and sectoral policies. on the design front, including development of
On the other hand, it makes use of the formal geotechnical/ geological data ownership and
instruments, such as physical development management systems that will enhance
planning or sectoral planning, for its knowledge sharing, while reducing costs [0].
implementation. Apart from thematic focuses To reshape urban cities into smart,
and depth of detail, integrated and strategic sustainable places as the ambition of the
urban development planning comprises a Government [0], a promising solution is to
number of recurring tasks. make use of underground space. The location,
In practice, these are not usually a simple scale and aesthetics of underground space can
linear series, but form the modules of an have a significant impact on the perception
interlinked system with numerous feedback and consequently the success of the urban
loops (Figure 9). development.
For Viet Nam to to achieve sustainable
urban targets, some recommended focus areas
include:
- Improve public transport (e.g. bus system,
metro rail system) to reduce the use of personal
vehicle.
-Synthesis and intergrated underground
synchronisation between the systems since
they were constructed in different periods of
time without coordinated planning.
- Use trenchless method in underground
Figure 9. Proposed integrated urban planning and construction for utility network instead of
development process [12] trench method (for examples in pipe jacking,
mini-tunneling, TBM, etc.).
A good example is to integrate - Incorporate convenience, user-friendliness
infrastructure by using multi-utility tunnel and aesthetics into infrastructures, transport
which inside all kinds of pipes and cables can and utilities in urban undergound space.
be laid, i.e. water pipes, sewer channels, To put the above recommended points into
power cables and telecommunication action, olicy-makers could allocate more
networks (Figure 10). resources on:

319
 - Identifying and understanding role of AHA Centre and JICA. 2015. Country Report
urban infrastructure system interdependencies. Vietnam - Natural Disaster Risk Assessment
- Design wholistic strategies and encourage and Area Business Continuity Plan
initiatives in policy-making, planning and Formulation for Industrial Agglomerated
outsourcing (for example, to encourage private Areas in the ASEAN Region, Japan
International Cooperation Agency.
sector participation in underground
Dimitris Kaliampakos. 2016. Underground
construction). development: a springboard to make city life
- Create data and methodologies to establish better in the 21st century, 15th International
a complete map for the entire urban scientific conference Underground
underground space. Integrate monitoring Urbanisation as a Prerequisite for Sustainable
systems and collect data in real-time during Development, Procedia Engineering 165, 205 –
construction and in operation. 213
+ Study and compile technical specifications Priscilla Nelson et al. 2012. Sustainability and
of design, construction and management of Resilience of Underground Urban
urban underground space, adapted to to be Infrastructure: New Approaches to Metrics
compatible with the Viet Nam condition. and Formalism, Geotechnical Special
Create long-term models, guidelines and Publication.
T.Yigitcanlar, S. Teriman. 2015. Rethinking
objectives as a guidance framework for private
sustainable urban development: towards an
and public actors. integrated planning and development process.
Int. J. Environ. Sci. Technol. 341–352,.
5. REFERENCES Wout Broere. 2012. Urban problem - underground
solution, Advances in Underground Space
ADB. 2010. Ho Chi Minh city adaption to climate Development – Zhou Cai & Sterling (eds),
change. Summary report. World Bank. 2012. Efficient logistics a key to
Godard. 2004. Urban Underground Space and Vietnams competitiveness, Technical report.
Benefits of Going Underground. World
Tunnel Congress 2004 and 30th ITA General
Assembly, Singapore.
Zargarian, Hunt, Braithwaite, Bobylev and Rogers.  
2016. A new sustainability framework for
urban underground space, Proceedings of the
Institution of Civil Engineers.
ITA Working Group Urban Problems -
Underground Solutions. 2012. Report on
Underground solutions for urban problems.
ITA report N°011.
ITU-T Focus Group on Smart Sustainable Cities.
2015. Multi-service infrastructure for smart
sustainable cities in new development areas.
WSP. 2018. Taking Urban Development
Underground, Future-ready solutions for
ensuring urban sustainability. WSP Global Inc.
Antonia Cornaro and Han Admiraal. 2017.
Sustainable urbanization through underground
development – towardss an urban
underground future, Advances in Structural
Engineering and Mechanics (ASEM17),
Korea.
Dean Cira and et al. 2012. Vietnam Urbanization
Review. World Bank Technical Report.

320
The 3rd Int. Conf. on Transport Infrastructure & Sustainable Development (TISDIC 2019). Construction Publishing House. ISBN: 978-604-82-2893-4.

Modal choice behavior of high school students in Danang

city, Vietnam

Ong Ich Truong1,*, Fumihiko NAKAMURA2, Shinji TANAKA3, Shino MIURA4

and Ryo ARIYOSHI5
1
Master of Philosophy in Infrastructure Management, Group 3B, HoaTho Dong, Cam Le, Danang, Vietnam.
2
Executive Director, Vice President, Graduate School of Urban Innovation, Yokohama National
University Yokohama city, Japan
3
Associate Professor, Graduate School of Urban Innovation, Yokohama National University,Yokohama
city, Japan
4
Assistant Professor, Graduate School of Urban Innovation, Yokohama National University,Yokohama
city, Japan
5
Associate Professor, Graduate School of Urban Innovation, Yokohama National University,Yokohama
city, Japan

* ONG Truong Ich, Email: [email protected]

ABSTRACT: This study investigates the travel mode choice behavior of the high school students in Danang
city, Vietnam. Based on the data collected on 1,350 students studying at a city’s downtown high school and a
city’s suburb one, a Multinomial Logit model (MNL) was developed to estimate the relationship between the
travel mode choice for going to schoolsamong the students and the influencing factors. The results show that
there are three significant factors including waiting time, school location, and traffic safety that affect the
students’ mode selection behavior.Basedon the study’s findings, measures such as bus waiting time
reduction, increasing the city’s bus coverage, and raising the awareness of bus as a safe transportation
modeare finally recommended for the purpose of encouraging the bus use among the city’s students.
Keywords: Mode choice, high school student, MNL, Danang, Vietnam.

1. INTRODUCTION Along with the popularity, the motorcycles

are now one of the major causes of urban
The motorcycle use has been increasing in
problems such as traffic accidents or air
recent years in developing countries such as
pollution. The WHO has reported thatabout
Vietnam, Indonesia, and Thailand. Because of 34% of the road traffic deaths in South-East
its convenience and cheap price, motorcycles Asia were the 2-3 wheelers. The traffic
become the most preferable choice of people accidents were the main cause of the death
for their daily trips. According to the report of among people who aged from 15 years to 29
the World Health Organization (WHO) in 2015 years (World Health Organization, 2015). Many
(World Health Organization, 2015), Vietnam of them were the high school students who
has nearly 40 million motorcycles which accounted for the high traffic demandin the
account for 95% of the registered vehicles in peak hours when they traveled to the schools.
the country compared to 83%, 59%, and 38% The Vietnamese educational system
in Indonesia, Thailand, and China respectively. comprises of four levels: Early childhood,

323
general education, vocational education, and recommended forpromoting the use of the city
higher education. The high school education public transport and mitigate the motorcycle
belongs to the level of general education use among the students to the city
together with the primary and secondary school transportation planners for their consideration.
education. According to the government This paper will be continued with the next
regulation, the city has to ensure that there are at section of the literature review, followed by the
least one primary school, one secondary school section of a brief description of the study area.
in each city’s commune, and one high school in The fourth part describes the data collection
each city’s district. Hence, the younger students methodology and results. Mode choice
could manage to go to the schools by walking or modeling’s description and results are
bicycles because the schools are close to their presented in the fifth part. The last part will
houses. However, when they become a high both recommend some measures for the city
school student at the age of 15, they are more public transport use improvement and conclude
dependent on the motorized vehicles such as this study.
motorcycles, cars, or buses because they have to
2. LITERATURE REVIEW
travel farther for their schools.
High school time is a transition phase for a Recently, modal choice behavior for the school
Vietnamese person from a parent dependent trips has attracted much attention from the
schoolboy to a more independent university transportation researchers. They have
man. This phase is very important for the discovered many influencing factors that
students to perceive the world and build their significantly affected the mode selection for the
attitude toward different aspects of the society. school trips among the studentsfrom many
Their behavior at this time will shape their places in the world. Alemu and
future behavior (Zhang et al., 2017). If their Tsutsumi(Alemu and Tsutsumi, 2011) have
travel mode choice behavior for the school found out that, to Japanese high school students
trips is well understood, the city transportation in Okinawa, there were several factors affected
planners could propose measures to encourage the modal choice pattern for going home from
the students to use the public transport instead schools like age, the distance between home and
of being escorted to the schools by their schools, traveling times or bus fare. They also
parents using motorcycles. This could help to found some different patterns in mode choice
form a good habit of using public transport between the urban school students and suburban
among the students before they get more school students. The downtown Japanese
dependent on motorcycles when they enter the students tended to use buses for their trips while
universities (Nguyen-Phuoc et al., 2018). the suburb ones were more dependent on the
Despite such importance, there are few cars. In another study, Daisy and Habib (Daisy
studieson the mode choice behavior of the high and Habib, n.d.) have discovered that to the
school students in a motorcycle dominant students in grade six to ten in Dhaka city,
country like Vietnam. This study is expected to Bangladesh, the factors that affected the mode
supplement the shortcoming. The study aims at choice for their school trips were found to be
investigatingthe relationship between the to- household income, the presence of sidewalk and
school travel mode selection among the high intersection around students’ houses in addition
school students in Danang city, Vietnam, and to age, traveling time, and distance.
the factors affect the choice. The data for this In China, with the concern with the school
study were collected from a survey on the travel environment of the students, Li and Zhao
students who were studying in two high (Li and Zhao, 2015) conducted a research on
schools in the city. The Multinomial Logit travel mode choice behavior for the journeys to
model was then used to estimate the relation school among children aged from thirteen to
between the probability of selecting travel fifteen in Beijing. Their study revealed that the
modes and various impact factors. Based on age of students, gender, household income,
the model’s findings, some measures could be travel distances, and Hukou policy in China

324
were the strong factors that influenced the 3. STUDY AREA
mode choice patterns of students. This study’s
This study’s area is Danang city which locates
results were analogous to those found in the
in the middle of Vietnam (Figure 1). The area
study of Zhang, Yao, and Liu (Zhang et al.,
of Danang is quite small, about 1,283km2. Its
2017) when they studied the students’ school
population is more than 1million people in
travel mode choice in Beijing. They also found
2015 which is much lower than that of other
some factors like departure time for school and
parental escorting that affected the modal big cities in Vietnam like Ha Noi and Ho Chi
choice pattern of the students. Minh with the population about 7.2 million and
In Western countries, the mode choice 8.2 million respectively in 2015.As same as
behavior for school trip among the students was other Vietnamese cities, motorcycles are very
also paid attention. Distance, car availability, popular in the city with 91.83% of the city’s
and the weather were found to be the most households own this kind of travel mode
influencing factors that affected the choice of (“Danang Statistical Yearbook,” 2016).
mode for going to school among students aged According to a survey of JICA in 2008, about
from ten to nineteen in Dresden city, Germany 77% of the daily trips were made by the
by the study of Müller, Tscharaktschiew, and motorcycles. Meanwhile, this share of the bus
Haase (Müller et al., 2008). On the other hand, was only 0.9%. Majority of the bus users were
to Canadian students, built environment factors the school students (JICA and Danang People’s
(i.e street intersections, block density, and Committee, 2008).
access to the transit) were significant factors As of the year 2015, the number of high
besidesother factors like distance and gender school students in the city was about 28,950
(Mitra and Buliung, 2015). which account for 2.81% of the city’s
However, in the context of motorcycle population. This number was quite small
dominant countries, there is a handful of compared to that of other cities like Hanoi and
studies that address the issue of mode choice Ho Chi Minh with the number of 180,879 and
among the high school students. In Thailand 192,684 respectively (“Vietnam Statistical
where motorcycles account for 59% of the Yearbook,” 2015). Currently, due to the shortage
registered vehicles in the country (World of the schools’ facilities, the high school students
Health Organization, 2015), the high school have to be divided into 2 groups, one group will
students mostly considered factors including be studying in the morning time and the another
travel cost, the transit service flexibility and will be studying in the afternoon time for the first
availability, the safety, vehicle convenience, semester. When the second semester begins,
and transit reliability when they select their these two groups will be interchanged. This
travel modes for the to-school journey. The resulted inhigh travel demand for the schools in
older students with the driving license were two periods of from 6a.m to 7a.m and from
mostly using motorcycles while the younger 11a.m to 12a.m (JICA and Danang People’s
ones were using the paratransit service Committee, 2008). With the objective of
(Wongwiriya et al., 2016). In Vietnam, a becoming a sustainable city in Vietnam, the city
majority of the researches were focusing on the has deployed more bus lines from 2016 and Bus
behavior of the working people (Tran et al., Rapid Transit (BRT) project to satisfy the travel
2016), or the Vietnamese university students demand of the citizens including the students as
whose mode choice were affected by their age, well as motivate the modal shift from private
their gender, income, motorcycle ownership, vehicles to public transport among people.
travel distance, travel cost, and safety Hence, understanding the mode choice behavior
(Nguyen-Phuoc et al., 2018) (Le and Trinh, of the commuters, particularly the high school
2016). The travel mode choice behavior of the students becomes essential to the city’s
school students has not been well understood. transportation planners to propose solutions for
Because of this matter of fact, it is necessary to improving the public transport service in addition
have more studies for filling this gap. to suchinfrastructure investment.

325
Figure 1. Study area and the surveyed schools for the study

4. DATA SURVEY their accessibility to thecity bus service. The

second section gathered the data on the
4.1 Methodology students’ travel characteristicssuch as the
The data was collected by the questionnaire frequent mode use, the reasons for selecting to-
survey conducted in the September 2017. Two school travel modes, bus waiting time.The
high schools, Phan Chau Trinh high school available travel modes were provided for the
which locates in the city downtown area and students’ choices including walking, bicycles,
Hoa Vang high school which locates in the city public buses, parental escort by using
suburbia, were selected for the survey (Figure 1). motorcycles (hereafter “motorcycles”),
In each school, the sampling was determined parental escort by using cars (hereafter “cars”).
by the randomly selected classes whose size is Other than the questionnaire survey, the data
about 45 students per one class unit. The class on the city bus travel time and the bus stop
monitors facilitated the questionnaire location were also collected using the portable
distribution and collection. Global Positioning System (GPS) device.
The survey required the students to read, Before conducting a full-scale survey, pilot
formulating the answers for the questions, and surveyswere necessarily conducted. The purpose
accomplishing the whole questionnaire by of this survey is not only to test the
themselves. The questionnaire mainly focused appropriateness of the questionnaire form and
on two sections. The first section collected the the question words, but also to provide the
students’ personal information such as gender, information for the sample size calculation
age, household size, the children number in the (Richardson et al., 1995). For the first pilot
household, household motorcycle ownership, survey, 40 high school students were selected.
the school location, and the school time. The results from this survey led to the revision
Especially, the students were asked to report of the questionnaire. The second pilot survey
their house’s addresses which in turn will be involved other 10 participants before the
used to geocode in ArcGIS to calculate the official full-scale survey took place in
distance from their houses to the schools, and September 2017.

326
4.2 Sample size select bicycles, buses, or walking to their
school more than their counterparts. The mode
The sample size was calculated using the
shares of bicycles, buses, and walking among
following formulae (Richardson et al., 1995):
the suburb school ones were 60.91%, 4.54%,
n and 6.14% respectively and higher than those
n (1)
 n  among the downtown school students. On the
1   other hand, the downtown school students were
N
asking their parents to escort them to the
Where: N = 28,950: Number of the city’s high school by using motorcycles and cars more
school students (“Danang Statistical Yearbook,” than the suburb ones.The use of motorcycles
2016). and cars among the downtown school students
n: Sample size for an infinite population. were 37.12% and 2.68% respectively compared
p 1 p  to 27.05% and 1.36% among the suburb school
n  (2) students.There was not any difference in the
s.e  p 
2
surveyed students’ age, household size, and
Where: p: proportion of a sample in motorcycle ownership. The average age of the
population. students was around 16 years old.There were 4
s.e(p): Standard error members in their house. And each household
had more than 2 motorcycles. The household
s.e  p   e / z (3) income of the downtown school students
Where: e: Margin of error or confidence limit. ranged within 500USD-700USD and was
z: z-score of normal distribution with the higher than that of the suburb ones.
mean of 0 and the standard deviation of 1. The distances from students’ houses to their
With the information obtained from the pilot schools and their closest stops were calculated
surveys, the sample size was calculated for the using the ArcGIS based on the student houses’
mode use proportion with the level of addresses reported in the questionnaire.
confidence of 95% and the sampling error of According to the calculation, the suburb school
5%. The results were summarized in table 1. students were living about 3.33km from the
school and a little farther than the downtown
4.3 Data description oneswith the distance of 3.20km. The
There were totally 30 classes corresponding to downtown school students easily accessed the
1350 high school students picked for the city bus service when they were living close to
survey. About 1038 out of 1,350 questionnaires the bus stops with the distance of 0.53km. On
were sent backed from the classes’ monitors. the contrary, the transit accessibility of the
This accounted for about 77% of the response suburb ones was more difficult when the
rate. The student and their household statistics distance from their houses the nearest bus stops
have been shown in the table 2. As shown in was up to 1.39km. If the students those were
the table, the downtown school students living within 800m from the city bus stops
accounted for the share of 57.61%. About the were assumed to be available to the transit
gender, the female respondents were more than service (Ben-Akiva and Lerman, 1985), it is
the male ones with the share of 58.86% calculated by using ArcGIS that 66.18% of the
compared to 41.14%. Regarding the mode use surveyed students had the bus service. The
in the sample, a majority of the surveyed downtown school students were served by the
students were using bicycles for their school city bus service better than their counterparts
trips with the largest share of 56.94%, followed when 80.27% of them could access the bus
by motorcycles, walking, buses, and cars. stop. On the other hand, this share of the
There were differences in the mode use suburb ones were only 47.05%. Despite that
between the downtown and the suburb school fact, the bus uses between them were almost
students. The suburb school students tended to equal (19/39 compared to 20/39).

327
 Table 3 shows the students’ and their more than their counterparts. The age,
households’ statistics classified by the mode household size, and motorcycle ownership
use. Based on this table, female students among the mode users were quite identical.
were using bicycles and buses more than the Regarding the distance, the bus users were
male students. The share of bicycles and those living farthest from their schools and
buses among the school girls were 59.41% close to the bus stops with the distance of
and 4.09% respectively compared to 53.40% 4.20km and 0.44km correspondingly. Those
and 3.28% among the schoolboys. of other mode users were not so different
Meanwhile, the schoolboys wereusing except for the walking students who were
motorcycles, cars, and walk to the schools living nearest to the schools.

Table 1. The sample size calculation

Mode use p (%) 100-p e (%) z(95%) s.e(p) n N n
Motorcycle 54.1 45.9 5 1.96 2.551 382 28,950.00 377
Bicycle 37.8 62.2 5 1.96 2.551 361 28,950.00 357
Car 2.7 97.3 5 1.96 2.551 40 28,950.00 40
Bus 2.7 97.3 5 1.96 2.551 40 28,950.00 40
Walking 2.7 97.3 5 1.96 2.551 40 28,950.00 40
From the table 1, the sample size for the full-scale survey was determined of no less than 377 individuals

Table 2. Student and their household statistics (Classified by school location)

Downtown Suburb Average/Total
No of students 598 (57.61%) 440 (42.39%) 1038 (100%)
Male 275 (45.99%) 152 (34.55%) 427 (41.14%)
Gender
Female 321 (54.01%) 288 (65.45%) 611 (58.86%)
Bicycles 323 (54.01%) 268 (60.91%) 591 (56.94%)
Motorcycles 222 (37.12%) 119 (27.05%) 341 (32.85%)
Mode share Buses 19 (3.18%) 20 (4.54%) 39 (3.76%)
Walking 18 (3.01%) 27 (6.14%) 45 (4.33%)
Cars 16 (2.68%) 6 (1.36%) 22 (2.12%)
Age 16.30 16.14 16.23
Household Size 4.38 4.49 4.43
Motorcycle ownership 2.49 2.55 2.52
Household Income 500$~700$ 300$~500$ 300$~500$
House-School Distance (Km) 3.20 3.33 3.26
House-Bus Stop Distance (Km) 0.53 1.39 0.68
687
Bus stop catchment < 800m 480 (80.27%) 207 (47.05%) (66.18%)
> 800m 118 (19.73%) 233 (52.95%) 351 (33.82%)

328
Table 3. Student and their household statistics (Classified by mode)
Bicycles Motorcycles Buses Walking Cars Average/total
228 150 14 22 13 427
Male
(53.40%) (35.13%) (3.28%) (5.15%) (3.04%) (100%)
363 191 25 23 9 611
Female
(59.41%) (31.26%) (4.09%) (3.76%) (1.47%) (100%)
Age 16.21 16.29 15.97 16.16 16.41 16.23
Household Size 4.46 4.39 4.38 4.36 4.55 4.43
Motorcycle
2.47 2.71 2.21 2.18 2.23 2.52
Ownership
Household Income 300$~500$ 300$~500$ 300$~500$ 300$~500$ 700$~1000$ 300$~500$
House-School
3.39 3.25 4.20 0.46 3.43 3.26
Distance (Km)
House-Bus Stop
0.79 0.68 0.44 0.41 0.65 0.68
Distance (Km)

5. MODE CHOICE MODELLING Viq = β0 + β1X1 + β2X2 + ... + βnXn (6)

5.1 Model description Where: β0: Alternative Specific Constant; (β1,
β2,..., βn): Vector of weighting coefficients
In this paper, the Multinomial Logit model
associated with the vector of influence factors
(MNL), which has been widely used in the
or independent variables (X1, X2, ..., Xn).
field of travel behavior, was used to analyze
the surveyed data to discover the mode choice 5.2 Travel mode choice set
behavior for the school trips among the city
high school students. Ortúzar and Willumsen (Ortúzar S. and
The model is based on the random utility Willumsen, 2011) have pointed out that, to
theorem which postulates that the alternatives apply the random utility theorem which the
are represented by a measure so-called utility MNL modelis based on, certain homogeneity
(U) and the individuals will select the in the population of interest is needed. This
alternative that maximizes their net personal means that all the individuals for modeling
utility subjects to legal, social and physical must share the same alternative set and face
constraints(Ortúzar S. and Willumsen, 2011). the same constraints. Therefore, it is
The utility is characterized by two components
important to segment the surveyed
as follows:
individuals into sub-groups based on their
Uiq = Viq + εiq (4) available travel modes for going to the
Where Uiq: Utility ofalternative i for individual school. To this end, some judgments were
q; Viq: The systematic, measurable component made as follows: (1) The students whose
of the utility function; εiq: The random household owned the motorcycles will be
component, assumed to be independent and available to motorcycles for going to the
Gumbel distributedwhich results in the general schools; (2) The students who were living
form of the MNLmodel as follows: within 800m from the bus stops could use the
Exp Viq  buses for school trips (Ben-Akiva and
Piq  (5) Lerman, 1985). And the reported feasible
 jCn Exp V jp  travel modes by the students will be included
Where Piq: The probability of individual q selects in the choice set as well. The segmentation
the mode i; Cn: The set of available modes. results were summarized intable 4.

329
5.3 Independent variables (mode choice to the schools in the morning when other
influence factors) people are going to work. This is increasing
the traffic volume in the morning which in
The independent variables which possibly affect
turn causes the traffic congestion in the urban
the probability of travel mode selection are were areas. Concerning the traffic accident during
comprised of the personal, journey, and travel the time of high traffic volume, the parents
mode variables and shown in table 5. will be more willing to escort their children to
5.4 Model results schools when their children ask. This is
consistent with what Zhang (Zhang et al.,
The MNL model estimation results are shown 2017) found in his study on travel mode
in table 6. In the model, the mode bicycle is choice behavior of school children in Beijing.
used as the reference. The results only show Larsen, Gilliland, and Hess (Larsen et al.,
the variables that are significant at the level of 2012) also discovered that it is the safety
5%. Other insignificant variables were concern associated with high traffic volume in
excluded from the final model. the rush hour that limited the use of active
There are three variables including the mode like bicycles.
waiting time, the school location, and the traffic Regarding the school location, this variable
safety found to be significant to affect the travel is significant to the bus choice only. To the
mode choice behavior of the city high school motorcycle choice, it is not significant at all.
students. From the results, it is found out that the The coefficient’s sign of the school location
students prefer using bicycles to using other variable is negative which means that; when
modes like buses and motorcycles. If all the considering the travel mode choices between
variables are zero, the utility functions of the bicycles and buses, the suburb school students
buses and the motorcycles will only have the tend to choose the buses instead of the bicycles
alternative-specific constant. They will become for going the schools. On the contrary, the
negative and smaller than that of the bicycles. In downtown school students tend to select the
fact, the highest use of bicycles revealed by the bicycles rather than the buses. This behavior
survey could prove this bicycle preferable result. could be explained by the fact that the average
The safety variable is significant at the distance from the suburb school students’
level of 5% to the bus choice and at 1% to the houses to their school is farther than that of
motorcycle choice. The positive sign of downtown ones. The pattern of bus
safety’s coefficient reflects the positive effect dependence among suburb school students
of the factor. This means that when the found in this study is similar to what was found
students consider their own safety to select the in the study of Merom (Merom et al., 2006) on
modes for the school trips, the students tend to the active communicating to school among
select buses or ask their parents to escort them school children in New South Wales, Australia.
to the schools using motorcycles instead of Finally, the waiting time is the significant
biking by themselves. This behavior could be factor that is found to disutilizes the travel
explained by the fact that traffic safety is one mode choice of the students. The negative sign
of the urgent problems in Vietnam. With the of the variable’s coefficient does mean that the
increase in the number of traffic accidents and utility of the mode will decrease if the waiting
the very high death rate of 0.94 per accident in time increases. The students will be more
Vietnam compared to those of 0.71, 0.17, and likely to select the travel modes that have the
0.16 of Indonesia, Thailand, and China shorter waiting time. This finding is consistent
respectively (“Vietnam National Traffic with the finding of Mohammed and Shakir’s
Safety Committee,” 2014), the safety is the study when they examined the effects of
most concerned issue of not only the students, waiting time on the bus use of the students in
but also of their parents. The students often go Malaysia (Mohammed and Shakir, 2013).

330
Table 4. Available travel mode choice sets (set of only one mode excluded).
Sets of available travel modes No of Individuals
Car and motorcycle 34
Car, motorcycle, and bicycle 106
Car, motorcycle, bicycle, and walking 47
Car, motorcycle, and walking 5
Motorcycle and bicycle 248
Motorcycle, bicycle, and bus 247
Motorcycle, Bicycle, and walking 191
Motorcycle and walking 25
In this paper, the choice set includes three modes of motorcycle, bicycle, and bus corresponds to 247
individuals was used to analyze the students’ mode choice behavior using the MNL model.

Table 5. The set of independent variables.

Variables Variable identification
Personal variables
Gender Dummy variable.
1 for male and 0 for female.
Age Years old.
Household size Number of family members living together in the house at the time of surveying.
Household children Number of school children living together in the house at the time of surveying.
Motorcycle ownership Number of motorcycle owned by each household member.
(Total household’s motorcycles)/(Household size).
Comfort Dummy variable.
1 if it is the reason for mode selection and 0 otherwise.
Convenience Dummy variable.
1 if it is the reason for mode selection and 0 otherwise.
Safety Traffic safety.
Dummy variable.
1 if it is the reason for mode selection and 0 otherwise.
Journey variables
House-school distance Distance from student’s house to his/her school (Kilometer, Km).
Calculated by ArcGIS based on the student reported addresses.
School time Dummy variable.
1 if the student went to school in the morning and 0 otherwise.
School location Dummy variable.
1 if the student was studying at the downtown school and 0 otherwise.

331
 Variables Variable identification
Travel mode variable
House-Bus stop distance Distance from the student’s house to the nearest stop (Kilometer, Km).
Calculated by ArcGIS based on the student reported address and GPS recorded
bus stop location.
Travel time Time for going from the student’s house to his/her school (Minute).
Waiting time Minutes
Travel cost By motorcycle:
= (Motorcycle operation cost unit)*House-school distance + parking fee (VND).
Where:
Motorcycle operation cost unit = 1500 (VND) (Kutani and Sudo, 2016).
Parking fee = 4000 (VND) (Kutani and Sudo, 2016).
By bicycle:
= parking fee = 2000 (VND).
By bus:
= Bus fare + egress cost (in case of motorcycle).
Bus fare = 5000 (VND) (current bus fare).
Egress cost = motorcycle operation cost unit * house-bus stop distance.

Table 6. MNL model estimation results.

Variable Coefficients Standard Error t-value Pr(>|t|)
Reference: Bicycle
Generic variable
Waiting Time -0.227 0.089 -2.541 0.011 *
Bus
Alternative Specific Constant -0.948 1.028 -0.922 0.357
School Location -1.159 0.538 -2.154 0.031 *
Safety 1.712 0.790 2.167 0.030 *
Motorcycle (escorted)
Alternative Specific Constant -2.853 0.578 -4.935 0.000 ***
School Location -0.014 0.398 -0.036 0.971
Safety 1.765 0.547 3.224 0.0013 **
Significant level: ‘***’ 0.1% ‘**’ 1% ‘*’ 5%
Log-Likelihood: -157.38
McFadden R^2: 0.089
Likelihood ratio test : Chi-square = 30.644 (p-value = 1.1e-05)

6. RECOMMENDATION AND CONCLUSION money to improve the city bus serviceand

thereby promote the public transport use among
6.1 Recommendation
the citizens. More bus lines have been deployed
A good public transport network plays a key role to cover more areas of the city. Cleaner and more
in forming a sustainable city (Reijsbergen and comfortable buses have been used to replace the
Gilmore, 2014). Danang city is on its way to old buses. The city is also implementing the Bus
becominga developed and sustainable city in Rapid Transit (BRT) project. It is expected to be
Vietnam. Recognize the importance of the public operated within a near future to serve the city
transport, the city has been investing a lot of travel demand better.

332
 However, along with such investment in the Other thanaforementioned measures, some
hardware infrastructure, it is necessary to actions should be taken for improving the
invest in understanding the citizen travel quality of the city bus service which in turn
behavior, particularly their travel mode choice encourages the bus use among students such as
behaviorto be able to propose appropriate providing the timetable as well as other
measures to promote the bus use among information of the city bus service like bus
people, especially the school students.Some routes, bus stops, and transfer buses to the
measures which resulted from this study should students, limiting the operation of the low-
be considered thoroughly as follows: quality intercity bus service in the urban areas,
Firstly, to the demand side, propagandas, using technology such as Public Transportation
safety classes should be held regularly to raise Priority System (PTPS) or the bus tracking
the awareness of the high risk of getting injured system for a smoother and faster bus service.
from riding motorcycles or bicycles and promote The abovementioned measures will more or
the benefits of using buses for the school trips less require the budget from both city
among the students. To the supply side, the bus government and the bus companies. Hence,
driver crews have to be trained on the safety from the beginning, to encourage the providers
issues throughtraining programs or safety to improvetheir service, the city should have
workshops. Some measures to secure the safety special policies for the companies such as
in the bus operation like alcohol detection system business preferential tax, special land rental
or driver recorder system needto be charges, or fare subsidy.When the companies
implemented. The training process and safety are able to live on the income from their service,
measures must be published in the social media these policies could be mitigated or withdrawn.
to gain more belief and interest from the students
as well as other people in the city bus service. 6.2 Conclusion
Secondly, more bus lines need to be addedto This study investigates the travel mode choice
increase the bus coverage toward the suburbia behavior of the high school students in Danang
to satisfy the demand from the students who city, Vietnam. The Multinomial Logit model was
are living in these places. One drawback of this used to discover the effects of the influencing
measure is its implementation cost. Adding factors on the possibility of selecting the travel
more bus lines means more budget has to be modes for the journey to schools among the high
spent. Hence, instead of adding more bus lines, school students. There are three factors found to
some current ones should be changed toward have the significant effects including the traffic
the suburb areas to spread the city bus service. safety, the school location, and the waiting
Although this measure also costs for bus time.According to the model results, traffic
stopconstruction and other relevant facilities, it safety affects significantly the likeliness of the
is cheaper than adding more lines. bus and motorcycle choice. If the students
Thirdly, the bus waiting time should be consider their safety for their school trips, they
reduced by shortening the bus headway. will tend to select buses or ask their parents to
Currently, the city bus headway is 10 minutes drive them to the schools instead of cycling by
in the peak hours and 20 minutes in the rest. It themselves.About the school location, it has the
is recommended that the headway out of the effects on the bus choice only. To the motorcycle
peak hours need to be mitigated to 15 minutes choice, it does not affect at all. Compared to the
at which,according to Desaulniers and bicycle choice, the suburb school students are
Hickman (Desaulniers and Hickman, 2007), more likely to select the buses for their school
the passengers will arrive the bus stop trips. On the other hands, the downtown ones
randomly and wait for the buses regardless of incline toward the bicycle choice.Finally, the
the bus schedule. However, this measure is waiting time affects the probability of travel
also costly due to the bus journey increase to mode choicethat the students will select the
meet the headway reduction. mode that possesses the less waiting time.

333
 With those key findings, some measures Mitra, R., Buliung, R.N., 2015. Exploring
such as bus waiting time reduction, increasing differences in school travel mode choice
the city’s bus coverage toward the suburbia, behaviour between children and youth.
Transp. Policy 42, 4-11.
and raising the awareness of bus as a safe Mohammed, a.a., shakir, a.a., 2013. factors that
transportation modecould be recommended for affect transport mode preference for graduate
the purpose ofpromoting public bus use and students in the national university of malaysia
mitigating the personal mode selectionamong by logit method 8, 13.
the students to the city’s transportation Müller, S., Tscharaktschiew, S., Haase, K., 2008.
planners for their consideration. Travel-to-school mode choice modelling and
This investigation is still limited because of patterns of school choice in urban areas. J.
Transp. Geogr. 16, 342-357.
the absence ofparental effects on the students’ Nguyen-Phuoc, D.Q., Amoh-Gyimah, R., Tran,
travel mode choice andalimited choice set. For A.T.P., Phan, C.T., 2018. Mode choice among
the future work, a deeper study considering the university students to school in Danang,
parents’ impactsand larger choice sets should Vietnam. Travel Behav. Soc. 13, 1-10.
be conducted. Ortúzar S., J. de D., Willumsen, L.G., 2011.
Modelling Transport, Fourth edition. ed. John
7. REFERENCES Wiley & Sons, Chichester, West Sussex,
United Kingdom.
Alemu, D.D., Tsutsumi, J.G., 2011. Determinants Reijsbergen, D., Gilmore, S., 2014. Formal
and spatial variability of after-school travel by Punctuality Analysis of Frequent Bus Services
teenagers. J. Transp. Geogr. 19, 876-881. Using Headway Data, in: Horváth, A., Wolter,
Ben-Akiva, M.E., Lerman, S.R., 1985. Discrete K. (Eds.), Computer Performance
choice analysis: theory and application to travel Engineering. Springer International
demand, MIT Press series in transportation Publishing, Cham, pp. 164-178.
studies. MIT Press, Cambridge, Mass. Richardson, A.J., Ampt, E.S., Meyburg, A.H.,
Danang Statistical Yearbook, 2016. 1995. Survey methods for transport planning.
Daisy, N.S., Habib, M.A., n.d. Modeling the Eucalyptus Press Melbourne.
factors affecting mode choice behaviour for Tran, M.T., Zhang, J., Chikaraishi, M., Fujiwara,
school trips in Dhaka City, Bangladesh. A., 2016. A joint analysis of residential
Desaulniers, G., Hickman, M.D., 2007. Public location, work location and commuting mode
Transit, in: Handbooks in Operations Research choices in Hanoi, Vietnam. J. Transp. Geogr.
and Management Science. Elsevier, pp. 69-127.
54, 181-193.
JICA, Danang People’s Committee, 2008. The
Study on Integrated Development Strategy for Vietnam National Traffic Safety Committee, 2014.
Danang City and Its Neighboring Area in the Vietnam Statistical Yearbook, 2015.
Socialist Republic of Vietnam. Wongwiriya, P., Nakamura, F., Tanaka, S., Miura,
Kutani, I., Sudo, Y., 2016. Addressing Energy S., Ariyoshi, R., 2016. User perception of
Efficiency in the Transport Sector through paratransit in Thailand: case study of journey
Traffic Improvement. to school in khon kaen city. J. Jpn. Soc. Civ.
Larsen, K., Gilliland, J., Hess, P.M., 2012. Route- Eng. Ser D3 Infrastruct. Plan. Manag. 72,
Based Analysis to Capture the Environmental I_801-I_807.
Influences on a Child’s Mode of Travel World Health Organization, 2015. Global status
between Home and School. Ann. Assoc. Am. report on road safety 2015: supporting a
Geogr. 102, 1348-1365. decade of action. WHO, Geneva, Switzerland.
Le, T.P.L., Trinh, T.A., 2016. Encouraging Public Zhang, R., Yao, E., Liu, Z., 2017. School travel
Transport Use to Reduce Traffic Congestion and mode choice in Beijing, China. J. Transp.
Air Pollutant: A Case Study of Ho Chi Minh Geogr. 62, 98-110.
City, Vietnam. Procedia Eng. 142, 236-243.
Li, S., Zhao, P., 2015. The determinants of
commuting mode choice among school children
in Beijing. J. Transp. Geogr. 46, 112-121.
Merom, D., Tudor- Locke, C., Bauman, A., Rissel,
C., 2006. Active commuting to school among
NSW primary school children: implications
for public health. Health Place 12, 678-687.

334
The 3rd Int. Conf. on Transport Infrastructure & Sustainable Development (TISDIC 2019). Construction Publishing House. ISBN: 978-604-82-2893-4.

An improved version of cell-transmission model for air

traffic flow

Tran Quang Khai1,*, Kunihiko HIRAISHI2

1
Faculty of Information Science, Japan Advanced Institute of Science and Technology, Nomi, Japan
2
Faculty of Information Science, Japan Advanced Institute of Science and Technology, Nomi, Japan

Corresponding author, E-mail: [email protected]

ABSTRACT: Previously, we can possibly describe the dynamics of every aircraft in the airspace using
Lagrangian method. Airspace traffic models using Lagrangian method might have high accuracy in terms
of predicting the trend of the air traffic flow. The most significant disadvantage of those models is that
they require a lot of computational effort (due to the spherical surface of the Earth and the complexity in
the dynamics of aircraft). Recent research shows that Eulerian method is a potential approach for
modeling the airspace traffic. Models using Eulerian method can be simple, low-order, and computational
friendly besides providing reasonable accuracy. In this paper, we first revise some popular models using
Eulerian method in order to extract the strong points of these models. Secondly, we propose a systematic
approach to form animproved version of the Modified Cell-Transmission Model. The proposed model
maintains the advantages of existing models such as vector-based aircraft count state, low-order system,
while it can still cover a larger air area.
Keywords: Big data, statistical analysis, airspace traffic, traffic assignment.

1. INTRODUCTION thereforemostly describe the traffic between

sectors and FIRs.
Airspace is the medium where the aircraft
travel. Like the mainland, which is divided into
several administrative areas, the airspace of a
country is also divided into a finite number of
areas called Flight Information Regions (FIRs).
Aircraft operate in a FIR will be provided flight
information services and safety guidance. Some
countries have only one FIR in their territorial
airspace, such as Japan (Fukuoka FIR, Fig. 1).
While some countries have two or more FIRs,
such as the United Kingdom (London, Scottish
and Shanwick Oceanic). An entire FIR can also
be divided into some smaller areas called
sectors. Thus, aircraft will travel between
sectors (domestic flights)and between FIRs Figure 1. Fukuoka FIR (Source: JapanMinistry of
(international flights) to get to destination Land, Infrastructure, Transport and
airports. Models that represent airspace traffic Tourism)

335
 There are two main methods for modeling the can be described as follows. Let ni(t) be the
air space traffic: Lagrangian method and number of vehicles in cell iat time step t. Under
Eulerian method. In this paper, we use the light traffic conditions,the number of vehicles in
termsLagrangian models and Eulerian models to cell i at time step t + 1 is calculated as follow:
refer to airspace traffic models using Lagrangian
ni  t  1  ni 1  t  (1)
method and Eulerian method, respectively. In
most Lagrangian models, the trajectory and Equation (1) denotes the most important
coordinates of an aircraft will be described by condition in CTM and its variations: under
some functions of time. Because of the spherical light traffic condition, every vehicle in a cell
surface of the Earth, it requires at least three (ni1) is expected to advance to the next cell
differential equations to describe the above (ni)between each time step. However, in case
characteristics of an aircraft. Also, due to the of traffic jam, Eq. (1) cannot be used to fully
small sample size (when comparing to land road describe the dynamic of the traffic in each
traffic), it is possible to model all aircraft in an cell.The evolution of traffic under constraints
observed airspace.Eulerian models, by contrast, is shown in Eq. (2):
forming some control volumes. Those volumes
have some important characteristics such as size, ni  t  1  ni  t   yti1  t   yii 1  t  (2)
density, and flow rate. The flow of aircraft is
Equation (2) can be explained as follow:
described between those volumes. The most
when congestion happens, the number of
major advantage of Eulerian models over
vehicles in cell i at time t + 1(ni(t + 1)) is not
Lagrangian models is that it can significantly
always equal to the number of vehicles in the
reduce the order of the system.
In this paper, the Modified Cell-Transmission previous cellat time t(ni-1(t)) as only a portion
Model (MCTM) and the Large-Capacity Cell- of vehicles in cell i1 will move to cell
Transmission Model (CTM(L)), two popular  yti1  t   , and also only a portion of vehicles in
cell will advance to cell i  1 yii 1  t   . In the
airspace traffic models based on the Cell-
Transmission Model arestudied. We analyze
these two models thoroughly in order to exploit original paper, Daganzo proposed that a cell
the advantages of them. Furthermore, we could contain a limited number of vehicles. We
propose anapplication of stochastic framework also denote Ni(t) as the maximum number of
into the model by utilizing the Aggregate vehicles that cell i can have at time t.
Dynamic Stochastic Model (ADSM). Parameter yti1  t  is determined by the
following constraint:
2. CELL-TRANSMISSION MODELS FOR
THE AIRSPACE TRAFFIC yit1  t   min ni 1  t  ,Qi  t  ,Ni  t   ni  t  (3)

2.1 The Cell-Transmission Model and its In Eq. (3), parameter Qi(t) represents the
variations maximum number of vehicles that can flow
from cell i1 into cell in one timer interval.
The Cell-Transmission Model (CTM) is a
Ni  t   ni  t  is the empty space in cell at
Lagrangian model originally built for land road
traffic. Proposed by Daganzo (1994),the idea time t.
behind CTM is to divide a one-way, single-entry To the best of our knowledge, MCTM is the
and single-exit road into some one-dimensional, first CTM-based model for the airspace traffic.
homogeneous parts called cells. The length of Menon et al. (2004) divide the airspace into
one cell is determined by the distance travelled some areas. Each area can contain one or two
by a typical vehicle in one time interval.Each cells. The relationship between number of
cell can hold a limited number of vehicles. vehicles in the same cell (Fig. 2) between two
Between each time step, vehicles in one cell will continuous time steps is expressed as:
advance to the next cell. The dynamics of CTM ni  t  1  ni  t    i qi 1  t    i qi  t  (4)

336
 merged into one stream qk , the relationship
between them is described as follow:
qk  qk 1  qk 2 (7)

Although it is possible to have more than

two outflows from a cell, in the MCTM, the
Figure 2. 1-D cell in MCTM merge/diverge flows are limited to two.
Another notable CTM-based model for the
In Eq. (4), i is the duration of one time
airspace traffic is the CTM(L) (Sun and
interval, qi 1  t  is the flow rate of vehicle from
Bayern, 2008). CTM(L) is a graph-theoretic
cell i  1 into cell i, and qi  t  is the flow rate model. CTM(L) only considers flights at
of vehicle from cell i into cell i  1. Equation certain altitude (24,000 ft. or above - Sun and
(4) is obviously equivalent to Eq. (2). In order Bayern, 2008). Based onthe geographical
to increase the accuracy of the model, Eq. (4) is location of sectors, some “links” are created to
modified so that it can handle the small change represent the flight path of aircraft between
in the traffic as follow: those sectors (Fig. 3). The length of a link is
decided based on the average travel time of
 v 
ni  t  1  1  i i i  ni  t   typical aircraft on that link. Then, a link will be
 i  (5) divided into several cells. The length of one
ACC  
  i qi t   i qi 1  t  cell is equal to the distance that typical aircraft
travel in one timeinterval.The discrete-time
In Eq. (5), i is the proportional factor dynamical system is expressed as follows:
corresponding to the portion of vehicles that
can travel from cell i to i  1, i is the xi  t  1  Ai xi  t   Biin fiin  t  
(8)
dimension of cell i, and vi is the aggregate  Bidesc fi desc  t   Bic limb fi c limb  t   Biu ui  t 
speed of aircraft in cell i. The major drawback
of the MCTM is that it is unclear to choose a
suitable cell’s size. Also, due to the large time
interval i (15 minutes in the original
literature), the size of one cell is big, thus a cell
can cover an entire sector. It leads tothe
accumulate error is bigand the model cannot
capture important characteristics of the
airspace traffic when it runs for a long time
(Smith et al., 1997). Sometimes, the flow of
aircraft in one cell can divide into two or more
stream, or two or more flow can also merge
into one streamand flow into one cell. The Figure 3. Airspace in the CTM(L)
MCTM handles these cases with the diverge
and merge nodes. For a diverge node where In Eq. (8), xi(t) is the state of cell i at time t It
stream qk is divided into two smaller streams is a column vector that has the form
T
qk+1 and qk+2, the mathematical expression of xi   xin xin 1  xi0  . Ai is called the system
the flows is: matrix.It is a square matrix of order n (n is the
qk 1   qk ; qk  2  1    qk (6) number of cells in link i). Vector fiin  t  is called
 is called the divergence parameter. For a the forcing input. The integer input fi desc  t  and
merge node where stream qk 1 and qk 2 are fic limb  t  is called the climbing input and

337
descending input. They represent the number of to another sector (B or C). A flight is called
aircraft entering or leaving a link from time t to arrival flight when it satisfies two conditions: it
t  1, respectively. Both Bic limb and Bidesc are must be on the boundary between sector A and
sector B at the time of observation, and it is
column vectors that have the form
going to land at an airport inside sector A.
B   0 0  1 0  0 . The kth element in B
T
In each sector, we introduce the term “route”.
that takes value 1 represents the entering/leaving A route is an imaginary road that allows all
cell.Vector ui(t) represents the number of aircraft aircraft that have similar flight’s type to travel.
that are under control by the ACC. Because of Therefore, we have three types of route in a
the altitude lower bound, the CTM(L) needs the sector: fly-over route, departure route and
parameters fi desc  t  and fic limb  t  to describe arrival route (Fig. 4). For examples, fly-over
flights that are travelling from sector A to sector
the aircraft that leaving or entering a cell in the
C through sector B, they are considered
middle of a sector. However, due to the
travelling on the same fly-over route in sector B,
characteristic of vector Bic limb and Bidesc , aircraft we denote this route as BAC . Without the loss
can only join a link at a certain cell. Method for of generality, we denote BC  A as the fly-over
choosing the suitablecell’s position for entering
route that allows flights from sector C to sector
and leaving aircraft is also not mentioned in the A (crossing sector B). Note that the length of
original paper (Sun and Bayern, 2008). BAC and BC  A can be different. Similarly,
2.2 Characterized cell-transmission model departure flights that departs from airport inside
sector A and travel to sector B will travel on the
In this section, we propose animproved version
same departure route, we denote this route as
of theModified Cell-Transmission Model
A A . Lastly, to represent the route for flights
called Characterized Cell-Transmission Model. 0 B
The model can cover a larger airspace that travel from sector B and land at airport
comparing to the MCTM and CTM(L). It also inside sector A, we use route A A . In the
B 0
eliminates the problem of determining MCTM, all three types of route are incorporate
corresponding cells for entering/leaving
as one link with the length of this link is equal
aircraft in the CTM(L).
to the airspace dimension (Menon et al. 2004).
2.2.1 Flight categorizing and route determination Thus, with the modified version, we can clearly
observe the flows of aircraft inside a sector.
In practice, any authorized flight would travel
on finite routes called airways to get to the
destination airport. Flights that have the same
departure airports and arrival airports are likely
to share the same set of airways (Sun and
Bayern, 2008). Using this feature, we can
categorize flights in a sector into three groups:
fly-over flights, departure flights, and arrival
flights. The definition of each group is as
follows. Let A, B, and C be sectors. We assume
that all of them are neighbors to each other. A Figure 4. Airspace in original MCTM and
flight is called fly-over flights if it satisfies two Characterized CTM
conditions: it is already airborneat the time of
observation, and it is travelling from sectorA to In this modified version, there is no
sector Cby crossing sector B without stopping. interconnection between routes inside a sector.
A flight is called departure flight when it However, routes between adjacent sectors can
satisfies two conditions: it is going to depart have connection with some strict rules (Fig. 5)
from an airport inside sector A, and it will travel as follows. All connetions are directional,

338
which meansthe order of routes in a connection 1994) and the MCTM (Menon et al. 2004) to
is irreversible. We denote P  Q as the calculatethe length of route and cell. The
connection between route P and Q, then: length of a route is determined by the average
- For fly-over routes: A fly-over route in travel time of typical aircraft on this route.A
sector A can connect with another fly-over route will be divided into several cells. The
route in sector B, the connection must length of one cell is equal to the distance
satisfy the follow rules: the connection travelling by typical aircraft in one time-
must satisfy the following order: interval (chosen as one minute in this paper).
AX  B  BAY ; a fly-over route in sector B The following assumptions are made in order
to describe the dynamics of the model:
can connect with a arrivalroute in sector
- Only regular flights are considered in the
A, the connection must satisfy the
model.
following order: BX  A  A A ; a fly-
B 0 - All aircraft have a constant speed on any
over route cannot be linked to any routes.
departure route. - Each route only allows traffic in one
- For departure routes:a departure route can direction.
only be connected to a fly-over route. The - There is no limit to the number of aircraft
connection must satisfy the following in a cell unless noted.
order: A A  BA X - All aircraft must depart from an airport and
0 B
land at another airport,that means there is
- For arrival routes: an arrival route cannot
no intermediate exist on any routes.
be connected to any other types of route.
The dynamics of the model is shown in
Eq (9):
nk  t  1  nk 1  t   qkACC   ACC  t 
1 t  qk (9)

In Eq. (9), nk  t  1 is the number of aircraft

in cell k at time t  1, nk 1  t  is the number of
aircraft in cell k-1 at time t, qkACC   and
1 t

qkACC  t  are the number of aircraft that are under

control from Area Control Centre (ACC) at time
t. Under light traffic condition (no control action
Figure 5. Connection between routes
qkACC  t   , Eq. (9) can be written as:
The following approach shows three main
nk  t  1  nk 1  t  (10)
advantages of the characterized CTM over the
MCTM and CTM(L): firstly, it can cover the Equation (10)satisfies the condition of the
entire observed airspace with no altitude original CTM (Eq. (1)). Also, the dynamics of
restriction (35000 ft in previous papers); the Characterized MCTM as shown in Eq. (9)
secondly, the time interval of the model is is totally equivalent to the original MCTM. To
small enough to capture the characteristics of see why, we consider Eq. (5). The coefficient
the air traffic while maintains the is always chosen as 1 as in the original
computational friendlypurpose; and thirdly, it literature (Menon et al., 2004). All aircraft
eliminates the route decision problem that cruise at constant speed then is a constant.
exists in previous models.  v
Therefore  i vi  i and 1  i i i  0. Equation
i
2.2.2 Model description
(5) now becomes:
With the above explanations, we use the same
method as in the original paper (Daganzo ni  t  1   i qiACC  t    i qi 1  t  (11)

339
 Parameter qi 1  t  as suggested by Menon et types of weather can happen, then we have to
al. (2004), is calculated as follow: consider the departure and arrival probability for
v n n
qi 1  t    i i ni 1  t   qiACC
i
1
t  . Then  i1  i   2n  1 possible combination of
 
 i qi 1  t   ni 1  t    i qiACC
1
t  . Equation (11) weather's types.
now takes the form: 2.2.3 Numerical results
ni  t  1  ni 1  t    i qiACC  t    i qiACC
1
 t  (12)
To validate the fidelity of the model, we use
It is obvious that Eq. (12) in the MCTM is CARATS (Collaborative Actions for
equivalent to Eq. (9) in the Characterized Renovation of Air Traffic Systems) Open Data
MCTM. The Characterized MCTM at this point provided by Japan Ministry of Land,
is largely deterministic. The impact of stochastic Infrastructure, Transport and Tourism (MLIT)
factors can be considered using the ADSM as the dataset. CARATS Open Data contain the
framework (Sridhar et al., 2003, 2008). Let A be raw radar flight information of regular flights
sector, then 0A is an airport inside sector A. We in Fukuoka FIR. Table 1 shows a portion of
denote s A  t  and s A  t  as the aircraft count sample of CARATS Open Data. To determine
0
for airport 0A and sector A, respectively. If a the corresponding sectors of each aircraft from
weather event W that affects the departure rate raw data, we use the method suggested by
happens from time step t, then the route input of Tokumaru et al. (2018).
any route that has the form A A (X is a We choose four sectors that represent
0 X
neighbor sector of A) will be calculated by: different density of the air traffic: Misawa
West Sector S01 (low density), Chugoku High
qin
A A
 t   s A  t  pW
A A . The component
0 X 0 o X Sector F11-2 (mediumdensity), Tohoku Sector
W
pA A is the probability that an aircraft will T02 (high density), and Kanto East Sector T04-1
o X (very high density). We observe those sectors
enter route A from airport 0A if event in four days in four different months. The
0AX
happens. Using the same reasoning above, if a observation period is 12 hours per day. The
weather event Z that affects the arrival rate from duration of one time interval is chosen as one
sector A to airport 0A, the flow rate control input minute. It is small enough to catch the
of any route that has the form A A will have important characteristics of the traffic
X 0
andcomputational friendly (Sridhar et al. 2006;
the form qkACC  t   nk  t   p AZ nA  t  .
X 0 A Sun and Bayern, 2008). Figure 6 shows the
Where p AZ is the probability that aircraft affection of different time interval to the
X 0 A accuracy of the model. In this paper, we assess
will pass the last cell of route A if event Z the sector count (number of aircraft inside a
X 0 A
happens. The above computation is just for one sector at specific time) to measure the
kind of weather event. If in sector A, there are n reliability of the model.

Table 1. Sample CARATS open data

Time Flight ID Latitude Longitude Altitude Aircraft Type
00:00:00.4 FLT0001 22.152099 127.175188 38025 B739
00:00:00.9 FLT0002 36.066731 45.189828 33975 B77W
00:00:00.9 FLT0003 36.274698 143.505599 35000 B772
00:00:01.3 FLT0004 33.434379 134.940998 34000 B773
00:00:01.4 FLT0005 35.634703 141.285142 32983 B77W

340
 As there is no intermediate exist in any most time steps is just one aircraft while the
route, aircraft will join a route by entering the maximum absolute error is less than three (Fig. 9
first cell of the corresponding route. For each and 10). The maximum absolute error of the
sector, we first determine all available types of proposed model is always less than or equal to
route, then calculate the number of routes for the resultsof the original MCTM and CTM(L)
each type. The length of each route is as shown by Sun et al. (2007).
calculated based on historical flight data. Due
to the significant difference in terms of Table 2. Number of each type of routes in
altitude, we process only the proposed model observed sectors
with the CARATS dataset. That is, if we limit Sector Fly-over Departure Arrival
the altitude, the Characterized CTM would route route route
now take the form of the MCTM and CTM(L), F11-2 22 2 2
and the performance and fidelity of the MCTM
S01 14 1 0
and CTM(L) are showed (Sun et al., 2007).
Results are calculated on a computer running T02 19 4 5
Windows 10 with the following specifications: T04-1 24 8 3
CPU IntelCorei5 7600 3.5GHz, 16GBof RAM.
The implementation program is written in Table 3. Some routes in Misawa West sector
Python Language (version 3.7.2). The size of (S01)
data file varies between 1.1 megabytes and 2.2
megabytes. The computational time is less than Route Average flight Number of
time (seconds) cells
five seconds in all sectors.
The proposed model successfully captures S01S032T01 710 12
the trend of traffic in the observed sectors (Fig. 7 S01S032S31 595 10
and 8). The maximum absolute error is capped S01T01S032 584 10
at four in very high-density traffic sector (T04-2). S01T01S31 504 9
In low-density and medium-density traffic
S01S31S01 304 6
sectors (F11-2 and S01), the absolute error for

Figure 6. Sector count for Tohoku sector (November 09, 2015) with different time interval

341
Figure 7. Sector count in different days for Kanto Figure 8. Sector count in different days for
East Sector Chugoku Sector

Figure 9. Sector count error in different days for Figure 10. Sector count error in different days for
Chugoku Sector Kanto East Sector

3. CONCLUSIONS the approach of CTM(L) to the MCTM and

proposed a modified CTM-based model that
Forecasting the traffic movement in air traffic
can exemplify the characteristic of aircraft's
at the present time is important due to the
increasing traffic scene.CTM is a powerful flow in a sector and it can cover a larger air
technique to modeling the traffic flow. area. It can lessen the difficulty to decide the
However, it is just applied to air traffic in corresponding cell for departing and arriving
recent years. In this paper, we have first aircrafts in the CTM(L). Additionally, the
revised some notable CTM-based models in model considers the departure airport and
order to find the advantages and difficulties of arrival airport as part of the model, thus it can
them in modeling air traffic. We have incorporated control the flow rate at a lower level. The

342
results based on CARATS Open Data show Arneson, H.M. and Langbort, C., 2007, December.
that the fidelity of the model is promising as Distributed control design for a class of
compartmental systems and application to
the aircraft distribution by the model
Eulerian models of air traffic flows. In 2007
successfully reflects the traffic trend in the 46th IEEE Conference on Decision and
observed sectors. This information is Control (pp. 2876-2881). IEEE.
supportive for the Area Control Centres to Roy, S., Sridhar, B. and Verghese, G.C., 2003,
manage the flights in a sector with suitable June. An aggregate dynamic stochastic model
actions such as flow rate control or adjusting for an air traffic system. In Proceedings of the
departure and arrival rate. Further work can be 5th Eurocontrol/Federal Aviation Agency Air
Traffic Management Research and
considered on the stochastic factors as the
Development Seminar, Budapest, Hungary.
transition probabilities required mining a large Canudas-de-Wit, C. and Ferrara, A., 2018. A
amount of historical data. variable-length Cell Transmission Model for
road traffic systems. Transportation Research
4. REFERENCES Part C: Emerging Technologies, 97, pp.428-
455.
Daganzo, C.F., 1994. The cell transmission model: Sumalee, A., Zhong, R.X., Pan, T.L. and Szeto,
A dynamic representation of highway traffic W.Y., 2011. Stochastic cell transmission
consistent with the hydrodynamic model (SCTM): A stochastic dynamic traffic
theory. Transportation Research Part B: model for traffic state surveillance and
Methodological, 28(4), pp.269-287. assignment. Transportation Research Part B:
Daganzo, C.F., 1995. The cell transmission model, Methodological, 45(3), pp.507-533.
part II: network traffic. Transportation Roy, S., Sridhar, B. and Verghese, G.C., 2003,
Research Part B: Methodological, 29(2), June. An aggregate dynamic stochastic model
pp.79-93. for an air traffic system. In Proceedings of the
Menon, P.K., Sweriduk, G.D. and Bilimoria, K.D., 5th Eurocontrol/Federal Aviation Agency Air
2004. New approach for modeling, analysis, Traffic Management Research and
and control of air traffic flow. Journal of Development Seminar, Budapest, Hungary.
guidance, control, and dynamics, 27(5), Sridhar, B., Soni, T., Sheth, K. and Chatterji, G.,
pp.737-744. 2006. Aggregate flow model for air-traffic
Menon, P.K., Sweriduk, G.D., Lam, T., Diaz, G.M. management. Journal of Guidance, Control,
and Bilimoria, K.D., 2006. Computer-aided and Dynamics, 29(4), pp.992-997.
Eulerian air traffic flow modeling and Tokumaru, S., &Hiraishi, K. (2019). Sector
predictive control. Journal of guidance, Identification for a Large Amount of Airspace
control, and dynamics, 29(1), pp.12-19. Traffic Data. IEICE TRANSACTIONS on
Sun, D. and Bayen, A.M., 2008. Multicommodity Fundamentals of Electronics,
Eulerian-Lagrangian large-capacity cell Communications and Computer
transmission model for en route Sciences, 102(5), 755-756.
traffic. Journal of guidance, control, and Smith, S. W., The Scientist and Engineer’s Guide
dynamics, 31(3), pp.616-628. to Digital SignalProcessing, California
Sun, D., Strub, I. S., &Bayen, A. M. (2007). Technical Publishing, San Diego, CA,
Comparison of the performance of four 1997,pp. 277–284.
Eulerian network flow models for strategic air
traffic management. Networks and
Heterogeneous Media, 2(4), 569.

343
The 3rd Int. Conf. on Transport Infrastructure & Sustainable Development (TISDIC 2019). Construction Publishing House. ISBN: 978-604-82-2893-4.

How do lifestyle behaviours affect risky driving

behaviors? a case study of app-based motorcycle taxi
riders in Danang, Vietnam

Vo Nhat Dinh Quang1,*, Le Thanh Quang2, Tran Phuong Anh Thi3,

Nguyen-Phuoc Quy Duy4
1
Civil Engineering Research & Training Group (CERT), University of Science and Technology - The
University of Danang, Danang City, Vietnam, [email protected]
2
Civil Engineering Research & Training Group (CERT), University of Science and Technology - The
University of Danang, Danang City, Vietnam, [email protected]
3
Faculty of Bridge and Road Engineering, University of Science and Technology - The University of
Danang, Danang City, Vietnam, [email protected]
4
Faculty of Bridge and Road Engineering, University of Science and Technology - The University of
Danang, Danang City, Vietnam, [email protected]

ABSTRACT: New technologies have introduced a number of new transport modes. An app-based
motorcycle taxi service which has contributed to a rapid increase in travel demand, is one of them.
However, the emergence of app-based motorcycle taxi riders is considered to contribute directly and
indirectly to the high proportion of road traffic crashes through their risky riding behaviours. The
correlations between unhealthy habits such as smoking and drink riding and other risky behaviours
among motorcyclists have received less concern. The aim of this study, therefore, is to explore the
effect of unhealthy lifestyle on risky riding behaviours among app-based motorcycle taxi riders. An
online survey and field surveys were carried out to obtain information about app-based motorcycle
taxi riders’ socio-demographics, work patterns and risky riding behaviours. More than 600 riders in
Vietnam completed a structured questionnaire. Binary logistic regression modelling showed that
riders who regularly smoked and drunk were more likely to engage in the risky behaviours of
smoking while riding and drink riding (Adj. OR=18.61, p<0.001 and Adj. OR=9.02, p<0.01
respectively). The results also indicated that those who sometimes smoked cigarettes were more
likely not to wear helmets when riding (Adj. OR=2.55, p<0.01). While smoking habits associated
with two risky riding behaviours including smoking while riding and not wearing helmet, most
surveyed risky riding behaviours were associated by drinking habits. Significant associations between
smoke/drink riding and all other risky riding behaviourswere also found in this study. A better
understanding of these relationships will help authorities and ride-hailing firms to improve safe riding
behaviours among the riders.

Keywords: App-based motorcycle taxi, risky riding behaviour, unhealthy habit, smoking, drinking,
ride-hailing.

344
1. INTRODUCTION country since it was first launched in Nov
2014 (Roscher, 2018). The reliability and
Motorcycles account for a high modal share in security of this transport mode has been
middle and low-income countries given its improved thanks to the transparent cost and
affordability and convenience. In 2006, contact information between riders and
motorcycle fleet was 313 million vehicles passengers via a mobile phone application.
comprising 77% from Asia, 14% from This led to the rapid increase in travel
Europe, 5% from Latin America and the rest demand. Hence, the number of GrabBike
4% from other regions (Rogers, 2008). This riders has jumped from 100 in late 2014 to
fleet reached 455 million in 2010 with 79% more than 50,000 riders, with hundreds
from Asia (Nguyen, 2013), and kept joining every day (VOA News, 2018).
increasing to over 500 million globally However, the emergence of app-based
(WHO, 2015). Particularly in Vietnam in motorcycle taxi riders is considered to
2016, approximately 95 percent of vehicle contribute directly and indirectly to a high
fleet was motorcycles which accounted for the proportion of road traffic crashes through
highest rate in South East Asia (WHO, 2015). risky riding behaviours.
Motorcyclists, bicyclists together with For car drivers, the relationships between
pedestrians are considered to be the most unhealthy lifestyle and risky driving
vulnerable road users and contributed to more behaviours or relationships among risky
than half of all road traffic fatalities (WHO, driving behaviours have been explored in
2018) (Shinar, 2012). Every year more than previous studies. Grout et al. (1983) found that
65% of road traffic accidents were results there was an association between smokers
from motorcycles (United Nations, 2018). In andthe non-use of seat belts. Those who
European countries, a report conducted by smoked while driving were also found less
WHO (2004) estimated that riding a used seat belts 3.5 times and had a 2.3 times
motorcycle was more dangerous than driving higher involvement rate in road traffic
a car as ten times per kilometre or nearly 20 accidents than drivers who had not engaged in
times per driven hour. In Australia, the rate of this behaviour (Koushki and Bustan, 2006).
motorcyclist deaths was about 30 times the The association of lifestyle with crash
rate for car occupants per distance travelled involvement has been received much concerns
(Johnston et al., 2008). Higher risky from scholars. DiFranza et al. (1986) found
estimation can be expected in low-income and that compared to non-smokers, smokers were
middle-income countries due to the 50% more likely to be involved in a crash,
heterogeneous traffic, lack of education, and whereas drinkers were nearly 45% more likely
weak enforcement for risky riding behaviours, to have had a crash than non-drinkers
resulting in further traffic danger to motorised (Hutchens et al., 2008). Similar to smoking,
two-wheelers. alcohol consumption was named as one of
While often having a larger engine and are main factors contributing to fatal crashes in a
used for leisure in developed countries, most study conducted by Lin and Kraus (2009).
motorcycles are low-powered and used as an For motorcyclists, there have been a few
essential mean of mobility in developing studies investigating the association between
countries (Rogers, 2008). In Vietnam, riders’ unhealthy lifestyle, particularly drinking
traditional motorcycle taxis (called ‘xe om’) habit, and risky riding behaviours as well as
have been operated as a type of public traffic crash/accident involvement. By using
transport mode which provides a flexible, exploratory factor analysis (EFA) approach,
cheap and door-to-door service (VOA News, more than 40 items related to riders’ lifestyle in
2018). Recently, the development of Serbia were classified into 8 groups including
technology has enabled ‘xe om’ become an sport activity, motorcycle as a hobby,
app-based motorcycle taxi service in the motorcycle addiction, aggression, culture,

345
alcohol and drug use, religiousness, and 2. METHODOLOGY
amusement (Stanojević et al., 2019).
Significant correlations between alcohol 2.1 Data collection
consumption and risky riding behaviours and The three largest cities of Vietnam including
traffic accident involvement were found in this Hanoi, Ho Chi Minh City and Da Nang were
study. Research on motorcyclist’s risky selected to collect data during May and June
behaviours in Vietnam can be found in recent 2018. An online survey was first designed on
studies (Truong et al., 2018, Nguyen-Phuoc et Google platform. A pilot test was carried out
al., 2019). The correlations between mobile with students studying in a faculty of civil
phone use with other risky behaviours such as engineering in Danang city. The survey link
drink riding, wearing helmets or running red- was then shared on Facebook-Groups of app-
lights were investigated by Truong et al. based motorcycle taxi riders in these three
(2018). They found that riders who rode a cities. More than 250 riders participated in the
motorcycle under the influence of alcohol were survey but only 172 respondents completed all
more like to calling and texting while riding. questions. Parallel to the online survey, a
However, data for this study was collected paper-based survey was also conducted to
from students in two university campuses so increase the diversity of the respondent
the results might not generalise for other sample. A number of research assistant groups
groups. The alcohol-involved motorcycle were organised and trained on how to
accidents have also been documented in a administer the survey, in addition to covering
number of studies (Williams and Hoffmann, the aims and objectives of the study before
1979, Peek-Asa and Kraus, 1996), resulting in carrying out field data collection. The
the common findings that drinking drivers questionnaires were distributed at locations
were more likely to perform a number of risky where app-based riders often gather to pick up
riding behaviours including speeding, losing or drop off passengers such as train stations,
control and not wearing a helmet and more airports or shopping areas. The research
frequently be involved in crashes than non- assistants approached the riders and invited
drinking drivers. them to participate in the survey. To prevent
While research related to the association duplicate responses, the research assistants
between motorists’ unhealthy lifestyle and had to check whether respondents had
risky driving behaviours has been undertaken attended the online survey or not. As the
by many scholars, few studies have been survey related to risky riding behaviours and
conducted for motorcyclists. The correlations crash involvement, some app-based riders
between smoking/drink riding and other risky were initially hesitant to attend. They were
behaviours among motorcyclists have also afraid that the results of the survey would
received less concern. Particularly, there has affect their business. Hence, research
assistants had to explain clearly the aim of this
been no study investigated the risky riding
research and emphasise that the survey was
behaviours and their correlations among app-
confidential and that their responses would be
based motorcycle taxi riders. The aim of this
anonymous. A total of more 600
study, therefore, is to answer the research questionnaires were distributed via this
questions of how such unhealthy lifestyle have method, from which 430 valid responses were
effects on risky riding behaviours among app- received, yielding a response rate of 71.7%.
based motorcycle taxi riders. The study also To encourage participation, prepaid mobile
explores the prevalence of risky riding phone cards of 20,000 VND (approximately 1
behaviours and their correlations. A better USD) were given to respondents who
understanding of these relationships will help completed the survey. Finally, a total of 602
authorities and ride-hailing firms to improve valid questionnaires collected from both
safe riding behaviours among the riders. channels were coded and used in the analysis.

346
Table 1. Details of the data collection process 2.1 Data analysis
City Data collection Number of SPSS Statistics 20.0 was used to descript the
method respondents survey data. Using descriptive statistics, the
Hanoi Online survey 20 prevalence of each risky riding behaviour was
Field survey 173 calculated with 95% confidence intervals. A
binary logistic regression model was then used
Da Nang Online survey 32
to investigate potential associations between
Field survey 166 lifestyle behaviour and a set of risky riding
Ho Chi Minh Online survey 120 behaviours among app-based motorcycle taxi
Field survey 91 drivers. Correlations between smoking while
Total 602 riding and drink riding and other risky
behaviours were also explored by logistic
The survey was a part of a wider project regression model. Variables in these models
related to risky riding behaviours and road were checked the correlations. If Person’s
traffic crash/fall involvement among app- correlations showed high collinearity (>0.70),
based taxi riders. This survey contains 50 only one variable of each pair was selected for
questions and it is expected to take the model. These statistical analyses were
participants approximately 15 minutes to conducted with the use of StataSE 12
complete. This paper only focused on statistical software.
questions related to risky riding behaviours,
particularly smoking and drink riding, and 3. RESULTS
factors associated with those.
Riders who participated in the survey were 3.1 Descriptive statistics
firstly asked about their socio-demographics Table 2 provides descriptive statistics about
such as age, gender, home town, place of the sample in terms of rider’s demographics,
residence, marital status, job, education level, working history and lifestyle behaviours. Of
and years of motorcycle license ownership. In the 602 valid survey responses, a high
addition, basic information about the type of majority of riders (95%) were male. The
taxi services (i.e. full-time or part-time), the average age of riders was relatively low at
type of motorcycle used (i.e. manual or 25.7 years old. Most surveyed riders were
automatic), perceived sufficiency of income migrants (over 80%), around half were
and daily travel distance was also collected. students (46.8%), and most were employed as
Information about riders’ unhealthy lifestyle app-based motorcycle taxi drivers on a part-
behaviours towards safety while riding was time basis (79.2%). Most riders used a manual
asked in section three. In the fourth section, motorcycle (83%) in contrast to an automatic
the riders were required to report the motorcycle (17%). More than 30% of
frequency of risk-taking behaviours while respondents reported to work as an app-based
riding a motorcycle (never = 1, seldom (few motorcycle rider more than 50 hours per
times a year) = 2, sometimes (few times a week. Regarding to lifestyle, approximately
month) = 3, often (few times a week) = 4, 60% of surveyed riders were non-smoker.
regularly (few times a day) = 5). These More than 30% of the riders stated that they
behaviours included using mobile phone, sometimes smoked cigarettes while 7.8%
neglecting signal, encroaching into car lanes, indicated that they do so on a regular basis.
running red lights, speeding, carrying more The rider who sometimes drunk were much
than one passenger, smoking while riding, not higher than smokers with approximately 58%.
wearing helmet, overtaking and riding under Only 18 respondents (about 3%) reported that
the influence of alcohol. they drunk every day.

347
Table 2. Characteristics of app-based motorcycle taxi riders in Vietnam
Demographics Working history Riders’ lifestyle behaviours
Variable n % Variable n % Variable n %
Age (years), Employment Smoking
mean 25.74 status status
Gender Part-time 477 79.24 Never 366 60.80
Female 31 5.15 Full-time 125 20.76 Sometimes 189 31.40
Male 571 94.85 Perceived Regularly 47 7.80
sufficiency
Marital status Drinking
of income status
Single 421 69.93
No 186 30.90 Never 235 39.04
Married 181 30.07
Yes 416 69.10 Sometimes 349 57.97
Type of riders
Motorcycle type Regularly 18 2.99
Non-migrant 105 17.44
Automatic 101 16.78
Migrant 497 82.56
Manual 501 83.22
Occupation
Weekly working
Non-student 320 53.16 hours
Student 282 46.84 < 50 hours/week 404 67.11
Education level ≥ 50 hours/week 198 32.89
Above high 107 17.77
school
High school 447 74.25
Less than high 48 7.79
school
Driving
experience 6.95
(years), mean
n: total number of respondent

Figure 1 and Table 3 gave information (20.9%, 95% CI, 17.7 - 24.2%), running red
about the prevalence of risky riding lights (19.4%, 95% CI, 17.6 - 24.3%) and
behaviours among app-based motorcycle taxi carrying more than one passenger (17.3%).
riders by frequency. Using a mobile phone The prevalence of smoking while riding,
while riding was found to be the most riding without a helmet and recklessly
common behaviour with 52% (95% CI, 48.0 - overtaking was lower at 12.6% (95% CI, 10.0
56.0%) of all respondents reporting to have - 15.3%), 11.1% (95% CI, 8.6 - 13.6%), and
engaged in this behaviour. It was followed by 9.1% (95% CI, 6.8 - 11.4%), respectively.
other risky behaviours consisted of neglecting Only 7.6% (95% CI, 5.5 - 9.8%) of riders had
to use turn signals (31.1%, 95% CI, 27.4 - admitted to ride a motorcycle under the
34.8%), encroaching car lanes (24.9%, 95% influence of alcohol, with very few (0.7%)
CI, 21.5 - 28.4%), exceeding speed limits doing so on a regular basis.

348
Figure 1. Frequency of risky riding behaviours among app-based riders in Vietnam

Table 3. Prevalence of risky riding behaviours 3.2 Logistic regression modelling

Prevalence Table 4 presents the results of logistic
Risky riding behaviour
Rate 95% CI regression modelling for risky riding behaviours
classified by socio-demographics, working
Mobile phone use 52.0 48.0 - 56.0
history and lifestyle behaviours among app-
Turn signal neglect 31.1 27.4 - 34.8 based motorcycle taxi riders. The results
Encroach into car lanes 24.9 21.5 - 28.4 showed older riders were less likely to smoke
Exceed speed limit 20.9 17.7 - 24.2 while riding a motorcycle (Adj. OR = 0.92, p <
0.05). Gender had no significant effect on all
Red-light running 19.4 17.6 - 24.3
risky behaviours. Those who got married were
Carry more than one 17.3 14.3 - 20.3 less likely to engage in drink riding behaviour.
passenger Migrant riders were found to be less likely to
Smokewhile riding 12.6 10.0 - 15.3 run red-lights, exceed speed limits and
No helmet 11.1 8.6 - 13.6 recklessly overtake other vehicles. Riders who
were currently students in universities were
Reckless overtaking 9.1 6.8 - 11.4
more likely to be involved in a number of risky
Drink riding 7.6 5.5 - 9.8 riding behaviours particularly overtaking and
CI = Confidence Interval drink riding (Adj. OR = 5.77, p<0.001 and Adj.

349
OR = 6.44, p<0.01 respectively). Riders who while riding and drink riding (Adj. OR=18.61,
perceived their income from their business to be p<0.001 and Adj. OR=9.02, p<0.01
sufficient were less likely to engage in a number respectively). The results also indicated that
of risky behaviours including encroaching into those who sometimes smoked cigarettes were
car lanes (Adj. OR = 0.49, p<0.001), running more likely not to wear helmets when riding
red-lights (Adj. OR = 0.59, p<0.05), speeding (Adj. OR=2.55, p<0.01). While smoking habits
(Adj. OR = 0.52, p<0.01) and riding under the associated with two risky riding behaviours
influence of alcohol (Adj. OR = 0.41, p<0.05).
including smoking while riding and not
Riders working more than 50 hours per week
wearing helmet, most risky riding behaviours
were more likely to engage in most risky riding
behaviours (except smoking while riding, not (excluding smoking riding and overtaking)
wearing helmets and drink driving) with were associated by drinking habits. For
adjusted odds ratios ranging from 1.64 to 5.54. instance, riders who drank alcohol on a regular
It can be seen that riders’ risky lifestyle basis were more than nine times more likely to
associated with a number of risky riding neglect turn signals when making a turn (Adj.
behaviours. It was not surprised that riders who OR=9.64, p<0.001) and nearly five times more
regularly smoked and drunk were more likely likely to run red lights (Adj. OR=4.83, p<0.01)
to engage in the risky behaviours of smoking than those who had never drunk.
Table 4. Logistic regression model results for risky riding behaviours
Carry
Mobile Turn Encroah Exceed Red- Smoke Reckles
more than No Drink
Variable phone signal into car speed light while over-
one helmet riding
use neglect lanes limit running riding taking
passener
0.01*
Intercept 0.08** 0.12*** 0.03*** 0.23 0.29 0.06* 0.09 1.17 0.09
*
Demographics
Age 1.04 1.05 1.02 1.02 0.99 0.99 0.92* 0.89 0.99 1.06
Gender
Female Ref Ref Ref Ref Ref Ref Ref Ref Ref Ref
Male 1.91 2.27 1.57 0.83 0.85 1.86 1.00 1.04 1.00 2.08
Marital status
Single Ref Ref Ref Ref Ref Ref Ref Ref Ref Ref
0.28
Married 1.30 1.53 0.76 0.69 0.73 1.32 0.95 0.67 1.02
*
Type of driver
Non-migrant Ref Ref Ref Ref Ref Ref Ref Ref Ref Ref
Migrant 0.74 0.87 0.71 0.58* 0.52* 0.75 1.20 0.70 0.46* 0.46
Occupation
Non-student Ref Ref Ref Ref Ref Ref Ref Ref Ref Ref
5.77** 6.44
Student 1.26 1.35 2.15* 2.79** 2.13* 1.67 1.14 1.45
* **
Education level
Above high
Ref Ref Ref Ref Ref Ref Ref Ref Ref Ref
school
High school 2.47* 3.13** 1.38 0.80 1.02 1.16 1.23 1.42 0.66 0.56
Less than high
2.43* 2.49* 2.55 0.66 1.61 1.18 0.59 1.80 1.11 0.96
school

350
 Carry
Mobile Turn Encroah Exceed Red- Smoke Reckles
more than No Drink
Variable phone signal into car speed light while over-
one helmet riding
use neglect lanes limit running riding taking
passener
Driving
0.93* 0.96 1.00 0.97 0.97 0.98 1.09 1.01 1.02 0.97
experience
Working history
Employment
status
Part-time Ref Ref Ref Ref Ref Ref Ref Ref Ref Ref
3.53* 3.62
Full-time 1.08 1.02 1.27 0.84 1.58 1.38 2.76* 1.07
* *
Perceived
sufficiency of
income
No Ref Ref Ref Ref Ref Ref Ref Ref Ref Ref
0.49** 0.41
Yes 0.87 1.05 0.52** 0.59* 0.78 0.57 0.73 0.60
* *
Motorcycle type
Automatic Ref Ref Ref Ref Ref Ref Ref Ref Ref Ref
Manual 0.89 0.80 1.82 0.86 1.16 0.72 0.53 0.50* 0.48 0.41
Weekly working
hours
< 50 hours/week Ref Ref Ref Ref Ref Ref Ref Ref Ref Ref
5.54**
≥ 50 hours/week 1.90** 1.64* 1.91* 2.16** 1.88* 3.56*** 1.58 1.67 1.58
*
Riders’ lifestyle
behaviour
Smoking status
Never Ref Ref Ref Ref Ref Ref Ref Ref Ref Ref
19.51 2.55
Sometimes 1.12 1.39 0.59 0.82 0.89 0.95 1.68 1.39
*** **
18.61
Regularly 1.33 1.87 0.91 1.24 1.21 1.12 1.49 2.42 0.92
***
Drinking status
Never Ref Ref Ref Ref Ref Ref Ref Ref Ref Ref
2.41** 2.43** 2.43** 8.15
Sometimes 1.93** 1.83* 2.34** 1.32 1.22 1.09
* * * ***
9.64** 9.56** 5.21* 9.02
Regularly 2.71* 2.91 4.83** 5.87** 2.09 3.15
* * * **
-
- -
Log likelihood -387.59 -340.03 -287.62 -280.70 -271.86 -239.09 -154.62 132.8
156.55 182.29
3
299.6
AIC 809.19 714.06 609.23 595.39 577.72 512.18 345.10 398.57 341.24
6
374.4
BIC 883.99 788.86 684.04 670.20 652.53 586.99 414.66 473.38 410.79
6
*p < 0.05, **p < 0.01, ***p < 0.001, Adj. OR = Adjusted Odds Rati

351
 To establish whether other risky riding turn (Adj. OR=1.950, p<0.001), run red lights
behaviours were associated with smoking while (Adj. OR=1.504, p<0.001), exceed speed limit
riding and drink riding, a number of logistic (Adj. OR=1.331, p<0.05), carry more than one
regression models were developed. Table 5 passenger (Adj. OR=1.510, p<0.01), not wear a
presents the results of these models. It can be helmet (Adj. OR=1.355, p<0.05) and undertake
seen that most of surveyed risky behaviours reckless overtake (Adj. OR=1.421, p<0.05).
were significantly associated with smoking Similar to the risky behaviour of smoking while
while riding except encroaching into car lanes. riding, the results showed that riders who rode a
The results indicated that riders who admitted to motorcycle under the influence of alcohol were
smoke while riding were more likely to use more likely to engage in all of risky riding
mobile phones while riding (Adj. OR=1.611, behaviours with adjusted odds ratios ranging
p<0.001), neglect turn signals when making a from 1.661 to 3.249.
Table 5. Correlations between smoking while riding, drink riding and other risky riding behaviours
Risky riding behaviour Intercept Smoke while riding Drink riding
Mobile phone use Estimate -1.676*** 0.476*** 1.099***
Standard Error 0.398 0.171 0.345
Adj. OR 1.611 3.002
Turn signal neglect Estimate -2.199*** 0.668*** 0.507*
Standard Error 0.280 0.153 0.209
Adj. OR 1.950 1.661
Encroach into car lanes Estimate -2.363*** 0.255 0.695***
Standard Error 0.257 0.133 0.197
Adj. OR 1.290 2.000
Red-light running Estimate -3.169*** 0.408*** 1.043***
Standard Error 0.305 0.139 0.228
Adj. OR 1.504 2.838
Exceed speed limit Estimate -3.063*** 0.286* 1.178***
Standard Error 0.309 0.141 0.241
Adj. OR 1.331 3.249
Carry more than one passenger Estimate -2.969*** 0.412** 0.655***
Standard Error 0.274 0.136 0.196
Adj. OR 1.510 1.925
No helmet Estimate -3.381*** 0.304* 0.720***
Standard Error 0.291 0.150 0.200
Adj. OR 1.355 2.055
Reckless over-taking Estimate -3.883*** 0.351* 0.882***
Standard Error 0.321 0.157 0.209
Adj. OR 1.421 2.417
*p < 0.05, **p < 0.01, ***p < 0.001, Adj. OR = Adjusted Odds Rati

352
3. DISCUSSION AND CONCLUSIONS non-smokers. In addition, the results showed
that riders’ smoking habits had a significant
This study firstly presented the prevalence of association with the non-use of helmets. This
risky riding behaviours among app-based finding was relatively similar to previous
motorcycle taxi riders. The results showed the research regarding car drivers and seat belt use
high prevalence rates for the lower age group (Grout et al., 1983, Eiser et al., 1979). They
(an average age of around 26 years). This is found that there was an association between the
consistent with previous studies which non-use of seat belts and drivers who smoked.
indicated that younger riders were more likely In terms of drinking habits, the results
to engage in unsafe riding behaviours (Truong indicated that riders who have consumed
et al., 2018, Tran et al., 2012). alcohol sometimes or on a regular basis were
While previous studies have explored the more likely to engage in more risky riding
relationship between drivers’ risky lifestyle behaviours than non-drinkers. A study
and crash involvement, there have been limited conducted in south western Nigeria showed
studies examining possible correlations that commercial motorcyclists who consumed
between lifestyle behaviours and risky riding alcohol experienced a higher number of traffic
behaviours among motorcycle riders. This accidents compared to those who did not
paper has investigated the association between consume alcohol (57.6% versus 37.6%)
unhealthy lifestyle behaviours and risky riding (Owoaje et al., 2005).
behaviours among app-based motorcycle taxi This investigation also provides an
riders in Vietnam. The results showed that understanding of the correlations between
unhealthy habits such as smoking and drinking smoking while riding, drink riding and other
alcohol were statistically associated with a risky riding behaviours. The results indicated
number of risky riding behaviours. The riders that riders who had admitted to smoking while
who were currently smokers were found to be riding were more likely to use a mobile phone
more likely to smoke while riding. Smoking while riding, neglect turn signals when making
while driving a vehicle has been associated a turn, run red lights, carry more than one
with traffic crashes because of the reduction of passenger, not wear a helmet and undertake
driving performance and night vision (Koushki reckless overtaking. The results also showed
and Bustan, 2006, Sacks and Nelson, 1994). that there were statistically significant
For motorcyclists, smoking while riding is associations between drink riding and all other
considered to be more serious as riders have to risky riding behaviours surveyed in this study.
ride a motorcycle by one hand. Riding with The riders who indicated that they rode
one hand impinges on the riders’ level of motorcycles under the influence of alcohol were
control of a motorcycle. In addition, smoking nearly twice as likely to use mobile phones
while riding might have negative impacts on while riding. These findings are consistent with
other road users around. For instance, a those of Truong et al. (2018) and Beck et al.
motorcyclist may lose their concentration if (2007) who found that riders who have engaged
cigarette ash flies into their eyes. The positive in drink riding behaviour were more likely to
association between smoking habits of drivers use mobile phones while riding.
and road traffic crashes has also received much Findings from the present study have a
serious attention (Grout et al., 1983, Hutchens number of practical implications. First,
et al., 2008, Bingham and Shope, 2005). Sacks although app-based motorcycles taxi riders
and Nelson (1994), by reviewing more than 90 were managed by ride-hailing firms, the
studies, stated that smokers have an increased prevalence rate of risky riding behaviour is still
risk of crash by approximately 50%. In another high. Increasing safety awareness for the riders
study conducted by Hutchens et al. (2008), through targeted training sessions organised by
current smokers were also found to be more rail-hailing firms could help to reduce the
than twice as likely to experience crashes as frequency of these risky behaviours. Second,

353
the results indicated that riders who have Third, the association of each risky behaviour
admitted to smoking while riding were more with crash involvement should be compared to
likely to engage in more risky riding identify which behaviour has more influence
behaviours and previous studies have revealed than other ones. Appropriate harm prevention
a strong relationship between this behaviour strategies would be proposed based on the
and crash involvement. However, there has seriousness of each risky riding behaviour.
been no police enforcement of this risky Fourth, the results of this study might be
behaviour in Vietnam. An efficient enforcement specific to Vietnam. Transferability of these
for tackling this behaviour should be findings to nations with distinctive road safety
considered by the authorities. Third, drink systems and culture might not be suitable.
riding was found to be associated with all risky Research has shown that differences across the
behaviours in this study. Stronger enforcement road safety system are associated with different
for this risky behaviour has been implemented patterns of risky behaviours across
from 2016 due to its seriousness (LawVietnam, jurisdictions due to cultural, infrastructural, or
2019). While strong enforcement of drink policy factors (Oviedo-Trespalacios and Scott-
riding has been implemented, the enforcement Parker, 2017, Scott-Parker and Oviedo-
for other risky riding behaviours has been Trespalacios, 2017) .
much weaker. For instance, a motorcyclist who In summary, this study has investigated the
engaged in mobile phone use while riding or factors, particularly lifestyle behaviours,
encroaching into car lanes would receive a fine associated with risky riding behaviours among
ranging from 100,000 - 300,000 VND (around app-based motorcycle taxi riders in Vietnam.
4 - 13 USD) while engaging in drink driving The significant association between risky riding
was fined from 1 million - 4 million VND behaviours related to smoking and drinking and
(around 40 - 175 USD). This explained why other risky behaviours was also explored in this
the prevalence rate of drink riding was much study. With the continual growth and expansion
lower in comparison to other risky riding of ride-hailing services in developing countries,
behaviours. Hence, there is a need to review targeted interventions are needed to reduce risky
existing enforcement practices in terms of riding behaviour, particularly smoking while
reducing all risky behaviours. In addition, riding and drink riding, among app-based
interventions at higher levels of the road safety motorcycle taxi drivers.
systems like motorcycle standards and road
infrastructure will also be beneficial (Salmon et 4. REFERENCES
al., 2019). Urgent action is needed to reduce
BECK, K. H., YAN, F. & WANG, M. Q. 2007.
risky driving behaviours. Cell phone users, reported crash risk, unsafe
There are several limitations in this study. driving behaviors and dispositions: A survey
First, this study was based on a self-reported of motorists in Maryland. Journal of Safety
survey so bias can occur in the data collection. Research, 38, 683-688.
As the survey asked about the prevalence of BINGHAM, C. R. & SHOPE, J. T. Adolescent
risky riding behaviours, this information may predictors of traffic crash patterns from
be underreported by respondents who might be licensure into early young adulthood. Annual
concerned that their responses could influence Proceedings/Association for the Advancement
their reputation or businesses. Secondly, the of Automotive Medicine, 2005. Association
present study has not explored the spatial for the Advancement of Automotive
difference in the prevalence of risky riding Medicine, 244-259.
DIFRANZA, J. R., WINTERS, T. H.,
behaviours among app-based motorcycle taxi
GOLDBERG, R. J., CIRILLO, L. &
riders. Further research should compare the BILIOURIS, T. 1986. The relationship of
prevalence of risky behaviours and associated smoking to motor vehicle accidents and traffic
factors among cities with difference social violations. New York State Journal of
demographic or level of traffic congestion. Medicine, 86, 464-467.

354
EISER, J. R., SUTTON, S. R. & WOBER, M. PEEK-ASA, C. & KRAUS, J. F. 1996. Alcohol use,
1979. Smoking, seat-belts, and beliefs about driver, and crash characteristics among injured
health. Addictive Behaviors, 4, 331-338. motorcycle drivers. Journal of Trauma - Injury,
GROUT, P., CLIFF, K., HARMAN, M. & Infection and Critical Care, 41, 989-993.
MACHIN, D. 1983. Cigarette smoking, road ROGERS, N. 2008.Trends in motorcycles fleet
traffic accidents andseat belt usage. Public worldwide. Committee Workshop on
Health, 97, 95-101. Motorcycling Safety.
HUTCHENS, L., SENSERRICK, T. M., ROSCHER, K. 2018. Presenting GrabBike with
JAMIESON, P. E., ROMER, D. & WINSTON, clear opportunities for growth in Vietnam.
F. K. 2008. Teen driver crash risk and Decision Lab, Vietnam.
associations with smoking and drowsy driving. SACKS, J. J. & NELSON, D. E. 1994. Smoking
Accident Analysis and Prevention, 40, 869-876. and injuries: an overview. Preventive
JOHNSTON, P., BROOKS, C. & SAVAGE, H. Medicine, 23, 515-520.
2008. Fatal and serious road crashes involving SALMON, P. M., READ, G. J., BEANLAND, V.,
motorcyclists, Australia THOMPSON, J., FILTNESS, A. J., HULME,
KOUSHKI, P. A. & BUSTAN, M. 2006. Smoking, A., MCCLURE, R. & JOHNSTON, I. 2019.
belt use, and road accidents of youth in Bad behaviour or societal failure? Perceptions
Kuwait. Safety science, 44, 733-746. of the factors contributing to drivers'
LAWVIETNAM 2019. Decree No. 46/2016/ND- engagement in the fatal five driving
CP dated May 26, 2016 of the Government on
behaviours. Applied Ergonomics, 74, 162-171.
administrative penalties for road traffic
SCOTT-PARKER, B. & OVIEDO-
offences and rail transport offences. Vietnam.
TRESPALACIOS, O. 2017. Young driver
LIN, M. R. & KRAUS, J. F. 2009. A review of risk
risky behaviour and predictors of crash risk in
factors and patterns of motorcycle injuries.
Australia, New Zealand and Colombia: Same
Accident Analysis and Prevention, 41, 710-722.
but different? Accident Analysis Prevention,
LU, H.-C., CHANG, C.-L. & HSIEH, J.-C. 2006.
Classification of PM10 distributions in Taiwan. 99, 30-38.
Atmospheric Environment, 40, 1452-1463. SHINAR, D. 2012. Safety and mobility of
NGUYEN-PHUOC, D. Q., PHUONG TRAN, A. vulnerable road users: Pedestrians, bicyclists,
T., DE GRUYTER, C., KIM, I. & NGOC SU, and motorcyclists. Accident Analysis and
D. 2019. Turn signal use when making a turn Prevention, 44, 1-2.
at intersections among car drivers and STANOJEVIĆ, D., STANOJEVIĆ, P.,
motorcyclists – A case study of Danang, JOVANOVIĆ, D., LIPOVAC, K. &
Vietnam. Accident Analysis and Prevention, SECURITY 2019. Impact of riders’ lifestyle
128, 25-31. on their risky behavior and road traffic
NGUYEN, H. H. 2013. A comprehensive review accident risk. Journal of Transportation
of motorcycle safety situation in Asian Safety, 1-19.
Countries. Journal of Society for TRAN, N. T., BACHANI, A. M., PHAM, V. C.,
Transportation and Traffic Studies, 4, 20-29. LUNNEN, J. C., JO, Y., PASSMORE, J.,
OVIEDO-TRESPALACIOS, O. & SCOTT- NGUYEN, P. N. & HYDER, A. A. 2012.
PARKER, B. 2017. Transcultural validation Drinking and driving in Vietnam: public
and reliability of the Spanish version of the knowledge, attitudes, and practices. Traffic
behaviour of young novice drivers scale Injury Prevention, 13, 37-43.
(BYNDS) in a Colombian young driver TRUONG, L. T., NGUYEN, H. T. & DE
population. Transportation Research Part F: GRUYTER, C. 2018. Correlations between
Traffic Psychology Behaviour, 49, 188-204. mobile phone use and other risky behaviours
OWOAJE, E., AMORAN, O., OSEMEIKHAIN, while riding a motorcycle. Accident Analysis
O. & OHNOFERI, O. 2005. Incidence of road Prevention, 118, 125-130.
traffic accidents and pattern of injury among UNITED NATIONS 2018. Road safety
commercial motorcyclists in a rural performance review Viet Nam. New York, the
community in south western Nigeria. Journal United States
of Community Medicine Primary Health Care, VOA NEWS 2018. Uber, others change Vietnam’s
17, 7-12. motorbike culture. VOA News.

355
WHO 2004. World report on road traffic injury
prevention. World Health Organisation.
WHO 2015. Global status report on road safety
2015. World Health Organisation.
WHO 2018. Global status report on road safety
2018. World Health Organization.
WILLIAMS, M. J. & HOFFMANN, E. R. 1979.
Alcohol use and motorcycle accidents.
Accident Analysis and Prevention, 11, 199-207.

356
The 3rd Int. Conf. on Transport Infrastructure & Sustainable Development (TISDIC 2019). Construction Publishing House. ISBN: 978-604-82-2893-4.

Development of safe bicycle infrastructure in Ukraine

Nataliia Kharytonova1,,Valerii Vyrozhemskyi2, Artem Bezuglyi3

1
Head of Ecology and Subgrade Department, M.P. Shulgin State Road Research Institute State
Enterprise - DerzhdorNDI SE, Kyiv, Ukraine
2
Head of Center for Standardization and International Cooperation, M.P. Shulgin State Road
Research Institute State Enterprise - DerzhdorNDI SE, Kyiv, Ukraine
3
Director, M.P. Shulgin State Road Research Institute State Enterprise - DerzhdorNDI SE, Kyiv, Ukraine

Corresponding author, E-mail: [email protected]

Abstract: In addition to health problems, air pollution, noise and vibration, motorization increases the
number of road accidents. Vulnerable groups of road users, namely bicyclists and pedestrians are often in
danger on the roads.
Bicyclists are particularly vulnerable to traffic conditions. High speeds and heavy traffic, as a rule,
interfere with cycling. If road conditions are not suitable for cycling, the measures should be taken to
design the streets more comfortable for bicyclists.
It can be noted that the design of a bicycle infrastructure is difficult taking into account the above criteria;
there is a number of restrictions in conditions of already existing transport infrastructure.
Properly designed bicycle infrastructure will stimulate the urban population to the use of bicycles for
both, the rest and for work trips. An effective bicycle infrastructure will allow: increasing the transport
mobility for all segments of the population regardless of social status; reducing the accident rate
considering that the bicycle transport is less accidental than the automobile one; improving the ecological
situation of cities and settlements.
Keywords: bicyclist1; infrastructure2; designing3; safety4

1. INTRODUCTION politics. Therefore, the bicycle has become

widespread including the northern countries,
The development of bicycle traffic is currently where the climate is not very different from the
being considered in many countries around the climate in Ukraine. The European experience
world as part of economic, environmental, shows that the availability of developed bicycle
social and security policies. In developed traffic infrastructure encourages the people to
countries, the bicycle is part of the transport use bicycles more often. It is especially
complex, and the government, on the regional advantageous for large cities, where a car trip
and state levels, stimulates the introduction of a for a distance of up to 5 kilometers is
bicycle in the daily life of the country accompanied by standing in traffic jams and
(Shelmakov Pyotr, Shelmakov Sergey 2013). consuming a significant amount of fuel. In
In Europe, where there is a severe shortage addition to saving the petroleum products, the
of natural resources, in conditions of high advantage of bicycle traffic is to improve the
population density, where the human impact on health of people who, in particular, in large
the environment is mostly evident, the cities, are having a sedentary lifestyle. Physical
environmental component is important in exercise helps to avoid cardiovascular disease,

357
obesity, diabetes and many other diseases. It car because of poor visibility at intersections,
should be noted that bicycle traffic is much drive around the parked cars and wait for
safer than the automobile one due to lower traffic lights signals. The most unpleasant thing
speed (Evseeva Anastasia, 2017). is not to go by the shortest route, because the
The availability of the rules that determine bicycle is not a quick vehicle, but the necessity
the advantages of a bicyclist to the drivers on to detour which often runs along the highways
the road, the existence of a developed bicycle where bicycle traffic is prohibited.
infrastructure which includes bicycle pathways, Figure 1 shows statistics for 2016-2018 on
safe bicycle crossing, passages and regulated accidents involving cyclists on public roads.
intersections that take into account the bicycle
traffic, lighting and special road signs
minimizes the danger for bicycle traffic on the
roads. The big advantage of using a bicycle is
the almost complete lack of road pavement
wear. A bicycle is a very cheap vehicle. But at
the same time, it can be used also for
commercial purposes, for example, to create a
bicycle traffic system, rental of bicycles and
the use of bicycle by the couriers during the
delivery of small-volume goods because the Figure 1. Data on the accident rate in Ukraine
bicycle capacity can be up to 300 kg. It is involving cyclists for 2016-2018 years
worth noting the creation of bicycle repair
shops and the sale of bicycle accessories as a For a large-scale introduction of bicycle
positive factor for the development of small traffic in our country, changes in the urban
business. In addition, the bicycle takes structure, in traffic rules, in the system of
considerably less space compared to a car, both marking a bicycle path are accepted. It is worth
on the roads and in parking spaces. It gives the considering the priority of a bicycle, as
advantage to use the bicycle in historical compared with the car, on the road. New road
centers of cities, where, especially in Ukraine, signs need to be introduced, since the existing
the housing development is very dense, and two signs which indicate the beginning of the
there is no space for traffic and car parking. bicycle path and its end, clearly do not give a
For Ukraine, the main issues are the lack of complete picture of how to move in certain
regulation of bicycle traffic in the regulatory specific situations. Moreover, they often
documents, the absence of clear paragraphs in provoke a bicyclist to violate the rules. Re-
the Road Traffic Rules, as well as the policy of equipping an intersection with separate traffic
motorization of the population. During the lights for bicyclists should also play an
construction of any facility, starting with important role in the development of a bicycle
schools, hospitals, etc., and ending with the infrastructure. The development of bicycle
restoration of historic buildings, the designs paths and lines due to the redevelopment of
should include not only accessibility for the parking places and the narrowing of the left-
cars, but also for pedestrians and bicyclists. At hand lane, the development of optimal bicycle
this, bicycle traffic should be safe, convenient traffic routes based on the existing bicycle
and beneficial in terms of routes. flow, as well as the survey of the bicyclists’
At present, bicycle traffic is not just opinion using a questionnaire, is the main task
dangerous, it is completely uncomfortable. of promoting a wide use of the bicycle.
Driving to the nearest big store will likely lead Such a laborious and important work, of
to the situation that the Road Traffic Rules are course, requires time. The main thing here is to
violated several dozen times, the bicyclist will solve the issue comprehensively. The
go down and get on the curbs, almost face the development of a bicycle traffic infrastructure

358
is one of the most important issues in our infrastructure should not interfere with other
country; therefore, the main issue was the road users and should not restrict the road and
development of a concept for building the pedestrian traffic.
bicycle infrastructure objects. The bicycle infrastructure should be reliable:
meet the requirements of regulations on riding
2. BASIC PRINCIPLES FOR THE qualities, geometric parameters of the plan,
DEVELOPMENT OF A BICYCLE longitudinal and transverse profiles, and tools
INFRASTRUCTURE of road traffic management and the elements of
The bicycle infrastructure should be safe and the road equipment. Bicycle racks should
convenient for all categories of the population. provide a reliable attachment of the bicycles to
Planning and design solutions for bicycle paths them. Bicycle parking should be a guarantee
should meet the safety and convenience from theft.
requirements of different groups of people The development of bicycle infrastructure
(children from 7 years old without parental should be accompanied by active promotion of
care, people with disabilities, the elderly). bicycle traffic and bicycles as an available,
The network of bicycle traffic paths is environmental friendly, economic and health
holistic and dense. The density of bicycle vehicle for travel.
infrastructure should be in compliance with the
bicycle traffic volume. Bicycle paths should 3. PLANNING THE SAFE NETWORK OF
run along the main passenger traffic directions. BICYCLE TRAFFIC
A bicyclist should safely drive to the The bicycle traffic network must be safe,
destination. On public roads, the network of comfortable and have the most direct routes for
bicycle paths and bicycle parking should all directions.
ensure the connection between the nearest
The bicycle traffic network should take into
settlements. On the streets and roads of
account the requirements of prevailing groups
settlements, the network of bicycle paths and
of road users (children / teenagers, adults,
bicycle parking should provide trips within the
elderly people and recreational bicyclists).
residential or planning area, the connection
between the planning areas. Planning and designing a bicycle traffic
Bicycle infrastructure needs to be network should be interpreted as a system that
integrated into the general transport network incorporates the elements such as bicycle
of the appropriate level (national, regional, infrastructure and service facilities.
district, city), so that it becomes its full The infrastructure that allows direct and
component. Combined traffic must be convenient bicycle traffic in safe traffic
provided: road or public transport and a conditions is the basis of the effective
bicycle. To do this, the possibility to park promotion of bicycle traffic. In addition to
bicycles on the bicycle parking near public bicycle traffic paths, it also includes attractive
transport, subway stations, rail and bus and safe parking for bicycles.
stations, and airports should be provided. Bicycle traffic service facilities can be
During the designing a bicycle provided, for example, by bicycle stops, repair
infrastructure, the principle of autonomy of the service objects, bicycle washing facilities,
road users should be provided: road, bicycle bicycle rental stations.
and pedestrian traffic flows should be When planning, it should be considered all
separated. The bicycle paths should be bicycle-related points of destination and the
physically or visually separated from the objects that can be reached by the bicycle.
sidewalks; they should be better separated from These requirements apply both to the streets and
the highway physically. Elements of bicycle roads of settlements, as well as to public roads.

359
 When planning a bicycle traffic network on 4. IMPROVING THE SAFETY OF THE
public roads passing via settlements, the main BYCICLISTS IN THE CITY OF
routes can be, for example, routes to the VINNYTSIA AS A RESULT OF
school, as well as the routes between IMPLEMENTATION OF A NEW
settlements, their separate parts and important APPROACH TO THE CREATION OF
objects. In the densely populated areas, in HIGH-SPEED INFRASTRUCTURE
addition to the educational and trade routes, the
work routes and evening leisure routes are The city of Vinnytsia was selected to evaluate
more likely to be in the priority. the effectiveness of new cycling organization
The stages of bicycle traffic network schemes. (according to Cycling Development
planning are based on the planning of transport Program in Vinnytsia for 2013-2020).
networks and do not depend on the area of the According to statistics, the number of accidents
planning (national, regional or local bicycle involving cyclists for 2010-2014 in Vinnytsia
traffic network (Yuliana Krause, 2010) was: 20 accidents during 2010, 18 accidents
The main stages of planning the safe bicycle during 2011, 23 accidents during 2012, 15
network are the following: accidents during 2013 and 17 accidents during
а) preliminary analysis with determining: 2014.
1) the level of integration of the bicycle traffic In order to improve the safety of cyclists on
network (for public roads: national, interregional, the streets and roads of Vinnytsia, new
regional, district, local; for the streets and roads schemes for organizing the bycicling traffic
of settlements: intercity, city, district) were proposed.
2) the purpose of the network based on the 4.1 Development of new schemes for
needs of the population: daily traffic, management of bicycle traffic in Vinnytsia
recreational or tourist bicycle traffic;
b) development of bicycle traffic safety The pilot project on the development of bicycle
measures; infrastructure of Ukraine has become the
c) information dissemination, discussion and bicycle traffic management project in the city
decision-making; of Vinnytsia. For this purpose, the following
d) implementation; standards for bicycle paths were determined:
e) monitoring the efficiency and quality of а) standard A1 - bicycle paths placed in one-
the made decisions. way direction on sidewalks. In this case, the
When planning the bicycle traffic network, optimal width of the bicycle path is 2.0 m, the
the following must be ensured: permissible width should not be less than 1.6
а) targeted general management of bicycle m. This width of the bicycle path allows
traffic; moving with the speed limit for the bicyclist
b) Staged implementation, in the combined and safely overtaking without exceeding the
with city districts or settlements on public size of the path, Figure 1;
roads, of bicycle traffic routes, routing access
to recreational areas and tourism objects;
c) the orientation of construction plans for
the development of a continuous bicycle traffic
network, the attachment of the development of
a bicycle traffic network and the corresponding
infrastructure to the construction,
reconstruction, overhaul of public roads, streets Figure 1. Standard А1
and roads of settlements;
d) development of incentives for social b) standard A2 - bicycle paths placed in one-
integration based on the use of bicycle way direction along the edge of the
infrastructure. carriageway of streets and roads. Mandatory is

360
their marking by lines. The optimal width of
such bicycle path is 1.5 m, the permissible
width is 1.25 m, Figure 2;

Figure 5. Standard B2

f) standard C - bicycle lanes are arranged in

Figure 2. Standard A2 the central part of the city on the carriageway,
the bicyclists move along with the road
c) standard A3 - bicycle paths which are vehicles. In this case, the speed of the vehicle
arranged in two-way directions on sidewalks must be limited up to 40 km / h, figure 6;
including by green plantations. In this case, the
optimal width of the bicycle path is 3.0 m, the
permissible width - 2.5 m, Figure 3;

Figure 6. Standard С

g) standard D - passage in residential areas

Figure 3. Standard А3 of the city, the bicyclists move along with road
vehicles. In this case, the speed of vehicles
d) standard B1 - bicycle paths located in
must be limited up to 40 km / h;
one-way direction on sidewalks. In this case,
the optimal width of the bicycle path is 2.0 m, h) standard E - recreation areas (parks,
the permissible width is not less than 1.6 m. forest parks, quays). In this case, the optimal
This standard should be applied on the streets width of the cycle path is 3.0 m, and the
where the use of any standard A is not permissible width is 2.5 m, Figure 7.
possible, Figure 4;

Figure 4. Standard B1 Figure 7. Standard Е

e) Standard B2 - bicycle paths are arranged in
a one-way direction along the sidewalk and In the city of Vinnytsia, the schemes for
giving the status of a pedestrian-bicycle path. In traffic management were developed according
this case, the optimal width of the bicycle path to the standard A - 10 pcs; B - 7 pcs; C - 5 pcs;
is 2.0 m, and the permissible width is 1.6 m, D - 30 pcs. This made it possible to evaluate
Figure 5; the effectiveness of the schemes developed.

361
4.2 Experience in organizing the cycling On controlled and uncontrolled
traffic at crossroads intersections, a bicyclist must dismount a
bicycle and pass the road following the Traffic
4.2.1 Type intersections Rules (to ensure their own safety and to pass
the carriageway on the pedestrian crossings),
The bicycle path runs parallel or along with the
Figure 10-11.
roadway, bicyclists have a "green zone" at the
intersections of 2 m in length, in the rightmost
lane or in front of all lanes, in order to leave
the intersection first. The bicycle traffic is
managed by the main or additional traffic
lights. The area in front of cars at the
intersections is called a bicycle box and is
usually colored differently from the color of
the bicycle path and the main road surface. An
example of the equipment of bicycle paths at
intersections, as bicycle box, taking into Figure 10. One-way bicycle path on the sidewalk -
account the European experience, is shown in passing by bicycle through the intersections
and exits from secondary streets
Figure 8-9.

Figure 11. Crossing the controlled and

uncontrolled intersections in
parallel with pedestrians

4.2.2 Equipment of bicycle paths at circular

intersections
Figure 8. Management of a intersections by giving the The movement on a circular intersections is
priority to bicyclists on the rightmost lane carried out by providing the standard priority
of the movement: the bicyclist, while moving
in a circle, remains in his area. The cars that
exit the intersection, should give priority to a
bicyclist moving in his lane, Figure 12-13
The rules for choosing a particular design
solution depend on the type of a bicycle route,
the spatial context (built or unbuilt area), the
traffic speed and volume of road vehicles.

Figure 9. The intersections with the priority of a

bicyclist on each traffic lane

Movement on controlled or uncontrolled

intersections should be carried out on an
existing zebra crossing or alongside it on a
special marking. Figure 12. Scheme of a circular intersetion

362
 development of the projects of overhaul,
reconstruction and new construction of public
roads, streets and roads of settlements on the
basis of the corresponding feasibility study.
Construction of the bicycle network is carried
out by arranging the separate bicycle paths and
bicycle paths on the pedestrian sections,
equipment of bicycle lanes. The equipment of
bicycle infrastructure involves designating the
bicycle routes, arranging the bicycle parking
lots, organizing the bicycle rentals, developing
Figure 13. Photo of circular intersetion the consulting and information points, etc.
In Vinnytsia, traffic management schemes The technical parameters of each bicycle
were designed for about 20 objects, including path or a bicycle lane (width, pavement, color,
traffic management schemes at the intersections, marking) are determined by the design
junctions and pedestrian crossings, as well as management individually for each street or
about 15 new cycling traffic management road taking into account the local conditions
schemes at circular crossroads. and requirements of the state standards.
There are a number of recommendations for
4.3 The effectiveness of the introduction of the development of the bicycle infrastructure
new traffic management schemes elements that do not contradict the current
standards but significantly improve the
As a result of the introduction of new cycling
efficiency and safety of bicycle infrastructure.
traffic management schemes, the number of
The main commonly accepted tools are the
accidents with cyclists has decreased by a total
following:
of about 25%; at that, the number of accidents
а) unambiguous and clear arrangement of
in residential areas decreased by 20%, in
the street space for bicycle traffic to all road
circular interchanges - by 30%, at crossroads -
users by marking and signing, the presence of a
by 30%. In the future, monitoring of the
prohibition for stopping the road vehicles;
effectiveness of the proposed traffic
b) zero level of kerbs at intersections;
management schemes is foreseen.
c) priority for bicyclists to ride first in front
of road vehicles;
5. FEATURES OF THE CREATION OF
d) ensuring the visibility for bicycle traffic
THE CYCLING INFRASTRUCTURE
and road traffic on both sides;
IN CITIES AND SETTLEMENTS
e) rainwater drainage;
Based on the analysis of the results of the f) providing the snow / dirt treatment.
introduction of new, effective bicycle traffic State building norms for the design of streets
MANAGEMENT schemes in Vinnytsia, an and settlements DBN V 2.3-5 contains
algorithm for creating a safe bicycle requirements regarding the width of the main
infrastructure in settlements of Ukraine was elements of the bicycle network of streets and
developed. roads of settlements (Table 1) and public roads
The engineering part of the project of the (Table 2).
bicycle infrastructure development is being During planning the bicycle traffic network,
implemented by the simultaneous development it is necessary to increase the connections
of design, construction of a bicycle network between the central destination points taking
and equipment of bicycle infrastructure. into account the intrinsic-territorial interests of
The design of the bicycle network and the the community since they determine the
elements forming it is carried out during the purposeful daily bicycle traffic.

363
Table 1. Minimum width of bicycle lanes and - In the built areas, bicyclists and
paths mechanical vehicles should be separated if
possible but they can be mixed while taking
Minimum width, m measures to prevent conflicts.
Type of bicycle In terms of In the unbuilt areas it is recommended to use
traffic management In terms of
new separate bicycle traffic paths on roads with
reconstruction
construction speeds above 80 km/h (Dirk Dufour, 2010). On
Bicycle lane 1,85 1,5 roads of local importance where vehicles move
at speeds up to 60 km / h, mixed use of the traffic
Bicycle path with
1,85 1,5 lane or arrangement of bicycle paths that can
one-way traffic
satisfy the traffic safety requirements is allowed.
Bicycle path with In the unbuilt areas, bicycle paths are
two-way traffic on recommended to arrange along the high-speed
2,5 2,0
both sides of the roads (speeds over 80 km/h). They are also
street recommended to arrange on roads with a lower
Bicycle path with speed (60-80 km / h) but with high traffic
two-way traffic on 3,0 2,5 volume (more than 2000 car / day).
one side of the street In the built area, it is not always possible to
A common bicycle separate the various road users and avoid
3,0 2,5 conflicts. Therefore, the most important in built
and pedestrian path
urban areas is the prevention of conflicts.
Table 2. Recommended bicycle path width Streets and bicycle routes should be designed
in such a way that all road users are visually
Type of bicycle warned on a potential conflict situation
Width of the bicycle path, m
path between different types of vehicles. In practice,
In the this means that bicycle traffic must be mixed
In the restricted-
restricted with the road traffic flow, where possible, and
free conditions
conditions must be separate where it is necessary.
Bicycle path: In the built area, the most connections in the
basic bicycle network must pass through streets
- one-way traffic 2,0 1,0 with a speed limit up to 30 km / h. On more
- two-way traffic 2,5 2,0 loaded streets and roads, as well as at difficult
Note. The restricted conditions include arranging intersections, especially with high road traffic
the bicycle paths in construction areas, speed and volume, it is necessary to separate
availability of capital building facilities or the bicycle traffic. Requirements for choosing
complicated terrain. the form of bicycle traffic management
depending on the category of the street are
For inter-regional connection at the distance given in table 3 (DBN B.2.3-5: 2018).
more than 10 kilometers for everyday bicycle Design of the plan, longitudinal and
traffic, high-speed special networks should be transverse sections of bicycle paths is carried out
provided. according to DBN B.2.3.4 (DBN B.2.3-5: 2018).
There is a significant difference in the The required list of elements of improvement
approaches to the design in the built and of bicycle paths consists of the following: type
unbuilt area, as well as depending on the of pavement, elements of a surface of a bicycle
functional purpose of the route. The following path with an adjacent territory.
main points can be highlight: Lighting should be provided on bicycle
- In the unbuilt areas it is always necessary paths located along the streets and roads, in
to separate bicyclists and mechanical vehicles recreational areas - green planting along
to prevent conflicts; bicycle paths.

364
Table 3. Choosing the form of bicycle traffic 6. CONCLUSIONS
management depending on the category of the
street and roads As a result of the implementation of the concept
of the development of bicycle infrastructure:
Form of bicycle traffic а) the number of road accidents involving
management bicyclists will decrease;
Bicyc Mixed б) the quality of transport services for the
Category of the street Bicyc le moveme
Bicycl population and its mobility for the movement
le pedest nt on
e path
lane rian carriage in public roads, streets and roads of the
lane way settlements will increase;
Arterial road + в) labour productivity will increase by
attracting a large part of the population, in
Arterial Continuou
streets of s traffic
+ + particular, young people to a healthy lifestyle
general and preventing cardiovascular diseases;
local controlled г) the total amount of emissions of harmful
+ + +
importance traffic substances into the atmosphere will decrease,
Arterial especially in the streets and roads of
streets of Controlled
+ + settlements;
district traffic д) the traffic capacity of streets and roads of
importance
settlements will increase.
Residentia
+ + +
l streets
8. REFERENCES
Streets
and roads DBN V.2.3-5: 2018 Streets and roads of settlements.
in DBN V.2.3-4: 2015 Highways. Part I. Design. Part II
scientific- Construction.
Streets and industrial,
roads of industrial + + + Dirk Dufour, 2010. Handbook of bicycling policy
local PRESTO. Infrastructure. Ukrainian version /
and
importance municipal- Dirk Dufour, Ligtermoet&Partners, The
warehouse Netherlands, 52 p.
areas Evseeva Anastasia, 2017 Monitoring of bicycle
traffic in a city / Electronic bulletin. Issue № 64.
pedestria
+ + Shelmakov S., Shelmakov P. 2013. Methodology
n streets
for evaluating the effectiveness of the bicycle
passages + route / International Research Journal, №9
(16) p. 131-134.
Green plantations along the bicycle paths The program of bicycle traffic development in
should not reduce their dimensions, the height Vinnytsia for 2013-2020. - 2013. - 43 p.
of the free space above the level of the bicycle Yuliana Krause. Principles of bicycle network
path should not be less than 2.5 m. planning (Publishing House «VDE HmbH»,
Pavement of bicycle paths is designed in Berlin Bracher, Haag, Holzapfel, Kiepe,
accordance with the requirements for pavement Lehmbrock, Reutter VDE Verlag GmbH
of light and transition type in accordance with Berlin Offenbach 60. Ergänzungslieferung
(DBN V.2.3-4: 2015). The pavement design of 04/11). - 18 p.
bicycle paths in settlements should ensure the
passage of treatment equipment.

365
The 3rd Int. Conf. on Transport Infrastructure & Sustainable Development (TISDIC 2019). Construction Publishing House. ISBN: 978-604-82-2893-4.

Pedestrian crossing behavior at signalized intersections in

Ho Chi Minh City

Doan Duc1, Nguyen Canh2, Nguyen Son3, Thai Dang4, Nhat Truong5
1,2,3,4,5
Department of Transport Planning, Faculty of Transportation Engineering, Ho Chi Minh City,
Vietnam

Corresponding author, E-mail: [email protected]

ABSTRACT: The understanding of pedestrian crossing behavior under mixed traffic conditions is
critical. In order to provide optimal service for infrastructure design, which is safe, convenient, fast and
more comfortable for pedestrians at signalized intersections. This article attempts to analyze the
pedestrian crossing behavior of 503samples at two signalized intersections in the Ho Chi Minh City. The
pedestrian crossing speed and other factors were observed and verified by statistical models. The initial
results show that the variation in pedestrian speed, such as: age groups, type of moving and the impact of
other vehicles, especially motorcycle. This studycould help researchers for proposing pedestrian safety
solutions, pedestrian behavior model at signalized intersectionsin motorcycle dominated environment.
Keywords: Pedestrian’s characteristics; pedestrian’s behaviors; signalized intersections; mixed traffic.

1. INTRODUCTION walking conditions and safety for pedestrians

like Ho Chi Minh City. Traffic lights designed
Ho Chi Minh City, a motorcycle-dependent
to eliminate conflicts between pedestrians and
city with 400 motorcycles per 1000 people, is
vehicles at the crosswalks, but traffic conflicts
facing many traffic challenges such as
often occur, due to intentionally non-
congestion, traffic accidents and environmental
compliance of pedestrian to signal lights. In
pollution. While researchers and authorities are
fact, there are many reasons for non-
focusing on finding ways to solve the
consequences rapidly the development of compliance behaviors of pedestrians such as
private vehicles [1]. The pedestrian problems degraded infrastructure, low quality traffic
have not really been concerned, especially at management, mixed traffic conditions and
the signalizedintersections, where there are length of cycle time. In addition, other factors
many potential risks of causing injury to that influences pedestrian compliance to
pedestrians. One of the main reasons is due to signalized intersections as well as the
the complexity of analyzing pedestrian interaction between pedestrians and vehicles
behaviors. This complexity arises from the are not mentioned in current studies at Ho Chi
many factors that affects pedestrian behavior Minh City.
and is difficult to identify. At signalized Pedestrian crossing speed is one of the
interactions, pedestrians tend to increase very important parameters for proposing pedestrian
quickly, especially in large cities and crowded safety measures at signalized intersections.
pedestrian areas without measures to improve Many of the existing studies have examined

366
the pedestrian crossing behavior at signalized District 1, where there are crowded
intersections in the developed country [2], [3], concentrations of pedestrian, regularly crossing
[4], [5], [6], [7]. Few studies have analyzed roads to move in/out of commercial areas. The
pedestrian behaviors in mixed traffic cameras were set up on the high buildings at
conditions with dominated motorcycles the walkway locations to collect data at
environment like Ho Chi Minh City. This intersections during the lunch peak hour, the
article focuses on identifying pedestrian most crowded pedestrian time in the study
characteristics and behavioral parameters in area. Determining the speed of pedestrians in
mixed traffic conditions for suggesting this paper is analysed through SEV software
pedestrian safety solutions and pedestrian [8]. SEV software helps to determine
behavior model. This study is divided into 4 pedestrian coordinates, speed and direction in
each frame position. The intersection
parts. After the first part - introduction is the
geometric data, walking dimension and
methodology. Next is the result and discussion
pedestrian data extracted from the camera will
and the last part is the conclusion.
be the input of the SEV software. After
declaring the input, a coordinate system
2. METHODOLOGY
corresponding to the location of fixed points
Traffic data collected by camera has been which is measured outside reality is very
widely applied in Ho Chi Minh city, to ensure important, to avoid errors in the analysis
data quality. The pedestrian data in this paper process. After the coordinates of each position
was collected at two signalized intersections in on the pedestrian's path are determined from
Ho Chi Minh city: (i) Le Lai-Nguyen Thi the actual coordinate system, their speed in
Nghia intersection; (ii) Phu Dong 6-leg each frame will also be determined easily. The
intersection. Nguyen Thi Nghia, Le Lai and whole process of data analysis from input and
Nguyen Thai Hoc street are the main routes in output is described in Figure 1.

Figure 1. Pedestrian’s crossing speed analysis process

367
Table 1. Total survey samples and geometric characteristics
Items Phu Dong 6-leg intersection Le Lai – Nguyen Thi Nghia
Survey time 11:30 - 12:30 11:30 - 12:30
Length  width of the walking line (m) 312 327
Number of survey samples [per /peak hour] 150 353
Direction 2 2
Average speed (m/s) 1.482 1.369
Standard deviation [SD] 0.219 0.126

Table 2. Pedestrians classified by their characteristics and behaviors

Observed parameters No. of observed pedestrians (%)
Characteristics Gender Male 326 64.8
Female 177 35.2
Age Group Child 7 1.4
Adult 482 95.8
Old 14 2.8
Crossing speed (m/s) < 1.0 6 1.19
1.0 - 1.2 21 4.17
1.2 - 1.4 237 47.12
1.4 - 1.6 192 38.17
> 1.6 47 9.34
Behaviors Crossing type Run 21 4.2
Walk 482 95.8
No. of pedestrian Group 280 55.7
Single 223 44.3
Crosswalk using Yes 365 72.6
No 138 27.4
Compliance Yes 458 91.1
No 45 8.9

All of survey samples and geometric 3. RESULTS AND DISCUSSION

characteristics at the intersection are indicated
in Table 1. Data from video provide the 3.1 Pedestrian crossing behavior
following information: pedestrian traffic, The speed data collected from the two
crossing time, appearance of pedestrians
intersections in HCMC has shown a big
(gender, age), crossing behaviors (running or
difference from the value of 1.2m/s of the US
walking, walking alone or walking in groups
and changing the speed when crossing the [2]. Pedestrians using the crosswalk during
road), location when crossing the road pedestrian green phase were considered as
(whether to use the walking line or not), signal compliant pedestrians while those who used
light (whether to comply with the signal light crosswalk during non-green phase were
or not) and the interaction between pedestrians considered as incompliant pedestrian. The
and motorized vehicles. The results of percentage of pedestrian compliance was
behaviors and characteristics of pedestrians are defined by the ratio between the number of
detailed in the next section. pedestrians using the crossing during green

368
light phase and the total number of pedestrians intersection. Total 503 pedestrians were
observed. Total 503 pedestrians were observed considered, ANOVA test was used to
directly from the video collected at the determine the factors of linear (single
intersections to find out detail information variable) affected crossing speed and
about their crossing behavior. Pedestrian multivariate regression was used to determine
characteristics and behaviors information are the degree of influence of variables to
shown in Table 2. crossing speed in traffic conditions with many
The results show that the number of male interactions with each other.ANOVA testing
pedestrians is higher than the number of was performed through SPSS software
female pedestrians during peak hours and the version 16. Gender, age, group, type of
frequency of adults is 95.8% compared to the approaching vehicle, signal compliance
elderly and children. Data indicate that behavior and pedestrian priority are factors
pedestrians prefer walking rather than running see the test for the test based on the
with 47.12% of pedestrians crossing the road theoretical basis. The test shows at the 95%
at 1.2-1.4m/s. The average speed found is confidence interval and the results are
1.39m/s. Only 72.6% of pedestrians use presented in Table 3.
walkways when crossing roads. The As expected, the above variables are
percentage of pedestrians complying with statistical significant. Male pedestrians often
traffic signals was 91%, there were nearly 50 have higher speed than female pedestrians
incompliant cases, and found that the with the average speed of 1.42m/s and
incompliance tends to increase, related to the 1.35m/s respectively. Realizing that the age
number of people walking in peak hours, 16% group has a relatively large influence with a
regression coefficient of 0.292. Children have
of incompliance at Phu Dong 6-leg
the highest speed compared to adults and the
intersection and 6% at Le Lai - Nguyen Thi
elder, the average speed of the elder is lower
Nghia intersection. than the average speed of adults, respectively:
3.2 Pedestrian crossing behaviors children (1.75m/s), adults (1.39m/s) and the
elder (1.21m/s). As expected, the group's
This analysis focuses on the change in speed crossing speed is lower than the crossing
of pedestrians during the crossing, the speed of single pedestrian, with regression
objective and subjective factors affecting coefficients of only 0.118 as follows: group
pedestrian crossing speed at signalized (1.37m/s), single walk (1.41m/s).

Table 3. Results of ANOVA test

ANOVA test

Factor Fcritical Ftable Pvalue Significant

Gender 15.393 3.842 0.000 Yes

Age 19.098 3.842 0.000 Yes

Group 4.442 3.842 0.036 Yes

Type of approaching vehicle 35.079 3.842 0.000 Yes

Signal compliance behavior 14.901 3.842 0.000 Yes

Pedestrian priority 0.697 3.842 0.404 No

369
 The cause identified is walkers who usually is faster and the safety of pedestrians will be
slows down to keep a safe distance from other ensured. No need to discuss anymore, the
people in the same group or other groups in the pedestrians who do not comply with the signal
opposite direction and speed of the group often light will face a direct conflict from the left-
depends on the speed of the leader in the side means, which causes them to lower speed
group. As expected, types of approaching mean to ensure safety, the average speed in both
are the main factor that changes the pedestrian
cases is respectively (1.4m/s) and (1.28m/s).
speed, with both critical coefficient Fcritical =
35.079 and regression = -0.421 is the largest 3.3 Pedestrian crossing compliance
and the most influent to the speed model. The
average speed for the respective access vehicle During the green light phase, there is still a
groups is as follows: motorcycles (1.36 m/s), case of pedestrians walking along the
cars (1.26m/s) and trucks (1.17m/s). It is easy crosswalk to cross the intersection. Notice that
to realize that, when pedestrians encounter this incompliant behavior affected by many
conflicts with bigger vehicles, the speed of different factors. This incompliant rate is 16%
crossing the road will be further reduced or we and 6% corresponding to the 2 intersections
can say that “the bigger vehicle approach the mentioned in the report. Pedestrian compliance
slower the pedestrians will move” in order to is analyzed through Pearson linear regression
keep a safe distance to avoid collision with the coefficients, ANOVA variance test and
vehicle. When a conflict between a pedestrian
verified two independent groups to assess
and a vehicle occurs, if the driver actively
factors affecting compliance with signal lights
reduces the speed to give priority to the
pedestrian, then the pedestrian will respond by of pedestrians. Gender, age, group and crossing
accelerating get out of conflict with vehicle. speed are factors considered for the test. All of
Realizing that in cases having pedestrian the tests are conducted with SPSS with 95%
priority, the crossing speed is always higher confidence intervals. The analytical results are
than the rest, the ability to exit the intersection presented in Table 4.

Table 4. Results of statistical test of factor

Factors Pearson correlations ANOVA Remarks
Coefficient P value F value Sig
Gender 0.114 0.010 6.630 0.010 Significant
Age -0.124 0.005 5.740 0.003 Significant
Group -0.183 0.000 17.355 0.000 Significant
Direction 0.274 0.000 40.757 0.000 Significant
Multivariate regression analysis
Model of non- Not standardized Standardized
compliant signal Characteristics coefficient coefficient OR t Sig.
lights B Std. Error Beta
Constant .835 .078 10.678 .000
Gender Male vs Female .063 .025 .105 2.511 2.486 .013
Old vs Child 0.167
Age -.139 .059 -.099 -2.342 .020
Adult vs Child 0.203
Group Group vs Single -.076 .025 -.132 0.290 -3.074 .002
Direction 1 vs
Direction .141 .025 .244 7.699 5.657 .000
Direction 2

370
 From Table 4, we can see that gender, age of pedestrians in direction 1 who are men and
group, group size and walking direction are 55% of pedestrians walking in this direction
factors that affects pedestrian signal are single pedestrians, all 7 violations of
compliance. From the results of ANOVA children are located in this direction.
analysis, it shown that gender is the factor that In addition to the all mentioned factors,
influences the signal compliant behavior. The length of cycle lights and the non-green light
OR (Odd Ratio) ratio in male and female phase for pedestrians are also factors that
incompliance in models is 2.511, Pearson influence the compliance. The number of
coefficient and regression coefficient were both incompliant pedestrians found to be directly
positive. This means that the risk of appearing proportional to length of cycle lights. The main
an incompliant male pedestrian is 2.511 times reason for this is pedestrians want to cut down
higher than female and when the number of on waiting time at intersections in adverse
male pedestrian increases, the likelihood of not weather conditions.
complying with the light phase increases. The
age group is also an influential factor, it is easy 3.4 Interaction between pedestrian and vehicle
to see that the elder tends to violate less than During the green light phase, conflicts between
children with OR ratio < 1, adults also have a pedestrians and vehicles may occur which due
lower rate of violation than children with OR to driver's incompliant behavior and vice versa,
ratio = 0.203. Pearson coefficient and conflicts may occur during the non-green light
regression coefficient were negative this means phase due to the incompliance of pedestrians.
the higher the number of children and adults, The interaction between pedestrians and
the higher the chance of incompliances with vehicles is analyzed based on the Pearson
the signal lights will increase. Group size is the coefficient and the ANOVA test to examine the
second most influential factor in the factors affecting pedestrian compliance.
incompliant behavior model, pedestrian in Statistical results are presented in Table 5.
groups are often depended on the crowd effect, From Table 5 and the collision model, we see
so the likelihood of a violation will be less than gender, group size, direction of approaching
a single pedestrian. The coefficient of 0.290 vehicle, type of approaching vehicle and signal
describes that the group pedestrian's ability to compliant behavior are factors that affects the
violate a lot less than a single pedestrian. When collision between pedestrians and vehicle.
the number of single pedestrian increases, the Pedestrians who refuses to comply with signal
likelihood of a violation will increase light are 1.12 times higher to have collision than
corresponding to the Pearson coefficient and compliant ones. Negative Pearson coefficients
regression coefficients were negative. show that the higher the number of pedestrians
From models, it is easy to see that the who refuses to comply with the signal light, the
walking direction is the most influential factor higher the chance a collision with vehicle will
in the pedestrian incompliance model, statistics occur corresponding to 100% of incompliant
show that up to 38 out of 45 cases of cases. Gender also is statistical significant, we
incompliance with signal lights came from can realize that men are 1.185 times higher in
direction 1 according to the camera's view. The experiencing collision, which indicates that
Odd Ratio in the incompliance model is 7.699, 64.8% of pedestrians at intersections are male
the Pearson coefficient and the regression and the proportion of signal lights incompliance
coefficient are both positive. This means that of men is 2,511 times higher. The group size
the likelihood of a pedestrian compliance with demonstrates the level of statistical significance
the signal light in direction 1 was 7.699 times in the collision model with a significance of
higher in direction 2 and if the number of 0.035 < 0.05, indicating that the pedestrian in
pedestrians in direction 1 is larger, the group is likely to have a higher likelihood of
likelihood of incompliance will be higher. To collision than single pedestrians 1,013 times,
explain this, according to statistics, there are 64% not too high.

371
Table 5. Results of statistical test of factor
Factors Pearson correlations ANOVA Remarks
Coefficient P value F value Sig Significant
Gender -0.261 0.000 14.148 0.000 Yes
Age 0.000 1.000 0.272 0.762 No
Group -0.021 0.766 0.088 0.766 No
Direction of 0.187 0.009 7.030 0.009 Yes
approaching vehicle
Type of approaching -0.228 0.001 7.876 0.001 Yes
vehicle
Signal compliance -0.163 0.023 5.279 0.023 Yes
behavior
Multivariate regression analysis
Not standardized Standardized
CollisionModel Characteristics coefficient coefficient OR t Sig.
B Std Error B
Constant 1.357 .157 8.662 .000
Gender Male vs Female -.132 .040 -.227 1.185 -3.275 .001
Old vs Child
Age -.116 .109 -.074 -1.066 .288
Adult vs Child
Group Group vs Single -.084 .040 -.153 1.013 -2.119 .035
Direction of
Through vs
approaching .039 .020 .139 0.890 1.965 .050
Turing
vehicle
Motorcycle vs
Type of Car 1.05
approaching -.086 .029 -.197 -2.913 .004
vehicle Motorcycle vs 1.438
Truck
Signal
Non-compliance
compliance -.087 .048 -.134 1.12 -1.815 .071
vs Compliance
behavior

The direction of approaching vehicle is highest compared to cars and trucks

statistical significant which impacts on the corresponding to more than 50% of the total is
collision between pedestrians and vehicles and motorcycle’s collisions. Pearson coefficient is -
the relatively high impact levels in the collision 0.228, which indicates a strong linear
model with high regression coefficient is correlation between motorcycles and collisions
0.139. The risk factor shows that in these two at both intersections. The minus sign shows
intersections the go straight collision rate is negative correlation, when the number of
0.890 times lower than turns, with over 65% of motorcycles on the road increases, the number
collisions caused by turns. The type of of collisions also increases. Realizing that
approaching vehicle has very high regression many cases of collisions occurred as
coefficients in the collision model is -0.197, pedestrians have not exit the intersection yet by
the collision chance of motorcycles is the the green light ending time. Large vehicles

372
turning heads at the walkway at the surveyed of turning vehicles at intersections or providing
intersection of Le Lai - Nguyen Thi Nghia pedestrian overpasses to reduce conflicts with
increases the exit time of the vehicles behind is vehicles. This paper can be further developed by
also the factors affecting conflict between increasing the number of survey samples to
pedestrians and vehicles. achieve better statistical results. Another factors
affecting pedestrian crossing such as
5. CONCLUSION communication and using mobile phone will be
explored in the further study.
In this paper, 503 crossing behaviors of
pedestriansat signalized intersections on the 4. REFERENCES
main roads of District 1 were analyzed. Basing
on the statistical test and direct observation Manfred Boltze& Vu Anh Tuan (2016),
through video, factors influencing the change “Approaches to achieve sustainability in
crossing speed of pedestrians, compliant Traffic Management”, Sustainable
behavior of pedestrians as well as the interaction Development of Civil, Urban and
between pedestrians and vehicles were Transportation Engineering Conference,
Procedia Engineering 142, pp. 205 – 212.
identified. The major findings of this paper are:
Highway Capacity Manual (2000)
pedestrian crossing speed of a male is faster than Marisamynathan&Vedagiri Perumal (2014) “Study
a female and single pedestrians are faster than on pedestrian crossing behavior at signalized
pedestrians in groups, the adult pedestrians are intersections”, Journal of Traffic and
faster than the elder pedestrians. The type of Transportation Engineering, 1(2): 103-110.
approaching vehicle and priority for pedestrians Mohamed Hussein & Tarek Sayed (2015),
also has an impact on their speed. Men have “Microscopic Pedestrian Interaction Behavior
higher rate of incompliance than women, a Analysis Using Gait Parameters”,
group has higher compliant rate than a single Transportation Research Record Journal of the
pedestrian due to the orientation effect. Men Transportation Research Board, No. 2519, pp.
have higher risk of collisions than women 28-38.
originate from the quantity of male at Ying Ni, Yingying Cao, Keping Li (2016),
“Pedestrians' Safety Perception at Signalized
intersections and their incompliant behavior.
Intersections in Shanghai”, WCTR 2016
The larger the group size, the higher the Shanghai.
probability of a collision.Turning directions of W.L.Wang, S.M.Lo, S.B.Liu, H.Kuang (2014),
vehicles with more motorcycle traffic will “Microscopic modeling of pedestrian
increase the number of collisions. Pedestrians movement behavior - Interacting with visual
who do not comply with the signal light attractors in the environment”, Transportation
correlates positively with the possibility of Research Part C, 44, 21-33.
conflict.This paper provides the knowledge of Wael K.M., Alhajyaseen& Miho Iryo-Asano,
pedestrian’s speed variation analysis under “Modeling pedestrian crossing speed profiles
mixed traffic conditions, to create initial considering speed change behavior for the
conditions for the development of a delay model safety assessment of signalized intersections”,
and model of the level of service of pedestrian at Safety Science, 91, 351 -360.
Chu Công Minh (2011), “Phát triển phần mềm tính
signalized intersections. The interaction model
toán các đại lượng chuyển động của xe nhằm
between pedestrians and vehicles and the xây dựng các ứng xử, trương tác khi xe khi
influencing factors analyzed in this article can lưu thông”, Đề tài NCKH cấp Trường, ĐH
provide some ideas to improve pedestrian safety Bách Khoa Tp.HCM. (Vietnamese version)
at signalized intersections, such as reducing
motorcycle traffic and measures to limit the flow

373
The 3rd Int. Conf. on Transport Infrastructure & Sustainable Development (TISDIC 2019). Construction Publishing House. ISBN: 978-604-82-2893-4.

Observation of traffic conflicts using the traffic conflict

technique: A case study in Ho Chi Minh City

Doan Duc1,, Pham Kha2

1,2
Department of Transport Planning, Faculty of Transportation Engineering, Ho Chi Minh City, Vietnam

Corresponding author, E-mail: [email protected]

ABSTRACT: Road traffic crashes are a major public health problem affecting those living in Ho Chi
Minh city, a motorcycle dependent city. There has been limited success measure in reducing traffic
crashes and the crash rate keeps increasing. High quality crash data is an important issue in Ho Chi Minh
city. Data do exist - but their reliability is questionable, with respect to the data collection methods used
and underreporting. This study undertaken in Ho Chi Minh City, used surrogate safety measures based on
observation of traffic conflicts to assess the behavior of road users at twointersections. Data is analyzed
using the Traffic Conflict Technique. This study highlights are applicability of surrogate safety measures
in traffic safety assessment at intersections under motorcycle-dominated traffic environment.
Keywords: Surrogate safety measure; traffic safety; un-/signalized intersections;traffic conflict technique;
mixed traffic condition.

1. INTRODUCTION seriously injured causes economic burden and

affects the overall development of society [1].
The World Health Organization indicated that Traffic conflict technique (TCT) was first
traffic accidents are one of the leading causes of mentioned in the US in the 60s of the twentieth
death for people, according to the organization's century. The traffic conflict analysis technique,
statistics, more than 90% of accidents occurred which was discussed in this study, originated
in medium and low income countries. In from Lund Technical University (Sweden) in
Vietnam, according to reports by authorities, the 70s and 80s [2]. "Traffic conflict" is
there are about 20,000 traffic accidents each defined as all situations of two or more road
year with the number of deaths recorded as users that can lead to an accident, if conflict
around 9,000 people, and nearly 20000 injured avoidance actions such as braking, navigation
people. Only in Ho Chi Minh City, in 2015 or speed change are not possible to perform.
there were nearly 4,000 accidents with about This definition was first introduced by Perkins
700 deaths and over 3,000 injuries. Compared & Harris in 1967 by observing the start of
to 2014, accident data in 2015 showed signs of brake lights, lane changes and violations,
decline, correspondingly (- 14.4% of accidents, discussed in Van der Horst, 1990. "Traffic
- 3.3% of deaths, - 18% of injuries). The Conflict Technique" method has been adopted
consequences of traffic accidents are extremely by many countries in Europe and North
heavy, the number of people killed and America as a model for studying accident data

374
The 3rd Int. Conf. on Transport Infrastructure & Sustainable Development (TISDIC 2019). Construction Publishing House. ISBN: 978-604-82-2893-4.

and continuing to improve on the content and intersections, include: (i) geographic data; (ii)
method of implementation to match the traffic data; (iii) type of traffic conflicts; (iv)
specific traffic conditions of your country. the speed of the road user (who makes an
Therefore, there have been many additional evasive action) at the moment the evasive
researches on traffic conflict measurement action starts; (v) the distance from the road
techniques implemented in the following user (who makes an evasive action) to the
decades [3],[4],[5], [6], [7],[8]. imaginary point of collision. Geographic and
The study was conducted based on traffic data were collected by manual
observing, collecting data and behavior of methods. Type of traffic conflicts were
road users in real-life traffic situations, identified by using camera. SEV software was
through video camera system that the user developed by Minh (2011), it supports to
placed at the survey location. This is a useful measure the speed of road users[8], as
means of investigation because it provides illustrated in Fig. 1. Handbook Swedish
more complete and accurate information Traffic Conflict Measuring Technique (2009)
about the behavior of traffic participants and is used to calculate the time to accident (TA)
the interaction of traffic participants, as well and evaluate the severity of a conflict. The
as the type of vehicles. Traffic conflict data is overall process to determine the severity of a
collected, recorded and analyzed reasonably conflict at intersections can be divided into
to provide the information needed in order to four steps below.
provide the reason why the incident occurred,
determine the severity, the factors affecting,
risks regarded to accidents. The important
parameters in the survey process need to be
collected is the speed and the distance of road
users when impending conflict, thereby
giving the counter measures such as
avoidance, acceleration and deceleration, ...
and time to collision value (TTC) is the
remaining time before a conflict occurs if the
vehicle does not respond.
The results of this study provide the first
scientific basis, and the orientation for further
studies on whether the traffic conflict
technique analysis method (TCT) is suitable to
apply to the traffic safety performance analysis
at un/signalized intersections in Ho Chi Minh
City or not?. This study is divided into four
parts. After the first part - the introduction is Figure 1. SEV software application
the methodology. Next is the result and
discussion, the last part is the conclusion. Step 1: Observation period
 When should we conduct the traffic
2. METHODOLOGY conflict survey and for how long?
Two intersections are used for this study,  At dry weather
include: (i) Un-signalized intersection –  Observation period can coincide with
Nguyen Gia Tri – Vo Oanh; (ii) Signalized point in time of accidents
intersection – Nguyen Gia Tri – D5, located at  Take into account road works, diversions
Binh Thanh District, Ho Chi Minh City. Four and events that can influence the
types of survey were conducted at these trafficsituation

375
The 3rd Int. Conf. on Transport Infrastructure & Sustainable Development (TISDIC 2019). Construction Publishing House. ISBN: 978-604-82-2893-4.

Step 2: Observation preparation Step 5: Data analysis

 Inform road maintenance authority  Analysis to type of road user, time and
 Location research causes serious conflicts and behaviour
 Comparison with accident data
 Use of video camera?
 Proposals to tackle the road safety issues
 How many observers are to be used? on the location.
 Training observers
 (collect traffic intensities, speeds and 3. RESULTS AND DISCUSSION
accident data)
3.1. Collision diagram
 Preparing for speed and intensity
measuring during observations Traffic conflict types at two intersections were
collected by manual method, and draw
 Draw up an observation form
collision types based on American Highway
Step 3: Observation Safety Manual (2010) and Tien et al. (2017).
Most frequently occurred collision types
 Explore location: get familiar with involving motorcycle are as following: (i) Type
dimensions of location, practicing speeds 1: sideswiped collision same direction by lane
 Taking a position from which to have a changing – traffic movements of different
good overview of the location directions use the same lane under mixed
 Taking into account position video camera traffic; (ii) Type 2: left turning other angle
and other observers collision – lacking left-turn protected signal
phase, left-turn vehicles makes left turn after
 After 1 hour of observation: take a
ending of green time; (iii) Type 3: right turning
15minute break other angle collision by right turning
 Note down as much information as movement – lack of guiding signs and
possible on the observation forms markings guided the right turning; (iv) Type 4:
rear-end collision – improper inter-green time.
Step 4: Data processing
). 172 potential conflicts were observed, and
 When applicable, process video footage drawn the collision diagram at Nguyen Gia Tri
 Verify data on the observation forms by - Vo Oanh intersection, as illustratedin Fig. 2.
means of the video footage 235 potential conflicts were observed, and
drawn the collision diagram at Nguyen Gia Tri
 Process accident data, speed and intensity
- Vo Oanh intersection, as illustratedin Fig. 3.
measurements Table 1, Table 2 illustrated the percentage of
 Represent results in graphs collision type of vehicles at two intersections.

Figure 2. Collision diagram at Nguyen Gia Tri - Vo Oanh intersection

376
The 3rd Int. Conf. on Transport Infrastructure & Sustainable Development (TISDIC 2019). Construction Publishing House. ISBN: 978-604-82-2893-4.

Table 1. Type of conflicts at Nguyen Gia Tri – Vo Oanh Intersection

Type of conflicts Mc-Mc Mc -C Mc –B Mc-Tr Mc-Oveh Mc-Ped C-C C-Oveh C-P
No. of conflicts 80 64 2 1 7 4 9 1 4
Frequency (%) 47% 37% 1% 1% 4% 2% 5% 1% 2%
Same direction 4 2 0 0 0 0 0 0 0
Opposite direction 23 28 1 1 2 2 3 0 3
Crossing 53 34 1 0 5 2 6 1 1
*Mc: Motorcycle, C: Passenger car, B: Bus, Truck: Tr, Oveh: Other vehicles, P: Pedestrian

Figure 3. Collision diagram at Nguyen Gia Tri - D5 intersection

Table 2. Type of conflicts at Nguyen Gia Tri - D5 Intersection

Type of conflicts Mc-Mc Mc -C Mc –B Mc-Tr Mc-Oveh Mc-Ped C-C C-Oveh C-P
No. of conflicts 80 127 2 1 7 4 9 1 4
Frequency (%) 34% 54% 1% 0% 3% 2% 4% 0% 2%
Same direction 0 0 0 0 0 0 0 0 0
Opposite
51 63 1 1 6 3 2 1 2
direction
Crossing 29 64 1 0 1 1 7 0 2
*Mc: Motorcycle, C: Passenger car, B: Bus, Truck: Tr, Oveh: Other vehicles, P: Pedestrian

3.2. Conflict severity diagram Type 1: (i) to provide advanced sign

Time to accident (TA) value is calculated by information; (ii) lane direction marking.
using speed and distance data. The results show Type 2: (i) left turn guidance; (ii) left turn
that the number of serious conflicts at un- protection signals.
signalized intersection higher than signalized Type 3: right-turn curb lane.
intersection in this study, as mentioned in
Type 4: longer yellow and all-read time.
Figure 4, Figure 5. Causes of traffic conflicts
mentioned in section 3.1. Some potential Time to accident (TA) of Nguyen Gia Tri –
measures are proposed to reduce the number of Vo Oanh intersectionis shown in Figure 4 with
conflicts and the severity of conflicts according 77 dangerous conflicts and 95 non-dangerous
to types of conflicts, as mentioned below. conflicts.

377
The 3rd Int. Conf. on Transport Infrastructure & Sustainable Development (TISDIC 2019). Construction Publishing House. ISBN: 978-604-82-2893-4.

traffic environment and as a result record all

serious and slight interactions. There could
have some variables or factors affecting
conflict severity which were not captured in the
study. To overcome these issues, further
research could employ a broader behavioural
observation to provide additional factors to
compare traffic behaviour and conflicts and to
measure conflict severity.

5. REFERENCES
Figure 4. Conflict severity diagram of Nguyen Gia
Tri - Vo Oanh United Nations (2018), Road Safety Performance
Review Vietnam, New York.
Time to accident (TA) of Nguyen Gia Tri – Lund University (1976), The Swedish Traffic
D5 intersectionis shown in Figure 5 with 52 Conflict Technique, Sweden.
dangerous conflicts and 183 non-dangerous M.R. Parker, Jr. và C.V.Zegeer (1989), Traffic
conflicts. Conflict Techniques for Safety and Operations
Observers Manual, FHWA-IP-88-027, U.S.
Department of Transportation.
S.M.S. Mahmud et al. (2018), Reviewing Traffic
Conflict Techniques for Potential Applications
to Developing Countries, Journal of
Engineering Science and Technology, Vol. 13,
No.6, p. 1869 - 1890.
A. Petermans et al. (2009), Handbook Swedish
Traffic Conflict Measuring Technique, Policy
Research Centre Traffic Safety, Hasselt
University.
Tran Quang Vuong (2017), Traffic Conflict
Technique Development for Traffic Safety
Figure 5. Conflict severity diagram of Nguyen Gia
Evaluation under Mixed Traffic Conditions of
Tri - D5
Developing Countries, Journal of Traffic and
Transportation Engineering 5, 228-235.
4. CONCLUSION Chinebuli Uzondu et al. (2018), Exploratory Study
Involving Observation of Traffic Behaviour and
In this paper, 407 traffic conflicts at two
Conflicts in Nigeria Using the Traffic Conflict
intersections were collected and determined the Technique, Safety Science 110, 273 - 284.
severity of conflicts. The result show that most Tien-Pen Hsu, Ku -Lin Wen (2017), Collision Types
frequently occurred collision types involving of Motorcycle Accident and Countermeasures,
motorcycle. Some potential measures were Proceedings of the 2th World Congress on
suggested to reduce the number of conflicts Civil, Structural, and Environmental
and the severity of conflicts. Although no Engineering (CSEE’17), Spain.
research has been conducted using the Traffic Chu Công Minh (2011), “Phát triển phần mềm tính
Conflict Technique (TCT) in motorcycle toán các đại lượng chuyển động của xe nhằm
dominated environment, such as Ho Chi Minh xây dựng các ứng xử, trương tác khi xe khi
City, the method adopted in this study provided lưu thông”, Đề tài NCKH cấp Trường, ĐH
the opportunity to observe road users in a real Bách Khoa Tp.HCM. (Vietnamese version)

378
The 3rd Int. Conf. on Transport Infrastructure & Sustainable Development (TISDIC 2019). Construction Publishing House. ISBN: 978-604-82-2893-4.

Development of trip generation model for Danang City by

utilizing multiple linear regression model

Cao Xuan My1, Phan Cao Tho2

1
University of Technology And Education, University of Danang, Danang, Vietnam.
2
University of Technology And Education, University of Danang, Danang, Vietnam.
E-mail: [email protected]
E-mail: [email protected]

ABSTRACT: This paper presents the results of developing trip generation models for Danang city-
Vietnam. Trip generation is the first step of 4-step model which has been used for transportation planning.
This study investigates the relationship between the household trip generated in Danang city with its
socio-demographics attributes. The data using in the study was collected on 459 households in the city by
home-interview method. The Multiple Linear Regression was used to analyze the collected data. The
results of the analysis have shown that the average daily trip per household mainly depends on household
size, the number of vehicles and the number of children and/or senior people in the family. 
Keywords: Trip generation model, linear regression, correlation coefficient, traffic demand forecasting,
four-step process.

1. INTRODUCTION The aim of this research is to develop trip

generation models to predict the number of
A very important key component in trips generated by households in Danang City
transportation design is travel demand as the case study. The models are developed
forecasting. Travel demand forecasting is using multiple linear regression analysis, which
essential for the design of transportation establishes relationship between the number of
facilities and services, and also for planning, trips generated by households and some
investment, and policy development. Trip socioeconomic attributes.
generation is the first step of the traditional
four-step travel demand forecasting process. It 2. LITERATURE REVIEW
is critical that this step produces an accurate
value as these values form the basis for the 2.1 Trip generation model
subsequent steps and the errors in this step can Trip generation analysis means understanding
propagate in the entire estimation process the trips generated in different traffic zones in
(Kikuchi and Rhee 2003). The reliability of an urban area. The trip for the purpose of
forecasting results influences the following analysis is defined as a one-way movement
steps such as trip distribution, mode split, and from an origin to a destination for a person.
traffic assignment. These steps are shown in The entire urban area is usually divided into
Fig.1. Therefore improved trip generation smaller traffic zones, where the points of
models are needed to improve forecasting origins and destinations are fixed as zone
precision (Golob 2000). centroid. This is illustrated in Fig. 2.

379
 These types of trips according to movement
are shown in Figure 3. Trip generation analysis
is mainly related to the internal to internal trips
and less to internal to external trips. Home
interview survey is the major tool to establish
trip generation models as will be discussed
later. The other kinds (internal to external,
external to internal, and external to external)
can be identified by cordon surveys, which
means surveys conducted at convenient points
on the cordon line or at points of intersection
of radiating road.
The principal task in trip generation
analysis is to relate the intensity of trip
Figure 1. Sequences of Activities in Transportation making (number of trips made from a point to
Analysis several other points) to and from traffic zones
Source: Principles of Urban Transport Systems to measures of the type and intensity of the
Planning (Hutchinson, 1974) land use in these zones and to other socio-
economic characteristics.
There are two types of trip generation
analysis that can be carried out; theseare:
1. Trip productionanalysis
2. Trip attractionanalysis
The term trip production refers to the trips
Figure 2. Origins and Destinations for Traffic
generated by residential zones, where these
Zones in an Urban Area trips are either trip origins or trip destinations
(Papacostas and Prevedouros, 2004). Trips,
After delineating the urban area boundary and which end at home, are called home-based trips
fixing the zone centroids as points of origin or or trip production, while the term trip attraction
points of destination, the trips can be classified is used to describe trips generated by activities
according to spatial movement into four kinds at the non-home ends (Hutchinson 1974).
(Micheal D. Meyer and Ericj. Miller, n.d.) Home-based trips and non-home based trips
1. Internal to internaltrip are analyzed separately, because it is difficult to
2. Internal to externaltrip combine these categories of trips in developing
3. External to internaltrip models. There is a need to develop separate trip
4. External to externaltrip generation models for home and non-home
based trips, as the type of variables that might
influence such trips are to be different, or if the
variables are the same, the effect of these
variables on trip making might be different, so
the models for these two types must be
separated (Papacostas and Prevedouros 2005).
The process of relating the trips produced by
households to the factors influencing trip
production by appropriate analytical technique
is termed as trip production modeling and the
Figure 3. Types of Trips according to Movement in process relating the trips attracted to non-
Study Area residential ends to the factors influencing trip

380
attraction by appropriate analytical technique is can be done using computer programs like
termed as trip attraction modeling (Papacostas Excel, SPSS, STATA or R (this study use R
and Prevedouros 2005). On the other side, trip statistical software)
production or home-based trips can be classified
into different categories based on trip purpose, 3. RESEARCH CONTEXT
which are: The research was undertaken in the city of Da
1. Work trips Nang, Vietnam. The city includes a major
2. Education trips harbour and is the largest urban centre in central
3. Shopping trips Vietnam with a population of 1,046,252 people
4. Social and recreationaltrips (Statistical Yearbook of Vietnam 2016). Da
5. … Nang includes six urban districts (Thanh Khe,
Finally, trip production or home-based trips Hai Chau, Lien Chieu, Ngu Hanh Son, Son Tra,
can also be classified into different categories and Cam Le) and one district (HoaVang) in
based on time of making the trip. which Thanh Khe and Hai Chau are considered
According to the level of analysis, it could be to be the city centre of Da Nang.
analyzed on an aggregate level (zone or area), or Number of households by household size and
disaggregate level (household or person). In this district is 204,035 households (JICA 2010). The
research, analysis on the level of households is average number of residents to a household
adopted, because depending on this level, more recently is 4.53 (SKM 2012). Approximately
accurate results can be obtained, through the 36% of households have four occupants, and
study of movements for each household. that 18% of households have five occupants,
2.2 Overview of linear regression method 24% of households have more than five
occupants.In Da Nang City, 62.3% of
There are several methods used in relating trips households own at least one motorcycle and
produced or attracted to the causal factors, 25.2% own more than 2 motorcycles. The ratio
which include the regression method and the of households with their own car(s) is still very
cross- classification method. low, approximately 3.2% (SKM 2012).
Linear regression method assumes that The urban transport system of Da Nang is
observations of the magnitude of the dependent primarily composed of both urban and intercity
variable, Y (number of trips), can be obtained roads. Although there are railways and airports,
for n observations of explanatory variables they are largely for intercity and regional
(Xs), and that an equation of the formis to be transportation services. Urban transportation
fitted to the data. This equation is for the single services are mostly provided by private
independent variable case. The equation may transportation, which ischaracterized by a
be for independent variables as in the dominating presence of motorcycles. Public
equation(Nguyen Van Tuan 2014): transportation services are composedof bus,
y =  + ixi + i (1) taxi, and Moto Taxi “xe om” - however, their
share in urban transportation is low.
where, for i = n observations:
In regards to the travel modes, these were
yi - dependent variable;
categorized into six groups: bicycle,
xi - expanatory variables;
motorcycle, car, bus, xe om and pedestrian. The
 - y-intercept (constant term); trips by motorcycle have the highest mode share
 - slope coefficients for each explanatory e (68.9%), next is bicycle and pedestrian
variable; (14.37%) (SKM 2012). There is no difference in
i - the model’s error term (also known as using vehicles between genders. Using
the residuals). motorcycle by age of 20-50 years isrelatively
Estimating the best regression line high; especially the rate in age group 20-29,
depending on the coefficient of determination which is more than 80% (SKM 2012). As age
(R²), the t-test for the parameters, and the F-test increases, then there is some change decrease in

381
the rate of female’s using a motorcycle, but this samples. In this study, this number is extended
rate remains thesame for males. For all public to 459 samples for the estimation process and
transportation in the city at this time, females the validation process.
tend to make greater use of publictransport than
males, their average bus share being 55.7% 4.1.2 Question naire design and required
(compared with 44.3% for males) (SKM 2012). information
According (SKM 2012), the average daily
To analyze the role of the household
trip rate of people living in survey area is 2.67
characteristics in trip production analysis,
trip/day while this number is 2.9 trip/day/person
including walking and 2.3 trip/day/person sevenvariables (independent variables) obtained
excluding walking according (JICA 2010) from previous studies and thought to play a
significant role in trip production rates were
4. METHODOLOGY selected depending on survey data. The
variable selected were vehicle ownership,
The data used in this research was obtained household size, family type,family structure,
from a cross-sectional observation survey housing type, housing ownership, licensed
undertaken over 7 districts of Danang city (see drivers, household income level, and total
Fig.4) during August-December 2016. number of employed persons in the household.
4.1 Field survey and data collection Also, a set of dependent variables (trip
purpose) were selected to be analyzed
4.1.1 Design sample size of study area againstindependent variables mentioned
A statistic formulation for sample size previously, total trips (TT), home based work
determination to find the smallest sample size trips (HBW), home based other trips (HBO),
that provides the desired level of precision. The non home based trips (NHB), home based
minimum amounts of samples required are 400 school and college (HBS).

Figure 4. Location of observation sites in Da Nang City, Vietnam

382
 Therefore, the questionnaire was divided 1. Two skilled enumerators (in this study,
into two parts as follow: enumerators are students) are chosen to help in
Part 1: Personal and household the survey process.
characteristics and identification: 2. Enumerators have to be trained by
It is composed of two sections with briefing them about the details of the survey
specified questions in each. and how to conduct the survey.
- The first section is to collect household 3. The enumerators will be given random
characteristics such as: living area, house type, household addresses divided according urban
house ownership, number of vehicles/cars districts.
owned by household and income level 4. The random sample of households in
- The second section is to collect different traffic zones will be numbered on a
characteristics of family members such as age, printed map and the same numbers are put on
gender, occupation, education level, car driving the questionnaire layout, where each number
license ownership. on the map and the questionnaire represents
Part 2: Trip data one household.
This part is to be filled out by each member 5. Since the actual survey may take place
of the household age 15 or older. This second any time during the day, the respondents are
part is a travel diary that is used to record required to answer the question about the travel
participant's travel behavior over one day period. details according to the previous day. This
previous day must be typical working day from
4.1.3 Sampling method Monday to Friday.
The sample that is drawn from the household 6. The enumerators have to first get
population is a stratified random sample that permission to be surveyed from the household.
includes 459 households. Stratification reflects Then, they ask the household members about
geographic locations defined as TAZ. the details required in the questionnaire. Each
Therefore, the study area is divided into 7 member of the household should give answers
TAZ's. The samples are randomly selected about his or her own travel details, except for
from each of the 7 TAZ's. The sample size of children below 15 years. Trip details of
each TAZ is shown in Table 1. children below 5 years are normally ignored.
Table 1. The sample size of each TAZ 4.2 Data assembly and analysis
No. TAZ. Rate No. samples 4.2.1 The linear regression analysis process
1 Lien Chieu 15,53% 71
Regression equations can be developed
2 Thanh Khe 18,31% 84 through the following sequence of steps
3 Hai Chau 20,27% 93 (Hutchinson 1974):
1. Examining the relationships between the
4 Son Tra 15,26% 70
dependent variable and each of the independent
5 Ngu Hanh Son 7,44% 34 variables in order to detect nonlinearities. If
6 Cam Le 10,65% 49 nonlinearities are detected, the relationship
7 Hoa Vang 12,54% 58 must be linearized by transforming the
dependent variable, the independent variable,
Total 100% 384 or both.
2. Developing the inter-correlation matrix
4.1.4 Conducting field survey
involving all the independent variables.
After determining the sample size, designing 3. Examining the simple correlation matrix
the questionnaire, and choosing the method of in order to detect the potential sources of
survey, the field survey in this study was multicollinearity between pairs of the
conducted according to the following steps: independent variables.

383
 4. After examining the correlation matrix, However, there are some consequences.
if there is multicollinearity between two First, regression coefficients are still best linear
independent variables (much closer to one), unbiased estimators (BLUE). Second, it is
then one of them must be eliminated from the more difficult to precisely identify the separate
regression process. effects of the correlated explanatory variables.
5. After choosing the correlated Finally, variances and standard errors are
independent variables, some of regression usually higher, making t-statistics lower and
equations are suggested, and then the possibly insignificant.
parameters of each of the potential regression Analysis often relies on what is called the
equations are estimated. Variance Inflation Factor (VIF) to detect
6. For every model built, relevant tests are multicollinearity more formally. The VIF
conducted to assess the goodness of the model shows how the variance of an estimator is
based on logic and statistics testing. The inflated by the presence of multicollinearity.
statistical tests include the coefficient of As R2, the coefficient of determination of a
determination (R²), the t- test for the parameters given explanatory variable with other
of the models, and the F-test for each model. In remaining explanatory variables in the model,
addition, there are logical aspects that must be increases toward one, the VIF also increases.
taken in consideration, because the model must The larger value of VIF, the greater the degree
be valid statistically and logically. of muticollinearity of one explanatory variable
7. After all the models or equations are with the other explanatory variables. As a rule
examined based on test results, the outcome is of thumb, if the VIF of a variable exceeds 10,
summarized. The best model is then chosen. that variable is said be highly collinear
(Kleinbaum and Kupper 1978).
4.2.2 Unit of analysis
4.2.4 R-Squared: Goodness of fit
The household is used as the unit of analysis in
this research. The R-squared (R2), also known as the
coefficient of determination, measures the
4.2.3 Correlation matrix and VIF: Testing for goodness of fit of the regression model. It
multicollinearity measures the proportion of the total variation
The problem of multicollinearity arises when in the dependent variable that can be explained
two or more explanatory variables included in by the explanatory variables included in the
the regression model have linear relationships. model. The value of R-squared lies between 0
There are two types of multicollinearity: exact and 1. A value of R-squared closer to 1
or perfect multicollinearity and inexact or indicates that the model has good fit, whereas a
imperfect multicollinearity. value closer to 0 indicates that the model has
If two or more independent variables have poor fit.
exact or perfect linear relationships, then we However, there is no standard on how high
have exact or perfect multicollinearity. On the R2 value is “good” enough. It depends on the
other hand, inexact multicollinearity occurs in application.
the case that two or more independent variables
are highly correlated. 5. RESULTS
In the case of exact multicollinearity, there
5.1 Descriptive statistics
is no unique solution to the normal equations
derived from the least squares principle. As a The average number of residents to a
result, the regression coefficients cannot be household recently is 3.99. The survey results
estimated. By contrast, when explanatory show that about 36% of households have four
variables have less than exact linear occupants, and that 26% of households have
relationships, the normal equations can usually three occupants, 25% of households have more
be solved to yield unique estimates. than five occupants.

384
 Approximately 40% of households own at Results (Figure 8) show that Home Based
least one motorcycle and 48% own more than 2 Work trips contribute for 39% of trips as
motorcycles. The ratio of households with mentioned with maximum of 16 trips/day and an
their own car(s) is still low, approximately 9%. average of 3.98 trips/day followed by Home
Based Other trips with 34% of trips and a
The distribution of total number of
maximum16 trips/day and a an average of 3.40
employed persons is shown in Figure 7. The trips/day. In the third place come Non Home
families with 2 employees came in the first Based trips with 17% of trips with maximum of 12
place followed by families with 1 and 3 trips/day and an average of 1.7 trips/day. In the last
persons. And also shows that when the one is Home Based School trips with 10%. Total
family grows bigger the number of families 4647 trips show a maximum of 31 trips/day made
decrease steadily. by household and an average of 10.12 trips/day.

            
Figure 5. Size of Household Figure 6. Number of vehicles owned by Household

Figure 7. Total number of employed persons Figure 8. Trip generation purposes

385
5.2 Logistic regression modelling Before using this method, the four
assumptions of linear regression are tested.
In running the regression analysis for Trip
Theyare: Linearity of residuals, Independence
Generation Models generated, all the
of residuals, Normal distribution of residuals
descriptors in Table 2 were entered at first and
and Equal variance of residuals. Besides, the
a multiply regression analyses and Bayesian
multicollinearity issuse is also tested to ensure
Model Averaging (BMA) were done.
that there is no exact collinearity between the
Table 2. Description of independents variables/ explanatory variables.
Parameters used in the analysis
6. DISCUSSION AND CONCLUSION
No. Independent Variables Symbol
1 Size of Household X1 6.1 Discussion
2 Number of vehicles owned by X2 The results of the analysis show that:
Household Household with more vehicles availability
3 Number of cars owned by Household X3 make more trips than household with less
4 Total number of employed persons X4 vehicles availability, both Total trips and
Home Based Work Trips.
5 Number of persons under 18-year-old X5
Smaller family sizes generate less trips than
6 Number of persons between a 18- X6 those with larger family sizes; that is trip
year-old and an 60-year-old generation decreases with decrease in family
7 Number of male X7 size for the study area.
8 Number of female X8 Notably, trip generation decreases with
9 Number of children and retiredpersons X9 increase the number of children and
retiredpersons in household of the study area.
10 Number of Students X10
This is because these people are usually carried
11 Number of Full-time employees X11 by others in their family.
12 Number of Part-time employees X12 Home based work trip makes the largest
percentage (39%) of people in study area
Table 3. Description of dependents variables/ compared with other purpose trips. Every one
Parameters used in the analysis Full-time employee increasing make
No. Dependent Variables Symbol Home Based Work Trips increase 0.59 trips.
1 Total trips YTT
➢ One of the most significant parameter in
estimating total trip generation according the
2 Home based work trips YHBW
finding of (Oyedepo and Makinde 2009) is
3 Home based other trips YHBO Household income disappears in this study.
4 Home based school trips YHBS Result of trip generation in Danang city is
5 Non home based trips YNHB useful for the further study on trip distribution,
mode choice and trip assignment in order to
Bayesian Model Averaging (BMA) is an forecast the travel demand for Danang city in
application of Bayesian inference to the the future time period
problems of model selection, combined It is recommended to develop trip generation
estimation and prediction that produces a models as the primary tool for travel demand
straight forward model choice criteria and less modeling process in order to get better
risky predictions (Nguyen Van Tuan 2014). modeling of traffic flow in the future for better
The results obtained give the proposed transportation planning processes in the
model for trip generation models generated as Vietnam cities.
shown in Table 5.

386
Table 4. Descriptive statistics independents and dependents variables used in the analysis
HH_size Veh_own T_employ TT HBW HBO HBS NHB
Mean 3.99 2.78 2.17 10.12 3.98 3.40 1.04 1.70
Std. Deviation 1.716 1.348 1.150 4.673 2.982 2.685 2.056 1.958
Minimum 1 0 0 0 0 0 0 0
Maximum 17 9 6 31 16 16 24 12
25 3.00 2.00 2.00 7.00 2.00 2.00 0.00 0.00
Percentiles 50 4.00 2.00 2.00 9.00 4.00 3.00 0.00 1.00
75 5.00 4.00 2.00 12.00 6.00 5.00 2.00 3.00

Table 5. Formulated Trip Generation Models

No. Model type Model R2
1 Total Number of Household Trip Generation YTT=1.92+1.53X1+0.99X2–0.92X9 0.51
2 Total Number of Home Based Work Trips YHBW=0.18+0.33X2+X4–0.38X5+0.59X11 0.43
3 Total Number of Home Based Other Trip YHBO=1.31+0.80X1–0.58X11 0.19
4 Total Number of Home Based School Trips YHBS=0.10+0.50X6+0.6X10–0.53X11 0.17
5 Total Number of Non Home Based Trips YNHB=0.56+0.78X5–0.23X8+0.4X12 0.17

6.2 Conclusion and recommendations vehicles owned by Household and the Number
of Full-time employees, respectively. The rest
It is concluded based on the findings: of variables have low correlation < 0.3.
It shows that Total trips model is the best Note that most of the mentioned variables
model with coefficient of determination R2 = listed above by trip purpose are variable with a
0.51 < 0.6 which is acceptable in transportation coefficient of correlation greater than 0.3.
consideration which required R2 to be greater That the result of the regression analysis of
than 0.3, followed by HBW trips with R2 = 0.43. total trips generated as dependent variable as
HBO trips, HBS trips and NHB trips model against other variables gives value of R2 as
came in the lower ranking with a coefficient of 51% (0.51) for the regression analysis of the
determination R2 = 0.19, R2 = 0.17, R2 = 0.17 socio-economic status of the people. R2 = 51%
(R2< 0.3), respectively, a weak relationship and means that 51% of the dependent variable is
it is considered not significant in transpiration explained by the independent variables
consideration. Therefore, these trips model includedin the regression analysis. Model
have to be modified using another statistical validation and optimization were done to test
model type such as non linear model. the usefulness of the proposedmodels. These
The most important variables that affect trip results confirm that the proposed models can
generation by trip purpose are as follow: be used to forecast future trip number for
- For total trips, 3 variables considered to be thestudy area and areas with similar household
influencing total trips but the most effective socioeconomic attributes.
variable was Number of vehicles with There are versus recommendations or
coefficient of correlation R = 0.66, followed by alternatives can be enhanced to improve the
Size of Household and Number of persons study:
under 18-year-old, respectively. A similar study for different possible
- Four variables affect HBW trips, the most independent variables or different division of
was Total number of employed person, with a zones according to different land uses should
high R = 0.63, followed by the Number of be produced. One of the results and data

387
interviewed from this study it can be JICA, Japan International Cooperation Agency.
recommended that further results can be 2010. The Study on Integrated Development
obtained if local study of trip attraction is Strategy for Danang City and Its Neighboring
performed on each trip attraction major center Area in the Socialist Republic of Vietnam
like super markets, commercial, educational, (DaCRISS), Final report/Annex 1.
Kikuchi, Shinya, and Jongho Rhee. 2003.
and recreational areas within the city of
Adjustment of Trip Rates in the Cross-
Danang to more effectively reflect its attraction Classification Table Using Fuzzy Optimization
size on the city transport volume. Besides, a Method. TRB 2003 Annual Meeting.
similar study in other parts of Danang cityis Kleinbaum, David G, and author.) Kupper
carried out for validation or comparison. Lawrence. 1978. Applied Regression Analysis
Danang Department of Transport should and Other Multivariable Methods. 2d ed.,
conduct Household Travel Survey (HTS) that PWS- Kent, Boston, Mass., 1988, PP. 210.
includes detailed data about trips made by Micheal D. Meyer and Ericj. Miller. 1986. Urban
households. This survey can be used by Transportation Planning. McGraw-Hill Book
researchers to estimate trip generation Company.
models. This date can be used to conduct the Nguyen Van Tuan. 2014. Data Analysis with R. Ho
research related to the temporal stability and Chi Minh General Publishing House.
Oyedepo, O. J., and Oluyemisi Opeyemi Makinde.
dynamics of travel behavior in order to
2009. Regression Model of Household Trip
understand how trip generation model Generation of Ado-Ekiti Township in Nigeria.
parameters change over time. European Journal of Scientific Research 28
(1): 132–140.
REFERENCES Papacostas, C. S, and P. D. Prevedouros. 2005.
Transportation Engineering and Planning.
Golob, Thomas F. 2000. A Simultaneous Model of
Third Edition. Prentice-Hall of India Pvt. Ltd.
Household Activity Participation and Trip
NewDelhi.
Chain GenerationJournal of the SKM, Sinclair Knight Merz Ltd. 2012. Feasibility
Transportation Research Board, No. 34, TRB.
Study for the Danang Sustainable City
Washington, USA, pp. 355-376. Development Project, Component 2 – BRT
Hutchinson, B. G. 1974. Principles of Urban
System (Final Feasibility Study Report).
Transport Systems Planning. First Edition
edition. McGraw-Hill Book Company, GSO.General Statistics Office. 2016. Statistical
NewYork. Yearbook of Vietnam, Statictical Publishing
House, Hanoi, Vietnam

388
The 3rd Int. Conf. on Transport Infrastructure & Sustainable Development (TISDIC 2019). Construction Publishing House. ISBN: 978-604-82-2893-4.

Advanced urban road intersection design: up-grade bike

extra lane

Marwa S. El-bany
A lecturer of transportation planning and traffic engineering, faculty of engineering, Port Said
University, Egypt

*Corresponding author, email: [email protected]

ABSTRACT: Traffic congestion is one of the most worrying problems at urban areasof developing
countries especially those with high population density and car ownership.Decision makers go towards
the sustainable transportation solutions such as biking. There are other solutions such as mass
transportation but these are not always the best because they require sacrifices in the living style of the
commuters. At intersections, the conflict between bikes and all vehicles is a big problem. This paper
proposeda smartup-grade extra lane design for bicycles atintersection with narrow roads that have no
possibility to enlargement. The recommendedroad intersection design consists of only one level
overpasses for small vehicles and mixed vehicles excluding one up-grade lane forbikes. Movement
separation by up-grade bike lane will be the effective solution as bridges made to avoid the conflict
between the through bike movement and opposite movements. According tothe average speed of bikes,
the minimum bypass lane distance needed to merge with the main roadis 62.5m for 2% up-grade lane.The
proposed design is a way to increase road traffic mobility and possible handling without disallowing
larger vehicles. Also, it is likely to increase safety of bikes movement because of the mode movement
separation for either signalized or non-signalized intersections.
Keywords: Road intersection, up-grade bike lane, sustainable transportation solution.

1. INTRODUCTION other solution is using bike to travel from origin

to destination. Bike as a mode of transportation
The traffic volumes and flow rates at road is widely feast in many parts of the world. Asia
intersection havea fast increasing in urban cities continues are the main market. Between 1997
which make many congestion problems, more and 2002, bikes made 52% of traffic, taking
frantic measures were proposed. Using the own superiority over all other vehicles, including
car becomes a usual mode to transport because cars. For example, 50% of all trips in the Dutch
it is easy and more safe than public transport cities are done on bikes [1]. Then the type of
that increasing the traffic volumes every day. traffic is called mixed traffic.
The own car handles one person but take a big At the intersections, mixed lanes are
area from road. Some solutions can solve this required as the roads intersect with each other
problems but with high cost like electronic 1- for mixed traffic. The intersection shapes;
driver car that will increase the passenger when two roads intersect, the intersection is
handling of the road transportation systems. The called simple 4-way, 5-way, and 6-way

389
The 3rd Int. Conf. on Transport Infrastructure & Sustainable Development (TISDIC 2019). Construction Publishing House. ISBN: 978-604-82-2893-4.

intersections. One lane from any direction can For each lane, the vehicle can go straight, right or
be eliminated to be up-grade for the bike left (basic three movement) as shown in Fig 1.
movements. The proposed design can use for
the 4-way intersection by taking one lane to be
up-grade lane for bikes at intersection area. At
peak periods, the best traffic handling is only
half of the best possible if bridges were used.
The traffic flow rates at 4-wayintersection
The total flow rate at4-way intersection is
needed to be calculated when bikes involved in
the traffic. The basic design equationdepends
on road traffic flow and time headway (gap).
Traffic flow (q) is the number of vehicles that
pass a reference point per unit of time. If his
the time headway(gap in time between
vehicles), so:
nt 1
 1 t ht 
n
q (1)
t hav Figure 1. Basic movement
Where: Actually there are bikes go right, though,
hav is the average headway. and left. For right turn, there is no problem if
Thus average headway is calculated as the there is a separated lane but for through and
inverse of flow. The microscopic character left they may be conflict with the vehicles
related to volume is the time headway or simply come from the other direction.
headway for bicycle will be the same equation. Although the bicycle is a popular form of
Time headway is definedas the time transport safety prevents most people who own
difference between two following vehicles bikes to ride them frequently. In developing
when they cross a given point. Practically, it countries, most city streets are designed keeping
engagesthe measurement of time between the in mind only the cars and cycling in the street
passage of one rear bumper and the next can be dangerous. The transport system must
passage a given point. If allheadways h in time have an independent network, as the bicycle is
period, t, over which flow has been measured, not considered an equal participant of city
are added then, Eq. (1) shows that traffic flow traffic as other vehicles in practice.
isdependent on the gap between two vehicles,
not on their speed [2]. 2. LITERATURE REVIEW
A safe gap for human drivers is two seconds
[3]. For such a gap, the traffic flow for one lane A few inventions allow road intersections
of road in one hour is 50 bicycles. Bikes should without using any traffic signals [4] uses traffic
therefore be able to increase traffic flow if they circles and flyovers to remove traffic
are able to operate safely at a lower time gap. congestion. It requires a lot of space so is not
For a 4-way intersection, it means that two suitable for a dense urban area. It has
road lanes cross. The total flow of the advantage for two road levels only. Another
intersection qi is doubled patent that uses three road levels [5]. It can be
qi  2.q (2) considered a hybrid because it uses a road
circle also. Although a roundabout does not
The total flow should equal this value for all require any traffic light, drivers need to wait
the possible traffic movements to reach before they can enter the circle. It is more
maximum operative action of the intersection. difficult than just merging two lanes into one.

390
The 3rd Int. Conf. on Transport Infrastructure & Sustainable Development (TISDIC 2019). Construction Publishing House. ISBN: 978-604-82-2893-4.

Classic textbook uses passenger capacity

unit (PCU) as a unit of measurement for the
traffic flow, q[6]. A larger vehicle such as a
bus is assigned as three PCUs. Eq 1 fails in this
case because when combining three vehicles,
the gap between them is zero, allowing a large
flow rate. The equation is to be used with care
but in this article, the main emphasis is in
designing for standard passenger cars.
The weaving section is used when bicycle
and cars are moved together [6], [2]. When two Figure 4. Amsterdam Central Station Tunnel, 2015 [8]
lanes merge into one, it is called merging.
When one lane diverges into two lanes, it is
called diverging. An example of a simple rule
of thumb in calculating the length of the
weaving section will be provided in Eq 5.
Although there are calls for small
underpasses [6], a search in the internet proved
elusive for small underpasses with a vertical
clearance of less than 3m, except for trains. An
even urban underpass does not propose small
underpasses. Small underpasses are widely Figure 5. Norman Foster bridge for elevated 'Sky
used in building car parks, which are only for Cycle' bike routes in London, 2014 [9]
vehicles that are less than 2m in heights [7].
4-leg Intersection with 3 lanes per each
It can be viewed as putting traffic lights at
direction
each driverless vehicle. Traffic flow is still
limited by the amount of real estate space just Here, the case study is any ordinary small 4-leg
like network hubs. Hubs cannot compete with intersection with 3 lanes per each direction.
network switches that allow point-to-point The beginning of thinking is the ordinary
communications similar to using layers of solutions. There are two ordinary solution
tunnels or bridges. scenarios for bikes, one is the main road 3
Problems of traffic jams are so bad that for lanes/direction with applicable (solution 1) and
people safety, the Amsterdam Central Station it has no problem to take one lane as bicycle
was built in the 1880s there is now a tunnel to lane Referring to or making any merging to
pass straight under the main road. It was finally increase width of road; if it is possible as seen
opened in November 2015. The tunnel had been in Figure 2, for bicycle driving on the right side
finished to load bikes and pedestrian to separate of the road, turning right don’t have any
movements between cars and bikes as seen in Fig problems because there is no conflict point. It
4 [8]. In Figure 5, amazing cycle super highways just needs a bypass road, indicated by two
for bikes had been implemented in 2014; the lanes going in the same direction.
Norman Foster design elevated bike paths [9]. Turning left requires control. One lane going
This kind of bridges has multi-lanes design and straight, left to right and vice versa, will need
cannot implement in small intersection. bridge lane. It will be up-grade lane to be far
Based on the previous work, there are many away from road vehicle movement and to go
solutions for safely bikes movement but need a through until intersection with mail lane and left
high cost of implementation and control or a turn with the nearest U-turn in the street far
big road width. So, this paper presents a away from the intersection. The other scenario is
solution for small intersections without the same intersection but with adding new lane
needing control by signals. for right turn (solution 2) as seen in Figure 2.

391
The 3rd Int. Conf. on Transport Infrastructure & Sustainable Development (TISDIC 2019). Construction Publishing House. ISBN: 978-604-82-2893-4.

road which called merging. For non-signalized

intersection, it will be safe for bikes. For
signalized intersections, it reduces time needed
for bikes phase.

Figure 2. the ordinary right turn bike route

solutions for 4-way road intersection

However, there is no need to traffic signals

for bicycle movement because two road lanes Figure 3. 4-way intersection with traffic lights
are able to merge to one road lane safely. For Source: https://io9.gizmodo.com/this-awesome-intersection
safe merging, the speed of bikes on both lanes -design-is-a-dream-come-true-f-1529805459
should be the same. It would require additional The distance ofthe bypass lane needed to
lanes to allow stationary bikes in the up-grade merge with the main road is calculated by the
lane to merge with the main lane, taking into following equations.
account that when corners are sharp; bikes will
need to slow down considerably. These The bypass lane design calculations
solutions are ordinary solution for bikes to take For the ordinary bypass lanecalculations,
right lane but it still has a conflict with the left assuming constant acceleration (a), distance
movement of the opposite direction. So, the travelled (D), time(t), a bike that is initially
proposed up-grade lane is a solution to increase stationary, it can be shown that:
safety and reduce intersection conflicts. Also,
these are effective only for right movement. D  1 at 2 (3)
2
In case of signalized intersection, the other
proposed design is to make the lane foronly Time, t, required to achieve terminal speed
bicycle, it is recommended to take all safety of vt can be show to be:
standards. It makesa separated phase for bikes t  vt / a (4)
such as in Figure 3. Finally,the ordinary design
is to make the lane for bicycle but the Substituting equation (3) into (4):
intersection still need traffic signal or any of D  1 vt2 / a (5)
traffic control. Figure 3 shows a design for a 4- 2
way intersection where traffic lights are still Eq 5 is to design the distance of the bypass
used. Only the right way can travel freely lane to merge with the main bicycle lane in that
without being obstructed for small vehicles. example. If vt = 5 km/hr and a = 0.2g (proposed
To avoid all previous control considerations values of bikes), the distance will be 62.5m. The
for go straight (through) movement of bikes, 5 km/hr is chosen because it is the average
the up-grade bypass lane design is proposed. speed of bikes on roads through roads. Bike
This lane will separate from road which is speeds with low traffic, range from 3 km/hr to
called diverging then merged with the main 8 km/hr [11] and [12]. This will be proposed for

392
The 3rd Int. Conf. on Transport Infrastructure & Sustainable Development (TISDIC 2019). Construction Publishing House. ISBN: 978-604-82-2893-4.

merging and diverging. It is recommended to 7%. For cars, up-grade height of 3m, it is
use it as the minimum distance for bypass lane. recommended to be two lanes (one for cars and
one for bicycle) in the reverse direction, the
3. THE PROPOSEDUP-GRADE BICYCLE underpass should be 12m long. The total length
LANE is therefore 60m. The geometric design of the
road should follow the standard approved for
In this paper, the proposed design is in the the locality [13].
form similar to a patent application. The To limit the levels to just two, there is no
diagrams and descriptions should be sufficient turn for up-grade lane. The crossing turns can
to allow a person skilled in the arts to construct be implemented with a U-turn after the
and use the apparatus as described far along. intersection or after merging area. The exit of
The proposed design consists of prior arts so the U-turn should be the left-turn. The
are feasible to be implemented. The prior arts minimum radius should allow the largest cars
are combined in an unobvious manner and yet that are allowed to use the U-turn underpass. If
allow a more efficient traffic flow without four lanes are used for the turning, assuming 3 m
requiring too much space. It is one up-grade per lane, the achievable turning radius is only
lane for bicycle for both directions. 6m, which is lower than the recommended
Two lanes are for other vehicles. Up-grade turning radius by [14], [15].
lane is recommended because it does conflict Fig 7 shows the side view of the entrance of
with the movement. For main road, traffic up-grade lane of 1%-7% upgrade to
volumes are high especially in through accommodate standards for bike bridges Fig 8
movements. One lane is can make for cars but [16]. As seen in the figure, the specific area of
with the proposed inclined design (see Figure 6) curve approved the grade with comfortable
and the rest of the central lanes can be reserved movement from 1.5-2.5%. The proposed
for cars turning left and right for one direction upgrade is 2% so that the design is be provided
to allow the movement of all other directions of to reduce the amount of water (windy
the secondary road to move without conflict. condition) if there is rainwater and been
comfortable. For example, the height can be
4m along 200m as the buses and trucks can
easily move in the perpendicular direction.
If it follows the 4-lane design, the minimum
turning diameter for the U-turn is only 9m
centre-to-centre. This is not a bad design choice
because in urban cities, a few U-turns actually
have this as the turning diameter because of the
3-lane space provided for the turn. It can also be
observed that most of the time, traffic
congestion occur only in one direction. This
solution is the cheapest and feasible solution for
crowded cities; for example Cairo city.
If more space and budget is available, the
protected bike lane designmay be attempted. It
does not allow large vehicles at all and require
Figure 6. the proposed through intersection up- six lanes although five lanes can still be
grade bicycle lane provided along the less congested direction.
Because of the extra lanes, the U-turn turning
Since the speed is limited to just 40 km/hr, diameter can be increased to 12m if the lane
the gradient of the entrance to the up-grade can width is 3m. Figure 9 illustrates theproposed
be steep. The recommended gradient is 1:8 or protected bike lane design.

393
The 3rd Int. Conf. on Transport Infrastructure & Sustainable Development (TISDIC 2019). Construction Publishing House. ISBN: 978-604-82-2893-4.

Figure 7. The proposed up-grade lanecross-section with grade 2%

Figure 8. The relationship between bikes up-grade and comfort ability [16]

Figure 9. One protected bike lane cross section [17]

4. DISCUSSION AND FUTURE WORK operate an embodiment of the invention.

Technical analysis is not compulsory but will be
The design that is presented is not as detailed as useful to establish the technical advantage of the
production drawings but this level of disclosure invention. It should be noted that these designs
of drawings and descriptions are usually do not require elaborate construction
accepted by patent offices, because what is techniques. However no design or patent had
required is that people skilled in the art can use been published that implement the road
the drawings and descriptions to construct and intersections as described in this paper.

394
The 3rd Int. Conf. on Transport Infrastructure & Sustainable Development (TISDIC 2019). Construction Publishing House. ISBN: 978-604-82-2893-4.

Future projects should go towards bicycle lanes. Some concluding observations of

activating the proposed design in Sustainable the design characteristics are:
Transport Project for Egypt [18]. This will be i. The design allows small cars to move
a solution forintersection controlling to reduce through lane without going through
fuel consumption and congestion in traffic conflicts. Because small cars
developing countries. contribute the most to the flow of road
traffic in urban areas, the proposed
5. CONCLUSION design will be suitable.
ii. It is recommended to use the proposed up-
Developing countries have a mixed traffic and grade lane for only bicycle to separate the
special features. All kinds of vehicles move movement and increase road safety level
together using the same lanes. At intersections, therefore there is no need toanalytical
many accidents may be happened if there is no methods used to justify the optimisation
traffic control especially for left turn unless up-grade bicycle lane standards.
movement when conflict with the opposite iii. The design permit to create up-grade U-
movements. Mixed traffic means cars, turn for left movement if needed.
motorbikes, heavy vehicles, bikes, and
pedestrians. The conflict points without control REFERENCES
among all movements are the main reason of
such problems. Sustainable transport such as UrošRadosavljević,Examples of Alternative
biking is the recent goal to reduce fuel Transportation Systems in Urban
Environments,Conference Paper, (2014).
consumption and air pollution. Decision
https://www.researchgate.net/publication/3273
makers use some ordinary solutions to increase 89361
such strategy by making a separated lanes but Fundamental Parameters of Traffic Flow, NPTEL,
it still has a problem at small intersection. In Chapter 30, May 3, (2007).
case of narrow roads, there is no possibility of Niiya, Futoshi, Niiya, Hiroshi and Kuroishi, Tomizo,
a separate lane execution. To avoid all previous Patent No: 7,234,891 b2 U.S.A., (2007).
restrictions, the up-grade bypass lane design is Slinn, Mike, Matthews, Paul and Guest, Peter,
proposed. This lane will separate from road Traffic Engineering Design: Principles and
which is called diverging then merged with the Practice, 2nd Edition, Oxford: Elsevier
main road which called merging. It is an ideal Butterworth-Heinemann, (2005).
Daily Times, Pakistan. More underpasses, flyovers
solution for congestion problemsthat solves
required to regulate traffic flow. Daily Times,
many of congestion and air pollution at Cited: November, (2017)
intersections. It is a capable design for both http://dailytimes.com.pk/sindh/16-May-
signalized and non-signalized intersections. 16/more-underpasses-flyovers-required-to-
For non-signalized intersection, it will be safe regulate-traffic-flow
for bikes. For signalized intersections, it The City of Calgary, Downtown Underpass Urban
reduces time needed for bikes phase. Design Guidelines. [Online] (2010).
The proposed design methodology of a road https://www.cip-
intersectionwas presented. The properties and icu.ca/Files/Awards/Planning-
parameters of the design are based on current Excellence/2011-Winner-Urban-Design.aspx
standards. This study proposes an example (a Urban Planning Affairs, Kingdom of Bahrain.
Guidelines for the Design of Off-street Car
case of 2% gradient) which does not allow
Parking Facilities, (2017).
speed driving above 5km/hr for bicycle. For http://websrv.municipality.gov.bh/ppd/doc/stu
vt = 5 km/hr and a = 0.2g, the calculated dy_carpark.pdf
distance of the bypass lane to merge with the Amsterdam Central Station Tunnel, Bicycle Dutch
main lanes is preferred to be more than 62.5m. Website, (2015).
Analytical methods are also used to justify the https://bicycledutch.wordpress.com/2015/11/2
optimisation used in the design for cars and 4/amsterdam-central-station-tunnel/

395
The 3rd Int. Conf. on Transport Infrastructure & Sustainable Development (TISDIC 2019). Construction Publishing House. ISBN: 978-604-82-2893-4.

Norman Foster unveils plans for elevated http://www.academia.edu/34511665/Roads_B

'SkyCycle' bike routes in London, (2014). ranch_Public_Works_Department_Malaysia_
https://www.theguardian.com/artanddesign/arc A_Guide_on_Geometric_Design_of_Roads
hitecture- design-blog/2014/jan/02/norman- Austroads Design Vehicle and Turning Path
foster-skycycle-elevated-bike-routes-london Templates Guide, Cited: July 25, (2017).
Global Ergometer Exercise Bikes Market http://www.austroads.com.au/road-
Manufactures and Key Statistics construction/road-design/turning-path-
Analysis,(2017). templates
https://www.egypt- S Columbium Way Protected Bike Lane Options,
business.com/...Bikes...Statistics.../175302 Seattle Department of Transportation, (2013).
New York State Department of Motor Vehicles, https://www.seattle.gov/Documents/Departme
Chapter 8: Defensive Driving, Cited: February nts/SDOT/NSF/15th_BoardsPBL.pdf
20, (2017) English summary of the CROW design
https://dmv.ny.gov/about-dmv/chapter-8- guide,”Brief Dutch Design Manual for Bicycle
defensive-driving, (2017). and Pedestrian Bridges”, June(2015).
Cedeno, Arturo Olivero, Patent No:2,941,454 https://www.fahrradlandbw.de/.../Brief_Dutch
U.S.A., (1960). _Design_Manual_for Bicycle and Pedestrian
Cycling in Cairo: Just use your head - Daily News Bridge.pdf.
Egypt, (2012). Cabinet of Ministers Egyptian Environmental
https://dailynewsegypt.com/.../bicycling-in- Affairs Agency, Funded by UNDP, (2014).
cairo-just-use-your-head/ http://www.eg.undp.org/content/egypt/en/hom
Public Works Department of Malaysia, a Guide on e/about-us.html
Geometric Design of Roads, (1989), from
website

396
The 3rd Int. Conf. on Transport Infrastructure & Sustainable Development (TISDIC 2019). Construction Publishing House. ISBN: 978-604-82-2893-4.

Development of bus models to evaluate road roughness

and passenger ride comfort

Nguyen Van Teron1, Patrick Swolana2, Bernhard Lechner3, and YiikDiew Wong4
1
01 Create Way, #10-02 CREATE Tower, Singapore 138602, email: [email protected]
2
Technische Universität München, Arcisstraße 21, 80333 München, Germany
3
Technische Universität München, Baumbachstraße 7, 81245 München, Germany
4
Nanyang Technological University, 50 Nanyang Avenue, Singapore 639798, Singapore

ABSTRACT: Improving ride comfort for bus passengers experiencing vibrations under the excitation of
road roughness is a meaningful solution to attract more public transport ridership. While many research
studies have been conducted to simulate car and truck dynamics, few studies have studied bus dynamics.
This study developed multi-degrees-of-freedom (3-DOF quarter, 5-DOF half and 9-DOF full)
mathematical bus models in Matlab/Simulink and compared these models in calculating passenger ride
comfort on buses operating along urban bus lanes. The developed algorithms can incorporate speed
changes that replicate the stop-and-go phenomenon of regular bus service, resulting in more reliable
outputs. The comparison results have shown that the 9-DOF full bus model is the best alternative to
measure passenger ride comfort within the error of 2%, as compared to 5-DOF half and 3-DOF quarter
bus model with 7% and 20% error, respectively. However, the modelling complexity is much higher in
cases of half and full bus models. The widely used quarter vehicle model for designing vehicle suspension
and for evaluating road roughness are reasonable and pragmatic to minimise the complexity.
Keywords: Bus lanes; road roughness; bus dynamics; ride comfort; numerical simulation; MATLAB/Simulink

1. INTRODUCTION the most accurate results with errors within 10%

of ride comfort indicator as compared to non-
Vehicle dynamics models have long been
correlated output from simplified vehicle models,
developed for different purposes such as to
in which the simulated QCS resulted in more
design and optimise suspension system [1], to
than double of the measured values.
evaluate passenger ride comfort and vehicle
Vehicle dynamics are inherently different by
handling [2], and to evaluate road roughness
models and types of vehicles. The study results
conditions [3]. The most commonly-used vehicle
models are quarter-car-simulation (QCS), half- from car and truck dynamics may be not
car-simulation (HCS) and full-car-simulation applicable for bus dynamics, with the latter being
(FCS), of which the QCS has been used most asymmetric and in larger dimensions, being
often due to its simplicity. A comprehensive equipped with air suspension, and able to carry
summary of current approaches to evaluating more passengers onboard. The bus dynamics
road roughness has shown that there are eight model to evaluate road roughness conditions
QCS, two HCS and one FCS models which have along dedicated bus lanes or bus rapid transit
been used to develop various indices [4]. In (BRT) lanes have not been well investigated yet.
another study of car dynamics [5], the FCS gave To fill the research gap, this study aims at

397
developing several bus dynamics models and The lumped parameters of three bus models
comparing their dynamics behaviours to identify are shown in Figure 1.
the best mathematical model for evaluating road The ordinary differential equations (ODE)
roughness and ride comfort of bus passengers. describing the dynamic behaviour of the 3-
DOF QBM (vertical motion of mH, mS, mU)
2. METHODOLOGY can be derived as follows:
z H  k H  z H  zS   d H  z H  z S 
m H (1)
To simulate bus dynamics, technical data was
referenced to the single-decker bus model zS  k H  z H  zS   d H  z H  z S  
mS
Mercedes-Benz Citaro operating in urban areas. (2)
 k S  zS  z U   dS  z S  z U 
The bus body is asymmetric where the front axle
can carry up to 7.245t and rear axle can carry 12t m Uz U  k S  zS  z U   dS  z S  z U  
(3)
when fully laden at 19t. Air suspension stiffness  k U  z U  z R   d U  z U  z R 
was calculated using the dynamic spring rate as
a function of the change in effective area, and for 6-DOF HBM (vertical motion of mH1,
volume, and pressure from the design height of mH2, mU1, mU1, mS; pitch motion S):
the airmount isolator [6]. Damping ratio = 0.3   k Z  d Z
m H1Z H1 H1 H1 H1 H1 (4)
was chosen in favour of passenger ride comfort.   k Z  d Z
Passenger seat characteristics were selected from m H2 Z H2 H2 H2 H2 H2 (5)
[7] concerning ride comfort optimisation.   k Z  d Z  k Z  d Z 
mS Z S H1 H1 H1 H1 H2 H2 H2 H2
The experimental set-up was conducted (6)
 k Z  d Z  k Z  d Z
S1 S1 S1 S1 S2 S2 S2 S2
along the following steps:
 Selecting bus lane segments with different ISy
θS  a Hx1k H1ZH1  a Hx1d H1Z H1  a Hx2 k H2 ZH2 
road roughness levels;  a Hx2 d H2 Z H2  a Sx1k S1ZS1  a Sx1dS1Z S1  (7)
 Scanning the bus lane by a multi-laser  a k Z  a d Z
Sx2 S2 S2 Sx2 S2 S2
road profiler;
 Measuring onboard ride comfort index z U1  k S1ZS1  dS1Z S1  k U1ZU1  d U1Z U1
m U1 (8)
using a seat-pad accelerometer;   k Z  d Z  k Z  d Z
m U2 Z U2 S2 S2 S2 S2 U2 U2 U2 U2 (9)
 Developing 3-DOF QBM, 6-DOF HBM
and 9-DOF FBM to estimate ride comfort And likewise for the 9-DOF FBM (vertical
index using the same measured road motion of mH1, mH2, mU1r, mU1l, mU2, mS; pitch
longitudinal profiles; and motion S; and roll motion S, U2):
 Comparing the measured and calculated   k Z  d Z
m H1Z H1 H1 H1 H1 H1 (10)
ride comfort levels to evaluate the
  k Z  d Z
m H2 Z (11)
developed bus dynamics models. H2 H2 H2 H2 H2

Figure 1. Three mathematical bus dynamics models with different DOFs

398
   k Z  d Z  k Z  d Z    k Z  d Z
mS Z S H1 H1 H1 H1 H2 H2 H2 H2 m U2 Z U2 S2r S2r S2r S2r  k S2l ZS2l 

 k S1r ZS1r  dS1r Z S1r  k S1l ZS1l  dS1l Z S1l  (12)  dS2l Z S2l  k U2r ZU2r  d U2r Z U2r  (17)
 k Z  d Z
S2r S2r S2r  k Z  d Z
S2r S2l S2l S2l S2l  k Z  d Z
U2l S2l U2r S2l

ISy 
S  k H1ZH1a Hx1  d H1Z H1a Hx1  k H2 ZH2 a Hx2    k Z b  d Z b 
I U2x Φ U2 S2r S2r Sy2 S2r S2r Sy2

 d H2 Z H2 a Hx2  k S1r ZS1r a Sx1  dS1r Z S1r a Sx1   k S2l ZS2l bSy2  dS2l Z S2l bSy2 
(13) (18)
 k Z a  d Z a  k Z a
S1l S1l Sx1 S1l S1l Sx1 S2r 
S2r Sx2  k U2r Z U2r b Uy2  d U2r Z U2r b Uy2 
 dS2r Z S2r a Sx2  k S2l ZS2l a Sx2  dS2l Z S2l a Sx2  k U2l ZS2l b Uy2  d U2r Z S2l b Uy2
ISx ΦS  k H1ZH1b Hy1  d H1Z H1b Hy1  k H2 Z H2 b Hy2 

Bus model parameters related to sprung
 d H2 Z H2 b Hy2  k S1r ZS1r bSy1  dS1r Z S1r bSy1  mass, unsprung mass and human-seat are listed
(14) in Table 1, Table 2, and Table 3, in which
 k S1l ZS1l bSy1  dS1l Z S1l bSy1  k S2r ZS2r bSy2 
weight is in kg, distance is in m, damping is in
 dS2r Z S2r bSy2  k S2l ZS2l bSy2  dS2l Z S2l bSy2
Ns/m, stiffness is in N/m and Moment of Inertia
  k Z  d Z  k Z  d Z
m U1r Z U1r S1r S1r S1r S1r U1r U1r U1r U1r (15) is in kgm2; COM is the centre of mass and ‘Eq’
m Z  k Z  d Z  k Z  d Z
U1l

U1l S1l S1l S1l S1l (16)
U1l U1l U1l U1l is the equivalent value.

Table 1. Bus model parameters: suspension & unsprung mass

Parameter Acronym Value Parameter Acronym Value
Unsprung mass front mU1 662 Unsprung mass rear mU2 1333
Tyre stiffness front axle kU1 1,000,000 Eq-tyre stiffness rear axle kU2 2,000,000
Tyre damping front axle dU1 150 Eq-tyre damping rear axle dU2 300
Distance COM-front tyre bU1 1.39 Distance COM-rear tyre bU2 1.02
x-Moment of Inertia front axle IBUx1 350 x-Moment of Inertia rear axle IBUx2 620
Tyre contact length lcontact 0.3045

Table 2. Busmodel parameters: human & seat

Parameter Acronym Value Parameter Acronym Value
Human mass mH1 / mH2 70 Eq-seat stiffness kH1 / kH2 10,000
Eq-seat damping dH1 / dH2 330
x-distance COM-seat 1 aHx1 3.71 x-distance COM-seat 2 aHx2 2.09
y-distance COM-seat 1 bHy1 0.64 y-distance COM-seat 2 bHy2 0.60

Table 3. Busmodel parameters: sprung mass

Parameter Acronym Value Parameter Acronym Value
PGW mPGW 19,000 Kerb weight mKW 10,770
y-Moment of Inertia IBSy 150,000 x-Moment of Inertia IBSx 13,000

Vehicle width bveh 2.5 Wheelbase lWB 5.8

front suspension stiffness kS1 332,409 Eq-rear suspension stiffness
kS2 664,818
(max./min.) 189,418 (max./min.) 378,836
front suspension damping dS1 50,220 Eq-rear suspension damping dS2 71,023
(max./min.) 29,674 (max./min.) 41,965
x-distance COM-front axle aSx1 3.71 x-distance COM-rear axle aSx2 2.09
y-distance COM-front bSy1 0.64 y-distance COM-rear bSy2 0.60
suspension suspension

399
3. EXPERIMENTAL RESULTS AND turns were used to match and compare with
DISCUSSION outputs from vehicle model simulation.
Four recorded measurements at different
A straight segment of 2.3km bus lane was
selected for field measurement to reduce the road length were selected for comparison based
impact of lateral acceleration on the on their most comparable GPS data between
measurement data. Multi-laser road profiler the longitudinal road profiles (by multi-laser
was used to measure bus lane longitudinal profiler) and on-board acceleration
profiles with the sampling rate of every 25mm. measurement (by SV100A). For illustration,
The bus lane has the average IRIleft = 4.85, Figure 2 shows the comparison of time-series
IRIright = 4.21 (m/km) in which detailed IRIs frequency-weighted acceleration awz(t) and awz
range from 2 to 7 (m/km) for every 100m, between simulation and field-measurement at
meaning this segment contains different road different velocity (m/s) at segment 3. Figure 3
conditions from smooth to very rough shows the ratios of awz values (r = awz-
surfaces. To measure vertical acceleration, a simulation/awz-measurement) of all segment
seat-pad accelerometer SV100A was seated measurements. The average ratios of r = 120%
onto the seat cushion above the rear axle and (std = 9%), r = 107% (std = 5%) and r = 102%
loaded by a 70kg human, which is one of the
(std = 6%) were found for the three cases of
most critical locations of vertical acceleration
QBM, HBM and FBM, respectively. The 9-
according to [7]. The device is controlled
wirelessly by a mobile application; DOF FBM provides the best matching results
acceleration data was collected at a sampling while 3-DOF QBM overestimates the ratio,
rate of 750Hz and GPS data at 1Hz, which in and 6-DOF HBM produces in between results.

QBM, r = 134% HBM, r = 104% FBM, r = 102%

Figure 2. The frequency-weighted awz-simulation(t) and awz-measurement(t)at segment 3

Figure 3. r = awz_simulation/awz_measurement ratios at different segments and the average ratios.

400
 Overall, FBM provides the most reliable phenomenon of regular bus service, resulting
results, however, its modelling complexity is the in more reliable outputs.
highest in terms of data requirement and Future study will analyse and establish a
modelling algorithms. The outputs from QBM new Bus Ride Index (BRI) that can evaluate
are higher than that of HBM and FBM since more precisely bus passengers ride comfort and
QBM only considers vertical oscillation while road surface irregularity, which has yet to be
HBM takes pitch motion, and FBM takes both investigated in current approaches [4]. BRI
pitch and roll motions into account, will serve as a benchmark to estimate bus
respectively. This consideration helps reduce lane/BRT lane condition using the onboard
vertical acceleration since when one tyre/wheel accelerometer, which is more efficient (faster,
goes over the hump, another tyre/wheel still easier and less expensive) as compared to the
stays on the ground in case of HBM and FBM, conventional method using road profilers.
as contrasted to QBM. In specific cases, the
output from HBM and FBM can be similar ACKNOWLEDGEMENTS
when the right and left road profiles are quite
uniform, resulting in no roll motion contributing This work is financially supported by the National
to bus dynamics behaviours (such as the results Research Foundation Singapore under its Campus
from segments 3 and 4 in this study). for Research Excellence and Technological
From the comparison of awz(t) in the time Enterprise (CREATE) programme.
domain, there is always the trend that simulated
outputs are higher than field-measured results. 5. REFERENCES
This can be explained by the tyre being modelled Sekulić, D., &Dedović, V. The Effect of Stiffness
as a point contact of the mathematical bus and Damping of the Suspension System
models, which is different from the dual-wheel Elements on the Optimisation of the
rear axle. The real dual-wheel helps to reduce the Vibrational Behaviour of a Bus. International
Journal for Traffic & Transport Engineering 1
impact of road roughness, for example when one (4): 231–44 (2011)
tyre goes over a small pothole while another tyre Prem, Hans, and Geoff Ayton. Improved
stays on flat ground to keep the whole dual- Techniques for Assessing Ride Quality on
wheel in the flat, balanced position. Concrete Pavements. In 8th International
Conference on Concrete Pavements, 733–54.
Colorado, USA (2005).
4. CONCLUSION AND OUTLOOK Cantisani, Giuseppe, and Giuseppe Loprencipe.
Road Roughness and Whole Body Vibration:
This is the first time a comprehensive Evaluation Tools and Comfort Limits. Journal
comparison between three mathematical of Transportation Engineering 136 (2010)
dynamics models of a heavy-duty city bus (19t) Múčka, Peter. Current Approaches to Quantify the
has been conducted, resulting in Longitudinal Road Roughness. International
awz(QBM)>awz(HBM)>awz(FBM) over the same Journal of Pavement Engineering 17(8):659–
79 (2016).
road segments. However, the output difference http://dx.doi.org/10.1080/10298436.2015.101
is not that significant in the case of QBM as 1782
being investigated in [5]. Obviously, the correct Hamersma, Herman A. A Comparison of Quarter,
behaviours of bus dynamics are well captured by Half and Full Vehicle Models with
Experimental Ride Comfort Data. Pp. 1–7 in
the FBM, but this is the most complex model. Proceedings of the ASME 2015 International
Therefore, each bus model can be utilised Design Engineering Technical Conferences &
depending on the specific study purpose. Computers and Information in Engineering
This experimental study has explored a new Conference. Boston, Massachusetts, USA:
method to compare between simulation and ASME (2017)
Firestone. Engineering Manual and Design Guide
field-measurement data by matching GPS (2014).
locations and speed profiles. Importantly, the http://firestoneip.com/-
developed algorithms can incorporate speed /media/www/fsip/files/ManufacturingIndustria
changes that replicate the stop-and-go l/Designmanuals/EMDG_Imperial_2014.pdf

401
The 3rd Int. Conf. on Transport Infrastructure & Sustainable Development (TISDIC 2019). Construction Publishing House. ISBN: 978-604-82-2893-4.

Invited lecture 1: Aerodynamic retrofitting of truss

stiffening girder of long-span suspension bridge

Hiroshi Katsuchi1,, Hitoshi Yamada1, Shigeki Kusuhara2 and Taku Hanai2

1
Department of Civil Engineering, Yokohama National University, Yokohama, Japan
2
Honshu-shikoku Bridge Expressway Co. Ltd., Kobe, Japan

Corresponding author, E-mail: [email protected]

ABSTRACT: Many long-span suspension bridges in Japan have a truss stiffening girder and open-
grating on the road deck for aerodynamic countermeasure. Through more than 30-year operation,
maintenance and operational problems of corrosion of small members and dropping of small objects to
the lower floor have risen. In order to solve the problems, the removal or closure of the open grating was
investigated by a wind-tunnel test and flutter analysis. By recent advancement and understanding for
long-span bridge aerodynamics, three-dimensional effects and higher mode contribution were considered.
As a result, it was found that the partial closure of the center open grating is possible and further closure
is possible if aerodynamic countermeasures such as a center barrier are installed.
Keywords: Suspension bridge, open grating, flutter analysis, wind-tunnel test, center barrier.

1. INTRODUCTION tests. The final solution for the suspension

bridges was the adoption of a truss stiffening
Many long-span bridges were constructed to girder and open-grating floors in the middle
connect islands in Japan in 1980s and 1990s. and both ends of the road deck as shown in
Of those, the Seto-ohashi Bridge is the Figures 3 and 4.
remarkable one which connects the main island
with the Shikoku island. It is consists of 3
suspension bridges, 2 cable-stayed bridges and
1 truss bridge, and was completed in 1985 as
shown in Figures 1. The longest span is
1,100m of the Minami-bisan Seto Bridge as
shown in Figure 2. Feature of the Seto-ohashi
Bridge is a road and railway combined bridge
where the upper deck is used for road traffic
and the lower deck is used for railway as
shown in Figures 2 and 3.
As well known, long-span bridges are
susceptible to wind dynamic effects. Detailed
investigation for the aerodynamic stability was
done at the design stage by using wind-tunnel Figure 1. Seto-ohashi Bridge

405
 Since the bridge open to traffic in 1985,
maintenance and operational problems have
risen that corrosion of small members of the
open grating started and small objects dropped
from the road floor to the railway floor through
the open grating. These problems were not
considered at the design stage but have been
identified for the first time for 30-year
maintenance experience.
In order to solve the problems, one solution
will be the removal or closure of the open
grating. However, the closure of the open
Figure 2. Minami-bisan Seto Bridge
grating will decrease the aerodynamic stability
35.0 m of the bridge. Since the investigation of the
Grating aerodynamic stability at the design stage was
conducted by a 2D section model wind-tunnel
test, recent advancement and understanding for
long-span bridge aerodynamics might evaluate
13. the aerodynamic stability more accurately by
0
considering three-dimensional effects and
m
higher mode contribution (Katsuchi, et al.,
1998). In addition, there may be some
30.0 m aerodynamic countermeasures to enhance the
aerodynamic stability.
Figure 3. Cross section of truss stiffening girder
Based on these, investigation for the
aerodynamic stability of the Seto-ohashi
Bridge was done when the open grating is
closed by using a wind-tunnel test and
aeroelastic analysis. In addition, possible
countermeasures were sought to increase the
aerodynamic stability (Kusuhara, et al., 2017).

2. WIND-TUNNEL TEST
The Minami-bisan Seto Bridge (Figure 6) was
Figure 4. Open grating floor investigated by a 2D section model test in this
study. Aerodynamic stability of the bridge was
checked by a section model test at the design
stage as shown in Figure 7. The reference wind
speed for flutter stability is given as a 150-year
recurrence wind speed, 79.1m/s for -3deg to
+3deg angles of attack and reduced to 25m/s at
+-7deg. It was confirmed that the flutter critical
wind speed decreased at the positive angles of
attack (Honshu-shikoku Bridge Authority, 1976).
Figure 8 shows test cases in this study.
Section A is the current cross section where the
open grating is installed in the middle and both
Figure 5. Falling object on railway floor sides of the road deck. The test result is used to

406
confirm the accuracy and as the base line.
Section B is the cross section where the center
open grating is closed. Practically, a steel plate
is attached on the open grating on site.
Furthermore, the grating may be replaced by a 80
solid deck so that maintenance labor for small
Stability 70
elements of the open grating can be saved. requirement
Another test case such as all open gratings 60
closed was considered, however it was
expected that all grating closure would be 50 Wind speed
worse than the Section B. Therefore, only the (m/s)
center grating closure case (Section B) was 40
tested in this study. Section C is the cross
30
section of Section B with a center barrier of 0.8
m high which is an aerodynamic counter
(Angle of attack)
measure to enhance the flutter critical wind
speed. This is based on previous studies Figure 7. Wind-tunnel test result (flutter critical
(Hederine, et al., 1971& Enami, et al., 1977) wind speed vs. angle of attack) at design stage
on the efficiency of the center barrier. A thin Section-A
steel or plastic plate can be easily installed on
the center guard fence.
Table 1 shows test conditions. The wind-
tunnel test was conducted with a 1/80-scaled
Section-B Closed
section model as shown in Figure 9. A closed-
circuit wind tunnel of Yokohama National
University was used.

Table 1. Test condition

Truss width 0.375 m
Truss height 0.1625 m
Unit Mass 6.48 kg/m
Polar moment of 0.182 (kgm2)/m
inertia
Figure 8. Test cases
Frequency Heaving 1.97 Hz
Torsion 3.29 Hz
Damping Heaving 0.025 - 0.032
Torsion 0.026 - 0.033

Figure 6. General plan of Minami-bisan Seto

Bridge (in m) Figure 9. Section model of wind-tunnel test

407
3. TEST RESULTS OF AERODYNAMIC design requirement. Countermeasures must be
STABILITY taken to close the center open grating for the
maintenance reason. Figure 12 shows the result
Test results are shown in Figures 10-12. The
of the Section C installing the center barrier. The
current cross section (Section A) shows good
performance where flutter did not appear up to flutter critical wind speed at the angle of attack
the reference wind speed for all angles of attack, of +5 degree was 69 m/s which is lower than
as shown in Figure 10. On the other hand, flutter those of other angles of attack. However, the
took place at positive angles of attack (+3 and + design requirement for flutter at +5 degree is
5 degrees) below the reference wind speed in the 37.5 m/s, which is much lower than 69 m/s.
Section B, as shown in Figure 11. This is Therefore, it is expected that installing the center
caused by the closure of the center open barrier will be the aerodynamic countermeasure
grating. Since this result does not satisfy the when the center open grating is closed.

600 Original 5 Original

-5 deg -5 deg
-3 deg Vr = 79.1 m/s
-3 deg
Vertical response (mm)

Torsional response (deg)

4
0 deg
400 +3 deg 0 deg
3 +3 deg
+5 deg
+5 deg
2
200
1

0 0
0 20 40 60 80 100 0 20 40 60 80 100
Wind speed (m/s) Wind speed (m/s)

Figure 10. Response amplitude vs. wind speed (Section A)

600 Center grating closed 5 Center grating closed

-5 deg -5 deg
-3 deg Vr = 79.1 m/s
-3 deg
Torsional response (deg)

4
Vertical response (mm)

0 deg
400 0 deg
+3 deg
3 +3 deg
+5 deg
+5 deg
2
200
1

0 0
0 20 40 60 80 100 0 20 40 60 80 100
Wind speed (m/s) Wind speed (m/s)

Figure 11. Response amplitude vs. wind speed (Section B)

600 Center grating closed 5

Center grating closed
-5 deg -5 deg + 0.8m center brrier
+ 0.8m center barrier
-3 deg
-3 deg Vr = 79.1 m/s
Torsional response (deg)

4
Vertical response (mm)

0 deg Vr = 37.5 m/s

400 0 deg
+3 deg (for ±5 deg)
3 +3 deg
+5 deg
+5 deg
2
200
1

0 0
0 20 40 60 80 100 0 20 40 60 80 100
Wind speed (m/s) Wind speed (m/s)

Figure 12. Response amplitude vs. wind speed (Section C)

408
4. ANALYTICAL INVESTIGATION FOR 25 eigen modes. Figure 15 shows the flutter
PARTIAL CLOSURE OF CENTER GRATING critical wind speeds for different length of the
center open grating closure at angle of attack of
The wind-tunnel test showed that the center +3 degree. Since the positive angles of attack
open grating closure does not satisfy the design
exhibited flutter at lower wind speed, the
requirement for flutter. However, recent
flutter analysis was performed with flutter
advancement and understanding for long-span
derivatives of +3 angle of attack. Changing the
bridge aerodynamics pointed out the necessity of
combination of flutter derivatives incorporating
consideration of three-dimensional effects and
to the equation of motion between Sections A
higher mode contribution. This consideration
and B, flutter critical wind speed was obtained.
can be realized by a 3D multi-mode flutter
analysis. Using self-excited force coefficients As shown in Figure 15, the flutter critical wind
(flutter derivatives) measured by a wind-tunnel speed decreases as the length of the Section B
test, an equation of motion of the bridge increases. The top bar in Figure 15 corresponds
incorporating self-excited forces via flutter to the Section A applied to the entire bridge
derivatives is solved by a complex eigenvalue while the bottom bar corresponds to the
analysis, which gives flutter critical wind speed. Section B applied to the entire bridge. Since
Figures 13 and 14 show flutter derivatives the design requirement for flutter is 79.1 m/s of
associated with heaving and pitching motions the critical wind speed, the Section B (center
for the Sections A and B, respectively. Flutter open grating closure) can be allowed for the
derivatives were measured with the same side spans and 130m sections from the tower
section model and a free-vibration method for the center span (About 50% of the entire
(Yamada, et al., 1996). Flutter analysis was length). Remaining section in the center span
performed with a multi-mode flutter analysis of cannot close the center open grating.

1.0 0.30
LθR
0.8 LθI 0.25 MZR
0.6 0.20
0.4 0.15
0.2 MθR
0.10
0.0
LZR 0.05 MθI
-0.2
-0.4 0.00
-0.6 LZI -0.05
MZI
-0.8 -0.10
-1.0 -0.15
0 2 4 6 8 10 12 14 16 0 2 4 6 8 10 12 14 16
U/fB U/fB
Reduced wind speed Reduced wind speed

Figure 13. Flutter derivatives for Section A

6.0 0.5
5.0 0.4 MθR

4.0 0.3 MZR

LθR
3.0 0.2
2.0 0.1 MθI
LθI
1.0 0.0
0.0 -0.1 MZI
LZR
-1.0 -0.2
LZI
-2.0 -0.3
0 2 4 6 8 10 12 14 16 0 2 4 6 8 10 12 14 16
Reduced wind speed U/fB Reduced wind speed U/fB

Figure 14. Flutter derivatives for Section B

409
Figure 15. Flutter critical wind speeds for different length of center open grating closure
(Angle of attack: +3 degree)

The closure of the center open grating for 6. REFERENCES

the entire span is desirable from the
maintenance point of view. However, even Enami,Y. & Kitagawa, K. 1977. Flutter and
pressure distribution of truss stiffening girder
partial closure can reduce the potential risk for
with composite steel deck, Proc. of 32nd
dropping objects and save the maintenance Annual Convention of JSCE, JSCE, pp.197-
cost. It is also expected to increase the closure 198 (in Japanese).
length of the center open grating with Hederine, A.& Silano, L.G. 1971. Newport bridge
aerodynamic countermeasures such as the superstructure, Journal of the Structural
center barrier. Division, Vol. 97, Issue 11, pp.2653-2678.
Honshu-shikoku Bridge Authority 1976. Wind-
5. CONCLUSIONS resistant design code for Honshu-shikoku
Bridges (in Japanese).
Investigation on the possibility of the closure Katsuchi, H., Saeki, S., Miyata, T. & Sato, H.,
of the center open grating for a truss stiffening 1998. Analytical assessment in wind-resistant
girder of long-span suspension bridges in design of long-span bridges in Japan, Bridge
Japan was performed in order to reduce the Aerodynamics (Proc. of the Int. Symp. on
maintenance cost and potential risk of Advances in Bridge Aerodynamics), Larsen &
dropping objects. Wind-tunnel tests and Esdahl (eds), Balkema, pp.87-98.
aeroelastic flutter analysis showed that the Kusuhara, S., Katsuchi, H., Naito, M., Yamane,
A.&Machida, A. 2017. Study on aerodynamic
center open grating closure in the entire span
stability of long-span suspension bridge
is difficult while the partial closure (about considering economically maintenance, J. of
50% of the entire length) is possible. Structural Engineering, Vol. 63A, JSCE,
Furthermore, it was found that further or pp.430-437(in Japanese).
complete closure is possible if aerodynamic Yamada, H. & Miyata, T.1996. Introduction of a
countermeasures such as a center barrier are modal decomposition and reassemblage
installed. The center barrier can be relatively method for the multidimensional unsteady
easily installed on site without large economic aerodynamic force measurement, Proc. of
burden. This will lead to the reduction of the BBAA III.
potential risk for dropping objects and save
the maintenance cost.

410
The 3rd Int. Conf. on Transport Infrastructure & Sustainable Development (TISDIC 2019). Construction Publishing House. ISBN: 978-604-82-2893-4.

Invited lecture 2: Dynamic analysis of FGM conical shells

surrounded by pasternak elastic foundations

Nguyen Dong Anh1, Le Quang Vinh2,*, Nguyen Manh Cuong3 and Vu Quoc Hien4
1
Institute of Mechanics, VAST, Hanoi, Vietnam
2,4
Viet Tri University of Industry, Phutho, Vietnam
3
Hanoi University of Science and Technology, Hanoi, Vietnam

*Corresponding author, E-mail: [email protected]

ABSTRACT: A new continuous element (CE) formulation has been established for studying the vibration
of thick FGM conical shells on elastic foundation. The dynamic stiffness matrix has been constructed using
the closed form solutions of the differential equations of thick FGM conical shell taking into account the
shear deflection effects in order to obtain natural frequencies of the investigated structures. Using the CE
formulation a Matlab program is writtenfor validating theproposed model. Numerical results of natural
frequencies are compared to those obtained by other approaches. This paper emphasizes advantages of CE
model in dynamics of complex structures in terms of precision in any range of studied frequencies.
Key words: Natural frequency of conical shells, thick FGM shell, dynamics stiffness matrix, continuous
element method, elastic foundation, FGM shell.

1. INTRODUCTION The study of functionally graded materials

(FGMs) conical shells on elastic foundations is
Nowadays, conical shells made by Functional very scarce.
Graded Materialand resting on elastic So far,different computational methods are
foundation play an important role in high-tech available for the vibration analysis of conical
applications such as: nuclear, aerospace, shells, such as the Rayleigh-Ritz method,
medical industries. Due to the complexity of Differential Quadrature (DQ) method, Galerkin
the problem, the comprehension of dynamic method, Finite element method, wave
behaviors of such structures is of great propagation approach: Tornabene et al. (2009),
important in order to design and manufacture meshless method: Tornabene (2009) and Qu
safer and more economic FGM shell structures. (2013), general domain decomposition method,
Different approaches for analyzing free and various hybrid methods Qu (2013). An
vibrations of the cylindrical and conical shell unified solution method for the free vibrations of
have been applied. functionally graded cylindrical, conical shells
Most earthen soils can be appropriately and annular plates with general boundary
represented by a mathematical model from conditions is presented by Su et al. (2014) by
Pasternak, whereas sandy soils and liquids can means of the first-order shear deformation
be represented by Winkler’s model Pasternak theoryto formulate the theoretical model.
(1954) and Kerr (1964). The static and Rayleigh-Ritz procedure is used to obtain the
dynamic analyses of conical shells on elastic exact solution base on the energy functions of
foundations have been studied in recent years. those structures. Xie et al. (2014) provided a

411
simple yet efficient solution for the free vibration effects of various parameters such as
analysis of functionally graded (FG) conical geometrical parameters, material grading index,
shells and annular plates. A solution approach mechanical boundary conditions, Winkler and
based on Haar wavelet is introduced and the Pasternak foundation moduli on the natural
first-order shear deformation shell theory is frequencies of the FGM truncated conical shell
adopted to formulate the theoretical model. are also investigated. The buckling analysis of
Sofiyev et al. (2012) studied vibration the simply supported truncated conical shell
analysis of FGM truncated and complete made of functionally graded materials (FGMs)
conical shells resting on elastic foundations is presented by Sofiyev (2010). The FGM
under various boundary conditions by applying truncated conical shell subjected to an axial
the Galerkin method. The elastic foundations compressive load and resting on Winkler-
considered include the Winkler- and Pasternak- Pasternak type elastic foundations. The material
type elastic foundations. The FGMs are properties of functionally graded shells are
assumed to vary as power and exponential assumed to vary continuously through the
functions through the thickness of the conical thickness. The modified Donnell type stability
shells. Sofiyev et al. (2012) presented solutions and compatibility equations are solved by
for the vibration analysis of truncated conical Galerkin method.
shells made of functionally graded material and From the review of the literature, it appears
resting on Winkler-Pasternak foundations. The that despite a variety of methods for analytical
governing equations according to the Donnell’s andcomputational analysis of conical shell
theory are solved by Galerkin method and the structures, it is still of need and great
fundamental frequencies with or without two- significance to develop a simple and efficient
parameter elastic foundation have been numerical method for vibration analysis of
investigated. Dung et al. (2012) presented an FGM shells especially for complex combine
analytical approach to investigate the shell structures and for medium frequencies. In
mechanical buckling load of eccentrically the last several decades, Continuous Element
stiffened functionally graded truncated conical Method have been developed as a new
shells surrounded by elastic medium and powerful tool for analyzing structures in
subjected to axial compressive load and external medium and high frequencies. Casimir et al.
uniform pressure based on the classical shell (2007) have succeeded in building the DSM
theory and Galerkin method. Sofiyev et al. for thick isotropic plate and shells of
(2011) applied Galerkin method to the vibration revolution. The CE models for composite
analysis of laminated orthotropic shells with cylindrical shells and conical shells presented
different boundary conditions and resting on in works of Thinh and Cuong (2013) imposes a
elastic foundations. A two-parameter foundation considerable advancement of the study on
model (Pasternak type) is used to describe the CEM for metal and composite structures.
shell-foundation interaction, from which Recently, Thinh et al. (2013) proposed new
Winkler foundation model can be easily Continuous Elements models for thick
obtained as a particular case. The modified composite conical shells.
Donnell type dynamic stability and The purpose of this paper is presenting a
compatibility equations have been obtained for new continuous element which takes into
laminated orthotropic truncated conical shells account both the rotational inertia and shear
resting on elastic foundations. Asanjarani et al. deformations effects for the vibration analysis
(2014) considered the free vibration analysis of of thick FGM conical shells resting on elastic
a truncated conical shell, which is functionally foundation. The natural frequencies obtained
graded in two directions. In this study, FSDT in by using this model have been compared with
conjunction with the Hamilton’s principle are those of other method and a good agreement
employed to extract the governing equations. was observed. The effects of parameter of
These equations are solved by using DQM and shells as well as foundation have been

412
envisaged. The advantages of the CE model in which the power-law exponent p is a
have been confirmed: high precision and positive real number (0  p ) and the
suitable for studying any range of frequencies parameters a, b, c dictate the material
variation profile through the functionally
2. THEORETICAL FORMULATIONS graded shell thickness. It is assumed that the
2.1 Description of the model sum of the volume fractions of the two basis
components is equal to unity, i.e., Vc + Vm = 1.
Let’s investigate the FGM conical shell with Therefore, according to the relations defined
(x, θ, z) coordinates, as shown in Fig. 1. Where in Eq. (2), when the power-law exponent p is
x is the coordinate long the shell generators set equal to zero (i.e., p = 0) or equal to
with the origin placed at the middle of the infinity (i.e., p = ), the FGM material
generators, θ is the circumferential coordinate,
becomes the homogeneous isotropic material,
and z is the perpendicular to the shell surfaces.
expressed as:
R1 and R2 are the inner and outer radius, h is
the thickness, L is the length of the shell and p  0  Vc  1, Vm  0  E ( z )  Ec ,  ( z )  c ,  ( z )  c
the radius coordinate R(x) of a point M inside p    Vc  0, Vm  1  E ( z )  Em ,  ( z )  m ,  ( z )   m
the shell is calculated as: R(x) = R1+x. Whereas the composition of ceramic(M1)
This shell is surrounded by a Winkler and metal (M2) is linear for p = 1. The
elastic foundation having a foundation variations of the volume fraction Vc through
stiffness kw or by a Pasternak foundation with the shell thickness for different values of the
the foundation stiffness kw and shear layer power-law exponent p are illustrated in
stiffness kp. Such shell is the basic continuous Figure 2. In Figure 2 the classical volume
shell element to contribute a steppedaxis-
fraction profiles, such as those reported in
symmetric shell surrounded by two above
literature Pradhan (2000), are presented as
types of elastic foundations.
special cases of the general distribution laws
Typically, FGM shellsmade from a mixture
by setting a = 1 and b = 0. As can be seen
of two material phases. In this paper, it is
from Figure 2a, for the first distribution
assumed that the FGM shells are made of a
FGMI(a=1/b=0/c/p) the material composition is
mixture of ceramic and metal. Young’s
modulus E(z), density ρ(z) and Poisson’s ratio continuously varied such that the bottom
 (z) are assumed to vary continuously through surface (z/h = -0.5) of the shell is M1 rich,
the shells thickness and can be expressed as a whereas the top surface (z/h = 0.5) is M2 rich.
linear combination: The volume fraction Vc decreased from 1 at
z/h = -0.5 to zero at z/h = 0.5. Figure 2b
E ( z )  ( Ec  Em )Vc  Em
(1) shows that for the second distribution
 ( z )  ( c  m )Vc   m
FGMII(a=1/b=0/c/p) the top surface (z/h = 0.5) of
 ( z )  ( c  m )Vc   m
the shell is M1 rich, whereas the bottom
in which the subscripts c and m represent the surface (z/h = -0.5) is M2 rich, instead. When
ceramic and metallic constituents, respectively, the volume fraction exponent is increased, the
and the volume fraction Vc follows two general content of M1 in FG layer decreases. More
four-parameter power-law distributions details on the material variation profile of
Tornabene (2009): FGMs with different parameters (i.e., p, a, b,

p
c) are available in Tornabene (2009).
1 z 1 z 
c

FGM I ( a / b / c / p ) : Vc  1  a    b    So far, all the needed parts of the first-order

  2 h   2 h  
(2) shear deformation shell theory (FSDT) are
p
 1 z 1 z 
c
presented, and they may be combined to obtain
FGM II ( a / b / c / p ) : Vc  1  a    b   
  2 h   2 h   the desired form of the equations of motion.

413
Figure 1. Geometry parameters of FGM conical shell surrounded by Pasternak elastic foundations

Figure 2. Variation of the volume fraction Vc through the thickness of a shell for different values of
power-law exponent p: (a) FGMI(a=1/b=0/c=2/p); (b) FGMII(a=1/b=0/c=2/p).

2.2 Kinematic relations and stress resultants respectively; x and  represent the transverse
normal rotations of the reference surface about
On the basis of the shell model presented the - and x-axis. t is the time variable.
previously and the assumptions of moderately The linear strain-displacement relations in
thick shell theory, the displacement the shell space are defined as:
components of an arbitrary point in the FG u0 
x  ; kx  x
shell for the first-order shear deformation x x
1 v  1  x  sin
theory are expressed in terms of the     u0 sin  0  w0 cos ; k x    
R   R  x R (4)
displacements and rotation components of the v 1 u0 sin 1  
 x  0   v0 ; k   x sin   
middle surface as given below Thinh and x R  R R  
Cuong (2013): w0  cos 1 w0
 xz  x ;  z  v0   
ux,  , z, t   u0 x,  , t   z x x,  , t  x R R 
vx,  , z, t   v0 x,  , t   z x,  , t  (3) Based on Hooke’s law, the stress–strain
wx,  , z, t   w0 x,  , t  relations of the shell are written as:
where u, v and w are the displacement  x  Q11 ( z ) Q12 ( z ) 0 0 0   x 
  Q ( z ) Q ( z ) 0     
components in the x,  and z directions,     12 11 0 0
(5)
 
respectively; uo, vo and wo are the middle 
 x    0 0 Q66 ( z ) 0 0   x 
   0 0   xz
 
surface displacements of the shell in the axial,  xz  
0 0 Q66 ( z )
 
circumferential and radial directions,  z   0 0 0 0 Q66 ( z )   z 

414
where the elastic constant Qij(z) are functions 2.3 Equations of motion
of thickness coordinate z and are defined as:
By means of Hamilton’s principle, the
E( z)  ( z)E( z) E( z)
Q11 ( z )  , Q12 ( z )  , Q66 ( z )  equilibrium equations of motion based on the
1   2 ( z) 1   2 ( z) 2[1   ( z )] (6) firs-order shear deformation shell theory (FSDT)
The stress and moment resultants are given as: in terms of the force and moment resultants can
h/2 be written as Casimir et al. (2007):
N x , N  , N x  , Q x , Q     xx ,   ,  x ,  xz ,   z dz (7)
h / 2  N x sin  1  N x
 ( N x  N )   I 0 u0  I1x ,
h/2
x R R 
M x , M  , M x     xx ,   , x zdz (8)  N x

2 sin 
N x 
1 N 

cos 
Q  I 0 v0  I1 ,
h / 2
x R R  R
where Nx, Nθ and Nxθ are the in-place force Q x 1  Q sin  cos   2w
  Qx  N  k w w  k p ( 2 (11)
x R  R R x
resultants, Mx, Mθ and Mxθ are moment 2
sin   w 1  w
resultants, Qx, Qθ are transverse shear force  
R x R 2  2
)  I0w 0 ,
resultants. The shear correction factor k is  M x sin  1  M x
 (M x  M  )   Q x  I1u0  I 2x ,
computed such that the strain energy due to x R R 
 M x 2 sin  1 M 
transverse shear stresses in Eq. (9) equals the  M x   Q  I1v0  I 2 .
x R R 
strain energy due to the true transverse stresses
predicted by the three-dimensional elasticity Where I 0 , I1 , I 2   h / 2  ( z )1, z1 , z 2 dz , in which
h / 2
theory Reddy (2003). Since the shear correction (z) is the density of the shell per unit middle
factor for a FGM shell depends on shell surface area. I0, I1 and I2 are the mass inertias.
parameters, such as material properties and
boundary conditions, it is difficult to obtain the 3. DYNAMIC STIFFNESS MATRIX
accurate value of the shear correction stresses. FORMULATION FOR FGM
In this paper, the shear correction factors k will CONICAL SHELL
be uniformly selected by 5/6 Tornabene (2009).
Substituting Eqs. (6)-(7) into Eqs. (8) and (9) The chosen state-vector isyT = {u0, v0, w0, φx,
following constitutive equations relating the φ, Nx, Nx, Qx, Mx, Mx}T. Next, the Fourrier
force and moment resultants to strains and series expansion for state
curvatures of the reference surface are given in variables is written as:
the matrix form: uo ( x, , t ), wo ( x, , t ), ( x, , t ), N x ( x, , t ), Qx ( x, , t ), M x ( x, , t )T
u 

 N x   A11 A12 0 B11 B12 0 0 0   x   m ( x), wm ( x), m ( x), N xm ( x), Qx m ( x), M xm ( x) cos me it
T
(12)
 N  A 0    m1
    12 A11 0 B12 B11 0 0
 N x   0 0 A66 0 0 B66 0 0  x  vo ( x, , t ), x ( x, , t ), N x ( x, , t ), M x ( x, , t )
T

     (9) 
 M x   B11 B12 0 D11 D12 0 0 0  kx 
v ( x), x m ( x), N x m ( x), M x m ( x) sin me it
T
     m
 M  B12 B11 0 D12 D11 0 0 0  k  m1

M x   0 0 B66 0 0 D66 0 0 kx 
    
0  xz 
where m is the number of circumferential wave.
 Qx   0 0 0 0 0 0 k.F44

Q   0
    0 0 0 0 0 0 k.F55  z  Substituting (13) in equations (10) and (11), a
system of ordinary differential equations in the
The structure materials employed in the x-coordinate for the mth mode can be expressed
following study are assumed to be functionally in the matrix form for each circumferential
graded and linearly elastic. So, the extensional mode m as Thinh and Cuong (2013):
stiffness Aij, the bending stiffnessDij, and the dy m
extensional-bending coupling stiffness Bij are  Am y m (13)
dx
respectively expressed as:
h/2 h/2
With Am is a 10x10 matrix. The dynamic
Aij  h / 2
Qij ( z ) dz Bij   h / 2
z.Qij ( z )dz (10) transfer matrix Tm is evaluated as:
h/2 h/2
Dij  z 2Qij ( z )dz , i, j  1,2,6 Fij  Qij ( z ) dz , i , j  4,5 L

h / 2 h / 2  Am  dx T T12  (14)

Tm    e 0   11
T21 T22 
where k is the shear correction factor (k=5/6)

415
 Finally, the dynamic stiffness matrix K(ω) 4.1 Validation of the present model
for annular plate is determined by Casimir et
al. (2007): The proposed continuous element model will
be validated by comparison with solutions
 T121T11  T121  (15)
K m     1 1 
available in the literature with other method
T21  T22T12 T11 T22T12 
results.First, to check the accuracy of the
Natural frequencies will be extracted from present method on the vibration analysis of
the harmonic responses of the structure by FGM conical shells, Table 2 and 3 shows the
using the procedure detailed in Thinh and first frequencies for Si3N4/ Ni FGM conical
Cuong (2013). shells with differentkw, kp exponents. The
geometrical parameters of the FGM conical
4. NUMERICAL RESULTS AND DISCUSSION shells are R1/h = 100; L = 2R1;  = 30◦.
The present exact procedure may be applied to In Table 2variations of the lowest
investigate the effects of various geometrical dimensionless frequency parameter for the
and material properties such as foundations FGM truncated conical shells with different
moduli, the power law index and different compositional profiles, versus the Winkler
boundary conditions. Two Configuration of foundation stiffness kw and the shear subgrade
functionally graded material are used with the modulus kp are presented. It is observed that
material properties listed in Table 1. dimensionless frequency parameter of
Table 1. Material properties of the used FG plate truncated conical shells increases gradually
with increasing kw or kp separately or together
Material FGM1 FGM2 for all compositional profiles. It is also obvious
Properties Si3N4 Nickel Zirconia Aluminum that for their small to intermediate values,
E(Gpa) 322.27 205.098 168 70 bothWinkler and shearing layer elastic
µ 0.24 0.31 0.3 0.3 coefficients have significant effects on the
(kg/m3) 2370 8900 5700 2707 lowest dimensionless frequency parameters.

Table 2. Comparison of the lowest dimensionless frequency parameter for FGM conical shells
resting on the Winkler-Pasternak foundations with moduli kw and kp different.
FGM1I(a=1/b=0/c=2/p=1) FGM1I(a=1/b=0/c=2/p=2) FGM1II(a=1/b=0/c=2/p=2)
kw Sofiyev Sofiyev Sofiyev
(N/m3) kp(N/m) (2012) CEM Erro (2012) CEM Erro (2012) CEM Erro
0 0 0.0997 0.0970 2.71 0.0887 0.0855 3.58 0.1139 0.1116 2.02
6
510 0 0.1103 0.1085 1.66 0.0988 0.0970 1.82 0.1255 0.1241 1.10
1105 0.1146 0.1116 2.62 0.1024 0.1001 2.22 0.1306 0.1283 1.77
2.5105 0.1198 0.1179 1.62 0.1072 0.1053 1.73 0.1362 0.1335 1.98
5105 0.1279 0.1262 1.33 0.1148 0.1137 0.97 0.1451 0.1429 1.52
7
110 0 0.12 0.1189 0.91 0.1078 0.1064 1.31 0.1361 0.1356 0.37
1105 0.124 0.1220 1.59 0.1112 0.1095 1.52 0.1408 0.1387 1.48
2.5105 0.1287 0.1272 1.13 0.1156 0.1147 0.75 0.146 0.1450 0.70
5105 0.1363 0.1356 0.52 0.1227 0.1220 0.55 0.1543 0.1533 0.63
7
510 0 0.1797 0.1815 0.99 0.1631 0.1648 1.04 0.2017 0.2044 1.35
1105 0.1823 0.1836 0.70 0.1653 0.1669 0.96 0.2049 0.2065 0.79
2.5105 0.1856 0.1867 0.59 0.1684 0.1700 0.96 0.2085 0.2107 1.05
5105 0.191 0.1930 1.02 0.1733 0.1752 1.11 0.2144 0.2159 0.70
*Error(%)=(CEM− Sofiyev)/ Sofiyev×100.

416
 However, for the large values of the Winkler Table 3. Variations of natural frequency for
elastic coefficient, the shearing layer elastic FGM truncated conical shells with and
coefficient has less effect and the results without elastic foundations versus the ratio
essentially become independent of it. The L/R1(h=0.01, R1/h = 100,  = 30).
effect of the compositional profiles on the
lowest dimensionless frequency parameters kw=0 kw=107N/m3 kw=107N/m3
reduce, as the considering affect of elastic L/R1 kp=0 kp= 0 kp= 2.5105N/m
foundations. Furthermore, for the small values FGM1I(a=1/b=0/c=2/p=)
of kw and kp the variations of volume fractions
0.5 305 313 323
of the FGM have a considerable influence on
the dimensionless frequency parameter. When 1 166 179 188
the shear subgrade modulus kp is zero and the 2 93 114 122
Winkler foundation stiffness kw is changed, the 3 68 95 101
effect on the lowest dimensionless frequency
FGM1II(a=1/b=0/c=2/p=2)
parameter. As the Winkler foundation stiffness
kw issmall, the FG has a considerable influence 0.5 351 361 371
on the dimensionless frequency parameter. 1 191 205 215
Natural frequencies computed by CEM are 2 107 130 139
compared withthose of Sofiyev (2012) and
obtained errors vary from 0.37% to 3.58%. 3 78 107 115
Therefore, this continuous element model is
reliable and effective to study FGM truncated Table 4. Variations of natural frequency for
conical shells surrounded by Pasternak elastic the FGM truncated conical shells with and
foundations. without elastic foundations versus the ratio,
R2/R1 (h = 0.01m, R1= 2;  = 30◦).
4.2 Influences of shell parameters
kw=0 kw=107N/m3 kw=107N/m3
In this section, the influences of various shell
R2/R1 kp=0 kp= 0 kp= 2.5105N/m
parameters and elastic foundations on the
dynamic behaviour of the FGM truncated FGM1I(a=1/b=0/c=2/p=)
conical shells will be studied such as length- 2 23 57 61
to-radius ratio, radius ratio, semi-vertex angle 3 13 54 55
(), and boundary conditions. Different
4 9.5 52 53
stiffness values of kw, kp are also taken into
account in this study. FGM1I(a=1/b=0/c=2/p=1)
2 33 74 78
4.2.1 Effect of shell parameters and elastic
foundations on natural frequencies 3 19 68 70
4 13 66 68
In Table 3, variations of the lowest natural
frequency of FGM truncated conical shells with Variations of lowest values of
or without elastic foundations, versus the ratio, natural frequency of FGM1I(a=1/b=0/c=2/p=∞),
L/R1, are presented. As the ratio, L/R1, increases, FGM1I(a=1/b=0/c=2/p=1) truncated conical shells
the lowest natural frequency decreases. The with or without the Winkler–Pasternak
effects of compositional profiles on the lowest foundations versus the ratio, R2/R1, are
dimensionless frequency parameter of the
tabulated in Table 4. As the ratio, R2/R1,
truncated conical shell resting on elastic
increases and (kw, kp=0), natural frequency and
foundations decrease, as the ratio, L/R1,
corresponding wave numbers decrease
increase. On the other hand, the effect of elastic
foundations on the lowest natural frequency is for both FGM1I(a=1/b=0/c=2/p=∞) and
increased as the ratio, L/R1, increase. FGM1I(a=1/b=0/c=2/p=1) truncated conical shells.

417
As the ratio, R2/R1, increases, the lowest 4.2.2 Effect of elastic boundary conditions on
natural frequency, (kw=107N/m3, kp=0) and natural frequencies
(kw=107N/m3, 5
kp=2.510 N/m), increase,
whereas, corresponding wave numbers
decrease for both homogeneous and FGM
truncated conical shells with the Winkler–
Pasternak foundations. Consequently, it can
be seen that increasing, R2/R1, the sensitive of
the natural frequency with respect to elastic
foundations coefficients reduce.

Figure 4. Variation of natural frequency of

FGM2I(a=1/b=0/c=2/p=2) conical shell
versus large and small radii (R1=1,
h/R1=0.01, kw=1107N/m3,
5
kp=2.510 N/m ).

In Figure 4, the variations of natural frequency

versus large and small radii ratio are investigated.
As one can observe,when increasing large to
small radii ratio, natural frequency decreases and
Figure 3. The variation of natural frequency versus the effect of boundary conditions for higher
Winkler parameters in terms of different values of the ratio becomes less. Increasing radii
values of Pasternak parameters and (C-C) ratio from large to small causes semi-vertex angle
boundary condition (R1=1, h/R1=0.01, increase and as previously noted, natural
L/R1=2,=30o, FGM1I(a=1/b=0/c=2/p=2), frequency parameter has reverse relation with
kw=5107 N/m3, kp=5106 N/m). semi-vertex angle which could describe non-
dimensional frequency parameter characterize.
Figure 3 depicts the variation of natural
frequency versus Winkler parameter in terms 4.2.3 Effect of semi-vertex angle and the indix
of different values of Pasternak parameters and p on natural frequencies
considered boundary conditions. For values of The variation of lowest values of natural
kw  1107 N/m3 we see that when kp increases, frequency of FGM truncated conical shells for
the natural frequency of the shell increases, different compositional profiles, with or
with 1107 N/m3 < kw < 11010 N/m3 when kw without the two-parameter foundation versus
increases, natural frequency increases rapidly the semi-vertex angle, , are given in Figure 5.
and kp has little effect on natural frequency The foundations moduli are kw = 5107N/m3 or
because the foundation is relatively hard, when kp = 5 × 105N/m in Fig. 5. As the semi-vertex
kw  11010 N/m3, the natural frequency is angle (), increases, the values of natural
constant due to elastic foundation turned to frequency increase as   45o and decrease as
rigid foundation. Significant effects of the  > 45o for the homogeneous and FGM
Winkler–Pasternak foundations on natural truncated conical shells with or without the
frequency of the investigated FGM conical Winkler–Pasternak foundations. As the semi-
shell could be distinguished when comparing vertex angle rises, the effects of the Winkler-
with solutions of the considered structure Pasternak foundations on the lowest natural
without foundation (kw = kp = 0). frequency increase.

418
 the natural frequencies of thick FGM conical
shells on elastic foundation with high
accuracy for any frequency range. Effects of
shell and foundation parameters are discussed
in detailed. The developed Continuous
Element can be used efficiently for the
analysis of FGM shell on elastic foundation in
medium and high frequencies where almost
other current methods meet grand difficulties
due to the important number of meshing
elements. The present Continuous Element
can be expanded to solve the problem of FGM
Figure 5. Variations of natural frequency of shells-containing flowing fluid, shell with
FGM2I(a=1/b=0/c=2/p) conical shells without an damping or shells with stiffeners.
elastic foundation versus the semi-vertex
angle, (R1=1, R1/h= 100; L = 2R1). 6. ACKNOWLEDGEMENT

The variation of lowest values of natural The authors express deep thanks to the
frequency of FGM truncated conical shells for Reviewer. The support of VAST under the
different compositional profiles, with or code NCVCC03.07/19-19 to the first author
without the two-parameter foundation versus is acknowledged.
the semi-vertex angle, , are given in Figure 5.
The foundations moduli are kw = 5107N/m3 or 7. REFERENCES
kp = 5×105N/m in Figure 5. As the semi-vertex Asanjarani, A., Satouri, S., Alizadeh, A.
angle(), increases, the values of natural &Kargarnovin, M. H. 2014. Free vibration
frequency increase as   45o and decrease as analysis of 2D-FGM truncated conical shell
 > 45o for the homogeneous and FGM resting on Winkler–Pasternak foundations.
truncated conical shells with or without the Proc IMechE Part C: J Mechanical
Winkler-Pasternak foundations. As the semi- Engineering Science, 1–22.
vertex angle rises,the effects of the Winkler- Casimir, J. B., Cuong, N. M. 2007. Thick shells of
Pasternak foundations on the lowest natural revolution: Derivation of the dynamic stiffness
frequency increase. matrix of continuous elements and application
to a tested cylinder. Computers & structures,
85, 1845-1857.
5. CONCLUSION
Dung, D. V., Hoa, L. K., Nga, N. T. 2013. On the
In this research, a new Dynamic Stiffness stability of functionally graded truncated
Matrix has been successfully constructed for conical shells reinforced by functionally
thick FGM conical shells resting on Pasternak graded stiffeners and surrounded by an elastic
medium. Composite Structures, 108, 77-90.
elastic foundation. This new FGM Continuous
Kerr, A. D. 1964. Elastic and visco-elastic
Element completes the library of existing
foundation models. J. Appl.Mech, 31, 491–498.
Dynamic stiffness matrix for metal and Pasternak, P. L. 1954. On a new method of
composite structures. Our model takes into analysis of an elastic foundation by means of
account the effect of shear deformations and two foundation constants, Gos. Izd. Lit.
rotational inertia. An excellent agreement is poStroit I Arkh. Moscow, USSR.
noticed between the results obtained by the Pradhan, S. C., Loy, C.T., Lam, K. Y., Reddy, J. N.
presented model and those of Finite Element 2000. Vibration characteristics of functionally
Method. Various test cases confirm that graded cylindrical shells under various
Continuous Element Method allows computing boundary conditions. ApplAcoust , 61, 111–129.

419
Qu, Y. G., Long, X. H., Yuan, G. Q. 2013. Meng Su, Z., Jin, G. Y., Shi, S. X., Ye, T. G., Jia, X. Z.
G. Aunified formulation for vibration analysis 2014. A unified solution for vibration analysis
of functionally graded shells of revolution of functionally graded cylindrical, conical
with arbitrary boundary conditions. Compos shells and annular plates with general boundary
Part B: Eng, 50, 381-402. conditions. Int J MechSci, 80, 62-80.
Reddy, J. N. 2003. Mechanics of laminated Thinh, T. I., Cuong, N. M. 2013. Dynamic stiffness
composites plates and shells. New York: CRC matrix of continuous element for vibration of
Press. thick cross-ply laminated composite
Sofiyev, A. H. 2010. The buckling of FGM cylindrical shells. Compos Struct, 98, 93-102.
truncated conical shells subjected to axial Tornabene, F. 2009. Free vibration analysis of
compressive load and resting on Winkler- functionally graded conical, cylindrical shell
Pasternak foundations. International Journal and annular plate structures with a four-
of Pressure Vessels and Piping, 87, 753-761. parameter power-law distribution. Comput
Sofiyev, A. H., Kuruoglu N. 2011. Natural Methods ApplMechEng, 198, 2911-35.
frequency of laminated orthotropic shells with Tornabene. F., Viola, E., Inman, D. J. 2009. 2-D
different boundary conditions and resting on differential quadrature solution for vibration
the Pasternak type elastic foundation. analysis of functionally graded conical,
Composites: Part B, 42, 1562-1570. cylindrical shell and annular plate structures. J
11. analysis of FGM truncated and complete Sound Vib ,328, 259-90.
conical shells resting on elastic foundations Xie, X., Guoyong, J. G. Y., Ye, T. G., Liu, Z. G.
under various boundary conditions. J Eng 2014. Free vibration analysis of functionally
Math, 77, 131-145. graded conical shells and annular plates using
Sofiyev, A. H., Schnack E. 2012. The Vibration the Haar wavelet method. ApplAcoust, 85,
Analysis of FGM Truncated Conical Shells 130-142.
Resting on Two-Parameter Elastic
Foundations. Mechanics of Advanced
Materials and Structures, 19, 241-249.

420
The 3rd Int. Conf. on Transport Infrastructure & Sustainable Development (TISDIC 2019). Construction Publishing House. ISBN: 978-604-82-2893-4.

Advanced solutions to increase the efficiency in the use of

pedestrian bridge systems in Hanoi

Nguyen Minh Ngoc1,*,Vu Anh2

1
Faculty of Urban Environmental and Infrastructural Engineering,Hanoi Architectural University,
Hanoi, Vietnam
2
Urban management Department, Hanoi Architectural University, Hanoi, Vietnam

* Corresponding author, Email: [email protected]

ABSTRACT: Traffic is extremely heavy in Hanoi, means of transport are interwoven, especially during
morning and afternoon in the rush hour, the psychology of passers-by want to move faster and awareness
of following traffic laws is not good, so the process of passing across a road which made for very
dangerous circumstances or it has disrupted traffic. Therefore, the pedestrian bridge systems have some
useful insights and creating a “civilized” urban environment. Promoting social mobilization concerning
build urban of public works, Hanoi city has planned to exchange the art on the footbridge (ten years on 45
footbridges) for public and social welfare works. However, there are many factors, which affect using the
footbridge at present like uncoordinated design, the illegal encroachment on footbridge for personal
purposes, awareness of people in using footbridges. This has reduced the practical effect of the footbridge
systems. This paper presents the significance and importance of footbridges in the urban, this factors has
led to limit the use of flyovers and the management of the footbridge today. At the same time, offering
some solutions on planning and managing effective footbridges in urban.
Keywords: Transport, footbridges, pedestrian bridge, urban, managing, planning.

1. URBAN TRAFFIC CONTEXT IN HANOI development. Traffic congestion, traffic

accidents, environmental pollution, psychological
Hanoi capital has the highest speed of inhibition when participating in traffic of people
urbanization in the country (Ngo Thang Loi are a widespread problem in Hanoi capital.
2009), with estimated growth in store numbers
of more than 3%, it is more than 10 million
people in 2020, In addition, personal vehicle
system also grows with motorbike growth of
about 13% / year, cars about 21% / year (Le Anh
Tu 2006), while the urban infrastructure system
has not changed much, a number of projects for
many years are still "on paper" and many
"bottlenecks" on the road can not be opened.
Thus, Hanoi's road system is in the overloaded
condition Because the development of transport
system is much slower than the population Figure 1. Ha Noi population growth chart
growth rate and the speed of socio-economic (source: https://vietnammoi.vn/)

421
 In 2018, the ratio of land for transport in
Hanoi is only about 9.38%, meanwhile, the
city’s proportion of transport land in the law
prescribed reach at 16% to 26% of the urban
land, the transport land was introduced about
20% to 25% in advanced countries
(Decision 519/QD-TTg2016).
Urban transport situation in Hanoi is
extremely chaotic with the high-volume of
traffic in the narrow road. The motor vehicle Figure 2. The footbridge on Vo Van Kiet road,
driver's consciousness always needs to go Hanoi, a place that combines going on
fast and benefit himself. Therefore, crossing foot, riding motobikes and bicycles
road has become a leading guide when
travelling to Hanoi, this guide is even in the
traveller's handbook.
It has arranged many crossroad solutions on
the streets in Hanoi. The typical example is the
zebra crossing (it is now fairly generalized), in
particular, there was a time when the priority
traffic lights were arranged for crossing road
(when pedestrians need to cross the street,
press the traffic light button to turn on, the
vehicles stop to let pedestrians cross the
Figure 3. The footbridge with roof source:
road).However, this solution was quickly http://www.baogiaothong.vn
eliminated because It has a big effect on the
traffic flow. Later, a new crossroad system was
builtthat it is footbridge with steel structures; it
is located at essential locations for crossing.
The first footbridge was built in front of Bach
Mai Hospital - Giai Phong Road in 2007 with a
value of 4.1 billion. Since then, the footbridge
systems have become an outstanding solution;
number of footbridges are gradually increasing.
At the same time, it has begun to refine and
extend their role. By 2018, the footbridge and
tunnel road system has reached 45 units.
At present times, the footbridge is used not Figure 4. The footbridge without roof and a banner
advertisement of Vinasing
only for pedestrians but also designed for use
by motorbikes and bicycles, for example, the However, currently the pedestrian bridge in
footbridge on Vo Van Kiet road, Hanoi city Hanoi has yet to promote its optimal role,
was inaugurated in 2018. people sometimes still hesitate to use this
The footbridge of the urban transport system transportation system, sometimes crossing the
is pre-eminent feature and a break-through to road right under the footbridge, or the
ensure traffic safety, reduce traffic congestion footbridges are misused (sales, sleep,
and show urban civilization. In the world, the advertising stickers and so on) especially
footbridge is also symbolic for the city with its sanitation and hygiene awareness of users of
modern architectural system and its own flyovers. Besides, the footbridges have almost
artistic features (Antonio et al. 2018). never reached design capacity (Vu Anh 2015).

422
 In order for the footbridges to promote its 2.2 Classification of material
inherent role and become symbol of civilized Concrete footbridge
urban traffic, it needs more technical and Steel footbridge
social solutions for promoting people to use the Footbridge of steel structure, bridge surface
footbridges more often (Mauro et al. 2007). covered with asphalt
2.3 Classification of purpose
Bridge for Pedestrian
Multi-purpose footbridge

3. CURRENT STATUS OF USING

FOOTBRIDGE IN HANOI
Assessing the extent of using footbridge, then
devising plan solutions and promoting the use
of footbridge in urban areas. During the
Figure 6. Traffic violation at the place where there research process, the authors have surveyed the
is the footbridge in Hanoi (Vu Anh 2019)
existing flyovers in Hanoi with statistics from
300 questionnaires of the people. The results
2. CLASSIFICATION OF FLYOVERS show that:
2.1 Classification of shape Some people used the footbridge about
59.7% of the respondents, it could be many
Footbridge with roof things, such as the footbridge is not much or
Footbridge without roof people was walking very little.

Figure 5. Identify some positions of the footbridgein Hanoi on google maps (Vu Anh 2019)

423
Figure 7. The study of using footbridge by questionnaires in 2018

The need to use flyovers is only the basic building footbridge. This leads to poor use of
level used for road crossing purposes (about footbridge, the inefficient assessment of
70% of studies), other purposes are very few. footbridge, accounting for 35% of the research.
The form of architecture and landscape of Some footbridges are made of steel, with
footbridge is not the main concern of designers plastic-covered surfaces without roofs have
and managers. A footbridge architectures been peeled (due to high temperatures), some
survey shows that 65.9% of respondents footbridges with roof, the part of the roof is
are very much concerned, and 60% of people not affected, the part without roof is peeling,
are not so much (Luong Minh Chinh 2016). leading to after the rain is stagnant, people are
The users of footbridge were mainly non- afraid to use it because stepping on stagnant
residents and students, at 40% and 30% of puddles. At present, some bridges have been
studies, people live on each side of the streets or peeled with plastic cover instead of steel
close footbridge are rarely used (20% of (footbridge on Nguyen Trai Street), but some
studies). Due to the psychology of people bridges have not been repaired (such as the
familiar with Heavy traffic and using foot footbridge at Trade union University).
bridge takes more time due to long distances or Currently, there are four units managing
going out of the house is to sit on a motorbike pedestrian bridge in Hanoi but there is no
for convenience, quickly whether near or far. specialized department for pedestrian bridge.
The time of using the footbridge is popular Therefore, the implementation of projects and
at rush hour in the morning (6:30 - 8:30) and in management of exploitation is ineffective. In
the afternoon (17:00 - 19:00), at 70% and 56%, 2018, Hanoi City People's Committee assigned
other times are only about 30%, this shows that Vinasing media and trading joint stock
people are aware of the difficulty of crossing company (Vinasing) to advertise and be
the roads at rush hour. responsible for maintaining and repairing 45
With the location, facilities and traffic footbridges in Hanoi for 10 years (according to
conditions at the place where the footbridgeis Decision No. 2856 / QD-UBND of Hanoi
placed, it highlights the importance of People's Committee on May 18, 2017 and
overtaking in the road traffic system. Official Letter No. 183 / SGTVT-QLKCHTGT
The quality of pedestrian bridgeis not really of the Department of Transport dated March
guaranteed, the survey shows that the 22, 2018), in return, the company will invest in
assessment is not good: 50% and normal: 29% the city of 1,000 public toilets, 10 specialized
researches. This shows that managers are only tank trucks (South Korea), 50 direct drinking
interested in the presence and absence of water purifiers, 200 cast iron seats for the city's
footbridge, and include the economic value of public benefit.

424
4. THE TRENDS IN THE DESIGN AND Reviewing the location of the footbridge is
USE OF FOOTBRIDGE IN URBAN ineffective, the appreciation causes inefficient
state to appropriate remedies.
The pedestrian bridge in urban must meet the * Form and feature of footbridge
demand for pedestrians crossing the roads, it is Form of bridge must have a separate
not just an architecture for linking two sidewalk, imprint of the municipality, each bridge is a
but must have sustainability with future urban unique structure, must consider the footbridge
development, so The pedestrian bridge needs to is not only using to cross the road but also
have the following basic features: viewing as an indicator of progress in technical
 Creating a walkway in a safe and capability, engineering skills as well as a
convenient manner for crossing the road in symbol of economic potential of a particular
urban(Paulo & Peter 2016). city or region.
 There are separate imprints on the Besides functional for pedestrians, should
architecture of pedestrian bridgein urban areas integrate the other function, it is used multiple
or each area where bridgesare placed. objectives to increase efficiency footbridge.
 Using simple materials, easy The footbridgecan be used as a place of
maintenance, suitable for local climatic sightseeing, walking, joggingor integrated
conditions and stable work structure (Peter works both walk and crossed the road for
Lam et al. 2013). motorcycles and bicycles, specially
 Regarded as objects footbridges represents is construction for people with disabilities.
the modern techniques of construction and * Needing immediately to be repaired and
home demonstration installations own street art upgraded the parts of the pedestrian bridge
(Marek &Klaudiusz 2016). that it is damaged, degraded
 Bring economic benefits to socialize Upgrading surface of walkways, stairs, use
practical investment capital for development of of materials to ensure durability, noise
footbridges and other public works. restrictions; slip resistance and good drainage.
 Integrating multiple demands on a system Renewing the part of dirtiness and graffiti,
footbridges (on foot, motorcycles, bicycles or using bright colours to increase the aesthetics.
jogging where sports or use for people with Ordinary repairing of damaged parts to
disabilities) (Hasan&Napiah 2017). increase the usability and feeling safe in traffic.
 Matching culture, consciousness, * Landscape embellishment, arrangement of
managed and owned by the government and public utilities on pedestrian bridge
local communities(Mauro &Gian 2007). Arrangement and amendment
 The location of footbridges ensure ofthe public orderly bin with a lid on the
ongoing linkages and performance using the pedestrian bridge, the bin is aesthetic and
footbridge between two parties regional of road practical. Arranging from 2 to 3 bins on a
is the largest. footbridge (placed at the top of the stairs up
and down or add one in the middle).
5. RECOMMENDED SOLUTIONS EFFECTIVELY Utility design for people with disabilities, at
USE FOOTBRIDGES IN URBAN the same time, advantages to the
5.1 Solutions in the planning and construction pedestrians with suitcases and goods.
of pedestrian bridge system Layout of message boards; paintings, funny
advertising and so on to increase the
* The solution in transportation planning attractiveness
Check, recalculate the traffic density of * Wide dissemination of legal documents for
pedestrians on roads important, the location of everyone who participates in transport
many public buildings and attractions, on that implements the traffic regulations seriously.
basis calculate the location, size and the Give emphasis on communication and
distance between the footbridges in the Dissemination of Traffic Law for strata,
metropolitan area of the city. emphasizing violations and penalties.

425
5.2 The raising of public awareness The major development Project Management
Unit in Hanoi will be responsible for managing
Increase education, communication to improve
construction investment projects, operate and
knowledge of and commitment to road safety
repair pedestrian bridges.
for road users in general, and the pedestrians in
The police of the districts coordinate with
particular.
the police of the ward to manage the traffic
Focus on typical factors, the examples of
order and safety, handling violations of
good people and good deeds in Aware of
encroachment on pedestrian space; violations
traffics law's implementation, aware of the
to the road is not the right place and so on.
importance of using pedestrian bridge,
District People's Committees direct People's
criticizing organizations and individuals that
Committees of wards, mass organizations such
committing law-breaking acts of traffic safety.
as the residential group, the Women's Union,
For example: for pedestrians who are
the Youth Union are responsible for managing
students, if they go to the wrong road and do
education, popularizing and propagandising to
not use the footbridge at places where there is a
raise people's awareness in using pedestrian
bridge, apart from the prescribed sanctioning
bridges.
forms, the student will be sent That violation
results to the managing unit such as a faculty 5.5 Solutions for community participation
or school to assess the student's activity point.
Since then, it helps people become more aware * Encouraging private sector participation in
of traffic. building pedestrian bridges
Financial problems are the main reason leading
5.3 Intensifying inspection, examination and to difficulties in investing in building
supervision of the implementation of violations pedestrian bridges, especially in the current
in the use of pedestrian bridges period when the state budget is being spent on
many areas, Therefore, the encouragement and
Strictly, Decree 46/2016/ND-CP on sanctioning
mobilization of private sector participation in
of administrative violations in the field of road
investment in buildingpedestrian bridges is
traffic for pedestrians walking in the wrong
essential. At the same time, the investor will
areas allowed, individual organizations occupy
exploit the advertising benefits on pedestrian
sidewalk space, using pedestrian bridge space
bridges to recover capital.
for business, trade and so on.
* Strengthening the capacity of communities
Ensuring security and order for people to
to participate in pedestrian bridge management
use pedestrian bridge.
Improve the capacity of the community
5.4 Solutions on policies and managements where there are footbridges, thereby exploiting
and using socialized resources in managing
Decentralized pedestrian bridge management is pedestrian bridges.
clearly effective as: To bring into full play the role of typical
Hanoi’s transport department will cooperate personalities, heads of street quarter groups,
with Hanoi Department of Planning and religious dignitaries, and the elderly as the core
Architecture and District People's Committees in mobilizing people to participate of rules of
in charge of planning, positioning, scale and using pedestrian bridges.
design, appraisal and approval of footbridge
design documents. 6. CONCLUSION AND DISCUSSION
Department of Planning and Investment was
responsible for capital and budget for Pedestrian bridge plays an important role in
Investment in the construction. For projects ensuring pedestrians safety, reducing the
outside the budget, depending on the scale, it number of traffic accidents caused by the
must be approved by the District People's reason people walk to the wrong road, at the
Committee or Hanoi People's Committee. same time, it is also a measure to reduce traffic

426
jam in rush hours. In addition, the footbridge 7. ACKNOWLEDGEMENT
also contributes to beautifying the urban
landscape and creates a new face for urban This research was supported by Hanoi
Architectural University.
transportation in Hanoi.
In recent years, Hanoi has to spend a big
8. REFERENCES
budget to invest in building footbridge, but a part
of people still do not use or use inefficiently. Antonio Occhiuzzi1, MariacristinaSpizzuoco&
Causes of inefficient use of footbridge Francesco Ricciardelli2018. Loading models
system: Plan, design and construction are not and response control of footbridges excited by
suitable; management is still weak; the running pedestrians.Struct.Control Health
Monit.,15, 349–368.
awareness of participants in traffic is limited; Luong Minh Chinh 2016. Đánh giá và phân tích
enforcement of laws is not strict and lack of kết quả khảo sát về mỹ quan các công trình
community participation. cầu vượt thành phố Hà Nội sau một thời gian
Stemming from this practice, the study has đưa vào khai thác. Journal of Water resources
proposed a basic solution to effectively exploit & Environmental engineering, 54, 34-41
and use pedestrian bridge systems in the inner (Vietnam).
districts of Hanoi Capital to minimize Marek Salamak & Klaudiusz Fross 2016. Bridges in
congestion and traffic accidents to contribute to Urban Planning and Architectural Culture.
Procedia Engineering, 161, 207 – 212.
sustainable development of the Capital. Mauro Eugenio Giuliani &Gian Carlo
Recommendations for using footbridge: Giuliani2007. Footbridge Design in the Urban
 Need a transport master plan with the Context.International Journal of Space
footbridge is an essential ingredient of traffic Structures, 22(1), 45-54.
and urgent component of investment. It is Ngo Thang Loi 2009. Đô thị hóa ở Hà Nội nhìn từ
necessary to re-evaluate the effectiveness of góc độ phát triển bền vững. Work shop on
existing footbridge, thereby finding the causes Sustainable development of the cultural,
civilization and peace in Hanoi capital
and solutions for new footbridge. (Vietnam).
 When building pedestrian bridge, Prime MinisterDecision 519/QĐ-TTgonApproving
synchronous solutions are required. It is the Hanoi capital transport planning until
accompanied by the construction of a big 2030, vision to 2050. March 31,
central reservation, it is made of hard material, 2016.Vietnam.
and so pedestrians cannot cross the road and Peter Lam & Francis Kung2013. Innovative and
must use the footbridge. Sustainable Construction for a Footbridge
System in Congested Mongkok, Hong Kong.
 The structure of the footbridge should HKIE Transactions, 11(1), 15-20.
ensure safety and aesthetics. It is Paulo Rui Anciaes & Peter Jones 2016.
recommended to use materials and structures Pedestrians' preferences regarding signalised
to ensure easy disassembly so that it can be crossings, footbridges, and underpasses. Street
easily adjusted when necessary, avoiding mobility and network accessibility series, 9,
waste in construction investment. 1-19.
R. Hasan& M. Napiah 2017. Utilization of
 There are solutions to enhance footbridges: influential factors and
community participation in the process of improvement proposal. Advances in
exploitation and use. Transportation Studies: an international
 There are appropriate policies to attract Journal Section A, 43-60.
capital from many economic sectors to invest in Vu Anh 2015. Cầu vượt bộ hành trong thành phố
footbridge or franchise used to increase resources Hà Nội thực trạng và giải pháp. Journal of
for other socialized investment sectors. construction & city, 9-10/2015 (Vietnam).
Vu Anh 2019. Giải pháp đến nâng cao hiệu quả sử
 Widely disseminating road traffic law, dụng cầu vượt bộ hành cho thủ đô Hà Nội.
raising people's awareness when participating Journal of construction & city, 62-63-2019,
in traffic, strengthen measures to handle 87-91 (Vietnam).
violations in using pedestrian bridges.

427
The 3rd Int. Conf. on Transport Infrastructure & Sustainable Development (TISDIC 2019). Construction Publishing House. ISBN: 978-604-82-2893-4.

Flexural behaviour of precast concrete members

connected by loop joints

Nguyen Quang Thinh1,, Hiroshi Mutsuyoshi2, Takeshi Maki3, Yosuke Ishihara4

1
Lecturer, Civil Engineering Dept., Da Nang Architecture University, Da Nang, Viet Nam
2
Professor, Civil & Environmental Engineering Dept., Saitama University, Saitama, Japan
3
Professor, Civil & Environmental Engineering Dept., Saitama University, Saitama, Japan
4
Civil Engineering Maintenance Designing Division, Metropolitan Expressway Co.Ltd, Japan

Corresponding author, E-mail: [email protected]

ABSTRACT: This paper presents an experimental study onthe flexural behaviour of precast concrete
members using loop joints of steel bars in the connection. Six specimens were tested in this study. Five of
them had loop steel bars in the connection and the last one had continuous straight reinforcements in the
joint. The overlapping length of the loop reinforcements, the bending diameter of the loop bars and the
distance between the loop bars in a pair wereinput variables. Cracking patterns and flexural strength of the
loop connection specimenswere observed and compared with the performance of the specimen without loop
steel bars. Based on the experimental data, mechanical behaviour and failure modes of the loop jointwere
discussed comprehensively.The results indicated that the overlapping length and the bending diameter of
the loop bars had a significant influence on the flexural strength and the deflection of the connection.
Meanwhile, the distance between the loop bars in a pair had a slight effect on the strength performance of
the loop connection; however, the parameter had a considerable impact on the flexural deflection.
Keywords: Loop joint, pre-cast concrete, loop steel bars, overlapping length, bending diameter.

1. INTRODUCTION distinct approaches to connecting PCa

members. Because connections using mortar
Precast concrete (PCa) deck systems have been
have higher tolerance they are easy to fix on
applied widely in bridge and viaduct structures
site when compared with alternatives such as
having significant advantages in comparison
bolt or weld connections. Thus, this paper will
with cast-in-place concrete structures, such as focus on connections between PCa members
shortening the construction time, securing that use a loop joint of steel bars in conjunction
concrete quality and reducing production costs. with non-shrinkage mortar.
However, PCa structures still exhibit some In fact, there are some studies regarding
challenges, in particular the cast-in-situ joint loop bar connection between PCa decks system
(which is usually used to connect the adjacent in bridge structure. Dragosavič et al. (1975)
PCa members), can become one of the main published a report with an investigation into
limitations of the structure with unusable strength and behaviour of the loop connection
problems such as cracks and leakage occurring between PCa slabs by mean of loop steel bars
as a result. and cast-in-place joint with 151 specimens.
Based on the principle material used at the Another studywas conducted by Hao (2004)
joint and the method to secure the continuity of with nineteen specimens with vertical loop
the longitudinal reinforcement, there are many steel bars in the connection. Recently,

428
Villalba‐Herrero and Casas (2016) conducted In this study, German Code DIN 1045 was
16 slabs experiment with static and fatigue used to calculate the bending diameter and the
loading test. Besides, the authors also proposed overlapping length of the loop steel bars in
expressions for calculating the moment the connection:
capacity of the loop connection. However, the
limitations of their researches were not cover dBreq  ( 1, 4  2,8. / e ). . e /  ck (1)
all the impact of configuration parametersand Lareq  f . sa / ( 4 0 a ). .( Ase / Asv ).k  1,5dB (2)
the interface treatment between the PCa
members and cast-in-situ joint. Where:
Therefore, an experimental study should be
- Φ: diameter of U-bar
conducted to understand more flexural
behaviour of the loop connection. Besides, the - e: distance between two adjacent pairs of
influence of the overlapping length (La), the loops
bending diameter (dB) and the distance - σe: Stress level of reinforcement at bending
between the loop bars in a pair (a) to failure point (180N/mm2)
mode of the connection will be discussed - σck: design strength of mortar (40N/mm2)
comprehensively in this study.
- τ0a: adhesion stress, basic on compressive
strength of mortar is 40 N/mm2, this value is
equal to 2,0N/mm2
- Ase/Asv: required rebar cross-sectional
area/Reinforcement placement cross section (in
this case, the value is taken 1.0).
- k: factor taking into the influence of the
Figure 1. PCa deck with ribbed loop bars shift of the joint reinforcement (the value is
taken 1.6).
2. EXPERIMENTAL WORK
2.1 Test specimens
In this study, five specimens with loop joint of
steel bars place on vertical direction in the
connection were tested, namely D-1, D-2, D-3,
K-1 and K-2. Another specimen with
continuous straight bars denoted as C-1 was
also tested for comparison. The loop steel bars
protruded from the PCa elements, as shown in
Fig. 1. Each specimen contained two P
Casegments connected to each other by a cast-
in-situ mortar joint, and the segments were Figure 2. Schemata of the loop steel bars in the
designed as symmetrical four-to-four loop steel connection
bars. The length and width of the specimens
were 2900mm and 520mm, respectively. The detail of loop connection is illustrated in
Meanwhile, the thickness had two values of Fig. 2. In this study, the overlapping length of
150mm and 190mm. The mortar joint was reinforcement (La), the bending diameter of the
reinforced with two transverse bars which were loop steel bars (dB) and the distance between
placed inside the core of the overlapping bar to the loop bars in a pair (a) are variables. The
provide confinement to avoid brittle failure of geometrical parameters of the loop connections
the specimen (Chapman 2010). have been summarized in Table 1.

429
Table 1. Detail of the six experimental specimens surface of the concrete segments was textured
to eliminate the secondary influences on the
No. h (mm) La (mm) a (mm) dB (mm)
experimental result.
D-1 190 280 62.5 117
D-2 150 280 62.5 77 2.3 Test set-up
D-3 150 280 13.0 77 Electrical strain gauges of 5mm were used to
K-1 190 170 62.5 117 monitor the strain of the loop bars (Fig. 3).
Meanwhile, the strains of mortar and concrete
K-2 150 170 62.5 77
surfaces were measured by strain gauges of
C-1 150 --- --- --- 30mm (PL-30) and 60mm (PL-60),
respectively. The opening of the interfaces
2.2 Test material
between PCa members and cast-in-place
The concrete segments were prefabricated in mortar was also monitored by four strain
smooth formwork and cured for six days gauges (PI-100). Besides, total of six linear
before casting of mortar joint. The uniaxial variable displacement transducers - LVDTs
compressive design strength of concrete is (CDP-100) were used to measure the mid-span
60MPa; meanwhile, that of mortar is deflection, as shown in Figure 4.
purposefully retained lower in order to ensure Four point bending test was applied and the
failure at the mortar join, 40 MPa. Loop bar of actual test set-up is illustrated in Fig. 5. Static
13mm diameter and the transverse bar of load was conducted by a hydraulic jack at the
10mm were used. Besides, the roughness of the rate of 2kN/step and 1mm/step corresponding
interface between PCa members and the cast- with load control and displacement control
in-situ mortar joint which are difficult to define processes. The load was progressively increased
in the construction site, has significant up to failure and the propagations of cracks on
influence on the bond and shear strength of the the specimen surface were marked at every step.
section (Jogensor 2014). Thus, before The data from electrical strain gauges and
connecting two PCa elements by mortar, the LVDTs were taken by a data logger system.

Figure 3. Position of 5mm electrical strain gauges on loop reinforcements

430
Figure 4. Positions of strain gauges and LVDTs

Figure 5. Actual test set-up

431
3. EXPERIMENTAL RESULT also on that of PCa members. The slope of the
force-deflection curves decreased appreciably
3.1 Force vs. mid-span deflection subsequent horizontal crack appearance.
Besides, the behaviour of the specimens had a
The relationships between load and deflection
little variation due to difference on
are shown in the Fig. 6. The curves can be
configuration parameters.
divided into three periods clearly:
Stage III: (yielding) Many cracks appeared
Stage I: (linear elastic stage) The mid-span
deflection increased linearly with the applied on the bottom surfaces of the mortar area and
load. In this stage, cracks only occurred at the the PCa members. The curves illustrated
interface between the PCa members and the considerable ductility with a small growth in
cast-in-place joint. Horizontal cracks in the load value, accompanied with significant
mortar joint had not appeared yet. increase in the deflection. The crack width at
Stage II: Some horizontal cracks appeared the mortar joint increased quickly at the end of
not only on the tensile side of mortar joint but period until the failure happened.

Figure 6. Force-deflection curves

a) Stage I b) Stage II c) Stage III

Figure 7. Development of cracks in the actual tests

432
3.2 Strain results positions reached yielding at the same load and
the others didn't yield. Meanwhile, the strain
Typical strain vs. load curves of the
data of specimen C-1 showed that both layers of
experimental specimens are plotted in Figure 8.
the continuous straight reinforcements just
In case of D-1, it may be seen that the U-bars
yielded in the cast-in-situ mortar joint area and
yielded at A, B, C and E positions; meanwhile,
did not yield at the precast concrete segments.
that of case D-2 is only at A, C and D. There are
a little difference in case D-3 with only yielding Finally, it is interesting to note that in all the
at A and C positions of the loop reinforcements. specimens, the loop steel bars were always
The typical strains of the loop steel bars in case yielding firstly at the interface between the PCa
K-1 and K-2 are almost same, only A and B members and mortar joint.

D-1 D-2

K-1

D-3

K-2 C-1

Figure 8. Typical strain vs. load curves

433
3.3 Cracking pattern and mode failure D-2

In case of D-1, D-2 and D-3,it is observed that

at small load value, cracks only appeared at the
bottom side and lateral surfaces. However, at
the ultimate stage, the configurations became
more complicated with appearance of diagonal
cracks (Fig. 9). A typical mode failure of the
specimens is yielding of the loop steel bars at
interfacesbetween mortar and concrete
corresponding to the opening of the critical D-3
section. The failure accompanied with crushing
of the mortar in the compressive zone.
Meanwhile, in case of K-1 and K-2 which
have smaller overlapping length, havea distinct
difference. It is visible the development of line
of fractures running diagonally between the
loops of adjacent U-bars. Besides, the strain
data of the loop reinforcements showed that the
bottom legs of the loop steel bars were only
K-1
yielding in mortar area. The two lacer bars
which were placed inside the mortar core of the
loop joint, were also yielding at the ultimate
load. Therefore, it can be concluded that the
mortar core in case K-1 and K-2 were failed
due to the failure by dowel action.
In case of C-1, the cracking pattern is simple
with an oblique crack on the bottom side at the
mortar joint. It is here interesting to notice that
both legs of the main reinforcements (both
tension and compression sides) were stressed K-2
in tension during the experimental process.
Furthermore, the legs only attained to yielding
in the mortar joint area. Finally, the specimen
failed as a normal bending component with
opening of crack in tension side and crushing
of mortar in compression side.

D-1

C-1

Figure 9. Practical failure of experimental specimens

434
Table 2. Summary of failure modes and ultimate Based on the strain data and load-deflection
load relationship, it can be concluded that the
overlapping length of 170mm is not enough to
Specimen Failure mode Ultimate load
transfer the tensile force between the two splice
(*) [kN]
bars in the specimens of 150mm thickness.
D-1 Y/CC 62.5 Meanwhile, the change in overlapping length
D-2 Y/CC 49.2 of the loop steel bar affects not much in case of
190mm thickness.
D-3 Y/CC 49.8
K-1 C/CC 59 4.2 Effect of distance between the overlapping
loop steel bars in a pair (a)
K-2 C/CC 44.2
C-1 Y/CC 50.5 In this research, the influence of the distance
between the loop steel bars in a pair is verified in
(*) C = mortar core failure, Y = yielding of U- case D-2 and D-3. The distance from center-to-
bars, CC = crushing of compressive mortar
centerwas changed from 62.5mm to 13mm,
respectively. As illustrated in Figure 6, the
4. FLEXURAL STRENGTH ultimate loads of case D-2 and D-3 are
approximately equal, 49.2kN compare with
4.1 Effect of longitudinal overlapping length (La)
49.8kN, respectively. Before the first crack
The influence of the lap length on the appearance in the mortar joint, the load-
performance of the loop connection is deflection curves are linear and almost similar.
evidently on comparing the experimental After crack appearance, non-linear behaviour
results of D-1 with K-1 and D-2 with K-2. In takes places and the response curves are
this experiment, the overlapping length was different gradually. In the case D-3, the mid-
changed from 280mm to 170mm. span deflection at the maximum load is smaller
As shown in the Fig.6, the effect of the significantly than that of D-2, 49mm and 74mm
overlapping length on the flexural performance respectively. It can be explained that because of
of the connection is slight in case of 190mm decrease the distance between the loop bars in a
thickness specimens (D-1 and K-1). The load- pair, the tensile force of the main reinforcements
deflection curves are almost similar during the can be transferred to the opposite bars directly
experimental process, and the load bearing by bond and friction. Meanwhile, in the case of
capacity difference is only 5.6%. D-2, the tensile force was transferred from the
On the contrary, it is clearly to see a loop steel bars to mortar area and then to the
significant impact of the overlapping length opposite loop bars. Thus, the strain of the mortar
on the behaviour of the loop connection by area contributed considerably to the total strain
comparing D-2 with K-2. As illustrated in of the connection.
Figure6, the initial parts of the response
4.3 Effect of the bending diameter of the loop
curves are similar. However, after the cracks
steel bars (dB)
appeared in the mortar joint, the slope of load-
deflection curves of case K-2 decreased The bending diameter of the loop steel bar in
quickly and significantly smaller than that of case D-1 and K-1 are 117mm; meanwhile that
case D-2. Finally, the ultimate load bearing of case D-2 and K-2 are 77mm. As shown in
capacity of case D-2 with a larger overlapping Figure 6, the strengths of the specimens are
length (280mm) were observed to be larger increase significantly when the bending
than that of specimen K-2 with shorter diameter of the U-bar increases, up to 33%. Of
overlapping length (170mm), with 11.3%. course, this is not surprise since the increase of
Besides, the mid-span deflections of case D-2 the U-bar bending diameter will correspond
are also larger than that of K-2, 74mm and with increase the cross-section area of the
52mm respectively. mortar inside the loop joint (mortar core).

435
Thus, the load bearing capacity of the mortar - The flexural strength of the loop joint
core will be improved significantly by increase increases considerably by the increase of the
of dowel action capacity. Besides, the increase bending diameter of the U-bar.
of the internal diameter also corresponds with - The cracks at the interface between the
the increase of thickness of the specimens cast-in-situ joint and PCa members always
obviously. As a result, rising of the effective occur early at a small load.
depth of the section to tension reinforcement
(the lever arm) greatly contributes to the 5. REFERENCES
strength increase of the loop connection.
Chapman and Chery Elizabeth 2010. Behaviour of
precast bridge deck joints with small bend
4. CONCLUSIONS
diameter U-bars, Master Thesis, University of
We havepresented and discussed test results Tennessee, Knoxville.
fromthe six specimens in this paper.The DIN 1045. Concrete and reinforced concrete
structures, Design and Construction. German
conclusions are as follows:
Standard.
- The overlapping length of 170mm is Dragosavič, M. 1975. Loop connection between
insufficient to transfer the tensile force precast concrete components loaded in
between the 13mm diameter loop bending. Heron, 20(3).
reinforcements, but in the case of 280mm lap Hao, J.B. 2004. Structural behaviour of precast
length, that is adequate for the transmission. components joints with loop connection. PhD
- In the case of 150mm thickness, increase Thesis, National University of Singapore.
of the overlapping length improves Jorgensen, H.B. 2014. Strength of loop connections
significantly the flexural performance of the between Precast concrete elements. PhD Thesis,
loop connection; meanwhile, the strength is University of Southern Denmark.
slightly improved by the increase in 190mm Villalba‐Herrero, S., Casas, J.R. 2016. New
thickness specimens. structural joint by rebar looping applied to
- The flexural strength of the loop joint is staged box girder bridge construction. Static
tests, Structural Concrete, 17(5).
not significantly impacted by the increase of
the distance between the loop steel bars in a
pair; however, the ductile behavior is greatly
affected by the change.

436
The 3rd Int. Conf. on Transport Infrastructure & Sustainable Development (TISDIC 2019). Construction Publishing House. ISBN: 978-604-82-2893-4.

Seismic performance of an infilled moment-resisting steel

frame during the 2016 central italy earthquake

Phan Hoang Nam1,2,, Fabrizio Paolacci1, Hoang Phuong Hoa2

1
Department of Engineering, Roma Tre University, Rome, Italy
2
Faculty of Road and Bridge Engineering, The University of Danang – University of Science and
Technology, Da Nang, Vietnam

Corresponding author, E-mail: [email protected]

ABSTRACT: A sequence of earthquakes occurred between the end of August 2016 and the end of
October 2016 in Central Italy causing significant damage and major disruption in a wide area. The
sequence of events is composed of five events with magnitude between Mw 5.5 to 6.5. As a consequence,
numerous residential buildings in the affected area was not particularly resistant tothe shaking, resulting
in the collapse and heavy damage. With a particular focus on masonry infilled steel frames, this paper
evaluates the seismic performance of an infilled moment-resisting steel frame located in Amatrice,
Central Italy, which suffered significant damage during the August 2016 Central Italy earthquake. The
aim is to investigate the effect of the masonry infill to the seismic performance of the building. The three-
dimensional (3D) frame building is modeled using the Opensees software, where the beam and column
elements are modeled by using a nonlinear hinge model and the infill is idealized as diagonal struts with
nonlinear hysteretic behavior. Nonlinear static and dynamic analysesare performed for both bare and infill
frames in order toassess the effect of the masonry infill on the overall seismic response and confirm the
actual damage pattern surveyed in the aftermath of the 2016 Central Italy earthquake of the case study.
Keywords: Steel moment-resisting building, seismic behavior, static pushover analysis, dynamic time
history analysis, infill modeling.

1. INTRODUCTION occurred on October 26th and the main-shock

of October 30th (Mw 6.5 and depth of 9km).
The 2016-2017 Central Italy earthquakes The latter was the strongest event since 1980.
consisted of several moderately-highmagnitude The particular event had an epicenter 5km from
earthquakes between moment magnitudes Mw Norcia, overlaying the northern part of the fault
5.5 and 6.5. The first main shock on August activated by the earthquake of August 24th.
24th, 2016 was of moment magnitude Mw 6.0 These events caused a total of 299 fatalities,
and occurred at a depth of 4km. The epicenter 386 injured and about 4800 homeless
of the August’s shock was located in the (Fiorentino et al. 2018). Most of the victims
North-Eastarea of the province of Rieti; this is were in the areas of Amatrice, Accumoli, and
a zone of high seismic hazard that has seen Arquatadel Tronto. In these municipalities,
damaging earthquakes in the recent past. A heavy damage and collapse of residential
series of aftershocks followed the above event buildings were reported (De Risi et al. 2018,
until the Mw 5.9 Ussita earthquake that Fiorentino et al. 2018).

437
 The observation of the damage has been 2. DESCRIPTION OF CASE STUDY
primarily reported in existing steel multi-story
residential buildings which exhibited low The sample steel frame is a two-storybuilding
energy absorption and inadequate dissipation located in Amatrice, Central Italy. It was built in
capacity. The insufficient horizontal stiffness of the early 90’s according to the 1996 Italian
the frame and the masonry infill led to seismic code and consists of a basement, the
ground floor, and two upper stories alongside
significant lateral drifts and buckling in the steel
ashorter top story that serves as a penthouse. The
components, especially in the columns. Local
building plan layout is trapezoidal; it is 22.5m
damage (e.g., buckling) has also been observed
long and 6.6 - 8.45m wide. The ground floor
at the connections due to the strut-action
height is 3.72 m and the interstory height of the
induced by the presence of the masonry infill.
first and second floor is 3.2m. The flooring
The issues of the existing steel moment
systems consist of concrete slabs on a steel
resisting frames (MRFs) are first related to the
corrugated 10mm thick sheet. The steel grade
accurate prediction of the building seismic
for all structural components of the frame is
performance that is due to the complex behavior
equivalent to S235 (fy = 235MPa). The columns
and the interaction between beam-to-column
are HEA 200, while the beams are HEA 300 and
connections, as well as between the frame and
HEA 160. The beam-to-column connections are
masonry infill (Di Sarno et al. 2017). Because of fully welded (i.e., moment-resisting frame). The
such complexity and the absence of a realistic, infills consist of a double layer of perforated
yet simple, analytical model, theeffect of
brick of the size of 12258 mm. The roof is
masonryinfill panels is often neglected in the
supported by a truss structure, whose elements
nonlinearanalysis of building structures. This
are made of 2L606mm profiles.
assumption may lead tothe inadequate prediction
Due to the lack of existing design drawings,
of the lateral stiffness, strength, and ductility of
the self-weight of the slab is assumed to equal
the structures. It also leads to the uneconomical to 4.88 kN/m2, while the live loads are assumed
design of the frame since the strength and equal to 2.00 kN/m2 for the intermediate floors.
stiffness demand on theframe could be largely The snow and wind loads are taken equal to
reduced (El-Dakhakhni et al. 2003). 3.76 kN/m2 and 0.96 kN/m2, respectively.
With a particular focus on masonry infilled
steel frames, the aim of this paper is to
investigate the seismic behavior of a steel MRF
building with masonry in fill panels. The
building is two-story andlocated in Amatrice,
Central Italy, which suffered significant
damage during the 2016 earthquake. The three-
dimensional model of the building is
implemented using Opensees software, where
both material and geometry nonlinearities are
taken into account for. The infill is modeled as
an equivalent diagonal strut, where its behavior
is followed by a hysteretic model of axial force
and displacement. The static pushover analysis
Figure 1. The sample building after the 2016-2017
is first performed for both bare and infilled Central Italy earthquakes
models, and the nonlinear time history analysis
is later performed with the Amatrice After the 24 August event, the building
earthquake signal. The analysis results mainly suffered major cracks in the brick infill
demonstrate the necessity of considering the panels, with only small local flange
infill in the steel frame model. instabilities observed at the top of two front

438
columns of theground floor. At the end of the
entire seismic sequence, the building
experienced evidentlypermanent deformation
along its longer direction. Such permanent
deformation was localized at the second level
of the building with a visible residual interstory
drift, as shown in Figure 1. Preliminary finite
element analyses of thebuilding confirmed that
the fundamental period of the structure is
approximately equal to 0.56s (Di Sarno et al.
2017). This was an uncoupled translational
mode along the long side, which was mainly
attributed to theorientation of the steel columns
with their strong axes aligned with the short
side of the building. Naturally, residual drift Figure 2. Macro-modelling of masonry infill
through equivalent struts
developed along the longitudinal axis.
The skeleton curve is step-wise linear in
3. NUMERICAL MODELING OF SAMPLE shape to describe the compression-only
BUILDING behavior of the strut. In particular, the model is
3.1 Infill modeling characterized by the following stress states (see
Figure 3):
In the past decades, numerous researches were - initial elastic behavior;
conducted to investigate the behavior of steel
- post-elastic linear responsecharacterized
frames infilledby masonry walls. Basically,
by a reduced value of stiffness;
two main approaches were proposed by various
- softening response of the panel after the
authors formodeling the masonry infill. The
maximum force;
first is based on a fullfinite element (FE)
modeling and the other one that is more - residual axial strength after a given value
simplified is based on the concept of of displacement.
“equivalent strut”.
Since FE models are generally unfit for
practice-oriented analyses of the seismic
response of steel frames, several macro-models
were proposed and are currently available in
the scientific literature to simulate the
globalseismic response of masonry infilled
steel frames (Asteris et al. 2011). Such models
are generally based on the assumption of two
diagonal equivalent struts connecting the two
opposite corners of the structural frame cell, as
shown in Figure 2. Although in principle, such
macro-modeling approaches generally result in
effective simplifications, the calibration Figure 3. Force-displacement curve according to the
ofequivalent struts is not an easy task. In the model by Panagiotakos and Fardis (1996)
following, a nonlinear force-displacement
The initial shear stiffness R1 of the
relationship to describe the response of
uncracked panel can be defined as follows:
equivalent strutsproposed by Panagiotakos and
Fardis (1996) is presented. R1  Gwtwlw / hw (1)

439
Where Gw is the shear modulus of the masonry
infill obtained in diagonal-compression tests
while tw, lw and hw are the thickness, the length
and the height of the masonry wall. Then, the
load value at the onset of cracking, namely
yielding force Fy, is defined as follows:
Fy  f ws twlw (2)

in which fws is the tensile strength evaluated by

diagonal-compression tests.
The axial post-cracking stiffness R2 can be
defined according to the following equation:
R2  Ewtw w / d (3)

in which Ew is the Young modulus of masonry

and the following equationcan be adopted for
evaluating the width w of the equivalent strut: Figure 4. Floor plan and front elevation of the
building model
0.4 d
w  0.175   hw  (4)
The beams and columns are modeled using
where and are defined as: fiber sections with a behavior represented by the
Ewtw sin 2
Menegotto-Pinto constitutive law. Two node-link
4 (5) elements are employed to model equivalent
4 Ec I c hw struts, which represent the nonlinear behavior of
the masonry infill in terms of both strength and
d  hw2  lw2 (6) stiffness. The stiffness degradation in both
loading and unloading phases is implemented
in which  is the slope angle of the infill’s
through the so-called Pinching 4 material model
diagonal, Ec is the elastic modulus of concrete,
available in Opensees. In detail, the floating point
and Ic is the moment of inertia of the column.
values defining force and deformation points on
The post-cracking branch keep growing up to
the skeleton compressive response are evaluated
the maximum force Fm = 1.3Fy. Then, a post-
from the geometric and mechanical properties of
peak softening branch follows, whose (negative) the masonry infills through the model by
stiffness R3 can be assumed within the range Panagiotakos and Fardis (1996), while the ones
0.005R1 < R3 < 0.1R1 and the residual strength on the response envelope in tension are assumed
Fu can range between 0 < Fu < 0.1Fy. equal to zero. The floor plan and front elevation
3.2 Structural modeling views of the model are shown in Figure 4.

The structural system is modeled with a refined 4. SEISMIC RESPONSE ANALYSIS

three-dimensional finite element model using
the finite element software Opensees (Mazzoni 4.1 Nonlinear static pushover analysis
et al. 2007). The rigid diaphragm is assumed at The presence of large residual damage in the
the first and second floors to account for the surveyed structure indicates the high level of
presence of the concrete slab on the steel seismic action to which the building was
corrugated sheet. It is noted that the roof does subjected during the Central Italian earthquake.
not have any diaphragm. The building is Consequently, nonlinear analyses of the
assumed fixed at the base and the beam-to- building are conducted by using the model
column connections are assumed rigid and described in Section 3. As the first level of
fullstrength (i.e., moment resisting). investigation, nonlinear static pushover analysis

440
is conducted to assess the static behavior of the floor is implemented by the reinforced concrete;
building with an increasing level of the action. there is no damage observed. The results of the
The results in terms of the total base shear and pushover analysis are in good agreement with
the roof drift ratio of the building exhibit the the real damage of the building, where the
significant increase both the stiffness and significant damage observed wasthe masonry
strength of the infilled frame as compared with infill on the first floor.
those of the bare frame, as shown in Fig 5(a).
4.2 Nonlinear time history analysis
The dynamic behavior of the building is
investigated considering the East-West
component of the ground motion recorded at the
accelerometricstation of Amatrice, as shown in
Figure 6(a), during the event M/S1, applied
along the longitudinal direction. The PGA is
equal to 0.85g with a resonance peak at 0.25 sec.

a)

a)

b)
Figure 5. Static pushover analysis results of bare
and infilled frame: (a) Roof drift - total
base shear and (b) Roof drift - infill
axial force

Figure 5(b) shows a representative result of

the pushover for the infill panels on three floors
at the second bay (see Figure 4 marked blue
b)
color). The infill panel on the first flooris
completely failed at a roof drift of 0.5%, while a Figure 6. East-West record (a)and its elastic response
much larger number of the roof drift, i.e., 1.0%, spectrum (b)of the 24 August mainshock
that can cause a total collapse of the one on the
second floor. From this point forward, the The modal analysis is first carried outfor both
pushover curves of the two cases are almost the cases, with and without infill panels. The results
same. The slight difference is due to the residual show that the fundamental period of the building
strength set for the infill to avoid the instability with the infill decreases two times, T1 = 0.5 s, as
of the model. The external wall of the ground compared with that of the bare frame, T1 = 1.0 s;

441
this confirms the influence of the infill in The local behavior of the infill panels on
increasing the overall stiffness of the building. three floors is shown in Figure 9. These three
The 5% damped response spectrum of the ground panels are in the second bay marked as blue
motion against the fundamental periods of the color in Figure 4. The most damaged infills are
bared and infilled frames is shown in Figure 6(b); clearly located on the firstfloor of the building,
this provides insight into the amplification of the where a large value of the interstory drift
spectral acceleration at T1 = 0.5 s. observed. The infill panel on the ground floor
The nonlinear time history analysis is then is still in the elastic range of the behavior, and
carried out with the Amatrice earthquake. the one on the third floor reaches the cracking
Figure 7 shows the timehistory of the total base strength. This result fits closely the observed
shear for the bare and infilled frames.The response of the real building after the first
responseof the building evaluated with the event, i.e., the 2016 Amatrice earthquake,
timehistory analysis with respect to the weak where most of the infill panels in the firstfloor
direction, i.e., longitudinal direction, of the
suffered a major cracking. The damage in the
building. A significant contribution of the infill
steel elements is limited at first and second
panels is observed. The peak base shear values
stories, which is similar to the real damage
for the bare and infilled frames are 800 kN and
1150 kN, respectively.The peak drift ratio pattern observed on site.
along the building height is shown in Figure 8,
where a large value of the drift is observed for
the first floor; this fits the real observation, as
shown in Figure 1.

Figure 9. Axial force - displacement of the

representative infill panels in the floors

4. CONCLUSIONS
Figure 7. Time history data in terms of total base
shear The paper presents a numerical assessment of
the structural performance of a case steel MRF
building located in Amatrice, Central Italy,
which sufferedextensive structural and non-
structural damage during the recent swarm of
events. Refined numerical models of the
existing steel building are developedwith and
without considering the effect of the perimeter
masonry infill. The primary aim is to assess the
effect of the masonry infill on the overall
seismic response of the building and to confirm
the actual damage pattern surveyed in the
Figure 8. Peak interstory drift along with the aftermath of the 2016 Central Italy earthquake
building height for the case study.

442
 The modal analysis results emphasize the Panagiotakos, T. B., and Fardis, M. N. 1996.
significant effect of the masonry infill on the Seismic response of infilled RC frames
global dynamic behavior of the structure. It is structures. 11th World Conference on
foundthat the fundamental period of the Earthquake Engineering.
vibration of the steel frame with the infill is De Risi, R., Sextos, A., Zimmaro, P., Simonelli,
A., Stewart, J. 2018. The 2016 Central Italy
nearly 1/2 of that of the bare steel structure, 0.5
earthquakes sequence: observations of
seconds versus 1.0 seconds. Thus, it is of incremental building damage. Proceedings of
paramount importance, when estimating the the 11th National Conference in Earthquake
seismic demand on steel framed buildings. The Engineering, Earthquake Engineering
presence of the infill tends to augment the Research Institute, Los Angeles, CA. 2018.
lateral stiffness of the structural system and, in Di Sarno, L., Paolacci, F., Sextos, A. 2017.
turn, to increase the seismicloads. Seismic Performance Assessment of Existing
The seismic performance of the case Steel Buildings: A Case Study, Key
studyis then assessed through advanced, Engineering Materials 763, Trans Tech
comprehensiveand efficient static and Publications, Switzerland.
dynamic nonlinear analyses, which based on a El-Dakhakhni, W. W., Elgaaly, M., and Hamid, A.
3D nonlinear model of the building. It is A. 2003. Three-strut model for concrete
masonry-infilled frames. Journal of Structural
found that the analyses can provide a realistic
Engineering, 129(2), 177–185.
damage observation when the masonry infill is Fiorentino, G., Forte, A., Pagano, E. et al. 2018.
included in the structural model and can thus Damage patterns in the town of Amatrice after
be utilized to perform a reliable risk August 24th 2016 Central Italy earthquakes.
assessment of the building. Bulletin of Earthquake Engineering, 16(3),
1399-1423.
5. REFERENCES McKenna, F., Fenves, G. L., and Scott, M. H.
2007. OpenSees: Open System for Earthquake
Asteris, P. G., Antoniou, S. T., Sphianopulos, D. Engineering Simulation. University of
S., Chrysostomou C.Z. 2011. Mathematical California, Berkeley, CA.
macromodeling of infilled frames: state of the
art.Journal of Structural Engineering,
137(12), 1508-1517.

443
The 3rd Int. Conf. on Transport Infrastructure & Sustainable Development (TISDIC 2019). Construction Publishing House. ISBN: 978-604-82-2893-4.

GPU acceleration in OpenFOAM simulation of

wind-over-top flow

Doan Sy Long, Hitoshi Yamada, Hiroshi Katsuchi

Department of Civil Engineering, Graduate School of Urban Innovation, Yokohama National
University, Yokohama, Japan

Corresponding author, E-mail: [email protected]

ABSTRACT: Large mesh size on a linear solver consumes several weeks for execution to give results. The
Graphics Processing Unit (GPU) is developed with thousands of CUDA cores so that the parallel simulation
on GPU would speed up. However, the suitable solver for GPU remains the key challenge for developers.
Besides, the wind-over-top (WOT) flow is generated among the high-rise building in the urban area. The
WOT flow is proved to increase the pressure near the top and positively contribute to the vibration of a
downstream building. The 3D simulation on the WOT flow with high quality of mesh requires a
hugenumber of mesh cells. The results indicate that the execution time in GPU solver is non-linear with
simulation time, while the linear execution time is observed in simulation on the Central Processing Unit
(CPU). Running on multiple GPU does not accelerate the process instead of a single GPU for cases of a
moderate number of mesh sizes. However, the solution time by GPU is beneficial when accelerating 2.5-3
times faster compared to CPU simulation in huge mesh. By using GPU in simulation for the WOT flow,
this flow contains the shear layer and the unsteady vortex under the shear layer. The shear lay ereffects to
the downstream buildings by increasing the pressure of front face and side faces near the top area. The
condition to form the WOT flow is governed by the distance of two buildings.
Keywords: RapidCFD, OpenFOAM, high-rise building, wind-over-top flow.

1. INTRODUCTION solver for GPU remains the key challenge for

developers in term of transfer data between
Computational Fluid Dynamics (CFD) has
CPU and GPU and the algorithm on GPU.
become a common tool among computational
Rapid CFD was designed by simFlow and
wind engineers for the simulation of wind
flows. The open-source program like Open based on OpenFOAM 2.3.0 code. With
FOAM is used to solve a variety of problems differences of concept in CPU and GPU, some
with different mesh sizes and algorithms. of solvers and algorithmsare not correlated.
Large mesh size on a linear solver consumes GPU is still at the beginning of the development
several weeks for execution to give results. The process to build the code and environment to
Graphics Processing Unit (GPU) is developed work with. The most difficulty is the code
with thousands of CUDA cores so that the productivity for each solver. GPU is ideal for
parallel simulation on GPU would speed up CFD tools in large-scale simulation, but it takes
compared to the simulation of the Central much time and human resources to implement
Processing Unit (CPU). However, the suitable all solvers, model updates, and optimizations for

444
GPU. Currently, many tips for GPU effects of WOT flow from the upstream
optimization are proposed. Also, easily and building to side faces of the downstream
highly productive tools like OpenACC is also building remain unclear.
available for guidelines. However, GPU In this paper, an attempt was made to
implementation of CFD tools like Open FOAM simulate the WOT flow in boundary layer
is still anenormous challenge. turbulent wind flows around square cylinders
Applying GPU in simulation is a using GPU with the RapidCFD based on the
considerable concern of many researchers. open-source program OpenFOAM. A high-
Yamagishi and Matsumura (2016) developed performance computing (HPC) computer with
the Kinaco program which is using GPU multiple GPU and CPU is used to compare in
simulation on the Fortran language. The simulation. Then, GPU simulation would be
simulation of large-scale oceanic circulation applied in the simulation of the WOT flow
indicates that the GPU implemented model is through the interference test with the Cartesian
faster about 4.7 times compared to the CPU coordinate system. Our main idea is to assess
model. Moreover, Rapid CFD is put into the the flow characteristics of WOT flow by the
continued development for speed reduction comparison of interference tests with the same
among the open-source code. In a simulation of height and double height interference models.
thermodynamic models using Rapid CFD, the The same height interference model produces
simulation run by the GPU is significantly the flow which contains the WOT flow from
faster compared to the CPU simulation the roof and the side flow generated from the
(Kurashov 2016). By using the GAMG side of the interference model. Meanwhile, the
algorithm, Open FOAM based RapidCFD double height interference model is high
proved the advantage of speed (2.6-3.5 times) enough to ignore the effect of the WOT flow
in the simulation of flow inside the centrifugal on the principal model. Comparisons of the
pump (Nocente and Nielsen 2016). same height and double height interference
Due to the limitation of Rapid CFD, tests in pressure coefficient and velocity vector
researchers have developed the new field are expected to show the influence of the
environment of GPU based on Open FOAM WOT flow.
platform. AmgX is developed by Rathnayake et
al. (2017) and allow the users to get access to 2. SPEED TEST ON GPU
more option in the selection of algorithm. 2.1 Computer specification
However, this method did not show the speed
acceleration compared to the original method of Deep Learning BOX(DL-BOX) computer is
Rapid CFD as the GAMG method. Rapid CFD integrated with multiple GPU and CPU. The
is still used as the common tool to simulate specification of this computer is listed in Table 1.
using GPU among computational engineers Noted that CPU is integrated with the latest
The WOT flow is a non-stationary development of CPU core working at high
complicated path and plays a different role frequency.
from Karman vortex generated on the side. Table 1. Computer specification
According to Taniike (1992), the principal
building receives more angular momentum Name Deep Learning BOX
which is generated from the shed vortices of an CPU Intel® Core™ i9-7920X [2.9 GHz,
upstream building. Kawai et al. (2009) showed 12core 2socket]
that the unsteady rooftop wind generates the GPU GeForce GTX 1080 Ti 12Gb x 4
arch-type vortex which stretches in a stream- Memory 2666Mhz, 128Gb
wise direction near the tip of the vortex. A OS Ubuntu 16.04 LTS 64bit
presence of an upstream building accelerates
CUDA 8.0
the flow and magnifies the pressure on the flat
roof of a downstream building. However, the MPI Open MPI

445
2.2 Speed test on the circular cylinder convergence at 10-8. When the mesh size
increases to 700,000 cells, 4GPU and 4CPU
The study is to compare the execution time
work at the same speed in solving the
during the simulation of GPU and CPU using
equations of the simulation. However, 1GPU
open-source program Open FOAM. All test
simulate faster compared to the simulation of
cases are single GPU simulation (1GPU), 4
4GPU. The parallel running of GPU could not
multiple GPU simulation using Rapid CFD for
provide the best performance in this case. The
solving the general equations of CFD, and
system needs some time to divide works to
OpenMPI for parallel processing (4GPU), 4
each GPU in parallel simulation. This may be
multiple CPU simulation using OpenFOAM for
the explanation for this fast running on a single
solving the general equations, and OpenMPI for
GPU compared to multiple GPU process. The
parallel (4CPU) and 24 multiple CPU
acceleration of 1GPU in case of 700,000 mesh
simulationusing Open FOAM for solving the
cells is 1.2 times faster to 4CPU simulation.
general equations, and Open MPI for parallel
At high computation effort with 3,000,000
(24CPU). Differences in the algorithm methods
cells, the GPU simulation shows its advantages
and transfer data among processing units give
in parallel simulation. The simulation of 4GPU
various results on the speed of simulation. The
is faster 2.7 times compared to the simulation
test on different mesh sizes would give more
by 4CPU. However, the computer is integrated
details in using GPU for CFD simulation.
with 24CPU. Then, higher computational mesh
The validation test with a cylinder in
size was proposed at 15 million cells. At this
laminar flow with structured mesh is tested
moment, parallel GPU simulation still active
with different mesh sizes. This simulation is
with the acceleration about 2.5 faster in
built on the structured mesh as Figure 1. The
simulation is already validated with the works convergence stage.
of literature in a wide range of Reynold The algorithm solution for Navier-Stokes
number (100-105). The mesh size is modified equation is not stable in GPU solver leading to
by changing the number of layers in the the non-linear execution time during the
thickness direction. The simulation keeps the simulation. In this case, CPU solver shows the
same algorithm and level of convergence for constant speed as every time steps when the
every part of Navier-Stokes equation. Reynold simulation gets convergence.
number is 104 for all cases of simulation. Figure 3 shows the summary of the
execution time at the same number of time step
(20,000 steps) at different mesh sizes. The
results show clearly that GPU simulation has
efficiency in high computational demand only.
If the requirement of number of CPU in
simulation is about 50,000 cells for 1 CPU, the
simulation of GPU strongly depends on the
memory of GPU and CPU.
Suggestion for using GPU in CFD simulation
could be: 1) With mesh size less than 700,000:
Using CPU for simulation and have more
options for solvers; 2) 1 GPU works efficiently
Figure 1. Mesh type of example test in the range of 700,000 - 3,000,000 cells. 3)
Multiple GPU is not suggested for small mess
The results are shown in Figure 2. At a low size (< 3 million cells). 3) At 15,000,000 cells,
level of mesh sizes, the CPU works better the speed ratio of 4GPU and 24 CPU is more
compared to the GPU simulation. The than 2.5 times. Meanwhile, single GPU
explanation for this phenomenon is that GPU simulation is equal to the simulation of 15CPU
accelerates the speed when the simulation gets in term of simulation speed.

446
Figure 2. Comparison of execution time and GPU acceleration different mesh sizes: i) 150,000 cells; ii)
700,000 cells; iii) 3,000,000 cells; iv)15,000,000 cells

Figure 3. Comparison of execution time at same time step (20,000) at different mesh sizes

447
3. GPU APPLICATION ON THE
SIMULATION OF THE WOT FLOW
3.1 Verification tests on isolated case
A simulation on OpenFOAM has modelled the
test of wind tunnel experiments at Yokohama
National University, Japan. More details on the
wind tunnel test could be found at Sy et al.
(2019), The domain (Figure 4) is simulated at
the same size of the wind tunnel section 1.8 ×
1.8m (width × height). A wall function is
applied near the wall of principal building to
reduce the computation efforts. The wall
function is applied for kinetic energy and
specific dissipation rate. Figure 5 indicates the
normalized wind velocity and turbulence
intensity which was used in the wind tunnel test
as well as the input velocity in simulation. The Figure 5. Velocity inlet and turbulence intensity
inflow boundary condition is tested with very
low turbulence intensity which is considered as The polyhedral mesh contains 806,156 cells
resulting in the value of Y+ at 5. The polyhedral
a smooth flow condition. The velocity in the
mesh could increase the size along with three
experiment at the top model height is calculated
axes while the structured mesh needs to keep
at U0 = 3m/s by the similarity law. The Reynold the small grid along the whole domain. The
number is defined by the width of the model polyhedral mesh has proved the advantage in
and reference velocity was 2.01×104. The mean reducing the number of mesh cells and
pressure coefficient c p of each pressure tap is disadvantage in increasing the number of faces
calculated by (Spiegel et al. 2011). However, polyhedral
mesh in this study shows the advantages in
p  p
cp  (1) both the number of cells and faces due to the
q strategy of increasing sizes in all directions.
where p is the total pressure at each tap, and p∞, To find out the best-fixed turbulence model
for the current study, all incompressible
q∞ are the reference static and dynamic pressures.
turbulence models in Table 2 are simulated in
the isolated cases and compared with the
Slip condition literature and current study on wind tunnel test.
The suggestion for URAN test in AIJ
Outlet Benchmarks test (Tominaga et al. 2016) was
Slip Renormalization group k-. While Wilcox
condition Wall
function (1998) suggests using k- SST model to
Velocity
(RANS) reproduce well the stagnation point. However,
inlet
the current wind tunnel test was performed
Slip with different turbulence intensity and
condition boundary condition compared to the
Benchmark test and works of literature. Thus,
the trial tests on different turbulence models of
Figure 4. Computational domain in isolated test URANS is needed to find out the most accurate
with the polyhedral mesh model for the current wind flow condition.

448
Table 2: Applied turbulence model in RANS simulation test
Turbulence model Note Reference
Standard k- kEpsilon (Launder and Spalding 1983)
Lien cubic non-linear low-Reynolds k- LienCubicKE (Lien et al. 1996)
Realizable k- turbulence model realizableKE (Shih et al. 1995)
Renormalization group k- RNGkEpsilon (Yakhot et al. 1992)
Shih's quadratic algebraic Reynolds stress k- ShihQuadraticKE (Shih et al. 1993)
Lien and Kalitzin's v2-f v2f (Lien and Kalitzin 2001, Davidson 2003)
Standard high Reynolds-number k- kOmega (Wilcox 1998)
Implementation of the k-SST kOmegaSST (Hellsten 1998, Menter and Esch 2001)

The results of mean pressure coefficient at the building model. Group of k- based models
2 show the various results along the measured
elevation of H0 in the simulation are
3 line. Standard k-, RNGkEpsilon, realizableKE
compared with the wind tunnel test (Sy et al. and v2f models could not produce the
2019) and previous studies on the isolated appropriate results when showing high pressure
building in Figure 6. List of the wind tunnel in the front face and low pressure in the side and
test on the Common wealth Advisory back faces. LienCubicKE and ShihQuadraticKE
Aeronautical Research Council (CAARC) model simulate the turbulence model in the test
standard building (Dragoiescu et al. 2006) with acceptable results on the local pressure.
including tests from City University, Briton The two equation models on k- model are not
University, Monash University, Tongji good guessers in this low turbulence intensity
University, and Cheng et al. (2010) are also case, while the nonlinear eddy viscosity models
plotted for comparison. could present the results better. In the other
hand, kOmega and kOmegaSST indicate
different accuracy level of pressure on faces.
The kOmega and v2f model are difficult to
avoid the stagnation point abnormally. The
turbulence model kOmega could not present the
accuracy pressure on the model while the
kOmegaSST well presents the results on faces
matching with literature and current study by
wind tunnel test.
The execution time is also observed in the
isolated test with different configuration on the
CPU and GPU. As recommendation from the
previous tests, the simulation with more than
Figure 6. Comparison of pressure coefficients of
700,000 cells, single GPU should be used.
all faces in isolated test to literatures and
wind tunnel test Figure 7 represents the results of execution
time during the simulation of isolated test. At
The means of pressure coefficients of all the convergence state, GPU simulation proves
faces in are represented in Figure 6 including its time-cost saving process when shortening
test of turbulence models in group of k- the time around 3 times compared to the
equations and group of k- equations. simulation by 4GPU. Noted that the computer
In general, all turbulent models can simulate can be easily to install more GPU, while the
the trend of pressure around all faces of the number of CPU is limited on each computer.

449
 Interference model

Principal model
U0

2H0

H0
Side flow

z
x

Figure 9. Configuration of double height

interference test

Figure 10 represents the mean pressure

Figure 7. Comparison of execution time in simulation
coefficient of pressure taps along with the
of CPU and GPU in isolated test
height of the front and side faces of the
3.2 Simulation on the Wind-over-top flow principal model in tandem arrangements
respectively. The data sets in the front face
The WOT flow is a flow which is generated present the trend of WOT flow clearly. In
from the rooftop of upstream building and interference tests, most of the pressure
affect to the downstream building in coefficients in the front face of the principal
interference effects. To test the characteristic of models show negative value compared to
WOT flow, a principal model is designed in positive pressures on the front face of the
the shape of a square cylinder with height H0 = isolated test. These positive pressures on the
600mm and width B = 100mm. Two front face and negative pressures on the rear
interference rigid models are also prepared. face makes the drag coefficient of the whole
One interference model is the same height as model turn to a positive value. When the
the principal model (Figure 8). While the other interference models located near the principal
is in double height to the principal model model at x/B less than 2, no significant
(Figure 9). Our main idea is to assess the flow difference between two different interference
characteristics of WOT flow via tests of the cases is found. The WOT flow has no effect on
same height and double height interference the principal model in these cases. From x/B =
models. The same height interference model 3 to 6, the pressure coefficients on the front
produces the flow which contains the WOT face of the principal model in same height
flow on the principal model. On the other hand, interference tests change gradually in each
the double height interference model is high distance. The areas near the top firstly turn to
enough to ignore the effects of the WOT flow positive pressures at distance x/B = 3. In this
on the principal model. case, the shear layer from the WOT flow could
move downward and reach the front face of the
Pitot tube
Interference model
downstream model. When the same height
Principal model interference models move away from the
B
U0 principal model, the area of positive pressure
on the principal model increased with the
H0

increase of the distance. From x/B = 7, the top

H0

Side flow
pressure of the front face of the principal
z

x
Interference model
building change the trend when backing to the
negative value. This phenomenon clearly
Figure 8. Configuration of same height interference presents the vortex from the roof top of the
test interference model.

450
Figure 10. Pressure coefficients on the principal model: i) Front face (Figure 1a – 9a); ii) Side face
(Figure 1b – 9b)

i) ii)

Figure 11. Different stage of rooftop wind over a single building: i) Initial stage; ii) Convergence stage
(Williamson and Govardhan 2004)

At the side face, the pressures along the face of the interference model becomes the 3D
building height in both cases of interference arch type vortex. This arch type vortex needs a
tests do not show the variation in the closed long distance enough to contribute to local
distance x/B<6 (Figure 10ii)). In this range, the pressure on the side face of the principal model.
wind from the rooftop could not fully develop to
contribute effects to the side face. However, in 3.3 Condition to form the WOT flow
further distance, x/B>6, the WOT in Through the simulation and the pressure
interference same height presents clearly when results on the front face of the downstream
showing the difference with the double high building. The WOT flow could be observed
interference test. The rooftop wind combined with two main parts (Figure 12). First, the
with the Karman vortex generated from the side shear layer is generated on the rooftop of

451
upstream building and intentionally moved In tandem arrangement with distance
downward to affect to the downstream 3  x/B  6, the downstream model play a
building. Secondly, the unsteady vortex is blocking role (Figure 13). The initial vortex
located under the shear layers tipped at the could not develop to the convergence stage
edge top of the upstream building. The size of with the Karman vortex. This vortex keeps
the vortex is changed instantaneously. growing up the size, release energy and return
to the smallest size. The tip of vortex is
Shear layer
always stay at the edge top of upstream
building. When the distance of two square
Interfe Princi cylinders increase x/B > 6, the large area in
rence pal between two models allow the vortex to
transform to the Karman vortex in shape of
Unsteady vortex
leg form (Figure 14).

4. CONCLUSIONS

In this paper, GPU application is applied in the

Figure 12. Two assumption parts of the WOT flow
CFD simulation of the WOT flow. GPU shows
Flow over the top of isolated building was its time saving advantages in CFD simulation
initially form as the vortex which tip at the in large number of mesh cells. Single GPU
edge top (Figure 11i). Then, this vortex is simulation should be used in the moderate
modified by Karman vortex which is generated number of mesh cell where the multiple GPU
from the side face of building. At convergence solver does not bring advantage to the
stage, the Karman vortex in form of leg vortex execution time. When simulating the large-
(Figure 11ii) is observed in research of scale simulation, multiple GPU accelerates the
Williamson and Govardhan (2004) speed of execution to 2.5-3 times. The
execution time of GPU solver is non-linear
with time steps due to the unstable of the
algorithm solution on GPU. In this case, a
simulation on a single GPU is equivalent to the
case of 15 CPU.
The WOT flow generated from square
cylinder high-rise buildings is investigated and
systematically studied through CFD
Figure 13. The flow pattern in same height simulation. WOT flow exits in the form of
interference x/B = 5 shear layer generated from the rooftop of
upstream building and the unsteady vortex
located under the shear layer tipped at the edge
top of building. The shear layer affects to the
downstream buildings by increasing the
pressure of front face and side faces near the
top area. The affected area from WOT flow
increases when the distance between two
buildings increases. The condition to form the
Figure 14. The flow pattern in same height WOT flow is governed by the distance of two
interference test x/B = 9 buildings 3  x/B  6.

452
5. REFERENCES Rathnayake, T., S. Jayasena and M. Narayana
2017. OpenFOAM on GPUs using AMGX.
Cheng, C.-M., C.-L. Fu and Y. Chen 2010. Proceedings of the 25th High Performance
Numerical simulation and validation of wind Computing Symposium, Society for Computer
load and structural responses of an isolated tall Simulation International.
building in turbulent boundary layer flow. The Shih, T.H., W. W. Liou, A. Shabbir, Z. Yang and J.
Fifth International Symposium on Zhu 1995. A new k-ϵ eddy viscosity model for
Computational Wind Engineering, USA May high reynolds number turbulent flows.
23-27, 2010. Computers & Fluids, 24(3), 227-238.
Davidson, L. 2003. Modification of the V2F model Shih, T.H., J. Zhu and J. L. Lumley 1993. A
for computing the flow in a 3D wall jet. realizable Reynolds stress algebraic equation
Turbulence Heat Mass Trans., 4, 577-584. model.
Dragoiescu, C., J. Garber and S. Kumar 2006. A Spiegel, M., T. Redel, Y. J. Zhang, T. Struffert, J.
comparison of force balance and pressure Hornegger, R. G. Grossman, A. Doerfler and
integration techniques for predicting wind- C. Karmonik 2011. Tetrahedral vs. polyhedral
induced response of tall buildings. Structural mesh size evaluation on flow velocity and
engineering and public safety proceedings of wall shear stress for cerebral hemodynamic
the structures congress. simulation. Computer methods in
Hellsten, A. 1998. Some improvements in Menter's biomechanics and biomedical engineering,
k-omega SST turbulence model. 29th AIAA, 14(01), 9-22.
Fluid Dynamics Conference. Sy, L. D., H. Yamada and H. Katsuchi 2019.
Kawai, H., Y. Okuda and M. Ohashi 2009. Three Interference effects of wind-over-top flow on
dimensional structures of flow behind a square high-rise buildings. Journal of Wind
prism. The Seventh Asia-Pacific Conference Engineering and Industrial Aerodynamics,
on Wind Engineering. 187, 85-96.
Kurashov, V. 2016. Modification of Taniike, Y. 1992. Interference mechanism for
RhoCentralFoam Solvers Based on enhanced wind forces on neighboring tall
OpenFOAM and RapidCFD to Calculate buildings. Journal of Wind Engineering and
High-Speed Flows. Speed Comparison of Industrial Aerodynamics, 42(1-3), 1073-1083.
GPU and CPU Solvers. Tominaga, Y., A. Mochida, R. Yoshie, H. Kataoka,
Launder, B. E. and D. B. Spalding (1983). The T. Nozu, M. Yoshikawa and T. Shirasawa
numerical computation of turbulent flows. (2016). AIJ Benchmarks for validation of
Numerical prediction of flow, heat transfer, CFD simulation.
turbulence and combustion, Elsevier: 96-116. Wilcox, D. C. (1998). Turbulence modeling for
Lien, F.-S. and G. Kalitzin 2001. Computations of CFD, DCW industries La Canada, CA.
transonic flow with the v2–f turbulence Williamson, C. and R. Govardhan 2004. Vortex-
model. International Journal of Heat and induced vibrations. Annu. Rev. Fluid Mech.,
Fluid Flow, 22(1), 53-61. 36, 413-455.
Lien, F., W. Chen and M. Leschziner 1996. Low- Yakhot, V., S. Orszag, S. Thangam, T. Gatski and
Reynolds-number eddy-viscosity modelling C. Speziale 1992. Development of turbulence
based on non-linear stress-strain/vorticity models for shear flows by a double expansion
relations, Engineering Turbulence Modelling technique. Physics of Fluids A: Fluid
and Measurements-3 (W. Rodi and G Dynamics, 4(7), 1510-1520.
Bergeles, Eds.). Rodi, W., Bergeles, G.(eds.), Yamagishi, T. and Y. J. P. C. S. Matsumura 2016.
91-100. GPU acceleration of a non-hydrostatic ocean
Menter, F. and T. Esch 2001. Elements of model with a multigrid Poisson/Helmholtz
industrial heat transfer predictions. 16th solver. 80, 1658-1669.
Brazilian Congress of Mechanical
Engineering (COBEM), sn.
Nocente, A. and T. K. Nielsen 2016. A Study of
Flow inside a Centrifugal Pump: High
Performance Numerical Simulations Using
Graphics Cards and Experiments.

453
The 3rd Int. Conf. on Transport Infrastructure & Sustainable Development (TISDIC 2019). Construction Publishing House. ISBN: 978-604-82-2893-4.

Dynamic responses of a functionally graded beam on

nonlinear foundation under moving harmonic loads by
using different theories

Huynh Van Quang

Faculty of Civil Engineering, Binh Duong University, Binh Duong province, Vietnam
E-mail: [email protected]

ABSTRACT: In the present study, dynamic responses of a functionally graded materials(FGMs) beams
on nonlinear foundation due to moving harmonic loads has been investigated by using Euler-Bernoulli,
Timoshenko and the third-order deformation beam theories. Material properties of the beam vary
continuously in thickness direction according to a power law form. The nonlinear behavior of foundation
is represented by the third order relationship. The governing equation of motion of the beam is derived
based on Hamilton principle expressed as Lagrange’s equations with specific boundary conditions
satisfied with Lagrange’s multiplier and sporadic by high order polynomial. The computer program using
Newmark- time integration and Newton Raphson procedure is written by MATLAB language. The
effects of the material distribution, velocity, excitation frequency, parameters of foundation as linear,
shear and nonlinear layer on the displacement of the beam have been examined thoroughly to draw some
useful conclusions.
Keywords: Beams, functionally graded beams, nonlinear foundation, moving harmonic loads, dynamic
analysis.

1. INTRODUCTION applications such as defense industries,

aircrafts, space vehicles, etc.
With the trend of development, materials Studies devoted to understand the static and
technology industry has introduced some new dynamic behavior of the functionally graded
material in practical applications. In terms of
beams on the ground under dynamic loads is one
mechanics, multi-layer composite material has
of the problems attracting many researchers
the advantages over base materials to create it
because they are distributed in the appropriate around the world. This problem is highly
position with the features of each type of practical, to describe behavior for many
material. About 3 decades ago, Japanese structures such as roads, railways, airport landing
scientists have introduced a material developed surfaces and horizontal fluid ducts. To survey
from multiple layers of composite material that behavior for many on the above structure, we
is functionally graded material. FGMs are need to select the model of beams and ground
special composites that have a continuous soil to suit the structural analysis problem.
variation of material properties from one Based on the above discussions, there have
surface to another. These new kinds of material been several investigated on free vibration
have been employed for a wide range of analysis of functionally graded beams. Simsek

454
(2010b) nonlinear dynamic analysis of FGMs moving load and moving mass supported by
beams subjected to moving harmonic loads Pasternak foundation based on Euler-Bernoulli.
based on Timoshenko theory combining Along with the development of computer
deformation and nonlinear relationship Von- science, humankind has proposed many models
Kerman and equations of motion of the system with different foundation are more parameters
obtained from the equation Lagranges, vertical recommended for accurate results, close to
and horizontal displacements are approximated reality Matsunaga (1999). The multi-parameter
by polynomial forms. foundation model was created to describe
Simsek (2010a) vibration analysis of FGM nonlinear behavior in the process of working
beams in different theory subjected to moving with the above structure. Recently, a number of
mass and the effect of shear deformation, the third-order nonlinear foundation models
inertia, Coriolis, the moving mass on the proposed the relationship between force and
dynamic displacements and the tresses of the displacement in the structural dynamics
beam. Simsek etal. (2006), (2007) the dynamic problem on the ground, giving results close to
behavior of simply-supported isotropic beams reality. Senalp etal. (2010) confirmed through
subjected to an eccentric axial load and a comparative research results turn deformed rails
moving harmonic load was analyzed by using foundation model of linear, nonlinear and
Euler–Bernoulli, Timoshenko and the third experimental foundation fact. In addition to
order shear deformation theory. Simsek (2010c) analyzing nonlinear behavior foundation, it
fundamental frequency analysis of FG beams by makes more viscous review, to describe more
different beam theories. More recently, realistically. Hung et al. (2017) study the
Mohanty etal. (2011) investigation of the oscillation frequency of Timoshenko beam with
dynamic stability of (FGSW) and (FGO) by any number of cracks. Cracks are replaced by
FEM on elastic foundation. Uzzal etal. (2012) springs with stiffness. Banerjee et al. (2016)
dynamic responses of beam subjected to dynamic response of cracked FGMs.

b : width of the cross section w0 : transverse displacement of any point on the

h : depth of the cross section neutral axis
v p : velocity of the moving harmonic load u0 : axial displacement of any point on the neutral
: excitation frequency axis
: the area of the cross-section t1 : time the load P( t ) comes onto the beam
 : natural frequency t2 : time the load P( t ) leaves the beam
 : dimensionless natural frequency P0 : amplitude of the moving harmonic load
 : Lagrange multipliers
P( t ) : concentrated moving harmonic load
z : z coordinate
x : x coordinate q( t ) : generalized coordinates
N : number of the terms in the displacement KijS : matrices due to Lagrange multipliers
functions
c( t ) : unit step function
M : mass matrix of the beam
L : length of the beam Bn : time-dependent generalized coordinate of the
K : kinetic energy of the beam axial displacements
I : moment of inertia of the cross section An : time-dependent generalized coordinate of the
w : transverse displacement transverse displacements
v : velocity of the moving harmonic load Cn : time-dependent generalized coordinate of the
V : potential of the external loads rotation displacements
U : strain energy of the beam  xx : normal stress
u : axial displacement  xx : normal strain

455
k : power-law exponent xP : location of the concentrated moving harmonic
 : lagrangian functional of the problem load
PT : material properties of the top surfaces k L : linear stiffness matrix
VT : the volume fractions of the top surfaces k NL : nonlinear stiffness matrix
0 : the rotation of the cross-sections I A ,I B ,I D : inertial coefficients
 xz : the shear strain f ( t ) : generalized load vector
K1 :dimension linear layer parameters Axx : extensional rigidity
K 2 : dimensionof shear layer parameter PB : material properties of the bottom surfaces
Ccr : drag coefficient tipping VB : the volume fractions of the bottom surfaces
F( t ) : the time-dependent generalized load vector  ( z ) :the characteristic shape function of
generated horizontal slip and stress distribution along the
 : ratio of resistance beam height
t : time kG : stiffness of shear foundation
K3 : dimension of nonlinear layer parameter

Timoshenko beam are obtained by a beam on nonlinear foundation under a moving

numerical technique using Ritz approximation. harmonic load. Newmark and Newton Raphson
Most recently, analyzed the dynamic behavior method is applied to the equations of motion of
of beams on the nonlinear foundation under to the beam. Also, examine the influence of
load Dahlberg (2002) and Ding et al. (2013). parameters such as the effects of the material
The stability analysis of beam cross section distribution, velocity of the moving load,
changes based nonlinear elastic foundation parameters of foundation as linear, shear layer
subjected to moving load by FEM, the results and nonlinear layer on the dynamic responses
indicate the influence of the boundary of the beam are discussed in detail. Numerical
conditions of the beam with parameters results show that the above mentioned effects
different foundation to the vertical play very important role on the dynamic
displacement of the beam Costa et al. (2015). deflections of the beam.
Kananiet al. (2014) and Fallahet al. (2011)
study nonlinear behavior of FG beams on 2. FUNCTIONLLY GRADED MATERIALS
nonlinear elastic foundation under to axial
force based on Euler-Bernoulli. Through this, We consider aFGMs beams: L, b, h, with
the foundation model is being researchers coordinate system (Oxyz) having the origin O
worldwide interest and also suggested in recent and respectively, resting on an nonlinear
times. Dynamic analysis of beam on a new foundation with cubic nonlinearity and
foundation model nonlinear behavior under to shearing layer.
a moving mass Nguyen et al. (2016). The
investigation of the effects of inertial,
centripetal, and Coriolis forces on the dynamic
response of a cracked beam under moving
mass load Pala et al. (2013). However, the
behavior of the beams on the nonlinear
foundation under a moving harmonic load has
not been studied much, especially the
application materials in the form of structural
FGMs has not mentioned much. The aim of Figure 1. AFGMs beams on nonlinear foundation
this paper is to investigate behavior of FGMs due to moving harmonic loads

456
Figure 2. Variation of the elasticity modulus E and the mass density along the thickness of the FGMs beams

The beam subjected to a moving harmonic From equation (4) we have

load P(t) which moves in the axial direction of h
the beam with constant velocity, vp is show in z   ,E  Em ,  m ,G  Gm ,    m
2
Figure 1.
It is assumed that the FGMs beams is made h
z  ,E  Ec ,  c ,G  Gc ,   c
of ceramic and metal. According to the rule of 2
mixture, the effective material properties of the
beam, i.e., Young’s modulus E and mass 3. THEORY AND FORMULATIONS
density  by Simsek (2010a). The problem considered in this study is
P  PT VT  PBVB (1) investigated under the following assumptions:
(a) The problem is both geometrically and
VT  VB  1 (2) materially linear, (b) the beam is initially at
The effective material properties of the rest, namely the initial conditions of the beam
FGMs beams is defined by the power-law are zero, and (c) the velocity of the moving
form. The volume fraction of the upper harmonic load is constant and the moving load
constituent of the beam by Simsek (2010a). is in contact with the beam at all times.
k Based on the third order shear deformation
 z 1 theory, the axial displacement, ux and the
VT     (3)
h 2 transverse displacement of any point of the
The variation of E and  along the thickness beam, uz.
of the beam (see Figure 2). Therefore, from w ( x,t )
u( x,z,t )  u0 ( x,t )  z 0 
Eqs. (1)-(3), the effective material properties of x
the FGMs beams can be expressed as follows: (5a)
 w ( x,t ) 
k
 ( z )  0  0 ( x,t ) 
 z 1  x 
E( z )  ( Ec  Em )     Em (4a)
h 2 w( x,z,t )  w0 ( x,t ) (5b)
k Different beam theory is determined
z 1
 ( z )  ( c   m )      m (4b) Euler-Bernoulli beam theory (EBT):
 h 2 
k
( z )  0 (6a)
 z 1
 ( z )  ( c  m )      m (4c) Timoshenko beam theory (FSDBT):
h 2 ( z )  z (6b)
E( z ) Third-order deformation beam theories
G( z )  (4d)
2[ 1   ( z )] (PSDBT):

457
  4z2  The deformation energy in the nonlinear
 ( z )  z 1   (6c) foundation is Winkler spring deformation, shear
 3h 2 
deformation, nonlinear deformation as follows:
The strain-displacement relations are given by 2
1 1  w 
u u0  w 2   w  
2 UF 
2L k L w2 dx   kG 
2 L  x 
 dx 
 xx    z 20   ( z )  20  0  (7a) (12)
x x x  x x  1
  k NL w4 dx
u w  ( z )  w0  4L
 xz      0  (7b)
z x z  x  From Eqs. (5a), (5b) that the velocities of
By assuming that the material of beam obeys any point on the beam take the form:
Hooke’s law, the stresses in the beam are u0  2 w0   2 w0 0 
 xx  E( z ) xx vx  u x  z   ( z )    (13)
(8a) t xt  xt t 
 xz  G( z ) xz (8b)
w0
vz  w  (14)
The strain energy of the beam at any instant t
in Cartesian coordinates is where (.) indicates the derivative with respect
1 to time. The kinetic energy of the beam at any
U    xx xx   xz xz dAdx (9)
2L instant is
1
2 V
Substituting Eqs. (7a,7b), (8a, 8b) into Eq. K  ( z )( vx2  vz2 )dAdx (15)
(9) we have
  u 
2    u  2 
 Axx  0    I A  0   
   x  
 2    t  

 2  E xx  Bxx  0 u  w 0   u0   w0  
 2( I E  I B )
 x x 2    
   t t  x  
u0  0 
 2 E xx   2
1  x x  (10) ( I  I  2 I )    w0   
U   F    
2 dAdx
1 
D H
2 L  2
 w0  K    t  x   dx (16)

  Dxx  H xx  2 Fxx     
2  2 L   w  

   x   2( I F  I H )  0  0  
 t  x  t 
  2 w0  0   
 2  Fxx  H xx  2    2

 x x   0  u0 0
   I H    2I E  
 w0 
2
  t  t t 
  Axz   0  
  x    2 
 w
 I A  0   
where   t  
 Axx ,Bxx ,Dxx    E( z )( 1,z,z 2 )dA By defining the following cross-sectional
A inertial coefficients:
( Exx ,Fxx )    ( z )E( z )( 1,z )dA ( I A ,I B ,I D )    ( z )( 1,z,z 2 )dA (17a)
A
A
2 (11)
( H xx )    ( z ) E( z )dA ( I E ,I F )    ( z )( 1,z )dA (17b)
A
A
2
  ( z ) 
( Axz )  G( z )dA ( I H )   [  ( z )] 2  ( z )dA (17c)
A
z  A

458
 Potential of the moving harmonic load at After substituting Eqs.(22a), (22b) and (22c)
any instant: into Eq. (20) and then using the Lagrange’s
equations given by Eq. (23).
V   P( t ).w0 ( x p ,t )[ c( t  t1 )  c( t  t2 )] (18a)
 d 
 .  0 n  1, 2 ,...,3 N  8 (23)
1 t  0 qn dt q n
c( t )   (18b)
0 t  0 The following definitions are made in
P( t )  P0 sin(  t ) (18c) Lagrange’s equations:
qn  An n  1, 2 ,3,...,N (24a)
where t1 is the time when the load P( t ) just
comes onto the beam ( t1 is considered as zero qn  Bn N n  N  1,..., 2 N (24b)
in this study), t2 is the time when the load qn  Cn 2 N n  2 N  1,...,3 N (24c)
P( t ) just leaves the beam and xP ( t ) is the
q3 N 1  1 ,q3 N  2   2 ,q3 N 3   3 ,q3 N  4   4 ,
location of the moving harmonic load at any (24d)
instant and represented as: q3 N 5   5 ,q3 N  6   6 ,q3 N 7   7 ,q3 N 8   8

L L L Yields the following coupled non-linear

xP ( t )  vP .t  ,   xP ( t )  , systems of equations of motion:
2 2 2 (19)
t1  0  t  t2  M qi ( t )  C qi ( t ) 
(25)
The Lagrange multipliers formulation of the   K L  K NL q( t ) q( t )   F( t )
considered problem requires constructing the
where q( t )   A( t ),B( t ),C( t ), ( t ) are the
T
Lagrangian functional as follows:
generalized coordinates.The non-linear equations
Gs  1w0 (  L / 2 ,t )   2 w0 ( L / 2 ,t )  of motion (25) in the time domain can be solved
w0 w by using Newmark- time integration and
 3 (  L / 2 ,t )   4 0 ( L / 2 ,t ) 
x x (20) Newton Raphson procedure is written by
 5u0 (  L / 2 ,t )   6u0 ( L / 2 ,t )  MATLAB language and then the displacements,
velocities and accelerations of the beam at the
 70 (  L / 2 ,t )   80 ( L / 2 ,t )
considered point and time are determined for any
The functional of the problem is time t between 0  t  L / v p .
  K  Gs  (U  U F )  V (21) The non-linear stiffness matrices  K NL  and
In order to apply the Lagrange’s equations, external load vector  F( t ) as zero. The
the unknown functions, w0 ( x,t ), u0 ( t ), 0 ( x,t ) following frequency equation is obtained
are approximated by space dependent which can be expressed:
polynomial terms, x 0 ,x1 ,x 2 ,...,x N 1 and time  K   q    2  M  q   0 (26)
dependent generalized coordinates: An ( t ),
The natural frequencies, i and
Bn ( t ) and Cn ( t ) .
corresponding mode shapes, are obtained by
N setting the characteristic determinant of the
w0 ( x,t )   An ( t )x N 1 (22a) characteristic equation from (26) equal to zero.
1
The following non-dimensional frequency
N
parameter is defined by Simsek (2010).
u0 ( x,t )   Bn ( t )x N 1 (22b)
h/ 2
1 I1
N
   L2 h/ 2
; I1    ( z )dz (27)
0 ( x,t )   Cn ( t )x N 1
(22c) h2  E( z )dz  h/ 2

1  h/ 2

459
4. NUMERICAL RESULTS performance of the present study. Firstly, to
further verify the present results, natural
In the numerical results, dynamic responses of a frequencies of FGMs beams composed of
FGMs beams on nonlinear foundation under to alumina and aluminum are calculated and
moving harmonic loads has been investigated. compared with before study for k = 0.3 and
L / h = (10, 30, 100) in Table 6. The following
Table 1. Material property of FGMs
material and beam properties: h = 0.5m.
constituents.
Alumina: Ec  380( GPa ) ,
Property Unit Aluminum Zirconia
(Al) ( ZrO2 ) c  3800( kg / m3 ) , c  0.23
E GPa 70 200 Aluminum: Em  70( GPa ) ,
 kg/m 3
2700 5700
 m  2700( kg / m3 ) , m  0.23
 - 0.3 0.3
Table 5. Examined the number of the time
Table 2. The dimensions and attribute of the steps (RL)
beam
RL 250 500 750 1000
L( m ) b( m ) h( m ) k ks
wmax (m) 0.068706 0.068603 0.068563 0.068549
20 0.5 0.5 1 5/6
Table 6. Comparison of non-dimensional
Table 3. Properties of foundation fundamental frequencies of FGMs beams
K1 K2 K3 (%) Boundar Author 
10 5 20 5 y L/h L/h L/h
condition
10 30 100
Table 4. Properties of load
S-S Simsek FSDBT 2.70 2.74 2.74
P0 ( kN )  ( rad / s ) vP ( m / s ) (2010c) PSDBT 2.70 2.74 2.74
1500 20 20 Kocatük FSDBT 2.77 2.81 2.81
(2006) PSDBT 2.69 2.73 2.74
The following non-dimensional variables
Article FSDBT 2.70 2.74 2.74
kG L2 k L L4
Ccr  2 kl  m A K2  K1  (28) PSDBT 2.70 2.74 2.74
 2 EI EI
C-F Simsek FSDBT 0.97 0.98 0.98
 k L6
  K3  NL (2010c) PSDBT 0.97 0.98 0.98
Ccr EI
Kocatük FSDBT 0.99 1.00 1.00
Displacements obtained according to EBT, (2006) PSDBT 0.96 0.97 0.97
FSDBT and PSDBTare compared with each
other. The material distribution, velocity of the Article FSDBT 0.97 0.98 0.98
moving harmonic load, excitation frequency, PSDBT 0.97 0.98 0.98
parameters of foundation as linear, shear and C-C Simsek FSDBT 5.88 6.18 6.21
nonlinear layer effects on the dynamic (2010c) PSDBT 5.88 6.18 6.21
responses of the beam are investigated.
The number of the time steps RL = 250, the Kocatük FSDBT 6.01 6.34 6.38
largest displacement between the beams (2006) PSDBT 5.81 6.16 6.21
converge and the subsequent calculations (see Article FSDBT 5.88 6.18 6.21
table 5). The numerical results are compared
PSDBT 5.88 6.18 6.21
with the previous works to demonstrate the

460
 The comparisons show that the agreement Mohanty (2011). The comparison is provided
between the present results and those of in Table 7. It is found that the present results
Simsek (2010c),Kocatük(2006) is satisfactory. are in good agreement.
FGMs Timoshenko beams analysis with the Chart of the beam displacement over time
following parameters: (see Fig. 4) relatively consistent with the
L  20m,b  h  0.5m,k  1,k s  5 / 6 , solution Ding et al. (2013). The article applies
P0  2000kN ,vP  20m / s,  0rad / s Timshenko beam theory with the generalized
coordinate method while the result Ding et al.
The present results (see Fig. 3) are in good (2013) applies Euler-Bernoulli beam theory
agreement with the results of Simsek (2010b). with the Galerkin and Runge-Kutta methods.
So the displacement line in 1/3 of the time at
the beginning and 1/3 of the last time, there is a
difference but not significant and it is
acceptable. Relatively small displacement
value less dangerous for the structure.
Fig. 5 shows the difference in oscillation
frequency results dimensionless between
different beam theory decreases when L / h
increases, this is consistent with the physical
properties when L / h increases, the effects of
Figure 3. Maximum transverse displacements shear deformation decreases. With L / h > 20,
curves for various values of N the effects of very small shear deformation
should result oscillation frequency
Survey beams (S-S) with the following dimensionless between theory beams are
parameters for k  0, L  b  1m, almost equal. Besides, as hard links, the
h  [ 0.1; 0.2; 0.5 ]m, E  206( GPa ), increasing frequency of oscillation. Therefore,
  7860( kg / m3 ) are calculated and consideration of additional shear deformation
compared with those of Matsunaga (1999) and is very important for structural analysis.

Table 7. Comparison of fundamental non-dimensional frequency

1  1L2  A / EI
L/h K1 K2 Article
Matsunaga (1999) Mohanty (2011)
FSDBT PSDBT
0 0 7.412 7.449 7.480 7.487
10 0 8.010 8.043 8.071 8.077
100 0 12.108 12.124 12.13 12.140
2
0 1 12.010 - 12.040 12.043
10 1 12.383 - 12.411 12.413
100 1 15.315 - 15.329 15.330
0 0 9.274 9.286 9.296 9.296
10 0 9.784 9.796 9.805 9.806
5 100 0 13.540 13.548 13.555 13.555
0 1 13.447 - 13.461 13.461
10 1 13.804 - 13.818 13.818
100 1 16.678 - 16.688 16.689

461
Figure 4. The vertical deflection of beam k = 0with time t

Figure 5. Variation of the first dimensionless frequency parameter with different boundary conditions

Figure 6. The largest displacement of the beam with L / h ratio for various values

462
Figure 7. Displacements of FGMs beams under to moving load P0 = 1500 (kN), vp = 20 (m/s),  = 20 (rad/s)

Table 8. Properties of the beam, foundation and load for Fig.4

Item Notation Value Nondimensional value

E (Gpa) 201 -

 (kg/m3) 7850 -

A (m2) 7.69  103 -

Beam
I (m4) 3.1  105 -

kf - 3.501  105

L 18 1

kL (N/m2) 3.5  107 7.0221

Foundation kNL (N/m2) 4  1014 3.6  1010

 (Ns/m2) 1732.5  103 99.879

Load P0 (N) 65  103 4.025  105

v (m/s) 10 0.001933

463
 Figure 6 increasing velocity of displacement From Figure 13. When the ratio Ec/Em larger
of the beam tends to decrease. The largest the beam contains many components in
displacement depends on the size of beam. ceramic. Large modulus of elasticity of the
Besides, velocity from 20 to 40 m/s, the danger ceramic, so the stiffness of the beams
for structure. Figure 7 with L / h =2, L / h = 5 increases, leading to the reduction of the
differences in displacement results between system's displacement.
EBT beam theory with two remaining theories
are significant, while two beams theory of
FSDBT and PSDBT give equal displacement
results. With L / h = 20, all three beam theory
gives almost equal results. Show that, when
L / h increases, the effect of cutting deformation
decreases. Figure 8 increasing the material
distribution coefficient, the displacement of the
beam increases. At that time, the ceramic
material gradually changed to metal and metal
with small hardness, so soft texture.
The influence of the excitation force Figure 9. Time history of midspan deflections of
frequency of moving harmonic loads is very (S-S) beamsfor k = 1 and
large. The higher the excitation force, the more   [ 20; 40; 60 ]( rad / s )
the oscillation of the beams increases. Besides,
when the excitation force frequency of the
moving load with the frequency of the beam
vibration, it is very dangerous for the structure
under consideration (see Figure 9)
When increasing coefficient K1, K2, K3 the
largest displacement of the beam decreases (see
Figure 10, 11, 12). It is shown that when increase
foundation coefficient, it increases the hardness
of the substrate and leads to the stiffness of the
beam also increases. At the same time, the Figure 10. Time history of midspan deflectionsof
number of oscillations of beams increases. (S-S) beams for k 1 and
K1  [ 10;100;1000 ]

Figure 8. Largest displacement of beams with Figure 11. Time history of midspan deflectionsof
L/h ratio for various values of the (S-S) beams for k  1 and
power-law exponent K 2  [ 5;10;15 ]

464
  A structural system when leaning on a
platform with a large hardness, the
instability of the system will be
significantly reduced Increasing
velocity of displacement of the beam
tends to decrease.
 The difference from previous research is
the simulation of the properties of the
soil more fully and the results are
relatively similar to reality.
Figure 12. Time history of midspan deflections of
 Through the above survey, the
(S-S) beams for k 1 and
coefficients of moving load, FGMs beams
K3  [ 104 ;106 ;108 ] and foundation play a very important role
in the behavior of structure.
Appendix A
The equations of motion Eq. (25) can be
written in the explicit form as follows:
 K L    KW    K12
L  K13
L  K14
S  
   11   11   N  N   N N   N N   N 8 
   An ( t )
  K 21
L   K 22
L   K 23
L   K 24
S   
  N N   N N   N N   N 8   Bn ( t )
  
  K31
L  K32
L   K33
L   K 34
S 
 C n ( t ) 
     N N   N N   N 8  
 n 
N N
( t )
  K 41
S   K 42
S   K 43
S   K 44
S  
  8 N  8N  8 N  88 
  K11

NL
( An ( t ))
NN
 0N  N  0 N  N 0N 8   An ( t )
Figure 13. Time history of midspan deflections of   0N  N  0N  N  0 N  N 0N 8   Bn ( t )
(S-S) beams for k 1 and   
Ec / Em  [ 1; 3; 5 ]
  0N  N  0N  N  0 N  N 0N 8  Cn ( t )
  
 08N 08N 08N 088   n ( t )
5. CONCLUSIONS C11 N  N  0 N  N  0N  N 0N 8   A n ( t )
 
Dynamic responses of a FGMs beams on   0N  N  0 N  N  0N  N 0N 8   B n ( t ) (A1)
  
nonlinear foundation due to moving loads has   0N  N  0 N  N  0N  N 0N 8  C n ( t )
been investigated by using Euler-Bernoulli,   0
 8 N 08N 08N 088   n ( t )
Timoshenko and the third-order deformation
beam theories. From the results analyzed above  M 11 N  N  0N  N  0N  N 0N 8   An ( t )  f ( t )
 
the following conclusions are reached:   0N  N  M 22 N N  M 23 N N 0N 8   Bn ( t ) 0 
   
 As hard as structural link, the oscillation   0N  N  M 32 N N  M 33 N N 0N 8  Cn ( t ) 0 
frequency of the system will rise and   0
 8 N 08N 08N 088  n ( t ) 0 
vice versa.When calculating the
structure, do not ignore the shear 6. REFERENCES
deformation of the beam.
 The content of ceramic that is more than Banerjee A, Panigrahi B, and Pohit G. 2016. Crack
the metal content, the hardness of the modelling and detection in Timoshenko FGM
system will be increased, when under beam under transverse vibration using
pressure, the structure is safe. The higher frequency contour and response surface model
the excitation force, the more the with GA. Nondestructive Testing and
oscillation of the beam increases. Evaluation, 31 (2), pp. 142-164.

465
Castro Jorge, P, Simões, F.M.E. & Pinto da Costa, Nguyen T.P, Pham D.T, Hoang P.H. 2016. A new
A. 2015. Dynamics of beams on non-uniform foundation model for dynamic analysis of beam
nonlinear foundation subjected to moving on nonlinear foundation subjected to a moving
loads, Computers and Structures, 148, 26-34. mass. Procedia Engineering, 142, 166 - 173.
Dahlberg, T.2002. Dynamic interaction between Pala Y. and Reis M. 2013. Dynamic response of a
train and nonlinear railway track model. cracked beam under a moving mass load.
Proceedings of the Fifth European Conference American Society of Civil Engineers.
on Structural Dynamic, Munich, Germany Simsek, M. 2010b. Non-linear vibration analysis of
1155-1160. a functionally graded Timshenko beam under
Fallah, A. &Aghdam, M.M. 2011. Nonlinear free action of a moving harmonic load. Composite
vibration and post-buckling analysis of Structures, 92, 2532-2546.
functionally graded beams on nonlinear elastic Simsek, M. 2010a. Vibration analysis of a
foundation. European Joumal of Mechanics- functionally graded beam under a moving
A/Solids, 30, 571-83. mass by using different beam theories.
Kanani, A.S, Niknam, H, Ohadi, A.R. &Aghdam, Composite Structures, 92, 904-907.
M.M. 2014. Effect of nonlinear elastic Simsek, M. &Kocatürk, T. 2007. Dynamic analysis
foundation on large amplitude free and forced of an eccentrically prestressed damped beam
vibration of functionally graded beam, under a moving harmonic force using higher
order shear deformation theory. StructEng-
Composite Structures.
ASCE, 133(12).
Khiem N.T. and Hung D.T. 2017. A closed-form
Simsek, M. 2010c. Fundamental frequency
solution for free vibration of multiple cracked
analysis of functionally graded beams by
Timoshenko beam and application. Vietnam
using different higher-order beam theories.
Journal of Mechanics, 39(4), 315-328.
NuclEng Des, 240.
Kocatürk, T. &Simsek, M. 2006. Vibration of Senalp, A.D. &Arikoglu, A, Ozkol, I, Dogan, V.
viscoelastic beams subjected to aneccentric 2010. Dynamic response of a finite length
compressive force and a concentrated moving Euler-Bernoulli beam on linear and nonlinear
harmonic force. Sound Vib, 291(1-2), 302-22. viscoelastic foundations to a concentrated
Matsunaga, H. 1999. Vibration and buckling of moving force. Journal of Mechanical.
deep beam-columns on two - parameter Uzzal. R.U.A., Bhat, R.B. & Ahmed, W. 2012.
elastic foundation, Department of Dynamic response of a beam subjected to
Architecture, Setsunan University, 17-8, moving load and moving mass supported by
Ikeda-naka-machi, Neyagawa, Osaka, 572- Pasternak foundation. Shock and Vibration.
8508, Japan. Composite Structures. Yang, Y., Ding, H. & Chen, L.Q.2013. Dynamics
Mohanty, S.C, Dash, R.R. & Rout, T. 2011. response to a moving load of a Timoshenko
Parametric instability of a functional graded beam resting on a nonlinear viscoelastic
Timoshenko beam on Winkler’s elastic foundations. ActaMechanicaSinica, 29(5), 718
foundation. Nuclear Engineering and design, - 727.
241, 2608-2715.

466
The 3rd Int. Conf. on Transport Infrastructure & Sustainable Development (TISDIC 2019). Construction Publishing House. ISBN: 978-604-82-2893-4.

Experimental study on ground ware propagations due to

the impact of road vibration rollers with a damping mass

Nguyen Lan1*, Hoang Phuong Hoa1, Hoang Thai An1

1
The University of Danang- University of Science and Technology

*Corresponding Author, E-mail: [email protected]

ABSTRACT: The operation of the road compacting by using vibrating rollers causes wave propagations
in the ground, which may affect neighbouring buildings. In recent years, many heavy road vibration
rollers are used in Vietnam for the new road construction or the road renovation, whichcan lead to the
damage of adjacent buildings near the construction sitedue to the vibration ground wave. In particular,
when the ground wave amplitude is large, it may cause significantdamage to nearby buildings as
demonstrated by several construction site observations. This paper presents an experimental study, where
the ground vibration data due to load cases of the vibration roller is measured and processed.Four load
cases on the ground due to vibrating rollers are proposedwithout and witha damping massto investigate
the effect of the damping mass on the ground wave velocity. Experimental results demonstrate the
capability of the damping mass in reducing the wave velocity of the ground by the vibrating rollers.
Keywords: Vibrator roller, road compacting, damping mass, ground wave, peak particle velocity.

1. INTRODUCTION for construction activities, e.g,TCVN 7378

In order to minimize the impact of the ground (2004) in Vietnam and DIN 4150-3 (1999).
vibration on the human living environment The regulations limiting the vibration effect
andinfrastructure, most of the countries have a caused by construction activities are
system of standards that limit the vibration level summarized as follows:

Table 1. Guidelines for assessing the impact of vibrations on buildings according to DIN 4150-3 (1999)
Type of structure Vibration thresholds for structural damage in terms of PPV (mm/s)
Short term Long term
At foundation Uppermost floor Uppermost floor
0-10 Hz 10-50 Hz 50-100Hz All frequency All frequency
Commercial/industrial 20 20-40 40-50 40 10
Residential 5 5-15 15-20 15 5
Sensitive/historic 3 3-8 8-10 8 2.5

467
 The base theory of the ground wave - Case 1: 1 HAMM 3410 vibratory roller
propagation problem has been mentioned in with the high vibration centrifugal force of
many literatures, e.g., Mark Svinkin (1996), 246 KN.
Arnold Verruijt (2010), Das and Ramana - Case 2: 1 vibratory roller DYNAPAC
(2011), and Prakash (1998).
CA30 IID with the high vibration centrifugal
force of 247 KN.
- Case 3: 1 HAMM 3410 vibrator and 01
DYNAPAC CA30 IID running simultaneously.
- Case 4: 1 HAMM 3410 vibrating roller (01
DYNAPAC CA30 IID fixed in damping mass
position). This load combination uses a
DYNAPAC CA30 IID vibratory roller with a
total weight of 12.55 tonsand a dimension of
Figure 1. Solution using trench seismic insulation
5.5 2.4 m using as a wave energy absorption
(Persson 2013) for the HAMM 3410 roller on the ground.
To obtain the most unfavourable vibration
To reduce the impact of the ground vibration intensity, the rollers move back and forthand
due to pile construction activities, many then stand at the position of the sensors to
measurement methods have also been obtain different intensity levels.
proposed, such as Trench seismic insulation or
changing the land surface (DIN 4150-3: 1999,
Tran Dinh Ngoc 2002, Peter Persson 2013,
Richart et al.1970). Figure 1 shows the solution
of the Trench seismic insulation method.
In road constructions, especially road
renovations with existing buildings on both
sides, the use of digging trenches is relatively
difficult and expensive. An alternative method
is the use ofmass spreaders to absorb the wave
energy and reduce the peak particle velocity
(PPV) on adjacent buildings which can
minimize the failure of the buildings due to the
ground vibration.

2. EXPERIMENTAL STUDY
2.1 Experimental setup
The experimental measurement of vibration
rollerswasin Hoa Phu road, Da Nang City; this
was the framework of the Hoa Ninh project.
The layouts of measuring points of different
load casesare shown in figures2, 3 and 4.
Geophone velocity sensors are used together
witha datalogger. The following load combinations
were implemented and received continuous
vibration data by a software built on the
LabVIEW environment: Figure 2. Layouts of the ground wave measuring

468
 having the distance to the road from 5 to 10 m
will be affected by the vibrating rollers.
The solution of using the damping mass
shows that the impact radius to neighbouring
buildings is reduced from 15% to 26%, as
shown in Table 3.

Figure 3. Layout of the vibration measurement with

a damping mass usinga Dynapac roller

Figure 5. Ground wave spectrum of roller HAM

3410 roller at point 1

Figure 4. Monitoring ground wave due to the

vibration roller at site

2.2 Experimental results

The maximum vibration velocities in the
vertical direction as a function ofthe distance
fromthe measuring points were obtained.
Statistical analysis was employed to build the
experimental curve showing the relationship
between PPV and the distance from the Figure 6. Example of FFT analysis
vibrating roller to the review point R, as shown
in figures 7, 8, 9, 10 for the different load
cases. The summary of the PPV measured from
the test of the Case 1 (1 01 vibration roller
HAMM 3410) is presented in Table 2. The
impact radius of the ground wave to
neighbouring buildings is then calculatedbased
on the vibration velocity, as shown in Table 3.
The experimental results in figures 7, 8, 9, 10
show that the impact radius of the ground wave
to neighbouring buildings with a vibration
velocity allowance of 3 mm/s is from 37 to 40 Figure 7. Velocity-radius relationship of1 HAM
m. With this radius, the neighbouring buildings 3410 roller

469
Figure 8. Velocity-radius relationship of 1 Figure 10. Velocity-radius relationship of1 HAM
DYNAPAC roller 3410 roller and 1 DYNAPC rollerusing
as a damping mass

Figure 9. Velocity-radius relationship of 1 HAM Figure 11. Vibration spectrum of1 HAM 3410
3410 roller and 1 DYNAPC roller roller and a fixed DYNAPC
rollerusing as a damping mass

Table 2. Experimental data in terms of PPV

Source Loading case Direction Point 1 (5 m) Point 2 (23 m) Point 3 (45 m)
Vertical 17.4 4.1 1.7
push - recede Radius 17.4 4.0 1.8
Total 23.7 5.7 1.9
Vertical 22.5 4.7 1.8
01 vibration push – fixed
roller HAMM vibration - Radius 20.9 4.6 1.7
3410 recede
Total 24.1 6.6 1.9
Vertical 24.1 4.1 2.1
push - fixed
vibration – turn Radius 21.6 4.2 2.1
off - recede
Total 26.5 5.84 2.2

470
Table 3. The impact radius and the damping mass effect
Limit 01 HAMM 01 DYNAPAC 01 HAMM 3410 01 HAMM 3410 roller (01 Damping
velocity 3410 roller CA30 IID roller roller DYNAPAC DYNAPAC CA30 IID fixed mass
CA30 IID roller as damping mass position) effect
3 mm/s 37.1 27.1 39.5 31.4 15.4%
3.6 mm/s 33.5 22.9 34.3 24.8 25.9%

3. CONCLUSIONS STIC CenterUD2019. Report the results of ground

vibration measurement of Hoa Ninh-Hoa Phu
This paper presents an experimental test on the road, Da Nang.
road construction site that uses vibration rollers DIN 4150-31999. Vibration in buildings- part
for compacting the ground. Based on the effects on structurers.
experimental results, it can be concluded that: Tran D.N.2002. The surface wave of the soil due to
pilling and affecting adjacent construction.
The impact radiusof the ground wave to
Doctoral thesis in engineering, Hanoi.
neighbouring buildings with a vibration Persson, P.2013. Reduction ground wave vibration
velocity allowance of 3 mm/s is from 37 to 40 using wave obstacles. Licentiate Dissertation,
m. With this radius, the neighbouring buildings Structural Mechanics, Lund University,
having the distance to the road from 5 to 10m Sweden.
will be affected by vibrating rollers. MarkSvinkin1996. Minimizing Construction
The solution of using the damping masshas vibration Effects. Practice Periodical on
shown that the impact radius to neighbouring structure design and construction. ASCE/May
2004; 108-115.
buildings is reducedfrom 15% to 26%.
Arnold Verruijt2010. An introduction to soil
Therefore, this solution can be used for several dynamic, Springer Netherlands.
cases where it is necessary to reduce the radius Das, B.M., Ramana, G.V. 2011. Principles of Soil
due to vibration to neighbouring buildings. Dynamics, Second Edition, Cengage India.
Prakash, S.1981. Soil Dynamics. McGraw-Hill
4. REFERENCES Book Company, New York.
Richart, F.E., Hall, J.R., and Woods, R.D.
TCVN 73782004. Vibration and vibration to the 1970.Vibrations of soils and foundations.
buildings - limited vibration level and PrenticeHall, Inc.
assessment method.

471
The 3rd Int. Conf. on Transport Infrastructure & Sustainable Development (TISDIC 2019). Construction Publishing House. ISBN: 978-604-82-2893-4.

Damage detection in plate-like structures using improved

modal strain energy method

Le Cao Thanh1,2, Ho Duc Duy3,

1
Ho Chi Minh City University of Technology, VNU-HCM, Ho Chi Minh City, Vietnam
2
Faculty of Civil Engineering, Nha Trang University, Nha Trang, Vietnam
3
Faculty of Civil Engineering, Ho Chi Minh City University of Technology, VNU-HCM, Ho Chi Minh
City, Vietnam.

Corresponding author, E-mail: [email protected]

ABSTRACT: This paper presents an improved modal strain energy (MSE) method for damage detection
in plate-like structures. Firstly, the theory of MSE method established for plate-like structures is briefly
outlined. Therein, two-step MSE-based procedure, which is global step and local step, is proposed to
enhance the accuracy for detecting the location of damages. An aluminum rectangular plate constrained
by fixed boundary conditions is investigated in order to verify the feasibility of the proposed method. The
plate’s finite element models are analyzed to obtain natural frequencies and mode shapes for before and
after the occurrence of damage. A damage index is presented to identify the location of crack in the plate-
like structures. The analytical results show that the global MSE step identifies the damage zone in the
plate; then, the local MSE step accurately detects the crack’s length in the damage zone.
Keywords: Damage detection, modal strain energy, plate, structural health monitoring, vibration.

1. INTRODUCTION two-dimensional curvatures. In their method,

the fractional strain energy of the plate before
In the last two decades, Structural Health
and after crack occurrence was used to
Monitoring (SHM) have been playing a very
calculateda damage index which can
important role for the safety, longevity and
successfully locate the area with stiffness
sustainable performance of construction
reduction as low as 10% using relatively few
structures. The nondestructive detection
modes. Furthermore, in their experimental
methods for diagnosing structural failures,
validation, two edge cracks with certain
especially methods based structural vibration
severities were identified by damage index.
responses, have received the attention of
many researchers. Modal Strain Energy Kim et al. (2003) applied the MSE method
(MSE) is one of the highly effective methods for simple beam structure and accurately
for detecting damages in structures (Fan and estimated the damage position in the middle
Qiao 2011). and in a quarter span with only two mode
Stubbs et al. (1995) first applied the concept shapes. The severity of damage in the beam is
of MSE method to detect the damage of also determined accurately.
structures based on the changesin MSE of Hu and Wu (2008) developed a
structures. Subsequently, Cornwell et al. nondestructive damage detection methodusing
(1999) extended MSE method for an index of experimental results to detect surface cracks in
damage in plate-like structure characterized by a thin aluminumplate under a completely free

472
The 3rd Int. Conf. on Transport Infrastructure & Sustainable Development (TISDIC 2019). Construction Publishing House. ISBN: 978-604-82-2893-4.

boundary condition. Only few mode shapes of Y (i,j) (Nx,Ny)

the plate before and after occurrence of Ly
crackare required to calculate the MSE and
damage index. The differential quadrature
method used to calculatethe second derivative
of the mode shapes in strain energy formula by Y j+1
using only displacement data at nodes. Hu et Yj
al. (2011) detected damage in circular hollow
cylinder using MSE and scanning damage X
index methods. 0 Xi Xi+1
Le and Ho (2015) developed MSE method
Figure 1. Schematic of plate structure
for thin plateunder different boundary
conditions. In this study, Central Differential For a particular mode shape k, the MSE of
Method (CDM) was used to calculate the the sub-region (i, j) is determined as follows
second derivative of the mode shapes. (Cornwell et al. 1999, Loendersloot et al. 2017):
Fu et al. (2016) established a two-step Dij
y j 1 xi 1
  2
2
   2k 
2

procedure using the MSE method for U k ,ij    [ 2 k

  x    2  
diagnosing plate structures using isotropic
2 yj xi    y 
(1)
homogeneous materials: The first step focuses   2   k 2    2k 
2

on detection of damage area; in the second 2  2 k

  x 
  2 y   2 1      ]dx dy
    xy 
step, the locations and extents of the damages
are identified. Loenders loot et al. (2017) Where Dij is the flexural stiffness of the sub-
applied damage index successfully on region (i, j); v is the coefficient of Poisson.
monitoring of a composite plate with stiffeners. Total strain energy of the plate during elastic
For most of previous proposed methods, the deformation is given by:
MSE method isdeployed for the structure at Nx N y

global level. This procedure requires a massive U k  U k ,ij (2)

i 1 j 1
displacement data throughout the surface of
The fractional strain energy the sub-region
plateto ensure the accuracy of the diagnosis
(i,j) is determined as follows:
results. It isn’t only expensive but also U k ,ij
complex in practice. In this study, MSE was Fk ,ij  (3)
Uk
improved and applied to the preliminary
 i x1  j y1 Fk ,ij  1
N N
damaged area. As a result, required data are where
reduced significantly but still ensure the Considering m mode shapes,the damage
accuracy of the diagnostic results. index of the sub-region (i, j) is defined as:
 k 1Fk*,ij
m

2. MODAL STRAIN ENERGY-BASED ij  (4)

 k 1Fk ,ij
m
DAMAGE DETECTION METHOD
where, the symbol * denotes damaged state of
2.1 Global modal strain energy method the structure.
For plate structures, the mode shapeis The damage index after normalization is
characterized by a two-dimensional surface. determined as follows:
Considering a platesubdivided into NxxNy sub- ij  ij
Z ij  (5)
regionsand each sub-region (i, j) islocated by  ij
the coordinates of two points (xi, yj) and where ij , ij are mean and standard deviation
(xi+1, yj+1) as shown in Figure 1. of the damage indices, respectively.

473
The 3rd Int. Conf. on Transport Infrastructure & Sustainable Development (TISDIC 2019). Construction Publishing House. ISBN: 978-604-82-2893-4.

The value Zij is used to detect the location of In order to assess the accuracy of the
cracks in the plate structure. In the MSE method, an Detect Capability Index (DCI) is
formula, there are second derivative values of proposed as follows:
the modeshape according to variables x and y. L pred
DCI  100% (8)
The center differential method (CDM) is used Lreal
to calculate these values (Hu et al. 2011). Where Lpred is the crack’s lengthdetermined
2.2 Localmodal strain energy method from the damage index chart; Lreal is the actual
crack’s length. Based on the DCI index, the
After deploying the global MSE method on the results are evaluated as follows:
whole plate structure with low resolution mesh,  If 0  DCI < 50%: detect capability is
the preliminary damaged area is determined. bad (B).
Next, the MSE method is applied separately to  If 50%  DCI  80%: detect capability is
each of localized failure zones (Figure 2) with a moderate (M).
higher resolution mesh to determine in more  If 80% < DCI  100%: detect capability is
detail the location and the severity of the cracks. good (G).
It should be noted that this is a reverse
y
diagnostic problem, without knowing the
damage information. The input data of the
(Ny+1,Nx+1)
(Ny+1,1)
problem is the modeshape of plate in the two
state: before and after the occurrence of cracks.
Local damage area with (Nx+1) Then, through the calculation steps of the SEM
nodes according to x-direction and
(Ny+1) nodes according to
(global and local), the diagnosis results are the
y-direction location and length of the cracks.
(i,1)

x 3. NUMERICAL VERIFICATION
(1,1) (1,j) (1,Nx+1)

Figure 2. Scheme of local damage area 3.1 Properties of plate

An aluminumplate is 3602402 mm in size,
The second derivative of the modeshape is
four edges are clamped; there are two cracks
also determined by the center differential
with the length of 40mm, the width of 1 mm
method (CDM), as follows:
and the depth of 1mm; one crack is in the
For the nodes Nij in x-direction, where i = 2,
middle and another is at the edge of the plate
3, ..., Ny; j = 2, 3, ..., Nx
k ,i j 1  2k ,ij  k ,i j 1
(Figure 3). Material properties of the plate: mass
 2k ,ij
2
 (6) per unit volume  = 2735 kg/m3, modulus of
 x ( sx )2
elasticity E = 70GPa, Poisson’s ratio  = 0.33
For the nodes Nij in y-direction, wherei = 2,
3, ..., Ny; j = 2, 3, ..., Nx Fixed
10

2
 k ,ij k , i 1 j  2k ,ij  k , i 1 j
 (7)
2 y ( sy )2
110

Surface cracks
where, sx and sy are grid space in x-direction with size of
40x1x1mm
and y-direction, respectively.
240

At the boundary nodes of the local damaged

zone, the second derivative of the mode shape
120

is calculated from the displacement data at the

true boundary points of the damaged zone. 160 40 160

This is different from the global MSE method, 360

which requires displacement data at virtual Figure 3. Rectangular plate with two cracks: in the
marginal points (Loendersloot et al. 2017). boundary and in the middle of plate

474
The 3rd Int. Conf. on Transport Infrastructure & Sustainable Development (TISDIC 2019). Construction Publishing House. ISBN: 978-604-82-2893-4.

The MSE method is deployed in two steps. Table 1. Vibration frequencies for first four
In the first step, the MSEis applied on whole modes.
plate with 10-mm-element mesh todetermine
the preliminary damaged area. In the next Frequency Frequency
Difference
step, MSE is applied on the area determined Mode before after damage
(%)
in first step with a 1-mm- element mesh to damage (Hz) (Hz)
determine the position and length of the
cracks in more detail. 1 222.82 220.12 1.21

3.2 Modal analysis 2 342.43 342.34 0.03

ANSYS software is used to simulate and 3 547.01 542.9 0.75

analyze free vibration for tested plate structure.
Figure 4 shows the first four modeshapes for 4 547.13 546.61 0.10
the undamaged state and damaged state. The
vibration frequency results are summarized in
Table 1. When there are damages, the mode 4. RESULTS AND DISCUSSIONS
shape almost does not change. Meanwhile, the
free vibration frequency decreases lightly. 4.1 Gloal detection
Figure 5 shows the damage index chart Zij
corresponding to the damage threshold
30% Z ijmax , where Z ijmax is the maximum damage
index. In this study, three types of combination
of modeshapesare examined: using only mode
shape; combination of two modeshapes and
(a) Mode 1 (b) Mode 1
combination of three mode shapes. The global
diagnostic results show that two region with
occurrence of crackare accuratelyzoned when
using the appropriate damage threshold and
mode shapes.

4.2 Local detection

(a) Mode 2 (b) Mode 2
After implementingthe global MSE method
on the whole pate, the preliminary failure
zones of the boundary crack and the middle
crack are identified as two rectangles with a
size of 60 × 20mm (Figure 6). Next, the
local MSE method is deployed on these
(a) Mode 3 (b) Mode 3
zones with the 1-mm element grid in order to
investigate the effectivenessof the method in
determining the location and length of
cracks. The analytical results are shown in
Figure 7, Figure 8 and summarized in Table
2, Table 3. It is clear from the results that the
(a) Mode 4 (b) Mode 4 MSE method determines accurately the
length of the crack when using the
Figure 4. First four modeshapes: (a) the undamaged
state, (b) the damaged state appropriate damage threshold.

475
The 3rd Int. Conf. on Transport Infrastructure & Sustainable Development (TISDIC 2019). Construction Publishing House. ISBN: 978-604-82-2893-4.

Table 2. DCI for the boundary crack

Threshold %Zijmax Mode 1 Mode 2 Mode 3 Mode 4 Mode 1+2 Mode 1+2+3
10% 98 (G) 98 (G) 98 (G) Noise 98 (G) 98 (G)
20% 93 (G) 97 (G) 93 (G) Noise 93 (G) 93 (G)
30% 85 (G) 95 (G) 85 (G) Noise 85 (G) 83 (G)
40% 70 (M) 93 (G) 78 (M) Noise 70 (M) 70 (M)
50% 90 (G) 95 (G) 85 (G) Noise 95 (G) 95 (G)

Table 3. DCI for the middle crack

Threshold %Zijmax Mode 1 Mode 2 Mode 3 Mode 4 Mode 1, 4 Mode 1, 2, 4 Mode 1, 2, 3
10% 93 (G) 87 (G) 58 (M) 95 (G) 95 (G) 94 (G) 90 (G)
20% 73 (M) 62 (M) 43 (B) 77 (M) 73 (M) 73 (M) 73 (M)
30% 95 (G) 95 (G) 48 (B) 95 (G) 95 (G) 98 (G) 95 (G)
40% 90 (G) 80 (G) 43 (B) 95 (G) 90 (G) 90 (G) 88 (G)
50% 85 (G) 75 (M) 38 (B) 90 (G) 90 (G) 85 (G) 85 (G)

(a) Mode 1 (b) Mode 1,2

(c) Mode 1,2,3

Figure 5. Damage index charts, threshold 30% Zijmax , 10-mm mesh.

476
The 3rd Int. Conf. on Transport Infrastructure & Sustainable Development (TISDIC 2019). Construction Publishing House. ISBN: 978-604-82-2893-4.

Fixed

10
Local damage

110
Surface cracks area detected
with size of preliminarily with
40x1x1mm size 60x20mm

240
120
(a) Mode 1

160 40 160
360

Figure 6. Preliminary damagedarea detected from

global diagnosis, 10-mm mesh

(b) Mode 1,2

(a) Mode 1

(c) Mode 1,2,3

Figure 8. Damage index charts, threshold

30%Z ijmax , 10-mm mesh, middle crack

5. CONCLUSIONS
In this study, the MSE method has been
(b) Mode 1,2 improved successfully fordetectingglobal and
local damagesin like-plate structures. From the
analysis results, the important conclusions
were drawn as follows:
(1) The proposed method has the ability to
accurately detect cracks in like-plate structures.
The global step locates the preliminary
damaged area. The local step identifies the
exact length of the cracks.
(2) In the local step, the plate’s boundary
(c) Mode 1,2,3 condition and the modal data at the nodes
outside the damageareadon’t need to consider.
Figure 7. Damage index charts, threshold 30%Zijmax Thus, the number of data decreases
10-mm mesh, boundary crack significantly and the analysis time is shortened.

477
The 3rd Int. Conf. on Transport Infrastructure & Sustainable Development (TISDIC 2019). Construction Publishing House. ISBN: 978-604-82-2893-4.

(3) The using of the first mode with a Hu, H.W., Wu, C.B., and Lu, W.J. 2011. Damage
damage threshold equal to 10% of the detection of circular hollow cylinder using
maximum damage index is good enough to modal strain energy and scanning damage
accurately detect the location and the length of index methods. Computers & Structures, 89,
the cracks in like-plate structures, for both 149-160.
Kim, J.T., Ryu, Y.S., Cho, H.M. and Stubbs, N.
crack at border and in the middle of the plate.
2003. Damage identification in beam-type
structures: frequency-based method vs mode
5. ACKNOWLEDGEMENT shape-based method. Engineering Structures,
25(1), 57-67.
This research is funded by Vietnam National
Le, T.C. and Ho, D.D. 2015. Damage detection in
University Ho Chi Minh City under grant plate-like structures using modal strain
number C2018-20-26. energy-based method Journal of Construction,
6, 100-105.
6. REFERENCES Loendersloot, R., Ooijevaar, T.H. and Warnet, L.
2017. Vibration based structural health
Cornwell, P.J., Dobeling, S.W. and Farrar, C.R.
monitoring of a composite plate with
1999. Application of the strain energy damage
stiffeners. International Conference on Noise
detection method to plate-like structures.
and Vibration Engineering, Leuven, Belgium:
Journal of Sound and Vibration, 224, 359-374.
ISMA, 2007-2015.
Fan, W. and Qiao, P.2011. Vibration-based
Stubbs, N., Kim, J.T. and Farrar, C.R.1995. Field
damage identification methods: A review and
verification of a nondestructive damage
comparative study. Structural Health
localization and sensitivity estimator
Monitoring, 10, 83-111.
algorithm. Proceedings of the 13th
Fu, Y.Z., Liu, J.K., Wei, Z.T. and Lu, Z.R. 2016. A
International Modal Analysis Conference,
two-step method for damage identification in
210-218.
plates. Journal of Vibration and Control,
22(13), 3018-3031.
Hu, H.W. and Wu, C.B. 2008. Nondestructive
damage detection of two dimensional plate
structures using modal strain energy method.
Journal of Mechanics, 24, 319-332.

478
The 3rd Int. Conf. on Transport Infrastructure & Sustainable Development (TISDIC 2019). Construction Publishing House. ISBN: 978-604-82-2893-4.

A study on behaviors of mass concrete due to heat of

hydration and its pipe cooling system

Vo Duy Hung1,*, Nguyen Van My1, Vo Van Viet2

1
Faculty of Road and Bridge Engineering, University of Science and Technology - The University of
Danang, Danang city, Vietnam;
2
Department of Planning and Investment, Hue City, Vietnam

Corresponding author, E-mail: [email protected]

ABSTRACT: Heat of hydration can cause the early cracks in mass concrete structure. The effects of
hydration heat has been studied so far. In Vietnam, many of large projects are being built, including long
span bridges with very high and large tower and huge foundation. Therefore, the study of the behavior of
concrete due to hydrothermal heat in massive concrete is particularly important. In this paper, The FEM
simulation will be carried out. The heat of hydration in massive concrete and its cooling system will be
discussed in detail. The effects of hydration heat can cause the potential of cracks in tower footing. In
addition the temperature behavior in foundation block from casting time to more than 1000 hours will be
clarified. Furthermore, the stress due to hydration heat in bridge foundation will be elucidated and
discussed. From the FEM analysis of concrete behavior and its pipe cooling system, the recommendations
for hydration heat effects regulation will be given.
Keywords: Heat of hydration, massive concrete, tensile stress, hydration temperature, pipe cooling system.

1. INTRODUCTION the exterior concrete tends to shrink and resist

The proper design and construction of mass interior concrete toexpend, thus causing
concrete can help prevent disasters such as thermal stress and surface crack. The high
crack or over dispacement. In Vietnam, many thermal gradient between the center and the
concrete dam and long span bridge projects are surface may cause thermal cracks when the
actively planned or are in progress. To thermal stress in concrete exceeds its tensile
ensure the safety and durability, mass concrete strength. Therefore, one of the challenges in
maintenances have great concern from mass concrete design and construction is the
designers and operators. The main difference maximum thickness of the concrete, which
between mass concrete and other typical types should be designed without creating thermally-
is its heat of hydration behavior. Depending on induced cracks as well as the time between the
concrete block thickness, the temperature placements of the next layer. The control of
generated from the hydration of cement in heat generated from cement evolution is one of
mass concrete can reach a true adiabatic the most effective ways to reduce the
condition in the interior if large blocksprevent temperature rise in mass concrete; methods
timely heat dissipation to thesurroundings. At a include the use of low-heat-generating cement
high temperature, the interior concrete tends to and the control of the amount of cement
expend, while at a low ambient temperature, content by using the large size of aggregate or

479
using cooling system. Many thermal stress A is the heat conductive area, perpendicular
analyses in mass concrete have been performed to the thermal conductivity direction;
by the finite element method (FEM). There are T
is the thermal gradient.
various approaches on how to use this method. x
In this paper, FEM software Midas 2011-V2.1
is used. The software is a multipurpose finite 2.2.2 Heat Convection
element analysis package that can perform 3-D The amount of heat transmitted by convection
analysis. These components can accurately on a solid block area unit is calculated as
simulate the behavior of different kind of follows:
structures, from small structures to large and
q= hc (T-T∞) (2)
complex models (Kim 2005). The main
objectives of this paper is to clarify concrete where hc is the heat transfer coefficient
behavior due to heat hydration and its cooling between solids and liquids or air. hc depends on
system. Single coolling systems, double many factors such as the geometry of the
coolling systems and four coolling systems will surface, the physical properties of the liquid /
be taken into account. air, the average temperature of the liquid
surface. When calculating the concrete blocks
2. HYDRATION HEAT ANALYSIS in the air, hc can be calculated as follows:
hc = 5.2 + 3.5v (v is wind speed; m/s);T is the
Hydration heat analysis includes heat transfer solids surface temperature;T∞ is the liquid / air
analysis and thermal stress analysis. Heat ambient temperature.
transfer analysis is the process of calculating The amount of heat inside the concrete
temperature changes over time related to heat arises due to hydration in a unit of time and
generation, convection, thermal conductivity volume is:
that occur during cement hydration. Thermal
stress analysis provides stress calculations for 1
g rcK ae  at / 24 (3)
mass concrete according to each construction 24
phase based on changes in temperature where
distribution over time obtained from heat K is the maximum temperature rise (oC);
transfer analysis. Thermal stress analysis also a is the reaction rate coefficient;
considers changes in material properties as r is the volume of concrete (kg / m3);
well as shrinkage and creep. Heat transfer t is time (day).
analysis consists of two main parts: thermal
conductivity analysis and thermal convection 2.2.3 Pipe cooling system
analysis (Gebhart 1993). Cooling with water pipe (pipe cooling) is a
2.1 Heat transfer analysis method to reduce the temperature in concrete
by hydration by using cool liquid pipes into
2.1.1 Heat conduction concrete. The essence of cooling here is the
Heat conduction is a form of heat transfer thermal convection between the liquid and the
associated with the exchange of energy from a and concrete. The amount of heat exchanged
high temperature region to a low temperature can be calculated as follows:
area. According to Fourier's law: Qconv = hpAs(Ts-Tm) (4)
T where
Q x   kA (1)
x Hp is the heat transfer coefficient of the liquid;
where Qx is the amount of heat transferred, the As is the area of the surface of the cooling
unit is kcal / h.m.oC; pipe;
k is the thermal conductivity coefficient, for Ts is the pipe surface temperature;
saturated concrete, k = 1.21-3.11; Tm is the temperature of the liquid cooled.

480
2.2 Thermal stress analysis
Stress in mass concrete at each construction
stage is calculated using the results of heat
transfer analysis, node temperature distribution,
as well as consideration of changes in material Figure 2. Location of some typical nodes
properties by time. These calculations relate to
some concepts such as the equivalent age of Table 1. Physical characteristic of concrete
concrete by temperature and time and
cumulative temperature. Physical characteristics Value
The compressive strength of concrete is
calculated according to the equivalent age and Concrete type C3000
cumulative temperature, calculated by the
Specific heat capacity (kcal.g / kg C) 0,25
following formula:
t Specific weight kg/m3 2400
f c' (t )  f c' (28) (5)
a  bt eq
Heat convection coefficient 12
where 2
(Steel form) (kcal/m .C)
f’c (t) is the concrete strength at the time of
calculation; Heat convection coefficient 15
f’c (28) is concrete strength at age 28 days; 2
(air) (kcal/m .C)
a, b are coefficients depending on the type
of cement; Ambient temperature (C) 25
t is the calculation time (day);
teq is the equivalent age of concrete. Temperature at concrete pouring (C) 20

3. EFFECT OF PIPE COOLING SYSTEM Concrete strength 28 days (kgf/m2) 3.e6

IN REDUCING HYDRATION HEAT
Elastic modulus of concrete (kgf/m2) 2,8.e9
3.1 Modelling of concrete block
Coefficient of thermal expansion 1,0.e-5
To figure out the characteristic of hydration
heat, the pier footing block with dimmension Poisson coefficient 0,18
44m34m5m will be simulated by Midas
Civil software (MIDAS Information Technology Cement content kg/m3 463
(2004). The detail of physical characteristic of
concrete and ambient environment can be seen 3.2 Hydration heat analysis
Table 1. Figures 1 and 2 illustrate the
modelling in FEM software. The characteristic 3.2.1 Hydration Temperature
also take references from ACI (2005) and
Temperature in concrete block is a vital
ASTM C 150 (2002).
characteristic of hydrothermal analysis. Due
to the hydration process, the temperature in
the concrete in the early stages increases and
then gradually decreases with time (Figure 3, 4
and 5). The highest temperature appeared at 80
hours after pouring with 71.14°C (Node 1265
center of concrete block). The high temperature
region is concentrated in the middle and
Figure 1. Modeling of foundation blocks for analysis gradually descends around pier footing.

481
 The temperature distribution at different
times shows that the initial time of the
temperature is quite evenly distributed inside
of concrete block. Then, temperature at
centerin creases. This can be explained as
follows: at the first time, concrete mortar is
still in liquid form. This make the convection
and heat transfer easily, so the temperature is
evenly distributed. But later, when the concrete
began to solidify, the heat generated from the Figure 3. Temperature at 10 hours
hydration reaction was accumulated in the
block and greatly increased around the center
of pier footing block.
Figure 6 illustrates the temperature
changingof 10 nodes. The highest temperature
is concentrated at the center of the concrete
block (Node 1261 and Node 1265). Node 2093
is located at the open surface, the temperature
increases rapidly in the period of 20 hours after
pouring concrete and begins to gradually
decrease to ambient temperature. Figure 4. Temperature at 80 hours
At the boundary, especially the free
boundary (the transient surface of the concrete
block) the temperature has a little bit increased
and quickly decreases the temperature to the
ambient temperature. However, the
temperature difference between the point on
the free boundary and the center of mass is
very large. At the age of 80 hours: the center
temperature is 71.14C while the temperature
at the free side is only 33.0C, the temperature
difference ΔT = 38.14C (Figure 6). Figure 5. Temperature at 120 hours

Figure 6. Changing Temperature of several nodes

482
Figure 7. Stress changing of 10 nodes in concrete block

3.2.2 Hydration Stress temperature between the center and the surface
of the concrete block. If the tensile stress value
From the analytical results, we find that the exceeds the tensile strength of concrete, the
stress in the concrete block at concrete pouring concrete will be cracked.
is mainly the compressive stress, while the From the stress diagram (Figure 7), it is
tensile stress only appears mainly in the open realize that after pouring concrete for 10 hours,
face and around the edges of the formwork. the stress at the surface starts exceeding the
This can be explained by the fact that this allowable stress, leading to the appearance of
period is the increase of temperature inside of
cracks. This period lasts until about 820 hours.
concrete block, then the inner concrete tends to
The highest stress of 7.45MPa coincides with the
heat up, so the stress generated mainly in the
time when the temperature in the concrete block
block is the compressive stress. Whereas
reaches the highest temperature (80 hours).
temperature on the open-surface and the corner
Furthemore, the maximum tensile stress
of concrete block was decreased rapidly (air
occurs at the concrete surface because the
and formwork). This make the concrete tend to
shrink, but this shrink will be kept by higher temperature difference at this position is the
temperatures of the inner layers. This is the largest. Therefore, the risk of surface cracks is
main cause of the tensile stress appearance. highest. This proves that the temperature
When the tensile stress exceeds the tensile change in concrete blocks greatly affects the
limit of concrete, the concrete will be cracked. stress of concrete blocks.
The development of the tensile strength are
shown in Figure 7 and 8. Through these
diagrams, it was found that in the temperature
increase phase of the concrete block (60-80
hours after concreting) the compressive stress
at the center increased gradually. However,
during the decrease phase (after 90-100 hours)
the compressive stress reduce due to the
balance with the tensile stress in the center of
block. The value of tensile stress increases
gradually depending on the difference of Figure 8. Tensile stress at 80 hours

483
3.3 Effect of pile cooling system on reducing Figure 12 shows the maximum temperature
hydration heat in the structure in the cases with and without the
use of a cooling system. When using 04 cooling
3.3.1 Heat reduction
systems, the maximum temperature decreases
According to above discussion, the hydration by about 16,93C after 80h. However, the
heat inside concrete block can cause the cracks highest temperature is still high. Furthermore,
on the surface. In this section, the pipe cooling Figures 13, 14, 15 show that the inner
system will be simulate and check the temperature for each case. Obviously, the case
effectiveness in reducing the hydration heat. of two and four pipe cooling system make
Single cooling system, double cooling system
temperature reduce a lot. The detail comparison
and four pipe cooling system will be simulated
and discussed. The pipe cooling system use the of temperature changing can be seen in Figure
water with ice with specific heat 16. This finding is coincided with Ballim and
100kcal.g/kN.C. The inlet temperature is 15C Graham(2004) and Akin (1994).
with flow rate 1,3m3/h. The inflow time start
from 0h to 60h. The diameter of pipe is 0,036m
and convection coefficient of pipe is 319,55.
The detail of simulation can see from Figure 9,
10, and 11.

Figure 12. Maximumtemperature when using pipe

cooling systems
Figure 9. Single pipe cooling system

Figure 13. The temperature distribution at 80 hours

for single pipe cooling system
Figure 10. Double pipe cooling systems

Figure 14. The temperature distribution at 80 hours

Figure 11. Four pipe cooling systems for double pipe cooling systems

484
Figure 15. The temperature distribution at 80 hours
for four pipe cooling systems
Figure 18. Maximum tensile stress when using
pipe- cooling system

4. CONCLUSIONS
The main objectives of this paper are to give the
general understanding about hydration heat in
mass concrete and its control methods by pipe
cooling system. It is clear that heat of hydration
is a dangerous issue which should be highly
paid attention during construction. In addition,
Figure 16. Temperature changing at Node 1394 the pipe cooling system can reduce the heat of
3.3.2 Tensile stress reduction hydration effectively. The two pipe cooling
systems and four pipe cooling systems shown
Figure 17 and 18 shows the maximum stress in high effectiveness in reducing the temperature
the concrete block in the cases with and due to hydration. The present study allows the
without cooling system. The maximum tensile following conclusions to be drawn:
stress for each case are 7,44MPa for normal
- The heat of hydration in mass concrete is
concrete block, 6,522MPa for single cooling
very high which can cause the crack in concrete.
system, 5,812MPa for double cooling system
and 4,794MPa for four pipe cooling system. - The pipe cooling system can be effective
When using 04 cooling systems, the maximum in mitigating the heat of hydration. The case of
stress decreases by about 2,65MPa after 80h two and four systems shown high effect.
and the tensile stress did not exceed 3,5MPa. - The tensile stress and temperature inside
However, the heat of hydration is still high. concrete are still high though we used pipe
Therefore, it should be use the pipe cooling cooling system, so it should be used in
system with other countermeasure to cope with combiation with the other countermeasures.
heat of hydration in mass concrete.
5. REFERENCES
ACI2005. Building Code Requirements for Structural
(ACI 318-05) Concrete and Commentary (ACI
318R-05). ACI Committee 318.
Akin, J.E. 1994. Finite Element for Analysis and
Design. Academic Press.
ASTM C 1502002. Standard specification for
Portland cement. Annual Book of ASTM
Figure 17. Maximum tensile stress for each case Standards.

485
Ballim, Y., Graham, P.C.2004. Early-age heat Gebhart, B.1993. Heat Condtion and Mass
evolution of clinker cements in relation to Diffusion. McGraw-Hill.
microstructure and composition: implications Kim,J.Y.2005. Heat Transfer Analysis 4th Edition.
for temperature development in large concrete Tae SungSofteware& Engineering.
elements. Cement and Concrete Composition, MIDAS Information Technology2004. Heat of
26, 417-426. Hydration- Analysis Analysis Manual Version
7.0.1.

486
The 3rd Int. Conf. on Transport Infrastructure & Sustainable Development (TISDIC 2019). Construction Publishing House. ISBN: 978-604-82-2893-4.

Invited lecture 1: Weathering performance of new

corrosion resistance steel at coastal area

Kiyonobu SUGAE1,, Hiroshi IKI2, Hiroyuki OKADA3, Tomoya TOMINAGA4,

Nguyen Thi Tuyet Trinh5
1
Steel Research Laboratories, Research & Development, Nippon Steel Corporation, Amagasaki, Japan
2
Plate Products Technical Service & Solution Dept.,Nippon Steel Corporation, Tokyo, Japan
3
Overseas Construction Products Engineering Dept., Nippon Steel Corporation, Tokyo, Japan
4
Nippon Steel Vietnam Co., LTd., Ho Chi Minh City, Vietnam
5
International Education Faculty, University of Transport and Communications, Hanoi, Vietnam

Corresponding author, E-mail: [email protected]

ABSTRACT: In order to reduce the life cycle cost of infrastructures, a new corrosion resistance steel for
painted steel structures has been developed. The corrosion behavior was investigated by both accelerated
corrosion test and exposure test in Japan. The developed steel displayed superior corrosion resistance
even underhigh saline conditions such as those at coastal areas. From the comparison of corrosion rate
and ISO corrosion category between Vietnam and Japan based on previous studies, it has been found that
this newly developed steel can show its superior corrosion resistance under the climate conditions in
Vietnam. This steel can be expected to prolong the lifespan of steel structures and reduce thelife cycle
cost (LCC) of infrastructures in Vietnam.
Keywords: Atmospheric corrosion, infrastructure, life cycle cost, maintenance.

1. INTRODUCTION required. Since the cost of maintenance lies

heavily on the budget of government or
For the continuouseconomic growth of
companies, novel methods and materials are
countries, construction of infrastructures
needed to reduce the life cycle cost (LCC) of
containing railways, highways, industrial
steel structures.
equipment, and port facilitations is
Weathering steel containing Cr, Cu, and Ni
indispensable. In Japan, a large number of
are used without coating in many steel bridges
infrastructures were constructed in the 1960s to
1970s, and these constructionshave now started since this type of steel showssuperior corrosion
todeteriorate. Steel is used for many types of resistance due to the formation of a protective
infrastructure because it is lightweight and rust layer. It is expected that weathering steel
easier to use for construction as compared to can reduce the LCC by omitting the coating
concrete. However, it is well known that the process and repainting maintenance. However,
steel structures under saline conditions such as weathering steel often shows an exceptional
those at coastal areasface corrosion and large corrosion loss under high saline
deteriorationdue to atmospheric conditions. In conditions (Japan Road Association, 1993). In
order to use the steel structures safely and for a fact, the use of weathering steel in Japan is
longer period, periodic heavy maintenance is limited to atmospheric conditions containing

489
small amounts of airborne sea salt particles less Therefore, the developed steel shows a superior
than 0.05 mg/dm2/day. Therefore, corrosion corrosion resistance around paint defects and
resistance steel such as weathering can be expected to reduce the LCC of painted
steelcannotexhibit its superior corrosion structures at coastal areas.
resistance under high saline conditions because This paper presents the results of accelerated
the protective rust layer cannot be formed. corrosion test in a laboratory and the results of
Hence, most steel structures under high saline long-term exposure test of the developed steel
conditions are coated on their surface. at costal area of Japan. Additionally, an
Coating is the most popular protection application example of the developed steel to
technique to shield steel structures from the steel girders of highways and continuous
corrosion. However, defects of paint film are unloaders of port facilitation in Japan is
often generated not only at the edge or angular introduced. Finally, we propose an application
parts but also at general flat parts of the steel of the developed steel for the steel structures at
plate even if the steel is coated appropriately. coastal areas in Vietnam.
Corrosion occurs at these paint defects and
causes delamination of the paint film 2. EXPERIMENTAL
delamination, and this becomes not only an
appearance problem but also a fatal 2.1 Test Specimens
architectonic one. Because of these reasons, Welding structural rolled steels with strength
novel materials showing superior corrosion of 490 MPa were produced by melting them in
resistance at the paint defects are desired. a furnace. The chemical compositions of these
The corrosion mechanism at paint defects types of steel are shown in Table 1. Test
has been reported (Kamimura et al., 2013). specimens were machined from the hot-rolled
From the mechanism, dissolution of metal steel plate, and their surface was treated by
(anodic reaction) proceeds inside the paint shot blasting according to ISO Sa 2 1/2. The
defect, while cathodic reaction proceeds under surface was coated by paint systems shown in
the paint film around the paint defect Table 2. The scribe reaching the steel surface
(Kamimura et al., 2013). The environment of was scratched to paint film by a plastic cutter.
the paint defect is changed to an acidic solution
containing a high concentration of chloride ion Table 1. Chemical compositions of steels
(Cl−) due to hydrolysis of Fe3+ ion and
electrophoresis of Cl−(Kamimura et al.,
2012;Kamimura et al., 2013; Kamimura, T., T.
Nishio, M., Maeda, T., Yoshida, N., Kashima,
K., Sugae, K., Miyuki, H., and Kudo, T. 2013).
A new corrosion resistance steel for painted
steel structures (Corrosion Resistance Steel for Table 2. Paint systems
rePaint Cycle Extension, CORSPACETM) has
been developed based on the corrosion
mechanism (Kamimura et al., 2012; Kamimura
et al., 2013; Kamimura, T., T. Nishio, M.,
Maeda, T., Yoshida, N., Kashima, K., Sugae,
K., Miyuki, H., and Kudo, T. 2013; Sugae et
al., 2015; Sugae et al., 2018). The corrosion
resistance of developed steel at coastal areas in
Japan has been clarified (Sugae. K.,Kamimura,
T., Kodama, M., &Iki, H., 2018; Aoki et al.,
2018; Sugae et al., 2017; Trinh et al., 2018).

490
2.2 Accelerated Corrosion Test 2.4 Evaluation of corroded samples
SAE J 2334 cycle test was used to verify the Evaluation of corroded samples including
atmospheric corrosion behavior of steels. SAE measuring the delamination area and corrosion
J 2334 cycle test is one of the accelerated depth was according to the method reported by
corrosion testsfor automobiles induced under previous study (Kamimura, T., Kashima,
the environment containing deicing salt (SAE K. Sugae, K., Miyuki, H. Kudo, & T, 2013).
J2334, 1998; Townsend et al., 1999). It is
3. RESULTS AND DISCUSSION
reported that the corrosion behavior of steel
corroded by SAE J 2334 cycle test is similar 3.1 Result of accelerated corrosion test
to the corrosion behavior of steel exposed to
Figure 2 shows the changes of paint
high salinity atmospheric conditions such as delamination area in SAE J2334 test cycles.
those at coastal areas (Nagano et al., 2004). Figure3respectively shows the corrosion depth
Figure 1 shows test cycleand condition of at the paint defect (Sugae. K.,Kamimura, T.,
SAE J2334. Kodama, M., & Iki, H., 2018). The paint
delamination was reduced depending on the
thickness of zinc-rich primer. If the primer
thickness is thinner, delamination occurs
easier, and faster cycles are noticed. These
results indicate that the zinc-rich primer
showed a superior corrosion resistance at the
paint defect due to the sacrificial protection of
zinc dust. However, corrosion depth increased
at the paint defect of thicker primer (coating C)
compared to its delamination progress. The
developed steel reduced the paint delamination
Figure 1. Test cycle and condition of SAE J2334. and corrosion depth at the paint defect of every
coating systems compared to the conventional
2.3 Exposure tests steel. From these results, it is considered that
the developed steel sufficiently shows a
The exposure test sites and their corrosive superior corrosion resistance even if it is
environmental conditions are shown in Table 3. applied to the painted steel structure exposed to
Hasaki in Ibaraki prefecture faces the Pacific the high saline condition such as coastal area.
Ocean, and its corrosion category of ISO
6000
12944 is C4 (Tame, S et al., 2013). Moreover, Conventional steel
Paint system A
Okinawa is located in southwest Japan. Its 5000 Developed steel
Delamination area (mm2)

Conventional steel
climate condition categorized in subtropics and 4000 Developed steel
Paint system B

corrosive category of ISO 12944(Tame, S et Conventional steel

Developed steel
Paint system C
3000
al., 2013) is C5-M.
2000
Table 3. Environmental condition of exposure
test site 1000

0
0 100 200 300 400 500
Number of corrosion test cycles

Figure 2. Changes of delamination area in test

cycles. (Sugae. K.,Kamimura, T.,
Kodama, M., &Iki, H., 2018)

491
 1600
Conventional steel Developed steel
and paint delamination did not occur at the
1400
300cyc. paint defect for 7 years. Meanwhile, corrosion
1200 and paint delamination occurred for the coating

Corrosion depth (m)
120cyc.
1000 system without zinc primer (coating A),.The
800 developed steel showed superior corrosion
600 resistance and reduced the paint delamination
400 and corrosion depth.
200 500cyc.
0
Paint system A B C
Primer
thickness 0 μm 15 μm 75 μm

Figure 3. Corrosion depth at paint defect. (Sugae.

K., Kamimura, T., Kodama, M., &Iki,
H., 2018)

3.2 Results of exposure corrosion test

To clarify the corrosion resistance of
developed steel under actual environments,
exposure tests were carried out at different
corrosive environmental test sites in Japan.
Figure 5. Corrosion depth at paint defect after
exposure test at Ibaraki for 7 years.
(Sugae. K.,Kamimura, T., Kodama, M.,
&Iki, H., 2018)

Table 4. Appearance after destructive

inspection of test samples exposed at Ibaraki
for 7 years.

Figure 4. Delamination area of painted specimen

after exposure test at Ibaraki for 7 years.
(Sugae. K.,Kamimura, T., Kodama, M.,
&Iki, H., 2018)

Figures 4 and 5 show the paint delamination

and corrosion depth at the paint defect of test
sample exposed for 7 years at the exposure test Figures 6 and 7 show the paint delamination
site in Ibaraki, respectively (Sugae. and corrosion depth at the paint defect of test
K.,Kamimura, T., Kodama, M., & Iki, H., sample exposed for 3 years at the exposure test
2018). Table 4 shows the appearance of test site in the coastal area of Okinawa, respectively
samples after destructive inspection (Sugae. (Sugae et al., 2017). Table 5 shows the
K.,Kamimura, T., Kodama, M., &Iki, H., appearance of test samples after destructive
2018). At Ibaraki whose corrosion category of inspection (Sugae. K.,Kamimura, T., Kodama,
ISO 12944 is C4, for the coating system M., &Iki, H., 2018). The Okinawa coastal
containing zinc primer (coating A), corrosion areacorrosion category of ISO12944 is C5. The

492
test samples coated with coating A were Table 5. Appearance after destructive inspectionof
exposed because coating C is categorized and test samples exposed to the coastal area of
recommended as the coating system for steel Okinawa for 3 years.
structures in extremely corrosive environments
(C5-M) by ISO12944 (Tame, S et al., 2013). At
the coast of Okinawa, corrosion and paint
delamination occur at the paint defect even in
coating A. The developed steel showedsuperior
corrosion resistance and reduced the paint
delamination and corrosion depth in the
corrosive environment C4 or C5. Therefore, it is
found that the developed steel reduces corrosion
at the paint defect even under highly corrosive
environments such as coastal areas. Hence, the
developed steel can be expected to prolong the
lifespan of steel structures and reduce the LCC
of steel structures even under high saline
conditions such as those at coastal areas.

Figure 6. Delamination area of painted specimen

after exposure test at the sea side of
Okinawa for 3 years (Sugae et al., 2017). 3.3 Application examples of developed steel
for steel structures in Japan

Figure 7. Corrosion depth at paint defect after

exposure test at the coastal side of Figure 8. Application example of developed steel for
Okinawa for 3 years. (Sugae et al., 2017). bridge over the port channel in Okinawa.

493
Figures 8 and 9show examples of developed that most environments in Vietnam are
steel for a bridge and a continuous unloader at categorized under C2 or C3 of ISO 12944.
the coastal area, respectively. The bridge was Miura et al. (2017) have reported the corrosion
constructed over the port channel in behavior of weathering steel in Vietnam.
UrasoeCity of Okinawa in 2017. The Additionally, they have also reported that most
continuous unloader has been operated in corrosive environments in Vietnam are milder
Nagoya since 2016. than thosein Japan (Miura, S. et al., 2017). By
comparison of the corrosion rate shown in
Table 6, it is found that the corrosion rates in
Ibaraki and the coastal area of Okinawa are
higher than thoseat exposure sites in Vietnam.
Hottaet al. (2014) have reported that the
coating system containing zinc-rich primer is
viable in almost all areas in Vietnam.

Table 6. Comparison of corrosive environment

data, corrosion rate of carbon steel, and ISO
category amongtest sites in Vietnam and Japan.

Figure 9. Application example of developed steel

for continuous unloader in Nagoya.

3.4 Corrosive Environment in Vietnam and

Expectation of the Developed Steel
Atmospheric corrosion test results of steel in
Figure 10 shows the conceptual diagram of
Vietnam have been reported by some
the effect of developed steel during in-service
researchers (Hotta, Y. et al., 2014; Lien, L. T.
H. et al., 2007; Miura, S. et al., 2017). The periods. Since the developed steel can reduce
corrosion rate and corrosive environmental corrosion deterioration, the first repairing timing
data reported by these previous studies are can be delayed compared to that of conventional
listed in Table 6. Especially, Lien et al. (2007) steel. It has been reported that the analysis
have reported the corrosion rate of steel and results of the changes of delamination area
corrosive environmental factor, including the occurred under corrosion test conditions by
amount of airborne sea salt particles, using the Gompert function (Ito et al. 2016;
temperature, humidity, and total time of Sugae. K., Kamimura, T., Kodama, M., &Iki,
wetness at the exposure test sites and have H., 2018). Therefore, it is considered that the
suggested the corrosion map of steel in developed steel can prolong the initial in-service
Vietnam. From their study, it has been found period by about 2 times as compared to

494
conventional steel and the lifespan of steel 5. REFERENCES
structures even under severe corrosive condition
Aoki, Y., Takada, Y. Sugae, K., Kamimura, T. Iki,
(Ito et al.,2016, Sugae. K.,Kamimura, T.,
H., Kodama, M., &Nasu, K. 2018.
Kodama, M., &Iki, H., 2018). Hence, the Proceedings of 73th Annual Academic Meeting
developed steel can be expected to show of Japan Civil Engineers, I-040.
superior corrosion resistance and contribute to Hotta, Y., Iwase, Y., &Sadaishi, K. 2014.
LCC reduction of steel structures in Southeast Evaluation of durability of coating film in
Asian countries including Vietnam. tropical region (in Japanese). Proceedings of
34th Academic Meeting of Japan Association
of Corrosion Control, 73-78.
Ito, Y., Hirohata, M., Okumura, T., Mori, T.,
&Kamimura, T. 2016. Corrosion and paint-
coat deterioration characteristics of coated Sn-
bearing steel and its application on steel
bridges. Journal of Japan Society of Civil
Engineers, 72, 407-419.
Japan Road Association: Specifications for
Highway Bridges, part I: General, Part II:
Steel Bridges (XX). 1993.3
Kamimura, T., Kashima, K. Sugae, K., Miyuki, H.
Kudo, & T.2012. The role of chloride ion on
the atmospheric corrosion of steel and
corrosion resistance of Sn-bearing steel.
Corrosion Science, 62, 34-41.
Kamimura, T., Kashima, K. Sugae, K., Miyuki, H.
Figure 10. Conceptual diagram of the effect of
Kudo, & T.2013. Corrosion resistance at
developed steel during in-service
defect of paint film for Sn-bearing steel.
periods.
Journal of the Society of Materials Science,
Japan, 62, 207-212.
4. CONCLUSIONS Kamimura, T., T. Nishio, M., Maeda, T., Yoshida,
In order to reduce the LCC of infrastructures, a N., Kashima, K., Sugae, K., Miyuki, H., &
Kudo, T. 2013. Develpoment of Sn-bearing
new corrosion resistance steel for painted steel
steel exhibiting superior atmospheric
structures containing small amounts of corrosion corrosion resistance at defect of paint film.
resistance elements has been developed based on Zairyo-to-Kankyo, 62, 187-191.
the corrosion mechanism at the paint defect Lien, L. T. H., San, P. T., & Hong, H. L. 2007.
under atmospheric condition. From the results of Results of studying atmospheric corrosion in
corrosion test and exposure test, it has been Vietnam 1995-2005.Science and Technology of
found that the developed steel shows superior Advanced Materials, 8, 522-588.
corrosion resistance even under high saline Nagano, H., Uchida, H., Yamashita, M. 2004.
condition such as those at coastal areas in Japan Environmentally Benign Materials–Corrosion
and is expected to contribute to the LCC & Anticorrosion Engineering for Protection of
reduction of steel structures. It is considered that Global Environment. Kyoritsu Shuppan Co.,
the developed steel also shows its superior Ltd., p74.
corrosion resistance under the climate conditions Miura, S., Murase, M., Okamoto, T. Tung, D.D., &
Iwasaki, E. 2017. Corrosion behavior and
in Vietnam from the comparison of corrosion
applicability of weathering steel in Vietnam.
rate and ISO corrosion category between Journal of Materials in Civil Engineering, 29,
Vietnam and Japan. Hence, the developed steel DOI: 10.1061/(ASCE)MT.1943-
can be expected to prolong the lifespan of steel 5533.0001817.
structures and contribute to the LCC reduction SAE J 2334: Cosmetic Corrosion Lab Test. 1998,
of infrastructures in Vietnam. SAE International, Warrendale, PA.

495
Sugae, K., Kamimura, T., Ando, R., & Tsuzuki, T., Tame, S., Aoki, R., & Tanabe, H. 2013. The trend
2015. Development and partical application of of international standards in corrosion
corrosion resistance steel for painting cycle protective paint systems. Journal of the
extension (CORSPACE TM). Nippon Steel & Surface Finishing Society of Japan, 64, 3-8.
Sumitomo Metal Technical report, 110, 77-83. Townsend, H.E., Simpson, M.W., Linde, W.B., &
Sugae, K., Kamimura, T., Tanaka, M., Iki, H., & McCune, D.C. 1999. License plate cosmetic
Kodama, M. 2017. Proceedings of 72th Annual corrosion test of automotive coated steel
Academic Meeting of Japan Civil Engineers, sheet.Corrosion, 55, 406-411.
I-025. Trinh.N. T. T., Sugae, K. 2018. Exposure test of
Sugae. K.,Kamimura, T., Kodama, M., &Iki, H., high corrosion resistant steel in the coastal
2018. Development of new corrosion area environment, The Transport
resistance steel exhibiting superior corrosion Journal,09/2018, 57-61.
resistance at paint defect.BoseiKanri, 10, 7-11.
Sugae, K., Kamimura, T. Miyuki, H & Kudo, T.
2018. Corrosion behavior of Sn-bearing steel
under wet/dry cyclic environments containing
Cl-. Material transactions, 59, 779-786.

496
The 3rd Int. Conf. on Transport Infrastructure & Sustainable Development (TISDIC 2019). Construction Publishing House. ISBN: 978-604-82-2893-4.

Invited lecture 2: Investigating mechanical characteristic

of ultra-high-performance fiber-reinforced concrete at
different strain rates

Ngo Tri Thuong1,, Tran Ngoc Thanh2, Chau Truong Linh3

1
Department of Transportation Engineering, Faculty of Civil Engineering, Thuyloi University, Viet Nam
2
Department of Structural Engineering, Faculty of Civil Engineering, HCMCUniversity of Transport,
Viet Nam
3
Department of Bridge and Road Engineering, University of Science and Technology, Danang City,
Viet Nam

Corresponding author, E-mail: [email protected]

ABSTRACT: The compressive and tensile strength of ultra-high-performance fiber-reinforced concrete

(UHPFRCs) at different strain rates were investigated. The static test was implemented using the Universal
Test Machine (UTM) while the strain energy frame impact machine (SEFIM) was used for high strain rate
tests. The compressive strength of Ultra-high-performance concrete (UHPC) and UHPFRCs reinforced with
3 vol.-% smooth steel fiber were 186 and 205 MPa, respectively. The tensile strength of UHPFRCs was
17.55 MPa at static rate (0.000667 s-1) increased to 34.04 MPa as the high strain rates up to 136 s-1. All
UHPFRC specimens failed with one major tensile crack accompanied by multiple crack formationeven at
high rate loading and the number of crack increased as the applied loading speed increased.
Keywords: UHPFRCs, tensile resistance, compressive strength, high strain rates, steel smooth fiber.

1. INTRODUCTION enhancement of the resistance of civil

infrastructure, buildings, especially military
Ultra - high - performance fiber - reinforced structures to extreme loads, such as seismic,
concretes (UHPFRCs) have demonstrated impact, and blast loads (Rong and Sun, 2012;
superior mechanical properties, in comparison Pyo and El-Tawil, 2015). For instance, the
with normal concrete (NC) and high- concrete wall in a military structure subjected
performance fiber - reinforced concrete to a single missile showed the shear failure and
(HPFRCs), including very high compressive local spalling on the surfaces (Ngo et al. 2017).
strengths (> 150 MPa), tensile strengths (> 13 However, very limited information is
MPa), tensile strain capacities (> 0.3%), and available regarding the resistance of
energy absorption capacities (> 30 kJ/m2) even UHPFRCs, even at static rate, owing to no
when containing only 1.5% volume content standard test method exists, design guides for
(1.5 vol.%) deformed steel fibers (Wille et al. engineers, and quality control method in
2011), as shown in Fig. 1. The response of manufacturing facilities as well their relatively
structures made by UHPFRCs under high rate high fiber contents and cost.
impacts and/or blasts was higher than that at Tran and Kim (2012) firstly built a strain
low rate owing to rate-sensitive material energy frame impact machine (SEFIM) to
characteristics (Park et al. 2016) as well as investigate the direct tensile stress versus strain
inertial effects. These properties favor the response of high-performance fiber-reinforced

497
cementitious composites (HPFRCCs) at high mmdiameter and length of 19 mm. The 3 vol.-%
strain rates between 20 and 80 s-1. Later, Pyo and used in this study is continued the experimental
El-Tawil (2015) built a modified SEFIM (M- plot had done in Tran and Kim (2015).
SEFIM) while Park et al. (2016) made an The UHPFRCs mixtures were mixed by a
improved SEFIM (I-SEFIM) to investigate the Hobart type laboratory mixer. Silica fume and
tensile response of UHPFRCs at high strain rates silica sand were first dry-mixed about 5 min
(90 to 200 s-1). The tensile strength and energy before cement (Type I) and silica power was
dissipation capacity of UHPFRCs greatly added and further mixed in 5 min. The water and
enhanced at the high strain rates, whereas the super-plasticizer then were gradually poured into
strain capacity was not sensitive to the applied the mixture and continuously mixed until it
strain rates. The post-cracking tensile strength of showed workability adequate for uniform fiber
UHPFRCs at high strain rates was approximately distribution. Finally, the fibers were added into
2.8 to 2.9 times higher than that at static strain rate. the mixture gradually and carefully by hand and
further mixed for approximately 2 min.
The mixture was poured into plastic molds
without vibration. All specimens were stored in
a laboratory at room temperature for 48 h prior
to demolding. The specimens were then cured
at 90 ± 3 ºC for 72 h in a hot water tank before
stopping the curing and storing the samples in
dry conditions. All the specimens were tested
after 28 days. Three thin layers of polyurethane
a) Compressive behavior of UHPFRCs were sprayed on the surfaces of specimens
before testing to precisely investigate the
cracking behavior of specimens with naked
eyes. Detail procedures regard to casting and
curing the UHPFRC specimens could be found
inPark et al. (2008) and Ngo et al. (2017).
Table 1. The composition of UHPC matrix by
weight ratio
Cement Silica Silica Super-
b) Tensile behavior of UHPFRCs Sand Water
(Type I) Fume powder plasticize
Figure 1. Compressive and tensile behavior of 1.00 0.25 1.10 0.30 0.067 0.2
UHPFRCs
2.2 Test setup and procedure
This study aims to develop a fundamental The tensile tests at static rates were performed
understanding of the mechanical characteristics using the Universal Test Machine (UTM) as
of UHPFRCs. The specific objectives are (1) to shown in Fig. 2. The detail of tensile test setup
investigate the compressive strength of refers to Tran et al. (2014). Both ends of the
UHPFRCs with or without fiber reinforcements, tensile specimen are bell-shaped and reinforced
(2) to investigate the tensile resistance of with steel wire meshes to prevent failure
UHPFRCs at static and high strain rates. outside of the gauge lengths, measuring 100
mm in this study, of the specimens. The speed
2. EXPERIMENTAL PROGRAM of loading was 1 mm/min and the data
2.1 Material and specimen preparation acquisition frequency was 1 Hz. Two LVDTs
were attached to the specimen by an aluminum
The composition of UHPC matrix by weight cage to measure the elongation of the specimen
ratio was listed in Table 1. The blended fiber during testing. The load signal was measured
was 3vol.% smooth steel fiber with 0.2 by a load cell at the top of the specimen.

498
 (50  50  50 mm3) were cast and tested. In
addition, the flow tests were used to measure
the effect of vol.% of fiber on the flowability of
mixtures. The flow test was in accordance with
ASTM C1437 (ASTM International, 2001).

Figure 2. Tensile test setup for UHPFRCs at static

rate

Figure 4. Compressive test set up and flow test

apparatus

3. RESULT AND DISSCUSIONS

Figure 3. Tensile test setup for UHPFRCs at high The flowability of mixtures and compressive
strain rate strength of UHPC and UHPFRCs are listed in
Table 2. As shown in this Table, the
A SEFIM was set up for the high-rate tensile compressive strength of UHPC was increased
tests, as shown in Fig. 3. In brief, as the applied by the corporation of steel fiber reinforcements,
load exceeds the capacity of coupler, the coupler whereas the flowability was decreased. The
is suddenly fractured and the elastic strain energy compressive strength and flowability of UHPC
stored within the energy frame will be rapidly and UHPFRC with 3 vol.-% smooth steel fiber
released. A high rate stress wave which is are 186 and 205 MPa, 221 and 208 mm,
generated by the released strain energy, respectively. The increase of compressive
propagates through the hinge grip to the strength of fiber-reinforced UHPC can be
specimen and eventually fails the specimen. The attributed to the interfacial bond stress between
speed of applied load was controlled by the fiber and matrix, which delayed the formation
capacity of coupler and types of energy frame: and propagation of cracks Wu et al. (2016). In
the coupler with 400, 800 N capacities and high addition, the decrease in the flowability of
strength steel energy frame with 30 and 45 mm mixtures has been reported by several previous
in diameter were used in this study. The tensile researchers(Wu et al. 2017; Yu et al. 2014).
stress was obtained from two dynamic strain Table 2. Compressive strength of UHPFRCs
gauges attached on the surfaces of the transmitter
bar, while the tensile strain of the specimen was Item UHPC UHPFRCs
measured from the relative displacement of Compressive strength (MPa) 186 205
marked points on the surface of specimen, as Flow (mm) 221 208
shown in Fig.3. Detail of high rate tensile test
setup could be found elsewhere(Park et al. 2016). Average tensile stress versus strain curves of
The compressive tests according to ASTM UHPFRCs at different strain rates are shown in
C109/C109Mwere implemented to check the Fig. 6 while those detail value was listed in
compressive strength of UHPC and UHPFRCs, Table. 3. The strain rates of group h1 were from
as shown in Fig. 4. Total six cubic specimens 34 to 42 s-1, while those of group h2 was from

499
76 to 99 s-1 and 127 – 134 s-1 for group h3. The The failure mode of UHPFRC specimens
tensile resistance including tensile strength (pc) was captured in Fig. 6. All specimens failed
and strain capacity (pc) of UHPFRCs increased with one major tensile crack accompanied by
as the applied strain rates increased. Thepc multiple crack formation. The number of
increased from 17.55 MPa at static to 27.25 minor cracks increased as the applied strain
MPa at h1, 30.14 MPa at h2, and 34.04 MPa at rate increased.
h3. Moreover, those values of pc are 0.98, 1.30,
2.93, and 2.68%, respectively. The higher 40
response of UHPFRC as the higher applied UHPFRC 3 vol.-% Static
strain rate can be explained by the increase of h1 (40/s)
clamping pressure of matrix surrounding fibers. h2 (89/s)
30

Tensile stress (MPa)

h3 (136/s)
When the mortar matrix surrounding fiber has
an acceleration by high rate loads, additional
pressure is generated on the surface of fiber by
20
inertial effect, in addition to the initial clamping
pressure generated by matrix shrinkage (Stang,
1996; Tran and Kim, 2015; Park et al. 2014).
The additional pressure due to the inertial effect 10
would increase the interfacial bond strength at
high rates.The tensile strength increased as the
fiber volume content increased. The average 0
tensile strength (17.55 MPa) of UHPFRC 0 1 2 3 4 5 6 7
reinforced with 3.0 vol.% in this study is much Tensile strain up to peak stress, 
greater than the 12.1 MPa of those with 1.5 vol.- Figure 5. Average stress-strain curves of UHPFRC
% in the previous study (Tran and Kim, 2015). at different strain rates

Table 3. The composition of UHPC matrix by weight ratio

Spec. Strain rate Tensile strength (pc) Tensile strain capacity (pc)
Type s-1 MPa %
SP1 0.000667 17.32 0.97
SP2 Static 0.000667 18.18 1.01
SP3 0.000667 17.14 0.96
Average 0.000667 17.55 0.98
SP1 42.38 29.57 1.28
High rate
SP2 43.49 24.87 1.30
1(h1)
SP3 34.88 27.30 1.32
Average 40.3 27.25 1.30
SP1 90.31 27.24 3.04
High rate 2
SP2 99.92 31.62 2.87
(h2)
SP3 76.20 31.57 2.87
Average 88.8 30.14 2.93
SP1 127.54 38.20 1.96
High rate 3
SP2 147.23 33.72 3.01
(h3)
SP3 134.31 30.21 3.05
Average 136.4 34.04 2.68

500
Figure 6. Failure mode of UHPFRC specimen at different strain rates

4. CONCLUSIONS 6. REFERENCES
The compressive and tensile resistance of ASTM C 1437 Standard Test Method for Flow of
UHPFRC at different loading speed was Hydraulic Cement Mortar’, pp. 1–2.
investigated. The following observations and ASTM C109M-02 (2007) ‘Standard Test Method
conclusions could be drawn from the for Compressive Strength of Hydraulic
experimental study: Cement Mortars’, Annual Book of ASTM
 The UHPFRCs were shown very high Standards, 04, pp. 1–6.
compressive strength (205 MPa) and high Ngo, T. T. et al. (2017) ‘Shear resistance of ultra-
tensile strength (17. 55 MPa) high-performance fiber-reinforced concrete’,
accompanied by multiple cracking and Construction and Building Materials, 151, pp.
high strain capacity (0.98%). 246–257.
 The compressive strength of UHPFRCs Ngo, T. T. and Kim, D. J. (2017) ‘Shear stress
were increased by the incorporation of versus strain responses of ultra-high-
steel fiber with 3 vol.%. performance fiber-reinforced concretes at high
 The tensile strength of UHPFRCs increased strain rates’, International Journal of Impact
as the fiber volume content increased. Engineering, pp. 1–23.
 As the applied loading speed increased, Park, J. J. et al.(2008) ‘Influence of the ingredients
both the tensile resistance and the number on the compressive strength of UHPC as a
of multiple crack the UHPFRC fundamental study to optimize the mixing
increased.These results signify that proportion’, Proceeding of the second
UHPFRCs is well suited for blast and international symposium on ultra high
impact applications. performance concrete, Kassel, Germany, pp.
105–102.
5. ACKNOWLEDGEMENT
Park, S. H. et al.(2014) ‘Effect of shrinkage
This project was supported by the National reducing agent on pullout resistance of high-
Foundation for Science Technology strength steel fibers embedded in ultra-high-
Development (NAFOSTED) under Grant performance concrete’, Cement and Concrete
number 107.01-2019.03 Composites. Elsevier Ltd, 49, pp. 59–69.

501
Park, S. H., Kim, D. J. and Kim, S. W. (2016) Tran, T. K. and Kim, D. J. (2013) ‘Investigating
‘Investigating the impact resistance of ultra- direct tensile behavior of high performance
high-performance fiber-reinforced concrete fiber reinforced cementitious composites at
using an improved strain energy impact test high strain rates’, Cement and Concrete
machine’, Construction and Building Research, 50, pp. 62–73.
Materials. Elsevier Ltd, 125, pp. 145–159 Wille, K., Kim, D. J. and Naaman, A. E. (2011)
Pyo, S. and El-Tawil, S. (2015) ‘Capturing the ‘Strain-hardening UHP-FRC with low fiber
strain hardening and softening responses of contents’, Materials and Structures, pp. 583–
cementitious composites subjected to impact 598.
loading’, Construction and Building Wu, Z. et al. (2016) ‘Effects of steel fiber content
Materials. Elsevier Ltd, 81, pp. 276–283. and shape on mechanical properties of ultra
Rong, Z. and Sun, W. (2012) ‘Experimental and high performance concrete’, Construction and
numerical investigation on the dynamic tensile Building Materials. Elsevier Ltd, 103, pp. 8–14.
behavior of ultra-high performance cement Wu, Z. et al. (2017) ‘Static and dynamic
based composites’, Construction and Building compressive properties of ultra-high
Materials. Elsevier Ltd, 31, pp. 168–173. performance concrete (UHPC) with hybrid
Stang, H. (1996) ‘Significance of shrinkage- steel fiber reinforcements’, Cement and
induced clamping pressure in fiber-matrix Concrete Composites, 79.
bonding in cementitious composite materials’, Yu, R., Spiesz, P. and Brouwers, H. J. H. (2014)
Advanced Cement Based Materials, 4(3–4), ‘Static properties and impact resistance of a
pp. 106–115. green Ultra-High Performance Hybrid Fibre
Tran, N. T., Tran, T. K. and Kim, D. J. (2015) Reinforced Concrete (UHPHFRC):
‘High rate response of ultra-high-performance Experiments and modeling’, Construction and
fiber-reinforced concretes under direct Building Materials, 68.
tension’, Cement and Concrete Research.
Elsevier Ltd, 69, pp. 72–87.
Tran, T. K. and Kim, D. J. (2012) ‘Strain Energy
Frame Impact Machine (SEFIM)’, Journal of
Advanced Concrete Technology, 10(3), pp.
126–136.

502
The 3rd Int. Conf. on Transport Infrastructure & Sustainable Development (TISDIC 2019). Construction Publishing House. ISBN: 978-604-82-2893-4.

Effect of fly ash and QuartzPowder on the properties of

high-performance concrete

Nguyen Duc Vinh Quang1,*, AleksandrovaOlga Vladimirovna2,

TkachEvgeniyaVladimirovna3
1
Lecturer, Hue Industrial College, Hue City, Vietnam; PhD student, Department of Technologies of
Cohesive Materials and Concretes, Moscow State University of Civil Engineering (National Research
University), Moscow, Russian Federation.
2
Associate professor, Department of Technologies of Cohesive Materials and Concretes, Moscow State
University of Civil Engineering (National Research University), Moscow, Russian Federation.
3
Professor, Doctor of technology science, Department of Building Materials, Moscow State University of
Civil Engineering(National Research University), Moscow, Russian Federation.

Corresponding author, E-mail: [email protected]

ABSTRACT: In this study, influence of fly ash (FA) in 1-5µm and quartz sand powder (QP) size of 5-
10µm on the properties of concrete containing silica fume (SF), coarse aggregate (CA) in 9.5 - 20 mm
and superplasticizer (SP) has been investigated. Concretes included fly ash as a cement replacement level
of 20%, 30% and 40% respectively and a partial replacement of fine aggregate (Fa) by quartz sand
powder. The water to binder ratio and total cementitious materials content were kept constant at 0.3 for all
mixes and 550 kg/m3 respectively. Concrete mixes were evaluated for compressive strength at 3, 7 and
28 days and tensile strength at 28 days. The result experimental indicates that the use of mineral
admixtures included 10% SF, 40% fly ash to partial replacement of sulfate-resisting Portland cement and
partial replacement of fine aggregate by 20% quartz sand powder respectively, improving both long term
durability and strength of HPC. The specimens were cast for finding mechanical properties of HPC was
studied and discussed.
Keywords: Fly ash; silica fume; quartz sand powder; high-performance concrete (HPC); mechanical and
durability properties of concrete.

1. INTRODUCTION ones using pulverized-coal combustion and the

In Vietnam, the industrialization and remainder using circulating fluidizing bed.
modernization of the country, the industry of Vietnam thermal power plants generated up to
Vietnam is being invested, developed strongly. 12.2 million tonnesof coal ash every year, while
Along with the growing production of products, one year treating only 4 million tonnes, meaning
energy, fuels serving the development of the inventory rose to 25.2 million tonnes, according
national economy, the development of many to the Department of Industrial Safety and
industries increases more and more waste. By Environment (DISE) under the Ministry of
which, the coal-fired thermal power industry is Industry and Trade. The amended master plan
one of the major sources of waste. The country for electricity development for the 2011-2020
has 21 operational thermal power plants, with 13 period with a vision to 2030 says that 52 coal-

503
fired power plants will be in operation by 2030 fired plants are polluting the sea and threatening
with a total designed capacity of 52,252 MW. aquaculture and tourism which are the livelihood
To produce 1kWh of electricity using coal dust, of millions of people. The volume of slag and
from 0.9 to 1.5kg of coal ash will be emitted. ash treated to make building materials remains
Thus, coal ash generated annually from the 23 very modest. Meanwhile, it is estimated that if
operational thermal power plants measures the slag and ash can be treated well, Vietnam
about 12.2 million tonnes. The North of the will be able to save tens of millions of minerals,
country is home to 60% of the plants, while the while hundreds of hectares of land will be
central region has 21% and the south 19%. liberalized as coal-fired plants will no longer
However, to date, the ash used as a construction need the area for slash grounds. More
material is very low compared to the amount importantly, Vietnam will be able to settle the
generated, about 10% is used, and 90% is problem of pollution caused by solid waste from
buried. This causes serious water pollution coal-fired power plants. The coal ash wastes
around the site burial ground. Most of this ash is from the coal-fired power plants consist of two
mixed with water and pumped out of the parts, the rough ash of coal ash at the bottom of
dumping site (such as Figure 1a and 1b). This in the kiln, as shown in Figure 1a, accounts for
addition to the environmental impact is a huge about 25% to 30%, and the rest is fly ash (fine-
waste of resources. Apparently, the problem of grained ash, spherical, smooth surface) accounts
ash handling at coal plants in Vietnam is very for about 25% to 30% as Figure1b is 65 ÷ 70%.
urgent, posing challenges for environmental Based on the results of the analysis of the
protection and thermal power plant chemical and physical properties, the results
development. The amount of untreated coal ash showed that the samples of fly ash were taken
is expected to reach 61 million tons by 2018, from the thermal power plants in the North of
109 million tons by 2020, up to 248 million tons Vietnam have the chemical composition almost
by 2025, and 422 million tons by 2030s. equivalent to some countries on the world, as
shown in Table 1.
Currently, Vietnam is facing environmental
problems from the development of coal-fired
power plants. It is up to the year 2030 to manage
the ash dumps with an area of more than 28,000
hectares (assuming a depth of 2m), distributed
along the length of the country through large
power centers from the region North, Central to
the Mekong Delta (articles in recent times). So,
in the next few years, if no solutions are 1a. Rough ash 1b. Fly ash
developed, the amount of ash could increase so Figure 1. Scanning Electron Microscope of Ash
much that there is no room for storage. Coal- coal slag (1000x magnification)

Table 1. Chemical composition of some kinds of fly ash

The result analysis chemical composition of fly ash(content % mass)
The thermal power plants on the
Parameter The thermal power plants in Vietnam
world
Pha Lai Uong Bi MongDuong NinhBinh Europe Chine American
SiO2 57.1-58.4 58.49 54.27 34.41-60.7 28.5-59.7 35.6-57.2 37.8-58.5
Al2O3 23.8-26.1 28.11 25.02 17.39-27.2 12.5-35.6 18.8-55.0 19.1-28.6
Fe2O3 4.7 - 7.2 6.11 4.71 4,8 – 5.61 2.6-21.2 2.3-19.3 6.8-25.5
CaO 0.7 - 0.81 0.82 0.91 0,4 – 1.21 0.5-28.9 1.1-7.0 1.4-22.4

504
 The result analysis chemical composition of fly ash(content % mass)
The thermal power plants on the
Parameter The thermal power plants in Vietnam
world
Pha Lai Uong Bi MongDuong NinhBinh Europe Chine American
MgO 1.16 - 1.2 1.10 1.22 0,8 – 1.11 0.6-3.8 0.7-4.8 0.7-4.8
Na2O 0.09 - 0.4 0.1 0.16 0.16-0.2 0.1-1.9 0.6-1.3 0.3-1.8
K2O 4.3 - 6.56 2.58 6.76 4,3 – 5.16 0.4-4.0 0.8-0.9 0.9-2.6
P 2O 5 0 - 0.13 - 0.16 0.08- 0.16 0.1-1.7 1.1-1.5 0.1-0.3
TiO2 0 - 0.78 - 0.78 0.13- 0.63 0.5-2.6 0.2-0.7 1.1-1.6
MnO 0 - 0.04 - 0.04 0.02- 0.06 0.03-0.2 - -
SO3 0,3 - - 0.3 0.3 1.0-2.9 0.1-2.1
H2O- 0 - 0.35 0.58 8.4- 14.02
Loss of ignition
4.36-16 ≈ 45 5.24 8,1-16.91 - - 0,2-11,0
(H2O+)

Vietnam is a developing country, so besides sand are getting depleted (A. Mardani 2014).
the need to develop energy sources to serve The prohibition on mining in some areas and
the development of society, then the the growing need for natural environment
development of infrastructures such as conservation further exacerbate the problem
bridges, roads, metro, tunnels, high-rise of river sand availability.
buildings ... it also play a key role in The use of FA in concrete can reduce the
economic growth and social development. consumption of natural resources and also
However, Vietnam is one of the most severely diminishes the effect of pollutant in
affected countries of climate change and sea environment. In recent studies, many
level rise, especially in the South of Vietnam researchers found that the use of additional
(Monre 2016). Saline intrusion, combined cementitious materials like FA in concrete is
with sources of waste from factories industrial economical and reliable. Fly-ash particles are
zones, is largely untreated, causing rapid almost totally spherical in shape, allowing
destruction of existing transport and them to flow and blend freely in mixtures. That
infrastructure works and also a challenge for capability is one of the properties making fly-
new construction works. ash a desirable admixture for concrete (T.G.S
Concrete is one of the most important Kiran et al. (2014), Anna et al. 2017, Sanjukta
materials in building construction and other et al. (2015), and Ruben 2012). Fly ash may
infrastructure works. About 2.7 billion m3 of have beneficial effects on both the fresh and
concrete was generated in 2002 worldwide, hardened properties of concrete mixtures, the
which is more than 0.4 m3 of concrete use of FA in concrete has proven to improve
generated per person once a year(T.R. Naik workability and long-term strength, reduce
2008). It is anticipated that the need for permeability, minimize risk of alkali silica
concrete will increase further to almost 7.5 reaction, lowering heat of hydration in mass
billion m3 (about 18 billion tons) a year by concrete, and enhancing durability
2050 (M.P. Monteiro 2015). Such an performance (resistance to chloride and
enormous utilization of concrete calls for sulphate attack) reported by Zuquanet al.
higher use of natural aggregates and cement, (2000), Shi et al. (2000), and also presented in
thus taking toll on the environment. At least research's Chindaprasirt et al. (2004, 2007).
three-quarters of the total volume of concrete Since quartz powder is available as a
consists of coarse and fine aggregates. natural material that is made from sawing
Obviously, natural resources such as river stones and it is almost useless (Song et al.

505
2013), and also by product of industries, it Vietnam, coarse aggregates with nominal
could be a good choice for using as a filler in maximum size of aggregates as 20mm in
concrete. Alaa. M. Rashad (2011) said that proportions of 60% and 10mm in proportions
replacing quartz powder by cement without of 40%, with specific gravity of 2.68 g/cm3,
silica fume did not change the strength of density of 2.63 g/cm3, water absorption 0,5%,
hardened concrete even with 30% replacement fineness modulus of about (6,5÷6,8) average
and just increased the slump of fresh paste, strength 105,6 MPa were used. Natural sand
and the hydration in early ages. But in the from Huong river in the central region of
other research, Q. Yang et al. (2000) Vietnam with a specific gravity of (2.6-2.65)
concluded that quartz powder does not help g/cm3, water absorption 1%, fineness modulus
flexural strength so much but it improves (FM) of about 3.0 was used as fine aggregate.
compressive strength and micro structures and The sand and coarse aggregate used for locally
adding mineral admixture such as crushed available in the market and usual materials
quartz can increase fire resistance in concrete. used conforming to standards of TCVN 7570-
By using quartz powder as filler instead of 2006, ГОСТ 8736-2014, and ГОСТ 8267-93.
cement, the same strength can be reached
more economically. In repost’s M. Courtial et 2.1.2 The cement
al.(2013s) evaluated the variable parameter Cement used was sulfate-resisting Portland
which was quartz powder to silica fume ratio cement -PCSR40 (type 5) of company Luks
and superplastisizer amount, he said that Cement (Vietnam) Limited. All properties of
replacing condensed silica fume by fine quartz cement were in accordance with TCVN 7711-
powder particles improved almost all the 2013 and ГОСТ 22266-2013. The physical
properties such as compressive strength and properties of the cement used including
flexural strength, with replacing 25% of silica compressive strength at 3 days (25,7 MPa) and
fume by quartz crushed powder, but more than 28 days (46,5 MPa), initial setting time 170
that does not have an efficient effect, this minute and final setting time 213 minute,
problem also described details in the article's content C3A is 2,49% and content C4AF
Eroshkina et al. (2017). +2C3A is 21,23%, specific gravity of cement
In this project objective is to study the
3,15 g/cm3 and Sulfate resistance 14 days
influence of using fly ash and quartz sand
expansion is 0,04% and chemical composition
powder as replacement of cement and fine
of cement as listed in Table 2 below.
aggregates (river sand) respectively, on the
compressive strength, flexural strength 2.1.3 Water
properties and durability of concrete at the age
of 3, 7 and 28 days was obtained. The water used for the study was free of acids,
organic matter, suspended solids, alkalis and
2. MATERIALS AND METHODS impurities when present may have adverse
effect on the strength of concrete Potable water
2.1 Materials with PH value of 7.0 confirming to TCVN
The physical properties of cement, fine 4506-2012 and ГОСТ 23732-2011 was used
aggregates, coarse aggregates, fly ash and for making concrete and curing this specimen
water used for mix design of M80 grade of as well.
concrete were tested in laboratory and are 2.1.4 Quartz sand powder (Qp)
mentioned below:
The Qp used in this study is a form of white
2.1.1 The aggregates powdered quartz flour, which replaces partial
All the coarse aggregates used in this research fine aggregate. The particle size used ranges
work were obtained from crushed granite rock from 5 to 10μm and the specific gravity of
from GaLoi stone-pit in Central region of quartz powder is 2.6 g/cm3.

506
2.1.5 Chemical admixture
Superplasticizer (SP)used in this study is Sika®
ViscoCrete®-151 is a high performance
concrete superplasticiser based on ablend of
3rd generation PCE polymers. Density 1.075 ÷
1.095 kg/liter and pH Value 4,0 ÷ 6,0.
Sika®ViscoCrete®-151 meets the requirements
of ASTM C494 Type G, andconformity with Figure 2. Scanning electron microscopy (SEM)
standard Russia ГОСТ Р 56592-2015. micrographs of fly ash from Pha Lai
thermal power plants, use NanoSEM450
2.1.6 Mineral admixture (enlarge x 500 and x 2.000)
Silica fume(SF)as mineral admixture in dry
densified form obtained from Sikacrete® PP1 –
Sika limited Vietnam, have particle size <
0.1μm and specific gravity approx. 2,15 g/cm3
complies with ASTMC1240 and ГОСТ Р
56592-2015.
In this study, the fly ash (FA) used was of
class F from thermal power plants Pha Lai,
conformity with standard TCVN 10302-2014,
the FEI Nova NanoSEM 450 scanning electron
microscope which delivers best in class imaging
and analytical performance were used to study
Figure 3. X-ray diffraction (XRD) image of fly ash
the morphology of the fly ash particles were
shown as in Figure 2. In the present study, the Table 2. Chemical composition of cement and
Model 5SDH-2 Pelletron Accelerator (NEC, mineral admixtures
USA) was used to determine the elemental
composition of fly ash. The chemical Sikacrete Cement
Component Quartz ® Fly ash
composition of fly ash samples was analyzed (%) powder (Phalai) PCSR40
using X-ray fluorescence spectrometry (Shimazu PP1
1800, Japan) was presented in Figure 3. The fly SiO2 99,7 92,48 57,43 20,59
ash sample was also evaluated for their Al2O3 0,044 0,86 24,05 3,77
mineralogical composition by the SIEMENS Fe2O3 0,040 1,91 6,06 5,06
D5005 X-ray diffractometer (Bruker, Germany).
CaO 0.052 0,32 0,68 62,12
The XRD analysis thus confirms the
predominant alumina and silica composition of MgO 0.036 0,85 0,96 1,72
Fly ash PhaLai thermal power plants. From the K2O 0.007 1,22 3,60 0,63
X-ray fluorescence spectrum of fly ash, the Na2O 0.020 0,38 0,27 0,135
analysis determined the chemical composition of
TiO2 0,044 - 0,70 0,12
fly ash samples at Pha Lai thermal power plant
such as in Table 2. Observe the chemical P2O5 - - 0,11 -
composition in Table 2, showed that the fly ash ZrO2 - - 0,03 -
of Pha Lai thermal power plant of type F. The Cr2O3 - - 0,02 -
samples consist of almost regular spherical BaO 1,89
(cenospheres) particles ranging from 1 to 8 µm
SO3 - 0,3 0,30 2,0
in diameter. In the present, fly ash PhaLai is
finer than the others. Usually, fly ash composed Loss of
0,042 1,68 5,74
of mostly small and spherical particles. ignition

507
 By using a Laser diffraction analyzer for The aim of the present study was to study
measuring the particle size distributions of the the effect of fly ash on strength and durability
size-fractionated fly ashes, such as in Figure 4. of concrete by partial replacement of cement
It can be seen from Fig. 4 that, more than 85% with 0%, 20%, 30% and 40% of fly ash. The
of the fly ash particles of Pha Lai samples are optimal dosage of quartz sand powder ratio is
smaller than 90µm. 20% including 10% size 5µm and 10% size
10µm by volume of fine aggregate, coarse
aggregate content maintained at 1080 kg/m3,
Silica fume/binder ratio 10%, W/B ratio 0.3 by
mass respectively for all mixes. The mixture
proportions and the measured slumps of the
different batches of concrete are shown in
Table 3.
Similarly, each set of 12 cubes concrete test
specimen were cast to determine the compressive
strength for 0%, 20%, 30% and 40%
Figure 4. The particle size distribution of Pha Lai replacement of cement with fly ash respectively.
fly ash. All freshly cast specimens were left in the molds
for 24 hours before being molded and then
2.2. EXPERIMENTAL METHODS submerged in water for curing until the time of
testing, such as in Figure 5. For the compressive
In the present study, the HPC mixes were strength, 1503mm cubes mold were used to cast
designed the existing mix design methods the cubes and 3 specimens were tested for each
namely, TCVN 10306:2014 and according to age in a particular mix (i.e. the cubes were
standard ГОСТ 26633-2015. To start, mix crushed at 3, 7 and 28 days respectively), the
proportions for M80 grade concrete were cylindrical specimens of 150mm diameter and
obtained without considering the addition of 300mm height were also cast from each mix for
any mineral and chemical admixtures and were determining the splitting tensile strengthand
treated as basic mix proportions. Compressive beam specimen of size 150150600mm were
strength proportioning based on trial batches, used to determine the flexural strength of
and calculated following Eq. (1) below: concrete at 28 days, and rest specimens in
corrosion environments such as seawater and
f cr  1,1 f c  4,8  1,1* 70  4 ,8  81,8MPa (1) acid sulfate water, was presented in Fig. 6.

Table 3. Concrete mixture proportions (kg/m3)

Supplementary cementitious (Water +
Aggregates, kg/m3
materials, kg/m3 Super Super
Water Slump
Mixture Cement Coarse Fine Quartz sand plasticizer plasticizer)
Silica fume Fly ash /Binder
Number PCCR40 agg. agg. powder, kg
10
kg % kg % kg kg C/F kg % 5µm kg % kg (W+SP)/B mm
µm
O1 (M0) 550 - - - - 1088 1.75 621.7 - - - 156 1.8 9.9 0.30 105
O2 (M1) 550 - - - - 1088 1.75 497.4 10 62.2 62.2 163 1.8 9.9 0.31 195
O3 (S1) 412.5 5.0 27.5 20 110 1088 1.75 621.7 0 0 0 156 1.8 9.9 0.30 235
O4(M2) 412.5 5.0 27.5 20 110 1088 1.75 497.4 10 62.2 62.2 152 1.8 9.9 0.30 230
O5 (M4) 385 10 55 20 110 1088 1.75 497.4 10 62.2 62.2 156 1.8 9.9 0.30 240
O6 (M7) 330 10 55 30 165 1088 1.75 497.4 10 62.2 62.2 156 1.8 9.9 0.30 245
O7 (M9) 275 10 55 40 220 1088 1.75 497.4 10 62.2 62.2 156 1.8 9.9 0.30 255

508
 In this experimental study, the durability by concrete at different ages. This ultimate
properties of concrete in the corrosion load divided by the cross-sectional area of
environment test was carried out on HPC cube the cube (1503mm) yields the compressive
specimens of size 1503mm size. The strength of concrete, conforms to the standard
specimens were cured in fresh-water tank for TCVN 10303:2014 and TCVN 3119:1993.
28 days, subsequently, they were taken out of All experiments on mechanical properties
the curing tank and kept in air at ambient and durability of concrete were conducted at
temperature for 2 days for constant weight, the laboratory LAS-XD578 directly under
after that, the specimens were weighed prior Vietnam Institute for building science and
to submerging them into the marine technology (IBST Vietnam), such as in
environment and high sulfate solutions having Figure 7.
pH value of 2.5 (the pH value was periodically
checked and was maintained at 2-3), for 360
days (Fig. 6). After that, the specimens were
taken out and were washed in running water
and kept in the atmosphere for 2 days for
constant weight. Afterward, the specimens
were weighed and loss in weight and hence
the percentage loss of weight was calculated.
On the other hand, after removal from the
corrosion environment, their compressive
strength also was tested. Figure 7. The properties of concrete tested in the
laboratory

3. RESULTS AND DISCUSSIONS

Based on the laboratory test it has been observed
that the strength and durability characteristics
and their results correlate with the study and
derive positive result and improvement.
3.1 Effect of Quartz sand powder on
compressive strength of HPC
Figure 5. Fabrication and curing of concrete samples
From the observations result in Table 4 and
Figure 8 show that, partial replacement of fine
aggregate by 20% quartz sand powder (where,
including 10% size 5µm and 10% size 10µm)
respectively, compressive strength increased
7.5% (3 days), 16.1% (7 days) and 9.2% (28
days) in normal water curing at change
temperature 28÷31oC. This is possible due to
increased proportion of fine fillers that enhance
Figure 6. HPC cubes in the marine environment (left)
and acid sulfate water environment (right) the packing density, pore fill in action and long
chains of C-H-S gel. Here quartz powder act as
Compression testing machine Matest effective filler material at normal water curing
model C089-17N capacity 3000kN was used and also it’s act as a pozzolane material at
to determine the total compressive load taken higher temperature.

509
Figure 8. Influence of dosage of quartz powder on compressive strength of HPC

Table 4. Mechanical properties of HPC mixtures

Splitting Flexural
Silica fume

Quartz powder,
Fly ash

Cement (W+S tensile strength

(% fine Average Concrete Strength, MPa
Mix No PCCR40 P) / B strength fct fr at 28
aggregate) at 28 days days
% % % 5µm 10µm 3 days 7 days 28 days 56 days MPa MPa
O1 100 0.3 - - - - 67.58 71.11 81.750 83.846 5.71 7.86
O2 100 0.3 - - 10 10 72.62 82.56 89.246 91.534 5.98 9.17
O3 75 0.3 5.0 20 - - 41.33 62.22 65.283 66.957 4.86 7.85
O4 75 0.3 5.0 20 10 10 52.44 75.97 81.812 83.910 5.72 8.78
O5 70 0.3 10 20 10 10 62.15 81.91 87.921 105.749 5.94 9.11
O6 60 0.3 10 30 10 10 69.32 74.96 90.764 109.169 6.07 9.35
O7 50 0.3 10 40 10 10 65.33 72.44 91.138 109.619 6.09 9.27

3.2 Effect of fly ash on mechanical strength cement and water because of minor difference
of HPC between their compositions. From table 4 and
Fig. 9, it can be seen that as expected, the
According to standard TCVN 10306-2014,
High strength concrete use fly ash type F strength develops gradually, and particularly,
mixtures were designed at the age of 56 days the strength development of fly ash containing
or 90 days, the fly ash component makes the concrete is slower than concrete made without
concrete slower to development. Because of, fly ash. The results obtained from Fig.9
the hydration reaction between fly ash and indicate that during the initial curing period the
water is significantly slower than that between strength gain is slower as the amount of fly ash

510
increases, the compressive strength of concrete
decreases with increase in fly ash content. The
reduction in compressive strength of concrete
at the age of 3 days was found to be 16.84%,
4.75% and 11.15%, while at age 7 days found
to be 0.8%, 9.3% and 3.5% for 20%,30% and
40% replacement of cement with fly ash, this
proved the strength development gradually.
However, the compressive strength increases
gradually than controlled concrete when rising
gradually content of fly ash in the concrete
mixes, at age 28 days. Specific example, partial
replacement of cement by 10% silica fume and
40% fly ash (type F) and 20% fine aggregates Figure 10. Flexural and split tensilestrength of
by quartz sand powder, compressive strength concrete mixtures at 28-day
increased 2.5% in comparison with strength of
3.3 Effect of fly ash on durability in corrosion
control concrete. The compressive strength of
environment
binary blended Portland cement mortars
containing fly ash was lower than the Portland There are many research results available in the
cement control and tended to decrease with reports of the national cooperative highway
increasing fly ash content. The reduction of research program about subject "use of fly ash
compressive strength of binary blended cement in concrete" (Woodrow J. H. 1986s) and in the
containing fly ash is due to its slow pozzolanic reports of © RILEM 2018 "State-of-the-Art
Reports 25"(N. De Belie et al. 2018), in the
reaction and the dilution effect (Papayianni et
majority of which researchers report that Fly
al. (2010), Duran-Herrera et al. (2011), and
ash may have beneficial effects on both the
Wongkeo et al. 2012). Figure 10 shows the fresh and hardened properties of concrete
variation of flexural strength (modulus of mixtures, fly ash improves the resistance of
rupture) and split tensile strength with fly ash concrete to chloride ion penetration, sulfate
different percentages in concrete mixtures after resistance, freezing and thawing resistance,
28-day curing in fresh water. alkali aggregate reactions...etc.
Dikeou studied the resistance to sulfate
attack of a number of cement types with and
without fly ash (Dikeou1975). Fly ash type
class F was used in this investigation, generally
this type of fly ash will improve the sulfate
resistance of any mixture in which it is
included, although the quantitative amount of
improvement may vary with either the cement
used or the fly ash. He reported the following
order of resistance from the most resistant to
the least resistant for the fly ashes and cements
used in his studies. As follows: type V plus fly
ash, type II plus fly ash, type V, type II, type I
plus fly ash, and least is type I.
Figure 9. Compressive strength fluctuation for Several attempts have been made in order to
different proportions of Fly ashat 3, 7, predict the sulfate resistance of a fly ash based
28, 56, and 90 days of curing on its chemical composition. Among them the

511
first is that of Dunstan (1980s). Dunstan has 3.4 Permeability, water absorption and abrasion
proposed an indicator of the relative sulfate
resistance of a fly ash termed the "R" value. The 3.4.1 Permeability, water absorption of concrete
"R" value is a ratio of calcium to iron oxide Cylindrical specimens shaped molds of
expressed as see for example Eq. (3) below: 15x15cm, after 24 hours of casting, the
%CaO  5 0.68  5 specimens were demoulded and cured for 28
R   0.73  1.5 (3) days; it is then taken out, wiped dry and tested,
%Fe2O3 6.06
was used to test permeability and water
According to Dunstan, a fly ash improves absorption of concrete, in there, water
the sulfate resistance of the mixture, those absorption test was carried out according to
having R ≤ 1,5 generally improved sulfate TCVN 3113-2012 and calculated using the
resistance, whereas, thisdecreasesif the R > 3. following Eq. (4) below:
In a later research Dunstan (1987) reported that Wa %   100 *  M1  M 2  / M 2 (4)
the sulfate resistance of mortars tested
according to ASTM C-1012 will be improved, where, M1 is weight in grams of the saturated
if fly ash with R < 3.5 is used. specimen, and M2 is weight in grams of the
Sanjukta Sahoo et al. (2015), in research their oven-dried specimen
about "acid, alkali and chloride resistance of Permeability of concrete was conducted as
high-volume fly ash concrete" to find variation per TCVN 3116:1993 and calculated using the
in compressive strength and mass of high- following Eq. (5) below:
volume fly ash concrete samples subjected to
different chemical solutions of sodium chloride,

K  Q / A T * H
L  (5)

sodium sulphate and sulphuric acid. All where, K is coefficient of permeability in

experimental investigations with respect to acid, cm/sec, Q is quantity of water in mm
alkali and chloride resistance were carried out percolating over entire period of the test after
on high volume fly ash concrete (40% steady state has been reached, A is area of
replacement with cement), and low volume fly specimen face in cm2, T is time in seconds over
ash concrete (25% replacement with cement) which Q is measured, H/L = Ratio of the
and their performances against control concrete pressure head to thickness of the specimen,
(mix design proportions for 1m3 of control both expressed in same unit.
concrete including cement (OPC, 43 grades)
3.4.2 Abrasion of concrete
370 kg, w/c ratio 0.38 including 140,6 liter
water and 0,68 liter super plasticizer, fine Specimens were tested according to TCVN
aggregates (FM = 2.23) 807 kg, coarse 3114:1993. Size of specimens tested was
aggregates (FM = 6.58) 1171,5 kg and concrete 70.770.725mm. After 28 days curing,
density is 2422 kg/m3). The result study showed specimens were weighed to the nearest 0.1gm
that, after certain curing duration in fresh water using electronic balance, calculated follow
with four levels of curing period of 28, 56, 90 Eq. (6) below:
and 120 days. The resistance of control concrete M a   M1  M 2  /  M1 * A * V (6)
to all the three chemical attack is better only up
to 28 days of water curing. At 56 days of water where, Ma is average loss in thickness in mm,
curing low volume fly ash concrete (LFC) M1 is initial weight of specimen in grams, M2
shows better resistance against the control and is final weight of specimen in grams, V is
high-volume fly ash concrete (HFC). However, initial volume of specimen in cm3, and A is
with prolonged water curing of cubes of 90 days surface area of the specimen in cm2.
and more, HFC has consistently shown highest Testresults on permeability, water
resistance; whereas the control concrete faced a absorption and abrasion of concrete was
great loss in strength. presented in Table 5.

512
Table 5. Test results on permeability, water 3. The fly ash concretes yielded better
absorption and abrasion of concretespecimens. resistance to chloride ion penetration both at
Permeability K
28 and 90 days. Thus, it is possible to design
Water
at 28 days, absorption
Abrasion high strength concrete of reduced permeability
Mix No 10 resistance, by including up to 40% Class F fly ash in the
K * 10 in 24 hours,
mm total binder.
(cm/sec) %
O1 (M0) 1.60 2.50 0.836
5. REFERENCES
O2 (M1) 1.18 2.39 0.861
O3 (S1) 3.16 3.23 1.079 Báo Sài gòn Giải phóng. 20/5/2017. Nhức nhối vấn
đề xử lý nước thải, khí, bụi, tro xỉ của nhiệt
O4(M2) 1.21 2.52 0.847 điện than.
O5 (M4) 1.10 2.43 0.874 Báo Môi trường. 5/9/2018. Phải đảm bảo an toàn môi
O6 (M7) 1.03 2.19 0.770 trường trong phát triển nhiệt điện than. Tổng cục
Môi trường, Bộ Tài nguyên và Môi trường.
O7 (M9) 1.09 2.33 0.843 Báo Dân trí. 16/10/2018. Yêu cầu xử lý hàng chục
triệu tấn tro, xỉ nhiệt điện gây ô nhiễm môi
4. CONCLUSIONS trường.
Báo điện tử Dai bieu nhan dan. 16/11/2018. Tọa đàm
According to the test results, it can be “Nhiệt điện than và vấn đề an ninh năng lượng,
concluded that the strength and durability of môi trường và sức khỏe con người”.
the concrete is increased when partial Báo greenid vietnam. 10/03/2019. Nhiệt điện than
replacement of cement by mineral admixtures trong một tổng sơ đồ năng lượng quốc gia đổi
such as fly ash, silica fume and quartz sand mới. Tham luận của GS. TS KH Nguyễn Ngọc
Trân trình bày tại Tọa đàm nhiệt điện than do
powder... etc.
Ban tuyên giáo Trung ương tổ chức tại Hà Nội
1. Considering the effect of binder content ngày 05/03/2019; 20/01/2016. Vietnam coal
on strength, it can be seen that higher power waste increases. Tran DinhSinh of the
compressive strengths are achieved with higher Green Innovation and Development Centre
binder contents. The effect of fly ash on (Green ID).
strength of concrete containing SF/B ratio Báo Mekong Eye (author’s Le Quynh).
10%, W/B ratio 0.3 and content of fine 07/03/2019. Vietnamese provinces say “no”
aggregate replaced by quartz sand powder at to coal plants–government and industry still
level 20% by mass respectively for all mixes, want more.
when partial replacement of cement with 20%, .... etc.
Bộ Tài nguyên và Môi trường. 2016. Kịch bản biến
30% and 40% of fly ash. The result
đổi khí hậu và nước biển dâng cho Việt Nam.
experimental, show that concrete mixes Nhà xuất bản Tài Nguyên môi trường và Bản
achieved 88, 91 and 92 MPa at 28 days, while đồ Việt nam, 11-2016, 188tr. [Ministry of
at 56 days they achieved 106, 109 and 110 Natural Resources and Environment of
MPa. When comparing these strength gains it Vietnam (MONRE 2016). Scenarios of
can be stated that the efficiency of the binder in climate change and sea level rise for
terms of compressive strength increases as the Vietnam,Publisher Environmental Resources
curing time increases. and Map of Vietnam,November 2016, 188p].
2. From table 4, show that an increase Climate Change, 17 January 2019. Climate change
in Vietnam. Environment and natural
gradual in the workability when used fly ash
resources, 2019.
partial replacement cement at level 20%, 30% A. Mardani-Aghabaglou, M. Tuyan, K. Ramyar,
and 40% respectively, this was increased the 2014. Mechanical and durability performance
plastic viscosity of concrete whilst the yield of concrete incorporating fine recycled
stress of concrete reduced with increase in fly concrete and glass aggregates. Materials and
ash content (Khan MI 2010). Structures, Vol 48, 2014s, 2629-2640.

513
Anna Galińska, Sławomir Czarnecki, 2017. The N. De Belie et al. (eds.) 2018. Properties of Fresh
Effect of Mineral Powders Derived from and Hardened Concrete Containing
Industrial Wastes on Selected Mechanical Supplementary Cementitious Materials.
Properties of Concrete. IOP Conference RILEM 2018, State-of-the-Art Reports 25.
Series: Materials Science and Engineering Papayianni, I., E. Anastasiou. 2010. Production of
245, October 2017, 032039, 8pp. high-strength concrete using high volume of
Chindaprasirtet al. 2007. Influence of fly ash industrial by-products. Construction and
fineness on the chloride penetration of Building Materials, Volume 24, Issue
concrete. Construction and Building 8, August 2010, 1412-1417.
Materials, Volume 21, Issue 2, February Rashad, A.M. &Zeedan, S.R. 2011s. A preliminary
2007, 356-361. study of blended pastes of cement and quartz
Chindaprasirt P, Homwuttiwong S, Sirivivatnanon powder under the effect of elevated
V. Cement and Concrete Research, Volume temperature. Construction and Building
34, Issue 7, 2004s, 1087. Materials, 672-681.
Chindaprasirt P, Kanchanda P, Sathonsaowaphak Ruben Snellings, 2012.Gilles Mertens and Jan
A, Cao HT. Construction and Building Elsen. Supplementary Cementitious Materials.
Materials, Volume 21, Issue 6, 2007s, 1356. Reviews in Mineralogy & Geochemistry,
Dikeou, J. T., 1975. Fly Ash Increases Resistance Volume 74, May2012, 211-278.
of concrete to Sulfate Attack. A Water Sanjukta Sahoo et al. 2015. Acid, alkali and
Resources Technical Publication, Research
chloride resistance of High-volume fly ash
Report No. 23, U. S. Bureau of Reclamation,
concrete. Indian Journal of Science and
Denver, Colorado (1975s).
Technology, Vol 8(19), 72266, August 2015,
Dunstan E. R., Jr. 1987. Sulfate resistance of fly
ISSN (Print): 0974-6846.
ash concretes “the R-value”. In: Proceedings
Sanjukta Sahoo et al., 2015. Acid, Alkali and
of the katharine and bryantmather
chloride resistance of High-volume fly ash
international conference on concrete
concrete. Indian Journal of Science and
durability, ACI SP-100, 2027–2040.
Technology, Volume 8(19), 72266, August
Dunstan, E. R., Jr. 1980. A Possible Method for
2015; ISSN (Print): 0974-6846, ISSN
Identifying Fly Ashes that Will Improve the
Sulfate Resistance of Concrete. Cement, (Online): 0974-5645.
Concrete and Aggregates, Vol. 2, No. 1 Shi C, Stegemann JA. Cement and Concrete
(1980s), 20–30. Research, Volume 30, Issue 5, 2000s, 803.
Duran-Herrera et al., 2011s. Evaluation of Song, T., Lee, S. H. & Kim, B. 2013s. Recycling
sustainable high-volume fly ash concretes. of crushed stone powder as a partial
Cement and Concrete Composites, November replacement for silica powder in extruded
2010 Volume 33(1), 39-45. cement panels. Construction and Building
Fib Bulletin 34. 2006. Model code for service life Materials, 105-115.
design. Fib (ISBN 2-88394- 074-6), February T.G.S Kiran et al., 2014. Fly Ash as a Partial
2006, ISBN: 2-88394-074-6. Replacement of Cement in Concrete and
Khan MI 2010s. Rheological characteristics of Durability Study of Fly Ash in Acidic (H2SO4)
HPC containing composite cementitious Environment. International Journal of
materials. Concrete Technology - Journal of Engineering Research and Development,
the Concrete Plant International, No 02/10, Volume 10, Issue 12, December 2014, 01-13.
Germany, 78–84. T.R. Naik. 2008. Sustainability of concrete
M.P. Monteiro. 2015. Concrete: Microstructure, construction. Practice Periodical on
Properties, and Materials, November 25, 2015. Structural Design and Construction, Volume
MireilleCourtialet al. 2013s. Effect of 13, Issue 2 - May 2008, 98–103.
polycarboxylate and crushed quartz in UHPC: Thomas MDA, BamforthPB, 1999. Modelling
Microstructural investigation. Construction chloride diffusion in concrete: Effect of fly
and Building Materials, Volume 44, July ash and slag. Cement and Concrete Research,
2013, 699-705. Volume 29, Issue 4, April 1999, 487-495.

514
Vagelis G. Papadakis. 2000. Effect of fly ash on TCVN 3114:1993 Heavyweight concrete - Method
Portland cement systems. Part II: High for determination of abrasion.
calcium fly ash. Cement and Concrete TCVN 3116:1993Heavyweight concrete - Method
Research, Volume 30, Issue 10, October 2000, for determination of watertightess.
1647-1654. TCVN 10306:2014 High strength Concrete -
Vagelis G. Papadakis. 2000s. Effect of Proportional design with cylinder sample.
supplementary cementing materials on TCVN 7711:2013Sulfate resistant blended
concrete resistance against carbonation and Portland cements.
chloride ingress. Cement and Concrete TCVN 7570:2006 Aggregates for concrete and
Research, Volume 30, Issue 2, February 2000, mortar – Specifications.
291-299. TCVN 4506:2012Water for concrete and mortar.
Wongkeo et al., 2012s. Compressive strength and Technical specification.
drying shrinkage of fly ash-bottom ash –silica TCVN 8826:2011. Chemical admixtures for
fume multi-blended cement mortars. Materials concrete (ASTM C494, type G).
and Design, Volum 36, 655-662. TCVN 10302:2014Activity admixture - Fly ash for
Woodrow J. Halstead. 1986. Published reports of concrete, mortar and cement.
the national cooperative highway research TCVN 8827:2011 Highly activity puzzolanic
program, use of fly ash in concrete. admixtures for concrete and mortar-Silica
Transportation research board national fume and rice husk ash.
research council 2101 constitution avenue, TCVN 10303:2014 Concrete - Control and
N.W. Washington, D.C. 20418, October assessment of compressive strength
1986, -72p. TCVN 3119:1993 Heavyweight concrete -Method
Yang, Q., Zhang, S., Huang, S. & He, Y. 2000s. for determination of flexural tensile strength.
Effect of ground quartz sand on properties of GОSТ 18105- 2010. Concretes. Rules for control
high-strength concrete in the steam-autoclaved and assessment of strength (ΓOCT 18105-
curing. Cement and Concrete Research, 1993- 2010. Бeтноы.
1998 years. Пpaвилаконтpоляиоцeнкпрочности)
Zuquan Jin, Wei Sun, Yunsheng Zhang, Jinyang GОSТ 10180-2012. Concretes. Methods for
Jiang, Jianzhong Lai. strength determination using reference
CementandConcreteResearch. Volume37(8), specimens (ГОСТ 10180-2012 Бетоны.
(2000s), 1223. Методыопределенияпрочностипоконтроль
Ерошкина Н. А., Коровкин М. О., Чамурлиев нымобразцам).
М. Ю.,Измельченный Кварцевый Песок – GОSТ 31384-2017. Protection of concrete and
Потенциальное Сырье Для Получения reinforced concrete structures against
Бесклинкерных Вяжущих,Электронный corrosion. Generaltechnicalrequirements
научно-практический журнал (ГОСТ 31384-2017. Защита бетонных и
«Молодежный Научный Вестник», Май железобетонных конструкций от коррозии.
2017, УДК: 691.5: 691.335. [Eroshkina N. A. Общиетехническиетребования.)
Eroshkina N. A., Korovkin M. O., Chamurliev GОSТ 12730.5-84 Concretes. Methods for
M. Yu. 2017. Crushed quartz sand is a determination of watertightness (ГОСТ
potential raw material for the production of 12730.5-84.
clinkerless binders.Electronic scientific and Методыопределенияводонепроницаемости
practical journal "Youth Scientific Vestnik", (сИзменением N 1))
May 2017, 6p.] GОSТ 10181-2014. Concrete mixtures. Methods of
TCVN 3113:1993/2012Heavyweight concrete - testing (ГОСТ 10181-2014. Смесибетонные.
Method for determination of water absorption. Методыиспытаний)

515
The 3rd Int. Conf. on Transport Infrastructure & Sustainable Development (TISDIC 2019). Construction Publishing House. ISBN: 978-604-82-2893-4.

The time - dependent effect over time of water-to-cement

ratio on the flow friction of concrete pressure pipe

Nguyen Van Vinh1, Mai Chanh Trung1,*, Nguyen Hoang Vinh1

Nguyen The Duong2
1
The University of DaNang-University of Science and Technology, 54 Nguyen Luong Bang, DaNang,
Vietnam
2
Faculty of Civil Engineering, Duy Tan University, 03 Quang Trung, DaNang, Vietnam


Corresponding author, E-mail: [email protected]

ABSTRACT: Several studies on concrete pumping technology have indicated that the movement of the
pumping flow in the pipe is due to a boundary layer formed at the interface between the fluid concrete
and the pipe wall. The rheological characteristics of this boundary layer are called the friction parameters,
which include the viscous constant and the interface yield stress, and the pumpability of concrete depends
mainly on these friction parameters. Thus, the purpose of studying the effect of concrete mix parameters
on friction parameters is to estimate the pumpability of concrete and to determine the variation of these
parameters. Previous studies have evaluated the effect of concrete composition on friction parameters, but
not considered the construction time. In reality, due to the fact that the construction sites have numerous
technical issues, concrete construction time was prolonged. For better understanding this issue, the
authors proposed the research topic: "The time-dependent effect over time of water-to-cement ratio on the
flow friction of concrete pressure pipe according to time”.
Keywords: Pumping pressure, tribometer, viscous constant, interface yield stress, time.

1. INTRODUCTION related to rheological characteristics of fresh

concrete (concrete grouting component,
Civil projects nowadays require more and more
additive use, construction time, characteristics
complex technical issues, so it also requires the
and properties of friction layer between fluid
development of corresponding construction
concrete and pipe wall ...).
technology. For buildings constructed using
Overall, the parameters related to site
concrete technology, the concrete considered to
conditions are almost predetermined, so the
be capable of construction requires the
necessary workability to mix, transport, pump, pump pressure mainly depends on the
pour, level, compact, and maintenance ... as parameters related to the rheological properties
presented by Best et al. (1980), Wallevick et of the fresh concrete. Previous studies by
al. (1990), Kaplan (2000) or Chapdelaine Kaplan (2000), Kaplan et al. (2005) or Ngo et
(2007). The specifications for in situ concrete al. (2010)indicated that: the movement of the
construction can be divided into: the pumping flowin the pipe is governed bya
parameters related to site conditions (mortar boundary layer formed at the interface between
mixer, transportation means, type of pump, the fresh concrete and the pipe wall, as shown
size, length, material of the pump, required in Figure 1. The rheological characteristics of
flow, compactors, etc...) and the parameters this boundary layer are called the friction

516
parameters, which include interface yield stress especially when considering the hydration
(τ0t) and viscous constant (), and the reaction due to longer construction time;
pumpability of concrete depends mainly on therefore, affect the rheological properties of
these friction parameters. The interface yield fresh concrete, in other words, pumpability.
stress, 0t, characterizes the inertia of the Thus, this study presents the time-dependent
concrete flow in the pump pipe, the greater effecton flow friction of concrete pressure
interface yield stress requires initial greater pipeaccording to the W/C ratio, which is the
pump pressure to be able to push the concrete ratio of the weight of water to the weight of
flow to move in the pump pipe. The viscosity cement used in a cube metre of concrete mix.
In the present study, friction parameters are
constant is specific to the inertia of the
measured by friction measuring devices
concrete flow when the concrete has moved in
(tribometer). Many friction measuring devices
the pump pipe. The smaller viscosity constant are proposed by the authors Morinaga (1973),
means that the concrete flow is easy to move, Best et al. (1980), Chapdelaine (2007) or
in other words, easy to pump Ngo et al. (2011). Kaplan (2000), this study used the friction
measurement device developed by Ngo (2009)
due to the simple use process, accurate results
and small cost. These tribometers generally
have the same goal - to measure friction at the
interface between the fluid concrete and the
pipe wall (usually are made of steel) based on
the concrete flow rate (in the pump pipe) or the
rotation speed of rotary cylinder in fresh
concrete (in tribometer measurement).
Figure 1. The model of concrete flow in pump pipe
2. BUILDING OF EXPERIMENTAL DATABASE
The evaluation of these friction parameters 2.1 Tribometrical measurement
will estimate the pumpability of fresh concrete.
Several previous studies have demonstrated The apparatus used in this study was developed
that the thickness and rheological properties of by Ngo (2009) is composed as shown in Figure
boundary layer mainly depend on the mix 2(a). The tribological test procedure is
proportion of the concrete Kaplan et al. (2005), composed of 5 steps: (1) Fill the first concrete
Estellé et al. (2008), Ngo (2009), Ngo et al. layer up to half of the container and pierce
(2012), Mai et al. (2014), Mai (2015) or Ngo et through this layer 25 times; (2) Locate the
al. (2018). Besides, the study on the effect of cylinder at the centre of the container and let it
parameters of concrete mix proportion on the stands on the concrete surface by its own
friction atthe contact interface can allow to weight, the cylinder is then fixed to the agitator
determine the development of these parameters axis; (3) Measure the friction torque
when the components of fresh concrete mix are corresponding to different imposed velocities,
varied. However, these studies have not as seen in Fig. 2(b); (4) Once step 3 is finished,
mentioned the time-dependent effect when fill the second layer on top of the first layer to
investigating the development of the friction the top of the container and pierce through this
parameters according to the parameters of layer 25 times; (5) Repeat step 3 and record the
concrete mix proportion, which are determined new results.
or measured immediately after mixing or The interface friction torque is obtained by
transporting to the construction site. In fact, the subtracting results of step 3 from those from
construction time sometimes has to be step 5. For each imposed speed level, the
prolonged due to unexpected reasons, which average interface friction on the cylinder
lead to the changes of concrete properties, circumferential area during the constant

517
rotation speeds is obtained. The obtained Using the values of T0 and V (fundamental
interface friction torque is recorded for the units) obtained from Eq. (2), in taking into
decreasing speed direction to identify the account the cylinder shape and its dimensions,
pumping parameters. The test results can be the interface parameters (0t and ) can be
approximated by an empirical linear function calculated by Eq. (3):
as follows:
T0 k
T  T0  k.V (2)  0t  and   (3)
2 R 2 h  2 3 R3h
where T (N.m) is the applied torque to the
revolving cylinder, T0 (N.m) is the initial 2.2 Experimental program
torque, k (N.m.s) is the linear coefficient, and V
In the present study, the commonly used local
(cycles/s) is the cylinder rotating speed.
ingredients were used to make concrete. All
studied ordinary concretes in this work are
produced with the cement PCB40 which is
based on clinker without additions. A granular
skeleton is formed by regular concrete sand
with grain-size distributions of 0/4 mm and
different crushed aggregates of grain-size
distributions 10/20. All coarse aggregate and
sand materials have specific gravities of 2.5. A
a) super-plasticizers Sika Plast®257 is used. For
every studied series in this program, in order to
highlight effect of each composition parameter
on the pumping parameters, only one
formulation parameter is varied while keeping
other parameters constant. For each mix
proportion, the tests will be performed to
measure friction parameters at times of 0, 30,
60, 90 minutes. To minimize the hydration
reaction and "hardening" phenomenon during
tests, fresh concrete in the mixing tank will be
b) covered by polycarbonate sheets and mixed for
Figure 2. The used tribometer and the imposed every 15 minutes. The summary of the
steed profile Ngo (2009). experimental program is presented in Table 1.

Table 1. Composition of tested concrete mixtures and results

Vpaste Cement Water Agregates Sand Sika 257

Specimens W/C A/S
(m3) (kg) (kg) (kg) (kg) (l/m3)

B1 0.4 440 175.5 995 888 4.1

B2 0.45 0.317 1.12 410 184.5 995 888 4.1

B3 0.5 385 192.5 995 888 4.1

B4 0.5 0.378 1.12 460 230 876 795 0

518
3. ANALYSIS OF THE EXPERIMENTAL plasticizer B1B3, the effect of reducing
RESULTS water and prolonging the setting time of
concrete is the reason to maintain the
The experimental results are summarized in
necessary concrete work ability when
Table 2 and Figure 3.
prolonging the construction time compared to
Experimental results from Figure 3(a)
concrete without super plasticizer.
showed that when the time increases, the
With respect to the effect of W/C ratio and
slump of concrete decreases, however, the
time factor on friction parameters, the
decline in slump of fresh concrete using
superplasticizer is slower than that not using experimental results in Figure 3(b, c) showed
additives. It wasthe fact that the amount of that the increasing of water retention time led
water used in the concrete mixture, partly to the increase ofparameters friction (interface
related to the hydration reaction, partly related yield stress and viscous constant) whether or
to the work ability of fresh concrete for the not the use of additives, thereby reducing the
construction process. Since the construction pumpability of fresh concrete.It is attributed to
process is prolonged, dehydration (due to the fact that W/C ratio plays an important role
many reasons, including evaporation) is one in concrete plastic viscosity that influences the
of the main reasons for the decline in concrete formation of the boundary layer at the interface
slump. With the concrete using super and thus the interface friction.

Table 2. The results of friction parameters

Time (min)
Specimens Pumping parameters Unit Vpaste W/C
(m3) 0 30 60 90

Slump cm 20 17.5 16 13

B1 Interface yield stress (0t) Pa 0.317 0.4 36.2 52.5 55.7 67.4

Viscous constant ( Pa.s/m 598 674 781 1027

Slump cm 21.5 19 17 15

B2 Interface yield stress (0t) Pa 0.317 0.45 29.6 42.5 41.5 54.3

Viscous constant ( Pa.s/m 495 537 637 792

Slump cm 23 21.5 18 17.5

B3 Interface yield stress (0t) Pa 0.317 0.5 21.4 33.5 36.7 47.1

Viscous constant ( Pa.s/m 407.2 472 579 715.8

Slump cm 10.5 8.5 6.5 2

B4 Interface yield stress (0t) Pa 0.378 0.5 52.6 64.8 72.5 #

Viscous constant ( Pa.s/m 693 827 1046 #

519
 From the Figure 3(b, c), when the W/C ratio
varied from 0.40.5 correspondings to B1B3,
with the time varied from 090 mins, while the
amount of cement and the A/S ratio remain the
same, the friction parameters are reduced. It is
the fact that when increasing the W/C ratio
while maintaining the other components, the
fresh concrete is diluted and thus reduce the
viscosity. Know that this layer is formed by
fine-grained components: cement, water, and
fine sand during concrete during testing, as
research by Ngo et al. (2010). Therefore, a
fresh concrete with less viscosity and thicker
will promote the formation of friction layer at
(a)
the contact surface, thereby reducing friction
and promoting pumpability. On the contrary,
the retention time of the mortar is prolonged,
the amount of water in the concrete decreases,
thereby increasing the viscosity of the fresh
concrete and thus increasing the friction and
reducing the pumpability. In B4 specimen, the
increase in surface friction is significant when
no additive is used, despite having the same
W/C ratio as B3, thereby, the role of super
plasticizer in work ability of concrete when
extending construction time is confirmed.
Moreover, during the survey period, the
average increase in friction parameters was
quite large, about 80% in interface yield stress
(b) and 65% inviscosity constant.

4. CONCLUSIONS
In the present study, the role of the W/C ratio on
friction atthe pipe wall over time has been
clarified. The friction measuring device
(Tribometer) developed by Ngo et al. (2010)
was used to investigatethe variationof friction
parameters (interface yield stress (τ0t) and
viscous constant ()) according to time. From
the research results, it can be concluded that:
(i) The experimental showed that the time-
dependent effects of friction parameters
between the fluid concrete and the pipe wall
could be evaluated.
(c) (ii) When the water retention time increases,
Figure 3. Experimental results: the developmentover the concrete slump decreases. This degree of
time of slump (a) and friction parameters (b, c) slump reduction is greater than concrete that

520
does not use additives and thus greatly affects Mai, C. T., Kadri, E. H., Ngo, T. T., Kaci, A.&
the work ability of fresh concrete, leading to Riche, M.2014. Estimation of the Pumping
the difficulty in the pumping process. Pressure from Concrete Composition based on
(iii) Increasing the W/C ratio generally the Identified Tribological
Parameters.Advances in Materials Science and
increases slump, reduces friction at the contact
Engineering, Volume 2014, Article ID
surface of the pump wall, leading to the 503850, 18 pages, p. 18.
convenient pump process. Morinaga, M. 1973. Pumpability of concrete and
(iv) Increasing mortar retention time pumping pressure in pipelines. Fresh
reduced the work ability of fresh concrete: Concrete: Importante Properties and Their
reduce slump, increase friction parameters, Measurement.Proceedings of a RILEM, vol.
making it difficult to pump freshconcrete. To 7,1-39.
overcome this situation, in addition to Ngo, T. T., Mai, C. T., Kadri, E. H.& Kaci, A. 2018.
increasing the W/C ratio, it is necessary to use Proceedings of the 4th Congrès International de
a superplasticizer. Géotechnique - Ouvrages -Structures: CIGOS
Funding: This research is funded by Funds 2017, 26-27 October, Ho Chi Minh City,
for Science and Technology Development of Vietnam: "Application of Empirical Models to
Optimizing Concrete Pumpabiltity", Springer
the University of Danangunder [project
Singapore, 2018. https://doi.org/10.1007/978-
number B2017-ĐN02-35]. 981-10-6713-6_33
Ngo, T. T. 2009. Influence de la composition des
5. REFERENCES bétons sur les paramètres de pompage et
validation d'un modèle de prévision de la
Best et Lane. 1980. Testing For Optimum
constante visqueuse. Thesis of Ph.D,
Pumpability of Concrete, ACI Concrete
University of Cergy Pontoise, France.
International, vol. 2, No.10, 9-17.
Ngo, T. T., Kadri, E. H., Bennacer, R.& Cussigh,
Chapdelaine, F. 2007. Étude fondamentale et
F. 2010. Use of tribometer to estimate
pratique sur le pompage du béton. Université
interface friction and concrete boundary layer
Laval de Montréal Google Scholar.
composition during the fluid concrete
Estellé, P., Lanos, C.& Duc, T. 2008.
pumping.Construction and Building
Caractérisation de l'écoulement du béton
Materials, vol. 24, n°7, 1253-1261.
frais.43ème Colloque du Groupe Français de
Ngo, T. T., Kadri, E. H., Cussigh, F.& Bennacer,
Rhéologie-Rhéologie et thermodynamique,
R. 2012. Relationships between concrete
139-142.
composition and boundary layer composition
Kaplan, D. 2000. Pompage des bétons, Ed. Etudes et
to optimize concrete pumpability.European
recherches des laboratoires des Ponts et
Journal of Environmental and Civil
Chaussées, OA 36, Rapp. LCPC, Paris.
Engineering, vol. 16, n°2,157-177.
Kaplan, D.,De Larrard, F. & Sedran, T.2005.
Wallevick, O. H.& Gjorv, O. E. 1990. Properties of
Disign of concrete pumping circuit.ACI
fresh concrete.
materialsjournal, vol. 102, No 2, 110-117.
Mai, C.T. 2015. Proposition, validation d’un
modèle du seuil d’interface à partir des
paramètres de formulation des bétons et calcul
de la pression de pompage. Thesis of Ph.D,
University of Cergy Pontoise, France.

521
The 3rd Int. Conf. on Transport Infrastructure & Sustainable Development (TISDIC 2019). Construction Publishing House. ISBN: 978-604-82-2893-4.

Resistance of rubberized cement-based composites to

acid attack

Pham Ngoc Phuong 1,2,*, Ahmed TOUMI 2, Anaclet TURATSINZE 2

1
Faculty of Bridge and Road Engineering, The University of Danang - University of Science and
Technology, 54 Nguyen Luong Bang Str, Danang, Vietnam
2
LMDC, INSAT/UPS Génie Civil, 135 Avenue de Rangueil, 31077, Toulouse Cedex 04, France

Corresponding author, E-mail: [email protected]

ABSTRACT: Rubberized cement-based materialsare suitable for pavement constructions due to their
improved strain capacity and lower propensity for cracking under restrained shrinkage condition, that are
required properties for pavements on sub-layers. However, bond defectsat rubber-cementitious matrix
interface are detrimental to the durability of the composites. Therefore, this investigation focused on the
effect of bond enhancement on resistance of the materials to acid attack. The acetic acid with pH 4.0
satisfying requirements of XA3 class in NF EN 206-1/CN (2012) standard was chosen to carry out the
test. The decreasing of degraded depth and of mass loss during acetic acid attack was observed on
rubberized mortars compared to the ones of control mix. The solutions to enhance the interface between
rubber aggregates and cementitious matrix by using air-detraining admixture or, especially, by copolymer
coating contributed efficiently to improve the sustainability of the rubberized composites under the acid
environment. The Scanning Electron Microscopy (SEM)-based microstructural analysis was also
conducted to clarify changes at the interfacial zone.
Keywords: Rubber aggregates, rubberized cement-based composites, rubberized pavements, acid attack,
microstructural analysis.

1. INTRODUCTION changing from 0 to 20%, by weight. After

Rubberized cement-based materials are immersing concrete samples under acid solution,
suitable for pavement constructions due to their authors found more losses in compressive
improved strain and lower propensity for strength and in mass forconventional concrete
restrained shrinkage cracking (Turatsinze et al. than rubcrete. However, the water absorption
2016) as well as higher resistance to freeze- was higher as observed on acid attacked-
thaw conditions (Pham et al. 2019a). However, rubcrete specimens. Hence, they concluded that
the durability of such pavement materials the rubcrete can be used in areas exposed to high
exposed to aggressive environments (acid/ salt risks of chemical attacks. The authors explained
attacks, etc.) is still a concern. that presence of RA in rubberized concrete helps
With regard to resistance of rubcrete to hold concrete constituents and to prevent
(rubberized concrete) to acid attack, Thomas et cracking initiation and materials separation. On
al., (2015) used three-size rubber aggregates the other hand, as for the conventional concrete
(RA) (2-4 mm, 0.8-2 mm and RA powder) as a or the rubcrete incorporating less RA contents,
replacement forsand at different contents, cracks were easy to develop, resulting in the

522
deterioration of the composites (Thomas et al. changes at RA-cementitiousmatrix interface
2016). Segre et al. (2004) compared the due to acid attack.
performance of conventional and the one of
rubberized mortars against chemical attack by 2. MATERIALS AND EXPERIMENTAL
immersing cylindrical specimens (45 mm in METHODS
height and 30 mm in diameter) in hydrochloric
2.1 Materials
acid (HCl) of 5% (1.4 M). Note that, to prepare
the rubberized mortars specimens, 10% of RA RA used as a partial volume replacement of sand
(size 200 µm) by weight saturated in NaOH in cementitious mixtures were obtained from a
aqueous solutions within 20-30 min before mechanically grinding process ofend-of-life tyres
incorporatingthem in cementitious mixtures. The that finally produces aggregates with selected
specimens were also cured for 40 days at room size in the range from 0 to 4 mm. The low stiff
temperature (25C) and at 100% relative aggregates were characterized by insignificant
humidity (RH) before the immersion. The water absorption and specific gravity of 1.2. Due
weight loss of each mortar specimens measured to hydrophobic property of RA, they repel water
confirmed higher resistance and better and entrap air bubbles on their surface, further
performance of the designed rubberized mortar partly releasing them into cementitious matrix
and by the way increasing the air content of the
against HCl attack. Ganesan et al., (2012) also
composite. This study used two cases of
pointed out that the mass loss of rubcrete
admixtures with the aim to enhance the
samples incorporating 15% of RA (size less 4.75
interfacial zone between RA and cementitious
mm) after an attack by sulfuric solution (H2SO4)
matrix such as air-detraining admixture (2.56
(3%) for 90 days was less than the one of kg/m3) and co-polymer (2% of RA by mass), as
control concrete. investigated by Pham et al. (2018b).
On the contrary, Azevedo et al. (2012) Four different mix designs presented in Table
investigated the behaviourof control and 1 were tested. Acronyms UR and CR refers to
rubcreteagainst sulphuric acid attack and the mixtures incorporating Uncoated RA and
concluded that higher RA contents resulted in Coated RA, respectively; letters P and A relates
more mass loss. Note that, RA (1-2.4 mm) to treatment materials forenhancement ofRA-
were incorporated as a sand replacement and cementitious matrix interface, namely co-
these concrete samples with a 150 x150x 150 Polymer bonding resin and Air-detraining
mm3 size were immersed in a 10% of sulphuric admixture. To improve workability and to
solution during 28 days. prevent the segregation of the mixtures when
As for a recent study, Pham et al. (2019b) RA were added, superplasticizer and
evaluated the deterioration of rubberized viscosityagent were used, respectively.
mortarsunder aggressive environments. Based According toprecoating method, as suggested
on damage variable, the authors concluded that by Pham et al. (2018a), RA and copolymer were
rubberized cement-based composites behaved initially mixed together. Then, the coated RA
better under acid/sulfate attacks, especially were transferred for 1 h in a room where
when the bond at rubber-cementitious matrix temperature and RH were controlled (20 °C and
interface was enhanced by incorporating RA 50%, respectively). This process aimed to
pre-coated with copolymer. It should be noted condense and stabilizecopolymer prior toshortly
that damage variable qualifiedthe variation of and lightly mixing the coated-RA with apre-
load-resisting area due to chemical attacks. mix. Amixing sequence of the pre-mixture was
Hence, this study is just dedicated to further the same forcontrol mix (0R) and for the ones
show distresses in the control and in rubberized incorporating uncoated-RA (30UR and 30UR-
mortarscaused by acid-attacking such A). In general, sand, cement, andRA (if
asdegraded depth and mass loss. The SEM applicable) were pre-mixed within 3 min, and
analysis was used to explore microstructural water together with admixtures were finally

523
added accordingly to finalizethe mixing The durability of the control mortar and the
procedure during 2 min. one of rubberized mortars attacked by acetic
acid is investigated by comparing corrosion
Table 1. Mix design (values in kg/m3) depth and mass loss as described below:
Compositions 0R 30UR 30UR 30CR
2.2.1 Corrosion depth
-A -P
Cement CEM I 500 500 500 500 The corrosion depth was determined based on a
52.5R colorless region close to the exposed surface
Sand (0-4 mm) 1600 1120 1120 1120 where mortar specimens reacted with acid. To
RA (0-4 mm) - 220 220 220 determine corrosion depth, two tested samples
Water 235 235 235 235 of each mortar were sectioned perpendicularly
Super-plasticizer 3.25 3.25 3.25 3.25
to their lateral surface. Phenolphthalein solution
was then sprayed on such sectioned area to
Viscosity agent 0.9 0.9 0.9 0.9
determine altered regions. The undegraded zone
was observed using camera and determined
2.2 Experimental methods
with a support of Image J software. The
Acid attack testson control and rubberized degraded depth illustrated in Fig.1 is then
mortars were carried out according toASTM calculated as Eq. (1), where d is the degraded
C267-01standard (2012). The NF EN 206-1/CN depth (mm), R is the radius of original
standard (2012) classifies aggressive cylindrical mortar specimen (14.1 mm), A is the
environmental conditions based on exposure undegraded area (mm2) (from Image J).
levels, changing from low risk (XA1) to the high
risk (XA3) of chemical attacks. Such d  R A/   (1)
classifications were based on the concentration
of chemicals (acid, sulfate, etc.) from Degraded depth (d)

groundwater or natural soils. According to the Surface after

Surface before
immersion
standard, the class XA3 is the most unfavorable immersion
exposure, corresponding to the conditions where
28.2 mm
aggressive conditions reaches the pH up to the
range of 4.0 to 4.5. Due to the stable Undegraded
concentration of the acetic acid in such pH scale Degraded area
layer
(Hegyi et al, 2013), the acid prepared with pH
4.0 was used.
The acetic acid (CH3COOH) solution with the Figure 1. Illustration of degraded depth determination
concentration of 0.28 M was prepared with the
liquid-tested specimens volume ratio changing 2.2.2 Mass loss
from 20 to 30. The acetic acid and the distilled The mass loss of each acid-attacked mortar was
water were firstly stirred. The obtained solution measured regularly until completion of the
was then added with 2 g/l NaOH to reach the pH experiment. Note that the balance used for
of 4. After the 28-day curing period in controlled weighting has an accuracy up to 0.0001 g. The
conditions (20C temperature and 95% RH), weight loss is calculated using Eq. (2), where
testedmortar samples (28.2 mm in diameter and m0 is initial mass of cylindrical mortar
60 mm in height) were wiped and then weighed specimen (before immersion) (g), mi is mass of
before an immersionin the acid solution. During cylindrical mortar specimen after time i of
whole experimental process, the stability of the immersion (g), and S is the lateral area of
concentration was controlled, thus the pH of 4 cylindrical mortar specimen (cm2).
was checked regularly. As monitored, the acid
mo  mi  g 
solution wasreplaced every 2 days for the first 10 mass loss   2 (2)
days and weekly until the testended (7 weeks). S  cm 

524
2.2.3 Microstructural observations 3.2 Mass variation
The rubber-cementitious matrix interfacial The mass loss versus duration of immersion in
zone in degraded areas was observed on acid attack is presented in Fig. 3. At the
fractured samples by SEM after 3 weeks of beginning, mass loss was not much different
acid attack and then compared to the for all mortars. It can be attributed to the wall
oneimmersed in water (without acid reaction). effect owing to compaction during casting
It aims to see microstructural changes at the process and by which cement paste
interfacial regions due to decalcification. One concentrated similarly on the lateral surface of
should note that the fractured samples were all mortar specimens, limiting acid diffusion.
dust-freeand then coated by a layer of carbon However, after damage of that surface layer,
justbefore SEM observations. mass loss of the control mortar exhibited
higher value than that of rubberized mortars.
3. RESULTS AND DISCUSSION This resultis in agreement with findings from
other studies (Thomas et al., (2016), Segre et
3.1 Degraded depth al., (2004), and Ganesan et al., (2012))andbut
For all composites, Fig. 2 shows an increase in not consistent with the one of Azevedo et al.,
altered depth versus time of immersion in (2012). The mortar using coated-RA showed
acetic acid. The control mortar exhibited a low the lowest mass loss. Usage of air-detraining
resistance to acid attack compared to admixture to reduce the porosity of the
rubberized mortars. It can be explained by interface seemed not efficient to improve
significant water absorption of sand compared significantly the resistance to acid attack.
to hydrophobicity of RA, and by higher water
capillary absorption and porosity of the control
mortar. The air-detraining admixture addition
or, especially rubber coating methodhelped the
rubberized mortars to be more resistant to acid
attack and to limit distresses induced by
decalcification phenomenon. Indeed, after 4
weeks of acid immersion, the degraded depths
of untreated rubberized mortar were 1.11 and
1.3 times greater than the ones using air-
detraining admixture and rubber coating,
respectively (Fig.2). It may be attributed to Figure 3. Mass loss versus duration of acid immersion
well-contact interface or bond enhancement at
RA-cementitious matrix interface due to the 3.3 Microstructural characterizations
effect of co-polymer coating.
Fig. 4 shows micro-structural changes in
rubber-cementitious matrix interface of the
composites before and after acid attack. Before
acid attack, bond defects can be observed at the
interface between RA and cementitious
matrixwhen untreated RA were used (30UR
and 30UR-A).In degraded area, such interface
appears to beporous (Fig. 4). As for mortar
incorporating RA coated by copolymer, the
deteriorated cement paste is still well-bonded
to the coated RA. Such behavior can explain
Figure 2. Effect of acid attack: Altered depth the decrease in the corroded depth and the
versus time mass loss of rubberized mortar incorporating

525
copolymer-coated RA, resulting in better reduced air permeability and the lower water
resistance of coated rubberized mortar to acid absorption of the coated rubberized mortar.
environment attack.
5. REFERENCES
ASTM C267-01(2012), 2012. Standard Test
Methods for Chemical Resistance of Mortars ,
Grouts , and Monolithic. ASTM International,
West Conshohocken, PA, 2012,
www.astm.org, pp.1-7.
Azevedo, F. et al., 2012. Properties and durability
of HPC with tyre rubber wastes. Construction
and Building Materials, 34, pp.186–191.
Available at:
http://dx.doi.org/10.1016/j.conbuildmat.2012.
02.062.
Ganesan, N., Bharati Raj, J. & Shashikala, A.P.,
2012. Strength and durability studies of self
compacting rubberised concrete. Indian
Concrete Journal, (950802003), pp.15–24.
György Hegyi, József Kardos, Mihály Kovács,
András Málnási-Csizmadia, László Nyitray,
Gábor Pál, László Radnai, Attila Reményi,
Figure 4. Microstructural changes in RA- I.V., 2013. Introduction to Practical
cementitious matrix interface in Biochemistry, Eötvös Loránd University.
degraded areas (C: Cement matrix; UR NF EN 206-1/CN, 2012. Concrete - Specification,
and UR-A: Untreated RA without and performance, production and conformity -
with air-detraining agent, respectively; National addition to NF EN 206-1,
CR-P: RA coated by copolymer) Pham, N.P., Toumi, A. & Turatsinze, A., 2019b.
Evaluating damage of rubberized cement-
4. CONCLUSIONS based composites under aggressive
environments. Construction and Building
This study has investigated the durability of Materials, 217, pp.234-241.
rubberized cement-based composites under https://doi.org/10.1016/j.conbuildmat.2019.05.
acid attack. In general, the materials hada 066
better performance in such aggressive Pham, N.P., Toumi, A. & Turatsinze, A., 2019a.
environment. It is expected to result in Effect of an enhanced rubber-cement matrix
improved sustainability of rubberized mortar/ interface on freeze-thaw resistance of the
concrete based-applications. From cement-based composite. Construction and
experimental work, major conclusions can be Building Materials, 207, pp.528-534.
https://doi.org/10.1016/j.conbuildmat.2019.02.
drawn as following:
147
(1) Acid diffusion into the rubberized mortars
Pham, N.P., Toumi, A. & Turatsinze, A., 2018b.
was limited. It resulted in a decrease of Effect of Styrene-Butadiene Copolymer
degraded depth and mass loss. Coating on Properties of Rubberized Cement-
(2) Enhancing the RA-cementitious matrix Based Composites. In V. Mechtcherine, V.
interface by using air-detraining admixture or Slowik, & P. Kabele, eds. Strain-Hardening
RA coatingby copolymer contributed efficiently Cement-Based Composites: SHCC4.
to increase resistance of rubberized Dordrecht: Springer Netherlands, pp. 342–
mortarsagainst acid attack. The improvement of 350. https://doi.org/10.1007/978-94-024-
thesedurability properties was explained by the 1194-2_40.

526
Pham, N.P., Toumi, A. & Turatsinze, A., 2018a. Thomas, B.S., Gupta, R.C. & Panicker, V.J., 2016.
Rubber aggregate-cement matrix bond Recycling of waste tire rubber as aggregate in
enhancement: Microstructural analysis, effect concrete: Durability-related performance.
on transfer properties and on mechanical Journal of Cleaner Production, 112, pp.504-
behaviours of the composite. Cement and 513.
Concrete Composites, 94(August), pp.1–12. http://dx.doi.org/10.1016/j.jclepro.2015.08.046.
https://doi.org/10.1016/j.cemconcomp.2018.0 Turatsinze, A., Measson, M. & Faure, J.P., 2016.
8.005. Rubberised concrete: from laboratory findings
Segre, N. et al., 2004. Rubber-mortar composites: to field experiment validation. International
Effect of composition on properties. Journal of Journal of Pavement Engineering,
Materials Science, 9, pp.3319–3327. 8436(December), pp.1–10.
Thomas, B.S. et al., 2015. Performance of high http://dx.doi.org/10.1080/10298436.2016.121
strength rubberized concrete in aggressive 5688.
environment. Construction and Building
Materials, 83, pp.320–326.
http://dx.doi.org/10.1016/j.conbuildmat.2015.
03.012.

527
The 3rd Int. Conf. on Transport Infrastructure & Sustainable Development (TISDIC 2019). Construction Publishing House. ISBN: 978-604-82-2893-4.

Improving properties of OPC at high temperature by fly ash

Do Thi Phuong1, Huynh Phuong Nam1*, Vu Minh Duc2

1
Faculty of Road and Bridge engineering , The University of Danang - University of Science and
Technology, Danang, Vietnam
2
Faculty of Building Materials, National University of Civil Engineering, Hanoi, Vietnam

Corresponding author, E-mail: [email protected]

ABSTRACT: This paper studies on properties of binder mixed of Ordinary Portland Cement (OPC) and
fly ash (FA) under high temperature. OPC was replaced by 0, 20, 25, 30 and 35% of FA mass. After
casting, specimens were cured under steam condition for 4 hours. Then, they were dried at temperature of
100oC for 24h and heated at temperature levels of 200, 400, 600, and 800oC for 2 hours. After that, the
specimens were cooled in air. Finally, their apparent density and compressive strength were measured.
The results showed that with FA replacement of 25% for OPC (bymass), such properties of hardened
binder paste at temperature of 800oC were maximum improved. Furthermore, the paper also showed
physicochemical analysis results of the cement matrices usingdifferential thermal analysis (DTA),
thermogravimetric analysis (TGA) and X-ray diffraction (XRD) methods.
Keywords: Fly ash, binder, high temperature, compressive strength, apparent density

1. INTRODUCTION (if it is cooled in air) makes the matrix expand

and destroy its structure (Seleem et al., 2011).
Ordinary Portland Cement (OPC) is hydrate From 600oC to 800oC, second phase of the
binder, which is mostly made of limestone and CSH is decomposed to form β-C2S. From
clay. It is a common binder used for 930oC to 960oC, CaCO3 is decomposed that
construction. However, at high temperature, makes the matrix porous. If the hardened
chemical composition and structure of cement paste is heated at higher temperature, it
hardened cement paste are changed, leading to starts to be destroyed and totally terminated at
reduction of compressive strength. When the temperature over 1200oC (Hager, 2013).
hardened cement paste is heated from 100oC to In order to increase performance of OPC at
250oC, compressive strength can improve due high temperature, many researchers supposed
to autoclaved process of minerals (Morsy et al., that mineral admixtures should be mixed with
2012). If the temperature is over 250oC, the OPC to react with free CaO. So far, some of
compressive strength reduces according to the fine mineral additives have been used to
loss of physical water. At 300oC, micro cracks improve properties of OPC at high temperature
are generated (Hertz, 2005). From 460oC to such as chamotte(Remnev, 1996), high-
540oC, water in CH mineral is lost and CaO is alumina fine powder (Hlystov et al.,
created (Hager, 2013), resulting in shrinkage of 2015),mullite, chromite, forsterite(Nekrasov
the matrix. The second hydration of free CaO &Tarasova, 1969)or industrial byproducts such

528
as silica fume, blast furnace slag, FA, composition and properties of OPC and FA were
metakaolin (Hertz, 1992; Wang, 2008; Aydın showed in Table 1 and Table 2, respectively.
& Baradan, 2007; Morsy et al., 2012). Some Physico-mechanical propertiesof OPC and FA
studies showed that if OPC was replaced by were determined by Vietnamese Standards
FA of 10% (bymass), all surface cracks of the TCVN 4030:2003, TCVN 6016:2011, TCVN
matrix would be eliminated after cooling it in 6017:2015 and TCVN 10302:2014.
the air from 600oC (Dias et al., 1990). FA can
also improve compressive strength of the Table 1. Chemical composition of OPC and FA
hardened binder paste, especially at the Chemical composition (%) OPCFA
temperature over 300oC (Grainger, 1980). If
OPC is replaced by FA at around 20% to 30% Silicon dioxide (SiO2) 21.12 53.70
(by mass), the matrix volume will be stable Aluminum oxide (Al2O3) 12.09 20.40
after cooling in high moisture condition Ferric oxide (Fe2O3) 1.44 6.10
(Rehsi& Garg, 1976). Calcium oxide (CaO) 60.73 0.92
Now, there are around 26 coal fired power Magnesium oxide (MgO) 0.86 1.50
stations in Vietnam. This number will increase Sodium oxide (Na2O) - 0.53
up to 43 in later 2020. As a result, FA - a by- Potassium oxide (K2O) - 3.30
product of coal fired power station - will reach Sulfur trioxide (SO3) - 0.13
to 109 million tons. Numerous FA requires wide Loss on ignition 0.84 10.7
landfill and it also causes seriously
environmental problem. Utilization of FA in Table 2. Properties of OPC and FA
construction is an effective way to solve these
problems. Nevertheless, most of studies on using Property OPCFA
3
FA as material just focus on the effect of FA on Specific density (g/cm ) 3.11 2.288
some properties of concrete or cement such as Bulk density (kg/m3) 973.2 982.1
workability, compressive strength, hydration Water content (%) - 0.4
temperature, water resistance, and corrosion Compressive strength (MPa)
resistance. Number of researches about effect of - 3 days 32.4 -
FA on OPC at high temperature is small. Some - 28 days 51.86 -
of them were conducted in Vietnam but FA used Standard water (%) 32.2 -
in these researches was in the North area (with Initial setting time (min) 110 -
low loss on ignition) (Phuong et al. 2018; Ninh,
Final setting time (min) 160 -
2018). Study about using FA in the South area to
On the sieve of 0.09mm (%) 0.97 17.5
improve characteristics of OPC at high
Soundness (mm) 0.22 -
temperature has been not found so far.
Strength activity index (%) - 89.8
Therefore, in this paper, FA in South Central
area (with high loss on ignition) was studied to Physico-mechanical properties of OPC
improve properties of OPC at temperature of satisfied the requirement ofTCVN 2682:2009
800oC so that it can be applied for mortar or standard. According toTCVN 10302:2014
concrete working at high temperature. standard, FA from Vinh Tan coal fired power
station was F class, so it can be used as an active
2. MATERIALS AND EXPERIMENTAL mineral admixture for mortar and concrete.
PROGRAMME
2.2 Experimental program
2.1 Materials
In this research, OPC was replaced by FA with
Binders used in this research include Song 0, 20, 25, 30 and 35% by mass (mixes were
GianhPC40 and FA from Vinh Tan coal fired signed as FA0, FA20, FA25, FA30 and FA35,
power station (Binh Thuan province). Chemical respectively). For each mix, standard water was

529
determined (conformed to TCVN 6017:2015 also decreased. However, fineness of FA is
standard). After that, 12 cubic specimens (for much more than that of OPC; therefore, FA
each mix) were made with the dimension of make setting time of binder longer.
222 cm. To measure compressive strength of
hardened paste at normal temperature, 6 3.2 Apparent density
specimens (among 12 specimens) were cured in Effect of FA content on apparent density of the
standard temperature of 27±2oC and air hardened binder paste at different temperature
moisture over 90%; the rest 6 specimens were levels was illustrated in Figure 1 and Figure 2.
cured in steam condition for 4 hours. To
determine properties of the specimens at high
temperature, after steam curing, all the
specimens were dried at temperature of 100C
for 24 hours. Then, each group sample of each
mix were separately heated in oven at
temperatures of 200, 400, 600 and 800C with
heating speed not faster than 5C/min, and
temperature at each level was maintained for 2
hours. After heating, they were cooled down in
the air to room temperature and their apparent
density and compressive strength were tested.
Physicochemical properties of the hardened Figure 1. Apparent density of the hardened binder
binder paste were measured by DTA, TGA and pastes at different temperature levels
XRD methods.

3. RESULTS AND DISCUSSION

3.1 Standard water and setting time of binders
Table 3 reports standard water (the amount of
water makes cement paste reach standard
consistency) and setting time of binders mixed
from OPC and FA.

Table 3. Standard water and setting time of

binders Figure 2. Comparison of apparent density of the
hardened binder pastes at different
Mix Standard water Setting time (min)
temperature levels to those at 100oC
(%) Initial Final
FA0 32.2 110 160 At the same temperature level, apparent
FA20 29.4 112 175 density of hardened paste decreased when FA
FA25 29.0 115 180 content increased (Fig. 1) because specific
FA30 28.8 124 189 density of FA smaller than that of OPC. At the
FA35 28.4 145 195
same FA content, apparent density of hardened
paste decreased when temperature increased.
According to Table 3, when FA content From 25oC to 400oC, apparent density
increased, standard water decreased but setting significantly reduced due to the vaporization of
time increased. OPC is a high active binder, so free water and physical water (Hertz, 2005;
it needs much water for hydration. When it was Hager, 2013) and their density at 400oC just
replaced by FA, active characteristic of OPC were around 92.4% to 93.8% of those at 100oC
reduced; thus, necessary water for hydration (Fig. 2).

530
 When temperature was over 400oC, At normal temperature, most of minerals in
chemical water in CH and CSH minerals (in FA were not hydrated; so, compressive
hardened binder paste) were decomposed strength of the hardened binder pastes
(Hager, 2013) leading to the reduction of decreased when FA content increased (Fig. 3).
apparent density to around 90% of those at When specimens were heated to 100÷4000C,
100oC (Fig. 2). the compressive strengths increased (Fig. 3). At
At 800oC, the density reduction of FA20, this temperature interval, free water and
FA25 specimens were smaller than that of FA0 physical water vapored leading to the shrinkage
specimen. It can be said that, with the of the matrix that made its density improve.
replacement of FA for OPC at around 20% to Furthermore, vapored free water would improve
25% (by mass), the resistance against loss of hydration of OPC (autoclave process) (Morsy et
al., 2012). At temperature of 200oC and 400oC,
weight at high temperature of OPC was
compressive strengths were around 105÷120%
improved the most significantly.
and 104÷116% of those at normal temperature,
3.3 Compressive strength respectively (Fig. 4).
At around 400oC to 600oC, the reduction of
Effect of FA content on compressive strength compressive strength of FA0 was the biggest
of the hardened binder pastes at different due to chemical water evaporation of CH, CSH
temperature levels is showed in Figure 3 and and sulfoaluminate hydrates. Decomposition of
Figure 4. CH to create free CaO and then, free CaO
reacted with moisture during cooling was the
reason of micro cracks and volume change of
the matrix (Hager, 2013), (Seleem, 2011).
When OPC was partly replaced by FA, the
reduction of their compressive strength was
smaller than that of FA0 due to pozzolanic
reaction (Tanyildizi & Coskun, 2008).
At temperature of 800oC, compressive
strength of FA0 was the smallest because of
decomposition of CSH second phase to form β-
C2S (Hager, 2013). It reduced to 37.5% of that in
Figure 3. Compressive strength of the hardened 100oC. The reduction of strength of FA25 was
binder pastes at different temperature the smallest. Compressive strength of FA25
levels decreased to 59% of that in 100oC. Clearly, 25%
replacement of OPC by FA (with high loss on
ignition) improved compressive strength at
temperature of 800oC the most significantly. The
same results were found in some researches
using FA with low loss on ignition (Phuong et
al., 2018; Rehsi, 1976). It means that FA can
prevent the loss of compressive strength of
binder due to pozzolanic reaction between SiO2
in FA and Ca(OH)2 in hardened cement paste
(Aydın & Baradan, 2007; Xu et al., 2003).
Moreover, new gels, whose compressive strength
are approximately 2 to 3 times higher than that of
Figure 4. Comparison of compressive of the CSH, were created. Therefore, binder, mortar or
hardened binder pastes at different concrete with FA can subjected to high
temperature levels to those at 100oC temperature (Xu et al., 2001; Poon et al., 2001).

531
3.4 Physicochemical analyses 3.4.2 X-ray diffraction analysis
3.4.1 Thermal analysis X-ray diffraction (XRD) analysis can
determine the change in composition of
Differential thermal analysis (DTA) and hydrates in hardened binder paste at high
thermogravimetric analysis (TGA) was temperature. Figure 6 presents XRD result of
effectively in determining mass change of FA25 specimen heated at 800oC. At
hydrates in hardened binder pastes at high temperature of 800oC, there were some
temperature. Figure 5 shows DTA and TGA minerals in hardened binder paste such as SiO2,
analyses of FA25 specimen at 900oC. When belite (C2S), calcium oxide (CaO) and alite
the specimens were heated from 25oC to (C3S). SiO2 appears at picks: dAo(1.8189;
900oC, there were thermal effects and mass 2.4045; 2.7535; 3.3564).C2S appears atpicks:
changes at the peaks of 113.1oC; 473.2oC; dAo(1.5539;1.5937; 1.5942; 1.6932; 1.7714;
564.8oC; 678.1oC; 783.9oC and 805.9oC. At 1.7613; 1.7959; 1.8997; 1.9343; 1.9725;
temperature of 113.1oC, a thermal effect due 1.9730; 2.0475; 2.0489; 2.0952; 2.1879;
to the loss of free water and physical water 2.2364; 2.2850; 2.3647; 2.7077; 2.7306;
occurred; and specimen’s mass reduced 2.7535; 2.7612; 2.2860; 2.9099; 3.1387;
3.49%. At temperature of 473.2oC and 3.3570; 3.8448; 4.6446;).CaO appears atpicksd
564.8oC, thermal effects because of the loss of Ao(1.6864; 4.2819).C3S appears atpicks:
chemical water appeared; and the mass dAo(1.5483;1.6328; 1.7613; 1.9345; 1.9732;
decreased 3.17% and 1.22%, respectively. 2.187; 2.2848; 2.6067; 2.7530; 2.941; 3.0299;
From 678.1oC ÷ 783.9oC and 805.9oC ÷ 3.5192; 4.2763). It can be seen that CH and
900oC, heat recovery effects due to carbonate CSH disappeared at this temperature. Strength
decomposition can be observed; and the mass increase of the specimen at 800oC might be due
lost 2.7% at 700oC and 6.15% at 900oC. The to dense structure of the matrix when its pores
huge loss of mass at 473.2oC implied that this was filled up by molten phases (Aydın &
temperature was the conflict point of hydrate Baradan, 2007). The existence of CaO
decompositions, which was the reason for intimated that mineral components of FA and
strength loss at high temperature. CaO did not react thoroughly.

DTA /(uV/mg) TG /%
 exo Peak: 678.1 °C
100
Peak: 564.8 °C
Mass Change: -3.49 %
Peak: 805.9 °C
0.8 98

[1]

96
Peak: 783.9 °C
0.6 Mass Change: -3.17 %

94
Peak: 473.2 °C Mass Change: -1.22 %

0.4 Mass Change: -2.70 %

92

Mass Change: 0.44 %

0.2 90

88

0
Mass Change: -6.15 %

86

-0.2 [1]
Peak: 113.1 °C 84

100 200 300 400 500 600 700 800 900

Temperature /°C

Figure 5. DTA andTGA analyses of FA25 specimen

532
Figure 6. XRD result of FA25 specimen

4. CONCLUSIONS 3. DTA, TGA and XRD results indicated the

minerals in FA25 specimens at temperature of
An experimental study was performed to 800oC as well the loss of mass and
evaluate the improvement of OPC properties at corresponding thermal effects.
high temperature when OPC was partly
replaced by FA with high loss on ignition. ACKNOWLEDGEMENT
Based on the obtained results, some
conclusions can be drawn as following: This research was funded by The University
1. When heated from 100oC to 400oC, of Danang – University of Science and
apparent density significantly reduced but Technology (DUT) under the project numbered
compressive strength increased; and specimens T 2019-02-12.
with FA had apparent density and compressive
strength larger than that without FA. From 5. REFERENCES
400oC to 600oC, apparent density decreased Aydın, S., & Baradan, B. 2007. Effect of pumice
and compressive strength also highly reduced. and fly ash incorporation on high temperature
From 600oC to 800oC, the loss of apparent resistance of cement based mortars. Cement
density and compressive strength continued; and Concrete Research, 37(6), 988–995.
and the loss of OPC specimens was larger than Dias, W.P.S., Khoury, G.A., Sullivan, P.J.E. 1990.
those of specimens with FA. Mechanical properties of hardened cement
2. At 800oC, compressive strength of FA25 paste exposed to temperatures up to 700oC.
specimen was 39 MPa (around 59% of its ACI Mater J, 87(2), 160–165.
strength at 100oC); which was 1.4 times to that Grainger, B.N. 1980. Concrete at high
of FA0 specimen. Its apparent density was temperatures. UK: Central Electricity
1790.7 kg/m3 (around 92% of that at 100oC). It Research Laboratories.
Hager, I. 2013. Behaviour of cement concrete at
can be said that if Vinh Tan FA (with high loss
high temperature. Bulletin of the Polish
on ignition) was partly replaced for OPC of Academy of Sciences: Technical Sciences,
25% in mass, it could improve some properties 61(1), 145–154.
of OPC at high temperature. Therefore, Vinh Hertz, K.D. 1992. Investigations on silica fume
Tan FA was suitable for heat resistant mortar concretes at elevated temperatures. ACI Mater
and concrete. J, 89(4), 345–347.

533
Hertz, K.D. 2005. Concrete strength for fire safety Seleem, H.E-D., Rashad, A.M., El-Sokay Tarek.
design. Mag. Concr. Res, 57(8), 445–453. 2011. Effect of elevated temperature on
Hlystov, A. I., Shirokov, V. A., & Vlasov, A. V. physic-mechanical properties of blended
2015. Efficiency Improvement of Heat- cement concrete. Construction and Building
resistant Concrete through the Use of Sludge Materials, 25, 1009–1017.
Technogenic Raw Material. Procedia Tanyildizi, H., & Coskun, A. 2008. The effect of
Engineering, 111, 290–296. high temperature on compressive strength and
Morsy, M. S., Al-Salloum, Y. A., Abbas, H., & splitting tensile strength of structural
Alsayed, S. H. 2012. Behavior of blended lightweight concrete containing fly ash.
cement mortars containing nano-metakaolin at Construction and Building Materials, 22(11),
elevated temperatures. Construction and 2269–2275.
Building Materials, 35, 900–905. TCVN 2682:2009. Portland cement – Technical
Nekrasov, K.D., Tarasova, A.P. 1969. Heat - requirements. Ministry of Science and
resistant concrete using portland cement. Technology.
Construction Publisher. Moscow. [in Russian]. TCVN 4030:2003. Cements - Test method for
Ninh, C.T.H. 2018. Research on manufacturing determination of fineness. Ministry of Science
technology and construction technology of and Technology.
fireproof - insulating lightweight concrete for TCVN 6016:2011. Cements - Test methods -
civil and industrial buildings (Doctoral Determination of strength. Ministry of
dissertation “Building and Industrial Science and Tecnology.
Construction Engineering"). NUCE. Ha Noi. TCVN 6017:2015. Cements - Test methods -
[in Vietnamese]. Determination of setting time and soundness.
Phuong, D.T., Tri, L.V., Duc, V.M., Hoa, N.N. Ministry of Science and Technology.
2018. Heat - resistant binder using fly ash. TCVN 10302:2014. Activity admixture - Fly ash
Journal of Science and Technology,the for concrete, moter and cement. Ministry of
University of Danang5(126), 51–55. [in Science and Technology.
Vietnamese]. Wang, H.Y. 2008. The effects of elevted
Poon, C.S., Azhar, S., Anson, M., Wang, Y.L. temperature on cement paste containing
2001. Comparison of the strength and GGBFS. Cem and Conc Compo, 30, 992–999.
durability performance of normal-and high- Xu, Y., Wong, Y. L., Poon, C. S., & Anson, M.
strength pozzolanic concretes at elevated 2003. Influence of PFA on cracking of
temperatures. Cement and Concrete Research, concrete and cement paste after exposure to
31, 1291–1300. high temperatures. Cement and Concrete
Rehsi, S.S., Garg, S.K. 1976. Heat resistance of Research, 33(12), 2009–2016.
Portland fly ash cement. Cement, 4(2), 14–16. Xu, Y., Wong, Y.L., Poon, C.S., Anson, M. 2001.
Remnev, V.V. 1996. Heat-resistant properties of Impact of high temperature on PFA concrete.
cement stone with finely milled refractory Cement and Concrete Research, 31, 1065–1073.
additives. Refractories and Industrial
Ceramics, 37, 10–11.

534
The 3rd Int. Conf. on Transport Infrastructure & Sustainable Development (TISDIC 2019). Construction Publishing House. ISBN: 978-604-82-2893-4.

Estimation of distribution of chloride ions in cover

concrete using electromagnetic wave and X-ray
fluorescence

Misaki Nakamura1,, Hitoshi Ito2, and Toshiaki Mizobuchi3

1
GraduateStudent, Department of Civil and Environmental Engineering, Hosei University, Tokyo, Japan
2
Doctal Student, Department of Civil and Environmental Engineering, Hosei University, Tokyo, Japan
3
Professor, Department of Civil and Environmental Engineering, Hosei University, Tokyo, Japan

Corresponding author, E-mail: [email protected]

ABSTRACT: Although concrete structures progress in deteriorationcaused by various factors, it is

possible to delay their deterioration by appropriate maintenance management. In the management of
maintenance, the concept of preventive maintenance is important and it is necessary to conduct
inspections from the early stage of deterioration and to take measures according to the state of
deterioration. The inspection method may damage the concrete structures among the inspection methods
carried out for investigations of deterioration. On the other hand, the nondestructive inspection is a
method that can inspect concrete structures without damage as much as possible.In this study, authors
focused on salt damage among various deterioration factors of concrete and investigatedon methods to
estimate the penetration of chloride ions to concrete without damaging the structure as much as
possible.In this paper, in order to evaluate the distribution of content of chloride ions during the latency
stage and the progression stage before corrosion induced cracks appear on the concrete surface, the
method to estimate the distribution of content of chloride ions in the cover concrete combining both
electromagnetic waves and X- ray fluorescent is proposed.
Keywords: electromagnetic wave, X-ray fluorescence, chloride ion, Non destructive test, latency stage

1. INTRODUCTION extend their service life. In order to carry out

appropriate maintenance, it is important to
Concrete structures deteriorate due to various conductthe inspections regularly and to take
factors such as environmental conditions and measures in the early stage of deterioration from
aging. In Japan, many infrastructures were built the viewpoint of preventive maintenance.
during the high economic growth period of the Among various causes of deterioration in
1950s to 1970 sand the damage caused by their concrete structures, it is important to detect salt
aging is beginning to occur all at once.As the damage as early as possible and to take
damage caused by the aging may cause major measures appropriatelybecause deterioration
accidents involving human life, it is necessary to after corrosion of rebar by salt damage
carry out to investigation, repair, strengthening progresses rapidly. If content of chloride
or rebuilding for degraded concrete structures. ionsexceedthe limit value of corrosion
However, it is impossible to rebuild a huge occurrence in neighborhood of rebar, expansion
amount of decrepit infrastructure stock at once. caused by corrosion of rebar occurs and cracking
Therefore, it is necessary to take measuresto is generated by the corrosion expansion in the

535
The 3rd Int. Conf. on Transport Infrastructure & Sustainable Development (TISDIC 2019). Construction Publishing House. ISBN: 978-604-82-2893-4.

concrete. After that, peeling in the cover In this paper, as the method to estimate the
concrete and cracking along the reinforcing bar penetration of chloride ionseasily for the
occur on the concrete surface.The peeling and purpose of early detection of deterioration
cracking caused by the salt damage do not show caused by salt damage, the method to estimate
any major change until they reach near the the distribution of chloride ions by combining
surface of the reinforced concrete structure. electromagnetic waves and fluorescent X-ray
In general, in order to evaluate the analysis is proposed.
progression due to progressive chloride
incursion in the cover concrete, from the results 2. ESTIMATION OF CONTENT OF
of chemical analysis such as potentiometric CHLORIDE IONS USING PORTABLE X-
titration using samples by cutting the core RAY FLUORESCENCE ANALYZER
drawn out from the target structure, it is possible
to investigate the content of chloride ions at that At present, the chemical analysis using
time near the rebar.However, this test method potentiometric titration by samples collected
has the problem of damaging the concrete with core drills at the surveyed fieldsis the
structure.In addition, because repeated testing at method that is carried out generally to evaluate
nearby locations is difficult, history of aging the content of chloride ions in concrete.
deterioration can’t be acquired. However, since the method of drawn cores has
On the other hand, the nondestructive problems such as that it takes time for sampling
inspection method is a survey to observe the and cutting, crushing and pretreatment for
change of the deterioration with the progress of chemical analysis after sampling take time and
time at the same place without damaging the cost, it is difficult that the estimate of the
structure as much as possible. Nondestructive content of chloride ions cannot be performed in
inspectionis possible to determine the content a short time. Therefore, using the combination
of chloride ions inside the reinforced concrete method of sampling by small diameter drill
structure. Therefore, it is possible to detect machine and analysis due to the fluorescent X-
the progress of deterioration due to salt ray analyzer, which has many past research
damage at an early stageand it will be possible results and its usefulness has been confirmed, it
to take the countermeasure. was carried out toestimate the distribution of
In order to solve these subjects, as an content of chloride ions at the survey fields.
estimation evaluation method of the content of When fluorescent X-rayanalyzeris used for
chloride ionsusing electromagnetic wave, the estimation of content of chloride ions, the
influence coefficients of various factors chlorine content ratio (ppm) in concrete is
considered to affect the attenuation of the measured, and the valuesare converted to
reflected waveform of the electromagnetic content of chloride ions (kg /m³) using a
wave were determined from results of the calibration curve for evaluation. Previous studies
laboratory tests.Furthermore, it is necessary to (Jyunichiro et al., 2018) have shown a single
estimate the content of chloride ions at the calibration curve generated from 861 data taken
rebar position as accurately as possible in order from eight sites. This equation has a high
to evaluate the progress of deterioration due to determination coefficient of 0.974 and is
salt damage in the early stage. relatively accurate. On the other hand, it is
Therefore, as a method for estimating the possible to obtain a more accurate calibration
distribution of content of chloride ions, the curve by preparing a calibration curve for each
authorsmeasured with the handy type site. Though there are various conditions such as
fluorescent X-ray analyzer using drilling mix proportions, environment, etc., in each site,
powders with the small diameter drill machine in this study, the investigation was conducted
from the structure to be surveyed and it was focusing on the conditions of mix proportion
conducted to study the method for estimating the and the influence of conditions of mix
distribution of chloride ions easily in the field. proportion on calibration curve was investigated.

536
The 3rd Int. Conf. on Transport Infrastructure & Sustainable Development (TISDIC 2019). Construction Publishing House. ISBN: 978-604-82-2893-4.

By using the sampling method by the small 3. INFLUENCE OF SAND-TOTAL

diameter drill proposed by this method, the AGGREGATE RATIO ON
drilling powders can be used for analysis using CALIBRATION CURVEINX-RAY
the handy type fluorescent X-ray analyzer FLUORESCENCE METHOD
without modifying them. From the past research
(Akio at al., 2009), it was confirmed that the In order to investigate the influence of the
analysis accuracy was improved by using the amount of aggregate, the samplesthat changed
drilling powders without processing rather than sand-total aggregate ratio and the content of
milling them. In addition, in order to facilitate the chloride ionswere prepared and the content of
sampling, special equipment that can be chloride ion was measured by the fluorescent X-
removable from the dust suction of the small ray analyserand the potentiometric titration. A
diameter drill machine and can sample directly total of 21 cases with 3 of sand-total aggregate
into the test container. Figure 1 shows the small ratio and 7 of sodium chloride concentration as
diameter drill machine and special equipment. parameters were adopted as study cases. The
Figure 2 shows the test container. When the drill mix proportions of each caseare shown in the
diameter is φ22 mm and the sampling depth is 10 Table 1. The sample was put in a dedicated
mm, one sample is about 9 g. In addition, the container shown in Figure 2 and the chlorine
sampling volume is reduced by about 95% content ratio was measured using the portable
compared to using core drill and the damage to fluorescent X-ray analyser shown in Figure 3
the structure is significantly reduced. for the fluorescent X-ray analysis. In addition,
the content of chloride ions obtainedby
potentiometric titration on the same sample as
that used for the fluorescent X-ray analysiswas
compared with the content of chloride ions of
the fluorescent X-ray analysis. Figure 4 shows
the relationship between the chloride ion
concentration (kg/m³) measured by
potentiometric titration and the chlorine content
ratio (ppm) measured by X-ray fluorescence.
From Figure 4, the measured value of X-ray
fluorescence was different due to the change of
the sand-total aggregate ratio to the same
chloride ion amount by potentiometric titration
method. Table 2 shows the mortar volume and
Figure 1. The small diameter drill machine and coarse aggregate volume for each sand-total
special equipment
aggregate ratio and Figure 5 shows the cross-
sectional photograph of the sample at each
sand-total aggregate ratio. From Table 2 and
Figure 4, the results of measurement by X-ray
fluorescence show a tendency that the
percentage of chlorine content (ppm) becomes
smaller as the mortar volume is smaller.
From these results, it was found that the
gradients and intercept of the calibration curve
were affected by the sand-total aggregate ratio
or mortar volume. Next, in order to evaluate
whether the main influence factor to the
calibration curve is the sand-total aggregate
Figure 2. The test container ratio or the mortar volume, when the mortar

537
The 3rd Int. Conf. on Transport Infrastructure & Sustainable Development (TISDIC 2019). Construction Publishing House. ISBN: 978-604-82-2893-4.

volume was constant, the influence of change

of the sand-total aggregate ratio on the
calibration curve was investigated.
As shown in Table 3, in this study, the
sand-total aggregate ratio was changed so that
the mortar volume was constant. The case s/a=35% s/a=45% s/a=50%
study wascarried out 12 cases with 3 levels of Figure 5. Sample at each sand-total aggregate ratio
water-cement ratio and 4 levels of sodium
chloride concentration. Figure 6 shows the
relationship between the content of chloride
ions (kg/m³) measured by potentiometric
titration and the content of chlorine ratio
(ppm) measured by X-ray fluorescence. As
shown in Figure 6, when mortar volume was
constant, it was not confirmed that the change
in fine aggregate rate affects the calibration
curve.From the above, it was confirmed that
the calibration curve was affected by the
mortar volume.

Figure 6. Comparison of chlorine content rate and

chloride ion concentration (mortar
volume constant)

Since the gradients and intercept of the

calibration curve change with mortar volume,
the calibration curve with mortar volume as a
variable was formulated. Figure 7 shows the
relationship between mortar volume and each
influence coefficient ( and ). Figure 8
shows the results of comparison between the
content of chloride ions when the influence
Figure 3. The fluorescent X-ray analysis factor to the mortar volume is considered and
the content of chloride ions when the mortar
volume is not considered. As shown in Figure
8, the accuracy of estimation of the content of
chloride ions is improved by considering the
mortar volume.
f  x    103 x   (1)

  6.4Vm 103  5.94

  7.9Vm 103  4.55

f(x) : amount of NaCl (kg/m3)
x : chlorine content ratio (PPM)
Figure 4. Comparison of chlorine content rate and
chloride ion concentration Vm : mortal volume (l/m3)

538
The 3rd Int. Conf. on Transport Infrastructure & Sustainable Development (TISDIC 2019). Construction Publishing House. ISBN: 978-604-82-2893-4.

Figure 7. The relationship between mortar volume Figure 8. Comparison of conversion results due to
and each influence coefficient differences in calibration curve

Table 1. Blending with sand-total aggregate ratio as a parameter

Target Target Unit amount(kg/m³)
W/C s/a NaCl Gmax
Air SL
(%) (%) (kg/m³) (cm) W C S G AE AEWR
(%) (%)
35.0 624 1199 0.0160
0,3,6,9,
50 45.0 20 4.5 8 160 320 820 1014 0.0128 3.2
12,15,18
50.0 891 922 0.0112

Table 2. Mortar volume and coarse aggregate volume

W/C (%) s/a (%) Mortar volume (L/m³) Aggregate volume (L/m³)

45 42.1 606
50 43.0 605 397
55 43.7 606

Table 3. Blending with sand-total aggregate ratio as a parameter (mortar volume constancy)
Target Target Unit amount(kg/m³)
W/C s/a NaCl Gmax
Air SL
(%) (%) (kg/m³) (cm) W C S G AE AEWR
(%) (%)
45 42.1 356 740 0.0087 3.6
50 43.0 0,3,6,9 20 4.5 8 160 320 769 1055 0.0096 3.2
55 43.7 291 793 0.0107 2.9

Table 4. Mortar volume and coarse aggregate volume

W/C (%) s/a (%) Mortar volume (L/m³) Aggregate volume (L/m³)

45 42.1 606
50 43.0 605 397
55 43.7 606

539
The 3rd Int. Conf. on Transport Infrastructure & Sustainable Development (TISDIC 2019). Construction Publishing House. ISBN: 978-604-82-2893-4.

4. ESTIMATION OF CONTENT OF close to the location where chloride ions

CHLORIDE IONS BY COMBINATION OF penetrate and are damaged, it was carried out to
ELCTROMAGNETIC WAVES METHOD collect by the small diameter drill machine and
AND X-RAY FLUORESCENCE METHOD to estimate the amount of chloride ion by X-ray
fluorescence analysis. The estimation results by
When an electrolyte such as sodium chloride is X-ray fluorescence analysis and the apparent
present in concrete, by focusing on the difference diffusion coefficients calculated based on the
in the attenuation of the reflection waveform of estimation results are shown in Figure10. As for
the electromagnetic wave from the rebar by the the position of the maximum content of chloride
difference in the content of chloride ions, it is ions, since the migration to the inside of the
possible to estimate the content of chloride ions concrete due to the carbonation etc. was not
in the reinforced concrete structure using the confirmed, in this study, the position of the
electromagnetic wave and authors have been maximum content of chloride ions was used as
researching and developing for implementation the concrete surface. Figure 11 shows the
of the method (Jyunichiro et al., 2015). However, mapping based on the estimation results of the
since the estimation of the content of chloride content of chloride ions at the position of rebar
ions by electromagnetic waves utilizes the from the 318 measurement points using the
attenuation characteristics of the electromagnetic calculated apparent diffusion coefficient.
waves, the estimated content of chloride ions From Figure. 11, the area where the content of
itself is the average content of chloride ions from chloride ions at the position of rebar is less than
the concrete surface to the position of rebar. 1.2 (kg/m³) is about 50% and the area less than
Therefore, as a method for estimating the 2.4 (kg/m³) is about 80%. Assuming that the
distribution of content of chloride ions, the limit value of corrosion initiation is in the range
authorsmeasured with the handy type of 1.2 (kg/m³) to 2.4 (kg/m³), although 50 to 80%
fluorescent X-ray analyzer using drilling of the bridge decks surveyed have been in use for
powders with the small diameter drill machine about 60 years, the content of chloride ions at the
from the structure to be surveyed and it was position of rebar may not exceed the limit value
conducted to study the method for estimating the of corrosion initiation. Although it may be
distribution of chloride ions easily in the field. judged that the target bridge is deteriorating from
In this study, an application example of the damage due to salt damage on the sea side, the
mapping of content of chloride ions in visual inspection of appearance of structure and
reinforced concrete combining electromagnetic years of service, from estimation results of this
waves and fluorescent X-rays to existing survey, it is estimated that corrosion of rebar
structures is shown below. does not occur in many parts of the target bridge.
The surveyed structure is the bridge
constructed at a distance of about 50m from the
coast and it has been 57 years since its service.
The survey point is the bottom of reinforced
concrete floor slab. As shown in Figure 9, the
peeling of concrete has occurred in the range of
about 1 m from the end facing the coast, and the
rebar is exposed in the exfoliated concrete. On
the other hand, almost no cracks were observed
inside the floor slab and no peeling concrete
was almost observed. As shown in Figure. 9,
measurements were carried out using
electromagnetic waves around the inside of the Figure 10. The estimated diffusion coefficients
undamaged floor slab and at the point relatively calculated based on the estimation results

540
The 3rd Int. Conf. on Transport Infrastructure & Sustainable Development (TISDIC 2019). Construction Publishing House. ISBN: 978-604-82-2893-4.

Mountain side

Mountain side 10(kg/m3)

9
8
7
6
5
4
3.5
3
2.5

12150
2
1.8
1.6
range of measurement
1.4
1.2
1
Bridge axis direction 0.8
0.6
0.4
0.2
試料採取位置
Sampling position
0

Sampling position

Rebar exposed Rebar

鉄筋露出部 exposed area
area
1850

630 浮き部分area
Peeling
Peeling area
Rebar exposed area
鉄筋露出部
Rebar exposed area

Sea side Sea side

Figure 9. The structure to be surveyed this time Figure 11. The result of mapping chloride ion amount

5. CONCLUSIONS covered by electromagnetic waves and the

content of chloride ions estimated by X-ray
In this paper, for the purpose of early evaluating fluorescence analysis become equivalent
the progress of deterioration due to salt damage fromthe results of the average content of
of reinforced concrete structures, a method was chloride ionsin the cover concrete by
proposed for the penetration of content of electromagnetic waves and the distribution of
chloride ions into concrete and the estimation of chloride ions obtained by portable X-ray
content of chloride ions at the position of rebar fluorescence analysis using drilling samples by
simply in the survey field. The method proposed small diameter drill machine,
in this paperis to estimate the content of chloride From the results of applying the mapping of
ions at the position of rebar by changing the the content of chloride ions at the position of
apparent surface content of chloride ions so rebar in reinforced concrete combining
that the total content of chloride ions in the area electromagnetic waves and fluorescent X-ray

541
analysis to the existing structures, in the stage Akio, W. &Seiji, H &Minobu, A.2009. Chloride
where almost no deformation is observed on ion concentration measurement of drilled
the concrete surface, it is possible to find out sample by X-ray fluorescence method.
the applicability of this method as an Concrete Engineering Annual Proceedings, 2-
estimation method for evaluating the progress 4 September, Fukuoka, Japan, Vol.31,
No.1,1987-1992.
of deterioration.
Jyunichiro, N. &Mami, U &Jyunichi, A &
Toshiaki, M. 2015. A study on estimation
6. REFERENCES method of salinity in concrete using
Jyunichiro, N. &Ruiko,W.& electromagnetic wave waveform. Concrete
Toshiaki,M.2018.Study on actual structure Engineering Annual Proceedings, 16-18
sampling by micro fracture and estimation of September Okayama, Japan, Vol.37, No.1,
chloride ion content in concrete structure by 1675-1680.
X-ray fluorescence analysis. Concrete
EngineeringAnnual Proceedings, Vol.40,
No.1, 1677-1682.

542
The 3rd Int. Conf. on Transport Infrastructure & Sustainable Development (TISDIC 2019). Construction Publishing House. ISBN: 978-604-82-2893-4.

Recycled rock as coarse aggregate of shotcrete for tunnelling

Tran Thu Hang1,, Nguyen Dong Tran Duc2

1
University of Transport and Communications, Hanoi, Vietnam
2
TEDI Incorporation, Hanoi, Vietnam

Corresponding author, E-mail: [email protected]

ABSTRACT: Shotcrete provides the initial support for tunnels, especially in NATM technology. The
utilization of the blasted rock after the tunnel’s excavation as the recycled coarse aggregate for shotcrete
is an action toward the sustainability in tunnel engineering. It helps to reduce the demand of a non-
renewable construction resource, and to mitigate the environmental burdens caused by the discharged
muck. In this paper, the substitution of the new coarse aggregate by the recycled crushed muck in a road
tunnel project in Vietnam was presented. Several shotcretes using recycled rock with different mixtures
and sources of materials were analysed and the comparison with the conventional ones were made. The
goal is to understand the possibility of using tunnel muck for shotcrete in tunnelling and to contribute to
the sustainable development practice in tunnel engineering.
Keywords: Tunnel muck, recycled aggregate, shotcrete, NATM, sustainability

1. INTRODUCTION been shared for the vast number of construction

With the expansion of the transport projects in the country (Prime minister of
infrastructure for the calling of the social- Vietnam, 2020). This fact leads to the short
economical development, the number of road supply of the raw rock as construction material
tunnels is increased quickly in Vietnam, in the road tunnel projects whereas the huge
recently. All of them have been built by NATM amount of tunnel muck wasted at dump areas.
technology with the application of modern Hence, the ideal of recycling the tunnel muck
construction equipment and techniques. for replacing a part of the raw material has been
Shotcrete provides the initial and permanent proposed. In some recent projects, the recycled
support for the tunnel structure. Since the first rock has been used for variouspurposes for in-
introduction in 1910, it has been widely used project use (Tran & Le, 2017). In this paper, the
parallel with and instead of the conventional muck utilization as the coarse aggregate for
concrete in construction engineering. In the shotcrete in a recent road tunnel project in
shotcrete mixture, the coarse aggregate is added Vietnam was studied. With the same blasted
for: a) Decrease of cementitious usage; b) rock aggregate, several shotcretes with different
Shrinkage reduction; c) Improvement of mixtures were designed and analysed, the
pumpability; and d) Advance of in-place density comparison with a new rock shotcrete was
(ACI, 2016). It is said to bring the better made. The goal is to understand the possibility
economical and sustainable gains. In Vietnam, of using tunnel muck for shotcrete in tunnelling
following the state plan up to the year 2020, and to contribute to the sustainable development
about 40.5 billion cubic meter rock resource has practice in tunnel engineering.

543
2. SHOTCRETE MIXTURE DESIGN USING The water absorption of the recycle rock was
BLASTED ROCK AS COARSE AGGRGATE varied from 0.96% to 1.58% in accordance with
The excavation zone of the studied tunnel was TCVN 7572-5:2006 standard (IBST(b), 2006).
dominated formed by the fractured low- Hence, the aggregate was said to be normal
weathered granite of Paleozoic era. Excavated by absorptive, well complied with international
explosive, the tunnel muck was considered to be regulations(Anchor, 1988)(EFNARC, 1993),
in the same quality with the tunnel’s bedrock whereas no requirement has been given in
except the crushing condition and the block Vietnamese standards. Normally, the water
dimension, and to be considerable good as new absorption is one of the weak points of the
rock exploited from nearby quarries which were recycled aggregates versus the virgin ones. The
stayed in the same rock medium. The recycling popular solutions are an extra water proportion
process of the blasted rock comprised of the in the mixture, the pre-soak (or water
selection (based on the surface cleanness, block injection)of rock in water before the rock
size, rock type), the crushing (for the target grain recycling process, etc.
size distribution with Dmax=10mm), and the The amount requirement of the coarse
screening (rejection of non conforming grains or aggregate in the shotcrete mixture varies
batches). Once the aimed aggregate had been between countries. For example, it is about 30
successfully obtained, the quality control to 45% of the combined aggregate in Japan
following the project specifications was (since the fine aggregate ratio is generally from
executed. Given that no specific standard for the 55 to 70% (JSCE Working group for Mountain
shotcrete materials existed, the current tunnels, 2006)) whereas it is recommended not
Vietnamese corresponding code for the greater than 30% of the combined one in
conventional concrete was used. Since the America (ACI, 2016). The recycled rock
crushing value in cylinder of the muck aggregate aggregate was combined with the new river
varied from 5.10% (in dry state) to 6.50% (in sand with the mixing ratio of 35:65 percentage
saturation state), the deductive grade of the by weight to form the combined aggregate. The
crushed aggregate was 140.00 following TCVN grain size analysis was carried out for the
7570:2006 standard(IBST(a), 2006).In Table 1, mixture and the requirements of the current
the typical characteristics of the studied recycled
American specifications(ACI, 2016) (ASTM,
coarse aggregate and the equivalent requirements
2013) were checked. The favourable result has
for the 30MPa concrete were presented.
been attained and presented on the grain size
Table 1. Typical characteristics of the studied distribution curves in Figure 1.
recycled coarse aggregate and the equivalent
requirements for the 30MPa concrete
Parameter (Unit) Obtained Requirement
value (IBST, 2006)
Specific gravity (g/cm3) 2.72 Not given
Compression strength 107.00 ÷ 20.00 ÷
(MPa) 114.00 140.00
Crushing value in 5.10 ÷ 0.00 ÷ 34.00
cylinder (% by weight) 6.50
Grade of aggregate 140.00 60.00 Figure 1. Grain size distribution curves of the
studied blasted rock, sand and combined
Softening ratio 0.94 Not given
aggregates in comparison with the standard
Content of dust, mud, 0.32  2.00 boundary for shotcrete (ACI, 2016)(ASTM, 2013)
clay (% by weight)
Content of elongated and 25.70  35.00 Aimed at the replacement of the new rock
flat particles (% by weight) aggregate by the recycled one, all shotcrete’s
Los Angeles abrasion 18.10 < 50.00 componentswere tested and quality controlled
(% by weight) as for a conventional shotcrete mixture. At the

544
end, four wet-mix shotcrete mixtures utilizing mentioned in Table 2, their estimated
the studied blasted coarse aggregate had been compressive strengths were obtained in Table
designed and the details were listed in Table 2. 3. They were all surpassed the designed
A wet-mix shotcrete mixture using new rock in compressive strength R = 30Mpa, the recycled
the same project was also presented for the mixtures were even possible in higher strength
comparison. than the new one.

Table 2. Five 30MPa wet-mix shotcrete Table 3. Predicted compressive strength of

mixtures(Unit: kilogram per 1m3 of shotcrete) 5 shotcrete mixtures
Shotcrete mixture Recycled 1 Recycled 2 Shotcrete mixture Predicted compressive
proportion strength (MPa)
Coarse aggregate 477.12 515.97 Recycled 1, 2, 3, 4 36.26
Dmax=10mm New 35.55
Fine aggregate 903.81 971.26
The test panels in accordance with ASTM
PC40 cement 560.00 510.00
C42/C42M (ASTM, 2018) of 5 mixtures
Water 224.00 204.00 wereformed. The shotcrete cores were taken and
Accelerator 44.80 40.80 the uniaxial compression tests following TCVN
Superplasticizer 8.59 7.82 3118-1993 standard (MOST(b), 1993) were
carried out. The obtained value of compressive
strengths at different ages (1, 3, 7 and 28 days)
Shotcrete mixture Recycled Recycled New been surveyed and the collation with the
proportion 3 4
minimal requirement given in the project
Coarse aggregate specifications was executed. The graph showing
523.00 562.59 618.00
Dmax=10mm the development trend of the compressive
Fine aggregate 984.04 1052.21 894.00 strength of 5 mixtures was in Figure 2.
PC40 cement 510.00 450.00 520.00
Water 204.00 180.00 213.00
Accelerator 30.60 36.00 31.20
Superplasticizer 7.82 6.90 6.86

3. DETERMINATION OF THE
RECOMMENDED RECYCLED ROCK
SHOTCRETE MIXTURE
Among the five shotcrete mixtures that had Figure 2. The development of the compressive
been designed in the previous part, the strength of 5 shotcretesof test panels
following ASTM C42/C42M
recommended one was selected based on two
criteria: (a) Qualification of the shotcrete’s If only the designed compressive strength of
compressive strength; (b) CO2 emission from the shotcrete (at 28 days) is considered, all
each cubic meter of shotcrete. Details were mixtures are found to be satisfied. However, if
described as follows. the compressive strength variability is
3.1 Variability of compressive strength examined, Recycled 4 mixture is concluded as
nonconforming since its early strength (before
Firstly, the predicted shotcrete strength was 10 days) was fallen under the margin. It was
studied. Taking 85% of the calculated noted that the mixture contained the minimal
compressive strength using the Bolomey cement content but the maximal combined
formulas (Brito, et al., 2018) of 5 shotcretes aggregate among the 5 mixtures. This fact can

545
be understood as the main concern for the 3.3 Commentary
failure, beside the other possible effects: the
After the examination on the compressive
non favourable curing condition, the improper
strength development of 5 shotcrete mixtures,
addition of accelerator admixture, the worker
only two first were qualified. Among them,
craftsmanship (Zhang, 2014). When the
because Recycled 2 used less water and chemical
requirement of using 1.2 times the required
admixtures (91.07% of Recycled 1), and created
minimum compressive strength was applied for
the smaller CO2 emission (92.10% of Recycled
the average compressive strength of all cores,
1), it was concluded to be more sustainable. Its
Recycled 3 and Recycled 4 shotcrete mixture
carbon footprint was even lighter than of the new
would be rejected for not passing the required
rock mixture. It was hence recommended among
values. Hence, only the mixture design of
the 4 recycled mixtures in this study. This
Recycled 1 and Recycled 2 were continued to
determination was agreed with the tunnel
be interested.
contractor’s proposal of using Recycled 2
3.2 Carbon dioxide emission mixture for the actual application at the project.
In the scope of this study, attention was not
The CO2 emission of one cubic meter of paid to the workability and sprayability. Besides,
shotcrete was calculated as the product of the the slump of five fresh mixtures had been
amount of the mainshotcrete’s components examinedin accordance with TCVN 3106-
(cement, rock, sand)in the mixture and their 1993(MOST(a), 1993) and ASTM C143/C143M
equivalent CO2 emission factor. The latter (ASTM, 2015) standard. The average value was
multiplier was used after D. Flower and J. equallyaround16cm. Through this record, no
Sanjayan(Flower & Sanjayan, 2007) and the indication of the two mentioned properties was
assumption that both rock types took the same desired, but the consistency of mix proportions
value was applied. The similar mixing and between five designed shotcretes was found and
batching process was manifested for all no deviation affected by the aggregate gradation
mixtures, therefore the comparison of the was tracked (Autralian Shotcrete Society, 2010).
greenhouse gas emission of the works was not Furthermore, because no fibre had been added in
needed. Table 4 presented the obtained results. the proportion, the energy absorption under the
load-deflection curve of the shotcrete was
Table 4. CO2 estimationof shotcrete mixtures
neglected in the study (Autralian Shotcrete
(Unit: ton per 1m3 of shotcrete)
Society, 2010).
CO2 emission Cement Rock Sand Total
Emission factor 0.82 0.04 0.01 4. CONCLUSIONS
Recycled 1 459.20 17.18 12.65 489.03 Beside the conventional concrete, shotcrete has
Recycled 2 418.20 18.58 13.60 450.37 been in widespread use for tunnelling in
Recycled 3 418.20 18.83 13.78 450.80 particularly and in construction engineering in
Recycled 4 369.00 20.25 14.73 403.98 general. The effort of recycling tunnel muck as
shotcrete coarse aggregate in road tunnel projects
New 426.40 22.25 12.52 461.16
in Vietnam has been started as a right execution
Even the smallest CO2 emission belonged to toward the sustainability. The practise of using a
Recycled 4 mixture, it was not prudent to blasted rock aggregate which had been recycled
conclude if it is the “greenest” because the at a recent road tunnel project in Vietnam for the
least cement and water but the biggest raw 30MPa wet-mix shotcrete was presented in the
sand amount were used. Although the new rock paper. Since the specifications on materials for
mix did not discharge the heaviest amount of shotcrete has not been promulgated in Vietnam,
carbon dioxide, it was criticized for using the the related requirements of conventional concrete
large amount of raw materials in the collation following the current Vietnamese standards and
with the other mixes. the appropriate foreign regulationswere applied.

546
The recycled rock was conformed to be used as a EFNARC. 1993. European specification for spray
shotcrete’s coarse aggregate. Four shotcrete concrete. European Federation of National
mixtures using the recycled tunnel muck had Associations Representing producers and
been designed and compared with a new rock applicators of specialist building products for
Concrete, UK.
one. The predicted and 28-day compressive
Flower, D.; Sanjayan, J.2007. Greenhouse gas
strengths of 5 shotcretes were all surpassed the emissions due to concrete manufacturing.
target value. Recycled 2 mixture of which the International journal of Life cycle assessment,
cement/water = 2.5, coarse/fine aggregate = 12(5), 282-288.
35/65%, 2.16% by weight of mixture IBST(a). 2006. TCVN 7570:2006 Aggregate for
(accelerator and superplasticizer) was concrete and mortar (technical requirement).
recommended to be used for the project. An Ministry of Science and Technology of
initial carbon footprint estimation was shown the Vietnam, Hanoi. (In Vietnamese)
prospective of the utilization of tunnel muck for IBST(b). 2006. TCVN 7572-16:2006 Aggregates
for concrete and mortar – Test methods - Part
shotcrete in collating with the conventional
5: Method for determination of apparent
use.Beside the better environmental effect to the specific gravity, bulk specific gravity and
project and its vicinity during and after the water absorption of original stone and coarse
construction phase, the tunnel muck recycling for aggregate particles. Ministry of Science and
shotcrete can also provide other gains for the Technology of Vietnam, Hanoi. (In
mechanical and workability properties of the Vietnamese)
mixture (Chan, 1996). This paper was hoped to JSCE Working group for Mountain tunnels. 2006.
contribute in promoting the sustainability in Standard specifications for tunnelling - 2006.
tunnel engineering. Japan Society of Civil engineer, Tokyo.
MOST(a). 1993. TCVN 3106:1993 Fresh
heavyweight concrete - Method for slump test.
5. REFERENCES Ministry of Science and Technology of
ACI. 2016. ACI 506R-16: Guide for shotcrete - Vietnam, Hanoi.(In Vietnamese)
Reported by ACI Committee 506. American MOST(b). 1993. TCVN 3118:1993 Heavyweight
Concrete Institute, Farmington Hills, MI. concrete - Method for determination of
Anchor, R. et al. 1988. BS 8007: the new code. The compressive strength. Ministry of Science and
Structural Engineer, 66(3). Technology of Vietnam, Hanoi.(In
Vietnamese)
ASTM. 2013. ASTM Designation: C1436-13
Prime minister of Vietnam. 2020. Decision
Standard Specification for Materials for
152/2008/QD-TTg dated 28/11/2008 on
Shotcrete. ASTM International, West
Approving the scheme for exploring,
Conshohocken, PA.
extracting, processing and using mineral as
ASTM. 2015. ASTM Designation: C143/C143M-
building materials in Vietnam by 2020.
15 Standard Test Method for Slump of Government of Vietnam, Hanoi.
Hydraulic-Cement Concrete. ASTM Tran, TH.; Le, ML. 2018. Tunnel Muck Recycling
International, West Conshohocken, PA. for Road Construction – A Case Study in
ASTM. 2018. ASTM C42/C42M-18a: Standard Vietnam. In: Tran-Nguyen HH., Wong H.,
Test Method for Obtaining and Testing Drilled Ragueneau F., Ha-Minh C. (eds) Proceedings
Cores and Sawed Beams of Concrete. ASTM of the 4th Congrès International de
International, West Conshohocken, PA. Géotechnique - Ouvrages -Structures. CIGOS
AutralianShotcrete Society. 2010. Recommended 2017. Lecture Notes in Civil Engineering, vol
practice: Shotcreting in Australia 2nd edition. 8. Springer, Singapore, 1153-1163.
Concrete Institute of Australia, Rhodes. Zhang, L. 2014. Variability of compressive
Brito, J., Kurda, R. & Silva, P. 2018. Can We strength of shotcrete in a tunnel-lining project.
Truly Predict the Compressive Strength of Shotcrete, Fall,22-27.
Aggregates? Appl. Sci., 8, 1095.
Chan, C.1996. Thesis for Master of Applied
science: Use of recycled aggregate in shotcrete
and concrete. University of British Columbia,
Vancouver.

547
The 3rd Int. Conf. on Transport Infrastructure & Sustainable Development (TISDIC 2019). Construction Publishing House. ISBN: 978-604-82-2893-4.

Early predicting compressive strength of foamed concrete

using an artificial intelligence model

Ngo Tri Ngoc1, Nguyen Quang Trung1, Truong Ngoc Son1, Tran Thi My Dung1
1
Faculty of Project Management, The University of Danang - University of Science and Technology,
Danang, Vietnam

Corresponding Author: Ngo Tri Ngoc

Address correspondence to First author: [email protected]

ABSTRACT: Foamed concretes are construction materials which improve the sustainability of
structures. Predicting its compressive strength is difficult due to its non-linearly influenced by factors
such as the density of foamed concrete, cement content, and foam volume. This study proposes a novel
hybrid Artificial Intelligence (AI) model which couples the Least Squares Support Vector Regression
(LSSVR) with the Grey Wolf-Inspired metaheuristic Optimization (GWO) to predict the foamed
concrete’s compressive strength. Inputs are the density, cement, sand/cement ratio, water/cement ratio,
sand size, and foam volume. The evaluation results show that the GWO-LSSVR model yielded a
correlation coefficient (R) of 0.991 and mean absolute percentage error (MAPE) of 3.54%, which shows
the good agreement between the actual and predicted compressive strength. Notably, the GWO-LSSVR
model was better than the other models with the improvement of 144.2 ~ 156.1% in MAPE. Thus, the AI
model is an effective tool for designing foamed concrete.
Keywords: Foamed concrete, sustainable concrete, compressive strength, artificial intelligence,
machine learning.

1. INTRODUCTION Various studies have performed to examine

the effects of foamed concrete aggregates on its
The foamed concrete is a sustainable
compressive strength (Nguyen et al. 2017,
construction material that made of cementitious
Kearsley and Wainwright 2001, Hilal et al.
binders with void space, with or without fine
2015, Falliano et al. 2018). For example,
aggregates. It has been widely used in the
Falliano et al. (2018) investigated the effects of
construction field such as in geotechnical
engineering, heat insulation, and roofs because foaming agent and dry density on the
it possesses advantages in terms of reducing the compressive strength at 28 days. They revealed
cost, insulation capacity, and fire resistance that the compressive strength is positively
(Hajimohammadi et al. 2018, Nguyen et al. related with the density. A study by (Nguyen et
2017). Among mechanical properties, the al. 2017) showed that the compressive strength
compressive strength is important in foamed of foamed concretes was affected considerably
concretes (Nguyen et al. 2017). Thus, predicting by the air-void distribution.
the foamed concrete’s compressive strength at Some empirical methods have been
the early stage is vital in designing mixture proposed to calculate the compressive strength
proportions of foamed concrete. of foamed concrete (Kearsley and Wainwright

548
2001). For instance, Pan et al. (2007) proposed The AI has been widely applied to model
Eq. (1) to present the compressive strength (fcc) material behaviors (Qi et al. 2018), analyze the
in foamed concrete as a function of the air building markets (Zhao and Lam 2012), and
content (A) and the compressive strength of the predict energy consumption in buildings (Chou
cement paste (fc) (Pan et al. 2007). The fc is and Ngo 2016). Various AI techniques have
determined by the water/cement ratio (w/c) and been used such as the Artificial Neural
curing time (t) as Eq. (2). Equation 3 Networks (ANNs), Support Vector Regression
determined the influence of the binder ratio on (SVR), Classification And Regression Tree
the compressive strength (Kearsley and (CART) (Chou et al. 2017). Chou et al. (2017)
Wainwright 2001). Kearsley and Wainwright implemented some combination strategies to
(2002) developed empirical equations to improve the performance of AI in modeling the
compute the compressive strength based on its pitting corrosion and marine corrosion rate of
porosity and age (Kearsley and Wainwright steel (Chou et al. 2017). They investigated
2002). Calculating fcc based on the dry density performance in combining single Machine
ratio (αd) was determined as Eq. (4) (Kearsley Learning (ML) models including the ANNs,
and Wainwright 2002) SVR, CART, and linear regression (LR)
models. Yaseen et al. (2018) used an extreme
f cc  1.048 fc (1  A)2.793 (1)
learning machine (ELM) for forecasting
f c  88.04  6.569 ln(t )  130.5w / c (2) compressive strength of foamed concrete
(Yaseen et al. 2018).
f cc  1.172 f c b3.7 (3) Among the ML methods, the SVR is
suggested as a powerful ML model which is
fcc  fc  0.324  1.325 d 
2
(4)
effective and reliable for regression problems
where fcc is the compressive strength in foamed (Jiang et al. 2017). It was developed from
concrete, fc is the compressive strength of the statistical learning theory and the principle of
cement paste and αb is the binder ratio. structural risk minimization (Vapnik V. N.
However, accurately predicting the foamed 1995). The Least Squares SVR (LSSVR)
concrete’s compressive strength is still difficult approach is an improvement version of the
because it is significantly influenced by various SVR (Suykens et al. 2002), which have an
factors in a complex and non-linear manner advanced generalization and efficiency. For
such as density, cement, sand/cement ratio, example, Chou and Ngo (2018) developed an
water/cement ratio, sand size, agents, and AI system based on the LSSVR to the
foam. such as the type of foaming agent, the engineering strength of randomly fiber-
water/cement ratio, and sand/cement ratio reinforced soil (FRS). In their study, the
(Ramamurthy et al. 2009). Because of the LSSVR was used to model the relationship
complexity, most of the previous studies focus between FRS properties and its strength (Chou
on some factors among influencing factors in and Ngo 2018).
modeling the relationship between ingredients The literature shows that there are few
of foamed concrete and its compressive studies that apply AI for predicting the strength
strength. This limits accuracy in predicting of foamed concretes. Thus, this study proposes
compressive strength and may result in a data-driven hybrid AI method to early
discrepancy the calculated and actual results. forecast the compressive strength of the
Meanwhile, Artificial Intelligence (AI) is a lightweight foamed concrete. The hybrid
potential to deal with such a complex problem. model couples the (LSSVR with the Grey Wolf
Therefore, this work develops a novel AI-based Optimization (GWO), in which the LSSVR is
prediction method to predict the compressive used to model the relationship between
strength at 28 days in the foamed concrete at ingredients of the foamed concrete and its
the early design stage.

549
compressive strength while the GWO 2. METHODOLOGY
optimizes the LSSVR hyperparameters to
improve predictive accuracy. Its predictive 2.1 Machine Learning for Regression
performance was compared against with The LSSVR is an ML technique (Suykens et
baseline ML models including the non-linear al. 2002) which maps input vectors to high-
ML models (i.e., SVR models) and the Linear dimension feature space by a kernel function.
Regression (LR) model that reported in The LSSVR is trained by using a least squares
previous studies. cost function to achieve linear equations in the
The first originality is the proposal of a dual space to reduce computational effort.
novel hybrid AI model couples an ML with Figure 1 presents the architecture of the
an optimization algorithm to capture the LSSVR in which the LSSVR is trained by
highly nonlinear relationship of mix minimizing the function as Eq. (5).
proportions of foamed concrete and its 1 1 N
 C  ek2 ; subject to
2
min J ( , e)  
compressive strength; this has yet to be done  ,b , e 2 2 k 1
literature. As the second originality, this
study provides an effective tool for designers yk  ,(xk )  b  ek , k  1,...N (5)
or civil engineers in the design process
through a better understanding of where J(,e) is the optimization function;  is
relationships between ingredients of the the parameter of the linear approximator; ek  R
foamed concrete and its compressive strength is an error; C  0 is a regularization constant; xk
for enhancing the design efficiency in is inputs; yk is outputs (i.e., compressive
sustainable materials. strength); and N is a number of samples.

Figure 1. Structure of support vector regression model.

550
 Lagrange multipliers (k) is used to solve this 2.2.1 Encircling Prey
optimization problem which results in as Eq. (6)
Equations (9) - (12) are mathematical
in which the kernel function is expressed in Eq.
(7). Gaussian radial basis functions (RBF) equations of encircling behavior wolfs during
kernel as Eq. (8) was used as the kernel function hunting. A wolf updates its position within the
in this study because it is simpler than the search using Eqs. (9) and (10).
polynomial kernel function, and it is effective    
D  C  X p (t )  X p (t ) (9)
for solving non-linear problems.
N
   
f ( x )    k K ( x , xk )  b (6) X (t  1)  X p (t )  A  D (10)
k 1
   
N A  2a  r1  a (11)
K ( x, xk )   g k ( x) g k ( xk ) (7)
k 1  
C  2  r2 (12)
2
K ( x, xk )  exp( x  xk / 2 2 ) (8) 
where X (t  1) is the grey wolf’s location vector
where are αk is Lagrange multipliers; b is the 
at the (t+1)-th iteration; X p (t ) is the location
bias term; K(x,xk) is the kernel function;  is 
the RBF width. vector of the prey at the t-th iteration; A and
 
As mentioned early, the predictive accuracy C coefficient vectors; components of a are
of the LSSVR is subjective to its linearly decreased from 2 to 0 through
 
hyperparameters including the  of the RBF iterations; r1 and r2 are random vectors
kernel and regularization parameter (C). Thus, belonging [0,1].
an optimization algorithm is useful to couple
with the LSSVR to improve its performance. 2.2.2 Hunting
The hunt is usually directed by the . The
2.2 Grey Wolf - Inspired Metaheuristic
and may participate in the hunt
Optimization Algorithm
occasionally. Because the location of the
The GWO is a novel nature-inspired optimum (prey) is unknown, the , , and are
metaheuristic algorithm (Mirjalili et al. 2014). assumed to have knowledge regarding the
The GWO inspires behaviors of grey wolves. potential location of prey. The , , and
Grey wolfs are structured hierarchically with estimate the positions of the grey, and other
four levels of alpha (, beta (), delta (), wolves such the omegas update their locations
and omega (ω). Alpha is the leader. Betas are randomly surrounding the three best solutions.
subordinate wolves who advise the leader and The following equations are developed in this
discipline for the pack. Beta disseminates the regard (Mirjalili et al. 2014).
leader’s ideas to the pack and sends feedback        
to the leader. Delta must report to alpha and D  C1  X   X , D  C2  X   X ,
beta, which are scouts, sentinels, elders,    
hunters, and caretakers. Omega wolves play a D  C3  X   X (13)
role as a scapegoat and they must report to    
   
other above wolves. GWO is presented in X1  X  A1  ( D ), X 2  X   A2  ( D ),
more details in (Mirjalili et al. 2014,    
Wikiversity 2018). The optimization is X 3  X   A3  ( D ) (14)
directed by , , and . Searching for prey,   
encircling prey, and attacking prey are three  X1  X 2  X 3
X (t  1)  (15)
steps in hunting prey. 3

551
2.2.3 Attacking Prey 3. DATASET AND MODEL

The value of a is decreased to model grey 3.1 Data Description

wolves approaching the prey. A is randomly A 150-sample dataset on lightweight foamed
within the range [-a, a] while a is reduced from concrete was prepared to evaluate the
2 to 0 through the optimization process. predictive performance of the hybrid AI model

Regarding Eqs. (9) and (10) as A are ranged in in comparing against with a baseline non-linear
[-1, 1], the next location of wolves is in ML model (i.e., SVR) and a linear prediction
anywhere between its current location and the model (i.e., LR). Data on foamed concreted has
prey’s position which means that |A| < 1 forces derived from previously published experiments
the wolves to approach the prey. (Abd and Abd 2017). The foamed concrete was
made by four types of materials which are
2.2.4 Search for Prey ordinary Portland cement, water, sand, and
When |A| > 1, the GWO forces wolves to foam. The density is ranged from 1406.81 to
diverge from the prey. This presents the global 2009.48 kg/m3, water/cement ratio were 0.30,
 0.35, 0.40 and 0.45 and sand/cement ratio were
exploration of the GWO. C also impact the
0.5, 1.0, and 2.0. Size of fine silica sand was
exploration capacity of the GWO, which varies
 0.6 mm, 1.18 mm, and 2 mm. Foam volume
in [0, 2]. C is a random weight which was ranged from 47 to 357 liter per cubic meter
stochastically emphasizes the influence of prey of concrete.
in determining the distance in Eq. (9) as C > 1 The 150 samples were produced in cubes
and vice versa. This assists the GWO algorithm and their compressive strength was measured
to avoid local optima. C is not linearly at 7 and 28 days. Inputs of the hybrid AI model
decreased in contrast to A. are data attributes of foamed concrete which
The searching is started by generating a wolf consist of the density of foamed concrete,
pack. The , ,  wolves predict the potential cement volume, sand volume, sand/cement
location of the prey through iterations. ratio, water/cement ratio, sand size, foaming
Exploration and exploitation are controlled by agent, foam, and the compressive strength at 7
the a and A. Wolfs tend to diverge from the days. The output of the hybrid AI model is the
prey as |A| > 1 and converge towards the prey foamed concrete’s compressive strength at 28
as |A| < 1. The GWO is completed as the days. Table 1 shows the descriptive statistics of
stopping criterion is reached. the dataset on the inputs and output.

Table 1. Descriptive statistics of lightweight foamed concrete dataset.

Parameter Min Max Ave. Std Correlation
(x,y)
Density of foamed concrete (kg/m3) x1 1406.81 2009.48 1742.77 171.02 0.81
3
Cement content (kg per m ) x2 439.20 992.80 727.45 123.68 0.67
Sand content (kg per m3) x3 374.40 1098.00 733.45 152.40 0.27
Sand/cement ratio x4 0.50 2.00 1.06 0.38 -0.23
Water/cement ratio x5 0.30 0.45 0.39 0.05 -0.48
Sand size (m) x6 600 4750 835.60 820.83 -0.38
Foaming agent x7 0.00 0.50 0.18 0.18 0.22
Foam (liter per m3) x8 47.00 357.00 208.85 80.33 -0.71
Compressive strength at 7 days (MPa) x9 2.55 43.50 21.42 9.81 0.96
Compressive strength at 28 days (MPa) y 3.23 48.88 28.27 11.70

552
3.2. Hybrid Artificial Intelligence Model determines the suitable values of the LSSVR
implementation hyperparameter through the optimization
In the AI model, the LSSVR predicts the process with the aid of the GWO algorithm.
foamed concrete’s compressive strength at 28 The test phase is to evaluate the predictive
days based on the inputs, while the GWO accuracy of the optimized GWO-LSSVR
optimizes hyperparameters of the LSSVR (C model after the learning phase. A sensitivity
and ). By doing this, the proposed hybrid analysis was performed to select an
model can be improved predictive accuracy. approximate value of parameters of the model.
Figure 2 describes the implementation of the For initial settings for the hybrid AI model, the
proposed model for early forecasting the range of C and  was set as [0.0001 – 100], the
compressive strength in foamed concrete, which kernel function of the LSSVR was the RBF
mainly consists of the learning and test phases. kernel, the number of wolves was 100 and the
The learning phase of the GWO-LSSVR model iterations was set as 10.

Figure 2. A hybrid AI model for predicting lightweight foamed concrete’s compressive strength.

553
 The proposed GWO_LSSVR model was lightweight foamed concrete’s compressive
built and evaluated using a k-fold cross- strength at 28 days while a traditional
validation approach in which dataset is divided measurement method requires a period of 28
randomly into 10 folds and the model is days to measure the compressive strength.
evaluated 10 times. In each of ten times, 9 of The evaluation results in Table 3 reveal that
10 folds (the learning data) are used to train the AI model performed well in predicting the
and learn the model and the remaining fold (the compressive strength of the foamed concrete.
test data) is used to test the accuracy of the Particularly, the obtained linear correlation
learned GWO_LSSVR model. (R_te) are ranged from 0.977 to 0.999 during
During the learning phase, 70% of the the test phase with 10 evaluations, which are
learning data (the training data) was used to train very close to 1. Figure 3 visualizes the actual
the model, 30% of learning data (the validation and predicted compressive strength of
data) was fed into the trained model to build up lightweight foamed concrete at 28 days that
the objective functions for the optimization obtained by the proposed hybrid GWO-LSSVR
process (i.e., root-mean-square error RMSE) model in the scatter plot in the learning and test
phases for two evaluation times. The linear
4. RESULTS AND DISCUSSION correlation patterns showed the good agreement
The values of C and  of the LSSVR were between actual and predicted values.
optimized by the GWO during the learning To provide multiple angles on evaluating the
phase, and the optimized model was evaluated predictive accuracy, the RMSE, MAE, and
during the test phase. The hybrid AI model was MAPE indices were calculated based on the
evaluated 10 times using the dataset. Table 2 actual and predicted compressive strength. In
summaries results of the hybrid AI model for the test phase, the GWO-LSSVR model
predicting the lightweight foamed concrete’s yielded 1.39 MPa in average RMSE, 0.91 MPa
compressive strength at 28 days. The ranges of in average MAE, and 3.54% in MAPE. In
C and σ were [31, 100] and [0.307, 4.116], addition, the predictive performance of the
respectively. The computational time was 1.78 model was stable and reliable during ten times
minutes for 10 evaluation times, which means of evaluation because its performance was
the hybrid AI model only took 0.178 minutes to ranged as 2.07 ~ 5.63 % in MAPE, 0.64 ~ 2.61
implement one evaluation. The results revealed MPa in RMSE, and 0.48 ~ 1.52 MPa in MAE.
that the GWO-LSSVR model is very efficient to This result presents that the proposed hybrid
determine the optimal hyperparameter. The AI model was effective in predicting the
model can quickly and early forecast the compressive strength at 28 days.

Figure 3. Actual and predicted compressive strength at 28 days obtained by the proposed GWO-LSSVR
model with all inputs.

554
Table 2. Accuracy of the proposed AI model.
Evaluation Learning phase Test phase Hyper-
number parameter
RMSE_le MAE_le MAPE_le R_le RMSE_te MAE_te MAPE_te R_te C 
(MPa) (MPa) (%) (MPa) (MPa) (%)
1 1.01 0.66 2.45 0.996 1.38 0.94 3.89 0.994 100 2.987
2 0.84 0.55 2.10 0.997 2.61 1.51 5.63 0.979 100 2.385
3 0.48 0.24 1.03 0.999 1.45 0.65 2.27 0.987 100 0.307
4 0.95 0.60 2.26 0.997 1.00 0.78 2.76 0.994 100 2.215
5 1.03 0.67 2.47 0.996 0.64 0.48 2.07 0.999 100 2.824
6 0.95 0.66 2.50 0.997 2.35 1.52 4.46 0.977 100 4.116
7 0.86 0.56 2.27 0.997 0.80 0.69 4.36 0.999 31 1.025
8 1.04 0.68 2.53 0.996 0.94 0.77 2.65 0.995 100 3.024
9 0.59 0.36 1.52 0.999 1.55 0.91 3.18 0.989 63 0.656
10 1.02 0.66 2.38 0.996 1.16 0.91 4.15 0.996 100 3.071
Ave. 0.88 0.57 2.15 0.997 1.39 0.91 3.54 0.991 89.4 2.261
Stdev. 0.19 0.15 0.50 0.001 0.65 0.34 1.14 0.008 23.5 1.223

Table 3. Performance comparisons among the proposed models and models in previous works.
Model Accuracy results Accuracy improvement rate (%)
RMSE MAE MAPE R RMSE MAE MAPE R
(MPa) (MPa) (%)
LR (Abd and Abd 2017) 2.40 1.80 9.07 0.979 72.7b 96.3b 156.1b 1.2b
a b b
SVR with radial basis function 1.81 1.49 8.65 0.991 30.0 63.2 144.2 0.1
(Abd and Abd 2017)
SVR with linear kernel function 5.03 - - 0.945 261.8b - - 4.6 b
(Abd and Abd 2017)
SVR with polynomial kernel 2.31 - - 0.986 66.6 - - 0.5a
function (Abd and Abd 2017)
SVR with sigmoid kernel 8.17 - - 0.877 488.1 b - - 11.5 b
function (Abd and Abd 2017)
GWO_LSSVR 1.39 0.91 3.54 0.991
Note: Bold values depict the best accuracy among the models. aSignificance levels higher than 5%.
b
Significance levels higher than 1%.

The accuracy of the proposed model was and Abd 2017) in predicting the compressive
also compared against those of models in strength of lightweight foamed concrete at 28
previous works. Table 3 presents the predictive days. A value marked with bold indicates the
accuracy obtained by non-linear machine best predictive accuracy in the corresponding
learning SVR models with four types of kernel performance index among the comparative
functions (radial basis function, linear, models. The results revealed that the GWO-
polynomial, and sigmoid) and the linear LSSVR model was superior to all comparative
regression (LR) model that reported in (Abd models in terms of all performance indices.

555
 In comparing with the LR model, the during the prediction. In contrast, in the
proposed hybrid AI model improved proposed model, the LSSVR hyperparameters
significantly the predictive accuracy with the were optimized by the GWO algorithm during
accuracy improvement rate of 72.7% in RMSE, the learning phase, thus, the predictive
96.3% in MAE, 156.1% in MAPE at the accuracy was improved relatively. To sum up,
significant level of 1%. This result is the comparison confirmed that the proposed
reasonable because the LR model is a linear hybrid AI model was effective and reliable in
model which assumes a linear relationship predicting the compressive strength at 28 days.
between each input such as the density and The GWO-LSSVR model can help designers
water/cement ratio and the output (i.e., the or civil engineers in effectively designing the
compressive strength at 28 days). In contrast, foamed concrete.
the GWO-LSSVR model is a non-linear model
5. CONCLUSIONS
which can capture well the non-linear
relationship between the inputs and the output. This study developed the data-driven hybrid
artificial intelligence model to early predict the
compressive strength of lightweight foamed
concrete at 28 days. The proposed hybrid AI
model integrated a machine learning technique
and an optimization algorithm. The LSSVR is
used to model the relationship of ingredients of
foamed concrete and its compressive strength
while the GWO optimizes the hyperparameters
of LSSVR to improve the predictive accuracy.
The results exhibit the good agreement
between the actual and predicted compressive
strength at 28 days with the average R of
0.991. During the test phase, the GWO-LSSVR
model yielded the RMSE of 1.39 MPa, the
MAE of 0.91 MPa, and the MAPE of 3.54%,
which shows the good prediction accuracy in
early predicting the compressive strength. In
addition, the model is efficient because it took
Figure 4. Accuracy measures among models. less than 2 minutes for 10 evaluation times.
The proposed AI model was superior to
Figure 4 depicts accuracy comparison other concerned models from the previous
among the models in forecasting the studies in of predictive accuracy. The proposed
compressive strength of lightweight foamed GWO-LSSVR AI model improved
concrete at 28 days. The GWO-LSSVR model significantly the predictive accuracy with the
was performed better than the SVR models. accuracy improvement rate of 72.7% in RMSE,
The accuracy improvement rate obtained by 96.3% in MAE, 156.1% in MAPE at the
the proposed model was from 30.0 to 488.1 % significant level of 1% as compared to the LR
in RMSE, 63.2 % in MAE, and 144.2 % in model. The proposed model was performed
MAPE. Both the SVR models and GWO- better than the SVR models with the
LSSVR models are non-linear models which improvement rate in a range from 30.0 to 488.1
can capture the non-linear characteristics of the % in RMSE, 63.2 % in MAE, and 144.2 % in
data. However, the predictive performance MAPE. The efficacy of the hybrid AI model
depends on hyperparameter settings of the was explained by the hyperparameter settings
models. Hyperparameter settings of the SVR of the LSSVR in which the GWO algorithm
models were set as defaults and the fixed value was used to improve the predictive accuracy.

556
 As a contribution, this study couples a Jiang, M., Jiang, L., Jiang, D., Xiong, J., Shen, J.,
Ahmed, S. H., Luo, J. and Song, H. (2017)
machine learning and a nature-inspired 'Dynamic measurement errors prediction for
metaheuristic optimization method to propose a sensors based on firefly algorithm optimize
novel data-driven model that enables to early support vector machine', Sustainable Cities
predict the compressive strength of at 28 days. and Society, 35, 250-256.
Kearsley, E. P. and Wainwright, P. J. (2001) 'The
The proposed model is an alternative tool for effect of high fly ash content on the compressive
designers or civil engineers in effectively strength of foamed concrete', Cement and
designing the foamed concrete. Future studies Concrete Research, 31(1), 105-112.
Kearsley, E. P. and Wainwright, P. J. (2002) 'The
can investigate other machine learning or deep effect of porosity on the strength of foamed
learning to improve predictive accuracy. To concrete', Cement and Concrete Research,
promote the practical application, development 32(2), 233-239.
Mirjalili, S., Mirjalili, S. M. and Lewis, A. (2014)
of a standalone or web-based hybrid AI system is 'Grey Wolf Optimizer', Advances in
suggested to support civil engineers or designers. Engineering Software, 69, 46-61.
Nguyen, T. T., Bui, H. H., Ngo, T. D. and Nguyen,
ACKNOWLEDGEMENT G. D. (2017) 'Experimental and numerical
investigation of influence of air-voids on the
compressive behaviour of foamed concrete',
This work was supported by The University of Materials & Design, 130, 103-119.
Danang, University of Science and Pan, Z., Hiromi, F. and Wee, T. (2007) 'Preparation
Technology, the code number of the Project: of high performance foamed concrete from
T2019-02-37. cement, sand and mineral admixtures', Journal
of Wuhan University of Technology-Mater.
Sci. Ed., 22(2), 295-298.
6. REFERENCES Qi, C., Fourie, A., Chen, Q. and Zhang, Q. (2018)
'A strength prediction model using artificial
Abd, A. M. and Abd, S. M. (2017) 'Modelling the intelligence for recycling waste tailings as
strength of lightweight foamed concrete using cemented paste backfill', Journal of Cleaner
support vector machine (SVM)', Case Studies Production, 183, 566-578.
in Construction Materials, 6, 8-15. Ramamurthy, K., Kunhanandan Nambiar, E. K.
Chou, J.-S. and Ngo, N.-T. (2016) 'Time series and Indu Siva Ranjani, G. (2009) 'A
analytics using sliding window metaheuristic classification of studies on properties of foam
optimization-based machine learning system concrete', Cement and Concrete Composites,
for identifying building energy consumption 31(6), 388-396.
patterns', Applied Energy, 177, 751-770. Suykens, J. A. K., Gestel, T. V., Brabanter, J. D.,
Chou, J.-S. and Ngo, N.-T. (2018) 'Engineering Moor, B. D. and Vandewalle, J. (2002) Least
strength of fiber-reinforced soil estimated by Squares Support Vector Machines, Singapor:
swarm intelligence optimized regression World Scientific.
system', Neural Computing and Applications, Vapnik V. N. (1995) The nature of statistical
30(7), 2129-2144. learning theory, Springer-Verlag, New York.
Chou, J.-S., Ngo, N.-T. and Chong, W. K. (2017) Wikiversity (2018) 'Algorithm models/Grey Wolf
'The use of artificial intelligence combiners Optimizer', [online], available:
for modeling steel pitting risk and corrosion https://en.wikiversity.org/wiki/Algorithm_mo
rate', Engineering Applications of Artificial dels/Grey_Wolf_Optimizer [Accessed 17 Oct
Intelligence, 65, 471-483. 2018].
Falliano, D., De Domenico, D., Ricciardi, G. and Yaseen, Z. M., Deo, R. C., Hilal, A., Abd, A. M.,
Gugliandolo, E. (2018) 'Experimental Bueno, L. C., Salcedo-Sanz, S. and Nehdi, M.
investigation on the compressive strength of L. (2018) 'Predicting compressive strength of
foamed concrete: Effect of curing conditions, lightweight foamed concrete using extreme
cement type, foaming agent and dry density', learning machine model', Advances in
Construction and Building Materials, 165, Engineering Software, 115, 112-125.
735-749. Zhao, J. and Lam, K. P. (2012) 'Influential factors
Hajimohammadi, A., Ngo, T. and Mendis, P. analysis on LEED building markets in U.S.
(2018) 'Enhancing the strength of pre-made East Coast cities by using Support Vector
foams for foam concrete applications', Cement Regression', Sustainable Cities and Society, 5,
and Concrete Composites, 87, 164-171. 37-43.
Hilal, A. A., Thom, N. H. and Dawson, A. R.
(2015) 'On void structure and strength of
foamed concrete made without/with additives',
Construction and Building Materials, 85, 157-
164.

557