Progressive collapse of structures

Copyright © ICE Publishing, all rights reserved.

 To Misuk, Rudi, Leo, Enno, and Lisa

Copyright © ICE Publishing, all rights reserved.

 Progressive collapse
of structures

Uwe Starossek
Hamburg University of Technology

Copyright © ICE Publishing, all rights reserved.

 Published by Thomas Telford Limited, 40 Marsh Wall, London E14 9TP.
www.thomastelford.com

Distributors for Thomas Telford books are

USA: ASCE Press, 1801 Alexander Bell Drive, Reston, VA 20191-4400
Australia: DA Books and Journals, 648 Whitehorse Road, Mitcham 3132, Victoria

Also available from Thomas Telford Limited

Blast Effects on Buildings, second edition, edited by David Cormie, Geoff C. Mays and
Peter D. Smith, ISBN: 978-0-7277-3521-8
Forensic Engineering: From failure to understanding, 4th International Conference on
Forensic Engineering, edited by Brian S. Neale, ISBN: 978-0-7277-3613-0
Modern Structural Analysis: Modelling process and guidance, by Iain A. MacLeod,
ISBN: 978-0-7277-3279-8

Cover photograph of the partial collapse of the Charles de Gaulle Airport Terminal
in 2004 — an example of a structural failure which did not lead to progressive collapse
owing to structural segmentation. Image courtesy of AP/Press Association Images.

First published 2009

A catalogue record for this book is available from the British Library

ISBN: 978-0-7277-3610-9

# Thomas Telford Limited 2009

All rights, including translation, reserved. Except as permitted by the Copyright, Designs and
Patents Act 1988, no part of this publication may be reproduced, stored in a retrieval system or
transmitted in any form or by any means, electronic, mechanical, photocopying or otherwise,
without the prior written permission of the Publisher, Thomas Telford Limited, 40 Marsh Wall,
London E14 9TP.

This book is published on the understanding that the author is solely responsible for the statements
made and opinions expressed in it and that its publication does not necessarily imply that such
statements and/or opinions are or reflect the views or opinions of the publishers. While every
effort has been made to ensure that the statements made and the opinions expressed in this
publication provide a safe and accurate guide, no liability or responsibility can be accepted in
this respect by the authors or publishers.

Typeset by Academic þ Technical, Bristol

Printed and bound in Great Britain by Antony Rowe Limited, Chippenham
Index created by Indexing Specialists (UK) Ltd, Hove

Copyright © ICE Publishing, all rights reserved.

 Contents

Preface ix

1 Introduction 1
1.1 Opening 1
1.2 Failure incidents 2
1.3 State of research 5
1.4 Standards and guidelines 7

2 Typology of progressive collapse 11

2.1 General 11
2.2 Types of progressive collapse 12
2.2.1 Pancake-type collapse 12
2.2.2 Zipper-type collapse 14
2.2.3 Domino-type collapse 16
2.2.4 Section-type collapse 18
2.2.5 Instability-type collapse 20
2.2.6 Mixed-type collapse 23
2.3 Classes of progressive collapse 25
2.4 Collapse-promoting features 25
2.4.1 Dynamic action and force concentration 25
2.4.2 Brittle material behaviour 26
2.4.3 Overstrength and ductile material behaviour 26
2.4.4 Continuity or discontinuity 27
2.4.5 Series or parallel load transfer 28
2.4.6 Spatial orientation, size, and slenderness 28
2.4.7 Structuredness 29
2.5 Summary 30

3 Current design procedures 32

3.1 Inadequacy of current procedures 32
3.2 Possible improvements 34

Copyright © ICE Publishing, all rights reserved.

 4 Design against progressive collapse 37
4.1 General 37
4.2 Two definitions 38
4.2.1 Robustness 38
4.2.2 Collapse resistance 39
4.3 Design criteria 40
4.3.1 Design requirements 40
4.3.2 Design objectives 42
4.3.3 Design methods 47
4.3.4 Verification procedures 48

5 Design methods 52
5.1 General 52
5.2 Prevent local failure 55
5.2.1 Specific local resistance 55
5.2.2 Non-structural protective measures 55
5.2.3 Discussion and further remarks 56
5.3 Assume local failure 58
5.3.1 General 58
5.3.2 Alternative load paths 59
5.3.3 Isolation by segmentation 62
5.3.4 Redundancy versus segmentation 71
5.3.5 Interaction and progressive collapse 77
5.4 Local failure: prevent or assume? 80
5.5 Prescriptive design rules 83

6 Applications 86
6.1 General 86
6.2 Design of collapse-resistant bridges 87
6.2.1 General 87
6.2.2 Continuous girder bridges 88
6.2.3 Cable-stayed bridges 89
6.2.4 Suspension bridges 97
6.2.5 Arch bridges 99
6.3 Design of collapse-resistant high-rise buildings 99
6.3.1 General 99
6.3.2 Specific local resistance 100
6.3.3 Non-structural protective measures 112
6.3.4 Alternative load paths 113
6.3.5 Isolation by segmentation 114
6.3.6 Fire protection, fire fighting, and evacuation 121

vi

Copyright © ICE Publishing, all rights reserved.

 7 Measures of robustness and collapse resistance 123
7.1 General 123
7.2 Areas of application 123
7.3 Requirements 124
7.4 Present approaches 125
7.5 Stiffness-based measures of robustness 126
7.6 Damage-based measures of robustness 127
7.7 Energy-based measures of robustness 130
7.8 Summary and future directions 131

8 Conclusions 133
8.1 Synopsis 133
8.2 Outlook 137

References 139

Index 147

vii

Copyright © ICE Publishing, all rights reserved.

Copyright © ICE Publishing, all rights reserved.
 Preface

I became involved with the topic of this book 16 years ago, when, as
a practising engineer involved in the design of the Confederation
Bridge, Canada, I was charged with the progressive collapse analysis
and design of that structure. Apart from a few papers, mostly on the
Ronan Point incident, and some references in building codes, little
was known about progressive collapse at that time. This was a blessing
in disguise, however, because my colleagues and I at J. Muller
International, San Diego, were free to develop our views and
approaches almost from scratch. I have remained involved with this
topic ever since. When entering academia a few years later, I had the
time to reflect on the previous project-related work and to develop it
further into a more formalised and generally applicable concept.
Progressive collapse is arguably the most dramatic and feared form
of failure in structural engineering. It usually occurs unexpectedly
and causes high losses. Although there has been an awareness for a
long time that such failures can occur, this has barely been reflected,
until recently, in a professional effort commensurate to the problem.
Since the bombing of the Alfred P. Murrah Federal Building in 1995,
and even more so since the events of 11 September 2001, research
on progressive collapse has clearly intensified. Nevertheless, most
publications only focus on particular aspects of the problem, and
there is still no generally accepted use of nomenclature and
procedures. The standard provisions that exist today lack general
applicability.
This book attempts to close that gap. Its purpose is to give a plain and
comprehensive introduction to the phenomenon of progressive
collapse, to offer a consistent and generally applicable set of nomen-
clature and procedures, to provide guidance to the practicing engineer
in both a systematic and pragmatic manner, and to give an outlook on
future developments. Because the book is self-contained and requires
only a basic understanding of structural analysis and design, it is also

ix

Copyright © ICE Publishing, all rights reserved.

 suitable and, I hope, particularly educational for advanced under-
graduate and graduate students.
The book is based on a previous contribution to Betonkalender 2008,
a yearbook popular among German-speaking structural engineers.
Thanks are due to Konrad Bergmeister, the editor who had invited
me to contribute, and to Ernst & Sohn Verlag, Berlin, the publisher
who has allowed the use of this previously published material.
Furthermore, I would like to gratefully mention a few individuals who
helped shape my early understanding of the problem, namely Gerard
Sauvageot, Jean Muller, Paul Mondorf, Daniel Tassin, and Gamil
Tadros. More recently, discussions with Bob Smilowitz, Mohammed
Ettouney, Franco Bontempi, and Luisa Giuliani have influenced my
views in one way or another, which is also appreciated.
Special thanks are extended to Hyun-Moo Koh, my host at Seoul
National University. During my sabbatical stay there, the collapse
typology presented in this book came into being. Warm thanks are
also due to his graduate student Yong-Suk Park for our discussions
on bridge collapses.
I would like to thank my students at Hamburg University of Tech-
nology, most notably Maren Wolff and Marco Haberland, who were
involved in some of the studies reported here. Moreover, they have
read various parts and versions of the manuscript and helped to
improve its clarity and correctness. The latter also applies to my
brother Bernd Starossek. Thanks are also due to Deima Aslan and
Todd Parry for helping with preliminary translations of some parts of
the book, and to Axel Seils for preparing most of the illustrations.
Finally, it is a great pleasure to acknowledge my indebtedness to the
one person without whom this project would not have been realised.
Stuart Alexander encouraged me to submit the proposal for this book
to Thomas Telford Ltd, and was instrumental in getting it accepted
and bringing it to print. He reviewed the entire manuscript, and not
only helped straighten out my English but also made many invaluable
and helpful comments on the contents. Any remaining errors or
misconceptions are my own.
Wholehearted thanks go to my wife Misuk and our children for their
patience during the work on this book.

Uwe Starossek

Copyright © ICE Publishing, all rights reserved.

 1
Introduction

1.1 Opening
If there is a pronounced disproportion between a comparatively minor
event and the ensuing collapse of a major part or even the whole of a
structure, then this is a disproportionate collapse. When the collapse
commences with the failure of one or a few structural components
and then progresses over successive other components, a fitting label
would be progressive collapse. Although the two terms are often used
interchangeably, a distinction can be made. The term ‘disproportionate
collapse’ is more appropriate in the context of design and performance
because a precise definition of ‘disproportionate’ requires the choice of
design objectives (see Section 4.2.2). When referring to the physical
phenomenon and mechanism of collapse, on the other hand, the
term ‘progressive collapse’ is more suitable. Nevertheless, dispropor-
tionate collapse mostly occurs in a progressive manner. The main
subject of this book, therefore, is progressive disproportionate collapse
or, for short, progressive collapse.
The nature of the triggering event is irrelevant to the qualitative
definition of disproportionate collapse given above as well as to the
quantitative definition discussed later. It can be a local action or a
local lack of resistance. Thus, the term ‘event’ includes all types of
potential triggering circumstances. Since these occur with a low
probability or even wholly unexpectedly, they are called accidental
circumstances. Traditionally, this term referred more to events like
accidents and natural disasters or design and construction flaws. In
the light of emerging new threats, however, it nowadays also includes
deliberate damage inflicted by explosion and other kinds of malicious
action.
Different structures are susceptible to progressive collapse to different
degrees. Such differences remain unrecognised, though, even in modern
verification procedures using partial safety factors. This follows, in parti-
cular, from not factoring in the structural response to an initial local

Copyright © ICE Publishing, all rights reserved.

 Progressive collapse of structures

failure. Additional considerations are therefore necessary concerning

both the initial local failure and the ensuing response of the structure.
Such considerations have in the past been made only in isolated cases
— such as for embassy buildings or very long bridges, that is, for obviously
exposed or susceptible structures — and for the most part at the
discretion of the design engineer. Codified procedures did not exist
until recently; and those that exist today lack general applicability.
The design measures resulting from such considerations can either aim
at increasing the level of safety against local failure or at limiting the total
damage following local failure. Structures that are insensitive to local
failure, which is denoted by a limited extent of total damage, are here
termed robust. For preventing disproportionate collapse, as per the
definition above, what ultimately is required, however, is not insensitivity
to local failure, that is, robustness, but insensitivity to accidental circum-
stances — a property for which the term collapse resistance is introduced
here. Collapse resistance can be achieved through robustness, but also
by other means. Robustness and collapse resistance are key terms that
will be defined more precisely later (Sections 4.2.1 and 4.2.2).
Disregarding a possible susceptibility to progressive collapse in struc-
tural design can be associated with such major disasters as the collapses
of the Alfred P. Murrah Federal Building (Oklahoma City, 1995) and
the twin towers of the World Trade Center (WTC) (New York,
2001), as well as with a large number of less dramatic failures. Hence
the potential risk is high. Progressive collapse seems comparable in
this regard to earthquakes and other natural disasters. A further
common feature shared with earthquakes is the importance of dynamic
forces. On the other hand, the problem considered here is clearly differ-
entiated from the problems of seismic engineering by the diversity of the
triggering circumstances and the large variety of possible collapse
scenarios.
In the rest of this introductory chapter, first a few incidents of struc-
tural failure are presented where accidental circumstances led to varying
degrees of collapse. This is followed by an account of the current states
of research and standardisation in the field of progressive collapse. A
detailed discussion of the theoretical and practical aspects of this
problem is given in Chapters 2 to 8.

1.2 Failure incidents

Naturally, structural failure provides the strongest impetus for investi-
gating the present problem. Accounts and investigations of incidents

Copyright © ICE Publishing, all rights reserved.

 Introduction

of failure can also be an important basis for research. Corresponding

publications are relatively scarce, however, and often lack the detail
required for scientific utilisation.
Wearne1 portrays a number of failure incidents of recent years,
ranging from Ronan Point (a high-rise building in London, 1968) to
the Sampoong Superstore (a department store in Seoul, Korea, 1995).
The courses of failure described are characterised by sensitivity to
local failure, and can mostly be identified as progressive collapses.
The triggering events include design and construction flaws (Sampoong
Superstore), accidental gas explosion (Ronan Point), bomb attack, and
a multitude of other circumstances. An abundance of failure cases of
bridges, buildings, and other structures is described in detail and
discussed by Scheer,2 occasionally referring to lack of structural robust-
ness as a concurring cause of collapse. Again, the triggering events are
manifold.
The Viadotto Cannavino, a four-span continuous girder bridge in
Italy, partially collapsed during construction in 1972.3 The triggering
event was a formwork failure. The ensuing collapse was facilitated by
a lack of structural robustness, at least at that particular stage of
construction. Another progressive collapse occurred during the
construction of the Haeng-Ju Grand Bridge, a continuous prestressed
concrete girder bridge in Seoul, Korea, in 1992. Presumably after the
failure of a temporary pier in the main span — the triggering event
and the exact location of the initial failure have not yet been officially
determined — the collapse progressed through the adjacent ten spans,
and an 800 m section of bridge was lost.2;4
In both cases, the continuous prestressing tendons in the superstruc-
ture of the bridge played a specific and disastrous role. When the
Haeng-Ju Grand Bridge collapsed, most tendons resisted the enormous
stresses caused by the rupture of the encasing concrete and the failure
and fall of structural components.4 The high degree of strength or
toughness of the material, coupled with the continuity of the tendons
over the length of the bridge, worked against the robustness of the struc-
ture. A chain reaction ensued where the forces transmitted by the
tendons led to the collapse of all 11 continuous spans between transition
joints. Remarkably, it did not progress further into the other spans. A
classification of the type of collapse is attempted in Section 2.2.6.
A different course of events developed during a failure incident
affecting the Tasman Bridge near Hobart, Australia, in 1975.2;5;6 Due
to the impact of a 7200 t ore carrier, two piers of the bridge were
destroyed, and three spans of deck supported by them collapsed. The

Copyright © ICE Publishing, all rights reserved.

 Progressive collapse of structures

other 19 spans remained intact. The bridge deck was made of precast
prestressed concrete beams. The absence of collapse progression and
thus the robustness of that bridge are apparently related to the discon-
tinuity of prestressing tendons between adjacent spans. Interestingly,
preventing progressive collapse was an original design intent. Hence
even though the deck slab was continuous over the supports, the long-
itudinal concrete reinforcing bars in the slab were interrupted there and
locally spliced with light reinforcement of limited well-defined yield
stress. Furthermore, the diaphragms over the supports were split in
half and the splitted parts were connected by keyed joints — a measure
that allowed the transfer of compression and shear but not of tension.6
These observations on the occurrence and non-occurrence of
progressive collapse of bridges hint at the possible effectiveness of struc-
tural fuses and segmentation — design approaches that are studied
further in Sections 2.4.3 and 5.3.3.
The collapse of the I-35W Mississippi River Bridge in Minneapolis in
2007, on the other hand, has been ascribed partly to the lack of alter-
native paths for load transfer.7 Alternative load paths, if present, are
activated when forces that were carried by failing structural components
are redistributed into the structure. Because they require continuity,
this case indicates that structural robustness can increase with the
degree of continuity. This contrasts with the conclusions reached in
the previous cases. The resolution of this apparent paradox requires a
quantification of robustness — as will be made later in Section 4.2.1.
Prendergast8 reports on the partial collapse of the Alfred P. Murrah
Federal Building in Oklahoma City in April 1995. It was triggered by
the detonation of a truck bomb outside the building. The high degree
of destruction, and the large number of casualties, are attributed to
insufficient structural redundancy, that is, to a lack of alternative
paths. Every second exterior column was indirectly supported by a
continuous transfer girder at the second floor instead of extending to
the foundation. In the corresponding Federal Emergency Management
Agency report9 and in Ref. 10, this and other particularities of design
and weakness of the structure are exposed, and recommendations for
structural design are derived from them. Among other provisions,
concrete reinforcement continuity is recommended to prevent the fall
of failed floor slabs.
In the light of some of the other failure incidents described above, an
undifferentiated and overall increase in continuity does not automati-
cally enhance the robustness and collapse resistance of a structure
and, in certain cases, can even be harmful. This recommendation,

Copyright © ICE Publishing, all rights reserved.

 Introduction

therefore, deserves further examination (see Sections 5.3.4, 5.3.5, and

5.5).
In terms of tragedy and loss, the above-mentioned incidents are far
exceeded by the collapse on 11 September 2001 of the twin towers of
the WTC. The impacts of the aircrafts and the subsequent fires led to
local damage. The ensuing loss in vertical load-carrying capacity
extended, however, over the complete cross-section of each tower.
The upper part of the structure started to fall. Its collision with the
lower part caused large impact forces, which resulted in the complete
loss of vertical load-carrying capacity in the area of impact. Failure
progressed in the same manner, and led to total collapse.11ÿ16 The
mechanism of collapse is further described in Section 2.2.1 in connec-
tion with a classification of the type of collapse.
The progressive collapse of the neighbouring WTC Building 7 was
triggered by the fires that followed the impact of debris from the collapse
of the north tower (WTC 1). The sequence of collapse seems to have
been more complex in that the failure progressed in both the horizontal
and vertical directions and involved various mechanisms of collapse.17

1.3 State of research

Progressive collapse has, time and again, been a topic of research and,
on a few occasions, the focus of conferences.18;19 The publications
address specific aspects of the phenomenon or its occurrence in parti-
cular kinds of structures.20ÿ25 Further contributions consider specific
actions that can trigger a progressive collapse, such as fire26 or impact
and blast loading.27;28
Yokel et al.29 examine the design of the US embassy building in
Moscow in terms of susceptibility to progressive collapse. They compare
analysis methods, consider alternative load paths, and recommend
measures to increase the collapse resistance. Other project-related
papers centre on the design of the 12.9 km-long Confederation Bridge,
Canada,30ÿ34 discussing the possibility of a collapse progressing over
many spans of the bridge and the corresponding design countermeasures.
The publications mentioned so far deal with specific projects or
particular kinds of structures under specific triggering actions. They
are independent of each other and do not lay the groundwork for a
comprehensive theory of progressive collapse. The approaches, results,
and recommendations vary from case to case.
A comprehensive account of the phenomenon of progressive collapse
and the derivation of general rules for design and verification have rarely

Copyright © ICE Publishing, all rights reserved.

 Progressive collapse of structures

been attempted. Leyendecker and Ellingwood35 suggest that progressive

collapse should be prevented either by providing alternative paths or by
increasing the local resistance of the key elements of the structure
(specific local resistance). Gross and McGuire36 also compare design
methods for avoiding progressive collapse. They recommend verification
of the strength of a structure after notionally removing selected elements,
and thus also embrace the engineer-like deterministic approach of
providing alternative paths. Examples of possible detailing solutions for
the design of so-called insensitive structures are given by Kersken-
Bradley.37 Ideas for enhancing robustness — applying this term not only
to strength but also to serviceability and durability — are presented by
Pötzl.38 First efforts towards probabilistic verification procedures in line
with reliability theory are made in Refs 39—42. Progressive collapses
mostly involve dynamic effects induced by local failure and failure
progression. Such effects are not yet considered in the papers mentioned
in this paragraph.
Since the events of 11 September 2001, research into progressive
collapse has been intensified and more closely coordinated.43ÿ46 The
intention to address all open questions, to define objectives for research
and codification, and to develop a collective research programme
becomes clear in the Multihazard Mitigation Council report43 that
contains the contributions of a workshop held in 2002. In contrast to
earlier publications, a number of aspects are now emphasised: the neces-
sity of establishing performance expectations as part of the design
criteria; a differentiation of requirements on collapse resistance and
accuracy of verification procedures depending on significance and expo-
sure of the structure (tiered approach); a distinction between prescrip-
tive design rules (indirect design) that can be implemented quickly —
such as the requirement of continuity of reinforcement — and more
elaborate methods that are based on design criteria and structural
analysis (direct design, performance-based design). The need to verify
and improve the analysis tools is also pointed out. As for verification,
analytical comparisons with full model tests, with structures that did
not collapse despite serious damage, and with controlled demolition
of structures are suggested.
The present state of research is reflected in three special editions of
relevant professional journals. Articles concerning explosive blast
loading and the corresponding local and global failure mechanisms
can be found in a 2005 issue of the Journal of Structural Engineering.47
The focus is on that particular triggering action and on the analytical
or experimental determination of the structural response. The article

Copyright © ICE Publishing, all rights reserved.

 Introduction

by Hayes et al.48 pursues the question of whether a design that meets

seismic requirements at the same time provides a higher resistance to
explosion-triggered progressive collapse.
An overview of current problems and methods of resolution is given
by a 2006 issue of the Journal of Performance of Constructed Facilities.49
Strategies for analysis and verification as well as design methods are
investigated, partly in connection with case studies. The problem of
impact loading caused by separated and falling structural components
now attracts more attention.50;51 Controlled demolition experts also
have their say.52 Even though their goal is the opposite of preventing
collapse, they provide some interesting insights into the structural
behaviour under extreme actions and impact loading. Only one out of
12 papers argues for an application of the probabilistic theory of relia-
bility, albeit without addressing the problems identified in Chapter 3
of this book.
A workshop organised by the Joint Committee on Structural Safety
and the Building Research Establishment in 200546 focused on the
robustness of structures with regard to progressive collapse; selected
contributions appeared in revised form in a 2006 issue of Structural
Engineering International.53 While some of these articles are concerned
with particular kinds of structures and specific actions, others54ÿ58
strive for a fundamental understanding of progressive collapse. One
way into the problem is to examine the robustness of a structure or its
complementary property, called vulnerability. Various definitions of
robustness and vulnerability are suggested, and how these properties
could be quantified and computed is investigated. These approaches
are discussed further in Chapter 7.

1.4 Standards and guidelines

The requirement of structural resistance to progressive collapse is not
yet incorporated in structural design codes in a consistent and effective
manner. The CEB-FIP Model Code 199059 states in its very first para-
graph: ‘1.1.1 General requirements: Structures should with appropriate
degrees of reliability, during their construction and whole intended
lifetime, perform adequately and more particularly: (a) withstand all
actions and environmental influences, liable to occur, (b) withstand
accidental circumstances without damage disproportionate to the
original events (this is called the insensitivity requirement).’ Require-
ment (b) calls for collapse resistance in the qualitative sense as
introduced in Section 1.1. Thus, the problem at hand is openly and

Copyright © ICE Publishing, all rights reserved.

 Progressive collapse of structures

correctly addressed, although in quite general terms. The practical

impossibility of completely precluding damage from accidental circum-
stances is accepted here. It is required, however, that such damage is
limited. Further rules for design and verification are not specified.
The implementation of these requirements in a particular construction
project is up to the discretion and capability of the design engineer and
the consent of the owner.
Eurocode EN 199060 does not, apparently, go much further by stressing
that the failure of one structural component shall not lead to the collapse
of the entire structure. Nevertheless, this requirement now calls for
robustness, instead of collapse resistance, in the qualitative sense as intro-
duced in Section 1.1. It is in line with the American Society of Civil
Engineers (ASCE) 7-02 guideline61 that states: ‘buildings and structural
systems shall possess general structural integrity, which is the quality of
being able to sustain local damage with the structure as a whole remaining
stable and not being damaged to an extent disproportionate to the
original local damage’. The term ‘general structural integrity’ thus defined
corresponds to robustness as understood here. Such a qualitative require-
ment has also been included in the German DIN 1055-100 standard.62
Structural integrity is also called for in the American Association of
State and Highway Transportation Officials specifications,63 together
with the recommendation to consider ‘the use of continuity and redun-
dancy to provide one or more alternate load paths’ and ‘external protec-
tion systems to minimise the effects of reasonably conceived severe loads’
(see Section 1.1.2 of Ref. 63).
More precise rules for design and verification have been specified to
date only for particular kinds of structures such as buildings and cable-
stayed bridges. For design of the latter, the Post-Tensioning Institute
recommendations64 state that the loss of any one cable shall not lead
to ‘structural instability’. For the verification of such conditions, it is
recommended that cable loss should be assumed to occur suddenly so
that an impulsive dynamic loading is produced. Similar recommenda-
tions are made in guidelines issued by the International Federation
for Structural Concrete.65
Concerning more precise code rules for building design, the British
standards have a long tradition that started with the Ronan Point
incident. Moore66 gives a survey of their historical and current develop-
ment, focusing on accidental actions. The most recent British provisions
can be found in Approved Document A.67 Buildings are classified
according to the consequences of a collapse. The design requirements
and applicable design methods differ for the various classes. They

Copyright © ICE Publishing, all rights reserved.

 Introduction

consist of prescriptive design rules (ties), provision of alternative paths,

and application of a specified accidental design loading to key elements
to increase their local resistance. These methods are further detailed in
BS 595068 for the design of steelwork in buildings and in BS 811069 for
the design of concrete structures, respectively.
Eurocode EN 1991-1-770 specifies design actions and describes
methods of computation for selected accidental circumstances like
internal explosions and vehicle impact. Annex A of that standard
strongly resembles Approved Document A.67 Its applicability is likewise
limited to buildings and stadia. Annex B provides guidance for
performing a risk assessment, as it is required for the design of buildings
of the highest consequences class. A detailed comparison between
Eurocode EN 1991-1-7 and the current UK standards with respect to
provisions on accidental actions and progressive collapse can be found
in a Department for Communities and Local Government guide.71
Building design guidelines in the USA and Canada are critically
assessed by Dusenberry and Juneja.72 Even when progressive collapse
or the need for robustness is mentioned — in particular in the Canadian
code,73 the ASCE 7-02 guideline,61 and the American Concrete
Institute Committee 318 requirements74 — it is done so without
providing much practical guidance. Precise and binding design rules,
where given, are of a prescriptive nature and limited to the detailing
of concrete reinforcement and connections. The design objectives
are not quantified, and applicable design methods and verification
procedures are not specified.
Nevertheless, arguably the most detailed rules for designing
collapse-resistant buildings, although not for the private sector, have
been codified and are in use in the USA. In the General Services
Administration guidelines for federal office buildings75 and in the
United Facilities Criteria for Department of Defense projects,76 require-
ments are defined and provisions for design and verification are made.
The latter includes prescriptive design rules and performance-based
design based on structural analysis. Efforts are currently under way to
revise both guidelines and to combine them into a single document
that applies to all building projects of the US federal government and
armed forces.77
The National Institute of Standards and Technology (NIST) issued
the report Best Practices for Reducing the Potential for Progressive Collapse
in Buildings.78 This document reviews progressive collapse provisions in
various national design codes, but is not intended as a design code itself.
It is directed towards owners and practising engineers. Practical means

Copyright © ICE Publishing, all rights reserved.

 Progressive collapse of structures

for reducing the likelihood of progressive collapse of buildings are

presented and design methods for enhancing the resistance to progres-
sive collapse are reviewed. Based on the NIST investigation of the WTC
disaster,12;17 changes are being incorporated in the 2009 editions of the
International Building Code79 and the International Fire Code.80 These
changes focus on fire protection, fire fighting, and evacuation, and not
on structural design.
New design guidelines for private sector buildings in the USA are
being prepared by the Progressive Collapse Standards and Guidance
Committee that was formed by the Structural Engineering Institute of
the ASCE. Meanwhile, a so-called PreStandard Prospectus81 has been
presented that delineates the intended scope and the basic approach
of the new guidelines, the publication of which is expected in 2010.82
The PreStandard Prospectus refers to an article55 by the author when
distinguishing between robustness and collapse resistance, although a
somewhat different use is made of these terms than set forth in
Ref. 55 and in this book.

10

Copyright © ICE Publishing, all rights reserved.

 2
Typology of progressive collapse

2.1 General
Progressive collapse of structures is typically characterised by a dis-
proportion between a small triggering event and the resulting collapse
of a major part or even the whole of the structure. Although the dis-
proportion between cause and effect is a defining and common feature,
there are various differing mechanisms of collapse that produce such an
outcome. The amenability to conceptual, theoretical, and computa-
tional treatment can vary accordingly. Collapse-promoting features,
possible or preferable countermeasures, and the suitability of indices
for quantifying robustness and collapse resistance will likewise depend
on the mechanism of collapse. Finally, different kinds of structures
are susceptible to different mechanisms of collapse.
It is thus useful to distinguish and describe the different types of
progressive collapse and to attempt a classification on that basis.83 In
the subsequent sections, six different types of collapse and, at a
higher level of abstraction, four classes of collapse are identified. The
characteristic features of each category are described and compared,
and a terminology is suggested. As an example of how the basis offered
here can be applied and used for further study, collapse-promoting
features and possible countermeasures are discussed with respect to
the various types of collapse. Some insight into possible design methods
follows from this discussion. Furthermore, the description of the types of
collapse gives hints to a number of analogies that should be pursued
further.
The term propagating action used in the subsequent discussion refers to
the action that in the course of a progressive collapse results from the
failure of one structural component and leads to the failure of one or
more further structural components. Collapse-promoting features can
be deduced, in particular, from an examination of the propagating
action. They are material related or inherent to the structural system.
Some hitherto little-noticed features can be extracted in this way.

11

Copyright © ICE Publishing, all rights reserved.

 Progressive collapse of structures

2.2 Types of progressive collapse

2.2.1 Pancake-type collapse
This type is exemplified by the collapse of the World Trade Center
(WTC) towers. The impacts of the aircrafts and the subsequent fires
led to damage and resulted in local failures in the areas of aircraft
impact. The ensuing loss in vertical load-carrying capacity was limited
to a few storeys but extended over the entire cross-section of each
tower. The upper part of the structure started to fall and to accumulate
kinetic energy. The subsequent collision with the lower part, which was
still intact, caused large impact forces that were far beyond both the
design and actual reserve capacities of the structure. This structural
impact thus again led to the loss of vertical load-carrying capacity
over the entire cross-section of the tower in the area of impact. Failure
progressed in the same manner and led to total collapse.11—16
(Dissenting explanations were suggested84,85 but have not found
general acceptance.)
The term suggested for this type of collapse originates from the
appearance after collapse of buildings smaller than the WTC towers
(Fig. 2.1). Their potential energy is smaller, and hence less destruction
occurs when compared with the WTC towers. The floor slabs of
these buildings thus pile on top of each other, resembling a stack of
pancakes.
The mechanism of a pancake-type collapse exhibits the following
features:

Fig. 2.1 Pancake-type collapse of a ten-storey building triggered by an earthquake

(Islamabad, 2005)

12

Copyright © ICE Publishing, all rights reserved.

 Typology of progressive collapse

. initial failure of vertical load-bearing elements (i.e. elements that

are vertically aligned and carry vertical loads)
. separation of structural components and their fall in vertical
rigid-body motion
. transformation of gravitational potential energy into kinetic energy
. impact of separated and falling components on the remaining
structure
. failure of other vertical load-bearing elements due to the axial
compression forces that result from the impact loading
. failure progression in the vertical direction.

Characteristic features are, in particular, the separation of structural

components (including non-load-bearing elements), the release of
gravitational energy, and the occurrence of impact forces. Depending
on the size of the falling components and the height of fall, the gravita-
tional potential energy released during the fall can far exceed the elastic
potential energy (strain energy) stored in the structure. If that energy is
reintroduced into the structure in a subsequent impact, large internal
forces ensue. Due to the dynamic nature of impact, they tend to con-
centrate in the load-bearing elements directly affected by the impact.
(Otherwise, the progressing failure fronts observed in the WTC
towers would be difficult to explain.) The failure progresses when
these forces cause the impacted elements to fail as well. Their actual
failure can be associated with any local failure mode, including
instability (buckling).
The propagating action is the vertical impact force. Another feature
of interest is that the principal forces in the failing elements (before the
onset of collapse), the propagating action, and the direction of failure
progression are all vertical and thus parallel. Furthermore, the structural
system is characterised by series primary load transfer. The principal
forces, the propagating action, and the direction of failure progression
being parallel to each other is related to the primary load transfer
being serial.
An alternative mode of pancake-type collapse, in which only the
floors or the cantilever floors of a building fail in a vertical failure
progression, is described in Section 6.3.2.3. Some important features
of such a collapse deviate from the description given above. Separation
of components can be partial instead of complete. Their fall does not
need to occur in a purely vertical rigid-body motion but can also be
associated with rotation. Furthermore, element failure is due to bending
or shear instead of axial compression and the failing elements are not

13

Copyright © ICE Publishing, all rights reserved.

 Progressive collapse of structures

linked through collective series load transfer. Such a collapse, although

limited to a building corner, occurred during the Ronan Point incident
(see Fig. 5.18). This alternative mode of pancake-type collapse exhibits
some features of the domino-type collapse described in Section 2.2.3.
More detailed discussion is left for future work.

2.2.2 Zipper-type collapse

For the design of cable-stayed bridges, the Post-Tensioning Institute
recommendations64 require that the loss of any one cable shall not
lead to structural instability. To prevent such an outcome, it is recom-
mended that loss-of-cable load cases be included in the structural
analysis. This requirement is intended, among other things, to prevent
a zipper-type collapse triggered by the initial failure of a single cable and
progressing due to the resulting overloading and failure of adjacent
cables (see Section 6.2.3 for further discussion).
Such a failure progression, albeit in a suspension bridge, can be
observed in the footage of the collapse of the original Tacoma Narrows
Bridge in 1940. After the first hangers snapped due to excessive wind-
induced distortions of the bridge girder,86 the entire girder peeled off
from the hangers and suspension cables, and fell (Fig. 2.2).

Fig. 2.2 Zipper-type collapse of the Tacoma Narrows Bridge, triggered by wind-
induced vibrations (Washington, 1940)

14

Copyright © ICE Publishing, all rights reserved.

 Typology of progressive collapse

A similar mechanism of collapse can be envisaged in anchored

retaining walls, where progressive collapse could be initiated by the
failure of one or a few anchors.87 The UK Standing Committee on
Structural Safety has drawn attention to a number of collapses of
heavy acoustic ceilings in cinemas and other buildings that seem to
correspond to zipper-type collapses triggered by the failure of one or a
few inadequate fixings to the supporting structure.88
In the examples just given, zipper-type collapse is related to the initial
failure of tension elements. This, however, is not a necessary condition.
Before identifying further possible cases, some characteristics of this
type of progressive collapse will be specified. The mechanism of a
zipper-type collapse exhibits the following features:
. initial failure of one or a few load-bearing elements
. redistribution of the forces carried by these elements into the
remaining structure
. impulsive dynamic loading due to the suddenness of the initial
failure and redistribution of forces
. dynamic response of the remaining structure to that impulsive
dynamic loading
. concentration of forces in load-bearing elements that are similar in
type and function to and adjacent to or in the vicinity of the initially
failing elements due to the combined static and dynamic structural
response to that failure
. overloading and failure of those elements
. failure progression in a direction transverse to the principal forces in
the failing elements.
Characteristic features are, in particular, the redistribution of forces
into alternative load paths, the impulsive loading due to sudden
element failure, and the concentration of static and dynamic forces in
the next elements to fail. The propagating action resulting from the
failure of an element is the negative of the sectional forces in that
element prior to failure acting as step impulsive loadings on the faces
of the ruptured cross-section (or possibly at the ends of the failing
element in the case of buckling). Impact forces, on the other hand,
do not typically occur — in contrast to the pancake-type collapse
discussed previously.
Another interesting difference is that now the principal forces in the
failing elements (before the onset of collapse) and the propagating
action on the one hand and the direction of failure progression on
the other are not parallel but more or less orthogonal. This corresponds

15

Copyright © ICE Publishing, all rights reserved.

 Progressive collapse of structures

to the structural system being characterised by rather parallel instead of

series primary load transfer. Failure progression occurs when the alter-
native paths, which in principle can form, and in reality do tend to
form, in such a system, become overloaded and fail.
Other cases exhibiting the features listed above can be envisaged: a
cable net or membrane structure that unzips following the occurrence
of local damage; a continuous girder in which one span fails in bending
leading to overloading and failure in bending of adjacent spans; or a
continuous girder supported by slender columns in which one column
fails leading to overloading and failure of adjacent columns. Also for
zipper-type collapse, element failure can be associated with any local
failure mode, including instability (buckling).

2.2.3 Domino-type collapse

A row of dominoes collapses in a fascinating chain reaction when one
piece falls at the push of a finger. The mechanism behind this type of
collapse is as follows:
. initial overturning of one element (i.e. of one domino)
. fall of that element in an angular rigid-body motion around a
bottom edge
. transformation of gravitational potential energy into kinetic energy
. lateral impact of the upper edge of the overturning element on the
side face of an adjacent similar element
. the horizontal pushing force transmitted by that impact is of both
static and dynamic origin because it results from both the tilting
and the motion of the impacting element
. overturning of the adjacent element due to the horizontal pushing
force from the impacting element
. failure progression in the overturning direction.
The occurrence and the importance of impact forces suggest a
relationship with the pancake-type collapse. On the other hand, the
principal forces in the failing elements — which are generally gravity-
induced — are now orthogonal to the direction of failure progression,
and the elements constitute a parallel load transfer system (at least
before the onset of collapse) — two properties that are also shared by
the zipper-type collapse (although in that case they hold for the
system after the onset of collapse as well). It thus appears reasonable
to distinguish the domino-type collapse from the previously discussed
types.

16

Copyright © ICE Publishing, all rights reserved.

 Typology of progressive collapse

A characteristic and distinguishing feature is the overturning of

individual elements. Another peculiarity is related to the propagating
action, which here is the horizontal pushing force that leads to the
overturning of the next element. The principal forces in the failing
elements (before the onset of collapse) are vertical. Thus, in contrast
to the previously described types of collapse, the principal forces and
the propagating action are not parallel. A domino-type collapse thus
exploits a particular weakness of the structure towards forces that do
not act in the direction of the principal forces.
Any group of individual structures that are susceptible to overturning
and are placed in a repetitive horizontal arrangement, similar to a row of
dominoes, could collapse in such a manner. Susceptibility to
overturning implies a certain degree of slenderness of the individual
structures and a lack of bracing. Such conditions could occur, for
instance, in a row of temporary scaffolding towers or in multi-span
bridges if the pier height exceeds the span length.
Another case exhibiting most of the features listed above is the
progressive collapse of the towers of overhead transmission lines,
which have occasionally occurred in a number of countries, sometimes
on an extremely large scale.89 In such an incident near Münster,
Germany, in 2005, more than 50 towers were affected. To include
such cases of collapse in the domino-type category, two amendments
are necessary to the above list of features.
First, it is conceded that the lateral impact of an element onto an
adjacent element can also be indirect, that is, mediated by other,
different elements. For the case just mentioned, the mediating element
is the transmission line that connects the individual towers. A corollary
to this is that the propagating action does not need to be a pushing force
but can also be a pulling force. Second, the failure does not necessarily
need to progress in the direction of overturning. Given that the
overturning of towers can also occur in a direction transverse to the
transmission line, the direction of overturning and the propagating
action can either be aligned with or orthogonal to the transmission
line and the direction of failure progression.
The list of features characterising a domino-type collapse is thus
amended as follows:

. initial overturning of one element, which can be an individual

structure
. fall of that element in an angular rigid-body motion around a
bottom edge

17

Copyright © ICE Publishing, all rights reserved.

 Progressive collapse of structures

. transformation of gravitational potential energy into kinetic energy

. abrupt deceleration of the motion of the element through sudden
activation, possibly effected by mediating elements, of discrete
other elements, which can be individual structures
. the horizontal force induced by that event is of both static and
dynamic origin because it results from both the tilting and the
motion of the decelerated element
. overturning of the suddenly activated other elements due to the
horizontal force from the decelerated element
. failure progression in the horizontal direction.

It is noted that this list of features also covers more general cases where
the consecutively affected elements or structures are neither similar nor
aligned.
When the individual elements or structures are anchored to the
ground, overturning and failure progression requires the failure of the
anchorages. Having said that, an overturning element does not need
to be a complete individual structure that turns around a bottom
edge at foundation level. The described failure mechanism is also
present when — after local structural failure above foundation level —
the upper part of an individual structure turns around a plastic hinge
at the level of local failure. This was apparently the case in the
above-mentioned progressive collapse of overhead transmission line
towers in Germany (Fig. 2.3).
Essentially unrestrained overturning motion and subsequent
deceleration are important features of domino-type collapse as they
enable a substantial amount of gravitational energy to be transformed
into kinetic energy and that energy to be subsequently reintroduced
into the structure.
From the list of features given above, another distinguishing pecu-
liarity follows that deserves attention. The horizontal force identified
as the propagating action acts exclusively on discrete elements and
not on a continuous structure as in the previously discussed types of
collapse. This could mean a larger concentration of that force on the
one hand and a better predictability of the structural response on the
other.

2.2.4 Section-type collapse

Consider a beam under a bending moment or a bar under axial tension.
When a part of the cross-section is cut, the internal forces transmitted

18

Copyright © ICE Publishing, all rights reserved.

 Typology of progressive collapse

Fig. 2.3 Domino-type collapse of overhead transmission line towers triggered by

ice accretion (Germany, 2005). (Source: picture-alliance/dpa/dpaweb. Photo-
graph: Franz-Peter Tschauner, # dpa)

until then by that part are redistributed into the remaining cross-
section. The corresponding increase in stress at some locations can
cause the rupture of further parts of the cross-section. In the same
manner, failure can progress throughout the entire cross-section.
While this kind of failure is usually not called progressive collapse but
brittle fracture or fast fracture, it is useful to include it in this description
of types of collapse in order to reveal similarities and to exploit analogies
that might exist.
When compared with the previously discussed types, a section-type
collapse exhibits certain similarities to a zipper-type collapse. Indeed,
the same list of features applies provided the terms ‘part of cross-section’
and ‘remaining cross-section’ are substituted for the terms ‘element’ and
‘remaining structure’, respectively. The cross-section assumes the role of
the structure. The main differences are that the cross-section of a beam or
tie rod is unstructured, continuously contiguous, and comparatively
homogeneous, whereas a structure, for instance a cable-stayed bridge,
is generally structured consisting of discrete elements that are inter-
connected in various ways and have different properties. Even so, the

19

Copyright © ICE Publishing, all rights reserved.

 Progressive collapse of structures

similarities might be close enough to apply by analogy the procedures for

treating section failure and brittle fracture, and, in particular, the
methods of fracture mechanics, to the zipper-type collapse.
According to the premises of linear-elastic fracture mechanics, the
stresses near a crack tip tend to infinity as the distance from the
crack tip approaches zero. In a structure that consists of discrete
elements, the distance between the elements consecutively engulfed
by the failure front, for instance the cables of a cable-stayed bridge, is
non-zero. Nevertheless, in the latter case there is also a force concentra-
tion in the elements that neighbour a failing element. This effect is
partly dynamic, which should be taken into account when adopting
methods of fracture mechanics.
The cross-section could also be a fillet or butt weld transmitting
tensile stresses perpendicular to the weld, and the same statements
apply. A somewhat different situation arises in a fillet weld in longi-
tudinal shear. In this case, the tendency to unzip is amplified by the
stress concentrations at the weld ends that already appear before the
onset of failure. The same applies to a lap splice of concrete reinforcing
bars. In both cases, the failure progresses in the directions of the
principal force and the propagating action, contrary to a zipper-type
collapse and the other kinds of section-type collapse discussed above.
Based on a postulated analogy between progressive collapse and fast
fracture, a verification procedure and a measure of robustness were
suggested by Smith56 (see also Section 7.7). Mechanisms of collapse
in buildings are studied based on evaluating and balancing the various
kinds of energy released and transformed during collapse. The effect of
separated and falling components that impact the remaining structure is
not considered, though, which might be attributed to fundamental diffi-
culties of the theory as presented. Thus, its application to buildings — that
is, structures that arguably are more susceptible to pancake-type collapse
— might not be as meaningful. As established above, the analogy between
progressive collapse and fast fracture is more applicable to structures
prone to zipper-type collapse, such as, for instance, cable-supported
bridges, cable nets, or membrane structures.

2.2.5 Instability-type collapse

Failure of structures due to instability is characterised by small pertur-
bations (imperfections, lateral loading) leading to large deformations
or collapse. Structures are designed such that stability failure will not
normally occur. This is typically achieved by providing additional

20

Copyright © ICE Publishing, all rights reserved.

 Typology of progressive collapse

structural components that brace or stiffen the structure. When the

discussion is limited to static stability failure under conventional
loading, the braced or stiffened parts of the structure are load-bearing
elements in compression, and the stability failure is called buckling.
Preventing stability failure by bracing, on the other hand, means that
the failure of a bracing element due to some small triggering event can
make a structure unstable and result in collapse. This applies, for
instance, to truss or frame structures in which bracing is used to stabilise
compression bars or compressed parts of cross-sections. Another
example is the initial failure of a plate stiffener that can lead to local
instability and the failure of the affected plate, and possibly to global
collapse. All these cases potentially exhibit a typical characteristic of
progressive collapse, namely a small triggering event resulting in
disproportionate collapse.
The mechanism of an instability-type collapse exhibits the following
features:
. initial failure of bracing or stiffening elements that have been
stabilising load-bearing elements (of the structure or of cross-
sections) in compression
. instability of the elements in compression that cease to be stabilised
. sudden stability failure of these destabilised compressed elements
due to small perturbations
. immediate collapse or failure progression.
If the element first affected by destabilisation is one of a few primary
load-bearing elements, say the corner leg of a truss tower, complete
collapse can ensue immediately without a progressive failure of consecu-
tively affected further elements. According to the distinction made in
Section 1.1, such an incident would be a disproportionate but not at
the same time a progressive collapse — in contrast to the other types
of collapse discussed in the previous sections.
On the other hand, the stability failure of the element first affected by
destabilisation can also set in motion the progressive stability failure of
consecutively affected further elements. This would be the case in a
continuous girder where the spans fail consecutively due to buckling
of the compression chord of the girder (lateral torsional buckling). In
this example, however, the subsequently affected elements fail due to
overloading as a result of a redistribution of forces — not because of
the discontinuance of stabilising elements. Such a failure progression
exhibits the features of a zipper-type collapse, discussed above, and
should be categorised as such.

21

Copyright © ICE Publishing, all rights reserved.

 Progressive collapse of structures

In order to specify a progressive instability-type collapse, the last item

of the above list should be replaced by the following two features (to be
appended to the penultimate item in the list):
. failure — caused by the stability failure of these destabilised com-
pressed elements — failure of bracing or stiffening elements of
other load-bearing elements in compression
. failure progression in the same repetitive manner.
Failure progression in the course of an instability-type collapse entails
progressive destabilisation. The destabilisation can progress consecu-
tively or even continuously. An example of the latter is propagating
buckles that can develop in deep-water pipelines and, once initiated,
can propagate over great lengths of pipe. A progressive destabilisation
occurs when the failure of destabilised elements leads to the failure of
stabilising elements. Though not a necessary condition, this can be
the case when load-bearing elements are at the same time stabilising
elements. In a deep-water pipeline, the load-bearing and stabilising
functions are both performed by the cylindrical shell that forms the
pipe.
Characteristic features of an instability-type collapse are compression
in stabilised elements, failure of stabilising elements, and subsequent
stability failure of the compressed elements thus destabilised. The
importance of compression loading and stability failure for this type of
collapse is rooted in the transformation of gravitational potential
energy into elastic potential energy (strain energy) that occurs during
the deformation of the destabilised elements in compression. Dynamic
action can play a greater or lesser role. The collapse of a truss tower after
the buckling of a corner leg can be explained by considering only static
forces. An appropriate modelling of propagating buckles in deep-water
pipelines, in contrast, requires the inclusion of dynamic forces.90
Likewise, the feature of a failure progression over multiple con-
secutively affected elements can be present or absent, as was discussed
above. The presence of this feature indicates a progressive instability-
type collapse. A propagating action, by definition, arises only in the
course of a progressive collapse. It is therefore absent in an instability-
type collapse without failure progression, such as in the immediate
collapse of a truss tower after the buckling of a corner leg. In
an instability-type collapse with failure progression, such as in
propagating buckles, the propagating action is predominantly a destabi-
lisation and less a force. In both cases, though, the compression loading
that causes the failure of destabilised elements is mainly a static force

22

Copyright © ICE Publishing, all rights reserved.

 Typology of progressive collapse

that already acts before the onset of collapse and does not propagate. It
is not a force that changes immediately before or during the course of
collapse. This peculiarity is another defining and common feature of
instability-type collapse. Furthermore, if there is a propagating action
that is not predominantly a destabilisation, the type of collapse is
different even if the first or subsequently affected elements fail due to
instability.
The notion of instability includes the concept of disproportion
between perturbations, which are small, and structural response,
which is large. This resembles the disproportion between a small
triggering event and the resulting widespread collapse of a structure
that characterises disproportionate and progressive collapse. This
similarity of concepts opens up an interesting line of thought. Just as
instability-type collapse is introduced here as one of several types of
progressive collapse, progressive collapse could conversely be conceived
as a particular kind of stability failure. This analogy might prove fruitful
for the development of approaches for treating progressive collapse
based on methods for treating instability.

2.2.6 Mixed-type collapse

The types of collapse considered so far are relatively easily discerned and
described. Some incidents of progressive collapse that have occurred in
the past do not neatly fit into these categories, however.
The partial collapse of the Alfred P. Murrah Federal Building (see
Section 5.1) exhibited features of not only a pancake-type but probably
also a domino-type scenario. A characteristic feature of the latter type is
the occurrence in the course of collapse of horizontal forces that lead to
the overturning of other elements. Horizontal tensile forces could have
been induced in that incident by partially separated falling components
that remained connected to the still intact parts of the structure
through continuous concrete reinforcing bars. The possible occurrence
and the importance of such forces are suggested by the fact that the
collapse stopped shortly after a reinforcement discontinuity in a major
horizontal element of the structure (see Section 5.3.4).
The collapse of the Haeng-Ju Grand Bridge (see Section 1.2) seems to
have involved features of the zipper-type and the domino-type cate-
gories. The importance of the latter type is underlined by the fact
that the debris of this continuous prestressed concrete bridge came to
rest in a longitudinally shifted position and remained connected
through the continuous post-tensioning tendons which mostly stayed

23

Copyright © ICE Publishing, all rights reserved.

 Progressive collapse of structures

intact. This indicates that the tendons served as a mediating element

transmitting the propagating action of a domino-type collapse. But
zipper-type features might also have contributed to the collapse because
the initial failure of one span must have induced an impulsive redistri-
bution of forces. This, in turn, possibly led to overloading and failure in
bending of adjacent spans.
The girders and towers of a cable-stayed bridge are in compression.
They are braced by the stay cables. Thus, the loss of one or more
cables can not only lead to unzipping, as stated above, but also to
global stability failure. In such a scenario, the features of zipper-type
and instability-type collapses will most likely interact and thus reinforce
each other in promoting collapse (see Section 6.2.3). Possible kinds of
interaction are (a) a progressive destabilisation through a zipper-like
failure of cables and (b) an increase in force concentration in the
next element to fail (a zipper-type feature) caused by the vertical displa-
cement increment due to compression (an effect associated with
instability). The propagating action in this case is both a force and a
destabilisation.
In certain kinds of structures, particularly certain kinds of buildings, it
even seems possible that features of the four basic categories — pancake-
type, zipper-type, domino-type, and instability-type collapse — interact
and jointly contribute to failure progression. The importance of
pancake-type and domino-type features for the collapse of buildings
has already been addressed above. Features of zipper-type collapse
could emerge, for instance, in a multi-bay frame structure when the
failure of columns leads to overloading and failure of adjacent columns.
Because failure progression also tends to reduce stiffness and to elimi-
nate stabilising elements in a progressive manner, and considering
that the columns in the latter example can fail due to buckling, the
propagating action can partly consist of destabilisation, a feature asso-
ciated with instability-type collapse.
Mixed-type collapse is less amenable to abstracting and generalising
treatment because the relative importance of the contributing basic
categories of collapse can, in principle, vary. Nevertheless, further
study could lead to the definition of other well-defined types of collapse,
which might turn out to be combinations of the basic types described
here or independent new varieties. This could be the case, for instance,
for the alternative mode of pancake-type collapse mentioned in
Section 2.2.1 in which only the floors of a building are involved in
vertical failure progression and which also exhibits features of a
domino-type collapse.

24

Copyright © ICE Publishing, all rights reserved.

 Typology of progressive collapse

2.3 Classes of progressive collapse

The preceding discussion of types of collapse based on their respective
mechanisms and characteristic and distinguishing features allows
further abstraction and an outline of a hierarchical classification. In
doing so, types of collapse are combined into classes. Both zipper-type
and section-type collapses are characterised by the redistribution of
forces carried by failing elements or cross-sectional parts into the
remaining structure or cross-section. These two types are thus
subsumed under one class of collapse, for which the term redistribution
class is suggested.
Pancake-type collapse and domino-type collapses, in comparison,
have fewer features in common. For all their differences, however,
they share an important characteristic: a substantial amount of gravita-
tional potential energy is transformed into kinetic energy during the fall
or overturning of components and is subsequently reintroduced into the
structure. The reintroduction of energy occurs more or less abruptly.
These two types of collapse are thus combined into one class of collapse,
for which the term impact class is suggested. The term ‘impact’ is
intended to be understood in a more general sense as it also refers to
the abrupt deceleration of overturning elements in a domino-type
collapse.
Instability-type collapse forms a class of its own. It is characterised by
the destabilisation of load-bearing elements in compression caused
by the discontinuance of stabilising elements. The transformation of
gravitational potential energy plays a role but in a different way than
for the types of collapse in the impact class. Finally, mixed-type collapse
also constitutes a separate class, for which, however, it is difficult to
identify common characteristics other than the fact that features of
various types of collapse interact and thus reinforce each other in
promoting collapse.

2.4 Collapse-promoting features

2.4.1 Dynamic action and force concentration
Despite the differences discerned in the preceding discussion, there are
some collapse-promoting features that are shared, to varying extents, by
the different types of collapse. Although dynamic action is necessary
only for the explanation of a pancake-type collapse, it also plays a role
in the other types of collapse.
A concentration of force in the next element to fail induced by the
failure of the previous element occurs in all the types of collapse

25

Copyright © ICE Publishing, all rights reserved.

 Progressive collapse of structures

discussed, except possibly in the instability-type collapse. This

force concentration is another feature promoting failure progression.
Both these features are characteristics not of a structure but of the
mechanisms of collapse.

2.4.2 Brittle material behaviour

Dynamic action and force concentration become more detrimental as
the material of the element that is prone to fail next becomes more
brittle. Brittle material behaviour can thus be a collapse-promoting
feature of a structure. Ductile material, on the other hand, is able to
absorb and dissipate kinetic energy, and permits a redistribution of
forces resulting in a reduction of force concentration.
For this reason, ductile material can be particularly beneficial — or
brittle material particularly harmful — when the design aims at
increasing the robustness of a structure through the provision of alter-
native load paths (see Section 5.3.2). This concerns, in particular, the
possibility of a zipper-type collapse and countermeasures to prevent it.
The beneficial effect of ductile material behaviour can show itself in
all types of collapse. This is less obvious regarding the prevention of a
domino-type collapse. But even then, it can help to dissipate kinetic
energy and to stop the collapse when the overturning of the element
concerned, say the tower of an overhead transmission line, would
require the failure of ground anchors or a local structural failure
above foundation level.

2.4.3 Overstrength and ductile material behaviour

Intriguingly, there are also instances where strength and even ductility
are detrimental. Consider a domino-type collapse in which the propa-
gating action is transmitted by mediating elements in tension (say a
transmission line). This action corresponds to a force in the mediating
elements larger than that occurring under normal conditions. To limit
that force, the strength of the mediating elements or their connections
should be limited, and thus overstrength should be avoided.
But even then, the actual strength of a mediating element can result
in an excessive force on the next element affected (a tower) that cannot
reasonably be designed for. The reason lies in the safety margin required
in the design of the mediating element for normal loading conditions
and in a possible change of position and force direction of that element.
In such a case, the mediating element or its connections should instead

26

Copyright © ICE Publishing, all rights reserved.

 Typology of progressive collapse

be non-ductile in order to minimise the duration of the action and the

impulse transmitted to the next element. The inertia of the next
element can thus be put to work to resist its overturning. These remarks
also apply to a mixed-type collapse with significant involvement of
domino-type features.
A mediating element that releases at a defined force or moment in a
non-ductile manner is a predetermined breaking point. It resembles a
fast-blow electric fuse, and can be considered a structural equivalent.
This analogy hints at the potential effectiveness of structural fuses for
also limiting a zipper-type collapse. In an electric power network, the
failure of a component (power station, transmission line, substation)
is usually compensated for by a redistribution of electric power flow to
other components. This redistribution, however, can lead to over-
loading and progressive failure of those other components and even
to the collapse of the entire network — a mode of failure that resembles
in its characteristic features the zipper-type collapse of a structure
described here. The redistribution of electric power flow is undesirable
in such a case, and must be precluded by activation of electrical protec-
tion devices. The collapse of the electric power network thus remains
spatially isolated and limited. A zipper-type collapse of a structure can
similarly be limited by structural fuses.
In more general terms, a well-defined strength and a non-ductile
material behaviour can be advantageous in isolating elements that are
intended to constitute segment borders. A segment border is an
expedient of a design method aimed at limiting the extent of collapse
through isolating the collapsing part by means of structural segmen-
tation (see Section 5.3.3). This method is applicable to all types of
collapse, although not always in association with structural fuses.

2.4.4 Continuity or discontinuity

The discussion in the previous two sections indicates that the
advantages and disadvantages of high strength and ductility should be
evaluated primarily with reference to the methods used in design and
less so in connection with the type of collapse. The same applies to
the structural features called continuity and discontinuity.
Continuity can be favourable when the design aims at increasing
the robustness of a structure by providing alternative load paths. If
collapsing parts are to be isolated by structural segmentation, on the
other hand, the elements designed as segment borders can achieve
their isolating effect not only by continuity and high local resistance

27

Copyright © ICE Publishing, all rights reserved.

 Progressive collapse of structures

but also by discontinuity. The choice between the two design methods
just outlined is a fundamental problem that should also be considered
with regard to beneficial or harmful interaction and the predictability
of the structural response (see Section 5.3.5).

2.4.5 Series or parallel load transfer

In the case of series load transfer, the initial failure of any one load-
bearing element can be expected to cause collapse. It can occur, for
instance, as a pancake-type collapse of a high-rise building. The initially
failing element here is the whole of the vertical load-bearing members of
a storey. The structure is characterised by a lack of robustness, at least
with respect to this assumed initial failure. Series load transfer is there-
fore undesirable in principle. On the other hand, it is inevitable in some
structures, such as slender high-rise buildings.
In the case of parallel load transfer, the forces carried by failing elements
can be redistributed into the remaining structure and carried by other
elements (alternative paths). Collapse can be avoided in this manner.
However, redistribution of forces is also a characteristic of zipper-type
collapse. Its occurrence requires parallel load transfer. To counter such
a type of collapse, a redistribution of forces must be permitted not only
by the topology of the structural system, that is, by an adequate degree
of connectivity and continuity, but also by making the resulting alter-
native paths sufficiently strong (see Section 5.3.4).

2.4.6 Spatial orientation, size, and slenderness

Structures are often aligned in a primary direction. High-rise buildings
are mainly vertically aligned, whereas bridges are mainly aligned along
a horizontal axis. Pancake-type collapse is more likely in a vertically
aligned structure. This follows from the characteristics of a pancake-
type collapse: the impact of separated and falling components on the
remaining structure and a failure progression in the vertical direction.
A domino-type collapse, in contrast, is characterised by the overturning
of elements and a horizontal failure progression; it therefore implies a
horizontal alignment of the structure or group of structures. A zipper-
type collapse typically progresses in the horizontal direction but does
not seem to require a distinct primary direction of structural alignment.
The latter also applies to the other types of collapse.
The susceptibility to progressive collapse increases with the size of a
structure. In buildings up to, say, ten storeys built in accordance with

28

Copyright © ICE Publishing, all rights reserved.

 Typology of progressive collapse

modern design standards, the impact loading resulting from the failure
of one storey can normally be tolerated by the remaining structure, and
does not generally lead to a pancake-type collapse (see Fig. 6.15). The
latter seems almost inevitable, though, in buildings higher than, say, 30
storeys (if special measures, such as those described in Section 6.3, are
not taken). The reason for this height dependency lies in the ratio of the
gravitational potential energy to the elastic potential energy, which
increases with height. The other types of collapse require a certain
number of elements and a certain structural size in order to be recog-
nised as progressive and disproportionate.
The slenderness of a high-rise building goes along with its vertical
alignment. It is similarly a structural feature that promotes pancake-
type collapse. This is based on the ratio between the extent of accidental
circumstances required for the failure of one storey and the suscept-
ibility to failure progression addressed in the previous paragraph. This
ratio decreases — and thus the structure becomes more susceptible to
a pancake-type collapse — with an increase in slenderness. Because
slenderness facilitates failure from overturning and instability, it is a
property that also favours domino-type and instability-type collapses.
In these cases, it is the slenderness of individual structures or elements,
not that of the overall structure, that is of main concern.

2.4.7 Structuredness
A further collapse-promoting structural feature appearing in some of the
types of collapse discussed is the structuredness (as opposed to compact-
ness) of a structure. This term is already in use in the context of software
development.91 Adopting it by analogy, structuredness is the degree to
which a structure possesses a definite pattern of organisation of its inter-
dependent load-bearing elements. In this sense, the discrete frame
structure of a high-rise building with its organisational pattern of hori-
zontal beams and slabs and vertical columns is highly structured
whereas an industrial chimney of the same height designed as a compact
reinforced concrete tube is comparatively unstructured.
Structuredness is a condition for, or at least favours, pancake-type
collapse. This follows from the observation that the characteristic
features of this type of collapse — that is, the release and subsequent
reintroduction of a large amount of gravitational energy associated
with the fall and impact of components — are difficult to develop in a
structure lacking structuredness. For a domino-type collapse, the
description of which is comparatively specific, structuredness is implied

29

Copyright © ICE Publishing, all rights reserved.

 Progressive collapse of structures

by the defining features. Structuredness is thus important for both types

of collapse in the impact class. For zipper-type, section-type, and
instability-type collapses, on the other hand, structuredness does not
seem to be a necessary condition.

2.5 Summary
Progressive collapse can be produced by various different mechanisms.
Based on a discussion of these mechanisms, five distinct types of collapse
are identified. The terms suggested for these five categories are pancake-
type, zipper-type, domino-type, section-type, and instability-type
collapse. These categories are relatively easily distinguishable through
their respective features, which are described here. These features can
interact, though, in various combinations and to varying degrees. The
resulting types of collapse are subsumed under a sixth category, called
mixed-type collapse. A higher level of abstraction is achieved by
combining the pancake-type and domino-type categories, on the one
hand, into a so-called impact class, and the zipper-type and section-
type categories, on the other hand, into a redistribution class.
It is hoped that the typology presented here will not only advance the
conceptual treatment of progressive collapse but also help in developing
analysis procedures and in quantifying robustness and collapse resis-
tance (see Chapter 7), as well as in developing, classifying, and selecting
countermeasures. If little progress has been made to date in achieving
these goals, one reason might have been the hitherto lack of differentia-
tion and classification of types of collapse on the basis of a description of
the underlying mechanisms.
Both the theoretical treatment and appropriate countermeasures will
differ for different types of collapse. For instance, if pancake-type
collapse can only be treated in a deterministic way, zipper-type collapse
might become tractable by probabilistic approaches. Some analogies are
given that seem promising to pursue further: the similarity between
section failure and zipper-type collapse hints at the applicability of frac-
ture mechanics to the latter type of collapse; the similarity between the
definitions of instability and progressive collapse could be explored to
develop concepts for treating progressive collapse from approaches for
treating instability.
As another example of how the basis offered here could be used for
future study, collapse-promoting features — and some countermeasures
— are outlined and their association with types of collapse are discussed.
Salient collapse-promoting features of the mechanisms of collapse are

30

Copyright © ICE Publishing, all rights reserved.

 Typology of progressive collapse

dynamic action and force concentration. Collapse-promoting features of

a structure are large size and slenderness, and a certain degree of struc-
turedness, a term suggested and defined here. If the structure is spatially
aligned in a primary direction, it is susceptible to particular types of
collapse. The influence of other structural features is more difficult to
generalise. Although brittle material behaviour can be collapse
promoting, it is shown that overstrength and ductile material behaviour
can also be detrimental in some cases. Similarly, continuity or disconti-
nuity, and series or parallel load transfer can be beneficial or harmful,
depending on the kind of structure, the type of collapse, and the
chosen method of design to prevent it. More insights might flow from
a further discussion of the types and mechanisms of progressive collapse.

31

Copyright © ICE Publishing, all rights reserved.

 3
Current design procedures

3.1 Inadequacy of current procedures

Modern design codes and verification procedures are based on the prob-
abilistic theory of reliability. Actions and resistances are determined
statistically from empirical data obtained from observations and
measurements and represented by their respective probability density
functions. Based thereupon, and after stipulating an allowable prob-
ability of failure, the design values for actions and resistances are
computed using probabilistic methods. Design using these values
promises a uniform level of safety.
Such an approach is based on a mathematically sophisticated and
sound foundation and is, as it seems, correct. It is reflected in the
design codes by partial safety factors and a series of load combination
schemes. The application of the accompanying code rules is often
cumbersome due to the multitude of load combinations to be consid-
ered. The design engineer might still take comfort in the idea that, by
working on a rational mathematical basis, a uniform level of safety is
achieved. Unfortunately, that idea is illusory because such an approach
fails with regard to the identification of a possible susceptibility to
progressive collapse and the development of appropriate counter-
measures. There are three reasons for this failure.31,92
The first reason lies in the focus on local instead of global failure.
Design equations for checking structural safety are usually defined
and applied at a local level only (check of stresses, sectional forces, or
stability of load-bearing elements). Structural safety, therefore, is also
accounted for only at a local level. The structural safety at a global or
system level, that is, the safety against the collapse of the entire
structure or a major part thereof, is a function of not only the safety
against local failure of all load-bearing elements but also of the structural
response of the entire system to local failure. The latter effect is
neglected. Different structures will respond differently to local failure.
The tacit underlying assumption that a uniform safety level of a

32

Copyright © ICE Publishing, all rights reserved.

 Current design procedures

structure is achieved by merely providing a uniform level of safety to its

elements is therefore not generally valid. When applied to non-robust
structures, such procedures produce unsafe designs. (See Section 4.2.1
for a definition of robustness.)
The problem can be illustrated by means of a simplified expression for
the probability, P(C), of a progressive collapse, C, due to an event, E:
PðCÞ ¼ PðCjLÞ
PðLjEÞ
PðEÞ ð1Þ
where P(E) is the probability of occurrence of E, PðLjEÞ is the
probability of local failure, L, given the occurrence of E, and PðCjLÞ
represents the probability of progressive collapse given the occurrence
of L.41 The factor PðCjLÞ is not reflected in the verification procedures
of current design codes. The possibility of a progressive collapse thus
remains out of consideration.
If the preceding appears to be critical of the application of the
probabilistic theory of reliability, it also holds that there is currently
no deterministic design code either that remedies the problem just
outlined. Nevertheless, it must be stated that the hope of achieving a
uniform level of safety by using reliability theory has not been fulfilled.
Although this deficiency is known to professionals familiar with
reliability theory, practising engineers are often unaware of it.
The second reason for the inadequacy of current design procedures is
that accidental circumstances are neglected. Accidental circumstances
are events that are unforeseeable or whose probability of occurrence is
very low, that is, events, E, for which P(E) is very small. In the context
of reliability-based design, neglecting them is necessary because the
supporting statistical data, derived from experience and observation,
are unavailable.93 For low-probability events, the source data are too
scarce. Unforeseeable events, in particular those that have never
occurred before, can at best be speculated on. In the case of a non-
robust structure, however, this neglect of accidental circumstances is
unacceptable. This follows from the first problem with current design
procedures discussed above.
For illustration, consider a structure characterised by series primary
load transfer, say a slender high-rise building. A relevant initial local
failure is the simultaneous failure of all the vertical load-bearing
elements of a storey. For the sake of argument, assume that such an
initial failure leads to impact forces that cause the building to collapse,
that is, PðCjLÞ ¼ 1. For buildings of sufficient height, this assumption
will hold for the initial failure of most storeys. The probability of collapse
is then of the order of the sum of the failure probabilities of all the

33

Copyright © ICE Publishing, all rights reserved.

 Progressive collapse of structures

constitutive elements, that is, of the individual storeys of the building.31

If the number of storeys is sufficiently large (put simply, if the area of
attack of the building is sufficiently large), even very low probabilities
of local failure resulting from accidental circumstances can add up to
a probability of global failure high enough to be taken seriously. For
structures with parallel primary load transfer, in contrast, the probability
of global failure is of the order of the product of the failure probabilities of
the constitutive elements, and therefore very low.
The third reason for the inadequacy of current design procedures is
that the underlying probabilistic concept requires the specification of
an acceptable probability of failure. The target failure probabilities of
probabilistic design codes are usually derived from calibration with
previous deterministic design codes. Hence, no new societal consensus
seems necessary when probabilistic design is adopted. Considering the
huge losses that can result from progressive collapse, however, it is diffi-
cult to see how an informed and true societal consensus on the numer-
ical value of the acceptable probability of such an outcome can be
reached — a problem that risks of the ‘low probability/high consequence’
type are typically up against.94 This problem can be evaded, but not
solved, by not bringing the question to the attention of the public.

3.2 Possible improvements

The first problem outlined in the previous section is not actually
inherent in the theory of reliability. As illustrated by Eq. (1) and further
explained in Refs 41 and 42, it is, in principle, possible to account for
progressive collapse within a probabilistic framework. The difficulty
arises from practical limitations that appear when the theory is applied
to actual structures. In this process, the structural response to local
failure needs to be examined. This response typically involves large
deformations and displacements, separation of structural components,
falling or overturning components striking other parts of the structure,
and other unusual kinds of interaction.
Proper treatment of all these effects generally entails a geometrically
and materially nonlinear dynamic analysis in the time domain. Even the
precise modelling of such scenarios is difficult because some structural
properties, such as the behaviour of joints or post-tensioning tendons
under extreme and unusual loading, are not well established. These
difficulties are compounded by the need to consider many different
cases of initial local failure and by the fact that, owing to the nonlinear
dependencies arising here, small errors in the modelling assumptions

34

Copyright © ICE Publishing, all rights reserved.

 Current design procedures

can produce large deviations in the computational outcome. Even a

deterministic analysis of the structural response to local failure poses
tremendous difficulties. A probabilistic analysis of this response together
with a computation of the safety against global failure adds further
dimensions of difficulty, and, apart from special cases, seems out of
reach of today’s analysis resources.
Regarding the problems of computational tractability and predict-
ability of the structural response, it is of interest that different kinds
of structures are susceptible to different mechanisms and types of
collapse (see Chapter 2). The latter, in turn, involve varying degrees
of computational effort. The amenability to computational treatment
in general, and to a probabilistic approach in particular, seems greatest
for structures where local failure is unlikely to result in major impact
incidents, for instance structures that are susceptible to zipper-type
collapse but not to pancake-type collapse.
If a probabilistic analysis of particular kinds of structures or structures
in general becomes feasible in the future, one could attempt to consider
the structural response to local failure by an additional system partial
safety factor on the resistance side of the design equations. The frame-
work of current design procedures could thus be maintained. The addi-
tional safety factor would take a value of 1 for robust structures and a
value smaller than 1 for non-robust structures. Provisions in some
design codes aim in this direction. In these cases, however, the reduction
of the design values of resistance for non-robust structures is based on
engineering judgement rather than on thorough probabilistic analysis.
Partial safety factors for considering structural robustness would have
to be specified on the basis of parametric analyses of the safety against
global failure for all the kinds of structures covered by the respective
design code. The robustness of all these structures will vary widely. A
classification of structures is thus required in order to assign to each
class of structure its respective reduction factor (see also Section 7.2,
point 5). Moreover, the reduction factor would have to be specified
differently even for different structural elements of structures of the
same class according to the respective global consequences of element
failure. Such differentiation can be considered as the semi-probabilistic
counterpart of the specific-local-resistance method, a deterministic
approach discussed in Section 5.2. Another possibility is to pursue a
fully probabilistic analysis for a given design task, thus dispensing with
the use of safety factors.
The second and third problems outlined in the previous section
are fundamental challenges to purely reliability-based design. If low

35

Copyright © ICE Publishing, all rights reserved.

 Progressive collapse of structures

probabilities of local failure resulting from accidental circumstances

add up to a large probability of global failure, then the probabilities of
occurrence of the accidental circumstances need to be known. They
are needed as input data for the probabilistic analysis, and cannot be
neglected. The determination of the actual values of these probabilities,
however, seems extremely difficult, if not impossible. Also, if a societal
consensus on the acceptable probability of a major disaster cannot be
reached, another basic ingredient to a probabilistic computation is
missing.

36

Copyright © ICE Publishing, all rights reserved.

 4
Design against progressive
collapse

4.1 General
It follows from the preceding discussion that the inadequacy of current
design methods are both fundamental and practical in nature, and can
only partly be overcome within the framework of reliability theory. The
possibilities for improvement that do exist require a major advance in
computational methods and extensive numerical studies, and may in
the end turn out to be insufficient. Nonetheless, there is a need for
immediate guidance. Both a general concept for handling the questions
raised by progressive collapse and a procedure for designing collapse-
resistant structures are urgently required. The following pragmatic
approach to tackling these tasks is suggested.55,92,95
As a starting point, the design procedures of current codes are
retained. They are based on reliability theory, which is reflected in
the codes by partial safety factors and load combination schemes.
Where necessary and possible, for instance concerning ship impact on
major bridges, this code-based design can be complemented by direct
probabilistic analysis and risk assessment. (In view of the inconsistencies
of current design procedures outlined above, one could argue that the
number of load combinations prescribed by some codes should be
reduced because it is exaggerated when compared with the accuracy
actually achieved.)
In addition, an assessment with respect to progressive and dis-
proportionate collapse is carried out and corresponding measures are
taken. The procedure is further described in the following sections. It
is anticipated here that this will not be based on the probabilistic
theory of reliability but on engineering judgement and on concrete,
deterministically defined design objectives that must be established in
a decision-making process. Structural analyses will similarly be carried
out deterministically. Emphasis is put on performance-based methods.

37

Copyright © ICE Publishing, all rights reserved.

 Progressive collapse of structures

This approach is pragmatic because it lacks the rigour of a purely

mathematical basis but still helps to solve the problem at hand. It
enables the design engineer to adequately address progressive collapse
in the sense that safety and economy are reasonably balanced and the
required structural analysis remains tractable. Such an approach has
implicitly been used already in the few reported cases where progressive
collapse has been considered in the design of actual structures.33 Recent
European and American regulations on accidental actions and progres-
sive collapse (see Ref. 70 (Annex A), and Refs 67, 75, and 76) indicate
that codification is already moving towards such a pragmatic approach.
The same holds for the ASCE Structural Engineering Institute PreStan-
dard Prospectus,81 which is influenced by this proposal. It is hoped that
future code development will lead to a clearer description of the design
criteria and a more precise and consistent use of language.
This presentation of the additional assessment and measures relating
to progressive collapse begins with two fundamental definitions. Then
follows a listing and discussion of design criteria that reflect the ques-
tions at hand, and the provisions to be established before and during
the design process. The approach presented here can be applied to
both new structures and retrofit projects.

4.2 Two definitions

The terms ‘robustness’ and ‘collapse resistance’ have already been
introduced (in Section 1.1). The term ‘robustness’ appears frequently
in discussions and publications on progressive or disproportionate
collapse. It is used differently by different authors,46,49,53,70 however,
and there is no general agreement to date on its precise meaning.
The definition of robustness given here and the introduction of the
new term ‘collapse resistance’ prove useful for the suggested approach
to handling progressive collapse. Before continuing to present that
approach, therefore, both terms are defined and discussed in more
detail in the next two sections.55,96

4.2.1 Robustness
Robustness is defined as insensitivity to local failure. This definition can
be understood both qualitatively and quantitatively. The latter requires
quantification of the defining terms ‘insensitivity’ and ‘local failure’.
These two specifics, however, are, in the end, not quantifiable by
resorting to first principles of natural sciences or mathematics but

38

Copyright © ICE Publishing, all rights reserved.

 Design against progressive collapse

have to be established in a decision-making process, mostly on a case-

by-case basis. The same holds for the design objectives listed in Section
4.3.2. It thus seems convenient to equate ‘local failure’ to the ‘assumable
cases of initial local failure’ (design objective 2) and to state ‘insensi-
tivity’ if no more than the ‘acceptable total damage’ (design objectives
3 and 4) results. (The term ‘total damage’ refers to the total amount of
all direct and indirect losses.) In this way, the quantitative definition of
robustness is rooted in the design objectives that have to be predeter-
mined in a decision-making process.
According to the definition suggested here, robustness is a purely
structural property in the sense that the cause and probability of local
failure — and thus also the nature, extent, and probability of the trig-
gering accidental circumstances — are immaterial. This is in contrast
to a broader definition of robustness — as is given, for instance, in Euro-
code EN 1991-1-770 — that refers to triggering accidental circumstances.
(Note that Eurocode EN 1991-1-7 uses the term in an undifferentiated
manner and also in the more specific meaning suggested here.) Such a
broader conception is close to the term ‘collapse resistance’ defined
next.

4.2.2 Collapse resistance

Collapse resistance is defined as insensitivity to accidental circum-
stances, that is, to unforeseeable and low-probability events. Again,
the quantitative interpretation of this term resorts to the design
objectives. Just as local failure was equated to design objective 2, the
‘accidental circumstances’ are now equated to the ‘assumable accidental
circumstances or actions’, that is, design objective 1. ‘Insensitivity’ is
quantified by design objectives 3 and 4, as before. The conjunction
with predetermined design objectives also provides a quantitative
basis to the definition of disproportionate collapse that was given in
Section 1.1 in a qualitative manner. A collapse is disproportionate
when assumable accidental circumstances lead to unacceptable total
damage.
Disproportionate collapse is avoided by ensuring collapse resistance.
The latter is a property that depends on both the structure and the
assumable accidental circumstances. In this regard, the structural prop-
erty ‘robustness’ is of particular importance: a robust structure is at the
same time collapse resistant. It would limit the range of design choices
unacceptably, however, if only those structures were permitted that are
clearly robust. Nor is such a limitation necessary, because a non-robust

39

Copyright © ICE Publishing, all rights reserved.

 Progressive collapse of structures

structure — whose response to an assumable local failure is unacceptable

total damage — could possibly be made less sensitive to accidental
circumstances, and thus collapse resistant, by other means, such as
increased structural resistance of so-called key elements. Furthermore,
collapse resistance will not be required for every structure.
The definitions of robustness and collapse resistance suggested here
bring about a clear distinction between the corresponding properties
— which is the essential point. As long as such a distinction is retained,
these properties could be named differently (if no consensus on a
redefinition of the term ‘robustness’ is reached).

4.3 Design criteria

In the assessment of a structural design with respect to progressive
collapse, the following design criteria are of importance:55,92,95
I. design requirements
II. design objectives
III. design methods
IV. verification procedures.
These criteria and their respective subcriteria are presented in Fig. 4.1
and further examined in the following sections. The design criteria
should generally be established according to codified provisions, as
long as these exist and are applicable. If applicable design criteria are
unavailable in codified form, they must be agreed upon on a project-
by-project basis by the contracting and other affected parties. For
major projects, a project-related agreement will probably be the rule
rather than the exception.

4.3.1 Design requirements

First, the design requirements — that is, in particular the question of
whether collapse resistance is necessary — should be clarified. The
necessity depends on the significance of the structure with respect to
the consequences of a collapse. These go beyond the immediate
human and material losses and include indirect consequences, such as
impairment of the infrastructure and of civil and national defence. In
this sense, a storage structure filled with hazardous material can be of
greater significance than a building of the same size. Likewise, a short
motorway bridge can have a greater significance than a much longer
bridge on a secondary road.

40

Copyright © ICE Publishing, all rights reserved.

 Design against progressive collapse

Design requirements Significance

Exposure

Design objectives Assumable accidental circumstances or actions

Assumable cases of initial local failure

Acceptable extent of collapse

Acceptable other damage

Partial safety factors/combinations of actions

Design methods Prevent local failure

Specific local resistance

Non-structural protective measures

Assume local failure

Alternative load paths

Isolation by segmentation

Prescriptive design rules

Verification procedures Exact procedures

Nonlinear dynamic analysis

Simplified procedures

Quasi-static analysis

Generalised single-degree-of-freedom system

Fig. 4.1 Design criteria for collapse resistance of a structure

Another criterion for establishing the design requirements is the

degree of exposure of the structure to accidental circumstances, such
as deliberate damage and assault (malicious action), accidents, and
human error. Regarding malicious action, a structure can be considered
exposed when it is large and vulnerable, is perceived as important, or
symbolises a major institution. The exposure to malicious and war

41

Copyright © ICE Publishing, all rights reserved.

 Progressive collapse of structures

action is particularly high for landmark structures, public buildings,

major bridges and tunnels, and other lifeline structures.
For secondary structures (i.e. those of low significance and exposure),
collapse resistance can be dispensed with, in particular when the
additional cost would be disproportionate. In the General Services
Administration (GSA) guidelines,75 a procedure called the ‘exemption
process’ is presented through which buildings with a low probability of
collapse or of low significance — this includes small buildings and
those with low occupancy — are exempted from further consideration.
In addition to answering the question of whether collapse resistance
is necessary, the significance and exposure of a structure can serve to
establish design and verification requirements that are more differen-
tiated. This leads to a classification with three or more classes of
structures. In this context, the term ‘importance factor’ appears in the
Multihazard Mitigation Council report.43 Such differentiation
resembles the ‘importance classification’ used in the seismic design of
bridges as set out in the American Association of State and Highway
Transportation Officials specifications.63 Annex A of Eurocode
EN 1991-1-770 provides a classification of buildings and stadia according
to ‘consequences classes’. These structures are classified into four groups
in an attempt to judge their significance with regard to the con-
sequences of a collapse. Class 1 structures, such as agricultural buildings
and single-occupancy houses not exceeding four storeys, do not require
an additional assessment or specific measures with respect to their
collapse resistance. For structures of higher classes, assessment and
measures are required to varying degrees. The British Approved Docu-
ment A,67 though not identical, here closely mirrors the Eurocode.
In the classifications according to Refs 67, 70, and 75, indirect
consequences for infrastructure and national defence and the degree
of exposure are not explicitly considered. These documents cover
solely buildings and stadia. A classification that factors in the degree
of exposure and indirect consequences for infrastructure, civil defence,
and national defence is given in an American Society of Civil Engineers
(ASCE) security guideline.97

4.3.2 Design objectives

If collapse resistance is deemed necessary, the following design
objectives must be established:
1. assumable accidental circumstances or actions
2. assumable cases of initial local failure

42

Copyright © ICE Publishing, all rights reserved.

 Design against progressive collapse

3. acceptable extent of collapse

4. acceptable other damage
5. applicable partial safety factors and combinations of actions.

The term ‘assumable’ means ‘to be assumed in the assessment of a

structural design with respect to progressive collapse’. Establishing the
assumable accidental circumstances or actions (design objective 1)
and the assumable cases of initial local failure (design objective 2)
requires a specification of their respective nature and extent. Accidental
circumstances include malicious action, war action, natural disasters,
fire, accidents, human error, and other hazards. They can involve
actions, such as explosive blast, vehicle impact, or simply static over-
loading, as well as a lack of structural resistance due to design and
construction flaws, corrosion or other causes. When choosing assumable
cases of initial local failure, the extent of initial failure occurring in each
case must also be defined. The cause of failure is irrelevant.
Note that design objective 3 refers to collapse and not merely to
failure. Design objective 4 refers to damage to the non-collapsed
remaining structure, even if extending to virtual failure and loss of
serviceability, damage to the surroundings, and indirect losses resulting
from an impairment of infrastructure and of civil and national defence.
Design objectives 3 and 4 are included in the term ‘acceptable total
damage’. The term ‘total damage’ thus refers to the total amount of
all direct and indirect losses. It is of particular interest that the design
objectives are established deterministically; the respective probabilities
of occurrence thus remain undefined.
It is anticipated that these design criteria can, at best, only partly
be derived from first principles. Quantifying assumable accidental
circumstances or actions that result from such circumstances (design
objective 1) is particularly difficult. The quantification of design objec-
tives 3 and 4 is not actually an engineering problem, but all the more
challenging. Establishing the acceptable total damage can be facilitated
by a cost-benefit analysis; however, this tends to become controversial
when human life is at stake. Advice can be taken from the insurance
industry, which has developed methods and parameters for such con-
siderations. Even so, there will remain the necessity for a judgement
and decision-making process when assets, the environment or human
life are affected on a large scale. That process can be supported by
professionals, but, in the end, must reflect public opinion.
Design objectives 1 to 4 can be understood as performance criteria to
be given by the client. Progressive collapse is a concern mostly for major

43

Copyright © ICE Publishing, all rights reserved.

 Progressive collapse of structures

projects, which means that public interest is usually affected. The term
‘client’ should then be taken in a more general sense as the public-
decision makers who represent public opinion as well as the parties
affected by the project. The decision of the client must be supported
by engineering judgement, especially concerning design objectives 1
and 2. To establish design objective 1, a risk assessment should be
performed. Annex B of Eurocode EN 1991-1-770 provides some
guidance in this regard, in particular a list of potential accidental
circumstances. Nevertheless, one must select from a multitude of
possible scenarios. That selection can only be done on a project-by-
project basis. It is difficult to generalise and to codify, and, therefore,
requires professional competence and experience.
The verification of collapse resistance can be carried out by verifying
robustness, that is, by invoking design objective 2 instead of design
objective 1. A definition of design objective 1, therefore, is unnecessary
in some cases. This facilitates the process and increases its reliability
because design objective 2 can often be established for a given project
by engineering judgement in a meaningful and safe manner — in the
case of the Confederation Bridge, for instance, as the failure of a
bridge pier (see Section 5.3.3). If both objectives are specified, the
assumable local failure (design objective 2) must not be smaller than
the damage resulting from assumable accidental circumstances
(design objective 1). When only design objective 1 is specified, design
objective 2 can be derived from it by structural analysis. This would
be necessary for the identification of key elements, given that it requires
a definition of the assumable cases of initial local failure (see Section
5.1).
Design objectives 1 to 4 can be specified in a continuous numerical
manner or as discrete conditions. The latter are often more meaningful
and better accommodate engineering judgement because structural
particularities can be considered. For instance, instead of specifying
the acceptable extent of collapse (design objective 3) as an overall
numerical ratio to the size of the structure, one could instead define
the location of collapse boundaries that constitute interception lines
preventing the collapse progressing further. This approach lends itself
especially to establishing the design objectives for a particular given
project. Such a more specific and descriptive definition also facilitates
the design process. Similar comments apply to design objective 2, as
already mentioned in the previous paragraph. Furthermore, it is noted
that design objectives 1 to 4 may have to be revised after examining
the design methods to be considered for a given project. That is, the

44

Copyright © ICE Publishing, all rights reserved.

 Design against progressive collapse

final specification of these design objectives can depend on the available

design options (see Section 5.3.3). Finally, it should be mentioned that a
further differentiation of these design objectives would be possible. They
could be given as more than one set of data or in an interrelated form,
for instance, with an acceptable total damage that depends on the
extent of initial local failure.
Design objective 5 is not actually a performance criterion. It is
included in the above list since its specification is likewise needed for
the design process. Reliability theory can be of assistance for this
specification, but here also engineering judgement will, in the end, be
decisive. This seems unavoidable but admissible. For large and unique
structures, which are the main concern here, achieving a uniform
level of safety is illusory (see Section 3.1). What is required, however,
is not necessarily a uniform level but a reasonable level of safety.
When applying the design method ‘specific local resistance’ (see Section
4.3.3), and when deriving design objective 2 from design objective 1 by
structural analysis, increased partial safety factors should be used (see
Section 5.2.3); when applying the design strategy ‘assume local failure’,
reduced partial safety factors are justifiable (see Section 5.4).
The concept presented here accommodates the need for public debate
by making the choices transparent. The design criteria for the Confedera-
tion Bridge, for instance, specified the acceptable extent of collapse as a
certain number of collapsed spans with a combined length of about 700 m
out of a total bridge length of 12 910 m.31,33 The only alternatives to
specifying a certain acceptable extent of collapse would have been to
disregard the possibility of progressive collapse in the design process or
to abandon the project altogether. These three choices are clear even
to a lay person. A societal consensus or an administrative or political
decision in regard to such choices is easier than in regard to choosing
an abstract mathematical quantity such as the safety index — even if
that consensus or decision means that certain construction projects are
better not pursued and some structures remain unbuilt.
If the design objectives 1 to 5 are viewed as a framework whose
content needs to be specified further, note that some of these speci-
fications are already being discussed43,93 or are in use. Eurocode
EN 1991-1-770 provides details on accidental actions (design objective
1). For some specific accidental circumstances, that is, internal explo-
sions and the impact of road vehicles, rail traffic, ships, or helicopters,
the corresponding accidental design actions are specified. Comprehen-
sive provisions on the impact of road vehicles on structures are also
given by a Swiss guideline.98

45

Copyright © ICE Publishing, all rights reserved.

 Progressive collapse of structures

Also with regard to design objectives 2 and 3, some specifications can

be found in the codes. Assumable cases of initial local failure in buildings
are treated in detail in the GSA guidelines75 and in the United Facilities
Criteria.76 Annex A of Eurocode EN 1991-1-770 specifies design
objective 2 by requiring that ‘the building should be checked to
ensure that upon the notional removal of each supporting column
and each beam supporting a column, or any nominal section of load-
bearing wall . . . (one at a time in each storey of the building) . . . any
local damage does not exceed a certain limit’. It remains unspecified
whether the ‘notional removal’ should be done slowly or suddenly,
that is, whether it should be examined by static or dynamic analysis.
The recommended limit of local damage — which corresponds to
design objective 3 — is ‘15% of the floor, or 100 m2, whichever is greater,
in each of two adjacent storeys’. The corresponding provision in
Approved Document A67 is similar but more narrowly limits the
acceptable damage to ‘15% of the floor area . . . or 70 m2, whichever
is smaller’. The corresponding limit set forth in the GSA guidelines is
likewise narrower than that of Eurocode EN 1991-1-7.
The standards and guidelines just mentioned mainly apply to
buildings. The Post-Tensioning Institute (PTI) recommendations64
state that ‘All cable-stayed bridges shall be capable of withstanding
the loss of any one cable without the occurrence of structural
instability’. Furthermore, loss of cable is to be assumed as sudden so
that an impulsive dynamic loading is produced. Design objective 2 is
thus explicitly defined as the sudden failure of any one cable at a
time. Given that an associated load combination and applicable partial
safety factors are stated, design objective 5 is also specified. The
definition of design objective 3, on the other hand, is less clear. If ‘struc-
tural instability’ is understood as a total collapse, then a certain extent of
local damage and large deformations would still be acceptable. Similar
recommendations are made in Ref. 65.
The acceptable damage to the non-collapsed remaining structure — a
constitutive part of design objective 4 — has not yet been specified in
codified form. This specification is related to the repairability of the
structure and the timescale for recovery of its serviceability. If public
interest is not affected, the decision on these issues can be left to the
owner (i.e. the client in the more specific sense). In regard to indirect
losses, however, the decision on the level of acceptability will generally
lie with public decision-makers.
Details on partial safety factors and combinations of actions (design
objective 5) can be found in Refs 64, 75, and 76.

46

Copyright © ICE Publishing, all rights reserved.

 Design against progressive collapse

4.3.3 Design methods

The following design strategies and methods to ensure collapse
resistance are described in the literature, and have at least partially
made their way into design codes:
1. prevent local failure
1.1. specific local resistance
1.2. non-structural protective measures
2. assume local failure
2.1. alternative load paths
2.2. isolation by segmentation
3. prescriptive design rules.
The two strategies specified in points 1 and 2 (design strategies 1 and
2, respectively, as they are called here for short) lead to so-called direct
design methods that start from given design objectives and, except for
non-structural protective measures, are based on structural analysis.
Such methods are performance-based approaches. Design strategy 1
is pursued by providing an increased level of safety against local failure
to the key elements of a structure. Design strategy 2, on the other hand,
is based on verifying and limiting the structural response to assumed
local failure. Only design strategy 2 aims at verifying or increasing the
robustness of a structure. The application of prescriptive design rules
is an indirect design method. It does not rely on the verification of
design objectives, nor does it require structural analysis. The design
strategies and methods are further discussed and compared in detail
in Chapter 5.
Applicable standards and guidelines, such as the Canadian code73 or
the ASCE 7-02 guideline,61 often focus on prescriptive design rules that
are intended to provide continuity and ductility, assuming that also
robustness or collapse resistance is thus enhanced. More detailed
specifications can be found in the GSA guidelines75 and in the
United Facilities Criteria,76 where prescriptive design rules are given
but also certain cases of assumable local failure are specified aiming at
the provision of alternative load paths (see also Section 4.3.4).
Eurocode EN 1991-1-770 classifies the approaches for handling acci-
dental design situations into strategies based on identified accidental
actions and strategies based on limiting the extent of local failure
(caused by unidentified accidental actions). Each of these two strategies
can be pursued by three different methods. One of the methods for
handling identified accidental actions is providing robustness — a struc-
tural property that remains undefined and unquantified. One method

47

Copyright © ICE Publishing, all rights reserved.

 Progressive collapse of structures

for limiting the extent of local failure is to design key elements for a
specified accidental design action — which aims not at limiting but at
preventing local failure. The scheme thus appears inconsistent, or at
least poses some questions. Nevertheless, the design strategies and
methods listed above are all addressed in one way or the other,
except for ‘isolation by segmentation’. The applicability of the various
methods is related to the consequences classes mentioned in Section
4.3.1 above.
The Eurocode strategy based on limiting the extent of local failure
aligns with Approved Document A.67 The latter refers to BS 595068
and BS 811069 for the design of steel and concrete structures,
respectively, concerning details of prescriptive design rules, checking
and limiting the effect of local failure, and providing specific local resis-
tance to key elements. The differences between the current European
and UK standards on accidental actions and progressive collapse are
summarised in Ref. 71.
Although the design method ‘isolation by segmentation’ is not yet
explicitly mentioned in standards or guidelines, it has already been
used in bridge design,33 and can be considered a fully fledged and gener-
ally applicable design alternative, as outlined in previous publications by
the author.55,99 Instead of segmentation, the term ‘compartmentalisa-
tion’ was used there. In this book, the term ‘segmentation’ is preferred,
as explained in Section 5.3.3.

4.3.4 Verification procedures

The assessment of the collapse resistance of a structure and the
application of direct design methods for ensuring this property require
suitable verification procedures. The structure is to be examined by
these procedures to evaluate its sensitivity to accidental circumstances.
The general approach follows from the quantitative definition given in
Section 4.2.2 that is rooted in the design objectives presented in Section
4.3.2. If assumable accidental circumstances or actions according to
design objective 1 lead to unacceptable total damage according to
design objectives 3 and 4, the collapse is considered disproportionate,
and the finding would be ‘susceptible to collapse’. Otherwise, the
structure is collapse resistant, and verification is complete. Given the
deterministic nature of the design objectives, the analyses required for
verification are also performed deterministically.
If collapse resistance is alternatively verified by verifying robustness,
the structure must notionally be modified by inflicting local failure

48

Copyright © ICE Publishing, all rights reserved.

 Design against progressive collapse

according to design objective 2, and its response to this must be

examined. Local failure can be modelled by the removal of one or
more load-bearing elements, by cutting connections or cross-sections
of members, or by releasing individual structural constraints. The ques-
tions about the actual cause and probability of local failure are irrelevant
to the assessment of robustness. Thus, increasing the level of safety
against local failure is also irrelevant to this assessment (unless this
increase is also provided to elements not assumed to fail in a particular
case of initial local failure). On the other hand, this approach eliminates
the need to make statements about the nature and extent of assumable
accidental circumstances. As noted before, this means a substantial
simplification, given the intrinsic difficulties in making such statements.
Even if the cause and probability of initial local failure do not need to
be specified, the assumable cases of initial failure must be properly
chosen and modelled realistically. The governing choice will not
normally be obvious at the outset. The various cases of initial failure
included in design objective 2 must thus be examined one after another.
The initial local failure can be modelled in various ways, as specified
above. In any case, the removal of elements or similar modifications of
the structure should normally be implemented as a sudden event and
examined by dynamic analysis. Assuming a delayed instead of a
sudden initial failure often, but not always, leads to a smaller dynamic
response of the structure (a counter-example is given in Section
5.3.3). If such a delay is justified and quantified based on design objec-
tive 1, it can be considered in the dynamic analysis, and a reduced
dynamic response possibly results.
The response of the structure to the static and dynamic actions
(removal of elements or structural constraints, permanent action,
and, possibly, parts of non-permanent action) must be examined as
realistically as possible. In view of the effects that are typically involved
and need to be considered — such as inertia forces, large deformations
and displacements, inelastic material behaviour, and damage, as well
as separation, fall, and impact of structural components — this examina-
tion generally requires a geometrically and materially nonlinear dynamic
analysis in the time domain. In particular cases, it can also be performed
by model or full-scale tests (see Ref. 30 for an example). The more
realistic and accurate the examination, the more confidence can be
had in the finding ‘collapse resistant’.
Such detailed investigation places high demands on the expertise of
the professionals in charge and on the modelling effort. The use of
simplified verification procedures should therefore be considered,

49

Copyright © ICE Publishing, all rights reserved.

 Progressive collapse of structures

although their accuracy must be checked. Simplification can be achieved

by assuming linear material behaviour or by performing less sophisticated
analyses. Dynamic action, for instance, could be handled in a simplified
way, and large deformations could be neglected. A nonlinear dynamic
analysis in the time domain can, in some cases, be replaced by a quasi-
static analysis that is performed in a linear or nonlinear fashion. In this
case, the static force in an element or constraint being suddenly removed
is multiplied by a dynamic amplification factor and then applied as a static
load. The dynamic amplification factor can possibly be determined using a
generalised single-degree-of-freedom system, if necessary resorting to
response spectra (see Section 5.3.3 for an example).
The accuracy of simplified verification procedures must be validated by
comparative numerical examination with more accurate methods. Such
examination is done for particular structures. The validation, therefore,
applies to other structures only if they are sufficiently similar to the
structures being examined. This, in turn, suggests that a classification of
structures with respect to the accuracy and admissibility of simplified veri-
fication procedures should be developed. Such a classification can be based
on the governing types and mechanisms of collapse (see Chapter 2). For
instance, simplified verification procedures that resort to quasi-static
analysis might be admissible for structures that are prone to zipper-type
collapse. They are more difficult to justify for structures whose collapse
behaviour is affected by impact forces, that is, structures that are
susceptible to pancake-type or domino-type collapse. When simplified
verification procedures whose accuracy has not been validated are used,
confidence can only be had in the finding ‘susceptible to collapse’.
Analysis-based but simplified verification procedures are described in
the GSA guidelines.75 Their application is limited to buildings where no
complex system response is expected. Otherwise, a more sophisticated
investigation, such as by nonlinear dynamic finite element analyses, is
required. The United Facilities Criteria76 distinguish design and verifi-
cation requirements according to basically three levels of protection. For
the lowest level, only prescriptive design measures are required. For the
other levels, alternative paths must be provided. In doing so, three
different kinds of analysis can be performed, namely linear static,
nonlinear static, or nonlinear dynamic. Static analysis has to be
performed for a factored load combination specified in such a way
that the analysis is effectively similar to a quasi-static analysis with a
dynamic amplification factor of 2.0. Requirements put on nonlinear
dynamic analysis are explained. It is not defined, though, which kind
of analysis is to be used in which case.

50

Copyright © ICE Publishing, all rights reserved.

 Design against progressive collapse

The PTI recommendations64 state that in the design of cable-stayed

bridges the effect of cable loss can be determined by performing a quasi-
static analysis with a dynamic amplification factor of 2.0. Alternatively,
a dynamic amplification factor determined by nonlinear dynamic
analysis can be used. These recommendations are further discussed in
Section 6.2.3. Eurocode EN 1991-1-770 gives guidance on the kind of
analysis to be used for determining the effect of vehicle impact. This
refers to the structural response to specific accidental actions and not
to initial local failure.

51

Copyright © ICE Publishing, all rights reserved.

 5
Design methods

5.1 General
Design measures for ensuring collapse resistance can be classified into
indirect and direct methods (see Section 4.3.3). Indirect design relies
on prescriptive design rules such as, for instance, the requirement of
concrete reinforcement continuity. Direct design is a performance-
based approach. It is more sophisticated in that it is based on the
verification of design objectives and requires structural analysis. In
the following, both approaches are presented, starting with the direct
design methods, which are the main focus here.
Direct design methods can be derived from the definition of collapse
resistance given in Section 4.2.2 that refers to the design objectives
outlined in Section 4.3.2. If assumable accidental circumstances or
actions lead to unacceptable total damage, then the structure is
deemed susceptible to collapse, and the design objectives are not met.
This deficiency can be corrected by pursuing one of two different
strategies. The design measures can be based either on increasing the
level of local safety and thus enhancing the effort to prevent an initial
local failure (design strategy 1) or on limiting the total damage following
an assumed local failure (design strategy 2). The latter requires verifica-
tion of and design for local failure events or ‘load cases’.
If design objective 1 (i.e. the assumable accidental circumstances or
actions) has been specified, both routes are viable. If design objective 1
has not been specified — in which case design objective 2 (i.e. the
assumable cases of initial local failure) must at least be established —
then only the second option remains: limiting the consequences of
local failure.
For both design strategies, the identification of the so-called key
elements is important. A key element is a structural member (or a part
of the structure) that meets two conditions: first, it is not larger than
the structural part assumed to initially fail at a time (design objective
2); second, its failure, if no countermeasures are taken, results in

52

Copyright © ICE Publishing, all rights reserved.

 Design methods

Fig. 5.1 Structural design concept of the Alfred P. Murrah Federal Building
(simplified)

unacceptable total damage (design objectives 3 and 4). Thus, by

referring to the design objectives, a quantitative and clear definition
is also obtained for the term ‘key element’. Applying this definition
generally involves structural analysis that should meet the quality
requirements outlined in Section 4.3.4. In some cases, key elements
can be identified by inspection.
Further discussion will be illustrated with examples. The Alfred P.
Murrah Federal Building (Oklahoma City, USA) partially collapsed in
1995, following the detonation of an ammonium nitrate—fuel oil
mixture equivalent to 1800 kg of TNT inside a truck in front of the
building (see Fig. 5.19).9,10 According to the structural design concept,
every second exterior column, here referred to as a secondary column,
was indirectly supported by a continuous transfer girder that ran across
the face of the structure at the second floor (Fig. 5.1). The other
exterior columns, here referred to as main columns, extended down
to the foundation.
For the sake of argument, it is assumed that the detonation caused
the immediate destruction of only one main column (Fig. 5.2). (In
reality, further main columns might have been directly affected.9)
The structure was not able to redistribute the force carried by that
column to the neighbouring main columns. The failure of the transfer
girder and the failure of the adjacent secondary columns resting on
the transfer girder (Fig. 5.3) led to a collapse progression that, possibly
promoted by impact loading and horizontal forces (see Section 2.2.6),
spread through the neighbouring spans and eventually affected a
major part of the structure.
For the Alfred P. Murrah Federal Building, the design objectives
could have been specified, for instance, as follows:

53

Copyright © ICE Publishing, all rights reserved.

 Progressive collapse of structures

Fig. 5.2 Detonation and destruction of one main column

. Design objective 1: blast loading corresponding to the size of an

explosive charge and the standoff distance as can be determined
according to, for example, the American Society of Civil Engineers
(ASCE) security guideline,97 or accidental design actions as
specified in, for example, Eurocode EN 1991-1-7.70
. Design objective 2: sudden failure of a main column.
. Design objective 3: no collapse progression, that is, no failure of
neighbouring main columns and no fall of floors.
. Design objective 4: any damage to the non-collapsed remaining
structure as long as safety and physical integrity of persons are
not compromised; light damage to the surroundings; light indirect
losses.
. Design objective 5: increased partial safety factors when using the
design method ‘specific local resistance’; reduced partial safety factors
when using the design strategy ‘assume local failure’; the combination
of actions, in the latter case, consists of the sudden loss of a structural
member, permanent loads, and reduced non-permanent loads.

Fig. 5.3 Failure of the transfer girder and secondary columns

54

Copyright © ICE Publishing, all rights reserved.

 Design methods

Fig. 5.4 Specific local resistance

Impact loading does not occur as long as design objective 3 is met. It is

assumed here that, with these specifications, the lower part of each main
column (i.e. the section between the foundation and the transfer girder)
becomes a key element.

5.2 Prevent local failure

5.2.1 Specific local resistance
Design strategy 1 (prevent local failure) centres on the structural
members identified as key elements. Their safety against initial failure
is increased. This is preferably achieved by designing the key elements
for the actions specified by or resulting from design objective 1. They
are thus provided with specific local resistance. In the case of the
Alfred P. Murrah Federal Building, this would mean that the lower
parts of the main columns are designed for blast loading, or the
corresponding accidental design actions, and are thus made particularly
resistant (Fig. 5.4).

5.2.2 Non-structural protective measures

If specific local resistance cannot be achieved or would require dis-
proportionate effort, the safety against failure of key elements can also
be increased by non-structural protective measures that avert harmful
events or attenuate detrimental actions (event control). Such measures
can include limitation or control of public access, barriers to resist
vehicle impact and to increase the standoff distance for reducing blast
loading from deliberate attack, sacrificial elements to mitigate blast
loading, and other protective measures. Again, the measures to be

55

Copyright © ICE Publishing, all rights reserved.

 Progressive collapse of structures

Fig. 5.5 Bollards protecting a building entrance against vehicle intrusion and
increasing the standoff distance (Canary Wharf, London). (Photograph: Stuart
Alexander)

taken are guided by the assumable accidental circumstances or

actions specified as design objective 1. The example shown in Fig. 5.5
illustrates that protective measures can be designed to be attractive
and unobtrusive.

5.2.3 Discussion and further remarks

Ensuring high safety against initial local failure (design strategy 1)
requires more than the use of high design loads or recourse to protective
measures. A local failure can also be caused by occurrences such as
corrosion or fire — events that are more effectively counteracted by
corrosion protection, regular inspection, fire protection, and fire-
fighting systems (which, in a wider sense, are also non-structural
protective measures) than by increasing design loads.
In general, the effectiveness of design strategy 1 eventually depends
on the completeness and validity of the accidental circumstances and
actions specified as design objective 1. Since these concern events
that are unforeseeable or occur with very low probability, completeness
and validity cannot be verified, and thus may not be ensured. This
uncertainty must be countered by encouraging, and enforcing as

56

Copyright © ICE Publishing, all rights reserved.

 Design methods

much as possible, a high level of safety consciousness throughout all

stages of a project.
Soil investigation, structural design, and construction must meet
enhanced quality requirements when key elements are concerned.
Possible hazards and all possible kinds of detrimental accidental circum-
stances should be identified using foresight and imagination. This
applies, in particular, to events that are associated with accidents or
malicious action, but also to local lack of resistance due to human error.
Because of both the inventiveness and fallibility of human nature,
foresight and imagination are particularly important for tracing those
detrimental circumstances that result from human action. Part of
such an approach, for instance, is to put oneself in the mindset and
position of potential aggressors, taking into account their intentions
and resources. Structural engineering knowledge on the part of the
aggressor must be assumed as a given. Tracing the corresponding
hazards is therefore not only the responsibility of risk consultants but
also of structural engineers.
The development and codification of accidental design actions to
account for accident-related or malicious circumstances are incomplete.
Further effort is necessary both in terms of accidental circumstances and
types of structures to be covered. Accidental design actions should also
be differentiated according to the criticality of the affected structural
members, that is, they should be increased for key (i.e. critical)
elements. This also applies to the ordinary design loading conditions,
such as dead load, live load, wind, snow, and earthquake, which, with
low but not zero probability, can occur with more severity than usually
considered. For bridges and other structures over water, this list further
includes loads from ship collision, waves, current, and drifting ice. The
influence of the criticality of the affected structural member can be
factored in by differentiating the partial safety factors on the load side
accordingly (design objective 5). The partial safety factors for material
and soil resistance should likewise be increased for the key elements,
which can include certain areas of the subsoil.
The construction stages are of particular importance, although it is
difficult to cover them in a standardised way. Temporary bracing,
auxiliary piers, and other construction devices can now also become
key elements. Instead of specifying accidental design actions, an overall
increase of safety factors might be more expedient. Alternatively, the
requirement can be made to consider specific accidental actions, such
as falling construction equipment or collapsing formwork, and to
design the key elements for the loading resulting from such occurrences.

57

Copyright © ICE Publishing, all rights reserved.

 Progressive collapse of structures

If such an approach is impracticable, the site safety requirements for the

construction stages identified as critical can be raised. In any case, the
site manager should be informed and aware of the key elements,
identified during design, that exist in the various construction stages
and the completed structure.
It must be noted that although high safety against initial local failure
can be achieved, this safety cannot be absolute, and in the face of
unknown future events may not even be as high as hoped for. Since
design strategy 1 does not explicitly aim at increasing robustness, a
local failure that, despite all provisions, does occur can still cause
collapse. Nevertheless, its application is justified in certain cases, in
particular when the significance and exposure of the structure are not
extremely high, the assumable accidental actions are relatively small,
or when the methods based on design strategy 2 (see Section 5.3) are
inapplicable — provided the key elements are clearly and fully identifi-
able. Design strategy 1 can be cost-effective as long as the combined
size of all key elements is small compared with the size of the structure.
Small to medium-sized buildings with transfer girders providing indirect
load transfer, for instance, can be made sufficiently collapse resistant
through local strengthening of the main columns (see Section 5.2.1) if
the alternatives ‘modification of structural system’ or ‘strengthening of
transfer girder’ (see Section 5.3.2) are unsuitable for architectural or
other reasons. Likewise, for small to medium-sized bridges up to, say,
1000 m in length, providing high safety against local failure might be prefer-
able to the ‘segmentation approach’ (see Section 5.3.3). For structures
characterised by series primary load transfer, the applicability of the
methods based on design strategy 2 can be impaired, in which case collapse
resistance might better be achieved by following design strategy 1, in
particular by providing specific local resistance to key elements (see Section
6.3). However, if design strategy 1 is applied to structures of extremely high
significance and exposure, such as, for instance, the Freedom Tower, New
York, or the Messina Bridge, Italy, an informed societal consensus must be
reached not only on the underlying design objectives but also on the
acceptability of the residual risk of collapse. A systematic comparison of
design strategies 1 and 2 is given in Section 5.4.

5.3 Assume local failure

5.3.1 General
Absolute safety against initial local failure cannot be achieved. Design
strategy 1 (prevent local failure), discussed in the previous sections,

58

Copyright © ICE Publishing, all rights reserved.

 Design methods

Fig. 5.6 Column failure introduced as load case ‘local failure’

should therefore be applied to extremely significant and exposed struc-

tures only when other methods are inapplicable, or in conjunction with
other methods. For other types of structures, a high level of safety against
local failure might be achievable, and the residual risk acceptable, but the
resulting costs of construction and protection are disproportionately high.
At least in such cases, it is better to strive for collapse resistance not by
high safety against local failure but by taking the alternative route, if
viable, offered by design strategy 2 (assume local failure).
Accordingly, initial local failure as specified by design objective 2 is
postulated, and it is ensured, during design, that the ensuing collapse
and damage do not exceed the limits set forth by design objectives 3
and 4. If the key elements have already been identified in accordance
with Section 5.1, the further design and verification process can be
based on the assumed failure, one at a time, of the key elements only.
In this way, collapse resistance is indirectly verified by verifying robustness.
In the case of the Alfred P. Murrah Federal Building, the lower part of
each main column was identified as a key element. The sudden failure of
each of these elements is to be considered as a triggering event, and
investigated separately as a dynamic load case. The dashed line in
Fig. 5.6 represents one such key element. Verification of robustness
and collapse resistance is deemed complete when design objectives 3
and 4 suggested in Section 5.1 are proven to be met. The verification
procedure and the corresponding structural analysis should meet the
quality requirements outlined in Section 4.3.4.

5.3.2 Alternative load paths

As suggested above, the requirement of ‘no collapse progression’ could
have been adopted for the Alfred P. Murrah Federal Building as design

59

Copyright © ICE Publishing, all rights reserved.

 Progressive collapse of structures

Fig. 5.7 Modification of the structural system

objective 3. This specification calls for limiting the collapse to the initi-
ally failing structural member, that is, to the lower part of a main
column. To achieve this objective, at least two different measures are
available. On the one hand, the indirect support of the secondary
columns by the transfer girder, identified as problematic, could be
replaced by direct support (Fig. 5.7).
In the modified structure, all columns extend down to the founda-
tion, and are main columns. In the case of failure of the lower part of
a column (see dashed line in Fig. 5.7), the upper part of that column
becomes a secondary column supported by the lower girder (i.e. the
girder at the position of the former transfer girder). The effective span
length of that supporting girder after failure initiation is the same as
that of the transfer girder of the undamaged original structure. It will
thus be possible to design and verify the girder for this loading condition.
Although this loading is significantly increased by dynamic action due
to the suddenness of the initial failure, the requirements on the design
are substantially relaxed, on the other hand, when adopting design
objectives 4 and 5 as suggested above. In particular, when damage to
the non-collapsed remaining structure is acceptable as long as human
safety is not compromised, verification can be restricted to the true ulti-
mate limit state. The ultimate load-bearing reserves due to, for instance,
catenary action can thus be fully utilised. Furthermore, reduced partial
safety factors can be used on both the load and the material side.
In the considered case of failure, the supporting girder becomes part
of an alternative load path for the forces otherwise carried by the lower
part of the affected column. The existence of alternative paths will
simply be referred to as redundancy. The specific meaning of this term
in this context is redundancy of the structure with respect to its ability
to carry loads. Redundancy is increased by the design modification just

60

Copyright © ICE Publishing, all rights reserved.

 Design methods

Fig. 5.8 Strengthening of the transfer girder

described. This assessment is, at least in this case, also intuitively

confirmed by the increase in static indeterminacy of the structural
system.
It should be noted that the suitability of this approach depends on the
design objectives. It might be reasonable, for instance, to change design
objective 2 for the modified structure by assuming the initial failure of
not one but three adjacent columns. In this case, the design modi-
fication brings little improvement (if all columns have the same strength
as the main columns of the original structure) or may even be
detrimental and increase the risk of failure progression (if the columns
are designed for smaller forces due to the higher number of columns).
Modifying the structural system might be undesirable, in particular
when there are architectural or serviceability reasons for indirectly
supporting secondary columns on a transfer girder. A second measure
could then be implemented as an alternative: the topology of the
structural system is retained but the transfer girder is strengthened by
designing it for the dynamic loading resulting from the sudden failure
of the lower part of a main column (Fig. 5.8). If the lower part of a
main column fails, the upper part of that column together with the
adjacent secondary columns are then supported by the transfer girder.
An alternative load path is provided, and thus structural redundancy
is increased also by this measure. Note that in this case, however, the
increase in redundancy is not accompanied by an increase in static
indeterminacy. In view of the now fairly large span after failure
initiation, it may prove difficult to strengthen the transfer girder
sufficiently. On the other hand, advantage can again be taken of relaxed
requirements on the design when verifying the transfer girder for this
extraordinary load case.

61

Copyright © ICE Publishing, all rights reserved.

 Progressive collapse of structures

If the transfer girder cannot be strengthened sufficiently, further

design alternatives are available that also retain the topology of the
structural system. Instead of supporting the upper part of the affected
main column by the transfer girder only, the exterior girders above
the transfer girder can be designed to participate in the alternative
load transfer and carry their share of column loading. The entire
structure, and not just the transfer girder, can now redistribute the
forces and bridge the failed member.
Furthermore, the structure can be designed not as an ordinary frame
but as a special moment-resisting frame,10 a concept also used in seismic
engineering. Also in this case, advantage can be taken of a substantial
amount of damage being acceptable in the non-collapsed remaining
structure. Ultimate total collapse will not occur until the plastic
reserves, which are large in such a specifically detailed and highly
redundant structure, are exhausted. Activating the other exterior
girders and designing the structure as a special moment-resisting
frame is also an option when the first approach initially outlined
above is taken, that is, when the structural system is modified.
Alternative load paths can develop in various ways, and there are a
variety of different modes of load transfer and structural members
involved. In the aforementioned case, the flexural capacity of the
main load-bearing elements and, if necessary, the plastic reserves of
the structure are mobilised. In other cases, alternative load paths can
form through the mobilisation of axial or torsional resistance. The
inversion of flexural load transfer, for instance from hogging to sagging
above a failing column, the transition from flexural to tensile load
transfer (catenary action), for instance in concrete floors or in steel
beams, and the transition from plane to spatial load transfer, for instance
in one-way slabs turning into two-way slabs, are all possible. If plastic
reserves are used, a sufficient degree of ductility must be ensured.
Note that the provision of additional alternative load paths, and thus
an increase in redundancy, is often accompanied by an increase
in continuity. As shown in the next section, the terms ‘redundancy’
and ‘continuity’ are nevertheless different and should not be used
synonymously.

5.3.3 Isolation by segmentation

For certain structures, the alternative-paths method will reach its limits.
The Confederation Bridge crosses the Northumberland Strait between
Prince Edward Island and the Canadian mainland. It is a continuous

62

Copyright © ICE Publishing, all rights reserved.

 Design methods

Fig. 5.9 The Confederation Bridge (Canada)

prestressed concrete box girder bridge consisting of 43 main spans of

250 m each and shorter approach spans on either side. The total
length is 12 910 m (Figs 5.9 and 5.10). If this structure were to be
made robust and collapse resistant by providing alternative paths, an
initial local failure to be considered would be the failure of a bridge
pier. This failure event, in turn, would require designing a prestressed
concrete frame with the double-span length of 500 m — arguably a
vain endeavour even if resorting to catenary action. The design
method chosen was to isolate collapsing sections, and thus to limit
the extent of collapse, by a segmentation of the structure.31—34
The specification of the design objectives for this particular structure
illustrates some of the remarks made in Section 4.3.2. The design
objectives were established by the design engineer in consultation with
the owner and the supervising authority, Public Works Canada. In doing
so, the considerations just outlined concerning the available design options
and the idea of isolating an incipient collapse were incorporated.
Accordingly, the sudden failure of a main bridge pier was to be
assumed in the design (design objective 2). For the structural system
Prince Edward Island
CL Navigation
New Brunswick
abutment

abutment
span

EI. 40.80 EI. 59.06 EI. 40.80

EI. 0.00 (CGD)

% +2.1%
% 0.0% –2.1 0.0% +2.1%
–2.3

Bedrock
1320 10 990 600

12 910

Fig. 5.10 Elevation of the Confederation Bridge (not to scale; dimensions in

metres)

63

Copyright © ICE Publishing, all rights reserved.

 Progressive collapse of structures

Fig. 5.11 Assumed cases of initial local failure and the acceptable extent of collapse

of the preliminary design, this led to two different cases of initial local
failure to be considered (Fig. 5.11, pier B or pier C). The acceptable
extent of collapse (design objective 3) was set by specifying the collapse
boundaries (pier D and hinge H1); any collapse starting within these
boundaries must not transgress them. Further design objectives were
not formally established. The verification of the remaining structure
outside the collapse boundaries, however, was restricted to the ultimate
limit state, and reduced partial safety factors were used. Furthermore,
the plastic cross-sectional and system reserves were used. This meant
that a certain level of damage to the parts of the non-collapsed
remaining structures that lie immediately adjacent to the collapsing
section was implicitly assumed to be acceptable (design objective 4).
The segmentation method is related closely to the concept of collapse
isolation. By specifying collapse boundaries, the structure becomes
notionally segmented already. The locations of the actual segment
borders, however, do not necessarily have to match those of the collapse
boundaries. The segment borders are chosen by the design engineer,
and can lie within the collapse boundaries, which are minimum require-
ments established as design objectives. The segment borders are formed
by isolating elements. The structural detail of the isolating elements and
their mode of operation as segment borders still need clarification.
This design method requires the verification of the segment borders
and the remaining structures for the forces resulting from incipient
collapse. Particular attention must be paid to the isolating elements
that form the segment borders. For the Confederation Bridge, further
study showed that they lie in the region between hinge H1 and pier
A and the region between hinge H2 and pier D (Fig. 5.11). Although
the isolating elements are not key elements in the special sense defined
in Section 5.1, their safe performance as collapse-isolating segment
borders is likewise crucial for the effectiveness of the method. They
are therefore referred to here as relevant elements. An obvious design
approach would be to provide these elements with high local resistance.

64

Copyright © ICE Publishing, all rights reserved.

 Design methods

Fig. 5.12 Bending failure at cast-in-place joint

Verifying that their resistance is adequate, however, can be difficult

because of the mostly large magnitude of the collapse-induced forces.
This difficulty is often compounded by uncertainties in the modelling
and analysis process.
This will be illustrated with the example of the Confederation Bridge.
Among other aspects, the response of the remaining structure to the left
of hinge H2 (Fig. 5.11) after a sudden loss of support at that hinge was to
be investigated. In the preliminary design, the initially affected region,
between hinge H2 and pier D, consisted of a 75 m-long cantilever
beam and a 100 m-long precast drop-in girder that was monolithically
connected to the cantilever beam by a cast-in-place joint and continuous
prestressing tendons. Static and dynamic analysis31,33 showed that the
sequence of collapse would be marked by the following distinct events:
. The girder fails in bending under its own weight at the cast-in-place
joint (Fig. 5.12).
. The drop-in girder rotates around this point, remaining connected
to the cantilever through the continuity tendons.
. The free end of the drop-in girder hits the water; the drop-in girder
fails in bending due to the inertial forces induced during its sudden
deceleration.
. Large forces are transmitted to the cantilever during this violent
event; shear failure occurs at the cantilever end.
. The continuity tendons cut through the cantilever bottom slab in
longitudinal direction, thus crippling the bending resistance of
the cantilever.
. The failure progresses throughout the cantilever towards pier D
(Fig. 5.13).
Note that the analysis had to consider not only large deformations
and dynamic forces but also unusual kinds of interaction such as the
drop-in girder hitting the water or the continuity tendons bearing
on the cantilever bottom slab and cutting through it. The results of
further analysis therefore depended more and more on the modelling

65

Copyright © ICE Publishing, all rights reserved.

 Progressive collapse of structures

Fig. 5.13 Failure progression towards pier D

assumptions. The analytical prediction of the further course of failure

was thus deemed to lack the accuracy and certainty required for
design and verification. In any case, it seemed possible that the collapse
could progress into the next span, beyond pier D, and then into the
further spans, similar to the collapse of the Haeng-Ju Grand Bridge
(see Section 1.2). Verification of adequate resistance of the isolating
element was not possible because of both the high loading and the
analytical uncertainties.
Both problems can be solved, or at least mitigated, by selectively
eliminating (i.e. breaking) continuity, by reducing stiffness or through
ductility. By inserting joints, break-away hinges, or structural fuses, or
by providing plastic hinges at the segment borders, the loading on
such isolating elements is reduced and the analysis simplified. The
mode of effectiveness of the isolating elements in this case is not
based on high resistance against forces but on the ability to accommo-
date large deformations and displacements, and the elimination or
reduction of forces or a large energy dissipation capacity thus achieved.
A combination of such measures was used in the design of the Tasman
Bridge, and proved successful (see Section 1.2).
For the Confederation Bridge, it seemed particularly important to
interrupt the continuity of the prestressing tendons to allow for an
early and safe separation of the falling drop-in girder from the remaining
structure, thus avoiding the problems described above. An attempt was
made to design a structural fuse within the cast-in-place joint between
the cantilever beam and the drop-in girder. However, no convincing
way was found to cut the continuity tendons automatically and safely
after the onset of collapse, and so the idea was abandoned. Hence, it
was deemed necessary to eliminate or, more precisely, reduce continuity
at that location.

66

Copyright © ICE Publishing, all rights reserved.

 Design methods

Monolithic connection

H2

Preliminary design

Hinge

H2

Final design
Fig. 5.14 Insertion of a hinge in every second span

This change was brought about by a permanent modification of the

structural system. Compared with the preliminary design, the final
design comprises one additional hinge in every second span of the
main bridge (Fig. 5.14). Each of these additional hinges replaces an
originally planned cast-in-place joint. This means that the drop-in
girder in every second span is connected to the cantilever beams by
hinges at both ends. In the other spans, which also consist of two canti-
lever beams and one drop-in girder each, the drop-in girders remain
monolithically connected to the structure by cast-in-place joints at
both ends.
The prestressing tendons are subject to a natural interruption of
continuity at the hinges. A reliable isolation of local failure is thus
achieved. If the support at hinge H2 is lost, the drop-in girder extending
between the inserted hinge and hinge H2 will fall and separate from the
remaining structure in a predictable way. The disengagement will be
forced and is defined by the geometry of the hinge corbel (Fig. 5.15).

θ
θ

Fig. 5.15 Forced disengagement of drop-in girder at hinge

67

Copyright © ICE Publishing, all rights reserved.

 Progressive collapse of structures

F(t) Fdyn(t)
12 MN
10
Fst
8

Time

Fig. 5.16 Vertical force at cantilever tip during the fall and separation of a drop-
in girder

This is also the case when the hinge bearing is not movable but
fixed. Hinge corbels with fixed or movable bearings are examples of
break-away hinges.
As a further measure, the length of the drop-in girders was reduced
from 100 m to 60 m, to ensure the separation of a falling drop-in
girder before its free end hits the water, and so to avoid the structural
and analytical problems associated with this event. This design modifi-
cation provides two additional benefits for the collapse resistance of
the structure. First, the dynamic loading on the remaining structure
during the fall and separation of a drop-in girder is reduced. Second,
the remaining cast-in place joints are shifted into a position where
the collapse-induced sectional forces from the collapse of an adjacent
span are smaller.
The remaining structure was checked for the impulsive dynamic
loading that acts on the cantilever tip during the fall and separation
of a drop-in girder. This event and the resulting impulsive loading
can now be accurately analysed (Fig. 5.16).31,34
For simplicity and to counter some remaining modelling uncertain-
ties, linear response spectra of a single-degree-of-freedom system were
developed for the acting impulsive loading (Fig. 5.17). From this, an
overall dynamic amplification factor was determined that was used in
a nonlinear quasi-static analysis and verification of the remaining
structure.
The dynamic amplification factor was 2.6. This is larger than the value
of 2.0 that would result for a simple step impulsive loading. The actual
load impulse would be more complex (Fig. 5.16). When the support at
hinge H2 fails, the vertical force on the cantilever tip suddenly drops

68

Copyright © ICE Publishing, all rights reserved.

 Design methods

3.00

2.00
–fdyn(t) = –Fdyn(t)/Fst
max η, min η

1.00

0.00

–1.00
0.00 0.50 1.00 1.50 2.00 2.50 3.00 3.50 4.00
Time: s, period of vibration: s

Fig. 5.17 Impulsive dynamic loading and spectra of extreme positive and negative
responses

to half of its pre-failure value. After rising again during the rotation of the
drop-in girder, it suddenly drops to 0 at the moment of disengagement. If
these two sudden events are tuned to a modal frequency of the remaining
structure, which would be the case here, a resonance amplification of the
structural response occurs.
In the nonlinear quasi-static analysis performed for the Confedera-
tion Bridge, the formation of plastic hinges was deemed acceptable;
plastic cross-sectional and system reserves were used when verifying
the collapse resistance of the structure. The validation of such a simpli-
fied verification procedure with more accurate methods, as suggested in
Section 4.3.4, was not performed because of a tight design schedule.
From today’s perspective, with the computational resources and analysis
programs now available, a more accurate investigation by nonlinear
dynamic analysis in the time domain seems desirable and possible. A
detailed account of the progressive collapse study undertaken during
the design of the Confederation Bridge is given in Ref. 31; abridged
versions can be found in Refs 33 and 34.
This section on the segmentation method will be concluded with
some more-general remarks. Limiting the effect of local failure without
considering the cause or the probability of its occurrence corresponds to
the design strategy ‘assume local failure’. It is ultimately also the basis of
the alternative-paths method, which is frequently mentioned in the
literature and recommended in standards and guidelines.60,67,75 The
same goal can be achieved by segmentation, however, which opens
up a design alternative that has received little attention so far.

69

Copyright © ICE Publishing, all rights reserved.

 Progressive collapse of structures

Segmentation is performed by specifying and implementing segment

borders. The locations of the segment borders are chosen by the design
engineer within the scope of the design objectives. The segment borders
are formed by isolating elements whose isolating effect is achieved by
one of three different modes of effectiveness:
1. the capability to accommodate large forces, that is, by high local
resistance
2. the capability to accommodate large deformations and displace-
ments, that is, by eliminating continuity or reducing stiffness,
which causes a reduction of forces
3. the capability to accommodate large forces and large displacements
at the same time, which usually means by high ductility and a large
energy dissipation capacity.
Segmenting a structure to isolate an incipient collapse has a func-
tional similarity to dividing a structure into fire compartments to
confine a fire to the compartment of origin. For this reason, the term
‘compartmentalisation’ was used in earlier publications by the
author55,99 instead of the term ‘segmentation’ used here. Referring to
the collapse resistance of a structure, however, the term ‘compartmen-
talisation’ appears in the literature with different meanings. The
mention in the Canadian code73 appears to refer to segmentation in
the sense understood here. It is used differently, though, by Corley
et al.10 and in the ASCE 7-02 guideline.61 In these publications, it
denotes a particular kind of residential building structure consisting of
small, box-like compartments made up of monolithically connected
reinforced concrete slabs and walls. Similarly to a moment-resisting
frame, the increased collapse resistance of such a structure would
result from the availability of alternative paths, that is, from redun-
dancy, and not from a deliberate isolation of incipient collapse at
defined segment borders. To avoid confusion, and also for brevity,
this book uses the term ‘segmentation’ to denote the structural design
method discussed in this section.
A local elimination of continuity — in the case of the Confederation
Bridge, by inserting hinges — reduces the degree of static indeterminacy.
The redundancy of the structure was not reduced by this measure,
however, because it did not remove any feasible alternative load path
— keeping in mind that designing a 500 m span, even though a
theoretical option offered by the original structural system, was
impracticable. On the other hand, the robustness of the structure,
that is, its insensitivity to local failure, and so its collapse resistance

70

Copyright © ICE Publishing, all rights reserved.

 Design methods

were increased. This shows that associating the degree of continuity or

static indeterminacy with redundancy and equating redundancy with
robustness or collapse resistance, even if possible for particular types
of structures, is not generally justified. It is therefore important not to
confuse all these terms, as is sometimes done, but to distinguish between
them carefully.

5.3.4 Redundancy versus segmentation

Apart from the Confederation Bridge and the Tasman Bridge (see
Section 1.2), the segmentation approach has so far only been mentioned
sporadically in the literature72,73 and without discussing the mode of
operation and the structural details of the segment borders. Neverthe-
less, other past applications in bridge design are probable, but may have
gone unpublished. The application to the design of the Confederation
Bridge was substantiated by the difficulty of providing alternative load
paths. This approach may still be preferable, though, even when
alternative paths can be provided. To word this more strongly, the
increase in continuity that often accompanies the provision of alterna-
tive load paths may, in certain cases, promote rather than prevent
failure progression.92,99
This view is supported by eye-witness accounts of fire fighters and
controlled demolition experts who have observed the collapse of
buildings.43,52 According to these accounts, local failure may actually
pull down a greater portion of the building when structural components
such as frames or diaphragms are tied together too well. Such observa-
tions seem plausible, considering that failure progression requires a
certain level of interaction, which, in turn, could mean a certain
degree of connectivity and continuity, between structural components.
In the light of these considerations, the Ronan Point incident, an
often cited example of progressive collapse, can be interpreted differ-
ently. Triggered by a gas explosion in one of the upper storeys, one
corner of the building collapsed over nearly its entire height. The
larger part of the building, however, remained undamaged (Fig. 5.18).
The progressive collapse of floor slabs in this incident has been ascribed
to a lack of continuity in the slab reinforcement. On the premise that
the collapse of a major part or the whole of a building must be
prevented, however, and contemplating the segmentation approach,
such a lack of continuity does not seem so unreasonable after all. At
least it would be an alternative in a design where alternative paths
cannot be provided safely.

71

Copyright © ICE Publishing, all rights reserved.

 Progressive collapse of structures

Fig. 5.18 Partial collapse of the Ronan Point building triggered by an internal gas
explosion (London, 1968)

Stimulated by the Ronan Point incident, requirements for enhanced

structural continuity have been included in building codes (e.g. Refs 67,
70, 73, and 74) in the form of prescriptive design rules (see also Section
5.5). These provisions are intended to lead to an overall increase in
robustness and collapse resistance. They appear to have been successful
in the UK against the kind of event experienced in the IRA bombing
campaigns. Nevertheless, they may not be appropriate for other kinds
of events and other types of structures. In any case, the objective of
increased robustness and collapse resistance is not achieved if the
alternative load paths resulting from enhanced continuity become
overloaded.
In this context, an observation made by Corley et al.9 concerning the
partial collapse of the Alfred P. Murrah Federal Building is of interest. If
only one main column was destroyed by direct blast action (one of the

72

Copyright © ICE Publishing, all rights reserved.

 Design methods

Fig. 5.19 Partial collapse of the Alfred P. Murrah Federal Building (Oklahoma
City, 1995)

possibilities discussed in Ref. 9), it is argued that the two neighbouring

main columns could have been pulled down by their connections to the
falling structural components in between them (see also Section 2.2.6).
This assessment is supported by the fact that the collapse stopped at
another main column located next to a discontinuity in the top
reinforcement of the transfer girder (Fig. 5.19). Regarding the actual
mechanism of collapse of the Alfred P. Murrah Federal Building, a re-
evaluation recently presented by Osteraas100 is of interest: this study
confirms the possibility that only one main column may have been
destroyed by direct blast action.
Two more examples where structural segmentation accomplished by
discontinuity has possibly prevented widespread disproportionate
collapse are the Pentagon Building in Washington, DC, and the Charles
de Gaulle Airport Terminal in Paris. The Pentagon Building consists of
three nested building rings, each divided into five segments, and thus
made discontinuous, by expansion joints. On 11 September 2001, the
aircraft impact near an expansion joint caused the destruction of several
columns on both sides of the joint.101 The more affected part of the
outer ring, to the right of the joint (Fig. 5.20), partly collapsed while
the less affected part, to the left of the joint, did not. Continuity instead
of joints would in all likelihood have promoted a collapse progression,
since the part to the left of the joint was heavily damaged as well and
would not have been able to carry additional loads. The isolation
of the collapsing section at the other end was achieved by strong
structural components with high local resistance that withstood the

73

Copyright © ICE Publishing, all rights reserved.

 Progressive collapse of structures

Fig. 5.20 Partial collapse of the Pentagon Building (Washington, DC, 2001)

collapse-induced forces and likewise acted as isolating elements and

formed segment borders.
The partial collapse of the Charles de Gaulle Airport Terminal 2E was
initiated by the failure of a portion of the roof, possibly owing to design
flaws and poor workmanship (the investigation into what caused the
incident is ongoing). The collapse came to a halt at joints on both
sides of the collapsing section, which was (unintentionally) isolated
from the adjacent sections by these structural discontinuities (Fig.
5.21). If continuity had been provided, it seems unlikely that the
forces that occurred during collapse could have been supported by
the adjacent sections since these were structurally deficient as well. A
further progression of the collapse would then have been possible.
Finally, the Tasman Bridge failure incident is again considered. The
bridge partially collapsed due to the impact of a ship (see Section 1.2).
The collapse was limited to the three spans that were supported by the
two bridge piers directly destroyed by the ship impact. The absence of
collapse progression apparently resulted from the interruption of conti-
nuity of the superstructure prestressing tendons at the bridge piers, that
is, from discontinuity and segmentation.
The potential value of continuity will not be called into question. It
must be kept in mind, however, that continuity can be harmful when

74

Copyright © ICE Publishing, all rights reserved.

 Design methods

Fig. 5.21 Partial collapse of the Charles de Gaulle Airport Terminal 2E (Paris,
2004). (Source: AP Photo. Photograph: Jerome Delay)

the alternative load paths that result from continuity are not provided
with the strength required to withstand the actual forces occurring in
an incipient collapse. This remark also applies to prescriptive design
rules that aim at providing alternative paths through an overall increase
in continuity (see Section 5.5). If it is impossible or prohibitively
expensive to provide alternative paths with sufficient strength, the
segmentation method — implemented, if necessary, by selectively
eliminating continuity — has the advantage.
This is also the case if alternative paths are strong enough, but the
corresponding structural analysis and verification prove difficult or
unconvincing. Such a problem can result from a variety of reasons:
high structural complexity, large displacements, unusual kinds of inter-
action, or high dynamic forces (see the discussion concerning the
Confederation Bridge in Section 5.3.3). Incidentally, a similar state-
ment can be made with respect to the three different kinds of isolating
elements used in the segmentation method: eliminating continuity,
when compared with the alternatives of providing high local resistance
or large energy dissipation capacity, will usually have the advantage of
facilitating analysis and verification.
The alternative-paths method, on the other hand, has the advantage
when the fall of structural components or debris must be prevented by

75

Copyright © ICE Publishing, all rights reserved.

 Progressive collapse of structures

any means — provided that alternative paths can be made available at all
and shown to be sufficiently strong. This applies particularly to cases in
which falling parts could lead to impact loading on key elements of the
remaining structure. This is because such loading is usually high, and
designing for it is difficult, afflicted with uncertainties, and normally
out of the question. Such conditions are typically found in structures
that are aligned along a vertical axis and susceptible to pancake-type
collapse, that is, in high-rise buildings. Impact loading is less of a
concern in structures mainly aligned along a horizontal axis such as
bridges and low-rise buildings (see Chapter 6 for a more differentiated
discussion). The suitability of the two design approaches compared in
this section thus depends on the type of structure and its orientation
in space.
The alternative-paths method requires an increase in either, or both,
continuity and structural resistance. Segmentation, on the other hand,
can be accomplished by less continuity or more resistance. Other differ-
ences between these two methods concern the spatial distribution of the
resulting design measures, the extent of collapse, and the dependency of
their effectiveness on the size of the initial local failure.
The alternative-paths method leads to design measures that are distrib-
uted throughout the structure; the extent of collapse increases and the
effectiveness of the method decreases with an increase in initial failure
size. It is therefore preferable when the size of the initial failure is assumed
to be small. The segmentation method requires design measures at
discrete locations within the structure; the extent of collapse and the
effectiveness of the method are insensitive to the size of the initial failure
(provided the segments are not too small); the extent of collapse,
however, is relatively large. This method is thus preferable when the
initial failure size is also assumed to be large. These dependencies indicate
that the choice between the two design methods compared here in the
end also depends on the given design objectives.
Furthermore, both methods can be combined in different ways. If the
structure is segmented and, additionally, alternative paths are provided
within the segments, then the extent of collapse will not be substantially
larger than a triggering initial failure for both small and large initial
failure sizes. This allows more differentiated design objectives to be
considered in which the assumable initial local failure and the accept-
able extent of collapse are given as more than one pair of data or in
an interrelated form. Another way of combining both design methods
is to resort to alternative paths for the design and verification of the
remaining structure outside a collapsing section that is isolated by

76

Copyright © ICE Publishing, all rights reserved.

 Design methods

segmentation. Such an approach was implicitly taken in the design of

the Confederation Bridge when using the plastic cross-sectional and
system reserves for the verification of the remaining structure (see
Section 5.3.3).

5.3.5 Interaction and progressive collapse

As stated in the previous section, failure progression requires a certain
degree of interaction between structural components. Because interaction
is of fundamental importance to the phenomenon of progressive collapse,
and a complex phenomenon in itself, a separate section is dedicated here
to this topic.102
Interaction is a term originating from sociology and psychology, and
refers to the interdependency between acting parties. When adopting
the term for technical systems, interaction denotes the interdependency
between the various components of a system and between a system and
the environment. Considering interaction according to this definition is
nothing new in structural engineering. The influence of the structural
members on each other in a statically indeterminate structure, for
instance, is taken into account as a matter of course. Another example
at an even more elementary level is the dependency of the applied loads
on the structural dimensions, which is likewise considered in design.
Nevertheless, interaction is usually accounted for in an incomplete
manner. On the other hand, many kinds of interaction are negligible
or difficult to quantify. It is thus essential in structural design to consider
all possible kinds of interaction, to discern the important from the
unimportant, and to describe the important ones with sufficient accu-
racy. In addition to this aspect of examination and analysis accuracy,
there is another problem that is more fundamental and therefore
more interesting, namely the question of whether and where interaction
is beneficial or harmful and whether it should be promoted or prevented
by design measures.
In progressive collapse studies, all of the important kinds of
interaction within a structure and between the structure and the
environment must similarly be adequately considered to ensure the
accuracy of examination and analysis. Compared with more ordinary
structural design tasks, however, the calculability and predictability of
the structural response can be compromised. This difficulty, in turn,
can have an indirect effect on the design.
These problems result from the fact that the interactions that develop
during a collapse can far exceed those that occur under normal

77

Copyright © ICE Publishing, all rights reserved.

 Progressive collapse of structures

construction and operation conditions. The structural analysis must

now consider many and unusual kinds of interaction, such as the partial
or complete separation of structural components and their fall and
impact on the remaining structure or the surroundings. This also
includes unusual kinds of action on the structure and material (e.g.
cutting action on a reinforced concrete slab, and lateral action on
prestressing steel: see below) and non-structural action — in particular,
fire. When considering such interactions, specifics must be taken into
account that are otherwise irrelevant or at least do not occur all at
the same time: dynamic action and behaviour, large deformations and
displacements, thermal action and behaviour, plastification, and partial
destruction of the structure.
For complex structures, an accurate and reliable analytical coverage
of all these possible kinds of interaction will only succeed, if at all, by
performing a geometrically and materially nonlinear dynamic analysis
in the time domain. If there remains any doubt concerning the validity
of analysis and verification, it may be necessary to reduce excessive
interaction to a manageable level by changing the design.
For illustration, reference is again made to the design of the
Confederation Bridge (see Section 5.3.3 and Refs 31 and 33). There
was some doubt as to the interaction of the continuous prestressing
tendons with the enveloping reinforced concrete box girder that
forms the cantilever. When studying a certain failure scenario and
analysing the corresponding sequence of collapse, it was assumed that
the tendons come to bear directly on the bottom slab of the box
girder, cut through the slab in the longitudinal direction, and thus elimi-
nated the bending resistance of the cantilever (see Figs 5.12 and 5.13).
The assumption that such cutting would occur, and not be prevented
by the rupture of the tendons due to the unusual lateral loading, is justi-
fied by the collapse of the Haeng-Ju Grand Bridge. In that incident,
most of the superstructure tendons resisted the stresses caused by the
rupture of the encasing reinforced concrete and the failure and fall of
structural components (see Section 1.2). A quantitative description
of such a process, however, is difficult. For the Confederation Bridge,
at least, it could not be achieved with the required accuracy and
certainty. It remained unclear how much energy would be dissipated
by the cutting of the box girder bottom slab, where this cutting and
the destruction of the cantilever would come to a halt, and, therefore,
what the effect would be on the next span.
By inserting hinges (see Fig. 5.14), and hence interrupting the
continuity of the tendons, the interaction described above between

78

Copyright © ICE Publishing, all rights reserved.

 Design methods

the tendons and the bottom slab was eliminated. Accurate analysis and
a reliable verification of collapse resistance were thus made possible.
The length of the drop-in girder (i.e. the section between the two
hinges in the lower part of Fig. 5.14) was reduced to such an extent
(and the position of the hinges was shifted accordingly) that the
drop-in girder separates from the remaining structure (see Fig. 5.15)
before its free end hits the water. This design modification was likewise
based on the intention to avoid a violent interaction (i.e. between the
structure and the water) that is difficult to determine by analysis.
It is of interest that, here, interaction influenced design decisions in
an indirect manner, that is, by compromising predictability. However,
the nature and degree of interaction first of all have a bearing on the
structural response and thus also directly relate to design issues, in
particular to the choice between the alternative-paths method and
the segmentation method.
While both approaches arise from the assumption of an initial local
failure, they follow contrary lines of thought. Alternative paths create
redundancy through the ability of redistributing forces. This approach
usually requires a certain degree of structural continuity and takes
advantage of the interaction between structural members. The attempt
to isolate collapsing sections by segmentation, on the other hand, aims
at reducing interaction. In this approach, isolating elements prevent the
interaction of the collapsing section with the remaining structure. The
reduction of interaction becomes particularly apparent when continuity
is eliminated by the insertion of hinges or joints.
Concerning the design of robust and collapse-resistant structures,
each of these two contrary design methods will find its respective
fields of application (see Section 5.3.4). Thus, depending on the
design task at hand, either a reduction or an increase in structural
interaction can be reasonable (provided excessive interaction does
not compromise predictability). The overall goal of the design measures
must be to promote beneficial interaction and to prevent harmful
interaction.
Incidentally, there are kinds of interaction concerning progressive
collapse that are beyond the usual structural engineering considera-
tions. The question of whether collapse resistance is necessary for a
given structure, for instance, must be decided after also considering
the exposure of the structure to circumstances such as malicious
action and war (see Section 4.3.1). From a technical standpoint,
however, such a relationship between design requirements and
exposure could simply be interpreted as a more general form of the

79

Copyright © ICE Publishing, all rights reserved.

 Progressive collapse of structures

above-mentioned dependency of the applied loads on the structural

dimensions.

5.4 Local failure: prevent or assume?

Direct design methods are based on one of two strategies. They can
either aim at preventing initial local failure (more precisely, making it
less probable) by locally increasing safety (design strategy 1) or at
limiting the total damage that follows an assumed initial local failure
(design strategy 2). Each of these two strategies leads to at least two
design methods, as outlined in the preceding sections.
If collapse resistance is to be achieved by preventing initial local
failure, the design methods ‘specific local resistance’ and ‘non-structural
protective measures’ are available to ensure a high level of safety against
local failure. These methods do not aim at increasing the robustness of
the structure (although locally increasing its resistance may also
increase its robustness). On the other hand, if a local failure is assumed,
the design methods ‘alternative load paths’ and ‘isolation by segmenta-
tion’ can be used, either separately or combined, to make the structure
robust and to limit the total damage to an acceptable extent.
In both cases, the structural behaviour and the safe performance of
key elements (design strategy 1) or other important elements (design
strategy 2) is crucial. The significance and role of these elements exhibit
interesting differences, which allow conclusions to be drawn regarding
the application and evaluation of design methods. The key elements
in design strategy 1 are to be designed for design objective 1, that is,
for assumable accidental circumstances or actions. The relevant
elements in design strategy 2 are alternative load paths or isolating
elements that are to be checked for the actions resulting from incipient
collapse. In the latter case, design objective 1 does not need to be
specified. Hence, the uncertainty associated with such a specification
has no bearing on the effectiveness of methods that are based on
design strategy 2. This justifies the use of different partial safety factors
for the different design strategies, as suggested in Section 4.3.2.
Another interesting difference is that the relevant elements in design
strategy 2, unlike the key elements in design strategy 1, can be chosen by
the design engineer within certain limits. They are defined by the choice
and the corresponding design of alternative load paths or segment
borders, respectively. The extent of this design freedom depends on
the design objectives. The number of relevant elements can be
minimised and be kept comparatively small, in particular when using

80

Copyright © ICE Publishing, all rights reserved.

 Design methods

the segmentation method whose relevant elements are the isolating

elements that form the segment borders.
Both aspects favour methods that are based on design strategy 2 in
terms of the safety-cost ratio.55,95 To illustrate, two fictitious structures
S and R are considered for which the following assumptions are made:
the collapse resistance of structure S is achieved by segmentation, and
that of structure R by specific local resistance; both structures are similar
and suitable in principle for both design methods; the probability of failure
of the isolating elements of structure S, in their role as collapse isolating
segment borders, is as high as the probability of initial failure of the key
elements of structure R. This leads to another tentative assumption:
the probabilities of progressive collapse of structures S and R, and thus
their levels of safety, are of the same order of magnitude.
For structure S, however, this safety comes at a lower cost. This is
because the actions to be applied in the verification of the relevant
elements of structure S are not due to accidental circumstances but
to collapse actions. These can be determined with a relatively high
degree of accuracy and certainty, and therefore be taken into account
with lower partial safety factors. This advantage would also benefit a
fictitious structure A, whose collapse resistance is achieved by alter-
native load paths. Structure S has the additional advantage, not
shared by structure A, that the isolating effect of the isolating elements
can be achieved not only by making them highly resistant but also by
eliminating continuity — a measure that, for the same probability of
failure, is attainable at a lower cost.
It was tentatively assumed in this example that the probabilities
of progressive collapse of structures S and R are of the same order of
magnitude. In reality, however, this probability will mostly be lower
for structure S (and structure A) than for structure R, which further
shifts the safety :cost ratio to the detriment of the latter. This is a
direct result of the underlying design strategies, according to which a
complete collapse of structure S requires the progressive failure of all
isolating elements, whereas a complete collapse of structure R could
be triggered by the initial failure of any one key element.
The mode of global load transfer (i.e. series versus parallel) and its
effect on the probability of a progressive collapse was discussed in
Section 3.1. Similar reasoning can be applied here to substantiate the
assertion of the previous paragraph. For structure R, the probability of
collapse is of the order of the sum of the initial failure probabilities of
all the key elements. This holds, at least, in the extreme case that the
failure of any one key element leads to complete collapse. The

81

Copyright © ICE Publishing, all rights reserved.

 Progressive collapse of structures

probability of collapse, therefore, increases with the number of key

elements, and can become relatively high. For structure S, in contrast,
the probability of collapse is of the order of the product of the failure
probabilities of the isolating elements, in their role as collapse-isolating
segment borders, multiplied by the probability of an initial local failure
anywhere in the structure. Thus, in this case, the probability of collapse
decreases with the number of isolating elements, and hence with the
number of segments, and will be relatively low.
For these reasons, design methods based on the assumption of an
initial local failure (design strategy 2) are preferable for structures of
high significance and exposure, especially if the assumable cases of
initial local failure (design objective 2) can be meaningfully and safely
established by engineering judgement, and provided the structure is,
in principle, suitable for these methods. In comparison with design
methods aimed at preventing initial local failure (design strategy 1),
they provide a high level of safety against progressive collapse at a
relatively low additional cost. Moreover, they are more satisfactory
from a reliability standpoint because their effectiveness is comparatively
insensitive to the probability of failure of relevant elements. Uncertain-
ties associated with accidental circumstances (design objective 1) are
altogether irrelevant. These methods, therefore, remain largely
unaffected by the difficulty of assessing failure probabilities.
On the other hand, design strategy 1, and in particular the specific-
local-resistance method, are of interest for structures that are poorly
suited for other methods. They can be preferable, for instance, for
providing collapse resistance to the primary load transfer system of
slender high-rise buildings, even those of high significance and exposure
(see Section 6.3.2). Furthermore, these methods can be preferable for
smaller structures, for structures with only a few and clearly identifiable
key elements, and in cases where the extent of assumable accidental
circumstances is small. More remarks on the application of design
strategy 1 can be found in Section 5.2.3.
The relationship between design strategies 1 and 2 is in some respects
analogous to the relationship between the alternative-paths method
and the segmentation method. As outlined in Section 5.3.4, these
two design methods differ with respect to their sensitivity to the size
of the initial local failure (design objective 2). When comparing
design strategies 1 and 2, their different sensitivities to accidental
circumstances (design objective 1) instead become apparent. The
choice between these two design strategies will again be influenced by
the design objectives.

82

Copyright © ICE Publishing, all rights reserved.

 Design methods

Again, design strategies 1 and 2, and their respective design methods,

can favourably be combined in different ways. While the primary load
transfer system of a slender high-rise building is possibly made collapse
resistant by specific local resistance (R) (a method in design strategy 1),
the design of its secondary load transfer system can be based on
providing alternative paths (A) or by segmentation (S) (methods in
design strategy 2) (see Section 6.3.2). When all three methods are
combined in the design of the same system or subsystem, dispropor-
tionate collapse can be averted for small (R), medium (A), and large
(S) accidental circumstances or initial local failure.

5.5 Prescriptive design rules

The design methods discussed above require detailed examination. The
response of the structure to accidental actions or to the removal of
selected structural elements is analytically (or, in exceptional cases,
experimentally) determined and compared with the design objectives
(direct design). Such a direct approach demands a great deal of time,
skill, and computational resources on the part of the design engineer.
For small to medium-sized structures, such a level of commitment
may become disproportionate. If, consequently, a detailed examination
is dispensed with, it may nevertheless still be desirable to achieve at least
a certain degree of collapse resistance, preferably by following prescrip-
tive design rules that are codified and simple to use (indirect design).
Such an indirect approach does not rely on an analytical verification of
specified design objectives, nor can it take design objectives into account.
Prescriptive design rules are mostly simple and all-encompassing direc-
tions concerning structural detailing. To what degree collapse resistance
is achieved remains unclear. If this approach proves unsuccessful, on the
other hand, the consequences are less serious, as long as its application is
limited to structures of relatively minor significance.
For building design, a number of prescriptive design rules have emerged
and made their way into standards and guidelines.61,67—70,73—76 An over-
view is given by Dusenberry and Juneja.72 The three most common rules
are:
1. providing tension ties
2. enabling catenary action
3. ensuring ductility.
The main structural members are required to be tied together
horizontally and vertically by tension ties. According to the American

83

Copyright © ICE Publishing, all rights reserved.

 Progressive collapse of structures

Concrete Institute Committee 318 requirements for the design of

reinforced concrete buildings,74 a certain portion of top and bottom
reinforcement in perimeter and spandrel beams is to be made contin-
uous to provide a continuous tie around the structure and enhance
its so-called ‘overall integrity’ (i.e. its robustness and collapse resis-
tance). Precast concrete elements are to be connected by tension ties
in all directions and around the perimeter of the structure to effectively
tie elements together. Eurocode EN 1991-1-770 requires horizontal and
vertical ties for the transfer of accidental design tensile forces, which are
derived from the ordinary design loads. Similar provisions can be found
in Approved Document A67 when read in conjunction with the
applicable British standards.68,69
By enabling catenary action, a transition from flexural load transfer to
tensile load transfer becomes possible. It can be activated in both beams
and slabs and in both steel and concrete structures. In the event of
failure of an intermediate column, catenary action allows the failed
column potentially to be bridged, and thus a failure progression to be
averted. Arguably, its most beneficial effect is in preventing the fall of
failed structural components. If, for instance, a floor slab is destroyed
in an incipient collapse, the debris is held in place by the tensile
forces within the sagging remnants of the slab. A pancake-type collapse
resulting from the impact of falling debris on the floor below is thus
avoided. Catenary action is enabled by providing for horizontal conti-
nuity within and between flexural members. In concrete members,
this is done by making the top and bottom reinforcement continuous.
The continuity of the bottom reinforcement is required because the
top reinforcement over a failing column tends to tear out of the
concrete.74 As an alternative, additional untensioned steel cables
within or outside floor slabs have been proposed by Astaneh-Asl.103
Ductility in structural members and connections is achieved through
proper detailing. In the event of a local failure, it enables plastic cross-
sectional and system reserves to be used and energy to be dissipated.
Stimulated by the ductility requirement, the suitability of seismic
design rules for design against progressive collapse has been discussed
by Hayes et al.48 and Corley.104
All prescriptive design rules, as far as specified in design codes, aim at
ensuring collapse resistance through continuity and the provision of
alternative load paths. Therefore, and in light of the discussion in
Section 5.3.4, these rules should be applied with prudence and only
to structures within the scope of the respective standard or guideline
(i.e. to particular types of buildings and construction methods). This

84

Copyright © ICE Publishing, all rights reserved.

 Design methods

is because prescriptive design rules may increase the number of potential

alternative paths but do not guarantee their quality, that is, they have
sufficient structural resistance. Thus, if potential alternative paths are
not provided with the strength required to effectively withstand and
transfer the forces occurring in an incipient collapse, then a collapse
progression is possibly not prevented but, rather, promoted.
When using direct design methods, such a possibility is identified and
excluded since the transfer of forces assigned to alternative paths is veri-
fied down to the foundations. Such a verification should preferably also
be performed when alternative paths are provided and forces are
assigned to them implicitly by prescriptive design rules. When tension
ties, catenary action, plastic hinges, etc. are provided through pre-
scriptive design, the forces whose transfer is to be verified should be
determined based on the overstrength of the respective tensile elements
or the overstrength plastic moments of the plastic hinges.
The idea of isolating collapsing sections by a segmentation of the
structure has been mentioned sporadically in the literature72,73 but
has not yet led to any prescriptive design rules. Such rules could be
useful when safety concerns, as outlined above with respect to prescrip-
tive design aiming at continuity and alternative paths, are more easily
met by specifying segmentation. This is possible, in particular, if the
isolating effect of the isolating elements is achieved by eliminating
continuity. For the development of prescriptive design rules based on
segmentation, the scope of application, the appropriate size of the
segments, and the detailing of the isolating elements would have to
be investigated and specified.
For structures of high significance and exposure, direct design
methods are generally preferable to prescriptive design rules. They
take into account the structural diversity and complexity inherent
in such structures and ensure collapse resistance in a more or less
reliable and verifiable manner. Moreover, not only safety but also
cost-effectiveness can be impaired by applying prescriptive design
rules. This is certainly true for large, unique, and expensive structures.
In such cases, it is altogether more economical to spend effort and cost
on detailed study and a design adapted to the conditions of the given
project, that is, on direct design.

85

Copyright © ICE Publishing, all rights reserved.

 6
Applications

6.1 General
In Chapters 4 and 5, an approach to the handling of questions arising
from the phenomenon of progressive collapse was presented, and
methods for the design of collapse-resistant structures were discussed.
Both direct design methods and prescriptive design rules were examined.
Direct design is more sophisticated in that it is based on design objectives
and structural analysis. Greater emphasis was put on the direct design
methods. They were illustrated by examples of building and bridge
structures. For the justification and comparison of these methods, failure
incidents and project-related design tasks were considered.
The suitability of this approach will now be tested through a
systematic application in two areas of structural engineering, namely
large bridges and high-rise buildings. In Section 6.2, the most important
types of large bridges are examined, and recommendations for the
collapse-resistant design of such structures are derived. A somewhat
different goal is pursued in Section 6.3, which is dedicated to the
design of high-rise buildings. Only extremely tall and slender structures,
comparable, say, to the collapsed towers of the World Trade Center
(WTC), are considered. The possibility of designing such structures
to be collapse resistant are explored using the tools developed in
Chapters 4 and 5. Economic aspects are treated here as secondary to
the question of the possibility, in principle, of achieving a collapse-
resistant design.
Common features of large bridges and high-rise buildings are high
significance and exposure, large internal forces, and a high degree of
structural interaction. Prescriptive design rules cannot normally meet
the particular demand for collapse resistance, and do not properly
account for the structural diversity and complexity inherent in such
structures. In addition, large bridges and high-rise buildings are gener-
ally unique and expensive. In terms of both safety and economy,
direct design methods are therefore preferable for such structures over

86

Copyright © ICE Publishing, all rights reserved.

 Applications

prescriptive design rules (see Section 5.5). The study which follows in
Sections 6.2 and 6.3 is therefore limited to direct design methods.

6.2 Design of collapse-resistant bridges

6.2.1 General
While there are now a number of design standards and guidelines for
buildings aimed at preventing progressive collapse (see Sections 1.4
and 4.3), almost no such regulations as yet exist for bridges. Notable
exceptions are the corresponding details in the Post-Tensioning Insti-
tute (PTI) recommendations64 and in Ref. 65, which, however, do
not exhaustively treat the problem and are relevant only for cable-
stayed bridges. Other regulations applicable to bridges are mostly
limited to general safety recommendations105 or the specification of
design actions for accidental circumstances. The focus of the latter is
the impact of road, rail, air, or water vehicles in order to prevent initial
local failure.70,98 Nonetheless, the Swiss guideline on the impact of road
vehicles98 advises choosing structural systems that are robust. The few
non-standard publications addressing the collapse-resistant design of
bridges are mostly limited to the prevention of initial local failure due
to previously disregarded accidental circumstances such as malicious
explosion.28,106,107 Exceptions to this are Refs 30, 31, and 96.
The discussion in Chapter 5, on the other hand, shows that there are
at least four generally applicable direct design methods to choose from
or to combine for the design of a structure against progressive collapse.
Furthermore, it was noted that the suitability of the different methods
depends on the design objectives, and on the type of structure and its
orientation in space (see, in particular, Section 5.3.4).
The understanding of these dependencies is still rudimentary. An
obvious characteristic feature of bridges, however, is that they are
mainly aligned along a horizontal axis. Impact loading caused by falling
structural components or debris is therefore less of a concern for bridges
than for buildings. It follows that in bridges there is less need to provide
alternative load paths and to tie together structural components in
order to prevent their fall and impact. This is just as well as it is usually
difficult to provide alternative paths in structures that are aligned along
one axis, as is the case for bridges.
Other general features and conclusions with respect to progressive
collapse are hard to find. The following discussion is therefore directed
at specific types of large bridges.

87

Copyright © ICE Publishing, all rights reserved.

 Progressive collapse of structures

6.2.2 Continuous girder bridges

Relevant examples of continuous girder bridges are the Viadotto
Cannavino and the Haeng-Ju Grand Bridge, which collapsed in a
progressive manner following an initial failure, and the Tasman
Bridge, where an initial failure did not trigger a progressive collapse
(see Sections 1.2 and 2.2.6). Another example is the Confederation
Bridge, for which the requirement to provide collapse resistance was
part of the design criteria (see Section 5.3.3). Out of this experience
and the preceding discussion in Chapter 5, two suitable design
approaches emerge to ensure collapse resistance in continuous girder
bridges.
The first one is the segmentation method. For the bridge type con-
sidered here, selected piers or the regions around those piers should
be chosen as segment borders. For short spans up to, say, 40 m, it will
be possible to provide these piers or regions with high local resistance
to enable them to act as isolating elements.
A historical example is the Chaumont Viaduct, France, a 600 m-long
railway bridge completed in 1856 consisting of 50 masonry arch spans.
Every fifth pier is designed as an abutment pier, with twice the
cross-sectional dimensions in both directions as the normal piers. This
arrangement facilitated construction because every five-span segment
could be built independently of the others. Even if providing robustness
was not an original design intent, the structure proved to be robust
during World War II, when four piers were dynamited but this
damage did not lead to progressive collapse.
For longer spans, the segmentation method requires selective elimi-
nation of continuity at the envisaged segment borders. This can be
accomplished in a reliable manner by the insertion of break-away
hinges. Further methods to eliminate continuity and to form isolating
elements and segment borders are described in Section 5.3.3.
The second approach aims at preventing initial local failure by
increasing the level of safety of the key elements — either by providing
specific local resistance or through non-structural protective measures.
Thought must be given to possible accidental circumstances that could
endanger key elements. The probability of occurrence of such circum-
stances should be minimised. Actions resulting from assumable acci-
dental circumstances must be considered in design.
Examples include: the impact of a ship or an aircraft on a pier or the
superstructure; unexpectedly strong ice floes striking a pier; scour
undermining a foundation; fire caused by a traffic accident that could
damage, for instance, the prestressing tendons in a prestressed concrete

88

Copyright © ICE Publishing, all rights reserved.

 Applications

bridge; a bomb blast at a vulnerable location; other accident-related or

malicious detrimental circumstances; deficiencies in design or construc-
tion; or, simply, corrosion.
For the design of major bridges over water, the probability of ship
impact and the corresponding design loading are usually evaluated by
specialist consultants. The decision to protect the bridge piers through
strengthening or through external impact-resistant barriers can be based
on such an assessment. The normal design loading for ship impact and
other actions should be increased when designing key elements. The
construction stages must also be considered. Further information on
the application of the design strategy ‘prevent local failure’ can be
found in Section 5.2.3. As noted there and in Section 5.4, the choice
between the two design approaches suggested here depends, in
particular, on the size of the structure and the design objectives.

6.2.3 Cable-stayed bridges

6.2.3.1 General
Cable-stayed bridges are a good illustration of the statement made in
Section 5.3.3 that the terms continuity, redundancy, and robustness
should be carefully distinguished. Although such structures possess a
high degree of static indeterminacy and internal continuity, this does
not mean that they are redundant or robust. The sudden failure of
one or more stay cables and a possibly ensuing zipper-type failure
progression and collapse (see Section 2.2.2) are of particular concern.
The cables should be seen as potential key elements (see Section 5.1)
of this type of bridge.
The problem is aggravated by a number of factors: the cables are often
easily accessible and exposed to accident-related or malicious detri-
mental action (Fig. 6.1); their cross-section is relatively small, which
makes it difficult to provide them with specific local resistance against
lateral action; cable failure could occur nearly instantaneously, and
thus produce an impulsive dynamic loading; unzipping of cables is
likely to interact with and be reinforced by a stability failure (i.e. buck-
ling) of the bridge girder and pylons that are in compression (see Section
2.2.6); and the structure could respond in a non-ductile manner.
To meet such concerns, the PTI recommendations64 require that a
cable-stayed bridge shall be designed so that the sudden loss of a
cable does not lead to collapse. A load combination associated with
such loss-of-cable events or ‘load cases’ and applicable partial safety

89

Copyright © ICE Publishing, all rights reserved.

 Progressive collapse of structures

Fig. 6.1 Exposure of stay cables to accident-related or malicious action

factors are specified. The loss of any one cable shall successively be
considered. Similar recommendations are made in guidelines issued
by the International Federation for Structural Concrete (fib).65 In
contrast to the fib guideline, the PTI recommendations also specify a
verification procedure: the structural response to the impulsive dynamic
loading that results from the sudden failure of a cable is recommended
to be determined in a quasi-static analysis using a dynamic amplification
factor (DAF) of 2.0.
These recommendations have been applied to the design of cable-
stayed bridges erected in the USA over the past years. It was found
that the loss-of-cable events in conjunction with quasi-static analysis
often become controlling load cases in the design of the bridge girder,
thus leading to increased construction costs. The corresponding
clause was therefore amended in the fifth edition of the PTI recommen-
dations:64 the DAF to be used in quasi-static analysis can alternatively

90

Copyright © ICE Publishing, all rights reserved.

 Applications

Fig. 6.2 Structural model for analysis of loss-of-cable events

be determined by nonlinear dynamic analysis; it should not be taken as

smaller, however, than 1.5.

6.2.3.2 Numerical study of loss-of-cable events

The effects of sudden cable failure and the corresponding verification
procedures were studied in a research project supervised by the
author.108—110 Numerical studies were performed for a five-span sym-
metrical cable-stayed bridge with a main span of 600 m, side spans of
210 m, and end spans of 60 m. In this structure, the two cable planes
are anchored along the edges of the bridge girder and slightly inclined
inwards in the transverse direction. They consist of 80 cables each in
a semi-fan configuration. The cable anchors in the bridge girder are
spaced at 15 m. The bridge girder consists of an orthotropic deck
supported on two 2.60 m-deep steel edge beams. The response of the
structure to the failure of any one cable was investigated by nonlinear
dynamic analyses of a spatial model of the structure (Fig. 6.2). Cable
sag, large deflections and their effect on equilibrium were taken into
account. Material nonlinearity was considered for collapse analyses,
which are addressed further below, but not for the computation of
DAFs.
The failure of a cable was modelled as a sudden event, causing a step
impulsive loading, and the ensuing extreme responses of the state vari-
ables (i.e. of the displacements and sectional forces) were determined.
By relating these extreme dynamic responses to the corresponding
static responses to cable failure, DAFs were derived that could be
used in quasi-static analysis. Because the comparison of dynamic and
static responses must be carried out case by case for individual cable
failures and individual state variables, each resulting DAF is valid
only for the case considered, and will generally be different for other
cable failures and state variables.

91

Copyright © ICE Publishing, all rights reserved.

 Progressive collapse of structures

4.0

3.0
DAF
2.0

1.0
Anchor pier Pylon Centre of bridge
0
0 60 270 570
(a)

–30
Bending moment My: MNm

–20
–10
0
10
20
30
Anchor pier Pylon Centre of bridge
40
0 60 270 570
Longitudinal bridge axis: m
(b)

All cables intact

Minimum dynamic response (envelope) Maximum dynamic response (envelope)
Corresponding minimum static response Corresponding maximum static response

Fig. 6.3 (a) The DAF for positive (h) and negative (g) bending moments. (b)
Bending moments in the bridge girder due to permanent loads and extreme bending
moments due to permanent loads plus the governing loss-of-cable load case

The numerical results confirmed that it is impossible to specify an

overall DAF. The DAF, in fact, strongly depends on both the location
of the failing cable and the location and nature of the state variable
considered. Regarding the vertical displacement of the bridge girder
at and in the vicinity of cable failure, the DAF is between 1.5 and
1.8, depending on the location of the failing cable. For the forces in
the cables adjacent to the failing cable, the DAF is between 1.35 and
2.0. It is between 1.3 and 1.6 for the positive bending moments, and
between 1.4 and 2.7 for the negative bending moments, in the bridge
girder in the vicinity of the failing cable (Fig. 6.3). More precisely,
these values refer to the bending moments in the bridge girder edge
beam in the plane of cable failure. For state variables at locations distant
from the failing cable, the DAF can generally be much larger than 2.0.
This, however, simply results from the fact that the corresponding static

92

Copyright © ICE Publishing, all rights reserved.

 Applications

response used as a reference value is usually small at distant locations —

and not from a particular large or governing dynamic response.
The numerical values of DAFs just stated were determined for an
undamped structure. The attenuating effect of damping was found to
be small.
Based on these results, on overall DAF of 2.0 seems appropriate for the
safe evaluation of design cable forces. Regarding the bending moments in
the bridge girder, the DAF is noticeably smaller than 2.0 for a wide range
of locations. Thus, the reduction of design bending moments sought by
the amendment in the fifth edition of the PTI recommendations is
partly possible. On the other hand, a DAF larger than 2.0 is obtained
for the negative bending moments at some locations. It thus seems
necessary to adopt values larger than allowed by the current PTI
recommendations64 for the corresponding design bending moments.
Furthermore, it stands out that the DAFs for the bending moments in
the pylons result much larger (i.e. by one order of magnitude) than 2.0.
The corresponding static responses used as reference values are small.
The situation thus seems to resemble that of state variables in the
bridge girder at locations distant from the failing cable. In contrast to
those findings, however, not only the DAFs but also the absolute dynamic
responses are large in the pylons, and can govern the design (Fig. 6.4).

200 All cables intact

Minimum dynamic response (envelope)
Corresponding minimum static response
Maximum dynamic response (envelope)
Corresponding maximum static response
150
Pylon height: m

100

50

0
–100 –50 0 50 100
Bending moment My: MNm

Fig. 6.4 Bending moments in pylon legs due to permanent loads and extreme
bending moments due to permanent loads plus the governing loss-of-cable load case

93

Copyright © ICE Publishing, all rights reserved.

 Progressive collapse of structures

Again, the results just stated were determined for an undamped

system. The attenuating effect of damping was found to be more
pronounced for the pylon moments than for the other state variables,
but not enough to eliminate the problem. When assuming a damping
ratio to critical of 1%, the design bending moments are reduced by
approximately 15%. These remarkable findings indicate that, for the
bending moments in the pylons due to cable failure, a quasi-static
analysis with a DAF of 2.0 can be unsafe and should be replaced by
proper dynamic analysis.
It was further determined that the sudden failure of one cable
can indeed result in bending moments that govern the design of the
bridge girder but does not lead to collapse (even if loss-of-cable load
cases were neglected in design) nor to particularly high cable forces.
In the case of the simultaneous sudden failure of two adjacent
cables, however, the ultimate load is notionally reached in the
adjacent cables. Nevertheless, when effects such as ductile material
behaviour or non-simultaneity of cable failures are taken into
account, failure progression requires a still larger number of initially
failing cables.
To simulate the actual collapse progression and to explore the
true ultimate capacity of the bridge, the material nonlinearity of all
structural members was incorporated in the analytical model. It
turns out that the simultaneous sudden failure of three adjacent
short cables results in immediate collapse due to buckling of the
bridge girder where it becomes unbraced (Fig. 6.5). Only one
additional cable snaps during buckling, so that there is no pro-
nounced zipper-type failure progression but mainly an instability-type
collapse.
It is maintained, though, that, for other cable-stayed bridge config-
urations, failure could also progress as a mixed-type collapse (see
Section 2.2.6) in which the features of zipper-type and instability-type
collapses interact. Further findings concerning the effect of sudden
cable failure are presented in Refs 108—110. The effect of delayed
(i.e. non-sudden) cable failure is examined in Ref. 110.
All these results were obtained for a long-span multi-cable bridge
with two cable planes and steel girders. It still needs to be examined
to what degree they are representative for cable-stayed bridges in
general. Results of dynamic analyses of cable failure for another,
though similar, cable-stayed bridge are presented in Ref. 111. Instead
of performing a quasi-static analysis with predetermined DAFs, it
might soon become possible and more convenient for the design

94

Copyright © ICE Publishing, all rights reserved.

 Applications

Fig. 6.5 Instability-type collapse triggered by the failure of three short cables
(pylon omitted for clarity)110

engineer to directly determine the design forces by nonlinear dynamic

analysis, at least for selected cases of cable failure.

6.2.3.3 Initial failure of multiple cables

When applying the PTI recommendations,64 the assumable cases of
initial local failure (design objective 2) follow from the condition that
the loss of any one cable at a time shall be considered. The question
arises whether the assumption of just one cable failing at a time is
always sufficient, in particular when the cable anchors in the bridge
girder are closely spaced. Considering traffic accidents as possible trig-
gering events, it seems sensible to assume the sudden and simultaneous
failure of all cables anchored within a segment of certain length placed
anywhere along the bridge girder.
In the design of the Taney Bridge, a recently erected cable-stayed
bridge in Ireland carrying two tracks of the Dublin Light Rail Transit
system, the length of that segment was set at 10 m, measured along
the cable anchors. Accordingly, the sudden and simultaneous failure
of any two adjacent cables needed to be assumed in the design of that
bridge.112 These dynamic load cases were examined by quasi-static
analysis using a DAF of 2.0. They were combined with a specified set
of additional actions, including live load on the track next to the failing

95

Copyright © ICE Publishing, all rights reserved.

 Progressive collapse of structures

cables. A specified set of load and resistance partial safety factors was
used. As a further combination of actions, the failure of any two
adjacent cables was considered statically in combination with live
load on both tracks. This was to account for the possibility that a
second train would cross the bridge soon after the failure of the cables
while the first train is still standing on the bridge. These design criteria
were developed jointly by the design engineer and the owner.

6.2.3.4 Other design methods

Verification of and design for loss-of-cable load cases is an application of
the alternative-paths method, one of the design approaches to providing
collapse resistance discussed in Chapter 5. When examining other
approaches, it turns out that the segmentation method is usually
inappropriate for cable-stayed bridges. Except for multi-span structures,
the minimum segment size, and thus the minimum extent of collapse,
corresponds to the size of the entire bridge (or at least half of the
bridge when a centre hinge is inserted in a three-span structure).
The specific-local-resistance method, too, seems inappropriate, at
least for the currently preferred multi-cable systems with narrow cable
spacing. This is due to the small cross-sectional area of the load-bearing
elements of the cables, their comparatively small resistance to accident-
related or malicious actions in the lateral direction, and their significant
impact on construction costs. This design method is suitable, at best,
after a change of structural system: if the multi-cable configuration
common today is replaced by a few widely spaced large-diameter
cables, these would also exhibit a higher local resistance to lateral
action. On the other hand, such change of system could impair the
formation of alternative paths in the case of cable failure.
Instead of by providing specific local resistance, the design strategy
‘prevent local failure’ can also be pursued through providing non-
structural protective measures. This method is applicable to cable-
stayed bridges, and should be employed more, not as a substitute for
but rather as a complement to the alternative-paths method, that is,
to verifying and designing for loss-of-cable load cases.
Local protective measures include barriers to fend off vehicles, such
as concrete barriers, guard rails, or crash cushions,98 fencing to deter
trespassers from approaching the cables, and measures that protect
against explosions. The last could be a thick composite-material
sheath to mitigate the effect of a shaped charge mounted directly on
a cable. The sheath would have to be applied to the cables up to a

96

Copyright © ICE Publishing, all rights reserved.

 Applications

certain height (say 5 m) above deck level. To prevent sudden or delayed

cable failure due to corrosion, efficient corrosion protection systems for
cables and anchors, and regular inspection, ideally complemented by
constant monitoring, are needed. For particularly significant and
exposed bridges, non-local measures such as aerial surveillance or
anti-aircraft defence may be required. Again, further information on
the application of the design strategy ‘prevent local failure’ is given in
Section 5.2.3.

6.2.4 Suspension bridges

The hangers of a suspension bridge are secondary load-bearing
elements, in contrast to the cables of a cable-stayed bridge, which are
constitutive elements of the primary load transfer system. Nevertheless,
the sudden failure of one or a few hangers would similarly produce an
impulsive dynamic loading on the structure that could lead to zipper-
like failure of further hangers and collapse.113 Such a failure progression
can be seen in the footage of the Tacoma Narrows Bridge collapse in
1940. After the first hangers snapped due to excessive wind-induced
distortions of the bridge girder, the entire girder peeled off from the
hangers and suspension cables (see Fig. 2.2).
In the case of a self-anchored suspension bridge, where the bridge
girder is in compression, any unzipping tendency is likely to be rein-
forced by the susceptibility of the bridge girder to buckling. As found
above for cable-stayed bridges, a mixed-type collapse can thus occur
in which the features of zipper-type and instability-type collapses
interact and reinforce each other.
Progressive collapse initiated by the sudden failure of one or more
hangers seems not to have been a major concern in the past. At least,
it is not regularly checked in design. With the recent advent of new
kinds of threats, a different stance should be taken. The challenge
can be met by making hanger failure less probable and by simultaneously
designing for such a scenario. The design and verification measures for
cable-stayed bridges outlined in the previous section are similarly
applicable.
For earth-anchored suspension bridges, there is a further possible
design approach that is impracticable for cable-stayed bridges and
self-anchored suspension bridges: the bridge girder could be provided
with a number of break-away hinges that give way at the beginning
of collapse and thus allow for a safe separation of the falling part of
the girder. In other words, the segmentation method can be used;

97

Copyright © ICE Publishing, all rights reserved.

 Progressive collapse of structures

the segment borders are formed by the inserted break-away hinges. A

collapse of the bridge girder initiated by hanger failure or another
event would thus be isolated by the hinges and limited to one
segment.
The primary load transfer system of a suspension bridge is formed by
the suspension cables, which generally are continuous over the length of
the bridge. The failure of a suspension cable will usually be equivalent to
collapse. Applying the segmentation method to the primary load
transfer system to confront this possibility is, at best, an option only
for multi-span systems. Each segment would consist of a self-contained
two-span or three-span suspension bridge.
The alternative-paths method, on the other hand, is more generally
applicable, as long as the force carried by a failing suspension cable can
be redistributed to neighbouring cables. To enable such a force redistri-
bution while at the same time limiting the resulting costs, a number of
cables would be required. This design method, therefore, is more
appropriate for large suspension bridges, where more than two cables
are already envisaged or where the number of cables can more easily
be raised without major disadvantages.
Applying the specific-local-resistance method to the primary load
transfer system would require a strengthening of the load-bearing
elements of the suspension cables. This can be an appropriate measure
for small suspension bridges. For large long-span suspension bridges, it is
not a reasonable option, because the design of such bridges is mostly cost
driven, and a major share of construction costs is related to the suspen-
sion cables. On the other hand, the suspension cables of long-span
bridges have large cross-sections and masses already, which provide
them with high local resistance, including against lateral action, even
without further strengthening. Furthermore, their resistance is less
affected by isolated wire breaks because each suspension cable consists
of many wires and even a broken wire will carry load again at a certain
distance from the breakage point.
In view of the emergence of new kinds of threats, however in
particular aggressive and well-resourced malicious action, it is advisable
to take non-structural protective measures to guard suspension cables
against such threats, regardless of cable size. For bridges of high signifi-
cance and exposure, these measures should include appropriate physical
shielding and electronic security systems to reliably deter trespassers
from approaching the suspension cables. This applies especially to
locations where the cables are easily accessible (mid-span) or where
their load-bearing elements are exposed (splay chamber).

98

Copyright © ICE Publishing, all rights reserved.

 Applications

6.2.5 Arch bridges

Arch bridges have similarities to suspension bridges in terms of topology
and the flow of forces. Much of what has been presented in the previous
section can be directly applied or readily adapted to arch bridges.
A through-arch bridge, in which the bridge girder is carried below the
arch, has hangers like a suspension bridge, and the same potential
problem of a zipper-type collapse. Similar design approaches can thus
be used. As for a self-anchored suspension bridge, segmentation of
the bridge girder is not an option for a tied-arch bridge. In a true
arch, the bridge girder is supported above the arch by columns and
not suspended under the arch by hangers. Since the columns, in
comparison with hangers, are more resistant to lateral action and not
exposed to traffic, column failure due to accidents is less probable.
On the other hand, they are also more hidden from public view,
which facilitates malicious action.
Regarding the primary load transfer system, there are two important
differences between the arches of an arch bridge and the suspension
cables of a suspension bridge. First, an arch can suffer global stability
failure, which can occur in the plane of the arch or in the lateral
direction. This opens up further possibilities of failure initiation and
progression.
Second, various kinds of material and cross-sections are possible and
in use for arches. These design options should be evaluated with respect
to assumable accidental circumstances, in particular those that induce
local lateral action. Local failure caused by lateral action can be facili-
tated by local instability (i.e. plate buckling), and can lead to global
instability and collapse. In this regard, solid cross-sections are more
resistant than thin-walled or hollow ones, and reinforced concrete is
preferable to steel. The susceptibility of a hollow steel tube arch to
local lateral action can, on the other hand, be reduced by partially filling
the tube with concrete. The filling can be limited to the regions that are
accessible to traffic or trespassers.

6.3 Design of collapse-resistant high-rise buildings

6.3.1 General
The Federal Emergency Management Agency (FEMA) report on the
collapse of the WTC towers11 concludes with a call for studies to deter-
mine ‘whether there are feasible design and construction features that
would permit such buildings to arrest or limit a collapse, once it

99

Copyright © ICE Publishing, all rights reserved.

 Progressive collapse of structures

began’. An attempt is made here to answer this call. The phenomena to

be dealt with are unforeseeable and low-probability events (accidental
circumstances), local failure, and progressive collapse. The tools for
handling these phenomena were provided in Chapters 4 and 5, which
include the discussion of generally applicable design methods.
These findings are now applied to structures that are comparable to
the collapsed WTC towers, that is, to high-rise buildings with a large
height-to-width ratio, a large number of storeys, and a high degree of
significance and exposure.114 Furthermore, it is assumed that the
design criteria call for assuming substantial accidental circumstances
or major local failure, as detailed later. The study is not limited,
however, to specific accidental circumstances, and its results apply to
all kinds of events, including fire. The primary objective is to identify
design alternatives that increase the collapse resistance as much as
possible and minimise the probability of a progressive collapse — even
if they seem unusual or are more costly in comparison with present
practice.
To this end, each of the four design methods ‘specific local resis-
tance’, ‘non-structural protective measures’, ‘alternative load paths’,
and ‘isolation by segmentation’ is examined in a separate section with
regard to its suitability to provide adequate collapse resistance of the
primary load transfer system, which is responsible for the global load
transfer of all vertical and horizontal loading. Measures and structural
systems that result from this examination are suggested where possible.
Indirect design (prescriptive design rules) seems inadequate for the type
of structure contemplated here, and is not considered further (see
Sections 5.5 and 6.1).

6.3.2 Specific local resistance

6.3.2.1 General
This design method aims at enhancing local safety by increasing the
structural resistance of key elements. These have to be identified first.
In principle, this is carried out based on a structural analysis in which
the assumable initial local failure is modelled, and the ensuing total
damage is then determined and compared with the acceptable total
damage (see Section 5.1).
In this process, both the assumable cases of initial local failure and the
acceptable total damage are design objectives that must first be estab-
lished. Alternatively, the assumable cases of initial local failure can be

100

Copyright © ICE Publishing, all rights reserved.

 Applications

derived by structural analysis from the assumable accidental circum-

stances (see Section 4.3.2). This implies that the assumable accidental
circumstances are clearly defined, for instance as an internal gas explo-
sion, and the corresponding actions can be determined or are directly
given. If the calculated total damage is acceptable, the structural
members initially modelled as failed do not qualify as key elements,
and they do not require specific local resistance. However, if the
acceptable total damage is exceeded, these members are key elements.
They can then be designed for the assumable accidental actions, and
thus be provided with increased structural resistance — again, on the
assumption that the assumable accidental circumstances are clearly
defined and the corresponding actions can be determined or are given.
It will often be difficult, however, to establish and agree on the
assumable accidental circumstances or actions. This applies, in
particular, to structures of high significance and exposure because
malicious action must be included in the list of assumable accidental
circumstances. Malicious action, however, is hard to anticipate due
to the inventiveness that any human aggressor must be assumed to
possess.
In such cases, the initial local failure should be chosen to be as large as
reasonably possible. For the type of structure considered here, this
means the initial failure of the structure is assumed to extend over
the whole of one storey. The adequacy of this assumption is emphasised
by the collapse of the WTC towers. It can be expected that such an
initial failure leads to a pancake-type collapse (see Section 2.2.1) if no
countermeasures are taken. Despite the apparent size of the assumed
initial failure, the collapse would, nevertheless, be disproportionate.
The part of the structure assumed to be initially failing here, that is,
all the vertical load-bearing members of one entire storey, thus qualifies
as a key element. This conclusion holds for all storeys (except for the
uppermost ones, whose failure does not necessarily result in intolerable
impact forces). All these key elements are contiguous. It follows that the
whole primary load transfer system must be provided with specific local
resistance over nearly the entire height of the structure.
The corresponding design ultimately does require a decision on the
assumable accidental circumstances or actions. Independently from
that decision, which must be made on a case-by-case basis, general
conclusions concerning the configuration of the primary load transfer
system can, however, be drawn from the preceding discussion. In
consideration of conceivable accidental circumstances — accidental or
malicious explosion, vehicle or aircraft impact, missile impact and

101

Copyright © ICE Publishing, all rights reserved.

 Progressive collapse of structures

blast, or fire — it is concluded that the structural members composing

that system should be as compact as possible. The surface area exposed
to attack is thereby reduced to a minimum and the resistance against
local action is increased to a maximum.
The design concept presented below results from opting to ensure the
collapse resistance of the primary load transfer system through specific
local resistance and from the ensuing demand for compactness of that
system. In the next section, the primary load transfer system itself is
treated. The other parts of the structure (secondary load transfer
system, façade) are covered in subsequent sections. It is found that
their collapse resistance must be ensured independently of the primary
system. The design approaches to be considered for this task are there-
fore not limited to the specific-local-resistance method.

6.3.2.2 Primary load transfer system

The primary load transfer system (or, for short, the primary structure) is
envisaged as a massive tube forming the vertical spine of the building
(Fig. 6.6). Because it is assigned the entire vertical and horizontal
load transfer, the tube must be larger in its outer dimensions than the
core of a core and outrigger system. To maximise its compactness, the
tube should have as few openings as possible and the openings should
be small. For this reason, the tube does not lend itself to forming the
façade but must be placed inside the outer perimeter of the building.
The usable floor space is located mainly between the tube and the
façade.
The tube consists of reinforced high-strength concrete or of structural
steel members embedded in reinforced concrete. It features a substantial
wall thickness of the order of 1 m or more to ensure adequate resistance
against local action. The wall thickness and the further detailing of the
tube depend on the assumable accidental circumstances and actions
that must be specified on a project-by-project basis.
The number and size of openings are limited to the minimum
necessary to allow access from the vertical lines of traffic and supply
(lifts, staircases, service risers) located inside the tube to the floor
space on the outside. The openings are staggered between successive
storeys, to minimise weakening of the global structural resistance. To
prevent the undetected placing of large amounts of explosives, which
could be done incrementally over an extended period of time, the
inner and outer faces of the tube should be permanently accessible for
inspection or be entirely open to the public.

102

Copyright © ICE Publishing, all rights reserved.

 Applications

Primary load
transfer system

A A

Secondary load
transfer system

Façade

Vertical
section

Section A–A

Fig. 6.6 Primary load transfer system: vertical and horizontal sections

An explosion inside the tube — caused accidentally or by smaller

amounts of explosives brought into the building shortly before detona-
tion — can lead to a substantial pressure build-up within the enclosed
inner space of the tube, a loading for which the tube should be designed.
This can be facilitated by choosing an annular cross-section for the tube

103

Copyright © ICE Publishing, all rights reserved.

 Progressive collapse of structures

Primary load transfer system

Façade

Section A–A

Fig. 6.7 Annular cross-section of the primary load transfer system

(Fig. 6.7). Prestressed reinforcement could be provided for carrying the

sectional forces in the circumferential direction. The pressure build-up
inside the tube could be reduced by permitting venting through the
access openings and by providing additional vents at the top and base
of the building.
The design features of the primary structure just outlined result from
applying the specific-local-resistance method that aims at preventing a
triggering initial failure. Although such a strategy does not exactly
correspond to the call of the FEMA report cited in Section 6.3.1, the
collapse resistance achieved in this way is high. The structure presented
here is deemed to exhibit the maximum collapse resistance achievable
by this design method. In terms of tolerable accidental circumstances, it
is surpassed only by the design alternatives presented in Section 6.3.5.2,
which are based on vertical segmentation. In addition, it is of interest
that the solid and compact structure described here, when compared
with non-compact column—beam structures, also seems more robust,
that is, less sensitive to local failure. This assumption is based on its
lower degree of structuredness, a collapse-promoting structural feature
discussed in Section 2.4.7.

6.3.2.3 Secondary load transfer system

The primary load transfer system is responsible for the global load
transfer of all vertical and horizontal loading. The primary structure
suggested in the previous section performs this task and, moreover,
exhibits a high degree of collapse resistance. In order not to compromise
this desirable property by possibly harmful interaction, the other parts of
the structure should not participate in the global load transfer. The
secondary load transfer system (or, for short, the secondary structure)
is therefore conceived as floors cantilevering out from the vertical
tube that forms the primary structure or spanning its inner faces
(Fig. 6.8).

104

Copyright © ICE Publishing, all rights reserved.

 Applications

Fig. 6.8 Secondary load transfer system

The current practice of floor design using beams or slabs is, in

principle, suitable. However, a self-contained progressive collapse of
the secondary structure is still possible, and must also be prevented.
Such a collapse could occur as a failure of floors that progresses in the
vertical direction without affecting the primary structure. This was
called an alternative mode of pancake-type collapse in Section 2.2.1.
Since the primary structure is independent of the secondary structure
with respect to its collapse resistance, the entire set of design methods
can be considered for providing collapse resistance to the secondary
structure.
Specific local resistance is more difficult to provide than it is for the
primary structure. This is because the elements of the secondary struc-
ture are intended for local load transfer only, and consequently have
smaller cross-sectional dimensions while at the same time they exhibit
large surface areas exposed to attack. Designing the secondary structure
for assumable accidental actions in all likelihood would lead to cross-
sectional dimensions that are impracticable. Because of the large surface
area exposed to attack and the multitude of conceivable accidental
circumstances and actions, non-structural protective measures seem
likewise inappropriate (see also Section 6.3.3). It is therefore better to
accept a local failure of the secondary structure as inevitable but to
limit the ensuing failure progression to an acceptable extent: in other
words, to resort to the design strategy ‘assume local failure’.
If the horizontal extent of the assumable accidental actions cannot be
convincingly limited, initial failure and fall of an entire cantilever floor
should be assumed. The loading of this floor can be analytically applied

105

Copyright © ICE Publishing, all rights reserved.

 Progressive collapse of structures

to the cantilever floor below. Thus, an alternative load path beneath the
initially failing element is provided for that loading. A two-step capacity
design approach is suggested.
Consider a cantilever floor that is deflected by the impact of the
failing floor above (Fig. 6.9(a)). Assume the outer edge of the deflected
floor barely touches the floor below, and that the collapse is arrested in
that position. For this to be a reasonable assumption, it must first be
ensured that the elastic and plastic energy required for the deflection
is larger than the released gravitational energy. The latter approximately
equals the total weight of one floor, including a reduced live load

Failing floor

Plastic hinge

Floor bent down by

the impact of floor above

(a)

Failing floors

Plastic hinges Floors bent down by

the impact of floors above

(b)

Fig. 6.9 Assumed initial local failure and the acceptable extent of collapse of the
secondary load transfer system

106

Copyright © ICE Publishing, all rights reserved.

 Applications

allowance, multiplied by the vertical distance of the floors and 1.5. The
factor 1.5 accommodates the additional gravitational energy released by
the lowering of the deflected floor itself.
Secondly, the rotational ductility capacity in the plastic hinge regions
must be sufficient to allow for such a deformation without a significant
reduction of the bending moment in the plastic hinge. Progressive
collapse initiated by the fall and impact of a floor is thus prevented.
Numerical studies show that the required rotational ductility capacity
is probably difficult to achieve with reinforced concrete beams but
that it is attainable with haunched steel girders.115
The analysis can be refined by assuming initial failure and the fall of
more than one floor, by increasing the weight of the falling floors by the
weight of a possibly incoming aircraft, and by considering larger degrees
of deformation when establishing the elastic and plastic energy contri-
butions. The impact of two consecutive floors above, for instance,
could be assumed to affect two consecutive floors below (Fig. 6.9(b)).
A refined analysis should also account for the interaction between
bending moments and axial forces in the plastic hinge regions. An
incoming aircraft would result in a concentrated additional vertical
load, and thus require sufficient circumferential load distribution
capacity. This can be provided by a perimeter beam, situated at the
outer edge of each cantilever floor, that works in bending and, for
higher loads, is capable of catenary action.
The design approach described here applies to the cantilever floors. A
similar capacity design approach can be developed for the floors span-
ning the inner faces of the tube.
If the horizontal extent of the triggering accidental actions is assumed
to be limited, then the collapse resistance of the cantilever floors can also
be ensured through load distribution in the circumferential direction.
The latter can be enhanced by, again, providing perimeter beams.
Such an approach corresponds to the provision of alternative paths
lateral to the initially failing elements. In this case, the fall of structural
components and impact loading do not need to be considered.
There is an alternative design option, in lieu of providing alternative
paths, to limit the failure progression. Collapsing sections can be isolated
by an appropriate segmentation of the secondary load transfer system.
As outlined in Section 5.3.3, segmentation is achieved by isolating
elements that can accommodate large displacements or large forces,
or both.
For the secondary structure suggested here, isolating elements that
allow for large displacements are relatively easily formed by joints in the

107

Copyright © ICE Publishing, all rights reserved.

 Progressive collapse of structures

Floor segment

Joint

Fig. 6.10 Horizontal segmentation of the secondary load transfer system by joints

cantilever floors. Each floor contains a certain number of joints running

perpendicular to the core (Fig. 6.10). In this way, each cantilever floor
is horizontally segmented and divided into a number of structurally inde-
pendent cantilever floor segments. The joints of all floors are vertically
aligned, that is, placed at the same location in plan. If one floor segment
fails and falls, the ensuing collapse would be limited to the floor segments
located below the failed segment — a partial collapse similar to the Ronan
Point failure (see Fig. 5.18). For this kind of horizontal segmentation, the
horizontal extent of the assumable accidental actions must be limited.
Compared with the alternative-paths and capacity design approach
discussed earlier, the extent of collapse is generally greater.
Both approaches can be combined: the floors can be designed to carry
impact loading following the alternative-paths approach outlined above,
and at the same time they can be segmented by joints, as described in
the previous paragraph. Circumferential load distribution is eliminated
through the insertion of joints. The concentrated vertical aircraft load,
therefore, does not need to be considered in the capacity design of the
cantilever floors, provided that for this accidental action the collapse of
all segments located below a failing segment is acceptable.
The isolation of collapse within the secondary load transfer system
can also be achieved through vertical instead of horizontal segmenta-
tion. In this case, the isolating elements must extend in the horizontal
direction instead of being aligned vertically. Furthermore, they must be
able to transmit large forces instead of allowing for large displacements.
This suggests that a certain number of floors should be selected as
isolating elements and designed as ‘strong floors’ that are able to bear

108

Copyright © ICE Publishing, all rights reserved.

 Applications

the impact loading from failing segments above. In the case of such a
vertical segmentation, the horizontal extent of the assumable accidental
actions does not need to be limited. The idea will be taken on in a
modified form in Section 6.3.5.2, applying it to the primary load transfer
system. Note that the approaches for horizontal and vertical segmenta-
tion of the secondary load transfer system described here can also be
combined with each other.

6.3.2.4 Façade structure

In addition to the primary and secondary load transfer systems, the
façade (or building envelope) structure requires attention. A progres-
sive collapse of the façade structure could remain self-contained or
occur in interaction with the secondary load transfer system.
Windows and glazing, cladding panels, and other façade elements can
be mounted either directly or by means of light framing on the canti-
lever floors (curtain wall). When the collapse resistance of the
secondary load transfer system is ensured by designing all or a few
selected floors for impact loading from failing segments above — two
of the design approaches outlined in the previous section — the contrib-
utory weight of the façade must be taken into account when deter-
mining the released gravitational energy.
For a capacity design of the secondary structure according to Fig. 6.9,
further measures concerning the façade need to be taken. When a
deflecting floor approaches an undeflected floor below, no excessive
forces must be transmitted from the façade to the floor below. This
can be achieved by a vertically overlapping or staggered arrangement
of the façade elements. Alternatively, the vertical force associated
with the destruction of the façade can be considered as a quasi-static
load, together with a dynamic amplification factor, in the elastic
design of the cantilever floors. When achieving collapse resistance of
the secondary structure through horizontal segmentation, the floor
joints that run perpendicular to the core and serve as isolating elements
must be complemented by continuous vertical joints in the façade.
Alternatively, the façade can be designed as a self-supporting struc-
ture. For the type of building considered here, a collapse of the façade
potentially causes great damage and compromises human safety in the
surroundings of the structure. Thus, a self-supporting façade must
usually be provided with collapse resistance of its own. This can be
achieved by providing alternative paths or by isolating collapsing
sections through horizontal segmentation brought about by the

109

Copyright © ICE Publishing, all rights reserved.

 Progressive collapse of structures

insertion of vertical joints. The provision of alternative paths, for

instance by designing the façade as a moment-resisting frame with
sufficiently stiff and strong horizontal spandrel beams, has the advan-
tage of a presumably smaller extent of collapse and other damage.
A self-supporting façade must be structurally connected to the canti-
lever floors of the secondary structure to provide for wind load transfer
and out-of-plane stability of the vertical load-bearing elements of the
façade. Only horizontal forces need to be transferred by these connec-
tions. Nevertheless, such connections could lead to an interaction
between the façade and the secondary structure that compromises the
collapse resistance of the latter. To avoid this, the connections should
preferably be detailed such that only horizontal load transfer takes place
and they freely disengage at a specified relative vertical displacement. A
full connection would also be possible, provided it releases safely at a speci-
fied vertical force (structural fuse). With such decoupling mechanisms,
the collapse resistance of the façade and the secondary structure can
largely be independently ensured by different design methods.
Without a decoupling mechanism and not being designed as a curtain
wall, the façade becomes a load-bearing part of a more complex
secondary load transfer system. It seems better, in this case, to design
for and make use of the vertical load transfer between the façade and
cantilever floors. The floors are then supported at both ends and
work in beam instead of by cantilever action. The collapse resistance
of such a combined secondary structure can be achieved by alternative
paths, by horizontal segmentation through the insertion of vertical
joints, or by a combination of both measures. The structural response
of such combined systems to assumable cases of initial local failure is
more difficult to assess than for a secondary structure that relies solely
on cantilever action.
Consider, for instance, the case that the horizontal extent of
triggering accidental actions and initial local failure is assumed to be
unlimited in such a combined floor—façade secondary structure. Its
collapse resistance is therefore based on the provision of alternative
load paths beneath the initially failing floor.
Accordingly, a capacity design approach similar to that introduced in
the previous section can be developed. In this case, however, the
vertical load-bearing elements of the façade must carry their share of
the forces induced by the fall and impact of one or more floors, a
corresponding part of the façade, and aircraft debris on the floor
below (Fig. 6.11). When the initial failure is assumed to be unlimited
in the horizontal direction and to occur everywhere at the same

110

Copyright © ICE Publishing, all rights reserved.

 Applications

Failing floors and

façade structure

Plastic deformation due

to impact from above

Primary load Secondary load

transfer system transfer system

Fig. 6.11 Assumed initial local failure and the acceptable extent of collapse of a
combined floor—façade secondary load transfer system

time, no alternative paths can be made available within the façade:

the impact forces act on a non-redundant structure, corresponding to
a column under axial compression, and must be transmitted elasti-
cally. The floor design, on the other hand, benefits from an increased
number of plastic hinges, thus requiring less ductility capacity per
hinge.
The floors above the failing floors lose their outer edge support from
the façade. Therefore, they must be able to carry their dead and live
load, as well as the impulsive dynamic loading due to the sudden discon-
tinuance of the outer support, in cantilever action. The primary load
transfer system must be capable of supporting these vertical loads but
also of carrying the concurrent horizontal loads. The distribution of
the impulse dynamic loading to all floors causes short-term vertical
tension in the façade above the zone of failure for which the façade
must be designed. On the other hand, the load-carrying capacity of
the façade in vertical tension can also be used to keep up the outer
edge support of the floors above the zone of failure. The vertical load-
bearing elements of the façade in this region convert from compression
members to hanger-type tension members. The vertical tension in the
façade can, in this case, be transferred to the primary structure by

111

Copyright © ICE Publishing, all rights reserved.

 Progressive collapse of structures

outrigger trusses at the top of the building — a load path that developed
in the towers of the WTC.11,12

6.3.2.5 Overall structure

The design alternatives discussed in the previous sections flow from
the initial choice to provide for collapse resistance of the primary load
transfer system by using the specific-local-resistance method. In
contrast to the other design methods discussed in the following
sections, a hierarchical arrangement of largely independent primary,
secondary, and, possibly, tertiary load transfer systems results from
that choice.
As stated above, the tube constituting the primary structure is placed
inside the perimeter of the building for reasons of serviceability. Such a
location is also advantageous, however, for the collapse resistance of the
primary structure: it is less exposed to actions from impact events or
external explosions because the façade and the floor structure between
the façade and tube serve to mitigate and distribute the resulting
effects.116 This benefit results from energy dissipation and the fragmen-
tation of incoming aircraft or projectiles. Another advantage of such a
hierarchical structural system is a large degree of architectural freedom
in the design and subsequent modification of the façade and floor plan.
This freedom is maximised when the façade is not made a load-bearing
part of a combined secondary load transfer system.

6.3.3 Non-structural protective measures

Non-structural protective measures for increasing the safety against
local failure include barriers to shield against vehicle impact or explosive
blast, control or limitation of public access, and other protective
measures, such as aerial surveillance or the provision of anti-aircraft
systems.
In that the structure must be protected against attacks from both the
outside and the inside, a combination of such measures would be
required. Relying on anti-aircraft defence will not find lasting favour
with those being protected in this way, that is, the users of the building,
who are generally civilians. Control or limitation of public access impairs
the serviceability and the cost-benefit ratio of the building. A large
amount of explosives could be smuggled into the building in small
quantities over an extended period of time. Trying to counter such a
tactic by scrutinising and searching every person entering the building

112

Copyright © ICE Publishing, all rights reserved.

 Applications

every day would cause the activities the building is intended for to grind
to a halt.
It is concluded that non-structural protective measures can
complement other (i.e. structural) measures but are unsuitable as
stand-alone methods to impart collapse resistance to large permanent
civilian structures such as those considered here. The particular role
of fire protection and fire fighting is addressed in Section 6.3.6 below.

6.3.4 Alternative load paths

This approach aims at creating a redundant load transfer system by
providing alternative paths for the forces normally transmitted by failing
load-bearing elements. The focus is on the elements identified as key
elements, that is, those elements whose failure, if no countermeasures
are taken, causes total damage greater than what is acceptable. For
the type of building considered here, it was argued in Section 6.3.2.1
that the whole of the vertical load-bearing members of any one storey
has to be designated as a key element if no agreement upon the
nature and extent of the assumable accidental circumstances or actions
can be reached. It is concluded that, in such a case, no alternative paths
can be made available for the global load transfer.
Again, the assumption ‘failure of an entire storey’, which underlies
the radical conclusion ‘no alternative paths available’, is corroborated
by experience, and is not far-fetched. It could be argued that much
has been learned from the collapse of the WTC towers and that for
the design of future high-rise buildings such an assumption is excessive
owing to improvements in structural design and fire protection. Such
arguments would refer not to the robustness of the structure but to
the probability and the extent of an initial local failure. The lessons
learnt from the events of 11 September 2001 can help to increase the
safety against local failure. It should be appreciated, however, that
these lessons are drawn from circumstances and actions that have
occurred in the past. The future is unknown.
Despite these reservations, assumable accidental circumstances or
actions could possibly be established. As outlined in Section 4.3.2,
such a specification requires a judgement and decision-making process
that reflects public opinion and, in the case of major structures, is
arrived at on a project-by-project basis. Selected structural members
can then be designed for the assumable actions, and other structural
members that are not designed to withstand such actions can be
specified as initially failing elements. The latter specification can also

113

Copyright © ICE Publishing, all rights reserved.

 Progressive collapse of structures

be arrived at directly within the judgement and decision-making

process, as design objective 2 according to Section 4.3.2, but entails
more arbitrariness in the type of structure considered here than in the
case of, for instance, continuous-girder bridges.
When the initial local failure determined in this way does not extend to
all the vertical load-bearing members of one storey, alternative load paths
can, in principle, exist or be made available. This applies to all structural
systems currently in use for high-rise buildings such as frame, shear wall,
core and outrigger, or tubular systems. The redundancy of core and
outrigger systems is evidenced by the pre-collapse response of the
WTC towers to aircraft impact.11,12 Identifying those structural systems
that are particularly suitable for providing alternative paths needs further
analytical study. Requirements for such verification are outlined in
Section 4.3.4. The starting points of the corresponding structural analyses
are the assumable accidental circumstances or cases of initial local failure
that are to be defined as design objectives. Conclusions on the suitability
of particular structural systems will, therefore, also depend on the design
objectives.
As stated above, preventing the initial failure of an entire storey is
likely to require the strengthening of selected structural members. In
the design of the primary load-bearing system of high-rise buildings,
therefore, the alternative-paths method can usually only be applied in
conjunction with the specific-local-resistance method.

6.3.5 Isolation by segmentation

6.3.5.1 Horizontal segmentation
The damage and collapse following an initial local failure can be limited
to an acceptable extent by another measure. Collapsing sections can be
isolated by an appropriate segmentation of the structure. Such an
approach seems most suitable for structures that are aligned along a
horizontal axis, such as continuous girder bridges, where the collapse
progression in the longitudinal direction is arrested at predefined
locations, and isolation is safely and easily achieved by a well thought
out insertion of hinges (see Section 5.3.3).
A similar approach was discussed in Section 6.3.2.3 for the secondary
load transfer system of a building. Instead of hinges, joints within the
floor structure were envisaged as limiting and isolating elements.
Extending this approach to the overall structural system leads to
horizontal segmentation brought about by continuous vertical joints

114

Copyright © ICE Publishing, all rights reserved.

 Applications

that run throughout the structure. In the case of an incipient collapse,

horizontal pushing forces can still be transmitted by the joints. Due to
the small width of the joint gap, however, these do not occur as
impulsive dynamic loading. It is thus improbable that a domino-type
collapse would be induced by such forces. A failure within one segment
therefore remains limited to the same segment, and there would be no
further collapse progression in the horizontal direction — at least not for
buildings with a small height-to-width ratio.
For tall and slender buildings, which are the focus here, this
conclusion might not hold. To keep the extent of the damage reason-
ably limited, the segments should be small and many. A horizontal
segmentation of the overall structure through the insertion of vertical
joints, therefore, is less appropriate here. The segments would become
too slender to be self-supporting. Detailing the joints to allow for the
cross-boundary interaction needed for stability and the transfer of
horizontal and shear forces is difficult. After the failure of individual
segments, these segments would cease to stabilise the remaining
structure. Furthermore, the tolerable accidental actions decrease with
the segment size, so that failure may not be limited to one segment.
If, again, the whole of the vertical load-bearing members of a storey is
designated as a key element, horizontal segmentation would, in the
end, be ineffective.
As indicated before, horizontal segmentation becomes more appro-
priate the smaller the height-to-width ratio of a building. Horizontal
segmentation through expansion joints apparently helped to prevent
a collapse progression in the Pentagon Building on 11 September
2001 (see Section 5.3.4). Instead of by inserting joints, horizontal
segmentation can also be achieved through isolating elements that
are sufficiently strong and capable of resisting the forces induced by
the collapse of an adjacent section — an alternative that likewise was
present and proved beneficial in the Pentagon Building. This opens
up an interesting line of thought that is pursued further in conjunction
with vertical instead of horizontal segmentation in the next section.

6.3.5.2 Vertical segmentation

The pancake-type collapse of the WTC towers progressed in a vertical
direction (see Section 2.2.1). Once set in motion, the collapse could
have been arrested only by horizontal lines of interception. A design
idea based on this observation is to isolate collapsing sections, and so
to limit the extent of damage, by means of vertical segmentation of

115

Copyright © ICE Publishing, all rights reserved.

 Progressive collapse of structures

the overall structure. The applicability of this approach is not limited to

specific structural systems within a segment. The structural characteris-
tics inside the segments are therefore not further considered here, and
attention is focused on the isolating elements that form the segment
borders.
Regarding the mode of effectiveness of the isolating elements, three
alternatives and corresponding design measures were stated in Section
5.3.3. Providing the capability to accommodate large displacements by
eliminating continuity is out of the question here. But strengthening
and the ensuing capability to accommodate large forces also does not
seem to be a viable option. The corresponding vertical resistance
would not only be required locally, that is, within the segment borders,
but also in the remaining structure above and below the failing segment,
which finally means it is required in the entire structure. This resistance
would have to be very high because it must be designed for impact
loading. The resulting dimensions of the vertical load-bearing elements
would further increase mass and impact loading, so that this alternative
seems unpromising.
Thus, only the third alternative for creating isolating elements
remains, which relies on the capability to accommodate large forces
and large displacements at the same time. Such capability brings
about high toughness at the segment borders. It can be achieved by
providing ductility and a large energy dissipation capacity. Although
the resulting forces are large, they are substantially smaller than those
arising in the second alternative discussed above.
The height of one segment is first assumed to be a certain fraction, say
one-tenth, of the building height (Fig. 6.12). Each segment comprises
the entire floor area, and is limited at the top and bottom by horizontal
segment borders that likewise extend over the entire floor area (except
the uppermost segment, which is limited by one border at the bottom
only).
When all the vertical load-bearing elements of one or more con-
tiguous storeys fail, two failure fronts will propagate in the vertical
direction — one downwards and the other upwards — from the location
of the initial failure to the two adjacent segment borders. At the same
time, the upper part of the building (i.e. the part above the failing
segment) moves downwards as a rigid body. The rotational motion
component (i.e. tilting) is presumably small — a preliminary assumption
suggested by the collapses of WTC 1 and 7 that, however, requires
further study (see also Ref. 14). During the fall of the upper part of
the building, the segment border below the failing segment is already

116

Copyright © ICE Publishing, all rights reserved.

 Applications

Segment height

Detail A
(see Fig. 6.13)

Shock
absorbing Segment border
zone

Fig. 6.12 Vertical segmentation of the overall structure

Shock Segment border

absorbing
device

Detail A
Fig. 6.13 Shock-absorbing segment border

117

Copyright © ICE Publishing, all rights reserved.

 Progressive collapse of structures

being loaded by the impacting debris of the failing floors above it. The
main impact occurs when the upper part of the building strikes the
accumulated debris on the lower segment border. Both the upper and
lower parts of the building will remain internally intact until then.
The impact causes large forces that act in addition to the gravitational
forces. The impact forces must be reduced by appropriate measures. For
this purpose, the segment borders incorporate shock-absorbing zones with
large energy dissipation capacity (Figs 6.12 and 6.13). A segment border
consists of two slabs, placed at a storey height or less apart, connecting
structural members, and shock-absorbing devices between the two
slabs. The remaining space between the two slabs of a segment border
can be used like regular floor space or it can house equipment and
appliances or serve as a storage area.
The slabs of a segment border must be stiff and strong enough to
distribute the occurring forces and to withstand them. As the points
of load application are not exactly known but an effective load distribu-
tion must nonetheless be achieved, they are envisaged as massive
reinforced or prestressed concrete slabs of substantial thickness (of
the order of 0.40 m). They are preferably made of high-strength light-
weight concrete to reduce weight and thus the impact forces. Openings
in the slabs are limited to the minimum required for the vertical lines of
traffic and supply.
The two slabs of a segment border are connected by vertical struc-
tural members. These must be designed for the forces from the ordinary
(i.e. non-collapse) loading conditions. A further requirement is that
these members fail under smaller vertical forces than the vertical
load-bearing elements outside the segment borders. It is thus ensured
that at the start of a collapse the parts of the structure outside the failing
segment remain intact, whereas the connecting structural members
inside the adjacent segment borders give way to the impact forces.
When this occurs, the two slabs of each adjacent segment border will
approach each other, that is, there will be a relative vertical displace-
ment between them.
The shock-absorbing devices are being compressed during this
approach movement, in which a large amount of energy must be
dissipated in a calculable way. To minimise the impact forces, the move-
ment should be smooth and take up the entire clear space between the
two slabs. The shock-absorbing devices should meet these require-
ments. They are preferably placed around the columns or close to
walls (Fig. 6.13). They can consist of telescoping large-diameter steel
tube segments provided with a compressible filling (Fig. 6.14). The

118

Copyright © ICE Publishing, all rights reserved.

 Applications

A A

Compressible filling

Section A–A

Fig. 6.14 Shock-absorbing device: telescoping steel tubes with a compressible filling

filling material must allow high volumetric compressive strain along

with large plastic or fracture energy. Candidates for such material are
scrap metal, metal foam, or even a mineral such as porous tuff gravel.
Metal honeycomb structures are suggested in Ref. 117 for similar
purposes.
Numerical studies confirm that shock-absorbing devices with suffi-
cient performance and moderate space requirement can, in principle,
be produced and that the underlying concept, at least with regard to
the required energy dissipation capacity, is applicable to the type of
structure considered here.115
The energy to be dissipated in the shock-absorbing devices of the two
adjacent segment borders corresponds to the released gravitational
energy of the failing segment and the impacting upper part of the
building minus the energy dissipated by the destruction of the failing
segment and the connecting structural members. The energy dissipated
in the failing segment is mainly fracture energy. It can be estimated from
the observed vertical acceleration of building collapses.15 This share of
energy is larger the more the acceleration falls short of the acceleration
of free fall.

119

Copyright © ICE Publishing, all rights reserved.

 Progressive collapse of structures

Some uncertainty in forecasting the failure progression and designing

the segment borders arises from the requirement, mentioned before,
that the connecting structural members are designed to fail under
smaller vertical forces than the vertical load-bearing elements outside
the segment borders. Instead of the failure progression described
above, in which first two failure fronts propagate inside the failing
segment, it is conceivable that the connecting structural members
give way almost immediately because the pressure waves induced by
the initial failure travel much faster than the failure fronts. The
shock-absorbing function of the segment borders would then be acti-
vated earlier than assumed above. Furthermore, the shock-absorbing
function of remote segment borders could be activated simultaneously.
These effects would probably have an attenuating influence on the
failure progression, which can now come to a halt inside the failing
segment before arriving at the segment borders.
In the design concept indicated in Fig. 6.12, each segment consists of
a number of storeys, and is comparatively large. The maximum extent of
damage corresponds to one segment and, therefore, is large as well. If
the maximum extent of damage is to be reduced, the segment size
must be decreased. This consideration leads to a design alternative in
which each segment consists of just one storey. In view of the reduced
drop height, there is less time now for tilting and less uncertainty
concerning the prediction of the failure progression, which is an
additional advantage.
In this alternative, every floor must be designed as a segment border,
that is, as a massive concrete slab. Because the failing segment and the
drop height of the upper part of the building are smaller, the floor slabs
can be of less substantial thickness (say of the order of 0.25 m). On the
other hand, shock-absorbing devices must now be provided in every
storey. Due to the reduced drop height, however, the released gravita-
tional energy is smaller. Therefore, the combined energy dissipation capa-
city requirement of all shock-absorbing devices is approximately the same
as before. This means that the cost of these devices and the ensuing usable
floor space reduction are also about the same. The total mass is likely to be
somewhat larger due to massive concrete slabs now being required for
every floor. In sum, this design alternative provides for a smaller extent
of damage and better predictability of the failure progression at the
same or only slightly increased cost, and thus seems preferable. A similar
concept and supporting numerical studies are presented in Ref. 117.
Partial mid-storey building failures occurred during the 1995 Kobe
and other earthquakes, in which the vertical load-bearing elements

120

Copyright © ICE Publishing, all rights reserved.

 Applications

Fig. 6.15 Failure of an entire storey without collapse progression (Kobe, 1995).
(Source: Eos, 1995, 76, No. 6. Photograph: Paul Somerville, # AGU)

of entire storeys failed without leading to pancake-type collapse

(Fig. 6.15). Such incidents show that it is, in principle, possible to
arrest a collapse in its early stages at horizontal lines of interception.
Finally, it is noted that vertical segmentation may seem unusual, but
it is the only approach that does not impose limits on the tolerable
accidental actions, at least not in the horizontal direction.

6.3.6 Fire protection, fire fighting, and evacuation

Whatever approach is taken in structural design, thought must be given
to non-structural protective measures for fire protection and fire
fighting. In addition, rapid and safe egress from the building must be
assured, as underlined by the WTC tragedy. The measures taken for
these purposes comprise physical provisions and emergency procedures.
Both must be developed further, not only considering effectiveness but
also in terms of robustness. In this context, robustness means that these
provisions and procedures remain operative even in the case of local
damage or partial collapse.
Fire fighting in high-rise buildings usually depends on water that is
pumped up from ground level through standpipes and supplied by
hoses. Even if complemented by secondary supply from water tanks
within the building, the automatic and manual fire-fighting capabilities

121

Copyright © ICE Publishing, all rights reserved.

 Progressive collapse of structures

of such a system might become inoperative through the same event that
ignites the fire. Such conventional fire-fighting systems seem best suited
to the specific-local-resistance method as applied in Section 6.3.2:
standpipes are fed through the massive tube that forms the primary
load transfer system, which means they are relatively well protected
against physical damage; the same holds for the staircases and lifts.
All these elements, however, could still be affected by accidental
circumstances inside the tube that leave the structure intact. To
create redundancy for, and increase the robustness of, fire fighting
and evacuation, the tube should therefore be partitioned by vertical
inner walls. Each compartment created in this way would house
independent standpipes, staircases, and lifts.
Nevertheless, it is desirable that novel fire-fighting systems are
developed that allow for a higher degree of robustness. Instead of
water, other fire-extinguishing agents such as carbon dioxide or foam
can be used. The constituents are relatively light and need little
energy and only light equipment to discharge them. They can be
distributed in a decentralised manner over the entire building, with
the effect of creating autonomous and redundant systems that are
independent of an external energy supply and the normal functioning
of standpipe systems.
Finally, the provisions and procedures for evacuation need to be
improved. As noted above, staircases should be physically protected
and be made redundant. The same applies to lifts, which could be
used not only for fire fighting but also for evacuation, in particular of
disabled occupants (see also the Institution of Structural Engineers
report on safety in tall buildings118). In addition, alternative evacuation
methods should be developed. Helicopter rescue from the roof might
become more of a viable option if special equipment (e.g. multiple-
person rescue baskets) is developed and procedures are improved. For
special structures, such as some towers and special manufacturing
environments, controlled descent devices are permitted by the Life
Safety Code119 to provide escape routes. Occupants could also be
evacuated to neighbouring buildings by zip lines and breeches buoys,
as used for the rescue of sailors from ships.

122

Copyright © ICE Publishing, all rights reserved.

 7
Measures of robustness and
collapse resistance

7.1 General
In Chapters 4 to 6, a pragmatic approach was presented that allows a
reasonable balance of safety and economy to be reached in the design
of structures susceptible to progressive collapse. This approach still
lacks an objective mathematical basis. Its use requires engineering
judgement, imagination, and experience. The further development of
this approach should, where possible, go beyond this subjective basis
towards more objective procedures in order to advance its applicability
and general acceptance.
The terms ‘robustness’ and ‘collapse resistance’ as defined in Section
4.2 offer the possibility for this development. According to these
definitions, robustness means insensitivity of a structure to local failure,
and collapse resistance is insensitivity to accidental circumstances.
The defining terms ‘insensitivity’, ‘local failure’, and ‘accidental
circumstances’ used here are quantifiable by referring to the design
objectives set out in Section 4.3.2. In the following, the formulation
of measures for describing robustness and collapse resistance quantita-
tively is discussed.

7.2 Areas of application

A quantification of these properties based on the definitions and design
criteria mentioned above would have various advantages and possible
applications:
1. Design requirements for robustness or collapse resistance could be
specified numerically in standards and guidelines. This could also
account for the type of structure and its significance and exposure.
2. Measures of robustness or collapse resistance could serve as verifica-
tion tools with the calculated value of a measure being checked

123

Copyright © ICE Publishing, all rights reserved.

 Progressive collapse of structures

against a required value. Project-related design criteria can be

factored into the calculated or required value.
3. It follows that such measures could also be used for optimising the
design concerning the balance between cost on the one hand, and
robustness or collapse resistance, on the other.
4. A measure of robustness could provide guidance on the choice of
design method. For markedly non-robust structures, for instance,
particularly those deemed significant and exposed, methods aimed
at limiting the extent of collapse could be indicated to be preferable
to methods aimed at preventing initial local failure.
5. As outlined in Section 3.2, it might be possible to augment conven-
tional design procedures, which use partial safety factors, with an
additional system partial safety factor on the resistance side of the
design equations to account for the response of the structure to
local failure and thus its robustness. If this additional factor could
be derived easily on the basis of a measure of robustness, then this
would constitute a further step towards a standardised and simplified
treatment of progressive collapse.
6. Finally, a quantification in connection with a classification of
structures opens up the possibility of cataloguing the numerical
values of measures of robustness and collapse resistance for the
various types of structures. The usability of the aforementioned
applications would be improved by such a classification.

7.3 Requirements
The validity and usefulness of measures of robustness and collapse
resistance are linked to certain requirements:120,121
1. Expressiveness: the measures should completely and solely express
robustness or collapse resistance and should not be influenced by
other aspects. The measures should provide clear distinctions
between robust and non-robust or collapse-resistant and non-
collapse-resistant structures.
2. Objectivity: the measures should be independent of user decisions.
Under unchanged conditions, the values of the measures should be
reproducible.
3. Simplicity: the measures should be defined in as simple a manner as
possible. This is in order to gain acceptance with users and in the
interest of objectivity and generality.
4. Calculability: it should be possible to derive the measures from
the properties or behaviour of the structure. All required input

124

Copyright © ICE Publishing, all rights reserved.

 Measures of robustness and collapse resistance

parameters must be quantifiable. A sufficiently accurate numerical

calculation should be possible and not require excessive effort.
5. Generality: the measures should, as far as possible, be applicable to
any kind of structure.

These requirements are partly in conflict with each other, so that it may
not be possible to meet them all at the same level at the same time. For
instance, an evaluation of the approaches proposed so far (see Section
7.4) suggests that a high level of expressiveness can be achieved, but
at the cost of calculability.
Thus, the above requirements may need to be moderated. This seems
justifiable and acceptable, at least for the generality requirement
because different kinds of structures are susceptible to different types
of collapse (see Chapter 2). Hence, different kinds of structures may
be better described by specifically defined measures that reflect different
mechanisms of collapse.

7.4 Present approaches

To date, there is no unified theory of progressive collapse nor is there
agreement on terminology. Accordingly, a large variety of different
approaches for quantifying the relevant characteristics and structural
features have been suggested in the literature. This includes both deter-
ministically defined54,56—58,120,122—125 and probabilistically defined54,121
measures. They are mostly based on assuming damage (an interesting
exception to this is presented in Ref. 126) and on either comparing
the properties of the damaged and undamaged structures or examining
the response of the structure to such initial damage.
When the properties of damaged and undamaged structures are
compared, the comparison is based on the respective load-carrying
capacities,54,58,122—124 stiffnesses,57,120,125 or probabilities of failure.54,121
When the response of the structure to initial local damage is examined,
the resulting extent of collapse or total damage can be quantified, and can
be further related, for instance, to the initial damage and the size of the
structure.120,125 This requires an analytical investigation of collapse
scenarios, which in most cases is complex and costly. The analysis of
the failure progression can be facilitated by topological considerations57
or by energy-based approaches.56,120,125
Furthermore, the measures suggested have different purposes. Verifi-
cation of whether the design objectives have been achieved is not
always sought. A detailed review120,127 shows that the definitions of

125

Copyright © ICE Publishing, all rights reserved.

 Progressive collapse of structures

the measures and analysis approaches suggested so far meet the require-
ments specified in Section 7.3 to varying degrees. No approach stands
out as clearly superior and practical. For further illustration, examples
of three basic approaches are presented below.

7.5 Stiffness-based measures of robustness

A simply defined and easily calculable measure follows from examining
the static stiffness of the structure:
det Kj
Rs ¼ min ð2Þ
j det K0
where Rs is the stiffness-based measure of robustness, Kj is the active
system stiffness matrix of the structure after removing a structural
element or constraint j, and K0 is the active system stiffness matrix of
the intact structure.
Further examination shows that Eq. (2) must still be calibrated, or at
least normalised, to obtain a practical and plausible measure with a
range of values between 0 and 1.120 A value of 1 represents the
maximum possible robustness, while 0 corresponds to a total lack of
robustness. Intermediate values are intended to be quantitative and
expressive measures of robustness.
Verification of this approach was carried out based on ultimate load
analyses of simple frame structures. It turns out that the reduction in
load capacity due to the removal of structural elements does not corre-
late very well with the corresponding values of Rs . So, the expressiveness
of the stiffness-based measure of robustness, at least in its present form,
is not sufficient. It is more a measure of the connectivity of the system
than of the robustness of the structure.120
The advantages of simplicity and ease of calculation, on the other
hand, make it seem worthwhile to follow up on the assessment of robust-
ness through system matrices. In addition to stiffness, other structural
properties, such as strength, load capacity utilisation, ductility, and
mass distribution, can also be described by corresponding matrices
and used in such an assessment. Further refinement is also possible by
taking into account the system topology.57
All these approaches seek to assess in a simplified manner the load
capacity remaining after the removal of structural elements, and thus
the capability of the structure to redistribute the forces carried by
these elements into alternative load paths. Based on the description
of mechanisms and types of collapse presented in Chapter 2, it can

126

Copyright © ICE Publishing, all rights reserved.

 Measures of robustness and collapse resistance

therefore be concluded that these approaches are, in principle, suitable

for structures that are susceptible to zipper-type collapse. They are not
suitable for structures whose collapse behaviour is governed by impact
loading, that is, structures that are susceptible to pancake-type or
domino-type collapse, because such loading is not reflected in structural
properties such as stiffness.

7.6 Damage-based measures of robustness

This is an approach that is based on the quantification of the extent of
collapse or total damage resulting from initial damage. It can easily be
aligned with the definition of robustness suggested in Section 4.2.1.
Consequently, and in contrast to the measure presented in the previous
section, adequate expressiveness can easily be ensured. In doing so, it
makes sense to include the design objectives listed in Section 4.3.2 in
the definition of the measure of robustness. A suitable formulation is
based on the complement of the dimensionless total damage:
p
Rd ¼ 1
ð3Þ
plim
where Rd is the damage-based measure of robustness, p is the maximum
total damage resulting from the assumable initial damage (which, more
precisely, can be any of the assumable cases of initial local failure), and
plim is the acceptable total damage. Note that p and plim refer to damage
occurring additionally to the initial damage. The quantification of
damage required here can be performed by reference to the affected
masses, volumes, or floor areas (in buildings), or even to the resulting
costs.
Since this measure is directly related to design objectives 2 to 4 in
Section 4.3.2, it can be used as a design and verification tool in
achieving these objectives. A value of 1 is obtained when no additional
damage occurs; it thus indicates perfect robustness. For values between
1 and 0, the design objectives are met to a greater or lesser degree.
Negative values are generated when p > plim , and indicate that the
design objectives are not met.
As said before, the measure of robustness defined by Eq. (3) is related
to design objectives, which ensures its usability as a design and verifica-
tion tool. The design objectives, however, must be predetermined in a
decision-making process and cannot be derived from first principles. If
expressiveness and objectivity are targeted not only within the context
of given design criteria but in a more general sense, then the relationship

127

Copyright © ICE Publishing, all rights reserved.

 Progressive collapse of structures

d(i)

A
d
B
d–i

i
i 1

Fig. 7.1 Maximum total damage d(i) as a function of the initial damage i

to design objectives must be abandoned. The question then arises: what

is the extent of the assumable initial damage? Pondering this question
leads to the following integral measure:
ð1
Rd;int ¼ 1
2 ½dðiÞ
i di ð4Þ
0

where Rd;int is the integral damage-based measure of robustness and dðiÞ

is the maximum total damage resulting from and including the initial
damage of extent i. Both dðiÞ and i are dimensionless variables obtained
by dividing the respective reference value (mass, volume, floor area, or
cost) by the corresponding value of the intact structure.
The dimensionless extent of initial damage i lies between 0 and 1.
Because the integration extends over the whole range of values of i,
selecting the assumable initial damage is no longer required. If only
direct structural damage is taken into consideration, and damage to
the surroundings and indirect losses are excluded, then the value of
dðiÞ also lies between 0 and 1.
Figure 7.1 shows three different curves for dðiÞ, each referring to a
different structure of different robustness. Curve A describes a non-
robust structure; even a small amount of initial damage results in
significant total damage. Curve B illustrates a comparatively robust
structure; major additional damage only occurs after significant initial
damage. The maximum total damage dðiÞ as a function of the initial

128

Copyright © ICE Publishing, all rights reserved.

 Measures of robustness and collapse resistance

damage i can only take values in the grey area above the straight line
described by the equation dðiÞ
i ¼ 0. The value of Rd;int therefore
lies between 0 and 1. Similarly to the stiffness-based measure of the
previous section, a value of 1 indicates maximum possible robustness,
and a value of 0 indicates a total lack of robustness.
Now consider curve C. Compared with curve B, it indicates a greater
sensitivity to minor initial damage. Since Rd;int is also a measure of the
convexity of the curve dðiÞ, the measures of robustness resulting from
curves B and C are of the same order of magnitude. Thus, Rd;int does
not adequately reflect the relatively greater importance of the effect
of minor initial damage. This weakness can be remedied by weighting
the initial damage so as to increase the importance of minor initial
damage in conjunction with renormalising the measure to a range of
values between 0 and 1. Modified definitions of damage-based measures
of robustness are presented and compared by Haberland.120
The requirements of expressiveness and simplicity, and in principle
also the requirement of generality, can be met by these types of
approaches. Incidentally, they are not only generally applicable to any
kind of structure but also generally compatible with any kind of
design method. Among the approaches presented here, they are possibly
the only ones that can capture the influence of the design method
‘isolation by segmentation’.
The question of calculability, on the other hand, is more difficult to
answer. Determining the total damage that results from initial damage
requires an examination of the failure progression, which in turn can
involve geometrically and materially nonlinear dynamic analyses in
the time domain (see Section 4.3.4). In particular, the initial damage
must generally be modelled not as a static removal of the affected
structural element or constraint but as a sudden or possibly delayed
event, that is, as a dynamic process. The analysis becomes particularly
difficult when the separation, fall, and impact of structural components
needs to be considered.
Depending on the type of structure and the governing type of
collapse, the analysis required here can become intractable or, for all
practical purposes, too complex. This applies especially to structures
that are susceptible to pancake-type or mixed-type collapse. When it
comes to calculability, and hence applicability in practice, therefore,
generality can also be affected.
Nevertheless, high computational cost is not necessarily a barrier to
the possible applications listed in Section 7.2. In particular, the possi-
bility of a classification of structures as mentioned under point 6 must

129

Copyright © ICE Publishing, all rights reserved.

 Progressive collapse of structures

be emphasised. The identification and cataloguing of numerical values

of measures of robustness would be done on the basis of selected repre-
sentative structures. This work would need to be carried out only once,
so that the anticipated high computational cost is of less importance.

7.7 Energy-based measures of robustness

The above discussion shows that stiffness-based and damage-based
measures of robustness are either easy to calculate or expressive, but
not, from the outset, both. Approaches based on energy considerations
might be capable of meeting both requirements to the same degree. The
following simple approach is based on the comparison of the energy
released during an initial failure and the energy required for failure to
progress:
Er; j
Re ¼ 1
max ð5Þ
j Ef;k
where Re is the energy-based measure of robustness, Er; j is the energy
released during the initial failure of a structural element j and contri-
buting to damaging a subsequently affected element k, and Ef;k is the
energy required for the failure of the subsequently affected element k.
Design objective 2 specified in Section 4.3.2 can be taken into
account when choosing the initially failing structural elements.
Design objectives 3 and 4, that is, the acceptable total damage,
cannot be accounted for in this simplest formulation. Instead, only
the possibility of a failure progression is examined. A value of 1 indicates
perfect robustness. Values between 1 and 0 are acceptable to a greater or
lesser degree. Negative values indicate failure progression. For the sake
of argument, it is assumed that failure progression is equivalent to
complete collapse.
The measure of robustness defined by Eq. (5) seems at first sight to
meet the simplicity requirement and, regarding the possibility of
complete collapse, also the requirement of expressiveness. A difficulty,
however, arises in determining the value of Er; j , which can be over- or
underestimated. The energy released during the initial failure of a struc-
tural element consists of several parts. In structures susceptible to
pancake-type or domino-type collapse, the gravitational potential
energy of separating or overturning elements that is transformed into
kinetic energy makes a major and possibly dominant contribution to
the total released energy. In such cases, a numerical estimation can
be easy and valid. However, in structures susceptible to other types of

130

Copyright © ICE Publishing, all rights reserved.

 Measures of robustness and collapse resistance

collapse, ordinary external forces such as gravity loads contribute

significantly to Er; j . They perform work over the elastic and plastic
displacements occurring during the initial failure. This contribution
can only be determined by a complete structural analysis.
Furthermore, Er; j should include only that part of the energy which
contributes to damaging the subsequently affected element k. This
condition again is most easily met in structures that are susceptible to
pancake-type or domino-type collapse. The separation, overturning,
and impact events that govern such scenarios occur in a discrete and
concentrated manner, so that the released energy is reintroduced into
the structure in a similarly concentrated manner, that is, without
major losses, and is therefore easy to calculate. Overall, it is anticipated
that such structures are the most suitable for the application of energy-
based measures of robustness.
In the simple form of Eq. (5), the measure is truly expressive only for
structures in which the failure of the subsequently affected element k
results in complete collapse. This applies, in particular, to structures
in which the elements j and k are similar, and which are composed
entirely of such elements. In such structures, the number of cases of
initial and subsequent failure to be examined is relatively small,
which enhances calculability. Examples are rows of overhead trans-
mission line towers, where each tower is an element, and high-rise
buildings, where an element could consist of all the vertical load-bearing
members of any one storey.
Equation (5) is, moreover, based on the assumption that the energy of
the event that triggers the initial failure is completely absorbed by that
failure, or at least that the excess energy does not contribute to
damaging the subsequently affected element. Modified definitions of
energy-based measures of robustness and collapse resistance that
attempt to overcome these limitations are presented by Haberland.120
A different energy-based approach in which the work done by external
forces and possible failure sequences are also examined is taken by
Smith.56 This leads to higher computational cost, but could make the
method applicable to structures that are susceptible to zipper-type
collapse.

7.8 Summary and future directions

Damage-based measures of robustness or collapse resistance account for
the actual response of a structure to initial local failure or accidental
circumstances. They can thus easily be aligned with the definitions of

131

Copyright © ICE Publishing, all rights reserved.

 Progressive collapse of structures

robustness or collapse resistance (which might require the choice of

design objectives) and, in principle, also meet the requirement of
generality. For the same reason, however, they are computationally
costly. Measures based on stiffness or other structural properties and
energy-based measures are not, a priori, expressive and general but are
easier to calculate. Improved formulations of such measures could be
used as alternatives to more precise damage-based measures, provided
they can be shown to be expressive.
This verification would have to be done by parametric analyses of
representative structures. Both the damage-based measures and the
improved alternative formulations based on properties or energy
would have to be computed and compared. Alternative formulations
prove useful when they correlate well with damage-based measures in
this comparison. Such analyses would be representative for the classes
of structures being represented by the examined structures. The
structures of each class are susceptible to different types of collapse,
which in turn will best be reflected in different measures. It can
therefore be expected that such improved alternative formulations
would be expressive only for specific classes of structures.

132

Copyright © ICE Publishing, all rights reserved.

 8
Conclusions

8.1 Synopsis
Progressive collapse of structures is a complex and multifaceted problem
for which the existing terminology and procedures are inadequate.
Perhaps the best example for illustrating this is the undifferentiated
classification of quite different collapse scenarios under the one term
‘progressive collapse’, as is customary today. Hence, the various types
and mechanisms of progressive collapse are first of all distinguished
and described in this book, and a terminology is suggested (Chapter 2).
Next, the terms ‘robustness’ and ‘collapse resistance’ are defined,
which brings about a further conceptual refinement. Thus, robustness
is insensitivity to local failure, and collapse resistance is insensitivity
to accidental circumstances. According to this definition, robustness
is a purely structural property, in the sense that the cause and
probability of local failure — and therefore also the nature, extent,
and probability of triggering accidental circumstances — are immaterial.
A robust structure is at the same time collapse resistant, but not vice
versa. A quantitative interpretation of these definitions can be obtained
by relating them to given design objectives.
The need for special considerations for progressive collapse can be
demonstrated by studying current design procedures and referring to
the probabilistic theory of reliability. It becomes clear that non-robust
structures are particularly susceptible and require special attention
(Chapter 3).
However, a purely reliability-based verification of actual structures
where progressive collapse comes into play is almost impossible, at
least for the time being, because of the number and complexity of
influencing factors that arise during a collapse. That this might be
more than a passing problem follows from the fundamental difficulties
in describing and quantifying accidental circumstances or actions and
in reaching an informed consensus on the acceptable probability of a
major disaster. The semi-probabilistic design codes and verification

133

Copyright © ICE Publishing, all rights reserved.

 Progressive collapse of structures

procedures in use today are based on reliability theory, and hence are
also inadequate to ensure collapse resistance. Possible future improve-
ments of procedures that are nonetheless based on reliability theory
are outlined.
On the other hand, both a general concept for handling the questions
raised by the phenomenon of progressive collapse and a procedure for
designing collapse-resistant structures are urgently required. To this
end, a pragmatic approach is suggested in which the design procedures
currently used are complemented by an additional assessment and
measures to ensure the collapse resistance of the structure (Chapters
4 and 5). This approach allows a reasonable balance of safety and
economy to be achieved without requiring excessive analysis effort.
The additional assessment for collapse resistance calls for the consid-
eration of design requirements, design objectives, design methods, and
verification procedures. It is no longer based on reliability theory but on
engineering judgement and on concrete, deterministically defined
design objectives that are to be understood as performance criteria
and must be established in a decision-making process. The most impor-
tant design objectives are the assumable accidental circumstances, the
assumable cases of initial local failure, and the acceptable extent of
collapse.
For ensuring collapse resistance, a number of methods are available,
which are presented and systematically compared. They are either based
on increasing the level of safety against initial local failure or aim at
limiting the extent of collapse following an assumed local failure. The
corresponding structural analyses required for assessment and verifica-
tion are carried out deterministically. Quality requirements for such
analyses are outlined. In general, nonlinear dynamic analysis in the
time domain is required. The feasibility and validity of simplified
analysis and verification procedures are discussed. Prescriptive design
rules, which do not require structural analysis, are also examined, but
it is found that they are not generally to be recommended for major
structures.
Recent European and American standards and drafts indicate that
codification is moving towards such a pragmatic approach already. It
is hoped that future code development leads to a clearer description
of the design requirements, design objectives, design methods, and
verification procedures, as well as to a precise and consistent use of
language.
Even if it is difficult to base the conceptual and practical treatment of
the present problem on reliability theory, it can play an important role in

134

Copyright © ICE Publishing, all rights reserved.

 Conclusions

the further development of the pragmatic approach presented here. A

more in-depth comparison of the design methods discussed and a
specification of design criteria such as the exposure of a structure, the
assumable accidental circumstances, and the applicable safety factors
could thus be made. Nevertheless, these and other design criteria
can, at best, only partly be derived from first principles. Instead, they
need to be established in a judgement and decision-making process
that can be supported by professionals, but in the end must reflect
public opinion. The concept presented here accommodates the need
for public debate by making the choices — for instance concerning the
acceptable extent of collapse — relatively transparent.
If collapse resistance is to be achieved by increasing the level of safety
against initial local failure, the design methods ‘specific local resistance’
and ‘non-structural protective measures’ are available. These methods
do not aim at increasing the robustness of the structure. If collapse
resistance is to be achieved by limiting the extent of collapse following
an assumed local failure, the design methods ‘alternative load paths’ and
‘isolation by segmentation’ can be used. In this way, the structure is
made both robust and collapse resistant.
In the specific-local-resistance method, key elements are designed for
the actions resulting from assumable accidental circumstances. A key
element is basically a structural member whose failure would trigger an
unacceptable extent of collapse. The safety of key elements can also be
increased by non-structural protective measures. The effectiveness of
both methods depends on the completeness and validity of the accidental
circumstances specified as one of the design objectives. Since these events
are unforeseeable or occur with very low probability, completeness and
validity cannot be verified and might not be ensured. Nevertheless,
these methods are of interest for structures that are poorly suited for
other methods. Furthermore, they can be suitable for smaller structures
and for structures with only a few and clearly identifiable key elements.
The alternative-paths method aims at increasing the redundancy of
the structure. It requires an increase in either continuity or structural
resistance, or both. This method, and the prescriptive design rules
based on the same idea, must be applied with prudence. Forces resulting
from alternative paths should be verified down to the foundations. In
the case of prescriptive design rules, they should be determined based
on the overstrength of the structural members and connections that
provide the alternative load transfer.
The segmentation method aims at isolating collapsing sections, and
thus limiting the extent of collapse, by a segmentation of the structure.

135

Copyright © ICE Publishing, all rights reserved.

 Progressive collapse of structures

The isolating elements can work in one of three ways: by high local
resistance, and thus the capability to accommodate large forces; by a
break in continuity, and thus the capability to accommodate large
displacements; or by high ductility and large energy dissipation capacity,
and thus the capability to accommodate large forces and large displace-
ments at the same time. For some structures, segmentation is the more
suitable approach to prevent progressive collapse, a fact that has been
given little attention so far. This may be because the terms ‘continuity’,
‘redundancy’, and ‘robustness’ are intuitively equated, a tacit assump-
tion that is justified, at best, for only particular types of structures.
The suitability of the different design methods depends on the design
objectives and on the type and features of the structure, for instance its
orientation in space. Bridges are mainly aligned along a horizontal axis.
Buildings can be aligned along a vertical or horizontal axis. Structural
components that separate from the structure and fall down could lead
to impact loading on key elements of the remaining structure below.
Such impact events are more of a concern for high-rise buildings than
for bridges or low-rise buildings. The need for continuity to prevent
separation, fall, and impact of components thus varies for different
types of structures, and different design methods will be preferable.
The effectiveness of the various design methods depends on the
accuracy of analysis and the adequacy of the specification of design
objectives, such as assumable accidental circumstances and assumable
initial local failure. This dependency is greatest for the specific-local-
resistance method and for non-structural protective measures (which
are both based on increasing the level of local safety). It is reduced
for the alternative-paths method, and at a minimum for the segmenta-
tion method (which both aim at limiting the extent of collapse). Further
information on the application of design strategies and methods can be
found in Sections 5.3.4 and 5.4 and in Chapter 6.
The discussion in Chapter 6 is directed at specific types of structures
and demonstrates the suitability of the concept presented in Chapters 4
and 5. In applying the concept to bridges, the most important types of
large bridges are examined, and recommendations for their collapse-
resistant design are developed. A different goal is pursued in applying
it to the design of high-rise buildings. The focus here is not so much
on the development of concrete recommendations but, rather, on the
possibility, in principle, of a collapse-resistant design of extremely tall
and slender structures by using the approach set forth in this book.
This approach is essentially based on engineering judgement, imagi-
nation, and experience. The possibility of moving to more objective and

136

Copyright © ICE Publishing, all rights reserved.

 Conclusions

quantifiable procedures is examined in Chapter 7. In particular,

describing robustness and collapse resistance by quantitative measures
seems appropriate and achievable. Simple expressions for stiffness-
based, damage-based, or energy-based measures of robustness are
presented. It turns out that the requirements to be made on such
measures cannot all be met at the same level at the same time. In
particular, measures that are easy to calculate lack expressiveness and
generality, and measures that are expressive and general lack calcul-
ability. Furthermore, it is found that structures that are susceptible to
different types of collapse are best described by differently defined
measures of robustness or collapse resistance.

8.2 Outlook
This book has set out to refine the terminology and to develop a sound
conceptual framework for the treatment of progressive collapse. This
approach has proved applicable and successful, so far. Future develop-
ments could aim at further refinements in terminology and the
identification of additional conceptual relationships. On the other
hand, they should provide insights and methods that are useful in
structural design practice.
Important relationships exist between the concepts discussed here,
and are illustrated in the diagram of Fig. 8.1. They are partly explained
in this book, but deserve further study and clarification. In this system
of mutual relationships, the types of collapse and the respective
propagating actions seem to hold a central position.
In any case, a careful description of types of collapse and propagating
actions leads to insights into collapse-promoting structural features, the
types of structures affected, and possible design methods, as demon-
strated in Section 2.4. As shown there, structural features that have a
bearing on the occurrence and type of progressive collapse are the
strength, ductility, continuity, mode of load transfer (series or parallel),
spatial orientation, and structuredness of a structure. This could be the
starting point for a future classification of structures.
The description of types of collapse also provides hints at a number of
analogies. These could be pursued further to advance the conceptual
framework and design procedures. This includes the analogy between
progressive collapse and instability (Section 2.2.5) and between
zipper-type collapse and fast fracture (Section 2.2.4).
Concerning the needs of design practice, the future development of a
comprehensive and sufficiently differentiated classification of structures

137

Copyright © ICE Publishing, all rights reserved.

 Progressive collapse of structures

Classification
of structures

Collapse-promoting
structural features

Types of collapse
Propagating actions

Design methods Measures of robustness

Verification procedures Measures of collapse resistance

Design objectives

Fig. 8.1 Conceptual relationships in the description and prevention of progressive

collapse

for progressive collapse is particularly important. As far as possible, such

a classification should be linked to recommendations on appropriate
design methods and simplified verification procedures.
Furthermore, objectively quantifiable criteria and procedures for
evaluating and classifying structures would be desirable for design
practice. It became clear in Chapter 7 that the types of collapse play a
leading role also in a quantitative description of the robustness and
collapse resistance of a structure. The further development of measures
of robustness and collapse resistance must be accompanied by
parametric analyses of representative structures. In this, the parameters
to be varied are the structural features that have a bearing on the
occurrence and type of collapse as well as the design objectives.
These analyses would serve to verify the envisaged measures of
robustness and collapse resistance, and to assign applicable definitions
of measures to each class of structure. Finally, they could be the basis
for cataloguing the numerical values of these measures and for cali-
brating them for use as part of a standardised and simplified treatment
of progressive collapse.

138

Copyright © ICE Publishing, all rights reserved.

 References

1. Wearne P. Collapse — When buildings fall down. TV Books, New York,

2000.
2. Scheer J. Versagen von Bauwerken. Ernst & John, Berlin, vol. 1, 2000;
vol. 2, 2001.
3. Wittfoht H. Ursachen für den Teil-Einsturz des ‘Viadotto Cannavino’ bei
Agro di Celico. Beton- und Stahlbetonbau, 1983, 78, No. 2.
4. Lee M. S. Personal communication, 1998.
5. Australian Bureau of Statistics. Feature article — The Tasman Bridge.
Tasmanian year book. ABS, Canberra, 2000. http://www.abs.gov.au/ausstats.
6. Mondorf P. E. Concrete bridges. Taylor & Francis, London, 2006.
7. National Transportation Safety Board. Safety recommendation H-08-1.
NTSB, Washington, DC, 2008. http://www.ntsb.gov/Recs/letters/2008/
H08_1.pdf.
8. Prendergast J. Oklahoma City aftermath. Civil Engineering, 1995, 65,
No. 10, 42—45.
9. Corley W. G., Sozen M. A., Thornton C. H. and Mlakar P. F. The
Oklahoma City bombing: improving building performance through multi-
hazard mitigation. Federal Emergency Management Agency Mitigation
Directorate, Washington, DC, 1996, FEMA Report 277.
10. Corley W. G., Mlakar Sr P. F., Sozen M. A. and Thornton C. H. The
Oklahoma City bombing: summary and recommendations for multihazard
mitigation. Journal of Performance of Constructed Facilities, 1998, 12, No. 3,
100—112.
11. Federal Emergency Management Agency — Federal Insurance and
Mitigation Administration. World Trade Center building performance
study. FEMA, Washington, DC, 2002, FEMA Report 403. Summary in:
Civil Engineering, 2002, 72, No. 5, 36—45.
12. National Institute of Standards and Technology. Federal building and fire
safety investigation of the World Trade Center disaster: final report on the
collapses of the World Trade Center towers. NIST, Gaithersburg, MD,
2005, NIST NCSTAR 1. http://wtc.nist.gov/NCSTAR1.
13. Bažant Z. P. and Zhou Y. Why did the World Trade Center collapse?
Society for Industrial and Applied Mathematics News, 2001, 34, No. 8.

139

Copyright © ICE Publishing, all rights reserved.

 Progressive collapse of structures

14. Bažant Z. P. and Zhou Y. Why did the World Trade Center collapse? —
Simple analysis. Journal of Engineering Mechanics, 2002, 128, No. 1, 2—6.
15. Bažant Z. P. and Verdure M. Mechanics of progressive collapse: learning
from World Trade Center and building demolitions. Journal of Engineering
Mechanics, 2007, 133, No. 3, 308—319.
16. Starossek U. Progressiver Kollaps von Tragwerken. WKP Ingenieurforum.
WKP König und Partner, Hamburg, 2001.
17. National Institute of Standards and Technology. Federal building and fire
safety investigation of the World Trade Center disaster: final report on the
collapse of the World Trade Center building 7. NIST, Gaithersburg, MD,
2008, NIST NCSTAR 1A. http://wtc.nist.gov/NCSTAR1.
18. Breen J. E. Research workshop on progressive collapse of building structures,
University of Texas at Austin, 1975. National Bureau of Standards,
Washington, DC, 1975.
19. Breen J. E. and Siess C. P. Progressive collapse-symposium summary. ACI
Journal, 1979, 76, No. 9, 997—1004.
20. Allen D. E. and Schriever W. R. Progressive collapse, abnormal loads, and
building codes. Structural failures: modes, causes, responsibilities. ASCE
national meeting on structural engineering, Cleveland, Ohio, Apr. 1972.
ASCE Press, Reston, VA, 1973.
21. Amrhein E. Progressive collapse of masonry structures. Proceedings of the
first North American masonry conference. Boulder, CO, 1978.
22. Hawkins N. and Mitchell D. Progressive collapse of flat plate structures.
ACI Journal, Proceedings, 1979, 76, No. 7, 775—808.
23. McConnel R. E. and Kelly S. J. Structural aspects of the progressive
collapse of warehouse racking. The Structural Engineer, 1983, 61A,
No. 11, 343—347.
24. Mitchell D. and Cook W. D. Preventing progressive collapse of slab
structures. Journal of Structural Engineering, 1984, 110, No. 7, 1513—1532.
25. Pekau O. A., Zielinski Z. A. and Lee A. W. K. Progressive collapse in
precast shear walls. Third international conference on tall buildings. Hong
Kong and Guangzhau, 1984.
26. Andrews K. R. F. and England G. L. A computer program for the pro-
gressive collapse analysis of steel frames subject to fire. Proceedings of
the international conference on mechanics of solids and materials engineering.
Singapore, 1995, 1010—1015.
27. Gilmour J. R. and Virdi K. S. Numerical modelling of the progressive
collapse of framed structures as a result of impact or explosion. 2nd
international Ph.D. symposium in civil engineering. Budapest, 1998.
28. Rutner M. P., Astaneh-Asl A. and Son J. Blast resistant performance of
steel and composite bridge piers. IABSE symposium. Budapest, 2006.
29. Yokel F. Y., Wright R. N. and Stone W. C. Progressive collapse: U.S.
Office Building in Moscow. Journal of Performance of Constructed Facilities,
1989, 3, No. 1, 57—75.

140

Copyright © ICE Publishing, all rights reserved.

 References

30. Ghali A. and Tadros G. Bridge progressive collapse vulnerability. Journal of

Structural Engineering, 1997, 123, No. 2, 227—231.
31. Starossek U. Zum progressiven Kollaps mehrfeldriger Brückentragwerke.
Bautechnik, 1997, 74, No. 7, 443—453. http://www.starossek.de.
32. Starossek U. and Sauvageot G. Discussion: Ghali, A. and Tadros, G.
Bridge progressive collapse vulnerability. Journal of Structural Engineering,
1998, 124, No. 12, 1497—1498.
33. Starossek U. Progressive collapse study of a multi-span bridge. Structural
Engineering International, 1999, 9, No. 2, 121—125. http://www.starossek.de.
34. Starossek U. Progressive collapse of multi-span bridges — a case study.
IABSE symposium. Rio de Janeiro, 1999. http://www.starossek.de.
35. Leyendecker E. V. and Ellingwood B. R. Design methods for reducing the risk
of progressive collapse in buildings. Building science series, vol. 98. National
Bureau of Standards, Washington, DC, 1977.
36. Gross J. L. and McGuire W. Progressive collapse resistant design. Journal
of Structural Engineering, 1983, 109, No. 1, 1—15.
37. Kersken-Bradley M. Unempfindliche Tragwerke — Entwurf und Kon-
struktion. Bauingenieur, 1992, 67, No. 1, 1—5.
38. Pötzl M. Robuste Tragwerke — Vorschläge zu Entwurf und Konstruktion.
Bauingenieur, 1996, 71, No. 11, 481—488.
39. Ellingwood B., Leyendecker E. V. and Yao J. T. P. Probability of
failure from abnormal load. Journal of Structural Engineering, 1983, 109,
No. 4, 875—890.
40. Bennett R. M. Formulations for probability of progressive collapse.
Structural Safety, 1988, 5, No. 1, 67—77.
41. Ellingwood B. R. and Dusenberry D. O. Building design for abnormal
loads and progressive collapse. Computer-Aided Civil and Infrastructure
Engineering, 2005, 20, No. 3, 194—205.
42. Ellingwood B. R. Mitigating risk from abnormal loads and progressive
collapse. Journal of Performance of Constructed Facilities, 2006, 20, No. 4,
315—323.
43. Multihazard Mitigation Council of the National Institute of Building
Sciences. Workshop on prevention of progressive collapse, Chicago, 2002.
MMC, Washington, DC, 2003. http://www.nibs.org/MMC/mmcactiv9.html.
44. American Institute of Steel Construction. Proceedings of the steel building
symposium: blast and progressive collapse resistance, New York, 2003.
AISC, Chicago, 2003.
45. Precast/Prestressed Concrete Institute. Blast and progressive collapse
resistance of precast and prestressed concrete structures. PCI workshop.
Chicago, 2004.
46. Joint Committee on Structural Safety. Robustness of structures. Workshop,
BRE, Watford, 2005. JCSS, IABSE, Zurich, 2005. http://www.jcss.ethz.ch/.
47. American Society of Civil Engineers. Journal of Structural Engineering,
2005, 131, No. 8.

141

Copyright © ICE Publishing, all rights reserved.

 Progressive collapse of structures

48. Hayes Jr J. R., Woodson S. C., Pekelnicky R. G., Poland C. D., Corley
W. G. and Sozen M. Can strengthening for earthquake improve blast
and progressive collapse resistance? Journal of Structural Engineering,
2005, 131, No. 8, 1157—1177.
49. American Society of Civil Engineers. Journal of Performance of Constructed
Facilities, 2006, 20, No. 4.
50. Dusenberry D. O. and Hamburger R. O. Practical means for energy-based
analyses of disproportionate collapse potential. Journal of Performance of
Constructed Facilities, 2006, 20, No. 4, 336—348.
51. Kaewkulchai G. and Williamson E. B. Modeling the impact of failed
members for progressive collapse analysis of frame structures. Journal of
Performance of Constructed Facilities, 2006, 20, No. 4, 375—383.
52. Loizeaux M. and Osborn A. E. N. Progressive collapse — an implosion
contractor’s stock in trade. Journal of Performance of Constructed Facilities,
2006, 20, No. 4, 391—402.
53. International Association for Bridge and Structural Engineering.
Structural Engineering International, 2006, 16, No. 2.
54. Maes M. A., Fritzsons K. E. and Glowienka S. Structural robustness in the
light of risk and consequence analysis. Structural Engineering International,
2006, 16, No. 2, 101—107.
55. Starossek U. Progressive collapse of structures: nomenclature and
procedures. Structural Engineering International, 2006, 16, No. 2, 113—
117. http://www.starossek.de.
56. Smith J. W. Structural robustness analysis and the fast fracture analogy.
Structural Engineering International, 2006, 16, No. 2, 118—123.
57. Agarwal J., England J. and Blockley D. Vulnerability analysis of structures.
Structural Engineering International, 2006, 16, No. 2, 124—128.
58. Wisniewski D., Casas J. R. and Ghosn M. Load capacity evaluation of
existing railway bridges based on robustness quantification. Structural
Engineering International, 2006, 16, No. 2, 161—166.
59. Comite Euro-International du Beton. CEB-FIP model code 1990. Bulletin
d’Information No. 203. CEB, Lausanne, 1991.
60. British Standards Institution. BS EN 1990:2002. Eurocode 0: basis of
structural design. BSI, London, 2002.
61. American Society of Civil Engineers. Minimum design loads for buildings and
other structures. ASCE, Reston, VA, 2002, ASCE 7-02.
62. Deutsches Institut für Normung. DIN 1055-100. Einwirkungen auf
Tragwerke — Teil 100: Grundlagen der Tragwerksplanung — Sicherheitskon-
zept und Bemessungsregeln. DIN, Berlin, 2001.
63. American Association of State and Highway Transportation Officials.
Standard specifications for highway bridges. AASHTO, Washington, DC,
2002, 17th edn.
64. Post-Tensioning Institute. Recommendations for stay cable design, testing and
installation. PTI Cable-Stayed Bridge Committee, Phoenix, AZ, 2007, 5th edn.

142

Copyright © ICE Publishing, all rights reserved.

 References

65. International Federation for Structural Concrete. Acceptance of stay cable

systems using prestressing steels. fib, Lausanne, 2005.
66. Moore D. B. The UK and European regulations for accidental actions.
Workshop paper, see Ref. 43, Appendix D, 2003.
67. Office of the Deputy Prime Minister. The Building Regulations 2000 —
structure: approved document A (2004 edition). ODPM, London, 2006.
68. British Standards Institution. BS 5950-1:2000. Structural use of steelwork in
building — part 1: code of practice for design — rolled and welded sections. BSI,
London, 2001.
69. British Standards Institution. BS 8110-1:1997. Structural use of concrete.
Code of practice for design and construction. BSI, London, 1997.
70. British Standards Institution. BS EN 1991-1-7:2006. Eurocode 1: actions on
structures — part 1-7: general actions — accidental actions. BSI, London, 2006.
71. Department for Communities and Local Government. Guide to the use of
EN 1991-1-7 — Accidental actions. DCLG, London, 2006.
72. Dusenberry D. O. and Juneja G. Review of existing guidelines and provisions
related to progressive collapse. Workshop paper, see Ref. 43, Appendix D,
2003.
73. National Research Council Canada. National building code of Canada
1995. NRC, Ottawa, 1996.
74. American Concrete Institute Committee 318. Building code requirements
for structural concrete (ACI 318-02) and commentary (ACI 318R-02).
ACI, Farmington Hills, MI, 2002.
75. US General Services Administration. Progressive collapse analysis and design
guidelines for new federal office buildings and major modernization projects.
GSA, Washington, DC, 2003.
76. US Department of Defense. United facilities criteria — design of buildings to
resist progressive collapse. DoD, Washington, DC, 2005, UFC 4-023-03.
77. Crowder B. and Marchand K. Progressive collapse: U.S. perspective and
existing guidelines and standards. SEI progressive collapse meeting. Imperial
College, London, 2007.
78. National Institute of Standards and Technology. Best practices for reducing
the potential for progressive collapse in buildings. NIST, Gaithersburg, MD,
2007, NISTIR 7396.
79. International Code Council. International building code. ICC, Washington,
DC, 2009.
80. International Code Council. International fire code. ICC, Washington, DC,
2009.
81. Steering Group of the SEI Progressive Collapse Standards and Guidance
Committee. A PreStandard Prospectus: robustness and collapse resistance
for buildings. SEI, Reston, VA, 2006. http://www.seinstitute.org/files/pdf/
progressivecollapse06.pdf.
82. Marchand K., Smilowitz R., Waggoner M. and Crowder B. SEI Progressive
Collapse Standards and Guidance Committee activity update. Structure

143

Copyright © ICE Publishing, all rights reserved.

 Progressive collapse of structures

Magazine, Dec. 2006, 71. http://www.structuremag.org/archives/2006/

Dec-2006/N-SEI-Dec06p70-71.pdf.
83. Starossek U. Typology of progressive collapse. Engineering Structures, 2007,
29, No. 9, 2302—2307. http://www.starossek.de.
84. Cherepanov G. P. Mechanics of the WTC collapse. International Journal of
Fracture, 2006, 141, Nos 1—2, 287—289.
85. Cherepanov G. P. and Esparragoza I. E. Progressive collapse of towers: the
resistance effect. International Journal of Fracture, 2007, 143, No. 2, 203—206.
86. Starossek U. Brückendynamik — Winderregte Schwingungen von Seilbrücken
[Bridge dynamics — wind-induced vibrations of cable-supported bridges].
Friedr. Vieweg & John, Wiesbaden, 1992. http://www.starossek.de.
87. Strom R. W. and Ebeling R. M. Methods used in tieback wall design and
construction to prevent local anchor failure, progressive anchorage failure,
and ground mass stability failure. Final report. US Army Corps of Engineers,
Washington, DC, 2002, ERDC/ITL TR-02-11.
88. UK Standing Committee on Structural Safety. Confidential reporting on
structural safety (CROSS). SCOSS Newsletter, Apr. 2008, No. 10. http://
www.scoss.org.uk.
89. Wood A. B., Fair R. I. and Liptrot F. J. Transmission-line design: the
ultimate load concept applied to the Java 500 kV transmission lines.
Power Engineering Journal, Mar. 1988, 63—70.
90. Song H. W. and Tassoulas J. L. Finite element analysis of propagating
buckles. International Journal for Numerical Methods in Engineering, 1993,
36, 3529—3552.
91. Boehm B. W., Brown J. R., Kaspar H., Lipow M., Macleod G. J. and
Merritt M. J. Characteristics of software quality. North-Holland, New
York, 1978.
92. Starossek U. Disproportionate collapse: a pragmatic approach. Proceedings
of ICE: Structures and Buildings, 2007, 160, No. 6.
93. Alexander S. Application of standard events in design against dispropor-
tionate collapse. IABSE symposium structures and extreme events. Lisbon,
2005.
94. van Breugel K. Storage system criteria for hazardous products. Structural
Engineering International, 1997, 7, No. 1, 53—55.
95. Starossek U. Progressiver Kollaps von Bauwerken. Beton- und Stahlbeton-
bau, 2005, 100, No. 4, 305—317. http://www.starossek.de.
96. Starossek U. Progressive collapse of bridges — aspects of analysis and
design. Invited lecture. Proceedings of the international symposium on sea-
crossing long-span bridges, Mokpo, 2006. http://www.starossek.de.
97. American Society of Civil Engineers. Structural design for physical security —
state of the practice. ASCE Task Committee, Reston, VA, 1999.
98. Bundesamt für Strassen. Richtlinie Anprall von Strassenfahrzeugen auf
Bauwerksteile von Kunstbauten, Ergänzungen zur Norm SIA 261 Einwirkun-
gen auf Tragwerke. ASTRA, Bern, 2005.

144

Copyright © ICE Publishing, all rights reserved.

 References

99. Starossek U. and Wolff M. Design of collapse-resistant structures.

Robustness of structures. Workshop, BRE, Watford, 2005. Joint Committee
on Structural Safety, IABSE, Zurich, 2005. http://www.jcss.ethz.ch.
100. Osteraas J. D. Murrah Building bombing revisited: a qualitative assess-
ment of blast damage and collapse patterns. Journal of Performance of
Constructed Facilities, 2006, 20, No. 4, 330—335.
101. Mlakar P. F., Dusenberry D. O., Harris J. R., Haynes G., Phan L. T. and
Sozen M. A. The Pentagon Building performance report. American Society
of Civil Engineers, Reston, VA, 2003.
102. Starossek U. Interaktion und progressiver Kollaps. Interaktionsprobleme
im Ingenieurbau. 9. Dresdner Baustatik-Seminar. Technische Universität
Dresden, Dresden, 2005. http://www.starossek.de.
103. Astaneh-Asl A. Progressive collapse prevention in new and existing
buildings. Proceedings of the 9th Arab structural engineering conference.
Abu Dhabi, 2003.
104. Corley G. Applicability of seismic design in mitigating progressive
collapse. Workshop paper, see Ref. 43, Appendix D, 2003.
105. Blue Ribbon Panel on Bridge and Tunnel Security. Recommendations for
bridge and tunnel security. American Association of State Highway and
Transportation Officials, Transportation Security Task Force, Washing-
ton, DC, 2003.
106. Rutner M. Der Widerstand von Stahl- und Verbundbrückenpfeilern
gegen hochdynamische Beanspruchung bei Explosion. Stahlbau, 2006,
75, No. 12, 976—982.
107. Son J., Astaneh-Asl A. and Rutner M. P. Blast performance of long
span cable-supported bridge decks. IABSE symposium. Budapest, 2006.
108. Wolff M. and Starossek U. Robustness assessment of a cable-stayed bridge.
Proceedings of the 4th international conference on bridge maintenance, safety,
and management (IABMAS’08). Seoul, 2008. http://www.starossek.de.
109. Wolff M. and Starossek U. Cable loss and progressive collapse in cable-
stayed bridges. Bridge Structures, 2009, 5, No. 1, 17—28.
110. Wolff M. Progressiver Kollaps und Schrägseilbrücken. Structural Analysis
and Steel Structures Institute, Hamburg University of Technology,
Hamburg, doctoral dissertation (in progress).
111. Park Y. S., Starossek U., Koh H. M., Choo J. F., Kim H. K. and Lee S. W.
Effect of cable loss in cable stayed bridges — focus on dynamic
amplification. IABSE symposium. Weimar, 2007.
112. O’Donovan J., Wilson K. and Dempsey T. The design and construction
of Taney Bridge, Dundrum. Tom McCormack Memorial Lecture. The
Institution of Structural Engineers, Republic of Ireland Branch,
Dublin, 2003.
113. Giuliani L. and Bontempi F. Dynamic analysis for structural robustness
evaluation. Proceedings of the 4th international conference on bridge mainte-
nance, safety, and management (IABMAS’08). Seoul, 2008.

145

Copyright © ICE Publishing, all rights reserved.

 Progressive collapse of structures

114. Starossek U. Avoiding disproportionate collapse of tall buildings. Structural

Engineering International, 2008, 18, No. 3, 238—246. http://www.staros-
sek.de.
115. Robohm M. Entwurfskonzepte für kollapsresistente Hochhäuser. Structural
Analysis and Steel Structures Institute, Hamburg University of Tech-
nology, Hamburg, diploma thesis, 2006.
116. Starr C. M. and Krauthammer T. Cladding-structure interaction under
impact loads. Journal of Structural Engineering, 2005, 131, No. 8, 1178—
1185.
117. Zhou Q. and Yu T. X. Use of high-efficiency energy absorbing device
to arrest progressive collapse of tall building. Journal of Engineering
Mechanics, 2004, 130, No. 10, 1177—1187.
118. The Institution of Structural Engineers. Safety in tall buildings and other
buildings with large occupancy. Report. IStructE, London, 2002.
119. National Fire Protection Association. Life safety code. NFPA, Quincy,
MA, 2006, NFPA 101.
120. Haberland M. Progressiver Kollaps und Robustheit. Structural Analysis
and Steel Structures Institute, Hamburg University of Technology,
Hamburg, diploma thesis, 2007. http://www.sh.tu-harburg.de.
121. Lind N. C. A measure of vulnerability and damage tolerance. Reliability
Engineering and System Safety, 1995, 48, 1—6.
122. Giuliani L. Beurteilung des Alfred P. Murrah Federal Building bezüglich der
Anfälligkeit für progressiven Kollaps. Structural Analysis and Steel
Structures Institute, Hamburg University of Technology, Hamburg,
diploma thesis, 2005.
123. Giuliani L. and Wolff M. Strategie per il conseguimento della robustezza
strutturale: connessione e compartimentazione. CRASC’06 — Convegno
Nazionale Crolli e Affidabilità delle Strutture Civili. Università degli Studi
di Messina, Messina, 2006.
124. Harte R., Krätzig W. B. and Petryna Y. S. Robustheit von Tragwerken —
ein vergessenes Entwurfsziel? Bautechnik, 2007, 84, No. 4, 225—234.
125. Starossek U. and Haberland M. Approaches to measures of structural
robustness. Proceedings of the 4th international conference on bridge
maintenance, safety, and management (IABMAS’08). Seoul, 2008. http://
www.starossek.de.
126. Beer M. and Liebscher M. Designing robust structures — a nonlinear
simulation based approach. Computers & Structures, 2008, 86, No. 10,
1102—1122.
127. Starossek U. and Haberland M. Measures of structural robustness —
requirements & applications. Proceedings of the ASCE SEI 2008 structures
congress — crossing borders. Vancouver, 2008. http://www.starossek.de.

146

Copyright © ICE Publishing, all rights reserved.

 Index

Page numbers in italics denote figures.

Acceptable extent of collapse 43, 54, 64, 111 zipper-type collapse, 15

acceptable total damage annular cross-sections, 103—104, 104
design objectives, 39, 43 applied design methods, 86—124
high-rise buildings, 100—101 Approved Document A 9, 46, 48
robustness measures, 127 arch bridges, 99
see also total damage ASCE (American Society of Civil
accidental circumstances, 1 Engineers), 8, 10, 42
assumable, 43—44, 101, 113—114 ‘assumable’, definition, 43
cable-stayed bridges, 89, 90 Assumable initial local failure, 42, 54, 64,
definition, 39 111
design actions, 45 assume local failure strategy, 47, 54, 58—83
design procedures, 33—34, 36 bridge design, 95—96
Eurocode EN 1991-1-7, 45 design objectives, 43—44, 46
exposure to, 41—42 high-rise buildings, 100—101, 106—107,
identified actions, 47 106, 111, 113—114
insensitivity to see collapse resistance
prevent local failure, 56—57 bending failure, cast-in-place joints, 65
see also triggering events bending moments, cable-stayed bridges, 92,
Alfred P. Murrah Federal Building, 2, 4 93, 94
collapse typology, 23 Best Practices for Reducing the Potential for
design methods, 53—54, 53—55, 59—62, Progressive Collapse in Buildings, 9
59—61, 72—73, 73 bracing, 21
alternative load paths, 4, 47, 59—62, 80, break-away hinges, 66, 97—98
82—83, 135 see also hinge insertion
bridge design, 87, 96, 98 bridges
high-rise buildings, 106—108, 110, cast-in-place joints, 65
113—114 collapse-resistant design, 86—99
interaction, 79 Confederation Bridge, 44—45, 62—67,
prescriptive design rules, 84—85 69—71, 77—79, 88
segmentation and, 71, 75—76 corrosion protection, 97
American Society of Civil Engineers damping effect, 93
(ASCE), 8, 10, 42 design methods, 62—66, 64—67, 69—71,
Analogies 19, 23 74, 77—79
analysis-based verification, 50—51 design objectives, 46
future directions, 132 failure incidents, 3—4
interaction, 78 Haeng-Ju Grand Bridge, 3, 23—24, 66,
present approaches, 125—126 78, 88
research publications, 6—7 instability-type collapse, 21
segmentation, 65—66, 69 Mississippi River Bridge, 4
anchored elements mixed-type collapse, 23—24
domino-type collapse, 18 section-type collapse, 19—20

147

Copyright © ICE Publishing, all rights reserved.

 Progressive collapse of structures

bridges (continued ) requirements of measures, 124—125

standards/guidelines, 8 verification, 44, 48—51, 59
Tacoma Narrows Bridge, 14, 14, 97 column failure, 53, 54, 59, 60—61
Taney Bridge, 95—96 combination of actions, 43, 46, 95
Tasman Bridge, 3—4, 66, 71, 74, 88 compactness, 29, 102
verification procedures, 51 compartmentalisation, 70
zipper-type collapse, 14 see also segmentation
see also individual named bridges compressible filling, shock-absorbers, 119,
British Standards, 8, 46, 48 119
brittle fractures, 18—20, 25 compression loading, 22
brittle material behaviour, 26 computational cost, robustness measures,
BS 5950, 8, 48 129—131
BS 8110, 8, 48 computational tractability, 35
buckling, 21—22 conceptual relationships, 137, 138
building design standards, 8—10 Confederation Bridge, 63
see also design . . . ; high-rise buildings collapse-resistant design, 88
building envelope, 109—112 design methods, 62—66, 64—67, 69—71,
77—79
cable net, 16 design objectives, 44—45
cable-stayed bridges consequences classes, 42
accidental circumstances, 89, 90 construction-stage design methods, 57—58
collapse-resistant design, 89—97 continuity, 27—28
design objectives, 46 failure incidents and, 4
mixed-type collapse, 24 redundancy vs, 62
section-type collapse, 19—20 segmentation and, 66—67, 70—76, 88
standards/guidelines, 8 continuous girder bridges, 3, 21, 88—89
verification procedures, 51 corrosion protection, bridges, 97
zipper-type collapse, 14 costs
calculability requirement, 124—125, 129 computational, 129—131
Canadian code, 9, 70 safety-cost ratio, 81, 85
cantilever floors, 106—109 cross-section collapse, 18—20
cantilever tip, drop-in girders, 68, 69 cutting action, 78
capacity design approach, 106—107, 109,
111 DAF (dynamic amplification factor),
cast-in-place joints, bridges, 65 90—95, 92
catenary action, 62, 84 damage see total damage
CEB-FIP Model Code 1990, 7 damage-based robustness measures, 125,
Charles de Gaulle Airport, 73—74 127—130, 131—132
Chaumont Viaduct, 88 damping effect, bridge design, 93
classes of collapse, 11, 25 decoupling mechanisms, 110
classes of structures, 137—138 deflection of floors, 106—107, 109
design procedures, 35 design criteria, 40—51, 41
design requirements, 42 see also design methods; design
robustness measures, 124, 129—132 objectives; design requirements;
verification procedures, 50 verification
clients, design objectives, 43—44 design methods, 47—48, 52—85
code-based design see design procedures applications, 86—122, 123—124
collapse boundaries, 64 collapse-promoting features, 27—28
collapse-promoting features, 11, 25—30 local failure, 2
collapse resistance, 2 research publications, 5—7
bridge design, 86—99 standards/guidelines, 8—10
continuity and, 4 synopsis, 134—136
definition, 38, 39—40, 123 see also design criteria; design procedures
design criteria, 40, 41 design objectives, 42—46
high-rise buildings, 86—87, 99—122, 136 Alfred P. Murrah Building, 53—54
interaction, 79 direct design methods, 52
measures of, 123—132, 137—138 key elements, 52—53

148

Copyright © ICE Publishing, all rights reserved.

 Index

robustness measures, 127—128, 130 event control, 55

see also design methods exemption process, GSA guidelines, 42
design procedures explosion-triggered collapse, 6, 102—103
accidental circumstances, 33—34, 36 see also detonation
current, 32—40 exposure to accidental circumstances,
improvement potential, 34—37 41—42
inadequacy of, 32—34 expressiveness requirement, 124, 126—127,
see also design methods 129
design requirements, 40—42, 123
destabilisation see instability-type collapse façade structure, high-rise buildings,
deterministic analysis, 35, 37, 125 109—112
detonation effects, 53, 54 failure incidents, 2—5
see also explosion-triggered collapse see also local failure
DIN 1055-100 standard, 8 failure probabilities see probabilistic theory
direct design methods, 47, 52 fast fracture, 18—20, 25
alternative load paths, 85 Federal Emergency Management Agency
applications, 86—122 (FEMA) report, 4, 99—100, 104
local failure, 80, 83 fib guidelines, 90
see also assume local failure strategy; fire fighting/protection, 121—122
prevent local failure strategy floor design, high-rise buildings, 105—111,
discontinuity, 27—28, 73—74 111, 120
see also continuity force concentration, 25—26
discrete conditions, 44—45 fracture energy, high-rise collapse,
disproportionate collapse 119—120
definition, 1, 39 fracture mechanics, 19—20
instability-type, 21, 23
domino-type collapse, 16—18, 19 General Services Administration (GSA),
classification, 25 42, 46—47, 50
energy-based robustness, 130—131 generality requirement, 125, 129, 132
mixed-type, 23—24 girders
promoting features, 26—30 Alfred P. Murrah Building, 53, 54,
drop-in girders, 66, 67—68, 69, 79 60—62
ductility strengthening, 58, 61
high-rise building design, 107, 116 global failure 1, 21
material behaviour, 26—27 arch bridges, 99
prescriptive design rules, 84 local failure vs, 32
dynamic action, 25—26 gravitational energy, 13, 22, 107
instability-type collapse, 22 Gross, J. L., 5—6
segmentation method, 68, 69 GSA (General Services Administration),
verification procedures, 49—51 42, 46—47, 50
dynamic amplification factor (DAF), guidelines, 7—10
90—95, 92 bridge design, 87, 90
design methods, 47
earth-anchored suspension bridges, 97—98 design objectives, 46
earthquake effects, 2, 12, 121 design requirements, 42
elastic energy, 13, 22, 107 prescriptive design rules, 83—84
electric power networks, 27 see also standards
Ellingwood, B. R., 5
energy-based robustness measures, 125, Haeng-Ju Grand Bridge, 3, 23—24, 66, 78,
130—131 88
energy dissipation, high-rise design, hangers, suspension bridges, 97
116—120 hierarchical structures, 112
Eurocodes high-rise buildings
EN 1990, 8 acceptable total damage, 100—101
EN 1991-1-7, 9, 39, 42, 44—48 collapse-promoting features, 28—29
evacuation procedures, 121—122 collapse-resistant design, 86—87,
events see triggering events 99—122, 136

149

Copyright © ICE Publishing, all rights reserved.

 Progressive collapse of structures

high-rise buildings (continued ) large bridges see bridges

load transfer systems, 82—83, 101—112, lateral action, local failure, 99
103—106, 108, 110, 111 Leyendecker, E. V., 5
see also individual named buildings Life Safety Code, 122
hinge insertion, 58 lifts, 122
interaction, 78—79 limiting local failure, 47—48
segmentation method, 66, 67, 70 load cases see local failure
suspension bridges, 97—98 load combination, 32, 43, 46, 95
horizontal load transfer, high-rise buildings, load transfer
102, 103, 110 arch bridges, 99
horizontal segmentation, 108—110, 108, collapse-promoting features, 28
114—115 design methods, 81—83
human error, 41, 57 high-rise buildings, 101—112, 103—106,
108, 111
identified accidental actions, 47 pancake-type collapse, 12—13
impact class collapses, 25 suspension bridges, 98
see also domino-type collapse; pancake- zipper-type collapse, 16
type collapse see also alternative load paths
impact forces local failure, 1—2
domino-type collapse, 16 assumable initial, 42, 54, 64, 111
pancake-type collapse, 13 column failure, 59
vertical segmentation, 116—118 definition, 38—39
importance factor, 42 design procedures, 32—36
see also significance of structure direct design methods, 52
impulsive loading, 68, 69 high-rise buildings, 100—101, 106—107,
indirect design methods, 47—48, 52, 83 106, 111, 113—114
see also prescriptive design rules insensitivity to see robustness
initial failure see local failure lateral action, 99
insensitivity, 123 limiting, 47—48
CEB-FIP Model Code, 7 multiple cables, 95—96
definition, 38—39 preventing, 47, 55—58, 80—83, 88—89,
see also collapse resistance; robustness 96—97
instability-type collapse, 20—23 segmentation method, 76
cable-stayed bridges, 94, 95 verification procedures, 49
classification, 25 see also assume local failure strategy
mixed-type, 24 loss-of-cable events, 89—90
promoting features, 26, 29 modelling, 91
suspension bridges, 97 numerical studies, 91—95
interaction, 77—80
International Federation for Structural McGuire, W., 5—6
Concrete (fib), 90 main column design, 53, 54, 55, 59—61
isolating elements, 64, 70 malicious action, 41, 57
bridge design, 64, 66, 70 cable-stayed bridges, 89, 90
high-rise buildings, 108—109 exposure to, 41—42
local failure, 80—81 high-rise buildings, 101
see also segmentation prevent local failure, 57
suspension bridges, 98
joints material behaviour
bridge design, 65 brittle, 26
high-rise buildings, 108, 114—115 ductile, 26—27
journal articles, 6—7 mediating elements, 17, 26—27
membrane structure, 16
Kersken-Bradley, M., 6 Mississippi River Bridge, 4
key elements, 52—53, 80, 101 mixed-type collapse, 23—24
Kobe earthquake, 121 bridges, 94, 97
classification, 25
Lap splice, 20 promoting features, 27

150

Copyright © ICE Publishing, all rights reserved.

 Index

modelling prevent local failure strategy, 47, 55—58,

local failure, 49 80—83, 88—89, 96—97
loss-of-cable events, 91 primary load transfer system
modification of structural system, 58, 60, arch bridges, 99
61, 67 design methods, 82—83
moment-resisting frames, 62, 70, 110 high-rise buildings, 101—104, 103—104,
Multihazard Mitigation Council report, 6 111—112
multiple-cable failure, 95—96 pancake-type collapse, 13
suspension bridges, 98
National Institute of Standards and zipper-type collapse, 16
Technology (NIST), 9 probabilistic theory, 32—37, 81—82, 125,
nonlinear analysis, 69, 91, 95 133—134
non-structural protective measures, 47, progressive collapse
55—56, 56, 80, 135 classes, 11, 25
bridge design, 96—97 conceptual relationships, 137, 138
high-rise buildings, 112—113, 121—122 definition, 1
numerical studies, 91—95 designing against, 37—51
see also quantification present approaches, 125—126
types, 11—31
objectivity requirement, 124, 127 propagating action, 11
orientation (spatial), 28—29 conceptual relationships, 137, 138
overhead transmission towers, 17—18, 19 domino-type collapse, 17—18
overstrength, 26—27 instability-type collapse, 22—23
overturning susceptibility, 17—18 mixed-type collapse, 24
see also domino-type collapse pancake-type collapse, 13
zipper-type collapse, 15
pancake-type collapse, 12—14, 12 protective non-structural measures, 55—56,
classification, 25 56, 80, 135
design applications, 101, 105, 115—116, bridge design, 96—97
121 high-rise buildings, 112—113, 121—122
energy-based robustness measures, PTI see Post-Tensioning Institute
130—131 public access, high-rise buildings, 102,
mixed-type, 23—24 112—113
promoting features, 25, 28—29 pylons, cable-stayed bridges, 93, 94
parallel load transfer, 16, 28, 34, 81
partial safety factors quantification
assume local failure, 54 design objectives, 43
design objectives, 45 robustness definition, 38—39, 123—124,
design procedures, 35 137
prevent local failure, 57 see also numerical studies
robustness measures, 35, 124 quasi-static analysis, 50—51, 90—95
Pentagon Building, 73—74, 74, 115
performance-based approaches see design redistribution class collapses, 25
methods; design objectives see also section-type collapse; zipper-type
plastic energy, 107 collapse
Post-Tensioning Institute (PTI), 46, 51, redundancy, 60—62, 70—77, 122
87, 89—90, 93 relevant elements, 64, 80—81
Pötzl, M., 6 reliability theory, 32—33, 35—37, 45,
power networks, 27 133—134
pragmatic approach, 38, 123, 134—135 research publications, 5—7
Prendergast, J., 4 resistance see specific local resistance
prescriptive design rules, 47—48, 52, 83—85 risk assessments, 44
continuity, 72 robustness, 2
direct design vs, 86—87, 100 acceptable total damage, 127
synopsis, 134—135 continuity and, 4
pressure waves, 120 definition, 38—40, 123—124, 137
PreStandard Prospectus, 10, 38 design objectives, 127—128, 130

151

Copyright © ICE Publishing, all rights reserved.

 Progressive collapse of structures

robustness (continued ) high-rise buildings, 100—112, 122

fire-fighting systems, 121—122 research publications, 5
measures of, 123—132, 137—138 stability failure, 22
requirements of measures, 124—127, see also instability-type collapse
129—130, 132 staircases, redundancy, 122
research publications, 6—7 standards, 7—10, 134
standards/guidelines, 8 bridge design, 87
Tasman Bridge, 3 design methods, 47
verification, 48—49, 59 design objectives, 46
Ronan Point incident, 2, 8, 108 prescriptive design rules, 83—84
collapse typology, 14 see also guidelines
design methods, 71—72, 72 static analysis, 50—51, 90—95
stiffness-based robustness measures,
safety 125—127
construction stages, 58 strain energy
cost:safety ratio, 81, 85 instability-type collapse, 22
index, 45 pancake-type collapse, 13
partial safety factors, 35, 45, 54, 57, strengthening
124 transfer girders, 58, 61
structural safety, 32—33 vertical segmentation method, 116
Sampoong Superstore, 3 see also overstrength
Scheer, J., 3 Structural Engineering Institute
secondary column design, 53, 54, 60—61 PreStandard Prospectus, 10, 38
secondary load transfer system, 83, structural fuses, 27, 66, 110
104—110, 105—106, 108, 111 structural integrity standards, 8
secondary structure requirements, 42 see also robustness
section-type collapse, 18—20, 25 structural safety, 32—33
segment borders, 27, 64, 70, 117—120, structure classes, 35, 42, 50, 124, 129—132,
118—119 137—138
segmentation, 47—48, 62—71, 80—83, structuredness, 29—30, 104
135—136 suspension bridges, 97—98
bridge design, 88, 96, 97—98 see also bridges
high-rise buildings, 107—110, 108, Swiss guidelines, 87
114—121
interaction, 79 Tacoma Narrows Bridge, 14, 14, 97
prescriptive design rules, 85 Taney Bridge, 95—96
redundancy vs, 71—77 Tasman Bridge, 3—4, 66, 71, 74, 88
self-anchored suspension bridges, 97 telescoping segments, 118—119, 119
self-supporting façades, 110 tension ties, 83—84
series load transfer, 13, 28, 33, 81 titling of structures, 116—117
ship impact, bridges, 89 total damage
shock-absorbing devices, 117—120, definition, 39, 43
118—119 high-rise building design, 100—101
significance of structure, 40—42 robustness measures, 127—129, 128
see also importance factor tractability, computational, 35, 66, 77—79
simplicity requirement, 124, 129—130 transfer girders
simplified verification procedures, 50 Alfred P. Murrah Building, 53, 54,
size of structure, 28—29 60—62
slender structures, 28—29 strengthening, 58, 61
see also high-rise buildings transmission towers, 17—18, 19
societal consensus, 34 triggering events, 1
soil investigation, 57 see also accidental circumstances
spatial orientation features, 28—29 tubular structures, 102—104, 112
specific local resistance, 45, 47, 55, 80—83, typology of collapses, 11—31
135
Alfred P. Murrah Building, 54, 55 United Facilities Criteria, 46—47, 50
bridge design, 96, 98 US standards/guidelines, 8—10

152

Copyright © ICE Publishing, all rights reserved.

 Index

verification World Trade Center (WTC), 2, 86

collapse resistance, 44, 48—51, 59 alternative load paths, 113—114
interaction, 78 collapse typology, 12—13
procedures for, 48—51 failure incident, 4—5
research publications, 5—7 FEMA report, 99—100
robustness measures, 123—126, 132 segmentation potential, 114—117, 121
segmentation method, 65—66, 69
standards/guidelines, 8—9 Yokel, F. Y., 5
vertical load transfer
high-rise buildings, 101—102, 103, zipper-type collapse, 14—16, 14, 19—21
110—112 anchored elements, 15
pancake-type collapse, 12—13 cable-stayed bridges, 89, 94
vertical segmentation, 109, 115—121, 117 classification, 25
Viadotto Cannavino, 3, 88 energy-based robustness, 131
vulnerability research, 7 mixed-type, 23—24
promoting features, 26—27, 28
Wearne, P., 2—3 stiffness-based robustness, 127
Weld, 20 suspension bridges, 97

153

Copyright © ICE Publishing, all rights reserved.