164

ARTICLE
Three-dimensional numerical analysis for rock slope stability
using shear strength reduction method
Jiayi Shen and Murat Karakus
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by University of North Dakota on 12/20/14

Abstract: Existing numerical modeling of three-dimensional (3D) slopes is performed mainly by using the shear strength
reduction (SSR) technique based on the linear Mohr–Coulomb (MC) criterion, whereas the nonlinear failure criterion for rock
slope stability is seldom used in slope modeling. However, it is known that rock mass strength is a nonlinear stress function and
that, therefore, the linear MC criterion does not agree with the rock mass failure envelope very well. In this research, a nonlinear
SSR technique is proposed that can use the Hoek–Brown (HB) criterion to represent the nonlinear behavior of a rock mass in the
FLAC3D program to analyze 3D slope stability. Extensive case studies are carried out to investigate the influence of the conver-
gence criterion and boundary conditions on the 3D slope modeling. Results show that the convergence criterion used in the 3D
model plays an important role, not only in terms of calculation of the factor of safety (FOS), but also in terms of the shape of the
failure surface. The case studies also demonstrate that the value of the FOS for a given slope will be significantly influenced by
the boundary condition when the slope angle is less than 50°.

Key words: three-dimensional (3D) numerical model, Hoek–Brown criterion, shear strength reduction method, slope stability,
boundary condition.

Résumé : La modélisation numérique de pentes en trois dimensions (3D) est généralement réalisée avec la technique de
réduction de la résistance au cisaillement (RRC) basée sur le critère Mohr–Coulomb (MC) linéaire, tandis que le critère de
For personal use only.

rupture non linéaire est rarement utilisé dans les modélisations de pentes. Cependant, il est reconnu que la résistance d'une
masse rocheuse est une fonction de contrainte non linéaire, ainsi, le critère MC linéaire ne correspond pas très bien avec
l'enveloppe de rupture de la masse rocheuse. Dans la présente étude, une technique RRC non linéaire est proposée, qui
permet d'utiliser le critère d'Hoek–Brown (HB) pour représenter le comportement non linéaire d'une masse rocheuse dans
le programme FLAC3D afin d'analyser la stabilité d'une pente en 3D. Des études de cas détaillées sont réalisées pour évaluer
l'influence du critère de convergence et des conditions frontières sur la modélisation de la pente en 3D. Les résultats
démontrent que le critère de convergence utilisé dans le modèle 3D joue un rôle important, non seulement dans le calcul
du facteur de sécurité (FS) mais aussi en termes de la forme de la surface de rupture. Les études de cas démontrent aussi que
la valeur du FS pour une pente donnée sera influencée significativement par la condition frontière lorsque l'angle de la
pente est inférieur à 50°. [Traduit par la Rédaction]

Mots-clés : modèle numérique en trois dimensions (3D), critère d'Hoek–Brown, méthode de réduction de la résistance au cisail-
lement, stabilité de pente, condition frontier.

Introduction However, it is known that 3D analysis provides a more realistic

model because it can take into account the appropriate geometry
Rock slope stability is one of the major challenges of rock engi-
and boundary conditions. Therefore, the development of 3D slope
neering projects, such as open-pit mining. Rock slope failure can
analysis has become a popular research topic in geotechnical en-
affect mining operations and result in costly losses in terms of
gineering in recent years. A list of 3D slope stability papers pub-
time and productivity. Therefore, the evaluation of the stability of
lished in the last 7 years is shown in Table 1.
rock slopes is a critical component of open-pit design and opera- Commonly used approaches for 3D slope stability analysis in-
tion (Naghadehi et al. 2013). clude the limit equilibrium method (LEM), limit analysis method
In most geotechnical applications, two-dimensional (2D) plane (LAM), and numerical modeling, such as the finite element
strain analyses are commonly used to simulate stability of earth method (FEM) and discrete element method (DEM). FEM is derived
structures (Basarir et al. 2005; Karakus 2007; Karakus et al. 2007; from the continuum theory and has the merit of simulating con-
Eid 2010; Tutluoglu et al. 2011). The majority of rock slope analyses tinuous bodies. DEM describes particulate materials and is partic-
in practical projects are still performed using 2D limit equilib- ularly well suited to the analysis of jointed media and multi-body
rium or plane strain analysis because 2D analysis is relatively interaction (Shen and Abbas 2013). The 3D LEM model involves
simple and yields a conservative factor of safety (FOS) compared various assumptions about internal force distribution, and it is
with three-dimensional (3D) analysis (Griffiths and Marquez 2007). difficult to locate the critical failure surface, as is well docu-
For comparative studies between 2D and 3D slope analysis, refer mented in the literature (Griffiths and Marquez 2007; Wei et al.
to Li et al. (2009), Michalowski (2010), and Zhang et al. (2013). 2009; Zhang et al. 2013). The 3D LAM model has been used for

Received 22 May 2013. Accepted 7 November 2013.

J. Shen. School of Civil, Environmental and Mining Engineering, The University of Adelaide, Adelaide, South Australia 5005, Australia; Institute of Port,
Coastal and Offshore Engineering, Zhejiang University, Hangzhou, Zhejiang, 310058, P.R. China.
M. Karakus. School of Civil, Environmental and Mining Engineering, The University of Adelaide, Adelaide, South Australia 5005, Australia.
Corresponding author: Murat Karakus (e-mail: [email protected]).

Can. Geotech. J. 51: 164–172 (2014) dx.doi.org/10.1139/cgj-2013-0191 Published at www.nrcresearchpress.com/cgj on 12 November 2013.
 Shen and Karakus 165

Table 1. 3D slope stability analyses using different Fig. 1. Instantaneous MC envelope of the HB criterion in the
methods. normal and shear stress plane.
Authors Methods
Cheng and Yip (2007) LEM
Griffiths and Marquez (2007) SSR
Farzaneh et al. (2008) LAM
Li et al. (2009) LAM
Michalowski and Drescher (2009) LAM
Wei et al. 2009 SSR/LEM
Li et al. (2010) LAM
Michalowski (2010) LAM
Detournay et al. (2011) SSR
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by University of North Dakota on 12/20/14

Stianson et al. (2011) SSR

Gharti et al. (2012) SSR
Zheng (2012) LEM
Nian et al. (2012) SSR
Michalowski and Nadukuru (2013) LAM
Nadukuru and Michalowski (2013) LAM
Zhang et al. (2013) SSR
Fig. 2. Correlations between MC parameters and ␴3.
Note: LAM, limit analysis method; LEM, limit equili-
brium method; SSR, shear strength reduction.

slopes with simple geometries. However, the construction of the

3D failure mechanism for LAM is not straightforward for compli-
cated slope models, which leads to this method being seldom used
for complex conditions (Wei et al. 2009).
Currently, 3D numerical modeling performed by the shear
strength reduction (SSR) technique is a very attractive and com-
For personal use only.

monly accepted approach among geotechnical researchers and

engineers because not only can it automatically locate the critical
failure surface, but it can also simulate the stress–strain behavior
and give the progressive shear failure of the slope in complex
geometries and loading conditions.
Although the SSR technique has the merits noted above, users still
must take into account the following limitations when using it for
the analysis of 3D isotropic rock slopes: (i) the existing 3D SSR tech-
nique is mainly based on the linear Mohr–Coulomb (MC) criterion. It
is known that rock strength is nonlinear, and many researchers
(Priest 2005; Jimenez et al. 2008; Li et al. 2008; Shen et al. 2012a; Tao
(1) ␴1 ⫽ ␴3 ⫹ ␴ci mb 冉 ␴3
␴ci
⫹s 冊
a

et al. 2013) showed that the MC criterion generally cannot represent

rock mass behavior very well, especially for slope stability problems
where ␴1 and ␴3 are the maximum and minimum principal
where the rock mass is in a state of low confining stresses that makes
stresses, respectively, and ␴ci is the uniaxial compressive strength
the nonlinearity more obvious; (ii) the selection of an appropriate
(UCS) of the intact rock. mb, s, and a are the Hoek–Brown input
convergence criterion is not easy for a 3D SSR model because the
parameters, which can be estimated from the Geological Strength
value of the FOS for a given slope can be considerably influenced by
Index (GSI), disturbance factor D, and intact rock constant mi.
the convergence criterion; (iii) boundary conditions play an impor-
tant role in the distribution of internal stresses in the slope model
and can affect the simulation results. (2) mb ⫽ mie[(GSI⫺100)/(28⫺14D)]
With the aim of achieving a better understanding of the funda-
mental rock slope failure mechanisms and improving the accu- (3) s ⫽ e[(GSI⫺100)/(9⫺3D)]
racy of the rock slope stability results of 3D numerical models, in
this research the authors propose a simple nonlinear SSR tech- e(⫺GSI/15) ⫺ e(⫺20/3)
(4) a ⫽ 0.5 ⫹
nique. The proposed method uses the Hoek–Brown (HB) criterion 6
(as described in the next section) to represent the nonlinear be-
havior of a rock mass in FLAC3D program (Itasca Consulting Group To use the HB criterion in conjunction with SSR methods for
2009) to analyze 3D slope stability. calculating the FOS of rock slopes, methods are required to deter-
Extensive case studies are carried out to investigate the influ- mine the instantaneous MC shear strength parameters of cohe-
ence of the convergence criterion and boundary conditions on the sion, c, and angle of friction, ␾, from the HB criterion (Fu and Liao
numerical results, which include rock mass shear strength, shape 2010). The HB criterion (see eq. (1)) is expressed by the relationship
of the failure surface, as well as FOS values. between maximum and minimum principal stresses. However,
it can also be expressed in terms of normal stress, ␴n, and shear
Instantaneous shear strength of the HB criterion stress, ␶, on the failure plane as shown in Fig. 1. The instantaneous
The nonlinear HB criterion, initially proposed by Hoek and cohesion, c, and angle of friction, ␾, can be calculated by locating
Brown (1980), has been widely used for predicting intact rock and the tangent of the HB envelope under a given value of normal
rock mass strength in rock engineering for several decades. The stress, ␴n, as illustrated in Fig. 1. The intercept with the ␶ axis gives
latest version of the HB criterion presented by Hoek et al. (2002) is the c value, and the slope of the tangent to the HB failure envelope
expressed as yields the ␾ value.

Published by NRC Research Press

 166 Can. Geotech. J. Vol. 51, 2014

Fig. 3. Flow chart of the HB criterion in FLAC3D using the nonlinear SSR technique. The values of RF can be adjusted using the bracketing
approach proposed by Dawson et al. (1999). Flow and Fup, lower and upper bracket values of FOS, respectively; FDiff, difference between upper
and lower FOS values.
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by University of North Dakota on 12/20/14
For personal use only.

Figure 1 also illustrates the stress state of an element where the

strength can be defined by the MC criterion. If the stress state
(␴1, ␴3) of an element is known, the corresponding instantaneous
c and ␾ values can be calculated using eqs. (5) to (8) proposed by
(6)
冋
␾ ⫽ arcsin 1 ⫺
2 ⫹ amb mb 冉
2
␴3
␴ci
⫹s 冊
a⫺1
册
Shen et al. (2012b).

冉
mb␴3
冊
(7) ␶⫽
␴ci cos␾
冉 mb
␴n
⫹s 冊 a

共 兲
a
sin␾ a ␴ci
⫹s 2 1⫹
␴n ␴3 ␴ci a
⫽ ⫹
冉 冊
(5)
␴ci ␴ci mb␴3 a⫺1
2 ⫹ amb ⫹s
␴ci (8) c ⫽ ␶ ⫺ ␴n tan␾

Published by NRC Research Press

 Shen and Karakus 167

The numerical slope model can be divided into elements using Fig. 4. Boundary conditions for a slope model. u, x-direction
mesh techniques. When the slope is modeled under the loading displacement; v, y-direction displacement; w, z-direction
condition, the stress states of the elements in the model will vary, displacement.
which leads to the elements having different values of c and ␾.
An example can be used to show the relationship between in-
stantaneous c, ␾, and minimum principal stress, ␴3, as shown in
Fig. 2. The following parameters were used for the calculation:
␴ci = 25 MPa, GSI = 80, mi = 15, and D = 0.5; the values of ␴3 range
from 0 to 25 MPa. Figure 2 illustrates that the values of instanta-
neous c increase and ␾ decrease with the increase of ␴3 values,
which reflects the nonlinear behavior of the HB criterion.
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by University of North Dakota on 12/20/14

Nonlinear SSR method for the HB criterion

The calculation of the FOS using the SSR technique is based on
reducing the MC shear strength parameters c and ␾ until the slope
fails, and then the value of the FOS can be defined as the ratio of
the actual shear strength to the minimum shear strength of the
rock or soil materials that is required to prevent failure (Duncan
1996). The reduced shear strength parameters cf and ␾f are given by

c
(9) cf ⫽
RF

tan␾
(10) ␾f ⫽ arctan
RF

where RF is a reduction factor, and the value of RF is equal to the

FOS when slope failure occurs.
For personal use only.

One of the most promising ways to use the HB criterion in

conjunction with SSR techniques is to estimate the instantaneous
MC shear strength parameters c and ␾ for elements in the slope.
The details of the application of this nonlinear SSR technique for The convergence criterion in FLAC3D is the nodal unbalanced
2D FEM slope analysis can be found in Fu and Liao (2010). Kumar’s force ratio, R, and the user must specify a number of calculation
(1998) solution was used by Fu and Liao (2010) to calculate the steps, N, to bring the model to a state of equilibrium. An example
instantaneous ␾ values, which requires Newton’s iteration for- of data from the paper by Hammah et al. (2005) can be used to
mula to calculate the ␾ values. It should be noted that eqs. (5) to (8) check the influence of the convergence criterion on the 3D slope
are an alternative form of Kumar’s (1998) solution. However, the model. The example has the following slope geometry and rock
equations offer the benefit of being able to calculate the instanta- mass properites: slope height, H = 10 m; slope angle, ␤ = 45°; ␴ci =
neous c and ␾ of an element from its stress state (␴1, ␴3) without 30 MPa; mi = 2; GSI = 5; D = 0; unit weight, ␥ = 25 kN/m3; deforma-
the need for iteration analysis. tion modulus, Em = 5000 MPa; and Poisson’s ratio, ␯ = 0.3.
In this research, the nonlinear SSR strategy, together with The model has 475 elements and the analyses were carried out
eqs. (5) to (8), was used to implement the HB criterion in FLAC3D using a 1 m unit width. The boundary conditions for the slope are:
for 3D rock slope stability analysis. Figure 3 presents a flow chart fixed x-direction displacement at the front and back faces of the
showing the steps for implementing the HB criterion in the slope model; fixed x, y and z direction displacement at the base
FLAC3Dslope model, as follows: face of the slope model; fixed y-direction displacement of the end
faces of the slope model (see Fig. 4).
• Step 1: Build the slope model according to slope geometry, rock Figure 5 compares the failure surfaces corresponding to the FOS
mass properties, loading, and boundary conditions. Mesh tech- values using various convergence criteria in the slope model. N is
niques are used to generate the grid elements for a slope. the iteration number and R is the convergence parameter.
• Step 2: Carry out the elastic stress analysis to determine the The results show that when the convergence criteria change,
stress state of each element in the slope model. the values of the FOS vary from 1.01 to 1.90. Calculation step N has
• Step 3: Use eqs. (5) to (8) to calculate the cohesion, c, as well as the a slight effect on the FOS when N is more than 1000. For example,
angle of friction, ␾, for each element. when R = 1E–3, the value of the FOS is equal to 1.82 for N = 1000 and
• Step 4: Reduce the c and ␾ values of all elements by a reduction FOS increases to 1.90 for N = 2000. The results also indicate a clear
factor, RF. trend in the increase of the FOS with the increase of R. For exam-
• Step 5: Use the reduced cf and ␾f values for the elastoplastic ple, when R values are increased from 1E–5 to 1E–4, N = 1000, the
analysis using the MC constitutive model. values of the FOS increase from 1.04 to 1.29, where the difference
• Step 6: Repeat steps 2 to 5 when a new reduction factor RF is is 19.4%. In this case, FOS = 1.29 is more realistic than FOS = 1.04 as
generated until slope failure. Finally, the value of the FOS of a the model cannot form the failure surface when the value of R is
given slope is equal to the reduction factor, RF. equal to 1E–5. We analyzed eight slope cases, which have the same
parameters as the slope from Hammah et al. (2005), with various
Convergence criterion in the 3D models slope geometries (␤ = 60° and 70°, H = 5 and 20 m) and rock
Research by Wei et al. (2009) demonstrated that the value of the properties (␴ci = 20 and 50 MPa, D = 0.3 and 0.5). It was found that
factor of safety of the 3D SSR model can be significantly influ- the eight slope models cannot produce a failure surface when R =
enced by the selection of the convergence criterion. Therefore, it 1E–5. However, the model will produce the failure surface, and the
is necessary to carry out some trial and error analysis to select an FOS value tends to stabilize when the convergence criterion is R =
appropriate convergence criterion for a slope model. 1E–4 and the value of N is more than 1000.

Published by NRC Research Press

 168 Can. Geotech. J. Vol. 51, 2014

Fig. 5. Comparison of failure surfaces corresponding to FOS values using different convergence criteria. N, iteration number; R, convergence
parameter).
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by University of North Dakota on 12/20/14
For personal use only.

Fig. 6. FOS values corresponding to computation time versus mesh Table 2. Input parameters of a
elements (computer used: Intel Core Duo, 2.53 GHz). slope case.
Input parameters Values
H (m) 20
␤ (°) 60
␥ (kN/m3) 27
␴ci (MPa) 5
GSI 40
mi 12
D 0.7

Boundary conditions in 3D models

The choice of appropriate boundary conditions is important for
3D slope stability analysis as boundary conditions play an impor-
tant role in the development of internal stresses in a slope, which
will change the shape of the failure surface corresponding to the
value of the FOS.
The commonly used boundary conditions for a 3D slope model
The choice of mesh techniques can also influence the FOS re- (see Fig. 4) are: fixing the x-direction displacement (u = 0) at the
sults in the SSR analysis. A comprehensive study of the influence front and back faces of slope model; fixing the (x, y, z) direction
of different mesh techniques and mesh elements on the calcula- displacement (v = u = w = 0) at the base of slope model. For the end
tion of the FOS using FLAC3D models has been conducted by faces, there are three types of boundary conditions as suggested
Zhang et al. (2013). Figure 6 compares the different mesh sizes for by Chugh (2003):
the calculation of the FOS corresponding to the computation time.
The results show that when the mesh size is more than 1400, the • Condition 1: fixing the y-direction displacement (v = 0), which
value of the FOS tends to equal 1.16, which is close to Hammah represents contact with a rigid, smooth abutment that can pro-
et al.’s (2005) result of an FOS of 1.15. vide a reacting thrust, but no in-plane shear restraint.

Published by NRC Research Press

 Shen and Karakus 169

Fig. 7. Comparison of failure surfaces: contours of c and ␾ and FOS of a slope model under various boundary conditions.
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by University of North Dakota on 12/20/14
For personal use only.

Table 3. Results of FOS and fB for the slope with different slope angles. Fig. 8. Correlations between fB and ␤ under different boundary
conditions for a slope case.
Boundary condition used Boundary
at end faces of weighting
slope model factor, fB
Slope angle, ␤ (°) FOSy FOSxy FOSxyz fB,xy fB,xyz
30 2.11 3.75 3.42 1.78 1.62
45 2.06 2.97 2.52 1.44 1.22
60 1.88 2.50 2.06 1.33 1.09
75 1.57 2.11 1.73 1.34 1.10
90 1.29 1.69 1.43 1.31 1.11

• Condition 2: fixing the (x, y) direction displacement (u = v = 0),

which represents a contact that provides side shear resistance.
• Condition 3: fixing the (x, y, z) direction displacement (u = v = w = 0),
which is used to characterize contacts with no movement.
A small-scale slope (Table 2) with modest rock mass damage
under good blasting (D = 0.7) was used to analyze the influence of The contours of the instantaneous c and ␾ shown in Fig. 7 are
boundary conditions on slope stability. Figure 7 compares the calculated using eqs. (5) to (8) together with the final stress states
slope failure surfaces corresponding to the FOS values, as well as of each element when the slope fails, which can be used to illus-
the contours of instantaneous cohesion and angle of friction in a trate the failure mechanics performed by the nonlinear SSR tech-
given slope under different boundary conditions. nique. For boundary conditions 1 and 2, where the z-direction

Published by NRC Research Press

 170 Can. Geotech. J. Vol. 51, 2014

Table 4. Results of FOS and fB for real slope cases.

Cases H (m) ␤ (°) ␥ (kN/m3) ␴ci (MPa) GSI mia Db FOSy FOSxy FOSxyz fB,xy fB,xyz
1 184 55 27 153 47 9 0.9 2.67 3.67 3.04 1.38 1.14
2 140 34 26 50 28 8 0.7 1.52 2.52 2.27 1.66 1.49
3 220 45 27 65 44 17 0.8 2.48 3.54 2.97 1.43 1.20
4 135 65 27 172 58 9 0.9 4.16 5.70 4.92 1.37 1.18
5 70 50 27 29 41 7 0.8 1.83 2.59 2.19 1.42 1.20
6 110 45 26.5 50 25 10 0.7 1.60 2.28 1.96 1.43 1.23
7 270 45 27 109 39 18 0.9 2.21 3.18 2.68 1.44 1.21
8 170 55 30 104 48 7 0.7 2.63 3.63 3.01 1.38 1.15
9 60 60 27 65 44 13 1 2.53 3.44 2.80 1.36 1.11
10 35 67 27 109 28 12 1 1.93 2.61 2.14 1.35 1.11
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by University of North Dakota on 12/20/14

11 63 35 27 109 28 12 1 2.04 3.15 2.78 1.55 1.37

12 70 49 27 3 49 24 1 1.20 1.72 1.46 1.44 1.22
13 58 50 27 5 55 22 1 1.80 2.53 2.12 1.41 1.18
14 60 48 27 5 54 22 1 1.75 2.48 2.11 1.42 1.20
15 60 52 27 5 56 22 1 1.83 2.50 2.09 1.37 1.14
16 40 71 27 50 33 14 1 1.70 2.29 1.87 1.35 1.10
17 110 50 27 50 25 14 1 1.05 1.48 1.26 1.40 1.20
18 41 50 27 3 46 24 1 1.34 1.90 1.62 1.42 1.21
19 41 55 27 3 49 24 1 1.42 1.96 1.64 1.38 1.16
20 46 55 27 3 50 24 1 1.40 1.93 1.62 1.38 1.16
21 57 49 27 3 48 24 1 1.26 1.81 1.55 1.44 1.23
22c 57 37 27 3 48 24 1 1.36 2.13 1.88 1.56 1.38
23c 57 40 27 3 48 24 1 1.33 2.03 1.79 1.52 1.34
24c 57 42 27 3 48 24 1 1.32 1.97 1.72 1.49 1.30
aValues of mi are estimated based on the information of uniaxial compressive strength for the general rock type (Shen 2013).
bExcavation methods for cases 9 and 10 were considered to be poor and cases 11–21 were obtained from open pit mines (the excavation method was assumed to be
production blasting), therefore, according to the guidelines by Hoek et al. (2002), D was assumed to be 1 for all cases.
cCases 22–24 are additional cases that have the same rock mass properties as case 21 except for slope angle.
For personal use only.

displacement of the end faces is not fixed, the slope surfaces have Fig. 9. Correlations between fB and ␤ under different boundary
relatively higher ␾ values and lower c values compared with the conditions for open-pit cases.
values at the bottom of the slope. This disparity is a result of
the fact that the stress state of the elements at the bottom of the
slope is greater than the stress state of the elements near the slope
surface; and the values of instantaneous c increase and ␾ decrease
with the increase of ␴3 values as shown in Fig. 2.
For boundary condition 3, where (x, y, z) direction displacement
of the end faces is fully fixed, the contours of c and ␾ are obviously
different from those of conditions 1 and 2. This can be explained
by the fact that the stress state in the slope under boundary con-
dition 3 is different from the stress state in the slope under bound-
ary conditions 1 and 2. Therefore, the values of instantaneous
c and ␾ will change, which leads to the change of the shape of the
failure surface as well as the FOS values.
The value of the FOS for boundary condition 1 is equal to 1.883,
which is lower than the FOS for boundary conditions 2 and 3
where the FOS is equal to 2.502 and 2.057, respectively. To inves-
tigate the possible correlation of the FOS under different bound-
ary conditions, the authors proposed a boundary weighting
factor, fB, as shown in Fig. 7, which represents the ratio of the FOS
from boundary conditions 2 and 3 to boundary condition 1. In this Our analysis of the database showed that there is a strong cor-
case, fB,xy = 2.502/1.883 = 1.329 and fB,xyz = 2.057/1.883 = 1.092. relation between fB and ␤ as shown in Fig. 9. Most of the data lie
Table 3 compares the FOS values under different boundary con- along the lines that have a trend of decreasing fB with the increase
ditions for the slope (see Table 2), with the slope angle, ␤, varying of ␤ when ␤ < 50°. When ␤ > 50°, fB tends to attain constant values.
from 30° to 90°. The correlations between fB and ␤ in Table 3 are The results that were presented demonstrate that the effects of
plotted in Fig. 8. The figure demonstrates that the boundary boundary conditions on the values of the FOS are more obvious
weighting factor fB decreases as the slope angle increases when for a slope with a low angle than a steep slope. The values of fB,xy
␤ < 50°. However, fB tends to reach stable values (fB,xy = 1.4 and and fB,xyz will go up to 1.7 and 1.5, respectively, when the slope
fB,xyz = 1.1) when ␤ > 50°. angle is less than 35°. On the other hand, when the slope angle is
Results presented in Fig. 8 and Table 3, however, provide only more than 50°, the values of fB,xy and fB,xyz tend to equal 1.4 and 1.1,
one example for specific rock properties and slope height. To respectively.
further check the correlation between fB and ␤, the authors con- The possible connections between fB and other parameters
ducted a comprehensive study using 21 real cases collected from (H, ␴ci, GSI, and mi) were also investigated as shown in Fig. 10.
Douglas (2002) and Taheri and Tani (2010), with various slope No strong relationship was observed between fB and these pa-
geometries and rock mass properties as indicated in Table 4. rameters.

Published by NRC Research Press

 Shen and Karakus 171

Fig. 10. Correlations between fB,xy and H, ␴ci, GSI, mi.

Can. Geotech. J. Downloaded from www.nrcresearchpress.com by University of North Dakota on 12/20/14
For personal use only.

Conclusions For example, a natural slope often has curvature, and concave
A simple nonlinear SSR method has been proposed to analyze surfaces appear in open-pit design. Therefore, further research is
the stability of 3D rock slopes satisfying the HB failure criterion. required to consider the effect of complex geometries on 3D nu-
This method is based on estimating the instantaneous MC shear merical analysis.
strength parameter c and ␾ values from the HB criterion for ele-
ments in a FLAC3D model.
Acknowledgements
The reliability of the proposed 3D slope model has been tested The Ph.D. scholarship provided by China Scholarship Council
using an example from Hammah et al. (2005). The value of the FOS (CSC) to the first author is gratefully acknowledged. The authors
calculated by the proposed slope model (fine mesh) is equal to 1.16, would like to express their gratitude to anonymous reviewers for
which is close to Hammah et al.’s results with FOS = 1.15. However, their constructive comments on the manuscript. The authors are
it was found that the convergence criterion used in the model also grateful to Barbara Brougham for reviewing the manuscript.
plays an important role not only in the calculation of the FOS, but
also in locating the failure surface. References
Basarir, H., Ozsan, A., and Karakus, M. 2005. Analysis of support requirements
Then, the proposed 3D slope model has been used to analyze for a shallow diversion tunnel at Guledar dam site, Turkey. Engineering
the influence of the boundary condition on the calculation of the Geology, 81(2): 131–145. doi:10.1016/j.enggeo.2005.07.010.
FOS using 21 real open-pit cases with various slope geometries and Cheng, Y.M., and Yip, C.J. 2007. Three-dimensional asymmetrical slope stability
rock mass properties. The authors have proposed a boundary analysis extension of Bishop’s, Janbu’s, and Morgenstern-Price’s techniques.
weighting factor, fB, to investigate the possible correlation of the Journal of Geotechnical and Geoenvironmental Engineering, 133(12): 1544–
1555. doi:10.1061/(ASCE)1090-0241(2007)133:12(1544).
FOS under different boundary conditions. Chugh, A.K. 2003. On the boundary conditions in slope stability analysis. Inter-
Our analysis demonstrates that there is a strong correlation national Journal for Numerical and Analytical Methods in Geomechanics, 27:
between fB and slope angle ␤. The value of fB will decrease with the 905–926. doi:10.1002/nag.305.
increase of slope angle when ␤ < 50°. However, fB reaches stable Dawson, E.M., Roth, W.H., and Drescher, A. 1999. Slope stability analysis by
strength reduction. Géotechnique, 49(6): 835–840. doi:10.1680/geot.1999.49.
values (fB,xy = 1.4 and fB,xyz = 1.1) when ␤ > 50°.
6.835.
It should be noted that, to obtain accurate results, the boundary Detournay, C., Hart, R., and Varona, P. 2011. Factor of safety measure for
condition of a slope model should be determined by considering Hoek–Brown material. In Proceedings of Continuum and Distinct Element
the real conditions. The results of this study demonstrate that the Numerical Modeling in Geomechanics - 2011. Edited by Sainsbury, Hart,
effects of boundary conditions on the values of FOS are more Detournay, and Nelson. Paper No. 13-07, pp. 765–772.
Douglas, K.J. 2002. The shear strength of rock masses. Ph.D. thesis, University of
prominent for a slope with a low angle than a steep slope. There- New South Wales, Australia.
fore, great care should be taken when selecting appropriate Duncan, J. 1996. State of the Art: Limit equilibrium and finite-element analysis of
boundary conditions during 3D numerical slope stability analysis slopes. Journal of Geotechnical and Geoenvironmental Engineering, 122(7):
with low slope angles (␤ < 50°) in the case of unavailability of real 577–596. doi:10.1061/(ASCE)0733-9410(1996)122:7(577).
Eid, H.T. 2010. Two- and three-dimensional analyses of translational slides in
boundary conditions at the early stage of a project. soils with nonlinear failure envelopes. Canadian Geotechnical Journal, 47(4):
The current 3D slope stability study is based on simple slope 388–399. doi:10.1139/T09-110.
geometry. However, the geometry of real slopes is more complex. Farzaneh, O., Askari, F., and Ganjian, N. 2008. Three-dimensional stability anal-

Published by NRC Research Press

 172 Can. Geotech. J. Vol. 51, 2014

ysis of convex slopes in plan view. Journal of Geotechnical and Geoenviron- Michalowski, R.L., and Nadukuru, S.S. 2013. Three-dimensional limit analysis of
mental Engineering, 134(8): 1192–1200. doi:10.1061/(ASCE)1090-0241(2008)134: slopes with pore pressure. Journal of Geotechnical and Geoenvironmental
8(1192). Engineering, 139(9): 1604–1610. doi:10.1061/(ASCE)GT.1943-5606.0000867.
Fu, W., and Liao, Y. 2010. Non-linear shear strength reduction technique in slope Nadukuru, S.S., and Michalowski, R.L. 2013. Three-dimensional displacement
stability calculation. Computers and Geotechnics, 37: 288–298. doi:10.1016/j. analysis of slopes subjected to seismic loads. Canadian Geotechnical Journal,
compgeo.2009.11.002. 50(6): 650–661. doi:10.1139/cgj-2012-0223.
Gharti, H.N., Komatitsch, D., Oye, V., Martin, R., and Tromp, J. 2012. Application Naghadehi, M., Jimenez, R., Khalokakaie, R., and Jalali, S. 2013. A new open-pit
of an elastoplastic spectral-element method to 3D slope stability analysis. mine slope instability index defined using the improved rock engineering
International Journal for Numerical Methods in Engineering, 91(1): 1–26. systems approach. International Journal of Rock Mechanics and Mining Sci-
doi:10.1002/nme.3374. ences, 61: 1–14. doi:10.1016/j.ijrmms.2013.01.012.
Griffiths, D.V., and Marquez, R.M. 2007. Three-dimensional slope stability anal- Nian, T.-K., Huang, R.-Q., Wan, S.-S., and Chen, G.-Q. 2012. Three-dimensional
ysis by elasto-plastic finite elements. Géotechnique, 57(6): 537–546. doi:10.
strength-reduction finite element analysis of slopes: geometric effects. Cana-
1680/geot.2007.57.6.537.
dian Geotechnical Journal, 49(5): 574–588. doi:10.1139/t2012-014.
Hammah, R.E., Yacoub, T.E., Corkum, B.C., and Curran, J.H. 2005. The shear
Priest, S.D. 2005. Determination of shear strength and three-dimensional yield
strength reduction method for the generalized Hoek–Brown criterion. Amer-
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by University of North Dakota on 12/20/14

ican Rock Mechanics Association (ARMA/USRMS 05-810). strength for the Hoek–Brown criterion. Rock Mechanics and Rock Engineer-
Hoek, E., and Brown, E.T. 1980. Underground excavations in rock. Institution of ing, 38(4): 299–327. doi:10.1007/s00603-005-0056-5.
Mining and Metallurgy, London. Shen, H., and Abbas, S.M. 2013. Rock slope reliability analysis based on distinct
Hoek, E., Carranza-Torres, C., and Corkum, B. 2002. Hoek–Brown failure crite- element method and random set theory. International Journal of Rock Me-
rion - 2002 ed. In Proceedings of NARMS-TAC 2002, Mining Innovation and chanics and Mining Sciences, 61: 15–22. doi:10.1016/j.ijrmms.2013.02.003.
Technology, Toronto, Ont. Edited by R. Hammah, W. Bawden, J. Curran, and Shen, J. 2013. Analytical and numerical analyses for rock slope stability using
M. Telesnicki. Available from www.rocscience.com [accessed 20 September generalized Hoek–Brown criterion. Ph.D. thesis, School of Civil, Environmen-
2011]. tal and Mining Engineering, The University of Adelaide, Australia.
Itasca Consulting Group. 2009. FLAC3D 4.0 manual. Itasca Consulting Group, Shen, J., Karakus, M., and Xu, C. 2012a. Direct expressions for linearization of
Minneapolis. shear strength envelopes given by the Generalized Hoek–Brown criterion
Jimenez, R., Serrano, A., and Olalla, C. 2008. Linearization of the Hoek and using genetic programming. Computers and Geotechnics, 44: 139–146. doi:
Brown rock failure criterion for tunnelling in elasto-plastic rock masses. 10.1016/j.compgeo.2012.04.008.
International Journal of Rock Mechanics and Mining sciences, 45: 1153–1163. Shen, J., Priest, S.D., and Karakus, M. 2012b. Determination of Mohr–Coulomb
doi:10.1016/j.ijrmms.2007.12.003. shear strength parameters from generalized Hoek–Brown criterion for slope
Karakus, M. 2007. Appraising the methods accounting for 3D tunnelling effects stability analysis. Rock Mechanics and Rock Engineering, 45: 123–129. doi:10.
in 2D plane strain FE analysis. Tunnelling and Underground Space Technol- 1007/s00603-011-0184-z.
ogy, 22(1): 47–56. doi:10.1016/j.tust.2006.01.004. Stianson, J.R., Fredlund, D.G., and Chan, D. 2011. Three-dimensional slope sta-
Karakus, M., Ozsan, A., and Başarir, H. 2007. Finite element analysis for the twin bility based on stresses from a stress-deformation analysis. Canadian
metro tunnel constructed in Ankara Clay, Turkey. Bulletin of Engineering Geotechnical Journal, 48(6): 891–904. doi:10.1139/t11-006.
Geology and the Environment, 66(1): 71–79. doi:10.1007/s10064-006-0056-z.
Taheri, A., and Tani, K. 2010. Assessment of the stability of rock slopes by the
Kumar, P. 1998. Shear failure envelope of Hoek–Brown criterion for rockmass.
slope stability rating classification system. Rock Mechanics and Rock Engi-
For personal use only.

Tunnelling and Underground Space Technology, 13(4): 453–458. doi:10.1016/

neering, 43: 321–333. doi:10.1007/s00603-009-0050-4.
S0886-7798(98)00088-1.
Tao, M., Li, X., and Li, D. 2013. Rock failure induced by dynamic unloading under
Li, A.J., Merifield, R.S., and Lyamin, A.V. 2008. Stability charts for rock slopes
based on the Hoek-Brown failure criterion. International Journal of Rock 3D stress state. Theoretical and Applied Fracture Mechanics, 65: 47–54. doi:
Mechanics and Mining Sciences, 45(5): 689–700. doi:10.1016/j.ijrmms.2007.08. 10.1016/j.tafmec.2013.05.007.
010. Tutluoglu, L., Ferid Öge, I., and Karpuz, C. 2011. Two and three dimensional
Li, A.J., Merifield, R.S., and Lyamin, A.V. 2009. Limit analysis solutions for three analysis of a slope failure in a lignite mine. Computers & Geosciences, 37(2):
dimensional undrained slopes. Computers and Geotechnics, 36(8): 1330– 232–240. doi:10.1016/j.cageo.2010.09.004.
1351. doi:10.1016/j.compgeo.2009.06.002. Wei, W.B., Cheng, Y.M., and Li, L. 2009. Three-dimensional slope failure analysis
Li, A.J., Merifield, R.S., and Lyamin, A.V. 2010. Three-dimensional stability charts by the strength reduction and limit equilibrium methods. Computers and
for slopes based on limit analysis methods. Canadian Geotechnical Journal, Geotechnics, 36(1–2): 70–80. doi:10.1016/j.compgeo.2008.03.003.
47(12): 1316–1334. doi:10.1139/T10-030. Zhang, Y., Chen, G., Zheng, L., Li, Y., and Zhuang, X. 2013. Effects of geometries
Michalowski, R.L. 2010. Limit analysis and stability charts for 3D slope failures. on three-dimensional slope stability. Canadian Geotechnical Journal, 50(3):
Journal of Geotechnical and Geoenvironmental Engineering, 136(4): 583– 233–249. doi:10.1139/cgj-2012-0279.
593. doi:10.1061/(ASCE)GT.1943-5606.0000251. Zheng, H. 2012. A three-dimensional rigorous method for stability analysis of
Michalowski, R.L., and Drescher, A. 2009. Three-dimensional stability of slopes landslides. Engineering Geology, 145–146: 30–40. doi:10.1016/j.enggeo.2012.
and excavations. Géotechnique, 59(10): 839–850. doi:10.1680/geot.8.P.136. 06.010.

Published by NRC Research Press