Engineering Geology: Roohollah Kalatehjari, Ahmad Safuan A Rashid, Mohsen Hajihassani, Mehrdad Kholghifard, Nazri Ali

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 12

ENGEO-03801; No of Pages 12

Engineering Geology xxx (2014) xxx–xxx

Contents lists available at ScienceDirect

Engineering Geology
journal homepage: www.elsevier.com/locate/enggeo

Determining the unique direction of sliding in three-dimensional slope


stability analysis
Roohollah Kalatehjari a, Ahmad Safuan A Rashid a,⁎, Mohsen Hajihassani a, Mehrdad Kholghifard b, Nazri Ali a
a
Department of Geotechnics and Transportation, Faculty of Civil Engineering, Universiti Teknologi Malaysia, 81310 UTM Johor Bahru, Johor, Malaysia
b
Department of Civil Engineering, Islamic Azad University, Yasouj Branch, Yasouj, Iran

a r t i c l e i n f o a b s t r a c t

Article history: The present study developed a new method to calculate the unique direction of sliding. In addition, a computer
Accepted 2 June 2014 code was written to generate a realistic three-dimensional model of slopes and perform slope stability analysis.
Available online xxxx The application of the presented method was verified by re-analyzing two benchmark slope stability problems
from the literature and a hypothetical slope stability problem including four different slopes. This method is
Keywords:
suitable for geometrically complex slopes since there is no need to define a symmetrical plane for the slope.
Slope stability
Factor of safety
Moreover, the automatic process of the presented method makes it suitable for use in advanced search
Direction of sliding algorithms to determine the critical slip surface of slopes. These outcomes may contribute to managing the
Limit equilibrium method problems of conventional two-dimensional and three-dimensional slope stability analyses in terms of complex
Slip surface slopes and resolve the shortcomings of the previous methods in determining the direction of sliding.
© 2014 Elsevier B.V. All rights reserved.

1. Introduction longitudinal direction, especially when the geometry and other charac-
teristics of the slope are complex in the third dimension. One of the
New analytical methods and the integration of new technologies main features of a 3D analysis is the 3D shape of the slip surface,
and instruments in slope stability analysis have provided advanced which has been studied by several researchers. This parameter can
analysis, monitoring, and precautionary methods for landslides and significantly affect the result of analysis by increasing the value of the
slope failures (e.g. Dai et al., 2002; Donati and Turrini, 2002; Jing and factor of safety (FOS) as a result of the end effect of the slip surface in
Hudson, 2002; Lee et al., 2004; Terhorst and Kreja, 2009; Jiao et al., 3D analysis (Baligh and Azzouz, 1975; Chen and Chameau, 1983;
2012; Zhang et al., 2012; Jiang et al., 2013; Jiao et al., 2013; Lowry Leshchinsky et al., 1985; Liang et al., 1999; Rassam and Williams,
et al., 2013; Massey et al., 2013). However, the limit equilibrium method 1999; Arellano and Stark, 2000; Huang and Tsai, 2000). The other signif-
(LEM) has remained a popular slope stability analysis method because icant feature of 3D analysis of slope stability is the direction of sliding
of its simple principles, understandable approach, and reliable results (DOS). The nature of 2D analysis requires the sliding body to move
(Kahatadeniya et al., 2009; Mendoza et al., 2009; Di Maio et al., 2010; through an assumed direction, but it is possible to calculate the DOS in
Ferrari et al., 2011; Zheng, 2012; Huang, 2013; Kalatehjari and Ali, a 3D problem. The DOS is an angle calculated anticlockwise from the
2013; Thiebes et al., 2013). Assessment of the stability of slopes by positive x-axis to the route of sliding which shows the direction of
LEM is applied in two-dimensional (2D) and three-dimensional (3D) movement of the sliding mass in the x–y plane (Figure 1). This param-
analyses. Both of the methods have limitations in terms of modeling eter was introduced by Huang and Tsai (2000) when calculating the
and analyzing the shape of slopes and real sliding procedure. Particular- 3D FOS. The importance of this parameter should be addressed as it
ly, two-dimensional analyses mainly reduce the problem to the plain- affects the value of the FOS and eventually the determined critical slip
strain condition. All of these 2D methods ignore the third dimension surface (CSS). Disregarding the DOS in the calculation of 3D FOS may
of the problem by considering infinite width for the slope. Obviously, cause miscalculations in the stability analysis of slopes.
slopes are generally limited in the third dimension and involve some Starting with Anagnosti (1969), several 3D methods based on LEM
changes in the longitudinal direction. Therefore, performing a 3D anal- have been established to analyze the stability of slopes by considering
ysis may best present the slope condition including the changes in its the third dimension. The majority of these methods are limited in prac-
tice, however, because of their assumptions and limitations. Adopting a
plane of symmetry for the sliding mass is a common assumption in 3D
⁎ Corresponding author.
methods which is equal to assuming the DOS (e.g. Hovland, 1977;
E-mail addresses: kjroohollah2@live.utm.my (R. Kalatehjari), ahmadsafuan@utm.my,
ahmadsafuanutm@gmail.com (A.S. A Rashid), mohsen_hajihassani@yahoo.com Chen and Chameau, 1983; Dennhardt and Forster, 1985; Leshchinsky
(M. Hajihassani), kholghifard.m@gmail.com (M. Kholghifard), nazriali@utm.my (N. Ali). et al., 1985; Ugai, 1985; Leshchinsky and Baker, 1986; Baker and

http://dx.doi.org/10.1016/j.enggeo.2014.06.002
0013-7952/© 2014 Elsevier B.V. All rights reserved.

Please cite this article as: Kalatehjari, R., et al., Determining the unique direction of sliding in three-dimensional slope stability analysis, Eng. Geol.
(2014), http://dx.doi.org/10.1016/j.enggeo.2014.06.002
2 R. Kalatehjari et al. / Engineering Geology xxx (2014) xxx–xxx

An improvised 3D method of slope stability analysis was developed


and introduced in this paper to calculate the unique DOS for general
slopes. This method is based on LEM as an eclectic method able to calcu-
late the unique DOS and its corresponding FOS for general slopes with
external forces, different layers, different water conditions, and complex
geometries.

2. The proposed three-dimensional method

Almost all generalized 3D slope stability analysis methods based on


LEM apply the method of columns (Yamagami and Jiang, 1996; Huang
and Tsai, 2000; Huang et al., 2002; Cheng and Yip, 2007). The first
step in this method is to discretize the selected sliding mass into a num-
ber of “columns”. A typical 3D column system is formed by applying a
square net on the sliding mass. The grid lines are usually parallel with
the horizontal axes and the grid width value determines the spacing
between grid lines as well as the width of columns. A column requires
slope material above the slip surface. All sides of the column are
Fig. 1. Discretizing of a sliding mass by using square grid.
assumed to be planar which is practically appropriate if grid width
values are small enough. The “active columns” are defined as the col-
Leshchinsky, 1987; Hungr, 1987; Leshchinsky and Mullet, 1988; Ugai, umn central points which are inside the boundaries of the sliding
1988; Xing, 1988; Hungr et al., 1989; Cavounidis and Kalogeropoulos, mass. The columns outside the sliding mass are ignored in the calcula-
1992; Leshchinsky and Huang, 1992; Lam and Fredlund, 1993; Jiang tions as “inactive columns”. Consequently, the discretized sliding mass
and Yamagami, 2004; Zheng, 2012). Consequently, the sliding body is is defined as the sum of active columns. A system of classification is
bound to move along the plane of symmetry to keep the force/moment used to indicate each column by its location in x- and y-directions.
equations in balance. This assumption eases analysis and speeds up Fig. 1 illustrates the typical discretization of a sliding mass by means
calculations by assuming that the sliding mass is a symmetrical body of a square grid. The following steps were designed and coded in the
and replacing the whole sliding mass with its half. However, there is computer program to discretize the sliding mass.
no guarantee of a true estimate of the DOS, since some asymmetric
parameters (such as different layers and local loads) may be engaged 1. The maximum and minimum boundaries of the slope model in x
with the slope. Moreover, the assumption of such planes is not applica- and y directions are provided by the user as [xmin, xmax] and
ble in complex slopes with asymmetric geometry. Therefore, the practi- [ymin, ymax].
cal application of a large number of 3D methods is limited to symmetric 2. The available 3D data of the slope surface are entered by the user as
problems. (x, y, z).
Some researchers have tried to calculate the DOS in asymmetric 3. The square grid width is provided by the user (d).
problems to generalize the application of 3D methods (Yamagami and 4. A square grid is generated in the x–y plane by dividing the distance
Jiang, 1996; Huang and Tsai, 2000; Huang et al., 2002; Cheng and Yip, between minimum and maximum boundaries by the grid width
2007). These methods also have limitations in basic theories and prac- in both x and y directions as [n = (xmax − xmin) / d] and [m =
tice. As an instance, Yamagami and Jiang (1996) proposed the first (ymax − ymin) / d].
method to calculate the DOS which was restricted to gentle slopes 5. Matrices of x and y coordinates of grid points are generated for later
with an inclination of up to 45°. This method was called into question use as [xg] and [yg]. The dimension of these matrices is (n × m).
by Cheng and Yip (2007), because of its extremely time-consuming pro- 6. The height of each grid point is obtained by either interpolation or
cess to find the DOS. They concluded that the limitations of this method extrapolation between the available 3D data of the slope surface
excluded it from being generally applicable in practice. Another method and the obtained z values are recorded in a new matrix as [zs].
belongs to Huang and Tsai (2000), who included the DOS in their 3D 7. Consequently each point of the square grid in slope surface can be
analysis without proposing a practical calculation method. In addition addressed by its x, y and z coordinates as (xg, yg, zs).
to this, this method ignored horizontal forces, which resulted in unreli- 8. A 3D equation of slip surface is assumed. This equation can apply to
able results for laterally highly asymmetrical slopes as reported by Chen ellipsoidal, spherical, or cylindrical shapes depending on the
et al. (2003). Huang et al. (2002) proposed a method to calculate the problem conditions and requirements. For instance, the radius (r)
DOS by considering the equations of directional factors of safety and coordinates of the central point of the spherical slip surface
(DFOS) for a sliding mass divided into a series of columns. This method (x0, y0, z0) are assumed in Eq. (1).
assumed different directions of sliding for column bases at the verge of
2 2 2 2
failure. Although this assumption accelerated the calculation of the ðx−x0 Þ þ ðy−y0 Þ þ ðz−z0 Þ ¼ r ð1Þ
overall DOS, it meant that the soil columns were detached on the
verge of failure. This assumption is in contrast with one of LEM's basic 9. The height of each point of the slip surface (z = zss) on the square
definitions for the rigidity of the failure mass. Cheng and Yip (2007) grid is obtained by solving the equation of the slip surface for x = xg
also observed some convergence problems with this method in analyz- and y = yg and the obtained z values are recorded in a new matrix
ing highly asymmetric slopes as well as slopes under transverse loads. as [zss].
They proposed a complicated procedure to analyze the stability of 10. Consequently each point of the square grid in slip surface can be
general slopes and calculate the DOS. This method used several inter- addressed by its x, y and z coordinates as (xg, yg, zss).
connected iterative processes to determine its unknown variables and 11. Finally, the produced matrices of [xg], [yg], [zx], and [zss] are used to
applied a discrete interval of angles to determine the DOS. Major disad- generate soil columns inside the sliding mass.
vantages of this method were its reliance on the interval of the DOS and
the increased calculation time required to improve the accuracy of the After discretization of the sliding mass, the force and moment equa-
DOS. Consequently, none of the existing methods could calculate the tions can be applied for the columns. Fig. 2 illustrates the internal and
DOS successfully. external forces acting on a typical column, where Wi,j is the weight of

Please cite this article as: Kalatehjari, R., et al., Determining the unique direction of sliding in three-dimensional slope stability analysis, Eng. Geol.
(2014), http://dx.doi.org/10.1016/j.enggeo.2014.06.002
R. Kalatehjari et al. / Engineering Geology xxx (2014) xxx–xxx 3

unique on the horizontal plane for all columns, but different angles
are produced as projections of the DOS on the base of columns.
Conversely, in the second and third assumptions the DOS is not a unique
angle on the horizontal plane and can be calculated as the mean angle of
directions of sliding of different columns. In this case, the space sliding
angles on the base columns can be either different or equal. Fig. 3
depicts the three assumptions.
As a basic principle of LEM, the sliding mass is assumed to be a rigid
body on the verge of failure. This principle is repealed if the soil columns
move towards or away from each other. Also, the sum of forces and
moments in static equations is no longer valid in the mentioned cases
(Cheng and Yip, 2007). Consequently, the first assumption is the only
true option which can obtain a unique DOS. The definition of FOS as
the ratio of resisting forces to driving forces and the Mohr–Coulomb
failure criterion can be used to calculate the value of FOS at the base of
each column.

Sfi; j C0i; j þ N0i; j tan ϕ0i; j


Fsi; j ¼ ¼ ð2Þ
Si; j Si; j

0 0
Ci; j ¼ ci; j Ai; j ð3Þ

0
Ni; j ¼ Ni; j −Ui; j ð4Þ

Ui; j ¼ ui; j Ai; j ð5Þ

where at the base of column (i,j) Fsi,j is the value of FOS, Sfi,j is the shear
strength, N′i,j is the effective normal force, C′i,j and c′i,j are the total and
average effective cohesion of the soil, respectively, Ui,j and ui,j are the
Fig. 2. Internal and external forces acting on a typical column.
total and average pore water pressure, respectively, Ai,j is the true
area, and ϕ′i,j is the effective friction angle.
the soil column, Lzi,j is the external vertical load, Lxi,j and Lyi,j are the The shear and effective normal forces at the base of column are
external horizontal loads in the x- and y-directions, Fevi,j is the vertical vectors and can be expressed by their components in three main direc-
force induced by seismic force, Fehxi,j and Fehyi,j are the horizontal forces tions as follows.
induced by seismic force in the x- and y-directions, Exi − 1,j and Exi,j
are the intercolumn normal forces in the x- and y-directions, Eyi,j − 1
n o n o
and Exi,j are the intercolumn normal forces in the y-direction, Xxi − 1,j Si; j ¼ Sxi; j ; Syi; j ; Szi; j ¼ Si; j f 1ij ; f 2ij ; f 3ij ð6Þ
and Xxi,j are the vertical intercolumn shear forces in the x-direction,
X yi,j − 1 and X xi,j are the vertical intercolumn shear forces in the
y-direction, Hxi − 1,j and Hxi,j are the horizontal intercolumn shear n o n o
0 0 0 0 0
forces in the x-direction, Hyi,j − 1 and Hxi,j are the horizontal inter- Ni; j ¼ Nxi; j ; Nyi; j ; Nzi; j ¼ Ni; j g1ij ; g2ij ; g3ij ð7Þ
column shear forces in the y-direction, and Si,j and Ni,j are the shear
force mobilized and the total normal force at the base of the column, re- where the components of each force are described in the x-, y-, and
spectively. The points of action of forces are clarified in the formulation z-directions as Sxi,j, Syi,j, and Szi,j for Si,j and N′xi,j, N′yi,j, and N′zi,j for
procedure and the convention signs of forces are positive when they act N′i,j, respectively. Moreover, the unit vectors of force are described in
in the positive direction of axes. the x-, y-, and z-directions as f1, f2, and f3 for base shear force and g1,
g2, and g3 for base normal force, respectively.
3. Calculation of the unique direction of sliding Fig. 4 shows different angles and shear forces at the base plane of a
typical column. The base inclinations in the x- and y-directions are
This paper utilizes the definition of DFOS introduced by Huang and calculated at the center of the base of the column and the unique DOS
Tsai (2000) to obtain different values of FOS in different directions. is calculated anticlockwise from the positive x-direction in the direction
This is possible by calculating the corresponding resisting and mobiliz- of S′i,j.
ing forces in each direction. A similar procedure to Huang et al.'s Where for column (i,j), a′ is the unique DOS on the x–y plane, ai,j is
(2002) was employed to calculate the directional FOS in the x- and the projection of DOS on the base plane, θi,j is the angle between the
y-directions. In order to discretize the acting forces in the x- and sides of the column on its base plane, axi,j is the inclination of the base
y-directions, the DOS and its space angle on the base of each column of the column calculated in the x-direction at its center, ayi,j is the
are needed. Three different assumptions can be made for a DOS on the inclination of the base of the column calculated in the y-direction at
verge of failure in a 3D slope stability analysis based on the column its center, Sxzi,j and Syzi,j are the components of Si,j in the x–z and y–z
method. First, soil columns move in the same order and have a unique planes, respectively, and S′i,j is the projection of Si,j on the x–y plane.
DOS. The second assumption states that soil columns move towards The definition of directional FOS was introduced by Huang and Tsai
each other and the third one is that soil columns move away from (2000). Based on this concept, the value of FOS can be calculated in
each other. If the first assumption is taken into account, the DOS is any direction as the ratio of components of resisting forces to deriving

Please cite this article as: Kalatehjari, R., et al., Determining the unique direction of sliding in three-dimensional slope stability analysis, Eng. Geol.
(2014), http://dx.doi.org/10.1016/j.enggeo.2014.06.002
4 R. Kalatehjari et al. / Engineering Geology xxx (2014) xxx–xxx

Fig. 3. Three assumptions of the DOS: (a) soil columns move in the same order, (b) soil columns move towards each other, and (c) soil columns move away from each other.

forces at the base of column along that direction. Consequently, the Sfi; j
Fsy ¼ ð9Þ
FOS in the x- and y-directions can be obtained by calculating the Syzi; j
components of resisting and driving forces at the base of the columns
in the x–z and y–z planes, respectively. The resultant values of FOS
On the other hand, the following equations can be obtained by using
in the x- and y-directions are known as DFOS and are obtained by
the law of sines for the force polygons of Si,j and S′i,j at the base of the
substituting the components of forces in the FOS equation. The DFOS
column and the x–y plane, respectively.
can be calculated by the following equations as Fsx and Fsy, respectively.
Since LEM assumes that the sliding mass is a rigid body, the values of Sxi; j
DFOS have to be the same for all columns in each direction to simulta- Sxzi; j ¼ ð10Þ
cos axi; j
neously drive the columns to the verge of failure.

Sfi; j
Fsx ¼ ð8Þ Syi; j
Sxzi; j Syzi; j ¼ ð11Þ
cos ayi; j

0 0
Sxi; j ¼ cos a S i; j ð12Þ

0 0
Syi; j ¼ sin a S i; j ð13Þ

 
Sxzi; j sin θi; j S sin a
Si; j ¼   ¼ yzi; j  i; j ð14Þ
sin θi; j −ai; j sin θi; j

Then, Eqs. (8) and (9) can be revised on the basis of the unique DOS
with the help of Eqs. (10) to (13) as follows.

Sfi; j cos axi; j


Fsx ¼ ð15Þ
cos a0 S0 i; j

Sfi; j cos ayi; j


Fsy ¼ ð16Þ
sin a0 S0 i; j
Fig. 4. Angles and forces at the base plane of a typical column.

Please cite this article as: Kalatehjari, R., et al., Determining the unique direction of sliding in three-dimensional slope stability analysis, Eng. Geol.
(2014), http://dx.doi.org/10.1016/j.enggeo.2014.06.002
R. Kalatehjari et al. / Engineering Geology xxx (2014) xxx–xxx 5

Fig. 5. Vertical cut of example 1.

Moreover, by means of the definitions of DFOS in Eqs. (8) and (9) and
substitution of the equivalents of Si,j from Eq. (14) into Eq. (2), Eqs. (17)
and (18) are obtained. Fsi,j is the value of FOS in DOS at the base of each
column. Noticeably, this value has to be unique for each column
base whether it is calculated by Eq. (17) or by Eq. (18) in the x- and y-
directions, respectively. Eq. (19) is achieved by equalizing the third
parts of Eqs. (17) and (18) which are used to find the satisfactory
value of ai,j.

   
Sfi; j sin θi; j −ai; j Fsx sin θi; j −ai; j
Fsi; j ¼   ¼   ð17Þ
Sxzi; j sin θi; j sin θi; j

Sfi; j sin ai; j F sin ai; j


Fsi; j ¼   ¼ sy   ð18Þ
Syzi; j sin θi; j sin θi; j

0  1
F
−1 @ sx
sin θi; j −ai; j
ai; j ¼ sin A ð19Þ
Fsy

Since the calculation process is iterative, it is started by adopting


initial values of directional factors of safety (Fsx, and Fsy) and sliding
direction (a′). These values should be selected properly for each prob-
lem to save the calculation time and avoid the convergence problem.
It was shown in the literature that the initial value of 1 is suitable for
Fig. 6. Generated models for example 1.
Fsx, and Fsy in general cases of symmetrical slope stability problems
(Huang et al., 2002). In addition, based on the experience of the authors,
initial a′ may affect the convergence of the method. In fact, this param-
eter is a starting point of the Secant method, which is used to calculate After adopting the initial parameters, the value of θi,j is directly
the DOS and it is sensitive to its starting points. However, to the knowl- achieved by Eq. (20) for all columns. Then, values of ai,j are calculated
edge of the authors there is no mathematical method that recommends for the base of different columns by Eq. (21). After the calculation of
a general set of these initial values. Consequently, these values should new values of Fsx and Fsy from Eqs. (8) and (9), new values of ai,j are cal-
be selected based on the conditions of each problem. culated by using the Secant method and iterative calculation of Eq. (19).
Finally, a new trial value for unique DOS is calculated by Eq. (22) where
ac is the total number of columns. The previous values of a′, Fsx, and Fsy
Table 1 are replaced with the new values in the next cycle of calculation. The
Properties of soil of example 1. main difference between the proposed method and the methods of
Layer c′ (kN/m2) ϕ′ (°) γ (kN/m3)
Huang and Tsai (2000) and Huang et al. (2002) is Eq. (22). This new
equation provides a unique DOS for all columns in each step of the
Slope material 24.5 20 17.64
iterative calculation process. Consequently, the rigidity of the sliding

Please cite this article as: Kalatehjari, R., et al., Determining the unique direction of sliding in three-dimensional slope stability analysis, Eng. Geol.
(2014), http://dx.doi.org/10.1016/j.enggeo.2014.06.002
6 R. Kalatehjari et al. / Engineering Geology xxx (2014) xxx–xxx

Table 2
Results of different studies for example 1.

3D method FOS

Present study Huang and Tsai (2000) Difference (%) Cheng and Yip (2007) Difference (%)

Simplified Bishop 1.776 1.766 0.57 1.801 1.39


Simplified Janbu 2.745 – – 2.782 1.33

body is maintained during the analysis as a principle of LEM and the Considering Eqs. (15) and (16), the values of DFOS in the x- and y-
unique DOS is directly calculated at the end. directions tend to infinity in the following cases:

  1. in Eq. (15) if a′ = π/2 then cos a′ = 0 and Fsx = ∞


−1
θi; j ¼ cos sin axi; j sin ayi; j ð20Þ 2. in Eq. (16) if a′ = 0, π then sin a′ = 0 and Fsy = ∞.

2 3 When Eq. (23) is satisfied, the overall FOS is obtained by dividing the
−1 4 sin θi; j sum of resisting forces (Sfi,j) by the sum of driving forces (Si,j) for all
ai; j ¼ tan   5 ð21Þ
cos θi; j þ cos ayi; j = tan a0 cos axi; j active columns. Based on Eq. (23), Fsi,j is unique in both directions.
Consequently, the values of Sfi,j and Si,j are unique for each column
whether they are calculated in the x- or the y-direction. Therefore the
0 2 31 overall FOS is calculated by Eq. (24) with the values of Sfi,j and Si,j from
m X
X
0
n
−1 sin ai; j cos ayi; j the x- or y-direction. In the case of infinite DFOS, however, the calcula-
a ¼ @ tan 4   5A=ac ð22Þ
j¼1 i¼1 sin θi; j −ai; j cos axi; j tion of overall FOS can be applied by using the forces in the direction of
finite DFOS.

A termination criterion is required to break the calculation cycle Xm Xn


Sfi; j
when the required solution is found. As mentioned before, the calculat- j¼1
FOS ¼ Xm Xn
i¼1
ð24Þ
ed value of Fsi,j has to be unique for each column base whether it is j¼1
S
i¼1 i; j
calculated from the x- or from the y-direction. Consequently, the differ-
ence between calculated values of Fsi,j for each column in the x- and y-
Through the proposed method, the unique DOS, and its correspond-
directions has to be smaller than an epsilon (ε) which is sufficiently
ing FOS are calculated with high accuracy for 3D slopes. Consequently,
close to zero. This condition is used to define a termination criterion
the minimum FOS and probable DOS can be predicted for possible
with ε smaller than 0.001 for the iterative process in Eq. (23).
slope mass failure.
Xm Xn  

F
i¼1 si; jðin x‐directionÞ
− Fsi; jðin y‐directionÞ  ¼ ε ð23Þ
j¼1 4. Numerical examples

The main parameters which account for the accuracy of the result
are the 3D method applied to find the FOS, the applied grid width, the
strategy for discretizing the sliding mass, and the shape of the slip
surface. In order to verify the results of the current study, several 3D
benchmark examples were analyzed by the proposed method and the
results were compared with those of previous studies. It should be
noted that a computer code was written in Matlab by the authors to
be used for a 3D model of the slope, generating 3D probable slip
surfaces, and performing the calculations.

4.1. Example 1

Huang and Tsai (2000) applied a sample model to test the perfor-
mance of its two-directional method in finding the DOS and 3D FOS.
Cheng and Yip (2007) re-analyzed the same model by extending 3D
methods based on simplified 2D methods (Janbu, 1954; Bishop, 1955)
to verify their study. Fig. 5 shows the geometry of the corner of a vertical
cut. This example was used especially to verify the performance of the
current study to find the DOS. This was because the corresponding
sliding mass has a symmetrical shape about an axis with inclination of
45° to the x-axis. This direction was predicted to be the DOS.
The example 1 included a five-meter-high vertical cut slope with a
spherical slip surface. The center point of the sliding surface was placed
at the corner of the vertical cut with a radius of 5 m. Table 1 shows the
properties of slope material in example 1.
The 3D model of the example was generated by the presented code.
Fig. 6 illustrates the generated models of example 1, where the vertical
cut, the sliding body, and the failure model are presented as a, b, and c,
respectively. In order to compare the results of the present study with
Fig. 7. Geometry and properties of example 2. those of the previous studies, two simplified methods including 3D

Please cite this article as: Kalatehjari, R., et al., Determining the unique direction of sliding in three-dimensional slope stability analysis, Eng. Geol.
(2014), http://dx.doi.org/10.1016/j.enggeo.2014.06.002
R. Kalatehjari et al. / Engineering Geology xxx (2014) xxx–xxx 7

simplified Bishop and 3D simplified Janbu were employed to re-analyze results of this study and the previous methods together with the same
the example. The two methods applied in the present study achieved a DOS in all studies demonstrated the accuracy of the current study.
DOS exactly equal to 45°, which was the same as Huang and Tsai's
(2000) and Cheng and Yip's (2007). Moreover, this DOS was found 4.2. Example 2
similar to the predicted direction of the axis of symmetry. These result
demonstrated the ability of the present study to find the DOS. In order to examine the performance of the presented method
Table 2 presents 3D FOS obtained in different studies for example 1. for calculating the unique DOS, a hypothetical slope stability problem
The applied methods of the present study produced results close was generated. Fig. 7 shows the geometry of this example with its
to those of Huang and Tsai (2000) and Cheng and Yip (2007). The geomechanical properties. This example involves four different slopes
difference between the results of this study and the previous results with spherical sliding surface. The boundaries of slopes were constant
for the 3D simplified Bishop method were 0.57% and 1.39%, respectively. and the face was rotated 10° clockwise in each step of the example.
Moreover, the difference between the results of the current study and Consequently, four different directions of sliding were predicted.
Cheng and Yip's (2007) for the 3D simplified Janbu method was 1.33%. All models were generated using the computer code. The top views
Although the same analysis methods were employed, the small dif- of generated models of slope and slip surfaces are shown in Fig. 8, which
ferences between the results could be the effect of applying different illustrates the intersection area between slope surface and slip surface.
methods of discretization and/or different grid widths. Such differences The sliding models are shown in Fig. 9. It can be seen that the shape
were also reported by Huang et al. (2002). Nevertheless, the similar and volume of sliding changed when the slope face angle altered. The

Fig. 8. Generated models of slopes with face angles of (a) 90, (b) 80, (c) 70, and (d) 60° from y-axis.

Please cite this article as: Kalatehjari, R., et al., Determining the unique direction of sliding in three-dimensional slope stability analysis, Eng. Geol.
(2014), http://dx.doi.org/10.1016/j.enggeo.2014.06.002
8 R. Kalatehjari et al. / Engineering Geology xxx (2014) xxx–xxx

Fig. 9. Generated models of sliding for slopes with face angles of (a) 90, (b) 80, (c) 70, and (d) 60° from y-axis.

calculated unique DOS and its corresponding 3D FOS for all conditions by sufficiently small grid width. The high accuracy in determining
are stated in Table 3. the unique DOS and the small difference ratios in calculated FOS
Clearly, the rotation of the slope face caused the unique DOS to demonstrate the ability of the proposed method and the computer
change. Expected and obtained DOS in Table 3 represent logically code to significantly contribute to calculating the unique DOS.
estimated DOS due to the rotation of the slope face and calculated
DOS by the proposed method, respectively. A maximum difference 4.3. Example 3
ratio of 0.4% between the expected and obtained angles of DOS shows
the accuracy of the presented method. According to this finding, the In order to compare the results of FOS between this current study
most unsafe downhill slope can be predicted, which is important for and the study of Huang et al. (2002) an extensive example of slope
both short-term and long-term neighboring engineering activities. As stability was employed. This example had been initially created by
predicted, the value of FOS was equal in all the models with a difference Baligh and Azzouz (1975) before being converted to 3D by Xing
ratio of less than 1.5%. This difference was because of the minor dissim-
ilarity in the modeling of the sliding body caused by the applied column
method. Since the column sides were parallel with the x–z and y–z
planes, any rotation of the slope face could cause a slight change in
the accuracy of the modeling. This problem can be minimized, however,

Table 3
Results of example 2.

Model Slope face angle FOS DOS to x-axis (°) Difference (%)
to y-axis (°)
Expected Obtained

1 90 2.44 90.00 90.00 0.00


2 80 2.46 80.00 79.88 0.15
3 70 2.47 70.00 69.72 0.40
4 60 2.47 60.00 59.81 0.32
Fig. 10. Geometry and properties of example 3.

Please cite this article as: Kalatehjari, R., et al., Determining the unique direction of sliding in three-dimensional slope stability analysis, Eng. Geol.
(2014), http://dx.doi.org/10.1016/j.enggeo.2014.06.002
R. Kalatehjari et al. / Engineering Geology xxx (2014) xxx–xxx 9

Table 4 of the slopes and perform the complicated calculations of 3D slope


Properties of soil layers of example 3. stability analysis. The definition of DFOS proposed by Huang et al.
Layer c′ (kN/m2) ϕ′ (°) γ (kN/m3) rua (2002) was used as the foundation of the present paper. Moreover, a
new relationship was introduced to calculate the unique DOS. The
Slope material 29 20 18.80 0.25
Weak layer 0 10 18.80 0.25 force and moment equilibriums of static forces were used to establish
a equations of DFOS in x–z and y–z planes for a mutual DOS. Then,
Applicable for cases 3 and 4.
the value of FOS along the DOS was calculated for each column base
by means of the relevant forces from the x- and y-directions. If
(1988) (Figure 10). Table 4 shows the properties of slope materials in they reach the true sliding angle, the values of FOS calculated from the
example 3. In this example, a 3D, symmetrical, circular failure surface x- and y-directions for each column base must be equal at the verge
in a cohesive slope was used. Six different 3D slope models were gener- of sliding to comply with the LEM principle. Consequently, the objective
ated to analyze all cases of the example. Table 5 shows the properties of was to minimize the difference of FOS calculated from x- and y-
the models for different cases. Different lengths of the sliding surface directions. Through this iterative process, the unique DOS and the
(Rx) were assigned to each case to evaluate the end effect or the effect overall 3D FOS were calculated. The main advantages of the presented
of the third dimension on the value of FOS. The range of Rx was selected method are as follows.
from the minimum applicable length to the maximum efficient length
of the slip surface. Fig. 11 illustrates the generated models by the current 1. In contrast to the majority of the existing 3D methods, there was no
study for case 6 as the most complex case. Huang et al. (2002) re- need to define a symmetrical plane for the slope. Thus, the method
analyzed four out of six possible cases of this example. was applicable for geometrically complex slopes.
Two methods were applied to calculate the value of FOS including 2. A global coordinate system was used in the analysis to reduce the
simplified Bishop and the rigorous methods, based on the applied complexity of the formulation.
methods in Huang et al. (2002). Table 6 shows the ranges of FOS for 3. The unique DOS was calculated with high accuracy without the use
all the six cases. It should be noted that the unique DOS was calculated of estimated, assumed, or interval values.
by the presented method as 90° from x-axis for all the cases, however 4. The corresponding 3D FOS of the sliding body was calculated with
Huang et al. (2002) did not report the sliding direction for this case. high accuracy through an understandable process within a main
Table 7 shows the results of the different 3D simplified methods. iteration.
Xing (1988) applied a simplified 3D method to analyze six different 5. The convergence of the method was not associated with the initial
cases of the slope stability. As mentioned above, Huang et al. (2002) values of variables such as initial DOS or initial DFOS.
analyzed four of the cases using the simplified Bishop method. Hungr 6. The overall process was controlled by a logical termination criterion
et al. (1989) and Lam and Fredlund (1993) re-analyzed the second to prevent immature and late convergence.
and the sixth cases using the simplified Bishop method. This current 7. The overall process was completely automatic and no interference by
study re-analyzed all the six cases. the user was required in the solving process.
Table 8 shows the results of the different 3D rigorous methods.
Huang et al. (2002) analyzed four cases of this example by the rigorous The presented method was verified by re-analyzing two benchmark
method. Chen et al. (2001, 2003), and Xie et al. (2006) re-analyzed slope stability examples from the literature and a hypothetical slope
the first two cases using their own rigorous methods. The current stability problem including four different slopes. The results shows the
study re-analyzed all the six cases. significant success of the presented method in an integrated work
The results of both simplified and rigorous methods of the current with the prepared computed code to generate the 3D model of slope
study were close to the findings of the previous studies. The maximum and slip surface, determine the unique DOS, and calculate the FOS in
difference of 7.67% between the results of the current study and Huang all the examples. The superiority of the presented method compared
et al.'s (2002) is believed to be due to the different assumptions in with two of the most general 3D slope stability methods (Huang et al.,
discretizing strategies and internal calculations of the methods. Sarma 2002; Cheng and Yip, 2007) are as follows.
(1973) showed that the mentioned factors might produce differences
as large as 18% in 2D analyses. By adding the third dimension to the 1. The conflict of Huang et al.'s (2002) method with the rigidity princi-
problems, there is a possibility that this difference grow larger. In ple of LEM was removed by introducing the equation of unique DOS.
addition, the different assumptions on intercolumn forces, the number 2. The accuracy of determined DOS and the solving time were signifi-
of columns, and the method of assigning soil columns may also affect cantly improved compared with the method of Cheng and Yip
the mentioned difference, as stated by Huang et al. (2002). Consequent- (2007). These enhancements were achieved by calculating the DOS
ly, the existence of difference is considered tolerable. rather than using an interval of sliding directions.
3. The difficulty of determining FOS when DOS met one of the main
5. Discussion and conclusion axes was resolved by a simple solution.

This paper developed a new method to calculate the unique DOS. In By avoiding the problematic situations and automation of the pro-
addition, a computer code was written to generate a realistic 3D model cess, the presented method is able to find the DOS and its corresponding

Table 5
Properties of produced models in different tests of example 3.

Case no. Slip surface Slip surface properties (m) Weak layer Water condition
length (m)
Xc Yc Zc Rx Ry Rz

1 100 50 13.4 27.4 10–67 24.4 24.4 – –


2 160 80 13.4 27.4 10–105 24.4 24.4 Yes –
3 200 100 13.4 27.4 10–130 24.4 24.4 – ru
4 250 125 13.4 27.4 10–160 24.4 24.4 Yes ru
5 160 80 13.4 27.4 10–100 24.4 24.4 – ps
6 160 80 13.4 27.4 10–107 24.4 24.4 Yes ps

Note: ps = piezometric surface and ru = coefficient of pore water pressure for both materials.

Please cite this article as: Kalatehjari, R., et al., Determining the unique direction of sliding in three-dimensional slope stability analysis, Eng. Geol.
(2014), http://dx.doi.org/10.1016/j.enggeo.2014.06.002
10 R. Kalatehjari et al. / Engineering Geology xxx (2014) xxx–xxx

Fig. 11. Generated models for case 6 of example 3.

Please cite this article as: Kalatehjari, R., et al., Determining the unique direction of sliding in three-dimensional slope stability analysis, Eng. Geol.
(2014), http://dx.doi.org/10.1016/j.enggeo.2014.06.002
R. Kalatehjari et al. / Engineering Geology xxx (2014) xxx–xxx 11

Table 6 References
The ranges of calculated FOS for all cases of example 3.

Case no. Length of slip Range of FOS Anagnosti, P., 1969. Three-dimensional stability of fill dams. Proceedings of 7th. In-
surface (m) ternational Conference on Soil Mechanics and Foundation Engineering, Mexico,
3D simplified 3D rigorous pp. 275–280.
Bishop method method Arellano, D., Stark, T.D., 2000. Importance of three-dimensional slope stability analyses in
1 10–67 2.438–2.085 2.707–2.344 practice. Geotechnical Special Publication, pp. 18–32.
Baker, R., Leshchinsky, D., 1987. Stability analysis of conical heaps. Soils Found. 27 (4),
2 10–105 1.979–1.568 2.142–1.758
99–110.
3 10–130 2.172–1.817 2.412–2.050
Baligh, M.M., Azzouz, A.S., 1975. End effects on stability of cohesive slopes. J. Geotech. Eng.
4 10–160 1.797–1.348 1.970–1.518 Div. 101 (11), 1105–1117.
5 10–100 2.288–1.913 2.563–2.186 Bishop, A.W., 1955. The use of the slip circle in the stability analysis of earth slope.
6 10–107 1.873–1.454 2.047–1.656 Geotechnique 5 (1), 7–17.
Cavounidis, S., Kalogeropoulos, H., 1992. End effects on the stability of cuts in normally
consolidated clays. Riv. Ital. Geotech. 2, 85–93.
FOS without any need for monitoring by users. This allows the method Chen, R., Chameau, J., 1983. Three-dimensional limit equilibrium analysis of slopes.
to be used in an automatic search for the CSS. It is hoped that with the Geotechnique 32 (1), 31–40.
Chen, Z., Wang, X., Haberfield, C., Yin, J., Wang, J., 2001. A three-dimensional slope
presented methods some of the problems owed to asymmetry of slopes stability analysis method using the upper bound theorem. Part I: theory and
can be addressed by a more realistic slope stability analysis. As a result, methods. Int. J. Rock Mech. Min. Sci. 38, 369–378.
generating a 3D model of complex slopes, finding their unique DOS to- Chen, Z., Mi, H., Zhang, F., Wang, X., 2003. A simplified method for 3D slope stability
analysis. Can. Geotech. J. 40, 675–683.
gether with the corresponding 3D FOS, and describing the 3D shape of Cheng, Y., Yip, C., 2007. Three-dimensional asymmetrical sloe stability analysis extension
slip surfaces can be a practical endeavor. Consequently, overestimated of Bishop's, Janbu's, and Morgenstern–Price's techniques. J. Geotech. Geoenviron. 133
and underestimated results of conventional slope stability analyses (12), 1544–1555.
Dai, F.C., Lee, C.F., Ngai, Y.Y., 2002. Landslide risk assessment and management: an over-
owed to estimated or assumed DOS can be corrected by using the pro-
view. Eng. Geol. 64 (1), 65–87.
posed method. Dennhardt, M., Forster, W., 1985. Problems of three-dimensional slope stability. Proceed-
The presented method is an innovative way to remove the limita- ings of the 11th International Conference in Soil Mechanics and Foundation Engineer-
ing, San Francisco, pp. 427–431 (Part 2).
tions of previous methods in calculating the unique direction of sliding.
Di Maio, C., Vassallo, R., Vallario, M., Pascale, S., Sdao, F., 2010. Structure and kinematics of
It is able to improve 3D slope stability analysis and provide safer results a landslide in a complex clayey formation of the Italian southern Apennines. Eng.
by predicting the direction of sliding. Moreover, it can remove the con- Geol. 116, 311–322.
flicts of previous methods with the basic limit equilibrium methods Donati, L., Turrini, M.C., 2002. An objective method to rank the importance of the factors
predisposing to landslides with the GIS methodology: application to an area of the
which commonly produce instability in analysis convergence over a log- Apennines (Valnerina; Perugia, Italy). Eng. Geol. 63 (3–4), 277–289.
ical FOS. Therefore, its framework can be applied in other 3D methods Ferrari, A., Ledesma, A., Gonzalez, D.A., Corominas, J., 2011. Effects of the foot evolution on
based on limit equilibrium principles. Overall, it can be considered as the behaviour of slow-moving landslides. Eng. Geol. 117, 217–228.
Hovland, H., 1977. Three-dimensional slope stability analysis method. J. Geotech. Eng. Div.
an innovative solution to improve the existing methods of 3D slope sta- 103 (9), 971–986.
bility analysis. Huang, C.C., 2013. Developing a new slice method for slope displacement analyses. Eng.
Geol. 157, 39–47.
Huang, C.C., Tsai, C.C., 2000. New method for 3D and asymmetrical slope stability analysis.
J. Geotech. Geoenviron. 126 (10), 917–927.
Acknowledgment Huang, C.C., Tsai, C.C., Chen, Y.H., 2002. Generalized method for three-dimensional slope
stability analysis. J. Geotech. Geoenviron. 128 (10), 836–848.
Hungr, O., 1987. An extension of Bishop's simplified method of slope stability analysis to
This study was conducted as part of a post-doctoral fellowship
three dimensions. Geotechnique 37 (1), 113–117.
scheme funded by Universiti Teknologi Malaysia (UTM) (PY/2014/ Hungr, O., Salgado, F., Byrne, P., 1989. Evaluation of a three-dimensional method of slope
01476). stability analysis. Can. Geotech. J. 26, 679–686.

Table 7
Comparison between the results of 3D simplified methods for example 3.

Case no. Sliding mass 3D simplified method Difference (%) between


volume the current study and
Xing Hungr Lam and Fredlund Huang and Tsai Huang Current
(m3)a Huang et al. (2002)
(1988) et al. (1993) (2000) et al. study
(1989) (2002)

1 11,106 2.122 – – – – 2.092 –


2 13,032 1.553 1.620 1.607 1.665 1.645 1.574 4.31
3 18,360 1.790 – – – 1.781 1.820 2.19
4 22,806 1.278 – – – 1.384 1.352 2.31
5 14,791 1.831 – – – 2.072 1.913 7.67
6 16,290 1.441 1.540 1.511 – – 1.461 –
a
Comparison volume is based on Xing (1988).

Table 8
Comparison between the results of 3D rigorous methods for example 3.

Case no. Sliding mass 3D rigorous method Difference (%) between


volume current study and
Chen Xie Chen Huang Current
(m3)a Huang et al. (2002)
et al. et al. et al. et al. study
(2001) (2006) (2003) (2002)

1 11,106 2.262 2.302 2.187 – 2.356 –


2 13,032 1.717 1.711 1.640 1.757 1.772 0.85
3 18,360 – – – 1.908 2.054 7.65
4 22,806 – – – 1.478 1.529 3.45
5 14,791 – – – 2.215 2.189 1.17
6 16,290 – 1.620 – – 1.669 –
a
Comparison volume is based on Xing (1988).

Please cite this article as: Kalatehjari, R., et al., Determining the unique direction of sliding in three-dimensional slope stability analysis, Eng. Geol.
(2014), http://dx.doi.org/10.1016/j.enggeo.2014.06.002
12 R. Kalatehjari et al. / Engineering Geology xxx (2014) xxx–xxx

Janbu, N., 1954. Application of composite slip surface for stability analysis. Proceedings of Lowry, B., Gomez, F., Zhou, W., Mooney, M.A., Held, B., Grasmick, J., 2013. High resolution
the Conference on Stability of Earth Slopes, Stockholm, Sweden, pp. 43–49 (Part 3). displacement monitoring of a slow velocity landslide using ground-based radar
Jiang, J.C., Yamagami, T., 2004. Three-dimensional slope stability analysis using an extended interferometry. Eng. Geol. 166, 160–169.
Spencer method. J. Japan. Geotech. Soc. Soils Found. 44 (4), 127–135. Massey, C.I., Petley, D.N., McSaveney, M.J., 2013. Patterns of movement in reactivated
Jiang, Q.H., Liu, X.H., Wei, W., Zhou, C.B., 2013. A new method for analyzing the stability of landslides. Eng. Geol. 159, 1–19.
rock wedges. Int. J. Rock Mech. Min. Sci. 60, 413–422. Mendoza, F.J.C., Gisbert, A.F., Izquierdo, A.G., Bovea, M.D., 2009. Safety factor nomograms
Jiao, Y.Y., Zhang, X.L., Zhao, J., 2012. A two-dimensional DDA contact constitutive model for homogeneous earth dams less than ten meters high. Eng. Geol. 105, 231–238.
for simulating rock fragmentation. J. Eng. Mech. ASCE 138 (2), 199–209. Rassam, D.W., Williams, D.J., 1999. 3-Dimensional effects on slope stability of high waste
Jiao, Y.Y., Wang, Z.H., Wang, X.Z., Adoko, A.C., Yang, Z.X., 2013. Stability assessment of an rock dumps. Int. J. Surf. Min. Reclam. Environ. 13 (1), 19–24.
ancient landslide crossed by two coal mine tunnels. Eng. Geol. 159, 36–44. Sarma, S.K., 1973. Stability analysis of embankments and slopes. Geotechnique 23 (3),
Jing, L.R., Hudson, J.A., 2002. Numerical methods in rock mechanics. Int. J. Rock Mech. Min. 423–433.
Sci. 39 (4), 409–427. Terhorst, B., Kreja, R., 2009. Slope stability modelling with SINMAP in a settlement area of
Kahatadeniya, K.S., Nanakorn, P., Neaupane, K.M., 2009. Determination of the critical the Swabian Alb. Landslides 6, 309–319.
failure surface for slope stability analysis using ant colony optimization. Eng. Geol. Thiebes, B., Bell, R., Glade, T., Jäger, S., Anderson, M., Holcombe, L., 2013. A WebGIS
108, 133–141. decision-support system for slope stability based on limit-equilibrium modelling.
Kalatehjari, R., Ali, N., 2013. A review of three-dimensional slope stability analyses based Eng. Geol. 158, 109–118.
on limit equilibrium method. Electron. J. Geotech. Eng. 18 (A), 119–134. Ugai, K., 1985. Three-dimensional stability analysis of vertical cohesive slopes. Soils
Lam, L., Fredlund, D., 1993. A general limit equilibrium model for three-dimensional slope Found. 25 (3), 41–48.
stability analysis. Can. Geotech. J. 30, 905–919. Ugai, K., 1988. Three-dimensional slope stability analysis by slice methods. Proceedings of
Lee, S., Ryu, J.H., Won, J.S., Park, H.J., 2004. Determination and application of the weights the International Conference on Numerical Methods in Geomechanics, Innsbruck,
for landslide susceptibility mapping using an artificial neural network. Eng. Geol. 71 Austria, pp. 1369–1374.
(3–4), 289–302. Xie, M., Esaki, T., Cai, M., 2006. GIS-based implementation of three-dimensional limit
Leshchinsky, D., Baker, R., 1986. Three-dimensional slope stability: end effects. Soils equilibrium approach of slope stability. J. Geotech. Geoenviron. 132 (5), 656–660.
Found. 26 (4), 98–110. Xing, Z., 1988. Three-dimensional stability analysis of concave slopes in plan view. J.
Leshchinsky, D., Huang, C.C., 1992. Generalized three-dimensional slope stability analysis. Geotech. Eng. 114 (6), 658–671.
J. Geotech. Eng. 118 (11), 1748–1764. Yamagami, T., Jiang, J.C., 1996. Determination of the sliding direction in three-dimensional
Leshchinsky, D., Mullet, T., 1988. Stability of vertical corner cuts. Proceedings of the Sixth slope stability analysis. Proceedings of the 2nd International Conference on Soft Soil
International Conference on Numerical Methods in Geomechanics, Innsbruck, Engineering. Hohai University Press, Nanjing, pp. 567–572 (Part 1).
Austria, pp. 1249–1256. Zhang, D., Wang, G., Yang, T., Zhang, M., Chen, S., Zhang, F., 2012. Satellite remote sensing-
Leshchinsky, D., Baker, R., Silver, M., 1985. Three-dimensional analysis of slope stability. based detection of the deformation of a reservoir bank slope in Laxiwa Hydropower
Int. J. Numer. Anal. Methods Geomech. 9, 199–223. Station, China. Landslides 10 (2), 231–238.
Liang, R.Y., Nusier, O.K., Malkawi, A.H., 1999. A reliability based approach for evaluating Zheng, H., 2012. A three-dimensional rigorous method for stability analysis of landslides.
the slope stability of embankment dams. Eng. Geol. 54 (3), 271–285. Eng. Geol. 145, 30–40.

Please cite this article as: Kalatehjari, R., et al., Determining the unique direction of sliding in three-dimensional slope stability analysis, Eng. Geol.
(2014), http://dx.doi.org/10.1016/j.enggeo.2014.06.002

You might also like