Engineering Geology 160 (2013) 2133

Contents lists available at SciVerse ScienceDirect

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

Analysis of stability of three-dimensional slopes using the rigorous

limit equilibrium method
X.P. Zhou a, b, c,, H. Cheng a
a
b
c

School of Civil Engineering, Chongqing University, Chongqing 400045, China

Key Laboratory of New Technology for Construction of Cities in Mountain Area, Chongqing University, Ministry of Education, Chongqing 400045, China
State Key Laboratory of Coal Mine Disaster Dynamics and Control, Chongqing University, Chongqing 400045, China

a r t i c l e

i n f o

Article history:
Received 6 June 2012
Received in revised form 18 March 2013
Accepted 30 March 2013
Available online 16 April 2013
Keywords:
Three-dimensional slope
Stability
Limit equilibrium method
Factor of safety
Least square tting

a b s t r a c t
Previously, quasi-rigorous limit equilibrium methods were applied to analyze the stability of three-dimensional
slopes or landslides, which only satisfy three direction force equilibrium, and one or two direction moment equilibrium. In this paper, the rigorous limit equilibrium column method, in which inter-column forces are taken into
account, is established based on six equilibrium conditions which include three direction force equilibrium conditions along coordinate axes and three direction moment equilibrium conditions around three coordinate axes.
The relationship between the width of sliding body and factor of safety is determined using trust-regionreective iterative algorithm. The value of the factor of safety is obtained using LevenbergMarquardt least
square method. Moreover, the present method can be applied to automatically search slip surface of threedimensional landslides and to determine the factor of safety of three-dimensional landslides with the known
arbitrary slip surface. Three examples are discussed to verify the robustness and precision of the present method
in detail. Comparing with quasi-rigorous limit equilibrium methods which only considered four or ve equilibrium conditions, the present method is more accurate and rigorous.
2013 Elsevier B.V. All rights reserved.

1. Introduction
Landslides are dened as the movement of a rock mass, debris or
earth down a slope (Cruden, 1991). They can be triggered by a variety
of external stimulus, such as earthquake shaking, intense rainfall,
water level change, articial disturbance or rapid stream erosion,
and so on. Landslides are one of the major natural hazards. As
shown in Fig. 1, Jipazi landslide which is mainly caused by intense
rainfall occurred in July of 1982 in the province of Sichuan. The volume of the Jipazi landslide was estimated to be 15,000,000 m 3 and
1700 buildings had been destroyed. Jipazi landslide caused enormous
losses in terms of both direct and indirect aspects.
At present, the conventional limit equilibrium method is still
playing a major role in practical slope or landslide engineering. In
particular, limit equilibrium methods are widely and maturely applied to analyze the stability of two-dimensional slopes or landslides.
For two-dimensional slopes or landslides, factor of safety is accurate
enough using limit equilibrium method which satises all equilibrium
conditions (Duncan, 1996). However, for practical slope or landslide
engineering, slopes or landslides are three-dimensional problems
which are not suitable for being simplied to two-dimensional ones.
Previously, when limit equilibrium columns methods were applied to
Corresponding author at: School of Civil Engineering, Chongqing University, Chongqing
400045, China. Fax: +86 23 6512 3511.
E-mail address: zhouxiaopinga@sina.com (X.P. Zhou).
0013-7952/$ see front matter 2013 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.enggeo.2013.03.027

study the stability of three-dimensional slopes or landslides, only partial

equilibrium conditions were satised, which were not in the framework
of rigorous limit equilibrium method. For example, only three equilibrium conditions are satised, which is only suitable for symmetric slopes
or landslides(Hungr, 1987; Zhang, 1988; Hungr et al., 1989; Lam and
Fredlund, 1993), only four or ve equilibrium conditions are satised
(Chen et al., 2001a, 2001b; Huang et al., 2002; Zhang et al., 2005). Moreover, the factor of safety, which is obtained from quasi-rigorous limit
equilibrium methods, was not accurate enough to meet the engineering
requirements (Baligh and Azzouz, 1975; Hovland, 1979; Hungr, 1987;
Zhang, 1988; Hungr et al., 1989; Lam and Fredlund, 1993; Feng et al.,
1999; Huang and Tsai, 2000; Chen et al., 2001a, 2001b; Cheng et al.,
2002; Huang et al., 2002).
When rigorous limit equilibrium method is applied to analyze the
stability of three-dimensional slopes or landslides, six equilibrium
conditions should be satised, that is three direction force equilibrium conditions and moment equilibrium conditions around three
axes should be satised. Up to now, fruitful results satisfying all of
the six equilibrium conditions are few. For example, rigorous limit
equilibrium method for three-dimensional slopes or landslides was
investigated based on a non-column method (Zheng, 2007). The difference between rigorous and quasi-rigorous limit equilibrium stability analysis for three-dimensional slopes or landslides is compared
using the method, in which normal stresses on slip surface are
modied by a function with ve parameters (Zhu and Qian, 2007).
However, inter-column forces were not taken into account in their

22

X.P. Zhou, H. Cheng / Engineering Geology 160 (2013) 2133

Fig. 1. Plane map of the Jipazi landslide.

works (Zheng, 2007; Zhu and Qian, 2007). Actually, inter-column

forces signicantly affect the factor of safety. In this paper, rigorous
limit equilibrium method for three-dimensional slopes or landslides
is proposed based on the principle of nonnegative normal force over
slip surface, in which inter-column forces are considered. The relationship between the factor of safety and widths of sliding body can
be determined using LevenbergMarquardt least squares method.
The factor of safety of three-dimensional slopes or landslides can be
obtained using trust-region-reective iterative algorithm.
Characteristics of the present method are summarized as follows:
(a) Six equilibrium conditions are strictly satised. Sum of forces
along three axes are zero, and all the moments of the entire sliding
body around three axes are less than 1 10 6 N m. The present
method can be referred to as rigorous limit equilibrium method.

(b) The assumptions of the present method, in which inclinations

of inter-column forces are not constant, are different from the
Spencer method (Spencer, 1967) in which inclinations of intercolumn forces are constant. In this paper, it is assumed that inclinations of inter-column forces in row direction and column direction are equal to and , respectively. The value of and
can strictly be obtained by numerical calculation. The values of
and represent comprehensive and average effects of the sliding
body.
(c) Numerical characteristics. According to Duncan's (1996)
works, non-convergence points exist in almost all examples of rigorous limit equilibrium methods. In this paper, the non-linear relationship between the factor of safety and the width of sliding
body is obtained using LevenbergMarquardt least square method. Factor of safety in arbitrary intervals can be obtained by the

X.P. Zhou, H. Cheng / Engineering Geology 160 (2013) 2133

23

tting curves. Even in some narrow intervals in which the numerical results are not convergent, factor of safety can be accurately
determined by the tting equation. Therefore, there exist good
numerical characteristics in the present method.
(d) In the present method, arbitrary slip surface not limited to
ellipsoidal slip one can be taken into account.
2. Principle of the rigorous limit equilibrium method
2.1. Forces acting on the columns
The sliding body is divided into a number of columns with vertical
interfaces, as shown in Fig. 2. Every column is labeled by letters
i and j, which represent the number of row and column, respectively.
There are m number of columns in the x direction and n number of
columns in the y direction. All forces acting on the column are
schematically shown in Fig. 3. The weight of the column is denoted
as W i, j. The normal force and shear force acting on the slip surface
are denoted as N i, j and S i, j, respectively. The pore water pressure
acting on the slip surface is denoted as U i, j. The inter-column force
between the column (i, j) and the column (i, j1) is denoted as Q i, j,
and the inter-column force between the column (i, j) and the column
(i1, j) is denoted as G i, j.
2.2. Basic assumptions
The column methods are indeterminate, in which the number of
unknowns is much more than the number of equations. So it is necessary to make various assumptions regarding inter-column forces to
render the problem determinate.
(a) It is assumed that the base of a column is plane, and the point of
the normal force Ni,j, the shear force Si,j and the pore water pressure
Ui,j act on the center (I) of the plane. Moreover, the weight Wi,j acts
on the centroid (H) of the column, which is point to (I), as shown in
Fig. 3.
(b) Assume that the inter-column forces of Qi,j and Gi,j are parallel to
the yz plane and the xz plane, respectively. According to the geometrical features of slip surface, the inclinations of inter-column forces
of Qi,j are , as shown in Fig. 3 (the inclination of Qi,j is when
the angle between Qi,j and the direction of z-axis is more than 90;
the inclination of Qi,j is when the angle between Qi,j and the direction of z-axis is less than 90). Similarly, the inclinations of
inter-column forces of Gi,j are , as shown in Fig. 3 (the inclination
of G i,j is when the angle between Gi,j and the direction of z-axis is

Fig. 3. Schematic diagram of the forces acting on a column.

more than 90; the inclination of Gi,j is - when the angle between
Gi,j and the direction of z-axis is less than 90).
(c) Assume that the direction of movement or the possibility of a
change in the direction of sliding body is opposite to the direction
of x-axis.
(d) Assume that there are no supporting structures in the threedimensional slope, such as retaining wall.
(e) Assume that tensile strength of soil is zero.
2.3. Geometrical model
As shown in Fig. 2, the coordinate system o-xyz is established, where
x-axis is opposite to the direction of movement of the sliding body,
y-axis is parallel to the direction of the width of the sliding body and
z-axis is parallel to the direction of the height of the sliding body. The
entire sliding body is placed into the rst quadrant. The ground
surface and the slip surface are described by equations z1 = g(x, y)
and z2 = f(x, y), respectively. The direction cosines of normal forces
over the slip surface are denoted as (nxi, j,nyi, j,nzi, j), and the direction
cosines of shear forces over the slip surface are denoted as (lxi, j,lyi, j,lzi, j).
Since the angle between normal forces and z-axis is less than 90,
the value of nzi,j must be positive. Therefore, the direction cosines of
normal forces over the slip surface are described by



  1 f
1 f 1
i; j
i; j
i; j
nx ; ny ; nz
;
;
x
y

where
s
 2  2
f
f
:

1
x
y

Since x-axis is parallel to the direction of movement of the sliding

body, we have

Fig. 2. Sketch of three-dimensional slip surface.






1
f
i; j i; j i; j
1; 0;
lx ; ly ; lz

24

X.P. Zhou, H. Cheng / Engineering Geology 160 (2013) 2133

The weight of the column is

where
W

s
 2
f
:
1
x

i; j

i; j

A gx; yf x; y

where the average unit weight of sliding body is denoted as and the
cross sectional area of the column is denoted as A i,j.

3. Limiting equilibrium equations

3.1. Force equilibrium equations of columns
The equation of force equilibrium along the x-axis is

i;j

N U

i;j


i;j
i1;j
i;j
cos G cos 0:
nx S lx G

The equation of force equilibrium along the y-axis is

i; j

i; j


i; j
i; j1
i; j
ny S ly Q
cos Q cos 0:

The equation of force equilibrium along the z-axis is

i; j

i; j


i; j
i1; j
i; j
i; j1
i; j
i; j
sin G sin Q
sin Q sin W 0:
nz S lz G

3.2. Moment equilibrium equations of the entire sliding body

According to the assumption (d), there are no supporting structures in the sliding body. Therefore, the boundary conditions of inter-column
force can be described as G0; j

1jn1

0, Gm1; j 1jn1 0, Q i;0 1im1 0, and Q i;n1 1im1 0. Because of the relationship of action and reaction

among other inter-column forces, the moments of inter-column forces around the axes are equal to 0. The moment equilibrium equations can be
obtained as follows.
Moment around the x-axis is

m X
n  h


i i; j i; j h


i i; j i; j
X
i; j
i; j
y
i; j y
G =Q
i; j
i; j
z
i; j z
G =Q
i; j H
N U ni; j S li; j z
N U ni; j S li; j y
W yi; j 0:

i1 j1

Moment around y-axis is


m X
n h


i i; j i; j h


i i; j i; j
X
i; j
i; j
x
i; j x
G =Q
i; j
i; j
z
i; j z
G =Q
i; j H
N U ni; j S li; j z
N U ni; j S li; j x
W xi; j 0

10

i1 j1

Moment around z-axis is

m X
n  h


i i; j i; j h


i i; j i; j 
X
i; j
i; j
x
i; j x
G =Q
i; j
i; j
y
i; j y
G =Q
0
N U ni; j S li; j y
N U ni; j S li; j x

11

i1 j1

where
xG

i; j

=Q i; j

i; j

i; j

yG =Q
i; j
i; j
zG =Q
H
xi, j
yi,Hj

is x coordinate of the forces point of G i,j or Q i,j (G i,j is the different value of inter-column force on the plane ABBA and the plane
CDDC in Fig. 3. Q i,j is the different value of inter-column force on the plane BCCB and the plane ADDA in Figure 3), which is equal
to the x coordinate of point H;
is the y coordinate of the force point of G i,j or Q i,j, which is equal to the y coordinate of point H;
is the z coordinate of the force point of G i,j or Q i,j, which is equal to the z coordinate of point H;
is the x coordinate of point H;
is the y coordinate of point H.

The factor of safety is dened as that factor by which the shear strength components must be reduced to bring the soil mass into a state of limiting
equilibrium along a selected slip surface. Furthermore, it is assumed that the factor of safety of the cohesive component of shear strength and the
friction component of shear strength are equal, and the shear force over the slip surface is determined by MohrCoulomb criterion. One obtains

i; j


Ni; j U i; j tan cAi; j
Fs

where and c are the effective internal friction angle and cohesion of soil, and factor of safety is denoted as Fs.

12

X.P. Zhou, H. Cheng / Engineering Geology 160 (2013) 2133

25

The different value of inter-column force on the plane ABBA and the plane CDDC in Fig. 3 is denoted as G i,j, then the following expression
can be written as
i; j

i1; j

i; j

G :

13

Similarly, the different value of inter-column force on the plane BCCB and the plane ADDA in Fig. 3 is denoted as Q i,j, then the following
expression can be written as
Q

i; j

i; j1

i; j

Q :

14

Substituting Eqs. (13) and (14) into Eqs. (6), (7) and (8) yields

i; j

i; j


i; j
i; j
nx S lx G cos 0

15

i; j

i; j


i; j
i; j
ny S ly Q cos 0

16

i; j

i; j


i; j
i; j
i; j
i; j
nz S lz G sin Q sin W 0:

17

Substituting Eq. (12) into Eqs. (15), (16) and (17), and eliminating Q i,j and G i,j, the following expressions can be obtained as

i; j





cos sin cAi; j li;x j Fsni;x j U i; j li;x j U i; j tan cos sin cAi; j li;y j Fsni;y j U i; j li;y j U i; j tan












cos sin Fsni;x j li;x j tan cos sin Fsni;y j li;y j tan cos cos Fsni;z j li;z j tan



i; j i; j
i; j i; j
i; j i; j
i; j
cos cos cA lz Fsnz U lz U tan cos cos FsW












cos sin Fsni;x j li;x j tan cos sin Fsni;y j li;y j tan cos cos Fsni;z j li;z j tan





i; j
i; j
i; j
i; j
i; j
i; j
nx cos sin cA 2U tan ny sin cos cA 2U tan










S
cos sin Fsni;x j li;x j tan cos sin Fsni;y j li;y j tan cos cos Fsni;z j li;z j tan



ni;z j cos cos cAi; j 2U i; j tan cos cos W i; j tan









 :

cos sin Fsni;x j li;x j tan cos sin Fsni;y j li;y j tan cos cos Fsni;z j li;z j tan

18

i;j

19

Eqs. (18) and (19) can be rewritten as

N

i; j

Bi;x j Bi;y j Bi;z j Di;1 j

Ai;x j Ai;y j Ai;z j
i; j

i; j

i; j

i; j

20

i; j

C x C y C z D2
Ai;x j Ai;y j Ai;z j

21

where



i; j
i; j
i; j
Ax cos sin Fsnx lx tan

22


i; j
i; j
i; j
Ay cos sin Fsny ly tan

23


i; j
i; j
i; j
Az cos cos Fsnz lz tan

24


i; j
i; j i; j
i; j i; j
i; j i; j
Bx cos sin cA lx Fsnx U lx U tan

25


i; j
i; j i; j
i; j i; j
i; j i; j
By cos sin cA ly Fsny U ly U tan

26


i; j
i; j i; j
i; j i; j
i; j i; j
Bz cos cos cA lz Fsnz U lz U tan

27


i; j
i; j
i; j
i; j
C x nx cos sin cA 2U tan

28


i; j
i; j
i; j
i; j
C y ny sin cos cA 2U tan



i; j
i; j
i; j
i; j
C z nz cos cos cA 2U tan

29
30

26

X.P. Zhou, H. Cheng / Engineering Geology 160 (2013) 2133

i; j

i; j

i; j

i; j

D1 cos cos FsW

D2 cos cos W

31
tan :

32

Eqs. (22)(32) give signicant physical implications which are as follows:

(a) Axi, j, Ayi, j and Azi, j are inuential parameters of the directions of the inter-column forces and the geometrical features of slip surface on the
normal forces and the shear forces in the x, y and z directions, respectively.
(b) Bxi, j, Byi, j and Bzi, j are inuential parameters of the direction of shear forces and pore water pressure on the normal forces in the x, y and z
directions, respectively.
(c) Cxi, j, Cyi, j and Czi, j are inuential parameters of the direction of normal forces and pore water pressure on the shear forces in the x, y and z
directions, respectively.
(d) D1i, j and D2i, j are inuential parameters of the weight of columns on the normal forces and the shear forces, respectively.
On the boundary of sliding body, there may exist negative normal forces. Therefore, negative normal forces should be modied to satisfy the
assumption (e). It is shown from the study cases that the number of the columns on which normal forces are negative is few. The numerical
result will only be affected slightly if these columns are neglected (Yang and Zhen, 2002), which is within practical tolerance.
Substituting Eqs. (20) and (21) into Eqs. (9), (10) and (11) yields
( " i; j
!
#
m X
n
X
i; j
i; j
B1 Bi;2 j Bi;3 j Di;1 j
i; j
y
i; j y
G =Q
ni; j S li; j z

U
M x
i; j
i; j
i; j
A1 A2 A3
i1 j1
" i; j
!
#
)
i; j
i; j
i; j
B1 B2 Bi;3 j Di;1 j
i; j
z
i; j z
G =Q
i; j H

n
y
0

S
l
W
y
i; j
i; j
i; j
Ai;1 j Ai;2 j Ai;3 j

33

(" i; j
!
#
i; j
i; j
i; j
i; j
i; j
i; j
i; j
m X
n
X
B1 B2 B3 D1
C 1 C 2 C 3 D2 x Gi; j =Q i; j
i; j
x
M y
ni; j
U
li; j z
Ai;1 j Ai;2 j Ai;3 j
Ai;1 j Ai;2 j Ai;3 j
i1 j1
" i; j
!
#
)
i; j
i; j
i; j
i; j
i; j
i; j
i;j
B1 B2 B3 D1
C 1 C 2 C 3 D2 z
i; j
z
Gi; j =Q i; j
i; j H
n
x
0

W
x

i; j
i; j
i; j
Ai;1 j Ai;2 j Ai;3 j
Ai;1 j Ai;2 j Ai;3 j

34

( " i; j
!
#
m X
n
X
B1 Bi;2 j Bi;3 j Di;1 j
C i;1 j C i;2 j C i;3 j Di;2 j x
i; j
x
Gi; j =Q i; j
M z
ni; j

U
li; j y
i; j
i; j
i; j
i; j
i; j
i; j
A1 A2 A3
A1 A2 A3
i1 j1
" i; j
!
#
)
i; j
i; j
i; j
i; j
i; j
i; j
i; j
B1 B2 B3 D1
C 1 C 2 C 3 D2 y Gi; j =Q i; j
i; j
y
n
x
0:

i; j
i; j
Ai;1 j Ai;2 j Ai;3 j
Ai;1 j Ai;2 j Ai;3 j

35

There are three unknown parameters , and Fs in the above three equations. Therefore, Eqs. (33), (34) and (35) can be rewritten as a set of
non-linear equations:
8
< F 1 ; ; Fs 0
36
F ; ; Fs 0
: 2
F 3 ; ; Fs 0
where
F 1 ; ; Fs M x

37

F 2 ; ; Fs M y

38

F 3 ; ; Fs M z :

39

The set of non-linear Eq. (36) can be solved using trust-region-reective iterative algorithm. The initial value is set as = 0, = 0, and
Fs = 1. Then the local optimal solutions can be obtained by iterating about 1020 times. The solutions can make the set of non-linear
Eq. (36) satisfy
8
6
>
< F 1 ; ; Fs10 Nm
F 2 ; ; Fs106 Nm
>
:
F 3 ; ; Fs106 Nm:

40

Therefore, the solutions can be regarded as satisfying the overall moment equilibrium conditions around three coordinate axes. The factor of
safety for this three-dimensional slope is the value of Fs.

X.P. Zhou, H. Cheng / Engineering Geology 160 (2013) 2133

4. Computation of the factor of safety of three dimensional slopes

or landslide with the known ellipsoidal slip surface

Substituting Eqs. (44) and (45) into Eq. (42) yields

2

The coordinate o xyz is established as shown in Fig. 4.

Assume that the height of the slope is H. The length of projection of
the slope surface along the x-axis is l. Equations of the slope are described by

0 x l :

42

To simplify the computation, the entire sliding body is placed into

the rst quadrant. A coordinate system o-xyz is established by
transforming the coordinate system o xyz.
In the coordinate system o-xyz, y-axis coincides with y-axis, and
let y 0 = b. Therefore, it is sure that y coordinates of the entire sliding
body are positive.
In order that the slip surface is tangential to the xy plane and the
z-axis is through the point A, as shown in Fig. 4. The coordinate system is transformed as
x xa0
z zb0


1 f
1 f 1
;
;
x
y

48

Where

xx0
yy0
zz0

1:
a2
a2
b2

43

q
a0 x0 a2 z0

44

b0 z0 a:

45

s
 2  2
f
f

1
x
y

f
x a0 x0

z0 b q
2
x
2
a x a x 2 a yb2
0

b2

f
a2 yb
q :
y b2 a2 x a x 2 a2 yb2
0

b2

From Eqs. (3), (4) and (47), the direction cosines of shear forces
over the slip surface are obtained as

where

1
1 f
; 0;


49

Where

8
>
<

b0 ;
x ba0
H
b0 x a0 ; a0 x la0 :
>
l
:
x > la0
b0 H;

f
x a0 x0

z0 b q
2
x
a2 x a x 2 a yb2
0

b2

s
 2
f
:
1
x

Substituting Eqs. (44) and (45) into Eq. (41) yields

Assuming that the equation of the ellipsoidal slip surface is

47

4.2. The direction of normal forces and shear forces over ellipsoidal slip
surface

41

x > l

From Eqs. (1), (2) and (47), the direction cosines of normal forces
over the slip surface are obtained as

xb 0

x a0 x0
yb
z b0 z0

1:
a2
a2
b2

4.1. Coordinate system

8
>
< 0;
H
z2
x;
>l
:
H;

27

46
4.3. Assumption of inclinations of inter-column forces
According to assumption (b), the inclinations of inter-column
forces can be considered as follows:
(1) When y > b, the inclination of inter-column force Q i, j is denoted as ; when y b, the inclination of inter-column force Q i, j is
denoted as ;
(2) When x > x 0 a0 , the inclination of inter-column force G i, j is
denoted as ; when xx 0 a0 , the inclination of inter-column
force G i, j is denoted as .
4.4. Determination of the factor of safety
The factor of safety Fs can be obtained when trust-region-reective
iterative algorithm is applied to solve Eq. (36). Then, a set of Fs corresponding to the different values of b is obtained. Therefore, the following non-linear equation can be applied to t the relationship between Fs
and b as

Fig. 4. Prole of calculation model.

a3 ba4

Fsb a1 a2 e

50

28

X.P. Zhou, H. Cheng / Engineering Geology 160 (2013) 2133

z'

The searching zone

The coordinate of searching

point (x0', z0')
The major semi axis of
searching ellipsoid (a)

x'

Fig. 6. Cross-section of sliding body at y = b for Example 1.

Fig. 5. Schematic diagram of searching slip surface for a three-dimensional slope.

The values of a1, a2, a3 and a4 can be obtained using the Levenberg
Marquardt least square method.
Then, the factor of safety can be obtained by substituting b into
Eq. (50).
When the parameters of the slope and slip surface (a; x0 ; z0 ) remain unchanged, the minimum factor of safety Fsmin can be computed
as
h
i
a ba4
Fs min lim a1 a2 e 3
a1 :

51

It is not difcult to realize from Eqs. (50) and (51) that the minimum factor of safety Fsmin is the two-dimensional factor of safety
degenerated from the corresponding three-dimensional slope. The
two-dimensional factor of safety is smaller than the corresponding
three-dimensional factor of safety.
4.5. Determination of the factor of safety in divergence interval
Because of the complexity of Eq. (36), the solution may be not
convergent in some narrow intervals (Duncan, 1996). For example,
Eq. (36) is not convergent when b = b1. Therefore, the factor of safety
should be determined by substituting b = b1 into the tting Eq. (50).
In this way, stability of three-dimensional slope can be investigated in
arbitrary intervals.

Table 1
The partial results of searching slip surface for Example 1.
The four parameters
(x0, z0, a, b)

Number of iterations

/rad

/rad

Fs

(0, 40, 40,40)

(5, 50, 53,40)
(3, 42, 47,40)
(3, 43, 51,40)
(3, 40, 48,40)
(3, 45, 50,40)
(3, 40, 43,40)
(6, 46, 57,40)
(7, 40, 46,40)
(8, 46, 55,40)

6
7
9
9
14
9
7
14
9
9

0
0.0549
0.1161
0.1926
0.3402
0.1050
0.1535
0.1826
0.1578
0.1500

0.5493
0.5659
0.6620
0.7704
0.8969
0.6533
0.7156
0.7688
0.6974
0.7069

1.2475
1.4286
1.4221
1.5614
1.5804
1.4352
1.3557
1.7765
1.5122
1.7170

5. The automatic search of the most critical failure surface for

three dimensional slopes or landslide
For slope stability analysis, it is required to determine the slip surface and the corresponding factor of safety. In most of the commercial
programs, only the systematic pattern search for critical circular failure surface is available to the engineers. In this paper, determination
of the factor of safety is divided into two kinds of cases. One is the
slope with the unknown location of slip surface, such as soil slopes
with the unknown location of ellipsoidal slip surface. The other is
the slope with the known location of slip surface, such as landslides
with an arbitrary slip surface not limited to ellipsoidal one. For the
slope with the unknown location of slip surface, location of the ellipsoidal slip surface can be searched and the factor of safety can be determined using the present method. In addition, for the slope with the
known location of the arbitrary slip surface, the factor of safety can
also be computed using the present method.
For the slope with the unknown slip surface, the location of slip
surface will be determined using the present method. For threedimensional soil slopes, the slip surface is generally ellipsoidal. The
automatic search of location of ellipsoidal slip surface will be
discussed in this section. In order to locate ellipsoidal slip surface,

Table 2
The numerical results using three-dimensional rigorous limit equilibrium method.
b/m

Number of iterations

/rad

/rad

Fs

20
30
40
50
60
80
100
120
140
160
180
200
240
260
280
300
350

7
6
6
6
6
6
6
6
6
6
6
6
7
7
7
7
7

0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0

0.4926
0.5277
0.5493
0.5585
0.5660
0.5729
0.5765
0.5788
0.5806
0.5817
0.5826
0.5831
0.5834
0.5837
0.5839
0.5840
0.5837

1.4248
1.3005
1.2475
1.2223
1.2069
1.1914
1.1840
1.1797
1.1769
1.1751
1.1739
1.1729
1.1721
1.1717
1.1714
1.1711
1.1709

X.P. Zhou, H. Cheng / Engineering Geology 160 (2013) 2133

29

Fig. 7. Fitting curve of factor of safety for Example 1.

the following searching approach is proposed in this paper. According

to Eq. (42), four parameters (x 0 ; z 0 ; a; b) should be determined to locate an ellipsoidal slip surface because y 0 does not affect the location
of an ellipsoidal slip surface. Therefore, the location of ellipsoidal slip
surface can be regarded as an optimization problem, and the optimal
equations are

optimal equations. The searching process of location of ellipsoidal slip

surface of three-dimensional slopes is plotted in Fig. 5.
For slopes or landslides with the known arbitrary slip surface not
limited to the ellipsoidal one, the determination of the factor of safety
will be discussed in Section 6.3.

8
< F 1 x0 ; z0 ; a; b; ; ; Fs 0
Constrains : F 2 x0 ; z0 ; a; b; ; ; Fs 0
:
F 3 x0 ; z0 ; a; b; ; ; Fs 0

6. Examples
52

6.1. Example 1: The automatic search of the most critical failure surface
of three-dimensional slopes

Objective function : min : Fs:

53

The searching approach of slip surface presented in Section 5 is applied to analyze Example 1. The four unknown parameters (x0 ; z0 ; a; b)
are needed to determine the location of the slip surface. The following
computation parameters are used (Li et al., 2003): the strength parameters of soil are c = 30 kPa, = 30 and = 22 kN/m3, parameters of
ground surface (H, l) = (40, 40). The entire sliding body is divided into
a number of columns with vertical interfaces. There are 50 columns in
the x direction and 50 columns in the y direction.
According to the optimal Eqs. (52) and (53), the partial numerical
results are listed in Table 1. It is shown from Table 1 that the optimal
parameters (x0 ; z0 ; a) are equal to (0, 40, 40) and the corresponding

The critical slip surface, whose safety factor is minimum among all
the available ones, can be determined by the objective function. As
illustrated in Section 4.4, the values of factor of safety decrease with
increasing b. When b , that is the two-dimensional slope, the
corresponding factor of safety is minimum. The parameters (x0 ; z0 ; a)
can be obtained from the optimal equations when b . For threedimensional slopes, the inuences of b on the parameters ( x0 ; z0 ; a)
and the factor of safety should be studied. In order to solve the objective
function, the value of b is assigned to be equal to the certain dened
value, and the numerical computation is performed, then the corresponding minimum factor of safety can be obtained from the optimal
equations as well as the corresponding parameters (x0 ; z0 ; a). For the
different value of b, the minimum factor of safety and the parameters
(x0 ; z0 ; a) of three-dimensional slopes can be determined. Therefore, effect of b on the parameters (x0 ; z0 ; a) and the corresponding minimum
factor of safety of three-dimensional slopes can be obtained from the

z'

Table 3
Comparison between the present results and the previous results for example 1.
b/m

General
column
method

3D Bishop
simplied
method

3D
Janbu's
method

3D
Spencer's
method

Li's
method

The
present
method

40
80
120
240
320

1.15
1.14
1.14
1.15
1.15

1.30
1.22
1.20
1.19
1.19

1.25
1.17
1.16
1.14
1.14

1.34
1.24
1.22
1.21
1.21

1.48
1.34
1.29
1.26
1.26

1.2534
1.1875
1.1763
1.1740
1.1740

x'

Fig. 8. Cross-sectional of sliding body for Example 2.

30

X.P. Zhou, H. Cheng / Engineering Geology 160 (2013) 2133

Table 4
Results of the slope stability using three dimensional rigorous limit equilibrium
method.
b/m

Number of iterations

/rad

/rad

Fs

50
60
70
75
80
90
100
120
140
160

26
26
11
14
12
17
14
17
15
20

0.9213
1.0106
1.0798
1.1087
1.1346
1.1789
1.2153
1.2714
1.3125
1.3439

1.0467
1.0523
1.0548
1.0557
1.0565
1.0576
1.0585
1.0596
1.0602
1.0606

2.1464
2.1301
2.1188
2.1147
2.1113
2.1062
2.1025
2.0976
2.0947
2.0928

According to Eq. (36), effects of b on the factor of safety of the

three-dimensional slope are listed in Table 2.
According to Table 2, the values of a1, a2, a3 and a4 in Eq. (50) are
determined using LevenbergMarquardt least square method, one
obtains
Fsb 1:174 0:4503e

0:04434b0:8722

The tting curve of factor of safety for Example 1 is depicted in

Fig. 7, and the tting results are listed in Table 3.
When the parameters of the ground and slip surface (x0 ; z0 ; a) remain unchanged, the corresponding two-dimensional factor of safety
Fsmin is
h
i
0:04434b0:8722
Fs min lim 1:174 0:4503e
1:174:

Table 5
Comparison between the present results and the previous ones for Example 2.

56

57

Factor of safety Zhang's result Chen's result Zheng's result The present result
Fs

2.122

2.262

2.140

2.092

factor of safety for the three-dimensional slope is minimum when

b = 40 m.
Equation of the ellipsoidal slip surface is obtained using the present method as:
x2 y402 z402

1:
402
402
402

54

According to Eq. (54), Example 1 becomes a homogeneous, cohesive

three-dimensional slope with an ellipsoidal slip surface. Equations of
ground and slip are depicted in Fig. 6. The corresponding factor of safety
is 1.2534.
In addition, the present method can be applied to study effects of b
on the factor of safety of three-dimensional slopes. When slip surface
(x0 ; z0 ; a) remains unchanged, equation of the ellipsoidal slip surface
can be written as:
x2 yb2 z402

1:
402
402
b2

55

In order to compare between the present results and previous

ones, the factors of safety of the three-dimensional slope are respectively calculated using different methods including General column
method, 3D Bishop simplied method, 3D Janbu's method, 3D
Spencer's method and Li's method (Li et al., 2003.). It is observed
from Table 3 that the present results are more conservative than the
previous ones.
6.2. Example 2: The determination of the factor of safety for
three-dimensional slopes with an ellipsoidal slip surface
Example 2 is a homogeneous, cohesive three-dimensional slope
with an ellipsoidal slip surface. As shown in Fig. 8, equations of
ground and slip surface have already known. The slip surface is ellipsoidal. The following computation parameters are used (Zhang, 1988;
Chen et al., 2001a, 2001b): the strength parameters of soil are: c =
29 kPa, = 20 and = 18.8 KN/m 3, parameters of ground surface
(H, l) = (12.2, 24.4), parameters of slip surface (a, b, x0 , z0) =
(24.4, 78.35, 5.11, 19.16).
As shown in Table 4, the stability of this three-dimensional slope
had been studied by Zhang (1988), Chen et al. (2001a, 2001b), and
Zheng (2007) and factors of safety are 2.122, 2.262 and 2.140,
respectively.

Fig. 9. Fitting curve of factor of safety for Example 2.

X.P. Zhou, H. Cheng / Engineering Geology 160 (2013) 2133

31

Fig. 10. The photograph of Xiangjiashan landslides.

The present method is applied to study the stability of this slope

for verifying the robustness and precision. The entire sliding body is
divided into a number of columns with vertical interfaces. There are
50 columns in the x direction and 50 columns in the y direction. The
results are listed in Table 5.
Eq. (50) can be determined using LevenbergMarquardt least
square method, the following expression can be obtained as
1:231b0:7146

Fsb 2:092 0:004794e

58

The tting curve of factor of safety for Example 2 is plotted in

Fig. 9.
Then, substituting b = 78.35 into Eq. (58), the factor of safety is
Fs = 2.0920.

Lower sliding mass

When parameters of the ground and slip surface (a; x 0 ; z 0 ) remain

unchanged, the corresponding two-dimensional factor of safety Fsmin
is
h
i
1:231b0:7146
2:092:
Fs min lim 2:092 0:004794e
b

59

Comparisons between the present results and the previous ones

for Example 2 are listed in Table 5.
In Table 5, Zhang's result and Chen's result were obtained using
quasi-rigorous limit equilibrium stability analysis for the threedimensional slope. Zheng's result was obtained using rigorous limit
equilibrium stability analysis for the three-dimensional slope, in
which inter-column forces were not taken into account. It is easily

Cross-section 3-3'
Upper sliding mass

Fig. 11. Plan view of Xiangjiashan landslides.

32

X.P. Zhou, H. Cheng / Engineering Geology 160 (2013) 2133

97

56

(m)
465
460
455
450
445
440
435

The slip surface of the

lower sliding mass

430
425
420

c o l+ d l

Q4

The slip surface of the

upper sliding mass

415
410
405
400

Z- 1 3 - 1 0
406. 00

J 1z

J 1z

T3x j

8. 60

395

del

Q4
10

390

T 8
17. 70

385
379. 60

380

del

Q4

375

11. 00

J 1z
33. 70

370

368. 00
12. 02( 367. 58)
13. 80( 365. 80)
14. 00( 365. 60)

365

J 1z
40. 80

359. 60

360

del

Q4

355

11. 29( 356. 71)

12. 80( 355. 20)

22. 80( 356. 80)

6. 58( 353. 02)

7. 80( 351. 80)

350
20. 80( 347. 20)

345
340

32. 80( 346. 80)

14. 80( 344. 80)

24. 80( 343. 20)
c o l+ d l
4

20. 30( 339. 30)

J 1z

39. 80( 339. 80)

29. 80( 338. 20)

335
330
32. 80( 326. 80)

Fig. 12. The 33 geological cross-section of Xiangjiashan landslides.

found from Table 5 that the values of factor of safety from the present
method are more conservative than those from other methods.
6.3. Example 3: A study case-determination of the factor of safety for
Xiangjiashan landslide with the known arbitrary slip surface
In order to analyze the stability of three-dimensional slopes with
the known arbitrary slip surface, a practical landslide-Xiangjiashan
landslide is taken into account, as shown in Fig. 10.
The Xiangjiashan landside is located on one side of Chongqing
Guizhou Expressway at the Nanping district in Chongqing. The landslide occurred on March 1998 following rainfalls and articial disturbance which appear during the period of roadway excavation for
ChongqingGuizhou Expressway. It is observed from Fig. 11 that the
strike of the landslide is in northsouth direction and the plan view
of the landslide is an irregular shape of horseshoe. Moreover, the
width and longitudinal length of the landslide are about 200360 m
and 230 m, respectively. The slope angle of the landslide is about
7080. The area of the sliding region is approximately 70,000 m 2,
and the volume of the sliding body is estimated to be 1,400,000 m 3.
Therefore, Xiangjiashan landslide is in the framework of large-scale
landslides.
According to the geological exploration for Xiangjiashan landslide,
there exist two sliding bodies (upper sliding body and lower sliding
body), which can be observed in Figs. 11 and 12. The upper sliding
body contains sandstone and weathered mudstone. The lower sliding
body also contains sandstone and weathered mudstone. Feng et al.

(Feng et al., 2006; Yin et al., 2007) had ever done a large number of
works on Xiangjiahsan landslide. According to Feng's works, shear
strength parameters of rock-soil are listed in Table 6. Moreover, the
factors of safety for Xiangjiashan landslide were calculated by Feng
et al. (2006) based on the unbalanced thrust transmission method.
The factors of safety for upper and lower sliding body from the unbalanced thrust transmission method are 1.11 and 1.65, respectively
(Feng et al., 2006).
Before the slope stability of Xiangjiashan landslide is analyzed
using the present method, the equations of the ground surface and
slip surface of Xiangjiashan landslide must rstly be obtained by
tting the geometric datum of the ground surface and slip surface,
which are used to determine the direction cosines of normal forces
over the slip surface. The upper and lower sliding masses are divided
into a number of columns with vertical interfaces. The upper sliding
body is divided into 50 columns in the x direction and 50 columns
in the y direction. The lower sliding body is divided into 150 columns
in the x direction and 150 columns in the y direction. The factors of
safety for the upper and lower sliding bodies from the present method are 0.9738 and 1.4571, respectively. Moreover, 10 iterations are
needed to arrive at convergence.
The factors of safety for Xiangjiashan landslide from the present
method and the unbalanced thrust transmission method are compared in Table 7. It is found from Table 7 that the values of the factor
of safety from present methods are smaller than those from the unbalanced thrust transmission method. The differences of the factors
of safety for the upper and lower sliding bodies obtained by the
above two methods are 12.27% and 11.69%, respectively.

Table 6
Calculation parameters of the Xiangjiashan landslide.
Masses

Unit weight
/KN/m3

Effective cohesion of
geomaterials c/KPa

Effective internal friction

angle of geomaterials /

The upper sliding

mass
The lower sliding
mass

24.92

55.0

26.0

24.92

60.0

25.0

Table 7
Comparison of factors of safety with different calculation methods.
Landslide masses

The previous results from unbalanced

thrust transmission method

The present
results

The upper sliding mass

The lower sliding mass

1.11
1.65

0.9738
1.4571

X.P. Zhou, H. Cheng / Engineering Geology 160 (2013) 2133

7. Conclusions
The previous limit equilibrium methods applied to analyze the
stability of three-dimensional slopes or landslides only satised
three directions force equilibrium, and one or two directions moment
equilibrium, which is called quasi-rigorous limit equilibrium method.
In this paper, the rigorous limit equilibrium method is proposed,
which satises both three directions force equilibrium conditions in
the three coordinate axes and three directions moment equilibrium
conditions around three coordinate axes. According to the geometrical features of slip surface, the inclinations of inter-column forces of
row direction and column direction are equal to and , respectively. On the basis of three force equilibrium equations and denition of factor of safety, normal force and shear force acting on the
slip surface can be expressed by , and factor of safety. Therefore,
a set of nonlinear equations is obtained by substituting normal force
and shear force into the three moment equilibrium equations. The relationship between the width of sliding body and factor of safety can
be determined by using trust-region-reective iterative algorithm.
The value of the factor of safety can be obtained using Levenberg
Marquardt least square method as well as values of and . Comparing with the previous methods, which only considered four or ve
equilibrium conditions, the present method is more accurate and rigorous. Moreover, the present method can go for an automatic search
of the most critical failure surface and determine the factor of safety
for three-dimensional slopes or landslides with a known slip surface
of arbitrary shapes.
Acknowledgment
This work was supported by the National Natural Science Foundation of China (Nos. 51078371 and 51279218).
References
Baligh, M.M., Azzouz, A.S., 1975. End effects on stability of cohesive slopes. Journal of
Geotechnical Engineering 101 (11), 11051117.
Chen, Z.Y., Mi, H.L., Wang, X.G., 2001a. A three-dimensional limit equilibrium method
for slope stability analysis. Chinese Journal of Geotechnical Engineering 23 (5),
525529.

33

Chen, Z.Y., Wang, X.G., Habereld, C., et al., 2001b. A three-dimensional slope stability
analysis method using the upper bound theorem part I: theory and methods. International Journal of Rock Mechanics and Mining Sciences 38 (3), 369378.
Cheng, Y.M., Zhu, L.J., Gu, Z.Q., 2002. Rotational failure of 3D non-symmetric slope predicted by Bishop's method. Chinese Journal of Geotechnical Engineering 24 (6),
706709.
Cruden, D.M., 1991. A simple denition of a landslide. Bulletin of the International Association of Engineering Geology 43, 2729.
Duncan, J.M., 1996. State of the art: limit equilibrium and nite element analysis of
slopes. Journal of Geotechnical Engineering 122 (7), 577596.
Feng, S.R., Feng, D.X., Ge, X.R., et al., 1999. 3D limit equilibrium method for slope stability and its application. Chinese Journal of Geotechnical Engineering 21 (6),
657661.
Feng, J.F., Yin, G.Z., Dai, G.F., et al., 2006. Stability analysis and comprehensive treatment for Xiangjiashan landslide. Soil Engineering and Foundation 20 (4), 1216.
Hovland, H.J., 1979. Three-dimensional slope stability analysis method. Journal of
Geotechnical Engineering 105 (5), 693695.
Huang, C.C., Tsai, C.C., 2000. New method for 3D and asymmetrical slope stability analysis. Journal of Geotechnical and Geoenvironmental Engineering 126 (10),
917927.
Huang, C.C., Tsai, C.C., Chen, Y.H., 2002. Generalized method for three-dimensional
slope stability analysis. Journal of Geotechnical and Geoenvironmental Engineering
128 (10), 836848.
Hungr, O., 1987. An extension of Bishop's simplied method of slope stability analysis
to three dimension. Geotechnique 37 (1), 113117.
Hungr, O., Salgado, F.M., Byrne, P.M., 1989. Evaluation of a three-dimensional method
of slope-stability analysis. Canadian Geotechnical Journal 26, 679686.
Lam, L., Fredlund, D.G., 1993. A general limit equilibrium model for three-dimensional
slope stability analysis. Canadian Geotechnical Journal 30, 905919.
Li, T.L., Wang, Y.X., Deng, H.K., 2003. An improved method for three-dimensional slope
stability analysis. Chinese Journal of Geotechnical Engineering 25 (5), 611614.
Spencer, E., 1967. A method of analysis of the stability of embankments assuming parallel inter-slice force. Geotechnique 17 (1), 1126.
Yang, M.C., Zhen, Y.R., 2002. Local minimum factor-of-safety method based on limit
equilibrium theory. Chinese Journal of Geotechnical Engineering 24 (5), 600604.
Yin, G.Z., Feng, J.F., Liu, L.M., et al., 2007. Study on the mechanism and stability of
Xiangjiashan landslide. Chinese Journal o f Underground Space and Engineering
3 (1), 167171.
Zhang, X., 1988. Three-dimensional stability analysis of concave slopes in plan view.
Journal of Geotechnical Engineering 114 (6), 658671.
Zhang, J.F., Wang, S.Y., Qi, T., 2005. Three-dimensional Spencer method for slope stability analysis. Chinese Journal of Rock Mechanics and Engineering 24 (19),
34343439.
Zheng, H., 2007. A rigorous three-dimensional limit equilibrium method. Chinese Journal of Rock Mechanics and Engineering 26 (8), 15291537.
Zhu, D.Y., Qian, Q.H., 2007. Rigorous and quasi-rigorous limit equilibrium solutions of
3D slope stability and application to engineering. Chinese Journal of Rock Mechanics and Engineering 26 (8), 15131528.