Available online at www.sciencedirect.

com

Procedia Engineering 10 (2011) 1402–1407

ICM11

Reliability analysis of slope Stability using stochastic finite

element method
Khaled Faraha, *,Mounir Ltifi a,Hedi Hassisa
a
Civil Engineering Laboratory. National Engineering school of Tunis, Tunisia BP 37 Le Belvédère 1002 Tunis
Tunisia

Abstract

This paper deals with the reliability analysis of slope stability with spatial variability of soil properties. Supposing an
elastic soil behavior until yield criteria, a stresses mobilized along slip surface are evaluated to formulate the
performance function. The results of SFEM (Stochastic Finite Element Method) and the limit equilibrium method,
such as Bishops simplified method are analyzed to check their efficiencies and accuracies. To look for critical slip
surface, an optimization strategy is performed, also a sensitivity studies are carried out in order to investigate the
influence of random field parameters used to model the spatial variability of soil properties.
© 2011 Published by Elsevier Ltd. Open access under CC BY-NC-ND license.
Selection and peer-review under responsibility of ICM11

Keywords: Slope stability; stochastic finite element method; limit equilibrium method

1. Introduction

The uncertainty in the calculation of slope stability results from a number of sources in particular; the
geotechnical properties of soils have several components of randomness. The values of parameter at any point are
uncertainly due to the spatial variability of soil; in addition they are estimated using a finite number of trials. The
analytical models adopted for slope stability problems are simplifications of more complex mathematic problems.
Recently, many studies of probabilistic slope stability have been presented, which takes into account the
uncertainties and randomness in soil properties. (Yang[1], SU Yong-huo and al[2];Tan[3]; Suchomel [4];Tan and
Wang[5]).
The objective of a deterministic slope stability analysis is to estimate the minimum safety factor which
corresponds to a critical slip surface. For simplicity we generally opt for computing a circular surface. However, the
slip surface can have a more complicated form for a slope with a heterogeneous soil l.
The search procedure of the slip surface uses an algorithm with a limited finite number of slip surfaces.

* Corresponding author. Tel.:+216 98 282 437

E-mail address: [email protected].

1877-7058 © 2011 Published by Elsevier Ltd. Open access under CC BY-NC-ND license.
Selection and peer-review under responsibility of ICM11 doi:10.1016/j.proeng.2011.04.233
 Khaled Farah et al. / Procedia Engineering 10 (2011) 1402–1407 1403

In practice, to find a circular critical slip surface we precede as follows, from a given centre, we give a different
radius and are looking for the one who gives the minimum safety factor, repeat the same operation for other centres
according to a predetermined grid. Since the limit equilibrium methods use different approximations, the location of
critical slip surface may change slightly. With the evolution of computation tools , the optimization techniques for
finding the critical slip surface were developed as dynamic programming (Baker[6]), Unconstrained minimization
technique sequential (Basudhar [7]), simplex reflection technique (Nguyen[8]) and more recently, Monte Carlo
technique
In all the algorithms listed above, the identification of deterministic surface associated with a minimum safety
factor is formulated as an optimization problem. By optimizing the various geometric variables of the slip surface,
the critical slip surface can be obtained. The determination of the probabilistic critical surface may be driving by the
same way, with the consideration of the random character of soil properties.
In this work the probabilistic approach proposed considers the spatial variability of soil mechanical parameters
and the Yong's modulus. By applying the stochastic finite element method SSFEM, we determine the stresses along
the slip surface necessary for the formulation of limit state function. The mechanical model of Bishop and Spenser
were adopted for evaluating the reliability index and the determination of the critical probabilistic slip surface. An
optimization procedure for finding critical probabilistic surface is proposed and is evaluated by numerical examples.
It uses the technique developed by (Celestino and Duncan [9]), which is an research of the slip surface associated
with a minimum safety factor. This procedure is modified to evaluate a maximum probability of ruin. A comparison
between the different approaches used to this reliability analysis is presented. The probabilistic calculation was
performed using the program FERUM which was developed by (Haukaas and Der Kiuregian [10]) and with a direct
coupling through a programation of the proposed algorithm written in Matlab to analyze the stability of a slope

2. Probabilistic approaches

The probabilistic approach allowed measuring the performance of a slope stability failure probability or
reliability index. It considers that soil properties are modelled by variables or random fields. By choosing as failure
criterion the safety margin, the performance function is defined by the following expression:

(1)

The safety factor can be evaluated by limit equilibrium methods or by applying the finite element method is a
function of random variable

2.1. Limit equilibrium method

The limit equilibrium method uses different assumptions and differing in the number of the equations satisfying
equilibrium, among used in this work: The simplified Bishop method was developed by Bishop. This procedure uses
the method of slices to find the factor of safety for the soil mass. It is based on the assumptions that the forces on the
sides of the slice are horizontal and the failure occurred by rotation of soil mass on a circular slip surface. Also, the
normal force is assumed to act at the centre of the base of slice. By using the Mohr Coulomb criteria and the
definition of the factor of safety. The Spencer‘s Method based on the assumption that the interslice forces are
parallel. It presented originally for the analysis of circular slip surfaces. By considering overall force equilibrium
and overall moment, two values of factor of safety respectively Ff and Fm are obtained
Note that the term of Fs is present in both sides of equilibrium equation. Hence, the algorithm proposed by (SU
Yong –hua and al [2]) based on checking point and iteration method is used to evaluate the probability of failure.
The expression of the factor of safety obtained by Bishop‘s simplified method and Spencer‘s Method is provided
in the literature.

2.2. Finite element method

The finite element method provides a distribution of stresses and displacements. It represents available tools that
satisfy all condition necessary for the evaluation of the solution of slope stability problem. Zheng [11]have used this
procedure.
1404 Khaled Farah et al. / Procedia Engineering 10 (2011) 1402–1407

The mean factor of safety is defined as a weighted average of shear strength over shear stress

(2)

Where:

n:the number of elements on a slip surface, c: cohesion, is the friction angle, is the normal stress and is the
shear stress, is length of the slip surface of the ith element. For an inclined at an angle to the horizontal, the
stresses are calculated from stress components ( by the following equations:

(3)

(4)

In the case of the stochastic finite element method SSFEM (Ganem and Spanos[12]based on a decomposition of
random fields to an order M using a discretization of Karhunem-Loeve and another decomposition of vector nodal
displacement solution of the problem based on polynomial chaos. This vector is truncated by a P term. P is given by
the equation:

(5)

The stresses are obtained from stochastic nodal displacement.

In considering the truncation of the decomposition of the Karhunem-Loève and the nodal displacement vector, the
convergence of the SSFEM method depends of the orders of expansion M and p
To determine the probability of failure several approaches can be adopted as First Order relibility Method FORM,
Second Order Reliability Methods SORM and Monte Carlo.
The method MVFOSM (mean value first order second moment reliability index) is a convenient to estimate the
reliability index. The reliability index is given by:

(6)

and are respectively the gradient of the performance function and covariance matrix

The probability of failure is estimated by the following equation:

(7)

Where is standard normal cumulative function.

3. Optimization procedure for evaluating probability of failure

The Optimization of the research algorithm of critical surface associated with the lowest index of reliability (or
maximum probability of failure) can be determined by the following steps:

a. From the displacement and contour of stress, we can identify an initial critical surface of failure. A deterministic
program of slope stability analysis for the evaluation of the safety factor is modified in order to determine the
reliability index or probability of failure.
 Khaled Farah et al. / Procedia Engineering 10 (2011) 1402–1407 1405

b. Determination of a critical circular surface by using the grid method. It represents an initial slip surface for the
search for a non-circular slip surface in reliability context. The technique developed by (Celestino and Duncan, [9])
to looking for critical slip surface is embedded in an algorithm for calculating the probability of failure or reliability
index

4. Evaluation of the proposed procedure

In order to verify the effectiveness of the procedure adopted and to find the effect of the variability of soil properties
considered as random, numerical examples and a comparative study are presented.
Consider a homogeneous slope, the Young's modulus, cohesion and friction angle are modelled by a Gaussian
random fields. The parameters values are listed in Table 1
Table.1 Input soil properties
Soil Properties c (kPa) (KN/m3) E( MPa)

Mean ( ) 7 20 19 20 0.3
Coefficient of variation
(covX) 0.10 0.10 - 0.10 -
Correlation length (m) [lx,ly] [60,6] [60,6] - [60,6] -

c: Cohesion. : Friction angle. : Unit weight. E: Young's modulus. : Poisson‘s ratio

For this slope stability analysis using SSFEM method, the finite element model considered a mesh composed of
1081nœud. The elements used are linear quadrilaterals QUAD4. The boundary conditions are: the bottom of the slop
is fixed also the horizontal displacement on the left and the right sides are fixed.

20

15
Height(m)

10

0
0 10 20 30 40 50 60
Length (m)

Fig.1. Meshing of the slope

In this example, we consider that the random fields have an exponential autocorrelation function. The soil behaviour
is assumed linear. The shear stress is calculated using the criterion of Mohr-Coulomb. The autocorrelation function is
given by the following expression:

(8)

Where, , :is the correlations lengths of a 2D domain, and the location of the soil mass

4.1. Determination of the reliability index and the critical probabilistic surface
1406 Khaled Farah et al. / Procedia Engineering 10 (2011) 1402–1407

The results of calculation the reliability index and safety factors obtained by different approaches are listed in
Table.1

Table1: reliability indexes and safety factors for both probabilistic and deterministic slip surfaces evaluated by the different methods

Probabilistic slip surface Deterministic slip surface

SSFEM Bishop Spenser SSFEM Bishop Spenser

(p=2, M=4) (p=3, M=4) (p=2, M=4) (p=3, M=4)

ȕ 3.783 3.785 4.312 4.489 2.759 2.758 2.768 2.771

F 1.767 1.765 1.658 1.651 1.871 1.869 1.874 1.881

Where ȕ the reliability index and F the factor of safety

The results of calculating the reliability index and safety factors obtained by different approaches are listed in Table
1.
From table1, It can be seen that the values obtained by the Bishop‘s and Spenser‘s model are comparable whereas
compared to the values obtained by SSFEM, a remarkable difference is observed
20

18

16
Bishop 's simplied method
14
Spenser 's method
12 SSFEM method
Height(m)

10

0
0 10 20 30 40 50 60
Length (m)

Fig1: Critical probabilistic slip surface obtained by the different methods

The corresponding critical probabilistic slip surface evaluated by the different approaches are presented in
Fig.1.The small difference between the values of the minimum reliability index presented in table1 is confirmed by
the positions of the critical probabilistic slip surface. This result demonstrate the effectiveness of the developed
method

4.2. Effect of the input random field

4.2. 1.Effect of the coefficient of variation

The minimum reliability indexes are evaluated by the proposed approaches for different coefficient of variation of
the random fields. The results are reported in Table 2

Table2: The minimum reliability indexes for different coefficient of variation and correlation length=[60,6]
0.1 0.20 0.25 0.3

SSFEM(p=3, M=4) 3.848 3.756 3.679 3.586

Bishop 4.227 4.205 4.121 4.023

Spenser 4.223 4.167 4.107 4.015

4.2. 2.Effect of the correlation length

 Khaled Farah et al. / Procedia Engineering 10 (2011) 1402–1407 1407

The reliability indexes are evaluated by the proposed approaches for different correlation length of the random
fields. The results are reported in Table 3

Table3: The minimum reliability indexes for different correlation length and coefficient of variation=0.1
Correlation length
[120,12] [240,24] [480,48] [960,96]

SSFEM (p=3, M=4) 3.945 3.845 3.764. 3.734

Bishop 4.345. 4.231 4.125 3.946

Spenser 4.325 3.221 4.113 3.926

From Table2 and, Table3 we can find that the results of the three methods are similar and the reliability index
decrease when the coefficient of variation or the correlation length increase. These results demonstrate the accuracy
and feasibility of the adopted procedure

5. Conclusions

In this work, a solution of the procedure for the analysis of slope stability problem is presented, in which it was
integrated an optimization technique to identify a critical probabilistic slip surface. We considered the spatial
variability of soil strength parameters and Young's modulus necessary for the determination of stress field mobilized
along the slip surface. These soil properties are modelled using the random field's theory. In this reliability study by
the stochastic finite element, soil mass is assumed elastic until the verification of Mohr-Coulomb criterion. A
comparison between the values of the reliability index and the position of critical surface obtained by the
probabilistic methods using limit equilibrium: Spenser and Bishop and SSFEM are presented. The results of the
proposed approaches showed a small difference between the values of reliability index and the position of critical
probabilistic slip surface.
The procedure established to search the probabilistic slip surface can be generalized to study the stability of a
heterogeneous slope and taking into account the effects of water and loading.

References

[1] Yang L. THE3-D Stochastic analysis of progressive failure of rock slopes. Geotechnical engineering for disaster mitigation and
rehabilitation.; 2008.
[2]SU Yong-Hua,ZHAO Ming-hua,ZHANG Yue-ying.Numerical method of slope failure probability based Bishop model.J.Cent.South Univ.
Technol;2008,15:100-105.
[3] Tan XH. Nonlinear Finite Element Reliability Analysis of slope stability. PhD Thesis, Hefei University of Technology, Hefei, China (in
Chinese); 2007.
[4] Suchomel R. and Masin D.Comparison of different probabilistic methods for predicting stability of a slope in spatially variable
Computers and Geotechnics; 2009.
[5] Tan XH. and Wang J. Finite element reliability analysis of slope stability .J Zhejiang Uinv Sci A; 2009, 10(5): 645-652.
[6] Baker R. Determination of the critical slip surface in slope stability computation. International journal for Numerical and Analytical
Methods in Geomechanics ;1980,04:333-59.
[7] Basudhar PK. Some application of mathematical programming techniques to stability problems in geotechnical engineering. PhD thesis,
Indian Institute of Technology Kanpur,1976, Kanpur ,India.
[8] Nguyen VU. Determination of critical slope failure surface .Journal of geotechnical engineering ASCE ;1987,111:238-50.
[9] Celestino TB and Duncan JM. Simplified search for non-circular slip surface. Proceedings of the 10 th international conference on soil
mechanics and foundation engineering; 1981, 391-394.
[10] Haukaas T.and Der Kiureghian A. A computer program for nonlinear finite element analysis. Proceedings of the eighth international
conference on structural safety and reliability, Newport beach,CA; 2001.
[11] Zheng H, Sun G, Liu D. A practical procedure for searching critical slip surfaces of slopes based on the strength reduction technique.
Computers and Geotechnics;2009. 36:1-5
[12] Ganem, RG and Spanos, PD. Stochastic finite elements. A Spectral approach. Springer Verlag; 1991.