See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/228995264

A Review of Current Methods for Slope Stability Evaluation

Article  in  Electronic Journal of Geotechnical Engineering · January 2011

CITATIONS READS
7 4,932

2 authors, including:

Behzad Kalantari
University of Hormozgan - Bandar Abbas
48 PUBLICATIONS   231 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Undrained behavior of clay-sand mixture with a large number of cycles View project

Ground improvement techniques View project

All content following this page was uploaded by Behzad Kalantari on 03 September 2017.

The user has requested enhancement of the downloaded file.

 A Review of Current Methods for Slope
Stability Evaluation

Amin Pourkhosravani
Department of Civil Engineering,
University of Hormozgan Iran
e-mail: aminpourkhosravani@yahoo.com

Behzad Kalantari
Assistance Professor, Department of Civil Engineering,
University of Hormozgan Iran
e-mail: behzad996@yahoo.com

ABSTRACT
Importance of slope stability problem is the reason why so many alternate methods have been
developed for evaluating the safety of a slope. The basic approach is determining a factor of
safety against failure by sliding for a given slope.
These methods differ in simplifications and accuracy. In this paper, authors study and discuss
current methods for slope stability analysis in detail. These methods are grouped under the
categories: Limit Equilibrium methods, Numerical analysis methods, Artificial Neural
Networks, and Limit Analysis method.
KEYWORDS: Slope stability, limit equilibrium, numerical methods, artificial neural
network, limit analysis.

INTRODUCTION
Slope stability is one of the most important and delicate problems in civil engineering,
particularly encountered in large and important projects such as dams, highways and tunnels.
Many techniques exist for evaluation of the stability of a given slope.

Earlier methods for slope stability analysis were generally based on hand-performed, and
therefore simplified computations. With more and more powerful computers becoming
commonly available, engineers have developed complicated but more accurate methods.

- 1245 -
 Vol. 16 [2011], Bund. L 1246

The most basic purpose of slope stability analysis is determining a factor of safety against a
potential failure, or landslide. If this factor of safety is determined to be large enough, the slope is
judged to be stable (safe). If it is 1.0 or less it is unsafe.
In this paper, we study and discuss all currently available methods for slope stability evaluation.

LIMIT EQUILIBRIUM METHODS

The approach that has been developed the earliest is generally termed the limit equilibrium
methods “LE” methods. These are based on determining applied stresses and mobilized strength
over an trial slide surface in the soil slope, then a factor of safety is determined by considering
these two quantities. Typically many trial failure surfaces are considered to find the most critical,
or the minimum value. There are various alternative methods that are available in this category.
The main difference between different limit equilibrium methods is in the assumptions made
about shape of slide surface (Circular, plane, logarithmic, etc.) and equilibrium equation that can
be satisfied (force or moment equilibrium or both). Although the “third” dimension, i.e.,
perpendicular to the plane of the cross-section, is sometimes considered (See for example, Ref.
1), it is usually assumed to be insignificant on the final results.

Slice methods are more commonly used in limit equilibrium approach to slope stability
analysis. A summary of current slice methods may be seen in the Table 1.

NUMERICAL ANALYSIS METHODS

Numerical analysis methods give reasonable approximations to the “correct” or “exact”
mathematical solution of the governing equations of the mechanics of slope stability. They are,
however, much more sophisticated and complicated than limit equilibrium methods: they take
into account deformations (strains) and not just forces (stresses) like the more conventional limit
equilibrium methods do.

Numerical methods have been extensively used in the past several decades due to advances in
computing power. In a broad sense, numerical methods can be classified into continuum and
discontinuum methods.

There is quite a large number of numerical methods that have been presented in the literature
to estimate the behavior of systems made of geomaterials. The most important and the most often
used methods are: (a) For continuum, Finite Difference Method (FDM), Finite Element Method
(FEM) and Boundary Element Method (BEM). (b) For discontinuum, Distinct Element Method
(DEM); Discontinuous Deformation Analysis (DDA), and Bonded Particle Model (BPM).

The need to use, for a particular problem, continuum or discontinuum methods depends on
the size, or scale, of the discontinuities with respect to the size, or scale of the problem that needs
to be solved [Antonio Bobet, 2010].

In the following, the continuum methods that are more commonly used in slope stability
analysis are presented.
Vol. 16 [2011], Bund. L 1247

Table 1: Summary of some LE methods [Abramson et al, 2002]

 Vol. 16 [2011], Bund. L 1248

Finite Difference Method

The Finite Difference Method (FDM) is based on the premise that governing differential
equations of elasticity theory can be adequately represented by finite differences. The method is
the oldest among the numerical methods in geomechanics and it was used even before the arrival
of computers with the FDM, the set of differential equations is reduced to a system of linear
equations, which can be solved by any of the classical methods.

With the method a grid is superimposed to the domain, as shown in Figure1.In the figure, the
sub-indices represent the position of the point in the grid; For example i,j represent a point with
coordinates (xi, xj, k) and so on. The method relies on the approximation of the field equations,
i.e. equilibrium, strain compatibility etc. by finite difference formulas [Antonio Bobet, 2010].

Finite Element Method

The finite element method is a powerful calculating method in engineering sciences. This
method is by far method used for analysis geotechnical problems, The FEM gives it power by
ability simulating physical behavior materials by use of a calculating tools without simplify
problem. Today, new analysis method in engineering can be studied with FEM as reference of
exact solution.
In the last two decades, several methods were presented for slope stability analysis by FEM.

Through this methods, gravity increase method and strength reduction method have widely
use. In the gravity increase method, the gravity forces, such weight, increase gradually until the
slope becomes unstable (gp). Then factor of safety define ratio between gravitational acceleration
in failure time and actual gravitational acceleration (g).

In the strength reduction method, decrease strength parameters of the slope until slope
become unstable. So, the safety of factor determine ratio between actual strength parameters and
critical strength parameters. The gravity increase method is well suited for analyzing the stability
of embankment is constructing because the rate of construction can be simulated with the rate at
which gravity loading on the embankment is increased [Colby C. Swan, 1999].

Boundary Element Method

With the Boundary Element Method (BEM), only the boundaries of the continuum need to be
discretized. See Figure 1, this is in contrast to the other two continuum methods, The finite
Difference and the Finite Element methods, where the entire medium has to be discretized.

In the BEM, the solution is approximated at the boundaries while equilibrium and
compatibility are exactly satisfied in the interior of the medium. The advantage of limiting the
discretization to the boundaries is that the problem is reduced by one order: From 3D to a 2D
surface problem at the boundary, and from 2D to a line problem [Antonio Bobet, 2010].

Requirements for slope stability analysis with numerical methods are: High-end computer;
have relevant software; understanding of advanced soil mechanics, including material models and
numerical methods.
 Vol. 16 [2011], Bund. L 1249

The following is a list of the codes most referenced in the literature:

Finite Difference Method: ABAQUS (Hibbit, Karlson and Sorenson, Inc ); PENTAGON-2D
and 3D (Emerald Soft); PHASE2 (Rockscience); PLAXIS (Plaxis BV) Boundary element
Method: BEFE (Coupled BEM-FEM, computer software and services (CSS).

Figure 1: Examples for FDM, FEM and BEM [Antonio Bobet, 2010]
Continues on the next page.
 Vol. 16 [2011], Bund. L 1250

Figure 1: Examples for FDM, FEM and BEM [Antonio Bobet, 2010]

ARTIFICIAL NEURAL NETWORK

As we saw in limit equilibrium methods, It’s necessary to have simplification assumptions
which questions the accuracy of these methods.

This problem will be solved by numerical methods, and we can analyze the slope geometry
with the most care that will be determined slope stability but in using the numerical methods there
are some factors that should be taken into consideration, they follow as: having the knowledge of
finite elements or finite difference, Leaving computer programs, to have enough time to prepare
input data and running computer programs.

As a suitable and substitute method for the above mentioned ways. Artificial Neural Network
will be used to analyze the stabilities of slopes and earth dam. In this case Neural Network with
high flexibility in learning problems have the advantage of LE and numerical methods together
and also improves the defect and short coning of LE and numerical methods.

In spite of the accuracy of numerical methods, it also includes the more simplicity and
acceleration compared with LE and removes the necessary of learning software and the
knowledge of numerical methods in calculation of fact or of safety.

Artificial neural networks based on biology models such as human brain rely on information
processing techniques based on establishing association between parameters.
 Vol. 16 [2011], Bund. L 1251

As with the human brain, ANNs are composed of a number of interconnected units called
neurons. Each neurons receives information, process the information, and send the result to other
neurons. The exclusive ability learning neurons, gives to this system the ability of training
difficult and nonlinear relations.

This procedure permits to user to “train” the network directly and store results as the
properties of the connections between neurons.

So after training stage, the artificial neural network can be used to simulate a real system to
predict the relation between known/given variables (input) an dependent variables (output).

The first generation of artificial neural networks started in the 1940s to 1960s. but it was not
until the 1980s when the introduction of new architectures and learning processes made ANNs
useful as practical tools. There are several types of Artificial Neural Network depending on the
characteristics of each neuron, the learning or training scheme, network topology, and network
function [Antonio Bobet, 2010].

Multi-Layer Perception Network (MLP) have been use in geotechnical problems very much.
This network compound two or several layers. Figure 2 shown one MLP with four layers. The
information stored in each neuron, often called the state of the neuron.

The process in the network works as follow: an array of input values is defined as state of the
neurons in the input layer. These value are transmitted to the second, hidden layer, following the
protocol defined Equation (1), where Xi (k) the input i of neuron k; wi dependent weight; σ
function of neuron k and y is output (or input to next layer).

Eq. 1 represents the manner of inserting data into layers [Bani Mahd et al, 2002]

V (k)= ∑ 𝑤 𝑋 (𝑘) → y (k)= σ (V (k)) (1)

The state of the neurons in the second layer is transmitted to the third layer where new
calculations are performed to obtain the state of the neurons in this layer. The process is repeated
until the output layer is reached. The state of the neurons in the output layer constitutes the output
of the system.

The weights gives from training network. Training of the network is done by comparing the
output provided by the ANN with actual results associated with a given input.

There are different strategies to minimize the error (difference between actual results and
network output) by changing the values of the weights. The most common strategy is the back
propagation algorithm or delta rule, where the derivatives of the error function with respect to the
weights are set to Zero.
 Vol. 16 [2011], Bund. L 1252

Figure 2: Architecture of a four-layer network [Shangguan et al., 2009]

There are no rules to design ANNs. The input and output neurons. In terms of numbers and
characteristics, are defined by the user. Thus, the user needs to decide what are the variables that
may affected the result and what are the results needed. The number of hidden layers and the
number of neuron per layer is problem-dependent. Increasing the number of neurons and/or
hidden layers does not necessarily result in better predictions [Antonio Bobet, 2010].

For design network, we need to experimental information and results (for example, in slope
stability problem, we need to have a number of slopes with exact F.O.S) The user often divide the
available data in two sets: one for training and the other for validation, then design a ANN. Some
of designed networks for slope stability analysis are: Samui and Kumar (2006); Sakellatiou and
Ferentinou (2005), P. Lu and M.S. Rosenbaum (2003).

LIMIT ANALYSIS METHOD

The limit analysis method models the soil as a perfectly plastic material obeying an
associated flow rule. With this idealization of the soil behavior, two plastic bounding theorems
(lower and upper bounds) can be proved [H. S. Yu et al., 1998].

The bound theorems of limit analysis are particularly useful if both upper and lower bound
solutions can be calculated, because the true collapse load can then be bracketed from above and
below. This feature is invaluable in cases for which an exact solution cannot be determined ( such
 Vol. 16 [2011], Bund. L 1253

as slope stability problems).because it provides a built-in error check on the accuracy of the
approximate collapse load [H. S. Yu et al, 1998].

In the upper bound theorem, should be assumption a allowable displacement field at the
firstly. The displacement field compound from ones with constant velocity, then have to compute
two rates:

1. Rate of work done by external loads (such as weight)

2. The rate of lose energy inside slide surface

According to the upper bound theorem, for each compatible lastic displacement collapse will
be occurred if the rate of lose energy inside slide surface. By standing different mechanisms, the
best (least) upper bound value may be found.

In the lower bound theorem, should be assumption a allowable stress field, firstly, this stress
filed is a discontinuum field and compound from several separate zone which covering the whole
soil mass.

According to the lower bound theorem, if this filed can satisfying equilibrium equations and
stress boundary conditions, and stress lower then yielding stress, so failure not will be occurred.

In this note that in the lower bound theorem, the strain and displacements are not considered
and that the state of stress is not necessarily the actual state of stress at collapse. By examining
different admissible states of stress, the best (highest) lower bound value may be found.

In recent years, many efforts have been made in the application of the plasticity limit
theorems to limit analysis of slope stability Donald & Chen (1997) proposed and energy work
balance approach (or the upper bound approach using the associated flow rule). Wang et al.
(2001) developed this methods to investigate the influence of a nonassociated flow rule on the
calculation of the factor of safety (FOS) of two-dimensional (2D) soil slopes. Chen et al (2001a,
2001b). recently extended the upper bound method for three –dimensional (3D) slope stability
analysis. Sloan (1988,1989), Sloan and Kleeman (1995) and Lyamin and Sloan (2002) have made
significant progress in developing new methods using finite elements and linear programming
(LP) for or nonlinear programming (NLP) for computing rigorous lower and upper bounds for 2D
and 3D stability.

CONCLUSION
This paper discusses and categorizes the available methods for slope stability analysis. LE
methods have less accuracy than another methods. Numerical methods and LA method have
acceptable accurate but these methods have required enough time and knowledge. ANN method
also have required much data to design network.

Accurate of choice method for analysis can be based on two important parameters: location
of slope and shape of probable slip surface.
 Vol. 16 [2011], Bund. L 1254

REFERENCES
1. Albataineh, N. (2006) “Slope Stability Analysis using 2D and 3D methods,” thesis for the
degree Master of Science, University of Akron, pp 1-36.

2. Abramson, L.W, T.S. Lee, S. Sharma, and G. M. Boyce, (2002) “Slope stability and
stabilization methods, " John Wiley & sons, Inc, New York, USA.

3. Bobet, A. (2010) " Numerical Methods in Geomechanics, “The Arabian Journal for
Science and Engineering, volume 35, number 1B.

4. M. Cala, J. Flisiak and A. Tajdus, (2002) “slope stability analysis with modified shear
strength reduction technique,” AGH University of Science Technology, Poland.

5. K. P. Aryal, (2008) “Differences between LE and FE methods used in slope stability

evaluations, " The 12th International and Conference of International Association for computer
methods and Advances in Geomechanics (IACMAG), India.

6. Colby, C. Swan and Young-Kyo Seo, (1999) “Slope stability analysis using finite element
techniques, " 13th Iowa ASCE Geotechnique Conference, USA.

7. Bani Mahd, M. and Yasrebi, S. SH. (2002) “Application of Artificial Neural Network in
Geotechnic Engineering, " Modarres Journal, volume 14, number 1.

8. M. G. Sakellariou and M. D. Ferentinou, (2005) “A Study of Slope Stability Prediction

Using Neural Network, " springer, Geotechnical and Geological Engineering, volume 23, pp 419-
445.

9. Sarvut, J., Somchai, C., Chusak, L. and Rittisak, J. (2002) “Neural networks: a tool for the
slope stability analysis, " 3rd International Conference on Landslides, Slope Stability & The
Safety of Infrastructures, Singapore.

10. Zichang Shangguan, Shouju Li and Maotian Luan, (2009) “Intelligent Forecasting
Method for Slope Stability Estimation by Using Probabilistic Neural Networks, " EJGE, volume
13, Bund .C.

11. Chen, J., Yin, JH., Lee, CF. (2003) “Upper bound limit analysis of slope stability using
rigid finite elements and nonlinear programming, " Canadian Geotechnical Journal, volume 40,
number 4, pp 742-752.

12. By H.S. Yu, R. Salgado, S. W. Sloan and J. M. Kim, (1998) “Limit Analysis versus Limit
Equilibrium for slope stability,” Journal of Geotechnical and Geoenvironmental Engineering,
volume 124, number 1.

© 2011 ejge

View publication stats