CHAPTER 1

CHAPTER 1
INTRODUCTION
 CHAPTER 1

CHAPTER 1
INTRODUCTION
Slope is an unsupported inclined surface of soil or rock mass which maybe
natural or manmade such as embankments and cuttings. The failure of a soil mass in
a slope is called a slide, which comprises of downslide and out slide of the entire soil
mass located above the sliding surface. It mainly occurs due to the forces involving
weight of the soil. The failure of slopes may occur along a plane or curved surface
when large mass of soil slides. The geometry and strength of the slope are the major
parameters that determine the shape of sliding surface. Stability analysis of a slope
on soft soils is one of the major problems of geotechnical engineering. The important
steps in stability analysis are to evaluate the potential failure surface and the
corresponding forces that appear to cause slip and to restore or stabilize the sliding
mass along with the safety factor calculation (Sengupta and Upadhayaya, 2009).
Many approaches such as limit equilibrium methods, finite element method and
optimization techniques have been proposed in the past for the stability of slopes.
The various methods used for slope stability analysis are briefly discussed below.

1.1 Limit equilibrium methods for slope stability analysis

Due to their simplicity and applicability, limit equilibrium methods are the
most common and popular methods for stability analysis. In these methods, a
number of potential sliding surfaces are selected and a stability analysis in terms of
safety factor is carried out on each surface by equating the forces causing sliding and
forces resisting sliding. A surface with least factor of safety is then selected as critical
sliding surface to obtain the factor of safety of a slope. Fellenius(1927) suggested
initial work on finding the critical failure surface in a slope stability study, in which a
trial slip surface is considered to be an arc of a circle and the material above the
assumed slip surface is divided into a number of vertical slices. Using overall moment
equilibrium of the forces causing and resisting the slide, about the center of the
assumed slip surface, safety factor is obtained. Later many methods were proposed

1
 CHAPTER 1

to account for more general failure of slope based on non-circular sliding surface
(Janbu (1957); Morgenstern and Price (1965); Spencer (1967); Sarma (1979)). Many
computer programs were also developed to facilitate the generation of large number
of slip surfaces required to obtain the realistic safety factor of slopes.

The limit equilibrium method is based on the assumption that there is similar
material and loading conditions throughout the slope geometry. In this method, the
forces resisting sliding is the shear strength of the soil and is generally obtained using
Mohr Coulumb equation *τ= c’+σ’tan(φ’)]. The forces causing sliding are the shear
forces developed along the slip surface due to the weight of the slope above the
sliding surface. Most commonly adopted method to determine the normal and shear
stresses required for stability analysis in this approach involves dividing the profile of
slope into number of slices, and determining tangential and normal components of
the weight of each slice along the sliding surface. Although this approach is straight
forward, it may not give reliable results for nonhomogeneous and anisotropic soil
mass. Also, stress-strain behavior of soil has a major role on distribution of stresses
in soil and hence the safety factor can vary significantly depending upon the
assumptions made in the analysis (Scott and Yamasaki, (1993)), and the conditions of
equilibrium used to attain the interslice forces. Since the shape of slip surface in
these methods is predetermined it may not be the weakest surface and is a
limitation when the slope is non-homogeneous.

The shape of the failure surface varies depending on the type of soil and can
be a plane, circular, curved or logarithmic or a combination of them. If the soil in the
slope are homogeneous, the failure surface will be close to the circle. But if the slope
is made of non-homogenous soil then the shape of slip surface may not be a circular.

An example of embankment constructed on non-homogenous foundation

soil (Ellithy and Stark, 2020) is shown in Figure 1.1. The non-homogeneity is in the
foundation soil with layers of silty clay having varying shear strength as shown in
Figure 1.1. The construction of embankment when reached a height of 2.4m, the

2
 CHAPTER 1

cracks were formed and once the embankment construction completed up to 43% of
height, failure was observed. The observed failure surface is as shown in Figure 1.1.

Figure 1.1 Details of failed embankment (Ellithy and Stark, 2020)

Another example of failed slope is in Hadong site (Oh and Lu, 2015), having
non-homogeniety as shown in the Figure 1.2. The slope is an engineered cut slope
along a road that has been in service without any instability problems for several
years following construction. The slope failed with the failure surface as shown in the
Figure 1.2. From computations, initial FOS is 1.5 but it decreased gradually 1.33 with
rainfall. Further with increase in rainfall there was abrupt decrease in the FOS to 1.05
thus causing failure.

Figure 1.2 Details of failed embankment (Oh and Lu, 2015)

The non-homogenous natural slope having complex soil layers as shown in

Figure 1.3 is analyzed by Zolfagari et.al (2005) using Morgenstern–Price method.
From the analysis it was observed that for a slope with a layered structure, the

3
 CHAPTER 1

methods assuming circular slip surfaces can over predict the factor of safety, which
might lead to conservative estimates of slope stability. The results showed that, the
non-circular failure surface represents the weakest layer and has a lower factor of
safety than the circular failure surface.

Figure 1.3 Slope details with the circular and non-circular slip surfaces (Zolfahari et. al, 2005)

A failed slope analyzed by Deng et.al (2019) having non-homogeneity in

embankment soil is as shown in Figure 1.4. The slip surfaces obtained by various
methods are non-circular as shown in Figure 1.4. The factor of safety for these slip
surfaces is less than 1.

Figure 1.4 Details of failed slope with circular and non-circular slip surfaces (Deng et.al, 2019)

4
 CHAPTER 1

1.2 Optimization techniques in slope stability analysis

In stability analysis of slopes using the limit equilibrium method, various slip
surfaces are assumed and their factor of safety is calculated. The critical slip surface
is then chosen from the assumed slip surfaces, as the one with least factor of safety.
While assuming the slip surfaces, there are chances that the actual critical slip
surface may not be considered. Optimization is a process of minimizing or
maximizing some parameters of importance and obtaining the best solution from a
bunch of alternatives. Thus, optimization techniques are applied to generate the slip
surfaces and minimize the factor of safety to obtain the critical slip surface. Similar to
earlier methods, the variation of material and loading condition are neglected in this
technique also, while calculating the normal and shear stresses.

For analyzing the slope stability, optimization techniques are found to be

convenient and efficient in both deterministic and probabilistic analysis. The use of
optimization techniques in slope stability analysis was first investigated by Baker
(1980), by combining the dynamic programming technique, an optimization
procedure with Spencer's method (Spencer, 1967) to locate the critical sliding
surface and calculate the associated safety factor. Over the past two decades,
several advanced approaches based on optimization techniques have been
established to identify the critical sliding surface through the minimization of the
safety factor. In addition to the application of nonlinear programming, genetic
algorithms, Monte Carlo techniques, evolutionary algorithms and artificial neural
network were also suggested for analyzing the slope stability.

Before the emergence of the deterministic method of nonlinear

programming in 1960, the only way to search for the minimum of a function having
several variables was the grid method known as Monte Carlo techniques (Greco,
(1996)). These approaches are based on random generation of solutions and
improving it by matching with best solution available. The random walk and random
jumping type are the two main classification of Monte Carlo techniques. Random
jumping techniques are based on a large number of trial slip surfaces being

5
 CHAPTER 1

generated and the critical slip surface is presumed to be the one with lowest safety
factor. The previous trial is not considered in each generation throughout these
methods. On the other hand, random walking techniques create random slip
surfaces similar to the current best solution by changing them slightly. In the second
approach, a sequence of enhanced approximations to the minimum is accomplished
with less iterative effort. To locate the critical sliding surface, Monte Carlo
techniques have been used by Greco 1996; Siegal et al. (1981); Cherubini and Greco,
(1987); Malkawi et al., (2001); Alkasawneh et al., (2008), Krishnamoorthy (2010,
2013), Krishnamoorthy and Smith (2016).

Further other methods based on optimization techniques such as Simplex

method (Nguyen, (1985)), Davidon -Fletcher-Powell method (Yamagami &
Ueta,(1988)), Sequential unconstrained minimization technique (Bhowmik and
Basudhar, (1989)) and genetic algorithm (Zolfaghari et al., (2005); Cheng et.al.,
(2007); Sengupta and Upadhyay (2009); Tran and Srokosz, (2010)) to predict the
critical sliding surface have been proposed. Kim and Lee (1997), proposed an
optimization technique where in the stress fields were obtained by finite element
analysis. Nonetheless, when comparing the various optimization techniques for
finding the global minima, genetic algorithms have been found to be more efficient
in determining the optimal solution for many complex problems (Goldberg, (1989)).
Genetic algorithms can provide more efficient results and can analyze geotechnical
failure problems involving many decision variables (Cui and Sheng, (2005)).

In several complex problems, evolutionary approaches such as the more

robust genetic algorithm are applied to find the optimal solution (global solution).
Genetic algorithms are ideal for solving complex optimization problems and are good
for applications involving strategies for adaptive problem solving. Genetic Algorithm
was developed by John Holland in 1975 to study the self adaptiveness in processes of
natural systems. Genetic algorithms are search algorithms based on the mechanics
of natural selection and natural genetics (Goldberg, 1989). To solve optimization
problems a computational model that emulates biological evolutionary theories is

6
 CHAPTER 1

used. Genetic algorithm is inspired by Darwin’s theory of evolution i.e., the survival
of the fittest, and has many applications of finding solutions for complex problems.
Genetic algorithm comprises a set of individual elements and a set of biologically
inspired operators defined over the population itself. In computing terms, genetic
algorithm records a problem on to a set of (binary) strings, each string representing a
potential solution. To search improved solutions, genetic algorithm then manipulates
the most promising strings and operates in four stages: i) creates a “population” of
string, ii) evaluates each string, iii) selects "best" strings, and iv) creates the new
population of strings using genetic manipulation.

Genetic algorithm find their applications in all fields of engineering design

because they are found to be brilliant to find the right materials, right shapes to
create stronger, faster and overall better products. Goh (1999), has used Genetic
algorithm to find the critical surface and the factor of safety using method of
wedges. McCombie and Wilkinson (2002), used Bishop’s simplified method and
Sabhahit et al., (2001) and Sabhahit and Rao, (2011), have used Janbu’s method to
search for the critical surface using Genetic algorithm. Optimization techniques
based on genetic algorithm have been proposed by many researchers (Zolfaghari et
al., (2005); Cheng et al., (2007); Sengupta and Upadhyay (2009); Tran and Srokosz
(2010)) to find factor of safety. It has been reported that Genetic Algorithm could
find better solution compared to other traditional optimization tools. However, the
genetic algorithm has been used in conjunction with limit equilibrium methods to
instigate the stability analysis of slope and determine the least factor of safety. The
stress-strain behavior of soil is not considered in these methods, and therefore is a
limitation when non homogeneous soil properties are encountered.

1.3 Finite element method in slope stability analysis

The slope stability analysis to obtain safety factor, when conducted using
limit equilibrium methods, depends on the assumptions based on interslice forces.
Also, the limit equilibrium methods in conjunction with the available optimization
techniques do not consider the stress-strain behavior of soil. Hence many

7
 CHAPTER 1

procedures based on finite element method which considers the stress-strain

behavior of soil are proposed due to its versatility, simplicity and reliability in slope
stability analysis. The complex boundary and loading conditions can be dealt easily in
finite element method when compared to limit equilibrium techniques (Kim et.al,
2002). Also, non-homogeneous soil properties and nonlinear stress-strain conditions
can be imposed easily in this method.

The finite element method is a numerical technique for finding approximate

but acceptable solutions to boundary value problems and problems involving
complex material properties. The finite element method can be systematically
programmed to accommodate complex and difficult problems such as non-
homogenous, non-linear stress-strain behavior and complicated boundary
conditions. Finite element method is the representation of a body or a structure by
an assemblage of sub-divisions called finite elements. These elements are
considered as interconnected at joints which are called as nodes or nodal points.
Simple functions are chosen to approximate the displacements over each finite
element. Such assumed functions are known as displacement functions or
displacement models which can be expressed as polynomials or trigonometric
functions. The equilibrium equation for each element is then developed and the
equation of equilibrium for entire body is obtained by combining equations for the
individual element in such a way that the continuity of displacement is preserved at
interconnecting nodes. The required boundary conditions are imposed and resulting
equation of equilibrium are solved for nodal displacement. After obtaining
displacement of each element, the stresses and strains are obtained using the
element properties.

With the availability of high-speed computers, finite element method has

become popular and are used widely in stability analysis of slopes. The finite element
method makes it possible to calculate the complete state of stress and strain in each
element, and thus the movements in the slope. Methods based on finite element
method have been proposed for the problems involving complex boundary and

8
 CHAPTER 1

loading conditions. Duncun and Dunlop (1969), and Donald and Giam (1988), have
derived factor of safety using nodal displacements obtained from finite element
method. Scott and Yamasaki, (1993) used finite element method to obtain stresses in
soil and a critical slip surface by joining the points of local failure in a slope. Delwyn
et al., (2003), combined finite element analysis and limit equilibrium method to
obtain the stresses in soil and factor of safety. Jiang and Magnan (1997) developed a
finite element method procedure to determine automatically the optimal failure
mechanism for geometrically complicated slopes under applied loads resulting in
factor of safety similar to those obtained by method of slices.

Krishnamoorthy (2010) has developed a method using finite element method

to obtain stresses at required points and used Monte Carlo technique proposed by
Greco (1996) to find critical slip surface and the corresponding factor of safety for an
embankment. In other approaches, finite element method is used to obtain the
stresses and the factor of safety is obtained either by gravity increase methods (Li et
al., 2009) in which the gravity forces are increased or by strength reduction method
(Matsui and San (1992); Kim et al., (1997, 2002); Chen et al., (2014)) in which soil
strength parameters are reduced until the slope becomes unstable. However, since
the equilibrium state of a soil mass with reduced soil strength has to be repeatedly
estimated, the shear strength reduction finite element method is usually more
expensive and time consuming than the traditional limit equilibrium method (Chen
et al., (2014)). In finite element method the constitutive model used to determine
the stresses describes the soil behavior. The linear constitutive models do not
consider the actual behavior of soils especially when the soil is non-homogenous and
deformations are observed due to application of load. Hence, a non-linear model is
required to analyze the soil behavior.

1.3.1 Non-linear behavior of soil in finite element method

The actual behavior of soils is very complicated as it behaves non-linearly,

and shows non-linear behavior below failure condition as well with stress dependent
stiffness. Soil undergoes plastic deformation and is unpredictable in dilatancy. In

9
 CHAPTER 1

finite element method, it is simple to enforce non-homogeneous soil properties and

nonlinear stress-strain conditions. In non-linear analysis, the vital constituent is use
of proper and realistic constitutive model for soil behavior. Several researchers have
proposed a number of constitutive models to describe different aspects of soil
behavior in detail and also how to apply such models in finite element modelling for
geotechnical engineering applications. But there is no constitutive model that can
completely describe the complex behavior of soils under all conditions. The most
common model used for non-linear study of soil is Elastic-Plastic model which
involves the elastic and plastic behavior based on the deformation of soil.

In Elastic-Plastic models the soil behavior is categorized by the recoverable

and irrecoverable deformations called elastic and plastic deformations respectively.
In soils there exists a yield surface wherein the response of the soil is observed to
change from stiff to less stiff. When the stress change reaches the yield surface, a
combination of elastic and plastic responses occurs. Models based on this theory are
widely accepted. To characterize an elastic-plastic model, four basic requirements
are elastic properties, yield surface, plastic potential and hardening rule.

Elastic properties, describes the way in which elastic recoverable

deformations of soil are to be described. Yield surface is a function of stress state σ
and state parameter k (called as the hardening or softening parameter), which
controls its size. When the material is considered to be perfectly plastic, then k is
constant. When hardening or softening plasticity is observed, then k varies with
plastic straining to represent the magnitude of the stress state at yield changes. The
behavior of the material is identified by the value of yield function F(σ,k). Purely
elastic behavior is observed when F(σ,k) < 0 and if F(σ,k) =0 it is elastic-plastic
behavior. Plastic potential describes the mode of plastic deformation that occurs
when soil is yielding. Usually for simplicity the plastic potential surface and the yield
surface are assumed to be identical.

There are two types of elastic-plastic models; Elastic-perfectly plastic models

where yield surface is fixed in the stress space, and Elastic-plastic models which are

10
 CHAPTER 1

characterized by existence of reversible and irreversible deformations. From the

literatures among the various constitutive models which can be implemented in FEM
program, elastic-perfectly plastic Mohr-Coulomb model which is simple and requires
less parameters compared to other models is found to be suitable. Krishnamoorthy
and Smith (2016) have applied non-linear Mohr-Coulomb method to represent a
realistic soil behavior, and to obtain the stresses in the finite element method to
calculate embankment safety factor.

1.3.2 Analysis of embankments constructed on consolidating soil

The factor of safety of the embankments built on soft saturated soils, vary
with time due to excess pore pressure dissipation after its construction. In the
stability analysis of the embankments built on these soils, consolidation plays an
important role. It has received greater attention after Terzaghi (1923) published his
consolidation theory and principle of effective stress.

At any stage of the consolidation process, the pore water pressure will vary
with time within the soil layer. When the embankment is constructed on the
consolidating soil, the initial part of load is taken by the pore water and the pore
pressure starts increasing. Once the construction is completed, the pore pressure
starts decreasing as the pore water escapes as shown in Figure 1.5 and finally
reaches to zero at the end of consolidation.

Figure 1.5 Pore pressure variation with time and load given by Gu et.al (2020)

11
 CHAPTER 1

Initially conventional preloading techniques were used to accelerate the

consolidation process in soft soils. With the advancement in techniques vacuum
preloading and conventional preloading with or without drains were introduced.
Preloading with vertical drains is a successful ground improvement technique. The
sand drains, stone columns, lime columns, prefabricated vertical drains, and stone
columns/ sand drains enclosed within geosynthetic materials are being used to
reduce the drainage path, thus increasing the rate of consolidation and also to
improve stability of embankments on soft consolidating soils.

In most cases, consolidation settlement is obtained by one dimensional

consolidation theory of Terzaghi (1943). A more general theory for three
dimensional consolidation, coupling soil deformation and pore pressure, was
developed by Biot (1941). The conventional method of finite element has also
proven to be an extremely effective analytical mechanism for solving problems of
consolidation. Sandhu and Wilson (1969) have developed an analysis using the finite
element method for the coupled consolidation problem. Lewis et al., (1976),
proposed a formulation for a nonlinear consolidation problem in which the
relationship between hyperbolic stress and strain proposed by Duncan and Chang,
(1970) has been used to model soil action.

The consolidation behaviour of strip footing was studied by Manoharan and

Dasgupta(1995), by modelling the behaviour of soil as elastic-perfectly plastic
satisfying the Mohr-Coulumb yield cretirion. Later many researchers (Bergado et. al.,
(1993), Indraratna et. al., (1994), Borges (2004), Krishnamoorthy (2010, 2013)) have
analyzed embankments on soft consolidating soil in order to analyze the
embankment settlement. The effect of rate of construction on stability of
embankments on consolidating soil is evaluated (Seo and Swan, 2001) using load-
based safety factors rather than resistance-based safety factors. The critical height of
embankment that can be constructed on a low bearing capacity soil has been
assessed (Lastiasih and Putu, (2015)).

12
 CHAPTER 1

Though, established numerical method such as plane strain finite element

analysis have been developed and recommended for the analysis of consolidation
problems, the method is limited to the investigation of the settlement behavior of
the embankment and pore pressure in underlying soft consolidating soil. There have
been no attempts made to find the safety factor of the slopes on soft soils. The
factor of safety and the critical slip surface are the basic results for the stability
analysis of slopes (Kim and Lee (1997)). It is noted that the safety factor of the slope
built on consolidating soil varies with time due to the dissipation of excess pore
pressure and there is a need to propose a method to obtain the factor of safety of
the slope on soft consolidating soils. However, stress-strain behavior of soil has a
major role in the stability analysis and the factor of safety can vary significantly
depending upon the assumptions made in the analysis (Scott and Yamasaki, (1993))
and conditions of equilibrium used to attain the interslice forces.

Among the various optimization techniques, genetic algorithm finds better

solution compared to other traditional optimization techniques. Genetic algorithms
are suitable for the solution of complex optimization problems, and consequently
are good for applications that require adaptive problem solving strategies. Genetic
algorithm is found to be sufficiently robust to handle layered soils with weak, thin
layers and as efficient and accurate as the conventional pattern search method (Goh,
1999). It can produce the results comparatively faster and can be applied to evaluate
the failure performance of geotechnical problems involving a large number of
decision variables. However, optimization techniques using limit equilibrium
methods do not consider the stress strain behavior of soil mass. The major
advantage of finite element method is that it can deal with complex boundary and
loading conditions (Kim et.al, (2002)).

In the present work genetic algorithm is used in conjunction with finite

element method to determine the critical slip surfaces and the factor of safety. The
genetic algorithm generates a number of slip surfaces and determines the critical slip
surface based on least factor of safety. The finite element method is used to

13
 CHAPTER 1

determine stresses along the generated slip surfaces and calculate the factor of
safety for each surfaces. In the consolidation analysis, the non-linear Mohr-Coulomb
model is implemented in FEM to accommodate the realistic behavior of soil.

1.4 Outline of the thesis

Chapter 1- Introduction: This chapter introduces about the available

methods to evaluate slope stability analysis. In this chapter a brief discussion on the
finite element method, genetic algorithms, non-linear analysis of soil and the
problems of embankments constructed on consolidating soil are explained. This
chapter finally give the outline of the thesis.

Chapter 2 – Literature review: This Chapter provides review of literature for

stability analysis of slopes using limit equilibrium methods, finite element method,
optimization techniques, non-linearity of soil and stability analysis of embankments
constructed on consolidating soil. This section highlights the work done by the
former researchers to develop various methods for stability analysis of slopes. This
literature review has given insight to the various approaches and also the limitation
of the previous researchers. The chapter ends with the identification of the research
gap and setting of objectives.

Chapter 3 – Methodology: This chapter discusses the approach and method

adopted for the research. The research method used has been briefly described with
the reasoning behind using this method. This chapter outlines the technique to
calculate the factor of safety combining genetic algorithm and finite element
method.

Chapter 4 - Stability analysis for non-homogenous slopes: This chapter

discusses the analysis and results of three non-homogenous slopes using the
proposed method combining finite element method and genetic algorithm. The
method proposed is validated and the factor of safety are compared with various
methods available. The parametric study involved in fixing the parameters of genetic
algorithm is also discussed.

14
 CHAPTER 1

Chapter 5 – Application of proposed method in Stability Analysis of

embankment constructed on soft consolidating soil: This chapter discusses the
validation of the problem when non-linear models are implemented. The stability
analysis of embankments on consolidating soil from end of construction to end of
consolidation is carried out using proposed method for different time intervals.

Chapter 6 – Research findings and Conclusions: This final chapter reports

the outcomes of the study. Based on the results of research, a new method
combining finite element method and genetic algorithm is validated and applied to
non-homogenous slope examples for both linear and non-linear analysis. This
chapter completes with conclusions and scope for future work of this research.

15