CHAPTER 3

THREE-DIMENSIONAL (3D) SLOPE STABILITY ANALYSIS

3.1 GENERAL

Three-dimensional slope-stability problems are rarely used in practice as they are

more complicated than the two dimensional methods. However, a two-dimensional

analysis can be regarded as conservative where 3D failure is expected and it is often

preferred in design (Cornforth, 2005). The three-dimensional slope-stability analysis

marks its importance where the nature of the slope is highly complex and it is difficult to

select a two-dimensional plane strain analysis. The importance of three-dimensional

slope-stability analysis can be found where the slope geometry and the slip surface differs

significantly in the lateral direction, the material properties are highly non-homogenous

and anisotropic, the slope is locally surcharged and to back calculate the shear strength of

a failed slope. It was Fredlund in 1970 who illustrated the benefits associated with

performing 3D slope stability. In the recent years, many 3D slope-stability methods were

researched ranging from method of columns based on variational calculus to the use of

dynamic programming. The increase in the importance of three-dimensional slope

stability is due to the fact that most of the slope failures are three-dimensional in nature

having a dish-shaped failure surface. Like the 2D methods, the 3D methods also require

some assumptions to achieve a statically determinate definition of the problem. Some 3D

methods do it by decreasing the number of unknowns, while some others achieve it by

increasing the number of equations or both, such that the two numbers tally with each

other.
3.2 LITERATURE REVIEW ON THREE-DIMENSIONAL SLOPE STABILITY

ANALYSIS

A literature review of the various researches done in the field of 3D slope stability

is discussed in the following sub-sections.

3.2.1 3D Limit Equilibrium Approach

The 3D slope stability have been used since 1969. Almost all the 3D limit

equilibrium methods (LEMs) were extended from 2D slice methods. The first 3D slope

stability method to calculate the Factor of Safety (FOS) was given by Anagnosti (1969).

This method was an extension of earlier Morgenstern and Price’s (1965) method. A

similar procedure of determining the 3D FOS was done by Sun et al. (2011). Hungr

(1987), Hungr et al. (1989), Ugai (1988), Huang et al. (2002) and Cheng and Yip (2007)

also extended the 2D LEMs to develop the 3D methods for determining the FOS. The

various 2D LEMs include Fellenius method (1936), Simplified Janbu method (1954),

Bishop’s method (1955), Generalized Janbu method (1957) and Morgenstern and Price’s

method (1965). The assumptions of each of these 3D methods followed the corresponding

assumptions of its 2D origin, but the slip surface was assumed different for different

slopes. Some researchers assumed it to be a rotational surface of circular cross-section

and some others assumed cylindrical cross-section. The FOS obtained by the 3D methods

was found to be higher than the 2D methods. Chen et al. (2003) presented a simplified

3D slope-stability analysis based on limit equilibrium theory which is basically an

extension of Spencer’s (1967) method. A parallel inter-column force inclination was

assumed similar to Spencer’s 2D method. This assumptions satisfies both the force

equilibrium and the moment equilibrium requirement about the main axis of rotation.
Jiang and Yamagami (2004) also extended the Spencer’s (1967) method and assumed the

direction of shear to be perpendicular to the longitudinal extent of the slope. They

established two different equations for FOS one with respect to horizontal force and the

other with respect to overall moment equilibriums. The FOS was then determined by

simultaneous solving of these equations with different values for inclination of inter-

column forces. The intersection point of two resultant plots, achieved from two equations

of FOS, resulted overall FOS.

3.2.1.1 Slope stability by LEM for cohesive soil

Many researchers have done many researches in the field of 3D slope stability for

cohesive soil. Baligh and Azzouz were the first to present a 3D method for cohesive soil

based on circular arc method in 1975 where the slip surface was assumed to be a

combination of cylindrical centre part with conical ends. Hovland (1977) also performed

a similar kind of work on 3D slope stability of cohesive soil. His method was basically

an extension of ordinary method of slices, but the method ignored all the inter-column

forces on the sides of the columns and the pore water pressure. The conclusion of his

work indicated that for cohesive soils, the 3D FOS is always higher than the 2D FOS.

Azzouz and Baligh again in 1978 made an attempt to expand their previous work of 1975.

The assumptions related to shear resistance force did not change; rather, two new

assumptions were introduced for the distribution of other forces. The first assumption

followed the method of slices (Fellenius, 1936) and to calculate the normal stresses based

on moment equilibrium of each slice and the second one assumed that the vertical

effective stress is the major principal stress and the horizontal stress is the minor principal

stress. Based on the analysis of four embankments done by Azzouz and Baligh, they
concluded that their new assumptions provide more reasonable results than the previous

assumptions of slope-stability method. Chen and Chameau (1983a, 1983b) presented a

3D method to analyze symmetrical homogenous cohesive and frictional slope. Chen and

Chamaeu considered both the force and moment equilibrium and different pore water

pressure conditions in the analysis. They finally found the 3D FOS to be higher than the

2D FOS in the presence of pore water pressure. A similar work on cohesive soil was done

by Gens et al.in 1988 where the slip surface was assumed to be similar to the work done

by Azzouz and Baligh (1983). The assumed slip surface was a combination of cylindrical

centre part followed by curve ends to calculate the FOS. The results of Gens et al. showed

that the ratio of the 3D FOS to the 2D FOS is more than unity and varies from 1.03 to

1.30.

3.2.1.2 Slope stability by LEM having external loads

Many other research methodologies include the application of various loading

conditions to know the effect of loading conditions on the stability of soils. Azzouz and

Baligh (1983) extended the 3D method of their previous work of 1975 to consider the

effect of applied loads on the stability of slopes. The geometry of the slope remained

simple and the slip surface was assumed to be a combination of central cylindrical and

ellipsoids at the ends. A comparative study was made between 2D and 3D slope-stability

problems due to the distribution of local loads for finding the FOS. All numerical

procedures were kept similar to their previous method of 1975. From the several practical

cases conducted by Azzouz and Baligh, they finally come to the conclusion that the effect

of 3D analysis could increase the capacity of critical load of 2D analysis between 5 and

10 times. Again Dennhardt and Forster (1985) proposed a 3D model to find the FOS of
symmetrical slopes with ellipsoidal slip surface by considering a symmetrical external

load on top of the slope. Dennhardt and Forster assumed a distribution of normal stress

throughout the slip surface to overcome the indeterminacy of the problem. The calculated

3D FOS by this method was found to be higher than corresponding 2D factor.

3.2.1.3 Slope stability for seismic condition

Very few 3D methods have been established based on limit analysis method to

determine the seismic stability of slopes. Some of the recent studies on 3D seismic slope

stability have been summarized. Ganjian et al. (2010) proposed a 3D method based on

upper bound theorem of limit analysis to determine the seismic stability of slopes under

local loading. Using the proposed 3D rotational collapse mechanism and applying the

energy dissipation method, seismic stability factors for non-associated slopes were

determined, and then the effects of dilatancy angle on the stability of locally loaded slopes

were investigated. On comparing the results with other analytical and numerical methods,

they finally came to a conclusion that the dilatancy angle is more important in 3D seismic

analysis of locally loaded slopes. Nadukuru et al. (2011) developed a 3D slope stability

analysis with quasi-static distributed force. The charts developed for calculating the FOS

was found to be very advantageous as it does not need any iterative procedure. The

analysis is found to be applicable in cases where the width of the mechanism is found to

be limited or when the mechanism is confined by local geology. They also developed an

analysis for calculations of critical acceleration coefficient and displacements due to

seismic excitation. However, the analysis was found to have a limitation that it is

applicable to slopes of inclination not smaller than 45°. Michalowski and Martel (2011)

carried out a 3D slope-stability analysis limited to steep slopes based on the kinematic
theorem of limit analysis. A rotational failure mechanism is used with the failure surface

in the shape of a curvilinear cone sector passing through the slope toe, typical of steep

slopes. Based on quasi-static approach, stability charts were developed to calculate the

safety factor and the charts were found to have high practical importance as it does not

require any iterative procedure for estimating the FOS. Nadukuru and Michalowski

continued to work on the kinematic theorem of limit analysis and in 2013 developed a

method to calculate the yield acceleration of slopes that fail in a 3D manner, with an

assumed width of the mechanism. An analysis was then carried out to arrive at the

displacements of slopes subjected to ground shaking. The outcome of the analysis was

found to be very convenient to use in practical applications. Tiwari et al. (2015) uses

Specfem 3D Slope, an open source spectral element method (SEM)-based program to

evaluate the stability of large-scale landslides. SEM technique is highly beneficial as it

drastically reduces the huge computational burden if it proceeds for p-refinement

techniques instead of h-refinement unlike FEM. Hence, the SEM technique seemed to be

very accurate and powerful in elasto-plastic modelling which deserves unique

computational scheme in large-scale landslide modelling.

3.2.1.4 Slope stability for unsaturated soil

Slope-stability analysis of unsaturated soil requires to simultaneously compute

deformation and groundwater flow with time-dependent boundary conditions. Very little

work has been found to be done on the 3D slope stability of unsaturated soil. Li et al.

(2006) described the implementation of strength reduction technique method for slope-

stability analysis using Finite Element Method (FEM). Strength reduction FEM can take

into account non-uniform distribution of metric suction and therefore has distinctive
advantages in dealing with 3-D stability of unsaturated soil slope compared to

conventional methods. Strength reduction FEM has been found to be a reliable numerical

approach compared to the stability-analysis of unsaturated slope. Yong et al. (2010)

computed the pore water pressure fields of unsaturated seepage to ascertain the stress and

strain distributions of 3D slope based on the constitutive model of unsaturated soils. The

contribution of matric suction to shear strength is considered and three-dimensional

stability analysis methods for unsaturated soil slopes were performed to evaluate the FOS.

Based on the results of three dimensional unsaturated seepage analysis, the variation rules

of pore water pressure of six observation points are in accordance with the change in

reservoir level, and compared to their changes, the pore water pressure shows some

hysteresis at different degrees. Zhang et al. (2015) carried out a comparative study

between 2D and 3D slope-stability analyses for unsaturated slopes and came to a

conclusion that for simple slopes with low slope angle, ∆Fs/Fs2D monotonically

increases with an increase in the value of c′ and ϕ′, whereas the value does not increase

for a simple steep slope. The difference of FOS between 2D and 3D analysis for a simple

steep slope is found to be larger for a simple slope having lower slope angle. They also

found that the difference between 2D and 3D stability analysis was most pronounced for

concave geometrics.

3.2.1.5 Slope stability by computer-aided programs

Many researchers worked on various computer aided and optimization programs

to study the stability of soils. Xie et al. (2006) and Tiwari and Douglas (2012) used the

GIS grid based 3D models to study the slope-stability analysis. Shen and Karakus (2013)

used the FLAC-3D program to analyze the 3D slope stability. A non-linear shear strength
reduction (SSR) technique was proposed that can use the Hoek–Brown (HB) criterion to

represent the non-linear behaviour of a rock mass in the FLAC-3D program. The result

of the proposed technique found to be very satisfactory. Rashid et al. (2015) presented

particle swarm optimization (PSO) in three dimensional (3D) slope-stability analysis to

determine the shape and direction of failure as the critical slip surface and a factor of

safety (FOS) was developed based on limit equilibrium method. A coding system was

developed in Matlab to work out the 3D form of the failure surface and calculate its FOS.

A 3D slope model beneath the vertical load was finally made and tested within the

laboratory. The results obtained from PSO were re-analyzed and compared with the code

results and it was found that the given codes were highly effective in determining the 3D

failure surface of the soil slopes.

Stark and Eid (1998) found that the commercially offered 3D slope-stability

software doesn’t take into account the shear resistance on the two sides of the sliding

mass. As a result, the 3D factor of safety may be underestimated whereas the back-

calculated shear strength may be overestimated. Arellano and Stark (2000) offered a new

technique for incorporating the shear resistance on the two sides of the sliding mass in

existing 3D software. Huang and Tsai (2000) developed a new 3D slope-stability method

which is based on 2D moment equilibrium method. They found that the new method is

very advantageous as it calculates the factor of safety as well as the possible direction of

sliding for semispherical and composite failure surfaces. Hence, the errors generated from

assuming a plane of symmetry is removed. Again Farzaneh et al.in 2008 based on the

upper bound theorem of limit analysis presented a new three-dimensional slope-stability

analysis for convex slopes. This method has the advantage of calculating both the 3D

factor of safety and the bearing capacity of foundations adjacent to such slopes. On
comparing the results of bearing capacity of foundations, they came to a conclusion that

the one located near the straight slopes has more capacity than the one located near convex

slopes. Zheng (2009) presented a rigorous 3D method that considered the whole failure

body rather than discretizing it into columns. The sliding surface was assumed to possess

a general shape with an arbitrary direction of shear. Zheng considered six equilibrium

conditions for the sliding mass along with a vector of integration equation. The unknown

values of these equations include FOS and total normal stress on the sliding surface that

was defined by a distribution form including five unknowns. Then, the distribution

function was substituted into the mentioned six equations and provided a system of non-

linear equations. The FOS and the distribution vector was found by solving the system of

non-linear equations. Michalowski (2010) approaches kinematically to calculate the 3D

factor of safety. He prepared stability charts using 3D failure mechanism for finding the

factor of safety. These stability charts are very helpful in calculating the factor of safety

as it does not require any iteration. Michalowski continued his research work in the field

of kinematic approach of limit analysis and in 2013, Nadukuru and Michalowski

described a three-dimensional slope-stability analysis applicable to slopes whose

geometry of failure patterns was physically constrained. Gao et al.in 2013 extended a

kinematically based 3D method of slope stability. In addition to toe failure, the extended

method incorporates face failure and base failure in both purely cohesive and frictional

soils. An analytical approach is derived afterwards to obtain the upper bounds on slope

stability and the corresponding type of the critical failure mechanism. The results are then

compared with a finite element analysis method and on comparing the results they found

that the 3D rotational failure mechanisms give the best estimate on the upper bound. Zhou

and Pond (2013) presented a rigorous approach for slot-cut stability analysis that is
applicable for slopes even with complex geometry, stratigraphy and surcharge loading

conditions. They considered the 3D effect by incorporating side-panel shear resistance in

the force limit-equilibrium equations. The result found to be very successful and the

application of slot-cut construction can be readily applied for removal and repair of a

landslide below an existing residence.

3.2.2 3D Finite Element Approach

Finite Element Analysis (FEA) uses the FEM to analyze a material or object and

find how applied stresses will affect the material or design. FEA is basically used in

Mechanical and Civil Engineering fields for analyzing and solving complex geometrical

problems. The methods most commonly used at present for slope-stability analysis are

the rigid-body LEM and the FEM (Bishop, 1955; Duncan, 1996; Griffths and Lane, 1999

and Chen et al., 2005). The former yields a safety factor determined by analyzing the limit

equilibrium status of a block. The method is characterized by simple calculations.

However, LEM cannot take non-linear structural deformation into account and the

method assumes that sliding surfaces reaches an ultimate state of failure simultaneously,

which does not reflect the actual stress status of slip surfaces (Lenchman and Griffths,

2000). FEM can be used to determine the stress field and displacement field of the slope

but cannot yield a specific value for the slope-stability safety factor (Liu et al., 2008).

Although many researchers have obtained slope stability safety factors using the strength

reduction method together with FEA (Jiang and Magnan, 1997; Dawson et al., 1999 and

Zhao et al., 2002). Jeramic (2000) presented a new approach for modelling of three-

dimensional slope-stability problems. A p-version of the FEM together with large

deformation hyper elastic–plastic formulation is utilized to model localized, continuous

deformation that has been observed in failure mechanisms of slopes. In particular, it is

shown how the new method can be used with a rather small number of finite elements to

model sharp deformation gradients resulting from shear localization during slope failures.

Tan and Sharma (2008) developed a new method for both homogenous and non-

homogenous slopes based on limit equilibrium method and used FEM to validate the new

procedure. The procedure was found to be in satisfaction with the FEM. However, some

differences were noticed when non associated flow rules were adopted in FEA. Li and

Shao in 2011 presented a three-dimensional finite element limit equilibrium method

(3DFLEM) based on the concept of strength reduction and the unique sliding direction.

They also clarified the physical meaning of factor of safety as well as the relationship of

FOS and the unique sliding direction. They compared the results of stability analysis

obtained from the proposed approach, 3D rigid limit equilibrium method (3DLEM) and

3D shear strength reduction method (3DSSRM) and found that the FOS and critical

sliding surface are generally in good agreement and the element size of slope and sliding

surface has certain effect on FOS by causing maximum difference of 2%. Nian et al.

(2012) analyzed a 3D slope-stability method using finite element strength reduction

method. They found that the concave-shaped vertical slope with a 90° corner angle is

markedly higher than that of a convex-shaped vertical slope with a 90° corner angle.

Moreover, they also found that a concave-shaped vertical slope with a 90° corner angle

can be replaced by a straight vertical slope for computation of the FOS. Liu et al. (2013)

used the multi-grid method to establish two grids, a structural grid for finite element

computation and a sliding surface grid for calculating a sliding surface’s stability safety

factor. This combination of grids makes it easy to determine the stability safety factor of

any sliding surface or sliding block, and it also considers the influences of non-linear
deformation and elastic–plastic stress adjustment on the stability safety factor. Zhang et

al. (2013) analyzed the effects of complex geometrics on 3D slope stability using an

elastoplastic finite difference method (FDM) with a strength reduction technique. The

results obtained from the analyses were useful for landslide hazard preparedness or safe

and economical design of infrastructures. Kelesoglu (2015) in his paper investigated the

effect of each contributing factor such as the curvature of the slope, the contribution of

the piles and the local loading of the slope by using SRM (Strength Reduction Method)

and FDM. He found that the stability of concave slopes is higher than those straight

slopes. The FOS values are increased up to 15 to 25% for slopes that have sharp concave

curvatures and 5 to 10% for smooth concave curvatures compared to a straight slope. He

also found that when there is local loading on top of the slope, due to the mobilized shear

strains under the surcharge, the pile row must move uphill towards the load to ensure the

local and global stability of the slope. If the surcharge is next to the crest (b = 0 m), then

the effective pile location is adjacent to the slope crest. If the surcharge moves away from

the crest (b > 0 m) the effective pile location is located between the no surcharge case and

the crest of the slope. In this case, if the pile row is located within these boundaries then

the FOS values differs only marginally.

Slope-stability analysis is an extremely important as the result of slope failure can

often be catastrophic, involving considerable loss of life and property. Hence, slope

instability is widely recognized as an ever-present danger and it is a continual source of

concern for geotechnical engineers and engineering geologists throughout the world.

Different methods of slope-stability analysis have been developed in the past and each of

them has advantages and limitations over the others. In the present study, literature review

is carried out on the three-dimensional slope-stability analysis based on two approaches–

limit equilibrium approach and finite element approach. The literature study of the limit

equilibrium approach reveals that most of the 3D slope stability LEMs are derived from

2D LEMs. The assumptions of the 3D slope-stability methods were kept similar to the

2D methods but the 3D FOS is found to be higher than the 2D FOS. Many literatures

were studied where the 3D FOS was determined for cohesive soil and surcharge loading

conditions. It was found that for both the cases the 3D FOS was found to be higher than

the 2D FOS. Very little work has been done on 3D seismic slope stability and 3D slope

stability for unsaturated soil. In majority of the methods, charts have been prepared to

calculate the FOS under seismic conditions which is found to be very helpful as it

eliminates the calculation of iterative method. The literature study of the FEM approach

reveals that shear strength reduction technique is mostly employed to analyze the stability

of slopes. Studies were also found to be done based on 3DFLEM, multi grid method,

elastoplastic finite difference method, etc. It is found that less attention is given towards

the delayed and/or time-dependent behaviour of slopes; hence, further research work

needs to be done to know the time-dependent behaviour of slopes. More emphasis have

to be placed on 3D slope stability for seismic and unsaturated soil conditions.

3.3 THREE DIMENSIONAL (3D) DESIGN CHARTS FOR THE

DETERMINATION OF FACTOR OF SAFETY (FOS)

Charts representing the slope stability analyses results require an iterative

procedure to arrive at the safety factor. Taylor’s (1937) stability charts were given in

terms of stability factor c/γHF, where c is the cohesion intercept needed to maintain limit

equilibrium and H is the height of the slope. The factor c/γHF is plotted against slope

inclinations β for a variety of friction angle φ. If slope angle β and friction angle φ for a

slope is given, one can easily find the value of F by calculating the value of c/γH (given
slope) from Taylor’s Stability Chart. But the safety factor needs to be applied to tan φ

also.

c tan φ
F= = …… (3.1)
c tan φ

Hence, the process becomes iterative. Taylor’s stability chart fails to define the location

of the slip circle. This problem was solved by Baker (2003) where he proposed design

charts that can compute the coordinates, centre and radius of the slip circle. Steward et al.

(2011) revisited Taylor’s stability charts and two modified design charts were presented,

one for undrained (φu = 0) clays and other for drained (c΄-φ΄) soils. The chart for

undrained clays consists of compound circles having two separate arcs connected by a

straight line at the interface with the stiff stratum which were not present in Taylor’s chart.

The chart for drained soils enables to compute the slope safety factor without any iterative

procedure. Michalowski (2002) produced a set of stability charts for slopes based on the

kinematic approach of limit analysis. These charts have the advantage of using them for

slopes subjected to pore water pressure and seismic forces. Michalowski (2010) extended

his 2D work to 3D and presented stability charts which do not require an iterative

procedure to arrive at the factor of safety. A 3D rotational failure mechanism constructed

by Michalowski and Drescher (2009) with the failure surface being a section of a

curvilinear cone (horn shaped) was adopted to develop the stability charts.

The work discussed herein required thousands of analyses of slopes on a micro-

computer having different geometry and soil parameters. All the analyses has been carried

out using a finite element software, PLAXIS 3D. Similar to Sec. 2.4, the software is first

validated by comparing the results with the one obtained from Abdelaziz et al. (2017).

On successfully validating the software, 3500 artificial slopes with different geometry

and soil parameters is studied and analyzed using PLAXIS 3D. For the analysis, a 15
nodal triangular element with very fine mesh is chosen and Mohr-Coulomb failure

criterion is considered for modelling the soil behaviour. The FOS values obtained from

PLAXIS 3D for a typical slope is shown in Fig. 3.1 (a), (b) and (c). In a similar fashion,

the remaining 3499 artificial slopes are analyzed using the FE software. Table 3.1 shows

only a part of 3500 artificial slope cases for FOS determination.

(a)

Soil Properties

C = 45 KN/m2; φ = 30°; γ = 20

(b)
 (c)

(d)

Figure 3.1: A Typical Example of Slope Stability Analysis Using PLAXIS 3D: (a)

Development of Soil Model; (b) 2D Finite Element Mesh Generation; (c) 3D Finite

Element Mesh Generation and (d) Active Pore Pressure Generation.

 Table 3.1. FOS for 200 Artificial Slope Cases for 3D Slope Stability Analysis
(Values of H, c, φ, β, γ and h are same as has been taken for 2D analysis, Table 2.1,
and all these values are arbitrarily chosen)
Angle Depth
Height Slope Unit
of Thickness of
of the Cohesion Inclin Weight
Sl. Internal of the water m' = FOS
slope (kN/m2) ation of Soil
No. friction slope (m) table h/B
(m) (°) (kN/m3)
(°) (m)
H c φ β γ B h F
1 15 10 35 30 18 7.5 3 0.4 1.793
2 15 0 37 30 17 7.5 0 0.0 1.640
3 9 5 35 30 17 4.5 1 0.2 1.807
4 10 10 35 30 18 5 3 0.6 1.952
5 11 17 26 30 18 5.5 3 0.5 1.755
6 12 30 26 30 20 6 1 0.2 2.125
7 16 35 19 30 16 8 0 0.0 1.943
8 8 35 25 30 18 4 0 0.0 2.966
9 12 32 10 30 20 6 4 0.7 1.322
10 15 32 10 30 20 7.5 7 0.9 1.050
11 10 35 24 30 18 5 2 0.4 2.302
12 16 32 10 30 20 8 3 0.4 1.113
13 15 32 10 30 20 7.5 0 0.0 1.177
14 15 32 10 30 20 7.5 3 0.4 1.522
15 16 55 25 30 16 8 9 1.1 2.265
16 16 37 22 30 16 8 5 0.6 1.305
17 16 33 12 30 17 8 0 0.0 1.442
18 10 0 39 30 17.5 5 0 0.0 2.113
19 19 0 31.5 30 18 9.5 0 0.0 1.060
20 22 0 36.5 30 17.8 11 0 0.0 1.072
21 13 17 31 30 17.7 6.5 10 1.5 1.252
22 13 14 31 30 17 6.5 0 0.0 1.970
23 20 27 31 30 18.9 10 6 0.6 1.883
24 20 22.5 27.5 30 19.5 10 14 1.4 1.158
25 11 33 36 30 17.8 5.5 2 0.4 2.996
26 11 26.8 31 30 18.5 5.5 0 0.0 2.548
27 13 55.5 22 30 16.6 6.5 0 0.0 2.910
28 18 49 25 30 17.2 9 5 0.6 2.013
29 18 57 37 30 18.7 9 0 0.0 3.125
30 14 45 25 30 19.2 7 4 0.6 2.359
31 15 10 35 30 18 15 3 0.2 1.795
32 15 0 37 30 17 15 0 0.0 1.617
33 9 5 35 30 17 9 1 0.1 1.815
34 10 10 35 30 18 10 3 0.3 1.934
35 11 17 26 30 18 11 3 0.3 1.748
36 12 30 26 30 20 12 1 0.1 2.123
37 16 35 19 30 16 16 0 0.0 1.946
 Table 3.1. FOS for 200 Artificial Slope Cases for 3D Slope Stability Analysis
(Values of H, c, φ, β, γ and h are same as has been taken for 2D analysis, Table 2.1,
and all these values are arbitrarily chosen) (Contd.)
Angle Depth
Height Slope Unit
of Thickness of
of the Cohesion Inclin Weight
Sl. Internal of the water m' = FOS
slope (kN/m2) ation of Soil
No. friction slope (m) table h/B
(m) (°) (kN/m3)
(°) (m)
H c φ β γ B h F
38 8 35 25 30 18 8 0 0.0 2.967
39 12 32 10 30 20 12 4 0.3 1.322
40 15 32 10 30 20 15 7 0.5 1.048
41 10 35 24 30 18 10 2 0.2 2.304
42 16 32 10 30 20 16 3 0.2 1.111
43 15 32 10 30 20 15 0 0.0 1.177
44 15 32 10 30 20 15 3 0.2 1.153
45 16 55 25 30 16 16 9 0.6 2.266
46 16 37 22 30 16 16 5 0.3 1.907
47 16 33 12 30 17 16 0 0.0 1.444
48 10 0 39 30 17.5 10 0 0.0 2.098
49 19 0 31.5 30 18 19 0 0.0 1.031
50 22 0 36.5 30 17.8 22 0 0.0 1.039
51 13 17 31 30 17.7 13 10 0.8 1.258
52 13 14 31 30 17 13 0 0.0 1.971
53 20 27 31 30 18.9 20 6 0.3 1.882
54 20 22.5 27.5 30 19.5 20 14 0.7 1.163
55 11 33 36 30 17.8 11 2 0.2 2.993
56 11 26.8 31 30 18.5 11 0 0.0 2.551
57 13 55.5 22 30 16.6 13 0 0.0 2.911
58 18 49 25 30 17.2 18 5 0.3 2.013
59 18 57 37 30 18.7 18 0 0.0 3.129
60 14 45 25 30 19.2 14 4 0.3 2.351
61 15 10 35 30 18 22.5 3 0.1 1.792
62 15 0 37 30 17 22.5 0 0.0 1.595
63 9 5 35 30 17 13.5 1 0.1 1.814
64 10 10 35 30 18 15 3 0.2 1.916
65 11 17 26 30 18 16.5 3 0.2 1.752
66 12 30 26 30 20 18 1 0.1 2.124
67 16 35 19 30 16 24 0 0.0 1.946
68 8 35 25 30 18 12 0 0.0 2.967
69 12 32 10 30 20 18 4 0.2 1.325
70 15 32 10 30 20 22.5 7 0.3 1.048
71 10 35 24 30 18 15 2 0.1 2.304
72 16 32 10 30 20 24 3 0.1 1.112
73 15 32 10 30 20 22.5 0 0.0 1.177
 Table 3.1. FOS for 200 Artificial Slope Cases for 3D Slope Stability Analysis
(Values of H, c, φ, β, γ and h are same as has been taken for 2D analysis, Table 2.1,
and all these values are arbitrarily chosen) (Contd.)
Angle Depth
Height Slope Unit
of Thickness of
of the Cohesion Inclin Weight
Sl. Internal of the water m' = FOS
slope (kN/m2) ation of Soil
No. friction slope (m) table h/B
(m) (°) (kN/m3)
(°) (m)
H C φ β γ B h F
74 15 32 10 30 20 22.5 3 0.1 1.153
75 16 55 25 30 16 24 9 0.4 2.265
76 16 37 22 30 16 24 5 0.2 1.907
77 16 33 12 30 17 24 0 0.0 1.444
78 10 0 39 30 17.5 15 0 0.0 2.083
79 19 0 31.5 30 18 28.5 0 0.0 1.003
80 22 0 36.5 30 17.8 33 0 0.0 1.006
81 13 17 31 30 17.7 19.5 10 0.5 1.262
82 13 14 31 30 17 19.5 0 0.0 1.974
83 20 27 31 30 18.9 30 6 0.2 1.885
84 20 22.5 27.5 30 19.5 30 14 0.5 1.163
85 11 33 36 30 17.8 16.5 2 0.1 2.993
86 11 26.8 31 30 18.5 16.5 0 0.0 2.554
87 13 55.5 22 30 16.6 19.5 0 0.0 2.911
88 18 49 25 30 17.2 27 5 0.2 2.013
89 18 57 37 30 18.7 27 0 0.0 3.129
90 14 45 25 30 19.2 21 4 0.2 2.351
91 15 10 35 30 18 30 3 0.1 1.790
92 15 0 37 30 17 30 0 0.0 1.572
93 9 5 35 30 17 18 1 0.1 1.781
94 10 10 35 30 18 20 3 0.2 1.919
95 11 17 26 30 18 22 3 0.1 1.753
96 12 30 26 30 20 24 1 0.0 2.121
97 16 35 19 30 16 32 0 0.0 1.946
98 8 35 25 30 18 16 0 0.0 2.967
99 12 32 10 30 20 24 4 0.2 1.324
100 15 32 10 30 20 30 7 0.2 1.048
101 10 35 24 30 18 20 2 0.1 2.306
102 16 32 10 30 20 32 3 0.1 1.113
103 15 32 10 30 20 30 0 0.0 1.178
104 15 32 10 30 20 30 3 0.1 1.153
105 16 55 25 30 16 32 9 0.3 2.265
106 16 37 22 30 16 32 5 0.2 1.909
107 16 33 12 30 17 32 0 0.0 1.450
108 10 0 39 30 17.5 20 0 0.0 2.068
109 19 0 31.5 30 18 38 0 0.0 0.974
 Table 3.1. FOS for 200 Artificial Slope Cases for 3D Slope Stability Analysis
(Values of H, c, φ, β, γ and h are same as has been taken for 2D analysis, Table 2.1,
and all these values are arbitrarily chosen) (Contd.)
Angle Depth
Height Slope Unit
of Thickness of
of the Cohesion Inclin Weight
Sl. Internal of the water m' = FOS
slope (kN/m2) ation of Soil
No. friction slope (m) table h/B
(m) (°) (kN/m3)
(°) (m)
H c φ β γ B h F
110 22 0 36.5 30 17.8 44 0 0.0 0.973
111 13 17 31 30 17.7 26 10 0.4 1.262
112 13 14 31 30 17 26 0 0.0 1.971
113 20 27 31 30 18.9 40 6 0.2 1.883
114 20 22.5 27.5 30 19.5 40 14 0.4 1.164
115 11 33 36 30 17.8 22 2 0.1 2.994
116 11 26.8 31 30 18.5 22 0 0.0 2.555
117 13 55.5 22 30 16.6 26 0 0.0 2.911
118 18 49 25 30 17.2 36 5 0.1 2.010
119 18 57 37 30 18.7 36 0 0.0 3.113
120 14 45 25 30 19.2 28 4 0.1 2.353
121 15 10 35 30 18 37.5 3 0.1 1.790
122 15 0 37 30 17 37.5 0 0.0 1.550
123 9 5 35 30 17 22.5 1 0.0 1.810
124 10 10 35 30 18 25 3 0.1 1.928
125 11 17 26 30 18 27.5 3 0.1 1.754
126 12 30 26 30 20 30 1 0.0 2.120
127 16 35 19 30 16 40 0 0.0 1.946
128 8 35 25 30 18 20 0 0.0 2.967
129 12 32 10 30 20 30 4 0.1 1.324
130 15 32 10 30 20 37.5 7 0.2 1.048
131 10 35 24 30 18 25 2 0.1 2.306
132 16 32 10 30 20 40 3 0.1 1.113
133 15 32 10 30 20 37.5 0 0.0 1.178
134 15 32 10 30 20 37.5 3 0.1 1.153
135 16 55 25 30 16 40 9 0.2 2.267
136 16 37 22 30 16 40 5 0.1 1.909
137 16 33 12 30 17 40 0 0.0 1.450
138 10 0 39 30 17.5 25 0 0.0 2.053
139 19 0 31.5 30 18 47.5 0 0.0 0.946
140 22 0 36.5 30 17.8 55 0 0.0 0.940
141 13 17 31 30 17.7 32.5 10 0.3 1.264
142 13 14 31 30 17 32.5 0 0.0 1.976
143 20 27 31 30 18.9 50 6 0.1 1.883
144 20 22.5 27.5 30 19.5 50 14 0.3 1.164
145 11 33 36 30 17.8 27.5 2 0.1 2.994
 Table 3.1. FOS for 200 Artificial Slope Cases for 3D Slope Stability Analysis
(Values of H, c, φ, β, γ and h are same as has been taken for 2D analysis, Table 2.1,
and all these values are arbitrarily chosen) (Contd.)
Angle Depth
Height Slope Unit
of Thickness of
of the Cohesion Inclin Weight
Sl. Internal of the water m' = FOS
slope (kN/m2) ation of Soil
No. friction slope (m) table h/B
(m) (°) (kN/m3)
(°) (m)
H c φ β γ B h F
146 11 26.8 31 30 18.5 27.5 0 0.0 2.555
147 13 55.5 22 30 16.6 32.5 0 0.0 2.911
148 18 49 25 30 17.2 45 5 0.1 2.010
149 18 57 37 30 18.7 45 0 0.0 3.132
150 14 45 25 30 19.2 35 4 0.1 2.353
151 15 10 35 30 18 45 3 0.1 1.790
152 15 0 37 30 17 45 0 0.0 1.527
153 9 5 35 30 17 27 1 0.0 1.810
154 10 10 35 30 18 30 3 0.1 1.928
155 11 17 26 30 18 33 3 0.1 1.754
156 12 30 26 30 20 36 1 0.0 2.120
157 16 35 19 30 16 48 0 0.0 1.946
158 8 35 25 30 18 24 0 0.0 2.972
159 12 32 10 30 20 36 4 0.1 1.324
160 15 32 10 30 20 45 7 0.2 1.048
161 10 35 24 30 18 30 2 0.1 2.306
162 16 32 10 30 20 48 3 0.1 1.114
163 15 32 10 30 20 45 0 0.0 1.178
164 15 32 10 30 20 45 3 0.1 1.153
165 16 55 25 30 16 48 9 0.2 2.267
166 16 37 22 30 16 48 5 0.1 1.909
167 16 33 12 30 17 48 0 0.0 1.450
168 10 0 39 30 17.5 30 0 0.0 2.038
169 19 0 31.5 30 18 57 0 0.0 0.917
170 22 0 36.5 30 17.8 66 0 0.0 0.907
171 13 17 31 30 17.7 39 10 0.3 1.264
172 13 14 31 30 17 39 0 0.0 1.976
173 20 27 31 30 18.9 60 6 0.1 1.884
174 20 22.5 27.5 30 19.5 60 14 0.2 1.163
175 11 33 36 30 17.8 33 2 0.1 2.994
176 11 26.8 31 30 18.5 33 0 0.0 2.256
177 13 55.5 22 30 16.6 39 0 0.0 2.910
178 18 49 25 30 17.2 54 5 0.1 2.023
179 18 57 37 30 18.7 54 0 0.0 3.113
180 14 45 25 30 19.2 42 4 0.1 2.353
181 15 10 35 30 18 52.5 3 0.1 1.790
 Table 3.1. FOS for 200 Artificial Slope Cases for 3D Slope Stability Analysis
(Values of H, c, φ, β, γ and h are same as has been taken for 2D analysis, Table 2.1,
and all these values are arbitrarily chosen) (Contd.)
Angle Depth
Height Slope Unit
of Thickness of
of the Cohesion Inclin Weight
Sl. Internal of the water m' = FOS
slope (kN/m2) ation of Soil
No. friction slope (m) table h/B
(m) (°) (kN/m3)
(°) (m)
H c φ β γ B h F
182 15 0 37 30 17 52.5 0 0.0 1.505
183 9 5 35 30 17 31.5 1 0.0 1.809
184 10 10 35 30 18 35 3 0.1 1.928
185 11 17 26 30 18 38.5 3 0.1 1.754
186 12 30 26 30 20 42 1 0.0 2.120
187 16 35 19 30 16 56 0 0.0 1.944
188 8 35 25 30 18 28 0 0.0 2.967
189 12 32 10 30 20 42 4 0.1 1.324
190 15 32 10 30 20 52.5 7 0.1 1.048
191 10 35 24 30 18 35 2 0.1 2.304
192 16 32 10 30 20 56 3 0.1 1.114
193 15 32 10 30 20 52.5 0 0.0 1.177
194 15 32 10 30 20 52.5 3 0.1 1.154
195 16 55 25 30 16 56 9 0.2 2.268
196 16 37 22 30 16 56 5 0.1 1.912
197 16 33 12 30 17 56 0 0.0 1.446
198 10 0 39 30 17.5 35 0 0.0 2.023
199 19 0 31.5 30 18 66.5 0 0.0 0.889
200 22 0 36.5 30 17.8 77 0 0.0 0.874

The process of iterations will be eliminated if the results are plotted as a function

of c cot φ. This is obvious that c cot φ is independent of the safety factor as from the

definition and illustrated in Fig. 3.2. No matter what the safety factor, product c cot φ will

always remain the same. It is hence necessary to plot the reciprocal of tan φ, Fm'/tan(φ)

versus the dimensionless parameter cm'/γBtan(φ) and produce stability charts for

different values of B/H at different slope inclination. Here, c is the cohesion intercept, H

is the height of the slope, B is the width of the slope, γ is the in-situ unit weight of the

soil, φ is the angle of internal friction and m' is a dimensionless parameter which is given

by ratio of water table depth (h) and the width of the slope (B). The water table depth (h)
is an alternative quantity for the active pore pressure. The pore pressure at a depth, z,

below the surface is given by:

p = γ (z − h) …… (3.2)

Where

h is the depth of water table, p is the active pore pressure (i.e. steady-state pore pressures

+ excess pore pressures) and γw is the unit weight of water.

Figure 3.2: Independence of Parameter c cot of the Safety Factor

3.3.1 Stability Charts for φu = 0 soil

The computational results are first presented for the undrained failure of the soil

(φu = 0) as shown in Fig. 3.3. The non-dimensional parameter (cum')/γBF is presented as

a function of slope inclination angle β for different ratios of B/H. The use of this chart is

very straight forward. If the slope geometry and hydrological condition is known, one can

easily calculate the value of cum'/γBF from which the FOS can be found out.
3.3.2 Stability Charts for c – φ soil

The remaining computational results are represented for soil strength

characterized by c and φ. Here, the results are presented as a function of Fm'/tan(φ) versus

cm'/γBtan(φ) for width to height ratios (B/H) of the failing slopes ranging from 0.5 to 4.0.

Each chart illustrates results of one inclination angle of the slope. The results of slopes of

inclination 30° and 35° are shown in Fig. 3.4 and Fig. 3.5 respectively, 40° and 45° are

shown in Fig. 3.6 and Fig. 3.7 respectively, 50° and 55° are shown in Fig. 3.8 and Fig.

3.9 respectively, 60° and 65° are shown in Fig. 3.10 and Fig. 3.11 respectively and 70° is

shown in Fig. 3.12. Once the B/H values of the mechanism exceed 4.0, the ratio of 3D

FOS to 2D FOS approaches to unity (≈ 1). However, this ratio exceeds beyond unity

(>>1) as the value of B/H lowers to 0.5.

Figure 3.3: Stability Charts for φu = 0 soil

Figure 3.4: Stability Chart for c – φ soil for Slope Angle β = 30°

Figure 3.5: Stability Chart for c – φ soil for Slope Angle β = 35°
Figure 3.6: Stability Chart for c – φ soil for Slope Angle β = 40°

Figure 3.7: Stability Chart for c – φ soil for Slope Angle β = 45°
Figure 3.8: Stability Chart for c – φ soil for Slope Angle β = 50°

Figure 3.9: Stability Chart for c – φ soil for Slope Angle β = 55°
Figure 3.10: Stability Chart for c – φ soil for Slope Angle β = 60°

Figure 3.11: Stability Chart for c – φ soil for Slope Angle β = 65°
 Figure 3.12: Stability Chart for c – φ soil for Slope Angle β = 70°

3.4 PROPOSED PREDICTION MODELS FOR 3D SLOPE STABILITY

With the introduction of artificial neural network (ANN) and multiple linear

regression (MLR), engineers and researchers from a variety of disciplines are

encouraging researches using these applications. The growing interest among the

researchers is due to the excellent performance provided by these learning machines in

pattern recognition and modelling of nonlinear multivariate dynamic systems. The

accurate estimation of the soil stabilization is a very challenging task for the geotechnical

engineers due to the intricacy and difficulty in determining the geotechnical input data

parameters. The slope stability analysis must be carried out by considering the various

important parameters like site sub-surface conditions, ground behaviour, applied loads,

etc. It is due to its practical importance that slope stability analysis has drawn the attention

of many investigators. This chapter deals with the extension of the prediction models of
slope stability analysis discussed in Chapter 2 by incorporating and modifying some of

the stability parameters so that the prediction models can predict the 3D FOS of the

slopes. For this, 3500 artificial slopes are studied using a finite element software PLAXIS

3D, by varying the geometrical and geotechnical parameters. The obtained FOS values

are used to develop the prediction models using MLR and ANN. The height of the slope

(H), cohesion (c), angle of internal friction (φ), angle of the slope (β), unit weight of soil

(γ) and dimensionless parameter (m′) are used as input parameters whereas the FOS as

the output parameter. The dimensionless parameter, ‘m′’ is defined as the ratio between

the water table depth (h) and the width of the slope (B).

3.4.1 Proposed Multiple Linear Regression (MLR) Model

The MLR model for predicting the FOS has been developed using Microsoft

Excel 2013. The summary of MLR for 3500 artificial slope cases is given in Table 3.2.

Table 3.2. Summary of MLR for 3500 Cases

Summary output of MLR

Regression Statistics
Multiple R 0.943
R Square 0.889
Adjusted R Square 0.889
Standard Error 0.168
Observations 3500
Stability Parameters Coefficients Standard Error t Stat P-value
Intercept 2.522 0.046 54.260 0.000
H -0.073 0.001 -83.049 0.000
c 0.023 0.000 118.896 0.000
φ 0.034 0.000 103.411 0.000
β -0.018 0.000 -83.689 0.000
γ -0.034 0.002 -15.032 0.000
m′ = h/B -0.259 0.014 -19.115 0.000
From Table 3.2, it has been found that the stability parameters are having a correlation

coefficient, R2 of 0.89, thereby, bearing a close relationship between them.

3.4.2 Proposed Artificial Neural Network (ANN) Model

The ANN model has been prepared in Matlab R2011a. Here also, multi-layer

feed-forward network having 20 neurons in hidden layer and 1 neuron in output layer is

used for developing the prediction model which is shown in Fig. 3.13.

Figure 3.13: Neural Network Showing Hidden Neurons for 3D Prediction Model

The network is developed by taking 3500 artificial slope cases having different

geometrical and geotechnical parameters. For cross validation technique, 80% of the data

set has been used for training and the remaining for validation of the model. The

validation performance of the network model and regression plot showing the value of R 2

for training, testing and validation is shown in Fig. 3.14 and 3.15 respectively. From the

regression plot, it has been found that the value of R2 has been found to be 0.99 which is

very close to unity. Hence, it can be stated that the prediction bear a close relationship

between the input variables.

Figure 3.14. Validation Performance for 3500 Slope Cases

Overall R2 = 0.993

Figure 3.15. Regression plot for 3500 Slope Cases

3.4.3 Performance of the Proposed Model and Design charts

The performance of the predicted models and design charts are checked by

examining the results by making predictions against case records which are not used

during training and testing. 40 vulnerable slope (Refer Table 2.3 presented in Chapter 2)

cases around Guwahati and its adjoining areas having different latitude and longitude are

selected. Laboratory tests are conducted to determine the geotechnical parameters. Total

station survey has been conducted to plot the contour map of the slope using Teraplot LT.

From the contour map different geometrical parameters are determined. These

geometrical and geotechnical parameters are used to determine the critical safety factor

of slopes and a comparison is made using the results of analytical, MLR, ANN and design

charts as shown in Table 3.3. It is evident from Fig. 3.16 that the results from design

charts is found to have highest correlation of over 92% followed by ANN and MLR

having 90% and 82% respectively. Hence, it can be said that the design charts can give

higher correlation compared to the other prediction models.

100
95
Correlation %

90
85
80
75
70
MLR ANN Design Charts
Prediction Methods

Figure 3.16. Correlation percentage by MLR, ANN and Design Charts for 3D
Slope Stability Analysis
The stability of the prediction models and design charts are further checked for error

analysis. The error analysis can be performed by computing RMSE and MAE. It can be

observed from Fig. 3.17 that RMSE and MAE values are found to be the lowest for

outcomes obtained from design charts compared to ANN and MLR and hence it can be

concluded that the design charts are able to predict the target values with higher degree

of accuracy.

25

20
Error %

15

10

0
MLR ANN Design Charts
Prediction Methods

Figure 3.17. Variation of error percentage for MLR, ANN and Design Charts for
3D Slope Stability Analysis
 Table 3.3. Case Study for 40 Vulnerable Slopes from Guwahati and its Adjoining Areas for 3D Slope Stability
Unit Depth of
Angle of
Slope weight Slope Slope water
Cohesion internal Dimensionless
Height of the Inclination Width table 3D-FOS
(kN/m2) friction parameter
(m) soil (°) (m) from G.L.
Sl. No. Site Name (°)
(kN/m3) (m)
From
From From From
H c φ γ β mˊ = h/B B h Design
FEM MLR ANN
Charts
1 Jorabat Site 1 20 39.5 30.2 17.6 50 0.8 20 30 0.986 0.976 1.006 0.986
2 Jorabat Site 2 23 39 30 17.3 50 0.8 20 28 1.038 0.994 1.012 0.954
3 Jorabat Site 3 18 38.7 30.5 17.8 60 0.8 22 21 0.971 0.971 1.113 0.948
4 Jorabat Site 4 25 39 31.2 17.9 55 0.5 20 13 1.010 0.920 0.983 1.013
5 Jorabat Site 5 18 39 30 17.3 50 0.5 15 12 1.101 0.935 1.076 1.015
6 Jorabat Site 6 22 39 30 17.3 50 0.4 33 12 1.185 0.795 1.178 1.066
7 Jorabat Site 7 29 38.5 30.7 17.5 48 0.3 18 10 0.986 0.790 1.052 0.995
8 Jorabat Site 8 26 37.9 30 17.3 45 0.4 22 11 1.210 1.100 1.007 1.194
9 Jorabat Site 9 20 38.5 29 17.5 50 0.5 17 16 1.123 1.076 1.252 1.127
10 Jorabat Site 10 20 39.2 29.7 17.5 55 0.8 30 26 1.210 1.179 1.167 1.224
11 Byrnihat Site 1 21 39 30 17.5 50 0.7 16 21 1.326 1.276 1.287 1.319
12 Byrnihat Site 2 30 39.8 31.3 17.8 45 0.4 20 11 1.257 1.333 1.214 1.289
13 Byrnihat Site 3 20 39 30 17.3 48 0.8 30 23 1.355 0.595 1.298 1.332
14 Byrnihat Site 4 23 39 30 17.3 53 0.7 17 22 1.289 1.259 1.235 1.232
15 Byrnihat Site 5 18 57 39 18 50 0.7 26 21 1.260 1.140 1.052 1.263
16 Byrnihat Site 6 25 57 39 18 52 0.2 30 6 1.700 1.721 1.810 1.704
17 Byrnihat Site 7 31 57.2 38.6 18.3 38 0.3 16 9 1.934 1.509 1.897 1.945
18 Byrnihat Site 8 22 57.5 41.3 19.8 61 0.5 24 12 1.543 1.599 1.542 1.527
19 Byrnihat Site 9 24 57.5 41.3 19.8 62 0.4 25 9 1.680 1.677 1.680 1.673
20 Byrnihat Site 10 15 13 41.9 18.5 55 0.6 25 13 0.987 0.864 0.989 0.967
21 Byrnihat Site 11 14 13 41.9 18.7 50 0.5 30 8 1.440 1.278 1.321 1.438
22 Byrnihat Site 12 15 14 42 18.6 65 0.4 30 6 1.100 1.276 1.257 1.107
23 Byrnihat Site 13 15 13.7 42.2 18.5 62 0.4 25 7 1.210 1.176 1.299 1.179
24 Umling Site 1 12 13 41.9 18.5 45 0.4 24 4 1.535 1.838 1.623 1.533
 Table 3.3. Case Study for 40 Vulnerable Slopes from Guwahati and its Adjoining Areas for 3D Slope Stability (Contd.)
Unit Depth of
Angle of
Slope weight Slope Slope water
Cohesion internal Dimensionless
Height of the Inclination Width table 3D-FOS
(kN/m2) friction parameter
(m) soil (°) (m) from G.L.
Sl. No. Site Name (°)
(kN/m3) (m)
From
From From From
H c φ γ β mˊ = h/B B h Design
FEM MLR ANN
Charts
25 Umling Site 2 16 13 41.9 18.5 52 0.3 28 5 1.307 1.432 1.523 1.301
26 Umling Site 3 14 13.5 42 18.5 40 0.5 30 9 1.743 1.474 1.646 1.604
27 Umling Site 4 17 26 42.2 20.4 50 0.6 35 18 1.387 1.227 1.620 1.273
28 Umling Site 5 20 25.5 42 20.5 50 0.6 35 18 1.260 1.243 1.589 1.266
29 Umling Site 6 15 25.5 42.2 20.4 40 0.5 36 15 1.093 0.948 1.455 1.094
30 Umling Site 7 20 25.5 42 20.7 45 0.7 20 22 1.110 1.078 1.248 1.138
31 Dhirenpara Site 1 15 15 35 18 60 0.2 15 3 1.154 1.218 1.297 1.155
32 Dhirenpara Site 2 18 18 35 17.9 60 0.3 54 5 1.016 1.051 1.027 1.021
33 Hengerabari Site 1 8 35 25 18 65 0.5 12 4 1.743 1.682 1.741 1.725
34 Hengerabari Site 2 11 48 22 18.5 45 0.5 16.5 6 2.017 1.991 2.154 2.148
35 Sunsali Site 1 18 35 24 18 70 0.2 30 2 1.495 1.489 1.381 1.492
36 Sunsali Site 2 17 0 37.5 18 45 0.1 34 2 1.105 1.104 1.111 1.109
37 Sunsali Site 3 28 46 15 18.7 60 0.1 80 2 1.047 0.888 0.997 1.072
38 Kharguli Site 1 18 36 0 18 50 0.1 24 1 1.272 1.222 1.267 0.677
39 Kharguli Site 2 25 47.8 0 18 35 0.2 30 3 1.020 1.233 1.120 0.805
40 Kharguli Site 3 20 27 22 17.8 40 0.3 60 6 1.253 1.028 1.227 1.212
3.5 COMPARISON OF 2D AND 3D SLOPE STABILITY

Slope analyses using 2D and 3D slope configurations has resulted into many

controversies over the last four decades showing both the importance and the difficulties

to handle the subject. The differences in the obtained results is due to the limited amount

of studies and limited cases handles in each study. Most of the studies done previously

were aimed at providing either example to validate the proposed method or to study

certain limitations of the proposed method. However, none of the studies addressed the

practical use and the limitations of the methods from the point of view of the engineers

and practitioners, and hence limited the use of such methods to the judgments of the user.

Therefore, it is important to study the efficacy of such methods and to compare the

available tools that are currently in use by engineers to help in bridging the gap between

the research community and the practitioners.

Since early 1960’s, different studies led to different results regarding the

difference between the 2D and 3D failure. In some cases, the results showed higher 2D

FOS compared to 3D (Baligh and Azzouz, 1975; Giger and Krizek, 1975; Leshchinsky

et al., 1985; Gens et al., 1988 and Leshchinsky and Huang, 1992) while in some cases, it

showed the opposite (Hovland, 1977; Chen and Chameau, 1983 and Seed et al. 1990).

The differences in studies is due to the difference in the methods that is used in each case.

In general, the 3D analysis was carried out for the most critical section from the 2D

analysis due to their simplicity in calculating the most critical slip surface using 2D

analysis. Therefore, the generated critical slip surface did not necessarily match the most

critical slip surface that can be generated from 3D analysis or field observations if a finite

width of the slope is considered. On the other hand, most of the 3D methods were derived

from the existing 2D methods and therefore carrying the limitations of the 2D method to
3D method. For example, Hovland’s method (1977) was an extension of the Ordinary

Method of Slices and therefore led to erroneous results in 3D analysis because normal

stresses on vertical surfaces were assumed to be zero and therefore the computed 3D FOS

were less than the 2D FOS. On the other hand, the high complex nature involved in the

extension of Spencer’s method by Chen and Chameau (1983) resulted in erroneous results

which found higher 2D FOS than 3D (Ugai, 1988).

Therefore, it can be concluded that, in general, slope stability methods that

produce higher FOS in 3D analysis than those in 2D analysis are more accurate while

those that produce the opposite are inaccurate due to the simplifying assumptions or the

erroneous derivations or extension of the methods.

In addition, some studies displayed a constant ratio between the 3D and 2D FOS.

Azzouz et al. (1981) submitted that the ratio between 3D and 2D FOS for slopes in

undrained cohesive soils using the Extended Swedish Circle was between 1.07 and 1.30.

Such results should not be adapted to other general cases due to the limited number of

cases that led to such conclusion, which in Azzouz et al.’s (1981) research was four. On

the other hand, 3D back calculated shear strength would be higher than 2D (Azzouz et

al., 1981 and Leshchinsky and Huang, 1992). Slopes in homogeneous cohesionless slopes

are expected to have shallow failure surfaces that are parallel to the surface of the slope

resulting in the same factor of safety regardless of the analysis dimension (3D or 2D)

(Hutchinson and Sharma, 1985 and Leshchinsky and Baker, 1986).

In this research, a few case studies have been done to have a comparison between

the 2D and 3D FOS using FEM shown in Table 3.4. The result shows that the ratio of 3D

FOS to 2D FOS is found to be greater than unity which implies that the 3D FOS is higher

than the 2D FOS.

 Table 3.4: Comparison between 2D and 3D FOS for 40 vulnerable sites from Guwahati and its Adjoining Areas
Slope Cohesion Angle of Unit Slope Slope Depth FOS
Height (kN/m2) internal weight of Inclination Width of
Sl. No. Site Name (m) friction (°) the soil (°) (m) water
(kN/m3) table
H c φ γ β B h 3D 2D 3D/2D
1 Jorabat Site 1 20 39.5 30.2 17.6 50 20 30 0.986 0.955 1.032
2 Jorabat Site 2 23 39 30 17.3 50 20 28 1.038 1.009 1.028
3 Jorabat Site 3 18 38.7 30.5 17.8 60 22 21 0.971 0.951 1.021
4 Jorabat Site 4 25 39 31.2 17.9 55 20 13 1.010 0.997 1.013
5 Jorabat Site 5 18 39 30 17.3 50 15 12 1.101 1.049 1.049
6 Jorabat Site 6 22 39 30 17.3 50 33 12 1.185 1.176 1.008
7 Jorabat Site 7 29 38.5 30.7 17.5 48 18 10 0.986 0.941 1.048
8 Jorabat Site 8 26 37.9 30 17.3 45 22 11 1.210 1.178 1.027
9 Jorabat Site 9 20 38.5 29 17.5 50 17 16 1.123 1.097 1.024
10 Jorabat Site 10 20 39.2 29.7 17.5 55 30 26 1.210 1.156 1.047
11 Byrnihat Site 1 21 39 30 17.5 50 16 21 1.326 1.211 1.095
12 Byrnihat Site 2 30 39.8 31.3 17.8 45 20 11 1.257 1.232 1.020
13 Byrnihat Site 3 20 39 30 17.3 48 30 23 1.355 1.297 1.045
14 Byrnihat Site 4 23 39 30 17.3 53 17 22 1.289 1.279 1.008
15 Byrnihat Site 5 18 57 39 18 50 26 21 1.260 1.250 1.008
16 Byrnihat Site 6 25 57 39 18 52 30 6 1.700 1.584 1.073
17 Byrnihat Site 7 31 57.2 38.6 18.3 38 16 9 1.934 1.784 1.084
18 Byrnihat Site 8 22 57.5 41.3 19.8 61 24 12 1.543 1.431 1.078
19 Byrnihat Site 9 24 57.5 41.3 19.8 62 25 9 1.680 1.570 1.070
20 Byrnihat Site 10 15 13 41.9 18.5 55 25 13 0.987 0.937 1.053
21 Byrnihat Site 11 14 13 41.9 18.7 50 30 8 1.440 1.410 1.021
22 Byrnihat Site 12 15 14 42 18.6 65 30 6 1.100 1.067 1.031
23 Byrnihat Site 13 15 13.7 42.2 18.5 62 25 7 1.210 1.109 1.091
24 Umling Site 1 12 13 41.9 18.5 45 24 4 1.535 1.511 1.016
25 Umling Site 2 16 13 41.9 18.5 52 28 5 1.307 1.276 1.024
26 Umling Site 3 14 13.5 42 18.5 40 30 9 1.743 1.678 1.039
27 Umling Site 4 17 26 42.2 20.4 50 35 18 1.387 1.311 1.058
 Table 3.4: Comparison between 2D and 3D FOS for 40 vulnerable sites from Guwahati and its Adjoining Areas
(Contd.)
Slope Cohesion Angle of Unit Slope Slope Depth FOS
Height (kN/m2) internal weight of Inclination Width of
Sl. No. Site Name (m) friction the soil (°) (m) water
(°) (kN/m3) table
H c φ γ β B h 3D 2D 3D/2D
28 Umling Site 5 20 25.5 42 20.5 50 35 18 1.260 1.245 1.012
29 Umling Site 6 15 25.5 42.2 20.4 40 36 15 1.093 1.047 1.044
30 Umling Site 7 20 25.5 42 20.7 45 20 22 1.110 1.094 1.015
31 Dhirenpara Site 1 15 15 35 18 60 15 3 1.154 1.057 1.092
32 Dhirenpara Site 2 18 18 35 17.9 60 54 5 1.016 1.008 1.008
33 Hengerabari Site 1 8 35 25 18 65 12 4 1.743 1.675 1.041
34 Hengerabari Site 2 11 48 22 18.5 45 16.5 6 2.017 1.987 1.015
35 Sunsali Site 1 18 35 24 18 70 30 2 1.395 1.321 1.056
36 Sunsali Site 2 17 0 37.5 18 45 34 2 1.105 1.096 1.008
37 Sunsali Site 3 28 46 15 18.7 60 80 2 1.047 0.996 1.051
38 Kharguli Site 1 18 36 0 18 50 24 1 1.272 1.18 1.078
39 Kharguli Site 2 25 47.8 0 18 35 30 3 1.020 0.994 1.026
40 Kharguli Site 3 20 27 22 17.8 40 60 6 1.253 1.183 1.059
3.6 CONCLUDING REMARKS FOR 3D SLOPE STABILITY

In this chapter, 3500 artificial slopes are studied using a finite element software

PLAXIS 3D, by varying the geometrical and geotechnical parameters. Design charts are

produced by varying the geometrical and geotechnical parameters to calculate the FOS.

The slip surfaces for this analysis are assumed to be spherical in both y and z direction.

A 15 nodal triangular element is chosen and Mohr-Coulomb failure criterion is considered

for modelling the soil behaviour. Two separate design charts have been presented for 3D

slope stability analysis: one for undrained soils where angle of internal friction, φu = 0

and other for drained soils (c – φ) soils. The use of these charts has been found to be very

straight forward. If the slope geometry and hydrological condition is known, the 3D FOS

of the slope can be easily computed from the design charts. Furthermore, this chapter also

includes the development of two prediction models using MLR and ANN. The

performance of the predicted models and design charts are checked by examining the

results by making predictions against case records which are not used during training and

testing. 40 vulnerable slopes were selected around Guwahati and FOS results are

compared with the predicted results obtained by MLR, ANN and Design Charts. Further,

a comparison study has also been done between 2D and 3D slope stability analysis using

FEM. The summary of the outcomes are given as under-

1. MLR, ANN and Design charts can act as a good prediction tool for predicting the

stability of slopes.

2. The design charts developed for calculating the 3D FOS are found to have the highest

correlation of 93.61 % as against 90.18% and 82.15 % for ANN and MLR

respectively.
3. The 3D FOS values obtained by the design charts are having the least percentage of

RMSE and MAE of 11.06 and 4.78 respectively. This shows that the proposed design

charts are a useful alternatives for slope stability analysis.

4. The use of these design charts eliminates the long iterative process of calculating the

FOS.

5. On comparing the prediction models developed by using MLR and ANN, it can be

concluded that ANN can predict the 3D safety factor with higher degree of precision.

6. If the width of the slide is very large compared to its height i.e., ratio of width (B) to

height (H) of slide is greater than four, the 3D effects (end effects) are negligible and

the stability of the slope is governed by the plane strain condition.

7. The ratio of 3D/2D FOS is found to be greater than unity which implies that the 3D

FOS is always higher than the 2D FOS.

8. In general, slope stability methods that produce higher FOS in 3D analysis than those
in 2D analysis are more accurate while those that produce the opposite are inaccurate
due to the simplifying assumptions or the erroneous derivations or extension of the
methods.