Slope Stability Assessments Considering Material Inhomogeneity

Slope Stability Assessments Considering Material Inhomogeneity
by
Zhiguang Qian
Submitted in fulfilment of the requirements for
the degree of
Doctor of Philosophy (Engineering)
Deakin University
School of Civil Engineering
July, 2017
Signature Redacted by Library




Slope stability assessments considering material inhomogeneity
ACKNOWLEDGEMENTS
First, I would like to extend my sincere gratitude to my supervisor, An - Jui Li, for all his
instructive and useful suggestions while I was conducting my research. He has exerted
considerable time and effort to direct my entire research and the course of my undergraduate,
postgraduate, and PhD studies. I am deeply grateful that he has spent time walking me through
all the stages of the drafting and completion of this thesis. Without his support, this thesis could
not have appeared in its present form.
I am deeply indebted to my co-supervisors, Professor Kanagaratnam Baskaran and Dr. Sui

Yang Khoo, whose valuable instruction and assistance have been greatly beneficial to me. My
gratitude also goes to Professor Andrei Lyamin for his constructive advice, and for giving me
the training course on numerical modeling and simulation software code from the University
of Newcastle.
I want to thank all my friends and my fellow colleagues at Deakin University, whose
continuous support has been remarkable. Particularly, I want to express my gratitude to David
Yong and his wife Ariel Hsieh, for their invaluable advice during the Higher Degrees by
Research confirmation term. Similarly, my appreciation is directed to Dr. Shu Li and Professor
Jega Jegatheesan for organizing the discussion meeting regularly during my postgraduate study
time. These discussions shall remain unforgettable. Particularly, I also wish to express my
gratitude to Kelvin Kah Wee Lim, my friend, who gave me his constructive comments and
discussions. The support and guidance provided by these friends and colleagues truly helped
me in the past few years.
This research has also been made possible with all the direct and indirect help I received from
the staff of Deakin University. Likewise, I wish to thank Deakin University for providing the
opportunities to attend conferences, which motivated me to improve and challenge myself.
I am also very grateful to my family because of their constant support. They always gave me
valuable suggestions and critiques.
Finally, I want to express my gratitude to my wife, Lin Wang, for giving me the courage,
motivation, and support to complete this research.
i
LIST OF PUBLICATIONS
Publications based on this thesis are as follows:
1. Paper 1 (Chapter 3):

Qian, Z.G., Li, A.J., Merifield, R.S. and Lyamin, A.V., (2014). “Slope stability charts
for two-layered purely cohesive soils based on finite-element limit analysis methods.”
Published in International Journal of Geomechanics, 15(3), p.06014022.

Qian, Z.G., Li, A.J., Chen W. C. and Lyamin, A.V., (2017). “Application of artificial
neural network to inhomogeneous soil slope stability assessments based on finite
element limit analysis methods.” Submitted to International Journal of Geomechanics.

Qian, Z.G., Li, A.J., Lyamin, A.V. and Wang, C.C., (2017). “Parametric studies of
disturbed rock slope stability based on finite element limit analysis methods.” Published
in Computers and Geotechnics, 81, pp.155-166.

Qian, Z.G., Li, A.J., Chen W. C. and Lyamin, A.V., (2017). “Seismic disturbed rock
slope stability assessments based on finite element lower bound limit analysis method.”
Submitted to Engineering Geology.
ii
ABSTRACT
Soil and rock slope stability assessment is one of the most important parts of geotechnical
engineering, as can be gleaned from the current literature and related applications. Numerical
simulation is a widely adopted approach in dealing with slope stability problems, particularly
for a large scale complex projects. However, for practicing engineers, performing numerical
simulations is not always feasible due to time and cost constraints. Therefore, stability chart
solutions are considered as convenient tools that can easily determine slope stability.
The aim of the present research is to apply finite element upper and lower bound limit analysis
to soil and rock slope assessments. The investigations include two - layered soil slope stability
assessment, disturbed rock slope analysis, rock slope stability analysis considering seismic
effects, and soil slope stability analysis using an artificial neural network. The differences in
the stability numbers of the upper and lower bound solutions in rock slope stability are
bracketed within ± 10.5% or better, whereas soil slope stability solutions are bracketed within
± 2.2% or better. The obtained stability numbers are presented in soil or rock slope stability
charts, which can be used in conducting convenient slope safety assessments. Moreover, for
comparison purposes, the limit equilibrium method is applied. Results show that the factors of
safety obtained from finite element limit analysis are better than those obtained from the limit
equilibrium method.
In order to predict the factors of safety of soil slopes, this thesis develops an artificial neural
network because of its advantages, namely, increased efficiency and reduced developing effort.
In addition, given that rock mass disturbance and damage zone have significant influence on
evaluating rock slope stability, this research also focuses on the influence of rock mass disturbance
and damage zone on inhomogeneous rock slope stability assessments. Finally, the influence of
earthquakes is also considered.
iii
TABLE OF CONTENTS
ACKNOWLEDGEMENTS .................................................................................................... i
LIST OF PUBLICATIONS ...................................................................................................ii
ABSTRACT ......................................................................................................................... iii
CHAPTER 1 INTRODUCTION ........................................................................................... 1
1.1 INTRODUCTION ....................................................................................................... 1
1.2 RESEARCH MOTIVATIONS AND OBJECTIVES ................................................. 1
1.2.1 Stability charts for civil engineering design ........................................................ 1
1.2.2 Rock slope stability analysis based on Hoek - Brown failure criterion ............... 2
1.2.3 Application of blasting damage zone .................................................................. 2
1.2.4 Seismic rock slope stability ................................................................................. 3
1.2.5 Slope stability analysis by artificial neural network ............................................ 3
1.3 THESIS OUTLINE ..................................................................................................... 3
1.4 REFERENCES ............................................................................................................ 5
CHAPTER 2 LITERATURE REVIEW ................................................................................ 9
2.1 INTRODUCTION ....................................................................................................... 9
2.2 LIMIT EQUILIBRIUM METHOD ............................................................................ 9
2.3 LIMIT ANALYSIS ................................................................................................... 10
2.4 FINITE ELEMENT METHOD................................................................................. 11
2.5 FINITE ELEMENT LIMIT ANALYSIS METHODS ............................................. 11
2.5.1 Finite element lower bound limit analysis methods .......................................... 12
2.5.2 Finite element upper bound limit analysis methods .......................................... 16
2.6 SOIL FAILURE CRITERION .................................................................................. 19
2.7 THE GENERALISED HOEK - BROWN FAILURE CRITERION ........................ 19
2.8 THE GEOLOGICAL STRENGTH INDEX ............................................................. 20
iv
2.9 DISTURBANCE FACTOR ...................................................................................... 21
2.10 APPLICATION OF MOHR - COULOMB FAILURE CRITERION TO ROCK

MASS .............................................................................................................................. 21
2.11. SLOPE STABILITY NUMBER AND CHARTS.................................................. 22
2.12 PREVIOUS STUDIES BASED ON THE PSEUDO STATIC METHOD ............. 23
2.13 ARTIFICIAL NEURAL NETWORK ..................................................................... 24
2.14 MESH ARRANGEMENT ...................................................................................... 26
2.15 LIMITATIONS ....................................................................................................... 27
2.16 REFERENCES ........................................................................................................ 29
2.17 APPENDIX ............................................................................................................. 48
CHAPTER 3 SLOPE STABILITY CHARTS FOR TWO-LAYERED PURELY

COHESIVE SOILS BASED ON THE FINITE ELEMENT LIMIT ANALYSIS
METHODS .......................................................................................................................... 60
3.1 ABSTRACT .............................................................................................................. 60
3.2 INTRODUCTION ..................................................................................................... 60
3.3 PREVIOUS STUDIES .............................................................................................. 61
3.4 PROBLEM DEFINITION ........................................................................................ 63
3.5 RESULTS AND DISCUSSIONS ............................................................................. 64
3.6 APPLICATION EXAMPLE ..................................................................................... 65
3.7 CONCLUSION ......................................................................................................... 66
3.8 REFERENCES .......................................................................................................... 68
3.9 APPENDIX ............................................................................................................... 72
CHAPTER 4 APPLICATION OF ARTIFICIAL NEURAL NETWORK TO

INHOMOGENEOUS SOIL SLOPE STABILITY ASSESSMENTS BASED ON FINITE
ELEMENT LIMIT ANALYSIS METHODS ...................................................................... 88
4.1 ABSTRACT .............................................................................................................. 88
4.2 INTRODUCTION ..................................................................................................... 88
4.3 PREVIOUS STUDIES .............................................................................................. 89
v
4.3.1 Slope stability analysis ...................................................................................... 89
4.3.2 Stability number and chart solutions ................................................................. 89
4.3.3 Slope assessment using ANN ............................................................................ 90
4.4 PROBLEM DEFINITION ........................................................................................ 92
4.5 RESULTS AND DISCUSSION ............................................................................... 93
4.5.1 LA results for stability number and stability charts........................................... 93
4.5.2 Comparison of results with LEM ...................................................................... 94
4.5.3 Comparison of results with two-layered purely cohesive soil slopes ................ 95
4.5.4 ANN training and validation analysis................................................................ 95
4.5.5 Application of ANN to fill slope stability assessments ..................................... 96
4.6 CONCLUSION ......................................................................................................... 96
4.7 REFERENCES .......................................................................................................... 98
4.8 APPENDIX ............................................................................................................. 102
CHAPTER 5 PARAMETRIC STUDIES OF DISTURBED ROCK SLOPE STABILITY

BASED ON FINITE ELEMENT LIMIT ANALYSIS METHODS ................................. 118
5.1 ABSTRACT ............................................................................................................ 118
5.2 INTRODUCTION ................................................................................................... 118
5.3 PREVIOUS STUDIES ............................................................................................ 119
5.3.1 Applicability of the generalised Hoek - Brown failure criterion ..................... 119
5.3.2 Mohr - Coulomb failure criterion .................................................................... 121
5.3.3 Stability number and stability charts ............................................................... 121
5.3.4 Limit analysis of rock slope stability ............................................................... 122
5.4 PROBLEM DEFINITION ...................................................................................... 123
5.5 RESULTS AND DISCUSSIONS ........................................................................... 124
5.5.1 LA solutions .................................................................................................... 124
vi
5.5.2 Comparisons of LEM solutions obtained using equivalent Mohr - Coulomb

parameters ................................................................................................................. 125
5.5.3 Comparisons of LEM solutions obtained using Hoek - Brown parameters .... 126
5.6 CASE STUDIES ..................................................................................................... 127
5.6.1 Rock slopes in Iran .......................................................................................... 127
5.6.2 Rock slopes presented by Douglas .................................................................. 127
5.6.3 Slope failure in closely jointed rock mass in barite open pit mine .................. 128
5.7 CONCLUSIONS ..................................................................................................... 128
5.8 REFERENCES ........................................................................................................ 130
5.9 APPENDIX ............................................................................................................. 136
CHAPTER 6 SEISMIC DISTURBED ROCK SLOPE STABILITY ASSESSMENTS

BASED ON FINITE ELEMENT LOWER BOUND LIMIT ANALYSIS METHOD ..... 151
6.1 ABSTRACT ............................................................................................................ 151
6.2 INTRODUCTION ................................................................................................... 151
6.3 PREVIOUS STUDIES ............................................................................................ 152
6.3.1 The Generalised Hoek - Brown failure criterion ............................................. 152
6.3.2 Mohr - Coulomb failure criterion .................................................................... 153
6.3.3 Seismic coefficient........................................................................................... 154
6.3.4 Rock slope stability number and stability charts ............................................. 154
6.3.5 Analysis of rock slope stability ....................................................................... 155
6.4 PROBLEM DEFINITION ...................................................................................... 155
6.5. RESULTS AND DISCUSSIONS .......................................................................... 156
6.5.1 LB solutions for seismic stability number and stability charts ........................ 156
6.5.2 Comparison of results with LEM using the Mohr - Coulomb Failure Criterion
.................................................................................................................................. 157
vii
6.5.3 Comparison of the difference between LA and LEM based on the Hoek - Brown
failure criterion ......................................................................................................... 158
6.5.4. Seismic effects based on the analysis of the stability of homogeneous rock slopes
.................................................................................................................................. 159
6.5.5 Case studies 1 .................................................................................................. 159
6.5.6 Case studies 2 .................................................................................................. 160
6.6 CONCLUSION ....................................................................................................... 161
6.7 REFERENCES ........................................................................................................ 162
6.8 APPENDIX ............................................................................................................. 168
CHAPTER 7 CONCLUDING REMARKS....................................................................... 195
7.1 GENERAL OVERVIEW ........................................................................................ 195
7.2 FILL SLOPES COMPOSED OF TWO - LAYERED PURELY COHESIVE SOILS

....................................................................................................................................... 196
7.3 APPLICATION OF ARTIFICIAL NEURAL NETWORK TO INHOMOGENEOUS

SOIL SLOPE STABILITY ASSESSMENTS .............................................................. 196
7.4 THE DISTURBED ROCK SLOPE STABILITY ANALYSIS BASED ON FINITE

ELEMENT LIMIT ANALYSIS METHODS ............................................................... 197
7.5 THE DISTURBED ROCK SLOPE STABILITY CONSIDER DIFFERENT

SEISMIC COEFFICIENT EFFECTS ........................................................................... 197
7.6 SUMMARY ............................................................................................................ 198
7.7 RECOMMENDATIONS FOR FUTURE WORK .................................................. 199
viii
LIST OF TABLES
Table 2.1. Summary table for the 2D limit equilibrium method ......................................... 48
Table 2.2. Summary table for the 3D limit equilibrium method ......................................... 49
Table 2.3. Summary table for the finite element method .................................................... 50
Table 2.4. Summary table for the 2D finite element limit analysis .................................... 51
Table 2.5. Summary table for the 3D finite element limit analysis .................................... 52
Table 2.6. Relationships between the blast damaged zone thicknesses and slope height
(Hoek and Karzulovic 2000) ................................................................................................ 52
Table 3.1. Stability numbers for  = 15o ............................................................................. 72
Table 3.2. Stability numbers for  = 30° ............................................................................. 74
Table 3.6. Results for Case 1 (cu1/cu2 = 0.2)........................................................................ 82
Table 3.7. Factors of safety for Case 2 ................................................................................ 82
Table 3.8. Results for Case 3 (cu1/cu2 = 0.8)........................................................................ 82
Table 4.1. Parameters investigated in this research .......................................................... 102
Table 4.2. The relationship between slope stability and d/H ............................................ 102
Table 4.3. The relationship between slope stability and  ................................................ 102
Table 4.4. The relationship between two layered purely cohesive soil slopes and  ........ 103
Table 4.5. Summary of slope data from cases studies ...................................................... 103
Table 4.6. Summary of F calculated by using NANN and N2ci ........................................... 104
Table 5.1. Parameters investigated in this study ............................................................... 136
Table 5.2. Average stability numbers for T = 0.5 H, mi = 5 and D0 = 0.7 ........................ 136
ix
Table 5.3. Average stability numbers for T = 0.5 H, mi = 5 and D0 = 1.0 ........................ 136
Table 5.4. The rock type, GSI, mi and other relevant information for Iran slope case
studies ................................................................................................................................ 137
Table 5.5. Assessment of the safety factor by using LA charts solution .......................... 137
Table 5.6. Summary of slope data from cases studies base on rock parameters ............... 137
Table 5.7. The factor of safety analysis by using different chart solutions ....................... 138
Table 6.1. Parameters investigated in this study ............................................................... 168
Table 6.2. Stability numbers (Ni) for T = 0.5 H, mi = 5, D0 = 0.7 or 1.0 under a range of
Kh. ....................................................................................................................................... 168
Table 6.3. Compared difference in F between Hoek - Brown (HB) and Mohr - Coulomb
(MC) failure criteria ........................................................................................................... 169
Table 6.4. Ni comparisons by using LA and LEM ............................................................ 169
Table 6.5. Seismic slope data selected based on different Kh magnitudes ........................ 169
Table 6.6. The factor of safety analysis by using different chart solutions based on D0 =
0.7....................................................................................................................................... 170
Table 6.7. Summary of slope data from cases studies base on rock parameters ............... 170
Table 6.8. The factor of safety analysis by using different chart solutions based on D0 =
0.7....................................................................................................................................... 171
Table 6.9. Parameters investigated in Daguangbao landslide based on Zhang et al. (2015)
............................................................................................................................................ 172
Table 6.10. The factor of safety analysis by using different condition ............................. 172
x
LIST OF FIGURES
Figure 2.1. The slope model and forces based on the research of Krahn in 2003 .............. 53
Figure 2.2. The different types of element based on lower bound analysis ........................ 53
Figure 2.3. The model of stress discontinuity in lower bound analysis .............................. 53
Figure 2.4. The stresses sign convention ............................................................................ 54
Figure 2.5. The stress boundary conditions ........................................................................ 54
Figure 2.6. The finite element upper bound limit analysis methods ................................... 54
Figure 2.7. The linear equation for Mohr - Coulomb presented graphically ...................... 55
Figure 2.8. The geological strength index chart for heterogeneous rock masses (Marinos
and Hoek 2001) .................................................................................................................... 55
Figure 2.9. The Geological strength index classification based on poor quality rock masses
(Marinos et al. 2005) ............................................................................................................ 56
Figure 2.10. The guidelines for estimating disturbance factor (Hoek et al. 2002) ............. 57
Figure 2.11. The seismic rock slope model based pseudo static method ............................ 58
Figure 2.12. The recommended horizontal seismic values (DeWayne and Julie 1997) ..... 58
Figure 2.13. Artificial neural networks based on single hidden layer neural network (Li et
al. 2016) ............................................................................................................................... 58
Figure 2.14. The mesh arrangement.................................................................................... 59
Figure 2.15. The relationship between results and mesh refine (Liu and Zhao 2012) ........ 59
Figure 3.1. Problem configuration for cohesive material filled on the purely cohesive soil
.............................................................................................................................................. 83
Figure 3.2. Typical two dimensional finite element meshes and boundary conditions used
in limit analysis .................................................................................................................... 83
xi
Figure 3.3. The chart solutions for  = 15° ......................................................................... 84
Figure 3.8. The transition of UB plastic zones for  = 15° and d/H = 1.5 .......................... 86
Figure 3.9. The transition of UB plastic zones for  = 15° and cu1/cu2 = 0.5 ...................... 87
Figure 3.10. The transition of UB plastic zones for  = 30° and cu1/cu2 = 0.66 .................. 87
Figure 4.1. The model based on the research of Qian et al. (2014) and Lim et al. (2015a)
............................................................................................................................................ 105
Figure 4.2. ANN based on single hidden layer neural network (Li et al. 2016) ............... 105
Figure 4.3. cu2 linearly increased based on  in Region 2 for purely soil material filled on
inhomogeneous cohesive soil ............................................................................................ 105
Figure 4.4. The trained ANN inputs forward and inverse to the outputs .......................... 105
Figure 4.5. LA solutions of stability numbers for  = 15° ................................................ 107
Figure 4.9. LA solutions of stability numbers for = 75° ................................................ 113
Figure 4.10. Transition of UB plastic zones for  = 0.25, = 15°, 30° and 45° and d/H =
1.5....................................................................................................................................... 113
Figure 4.11. Transition of UB plastic zones for  = 1.5 and cu1/cu2 = 5.0 ........................ 114
xii
Figure 4.12. UB plastic zones for different  (0.25, 0.75 or 1.5) when = 45°, d/H = 2 and
cu1/cu2 = 1.5 ........................................................................................................................ 114
Figure 4.13. Stability number difference based on different method ............................... 114
Figure 4.14. Comparisons of slope failure mechanisms between UB and the LEM ........ 115
Figure 4.15. Comparisons slope failure mechanisms for different  ................................ 116
Figure 4.16. Results of ANN slope stability trained with 3200 data set ( from 0 to 1.5)
............................................................................................................................................ 116
Figure 4.17. Results of ANN slope stability trained validation data with 800 data set (
from 0 to 1.5) ..................................................................................................................... 117
Figure 5.1. Comparison between the Hoek - Brown failure criterion and equivalent Mohr -
Coulomb failure envelope. ................................................................................................. 139
Figure 5.2. D linearly decreased with depth in Region 1 for rock slope model. .............. 139
Figure 5.3. Limit analysis solutions of stability numbers for inhomogeneous rock slopes β
= 15o ................................................................................................................................... 140
= 30o ................................................................................................................................... 141
= 45o ................................................................................................................................... 142
= 60o ................................................................................................................................... 143
= 75o ................................................................................................................................... 144
Figure 5.8. Stability numbers for GSI = 60, T = 2 H and different β ............................... 145
Figure 5.9. Stability numbers for GSI = 90, T = 0.5 H and different β ............................ 145
Figure 5.10. Stability numbers for β = 60o or 30o, D0 = 1.0, GSI = 50 and different T .... 145
xiii
Figures 5.11. Failure surfaces for β = 60o, GSI = 70, D0 = 1.0, mi = 10 and different T
............................................................................................................................................ 146
Figures 5.12. Failure surface for GSI = 60, D0 = 1.0, mi = 20, T = 0.5 H and different β
............................................................................................................................................ 147
Figure 5.13. Comparison of the upper bound plastic zones for β = 75o, GSI = 50, D0 = 1.0
and T = 1.5 H ..................................................................................................................... 148
Figure 5.14. Failure surface comparisons between the LEM and upper bound results for T
= 2.5H, mi = 10, GSI = 50, D0 = 1.0 .................................................................................. 148
Figure 5.15. Comparisons of the upper bound, lower bound and LEM solutions for T = 1.5
H, GSI = 50 and different β ............................................................................................... 149
Figure 5.16. Disturbance factors relation between Sonmez et al. (2004) and Hoek et al.
(2002) adapted from Sonmez et al. (2004) ........................................................................ 150
Figure 5.17. Comparisons of the slip area present by D linearly decreased with T (band),
plastic zones given by Li et al (between two red lines) and Sonmez and Ulusay (black line)
............................................................................................................................................ 150
Figure 6.1. Comparison between the Hoek–Brown failure criterion and equivalent Mohr–
Coulomb failure envelope .................................................................................................. 173
Figure 6.2. Design Kh for pseudo-static analysis .............................................................. 173
Figure 6.3. D0 linearly decreased with depth in Region 1 for seismic rock slope model
............................................................................................................................................ 173
Figure 6.4. Limit analysis solutions of stability numbers for rock slopes β = 15o, Kh = 0.1
............................................................................................................................................ 174
............................................................................................................................................ 175
............................................................................................................................................ 176
xiv
............................................................................................................................................ 177
Figure 6.8. Limit analysis solutions of stability numbers for rock slopes β =30o, Kh = 0.2
............................................................................................................................................ 178
............................................................................................................................................ 179
............................................................................................................................................ 180
............................................................................................................................................ 181
............................................................................................................................................ 182
............................................................................................................................................ 183
............................................................................................................................................ 184
............................................................................................................................................ 185
............................................................................................................................................ 186
............................................................................................................................................ 187
............................................................................................................................................ 188
Figure 6.19. Ni for GSI = 60, D0 = 0.7 or 1.0, Kh = 0.1 with different T .......................... 189
Figure 6.20. Failure surfaces for β = 45o, GSI = 70, D0 = 1.0, mi = 10 and different T ... 189
xv
Figure 6.21. The LB failure surface based on plastic points for different slope angles ... 190
Figure 6.22. Assumed 2D seismic rock slopes model based on LEM .............................. 190
Figure 6.23. The failure surfaces difference between MC and HB .................................. 191
Figure 6.24. Comparison seismic failure surface when T = 0.5, GSI = 50, mi = 10 and D0 =
0.7....................................................................................................................................... 192
Figure 6.25. Comparison seismic failure surface when T difference between

inhomogeneous and homogeneous rock slope ................................................................... 193
Figure 6.26. Problem configuration for Daguangbao landslide analysis .......................... 193

inhomogeneous and homogeneous rock slope when Kh = 0.2 or 0.3 ................................ 194
xvi
CHAPTER 1 INTRODUCTION
1.1 INTRODUCTION
Based on the current literature and some related applications, several geotechnical engineering
assessment techniques have been proposed and used, including trench stability (Morgenstern
and Amir-Tahmasseb 1965; Fox 2004; Li et al. 2013), tunnel stability (Lee et al. 2006;
Chambon and Corte 1994; Barton et al. 1974) and slope stability (Bishop 2008; Griffiths and
Lane 1999; Dawson et al. 1999; Li et al. 2009). The stability problem of natural slopes is
commonly encountered in civil engineering projects. Meanwhile, slope design includes
investigations on reliability, economics, failure mechanism, and possible remedial measures.
Although slope stability analysis has drawn the attention of many investigators (Duncan 1996;
Chang 2002; Baker et al. 2006; Michalowski and Tabetha Martel 2011), the slope stability
problem still presents a significant challenge to safety design.
In the current research, the main outcomes are presented through stability charts. Chart
solutions can be convenient tools used to determine slope stability during preliminary designs
(Li et al. 2009; Qian et al. 2014; Lim et al. 2015). The literature review reveals that soil and
rock have different properties (Fu and Liao 2010; Qian et al. 2017; Hoek et al. 2002). Therefore,
in slope stability analysis, the present research adopted different theories based on different
conditions. All slope stability analyses were performed using a two - dimensional (2D) plane
strain analysis, and the investigation of slope stability featured various soil and/or rock
materials. As outcomes, a series of special slope stability charts were made, including soil slope
stability charts based on two different layers of purely cohesive soils, inhomogeneous cohesive
stability charts based on two different layers, rock slope stability charts based on
inhomogeneous rock mass, and rock slope stability charts based on seismic effects.
1.2 RESEARCH MOTIVATIONS AND OBJECTIVES
1.2.1 Stability charts for civil engineering design
In 1937, Taylor proposed the first set of stability charts for purely cohesive uniform soil slopes
(Taylor 1937). Since then, the use of slope stability charts as design tools have become
widespread among researchers and practitioners (Li et al. 2009; Michalowski 2002; Baker et
al. 2006). Most existing stability charts are designed exclusively for soil slopes and are based
on the conventional Mohr – Coulomb equivalent parameters (Qian et al. 2014; Taylor 1937;
1
Baker 2003; Steward et al. 2010). One of the objectives of the current study is to investigate
soil slope stability charts based on two layered purely cohesive soils and soil slope stability
charts for fill materials. To the bets of our knowledge, no previous case studies or literature
regarding these research topics are available.
In addition, slope stability charts rarely use rock parameters or are based on a nonlinear failure
criterion. Therefore, another aim of this research is to investigate the inhomogenous rock slope
stability charts and the rock slope stability charts based on seismic effects. The rock slope
stability charts in this research are efficient and easy to read. Based on this type of charts,
engineers can make an immediate analysis and understand basic information quickly. However,
finding a comprehensive stability chart for inhomogeneous rock slopes is difficult.
The results from this research are presented as chart solutions based on the form of stability
numbers. The definition of factor of safety (F) in the stability number is generally different
from that of other methods, such as limit equilibrium method (LEM) (Li et al. 2012). This study
fills the existing gap in the soil and rock slope stability literature.
1.2.2 Rock slope stability analysis based on Hoek - Brown failure criterion
Rock slope stability is typically regarded by geotechnical engineers as a difficult problem. For
rock slope stability analysis, most software are based on linear criteria and still require Mohr -
Coulomb strength parameters (Han and Leshchinsky 2004; Zanbak 1983; Taheri and Tani
2010). Past estimates of rock slope stability based on soil parameters are inaccurate, because
they fail to consider the non - linear nature of rock mass failure envelope (Fu and Liao 2010;
Li et al. 2011). Therefore, in this research, rock mass strength is estimate based on nonlinear
Hoek - Brown failure criterion. The results are more accurate and reliable than those obtained
using the Mohr-Coulomb strength parameters.
1.2.3 Application of blasting damage zone
Disturbance factor (Hoek et al. 2002) is a significant concern regarding slope safety or stability
(Li et al. 2011). In this context, knowing how the disturbance factor influences a rock slope is
important. Based on the research of Hoek in 2012, a consistent disturbance factor (D) could
seriously underestimate the strength and stability of the entire rock mass. Thus, D should only
be confined in the damaged zone. The blasting damage zone thickness (T) can be very shallow
or very deep, so different ranges are considered in this research. As one of the research
2
objectives of this study, D that linearly decreases with depth is taken into consideration when
creating rock slope design charts.
1.2.4 Seismic rock slope stability
Seismic activity is a common and difficult problem in engineering. It can trigger landslides
and/or slope collapses, such as the Wenchuan Earthquake in 2008 (Huang et al. 2009) and the
Niigata Earthquake in 2004 (Wang et al. 2007). Thus far, recent investigations (Li et al. 2011;
Li et al. 2012; Li et al. 2009) have shown that complete systematic inhomogeneous rock slope
charts on seismic effects remain lacking. Thus, this part of the study presents seismic charts for
assessing seismic rock slope stability by considering different seismic coefficients. The
objective of seismic rock slope design is to provide straightforward and accurate basis through
which to estimate the F under seismic conditions.
1.2.5 Slope stability analysis by artificial neural network
Over the last few years, artificial neural networks (ANN) have been applied to solve
geotechnical problems in many areas (Lee et al. 2004; Shahin et al. 2001; Ermini et al. 2005;
Gomez and Kavzoglu 2005; Li et al. 2016; Lee and Lee 1996; Goh 1994; Li et al. 2014). For
this part of the investigation, the ANN is adopted to estimate two - layered undrained soil
slopes. Once the data are trained in ANN, they can be used to predict the slope failure of the
two-layered undrained clay slopes. Such data can also be used to predict stability numbers, F,
and the potential failure mechanism of the slopes. Therefore, conducting a thorough study on
how to make straightforward and accurate slope safety assessments is important.
1.3 THESIS OUTLINE
As an overview, the whole research presented in this thesis can be divided into four principal
sections, two for soil slope stability analysis and another two for rock slope stability analysis.
The Chapters 1 and 2 present the introduction of this thesis and the previous literature review,
respectively.
Chapters 3 – 6, highlight the main body of the thesis and present papers for the studies of this
research. After these chapters, relevant materials are all presented in the appendix.
3
The chapters are as follows:
 Chapter 3: Slope stability charts for two-layered purely cohesive soils based on finite
element limit analysis methods
 Chapter 4: Artificial neural network for inhomogeneous soil slope stability using finite
element limit analysis methods
 Chapter 5: Parametric studies of disturbed rock slope stability based on finite element
upper bound and lower bound limit analysis methods
 Chapter 6: Seismic disturbed rock slope stability assessments based on finite element
lower bound limit analysis method
Chapter 7 discusses a brief summary and discussion of directions of future research.
4
1.4 REFERENCES
Morgenstern N and Amir-Tahmasseb I (1965) The stability of a slurry trench in cohesionless

soils. 1965.
Fox PJ (2004) Analytical solutions for stability of slurry trench. Journal of geotechnical and
geoenvironmental engineering2004. p. 749-58.
Li A, Merifield R, Lin H and Lyamin A (2013) Trench stability under bentonite pressure in
purely cohesive clay. International Journal of Geomechanics2013. p. 151-7.
Lee C, Wu B, Chen H and Chiang K (2006) Tunnel stability and arching effects during
tunneling in soft clayey soil. Tunnelling and Underground Space Technology2006. p. 119-32.
Chambon P and Corte J-F (1994) Shallow tunnels in cohesionless soil: stability of tunnel face.
Journal of Geotechnical Engineering1994. p. 1148-65.
Barton N, Lien R and Lunde J (1974) Engineering classification of rock masses for the design
of tunnel support. Rock mechanics1974. p. 189-236.
Bishop AW (2008) The use of the slip circle in the stability analysis of slopes. The Essence of
Geotechnical Engineering: 60 years of Géotechnique: Thomas Telford Publishing; 2008. p.
223-33.
Griffiths D and Lane P (1999) Slope stability analysis by finite elements. Geotechnique1999.
p. 387-403.
Dawson E, Roth W and Drescher A (1999) Slope stability analysis by strength reduction.
Geotechnique1999. p. 835-40.
Li A, Lyamin A and Merifield R (2009) Seismic rock slope stability charts based on limit
analysis methods. Computers and Geotechnics2009. p. 135-48.
Duncan JM (1996) State of the art: limit equilibrium and finite-element analysis of slopes.
Journal of Geotechnical Engineering, ASCE1996. p. 577-96.
Chang M (2002) A 3D slope stability analysis method assuming parallel lines of intersection
and differetial straining of block contacts. Can Geotech J2002. p. 799-811.
5
Baker R, Shukha R, Operstein V and Frydman S (2006) Stability charts for pseudo-static slope
stability analysis. Soils dynamics and earthquake engineering2006. p. 813-23.
Michalowski RL and Tabetha Martel SM (2011) Stability charts for 3D failures of steep slopes
subjected to seismic excitation. Journal of Geotechnical and Geoenviromental Engineering,
ASCE2011. p. 183-9.
Qian Z, Li A, Merifield R and Lyamin A (2014) Slope stability charts for two-layered purely
cohesive soils based on finite-element limit analysis methods. International Journal of
Geomechanics2014. p. 06014022.
Lim K, Lyamin A, Cassidy M and Li A-J (2015) Three-dimensional slope stability charts for
frictional fill materials placed on purely cohesive clay. International Journal of
Fu W and Liao Y (2010) Non-linear shear strength reduction technique in slope stability
calculation. Computers and Geotechnics2010. p. 288-98.
Qian Z, Li A, Lyamin A and Wang C (2017) Parametric studies of disturbed rock slope stability
based on finite element limit analysis methods. Computers and Geotechnics2017. p. 155-66.
Hoek E, Carranza-Torres C and Corkum B (2002) Hoek-Brown failure criterion-2002 edition.

Proceedings of NARMS-Tac2002. p. 267-73.
Taylor DW (1937) Stability of earth slopes: Boston Society of Civil Engineers; 1937.
Michalowski RL (2002) Stability charts for uniform slopes. Journal of Geotechnical and
Geoenvironmental Engineering2002. p. 351-5.
stability analysis. Soil Dynamics and Earthquake Engineering2006. p. 813-23.
Baker R (2003) A second look at Taylor’s stability chart. Journal of geotechnical and
Steward T, Sivakugan N, Shukla S and Das B (2010) Taylor’s slope stability charts revisited.
International Journal of Geomechanics2010. p. 348-52.
6
Li A, Cassidy M, Wang Y, Merifield R and Lyamin A (2012) Parametric Monte Carlo studies
of rock slopes based on the Hoek–Brown failure criterion. Computers and Geotechnics2012.
p. 11-8.
Han J and Leshchinsky D (2004) editorêditors. Limit equilibrium and continuum mechanics-
based numerical methods for analyzing stability of MSE walls. 17th ASCE Engineering
Mechanics Conference; 2004.
Zanbak C (1983) Design charts for rock slopes susceptible to toppling. Journal of Geotechnical
Engineering1983. p. 1039-62.
Taheri A and Tani K (2010) Assessment of the stability of rock slopes by the slope stability
rating classification system. Rock mechanics and rock engineering2010. p. 321-33.
Li A, Merifield R and Lyamin A (2011) Effect of rock mass disturbance on the stability of rock
slopes using the Hoek–Brown failure criterion. Computers and Geotechnics2011. p. 546-58.
Hoek E (2012) <Blast-Damage-Factor-D-Hoek.pdf>. Technical note forRocNews2012.
Huang R, Li and WL (2009) Analysis of the geo-hazards triggered by the 12 May 2008
Wenchuan Earthquake, China. Bulletin of Engineering Geology and the Environment2009. p.
363-71.
Wang H, Sassa K and Xu W (2007) Analysis of a spatial distribution of landslides triggered by

the 2004 Chuetsu earthquakes of Niigata Prefecture, Japan. Natural Hazards2007. p. 43.
Lee S, Ryu J-H, Won J-S and Park H-J (2004) Determination and application of the weights
for landslide susceptibility mapping using an artificial neural network. Engineering
Geology2004. p. 289-302.
Shahin MA, Jaksa MB and Maier HR (2001) Artificial neural network applications in
geotechnical engineering. Australian Geomechanics2001. p. 49-62.
Ermini L, Catani F and Casagli N (2005) Artificial neural networks applied to landslide
susceptibility assessment. Geomorphology2005. p. 327-43.
7
Gomez H and Kavzoglu T (2005) Assessment of shallow landslide susceptibility using

artificial neural networks in Jabonosa River Basin, Venezuela. Engineering Geology2005. p.
11-27.
Li A, Khoo S, Lyamin A and Wang Y (2016) Rock slope stability analyses using extreme
learning neural network and terminal steepest descent algorithm. Automation in
Construction2016. p. 42-50.
Lee I-M and Lee J-H (1996) Prediction of pile bearing capacity using artificial neural networks.
Computers and geotechnics1996. p. 189-200.
Goh AT (1994) Seismic liquefaction potential assessed by neural networks. Journal of

Geotechnical engineering1994. p. 1467-80.
Li A, Khoo S, Wang Y and Lyamin A (2014) editorêditors. Application of neural network to

rock slope stability assessments. Proceedings of the 8th European Conference on Numerical
Methods in Geotechnical Engineering (NUMGE 2014); 2014.
8
CHAPTER 2 LITERATURE REVIEW
2.1 INTRODUCTION
The slope stability analysis plays an important role in geotechnical engineering (Schmid et al.
2016; Lim et al. 2014). A successful slope stability assessment requires accurate geological
information and appropriate analysis theory. In general, LEM is the most popular approach
used to evaluate slope stability problems (Cassidy et al. 2008; Duncan et al. 2008). However,
the best solutions cannot be guaranteed by using the LEM because the failure surface must be
assumed (Li et al. 2012). The finite element upper bound (UB) and lower bound (LB) limit
analysis (LA) have been recently developed for slope stability assessment (Lyamin and Sloan
2002a; Lyamin and Sloan 2002b; Krabbenhoft et al. 2005). For LA, the failure mechanisms
can be automatically generated (Lyamin and Sloan 2002a; Krabbenhoft et al. 2005).
For slope stability design, numerical simulations must still be performed. For practicing
engineers, the sophisticated numerical modelling may not always be warranted or feasible due
to time and cost constraints. Therefore, chart solutions can be used as convenient tools to
provide an easy way to determine the F of a slope. The investigations in this research include
two main areas: soil slope stability analysis and rock slope stability assessment.
The results obtained in soil slopes are based on the Mohr - Coulomb failure criterion. This
failure criterion can be written as a linear equation. Moreover, the results are obtained based
on the Hoek - Brown failure criterion for rock slopes analysis.
In different geological conditions, engineers must consider varying strengths, disadvantages,

and limitations inherent in each methodology and, choose the best analysis technique for the
design. This research considers different methods based on different situations, and the method
chosen is based on either the Hoek-Brown failure criterion or the Mohr - Coulomb failure
criterion.
2.2 LIMIT EQUILIBRIUM METHOD
The LEM is the most popular approach used to evaluate slope stability problems (Cassidy et
al. 2008; Duncan et al. 2008). It is a well - established method that is widely used by civil
engineers (Cheng et al. 2007). The LEM can estimate slope F without considering the stress
strain relationship. It is a statically method based on a physical theory, which is presented as a
9
translational collapse mechanism (Michalowski 1995). The F in LEM is defined in Equation

(2.1)
∑(resisting actions)
Factor of safety = ∑(driving actions)
(2.1)
Currently, the LEM has been applied to the 2D (Bishop 1955; Morgenstern and Price 1965;
Fredlund and Krahn 1977) and three - dimensional (3D) slope stability analyses (Chang 2002;
Huang et al. 2002). Table 2.1 and Table 2.2 provide a summary of other works that relate to
the LEM.
Basically, the LEM is used to seek the F or the maximum external load. This method, assumes
the inter column shear force (Petterson 1955; Bishop 1955). Based on the research of Fellenius
in 1936, the slope should be divided into many slices and a continuous surface passes all slices.
The F can be considered as the sum of resisting actions, which is divided by the sum of the
driving actions (Steward et al. 2010; Zhu et al. 2003), as shown in Figure 2.1. The forces on
the failure surface should be equal to the resistance or the shear strength (Chen and Chameau
1983).
In fact, the LEM needs to assume the inter column shear force. Moreover, the potential slip
surface must be assumed before calculating the stability of slopes (Cheng et al. 2007). Many
relevant studies have been performed to determine the failure surface, such as Greco (1996),
Zhu (2001), Cheng et al. (2008) and Hajiazizi and Tavana (2013). Past research on 3D slope
stability analysis based on Huang and Tsai (2000) and Huang et al. (2002) also show that prior
assumptions of the sliding direction can be solved by considering the sliding direction of the
sliding mass.
However, according to the assumption mentioned above, the literature review results obtained
by the LEM could affect the result accuracy (Lim et al. 2015; Yu et al. 1998).
2.3 LIMIT ANALYSIS
The limit analysis method has attracted research attention for several decades (Michalowski
1995; Donald and Chen 1997; Farzaneh and Askari 2003; Michalowski and Nadukuru 2012).
In addition, the slope stability analysis based on 3D limit analysis method has also gained the
attention of researchers (Michalowski 1989; Chen et al. 2001a; Chen et al. 2001b; Michalowski
2010; Michalowski and Martel 2011; Gao et al. 2012). Similar with LEM, the conventional
limit analysis method also requires the assumption of the slip surface before calculation (Gao
10
et al. 2013). The limit analysis method is based on two different limit theorems, the upper
bound and lower bound. The F obtained using lower bound analysis is generally smaller or
equal to the actual value. In comparison, the F is larger than or equal to the actual value when
upper bound analysis is applied (Leshchinsky and Ambauen 2015).
However, using lower bound theorem limit analysis to compute the stress field in a slope is
difficult. Therefore, only the upper bound analysis is widely used to solve the slope stability
problem (Michalowski 1995, Donald and Chen 1997; Chen et al. 2004). An investigation based
on Lim et al. (2016) shows that the use of only the upper bound method may cause conservative
solutions. In addition, the critical failure mechanism may highly influence the accuracy of the
slope analysis (Gao et al. 2013). The problem regarding failure mechanism has attracted many
researchers, such as Chen et al. (2001), who divided the failure mass into prisms; and, Gao et
al. (2013), who introduced a toe failure with rotational failure mechanism based on the research
of Michalowski and Drescher .
2.4 FINITE ELEMENT METHOD
In 1943, Courant introduced the finite element methods (Dhatt et al. 2012). The finite element
analysis is a numerical technique (Bathe and Wilson 1976) that has been well accepted in
geotechnical engineering (Hammah 2005; Zienkiewicz and Taylor 2000; Zienkiewicz and
Taylor 2005). The finite element methods can be applicable on a series of complex conditions
that are difficult to achieve using LEM alone, such as slope movements and soil displacement
(Potts et al. 2009; Troncone 2005). The F obtained by the finite element method are based on
the strength reduction method (Matsui and San 1992). The results obtained from the research
of Hammouri et al. in 2008 and Ozbay and Cabalar in 2015, show that the F generally agreed
with those of the LEM. Furthermore, this method is widely accepted by many researchers
(Jianxun et al. 2004; Zhao et al. 2005; Dawson et al. 1999; Hammah et al. 2005; Shangyi et al.
2003). A brief overview of the finite element method is shown in Table 2.3.
2.5 FINITE ELEMENT LIMIT ANALYSIS METHODS
For the past several years, the LA has been developed rapidly and has been proven to be robust
and efficient in estimating the strength of slopes (Sloan 1988; Lyamin and Sloan 2002b; Sloan
1989; Lyamin and Sloan 2002b; Krabbenhoft et al. 2005; Sloan 2013). The LA are based on
two separated methods. Moreover, the LA can be used to analyses soil and rock mass analysis
(Bottero et al. 1980;Li et al. 2008), and do not require prior assumptions of the failure
11
mechanisms (Sloan 2013), therefore, it can be applied to various geotechnical problems

(Sutcliffe et al. 2004; Merifield et al. 2006; Soubra 1999; Shahin et al. 2001; Chai and Carter
2009). Huang et al. (2013) introduced the volume of sliding mass analysis based on LA and
estimated landslide based on sliding mass. Shiau et al. (2011) investigated the effects of
external loading on undrained slopes. Table 2.4 and Table 2.5 provide a brief overview of the
LA development.
2.5.1 Finite element lower bound limit analysis methods
The LB limit analysis is uses static theorem (Lyamin and Sloan 2002a; Krabbenhoft et al.
2005). The LB is based on linear programming developed in 1970 based on Lysmer . However,
the original method cannot solve complicated problems due to limited computational
performance. In 1988, Sloan used an active set algorithm to avoid traditional simplex strategies.
Thus, the LB based on an efficient linear programming has witnessed great improvement.
In addition, Lyamin and Sloan (2002a) developed a new formulation for LB. The new LB
theorem is based on linear finite elements with nonlinear programming. Based on the research
of Lyamin and Sloan (2002a), the new approach provides accurate solutions that are
significantly better than the linear programming approach.
Three types of elements are used for LB, as shown in Figure 2.2, and can be expressed as
Equations (2.2) to (2.4).
𝜎𝑥 = ∑i=3
i=1 𝑁𝑖 𝜎xi (2.2)
σy = ∑i=3
i=1 𝑁𝑖 𝜎𝑦𝑖 (2.3)
τxy = ∑i=3
i=1 𝑁𝑖 𝜏𝑥𝑦𝑖 (2.4)
where
Ni represents the linear shape functions and (𝜎𝑥𝑖 , 𝜎𝑦𝑖 , 𝜏𝑥𝑦𝑖 ) are the nodal stresses. Ni may be
expressed in terms of the nodal coordinates (xi, yi) according to Equations (2.5) – (2.7)
𝑁1 = [(𝑥2 𝑦3 − 𝑥3 𝑦2 ) + 𝑦23 𝑥 + 𝑥32 𝑦]/2A (2.5)
𝑁2 = [(𝑥3 𝑦1 − 𝑥1 𝑦3 ) + 𝑦31 𝑥 + 𝑥13 𝑦]/2A (2.6)
𝑁3 = [(𝑥1 𝑦2 − 𝑥2 𝑦1 ) + 𝑦12 𝑥 + 𝑥21 𝑦]/2A (2.7)
12
where
A represents the element area
2A = |𝑥13 𝑦23 − 𝑥32 𝑦31 | (2.8)
𝑥32 = 𝑥3 − 𝑥2 ; 𝑦23 = 𝑦2 − 𝑦3 (2.9)
𝑥13 = 𝑥1 − 𝑥3 ; 𝑦31 = 𝑦3 − 𝑦1 (2.10)
𝑥21 = 𝑥2 − 𝑥1 ; 𝑦12 = 𝑦1 − 𝑦2 (2.11)
For LB collapse, the stresses must be in equilibrium. In addition, LB needs to satisfy the stress
boundary conditions and the yield criterion, which separately require a set of constraints on the
nodal stresses. Therefore, LB must consider all of these factors in the calculation.
Equilibrium conditions
The equilibrium conditions for plane strain must satisfy Equations (2.12) and (2.13), as
shown in Figure 2.3,
𝜕𝜎𝑥 𝜕𝜏𝑥𝑦
𝑘= + (2.12)
𝜕𝑥 𝜕𝑦
𝜕𝜎𝑦 𝜕𝜏𝑥𝑦
𝛾= + (2.13)
𝜕𝑦 𝜕𝑥
where
k can be 0 or equal to a portion of γ. For example, in pseudo static method, k can be a portion
of γ, which implies that the horizontal force is a portion of vertical force for k and γ. Then,
considering Equations (2.2) – (2.4) into Equations (2.12) and (2.13) can obtain
𝑎1 𝑥 = 𝑏 (2.14)
In Equation (2.14)
𝑦23 0 𝑥32 𝑦31 0 0 𝑥13 𝑦12 0 𝑥21

𝑎1 = [ 𝑦23 0 𝑦31 0 ] (2.15)
0 𝑥32 0 𝑥13 𝑦21 𝑥12
𝑥 𝑇 = {𝜎𝑥1 , 𝜎𝑦1 , 𝜏𝑥𝑦1 ⋯ 𝜎𝑥3 , 𝜎𝑦3 , 𝜏𝑥𝑦3 } (2.16)
𝑏1𝑇 = {𝑘𝛾} (2.17)
For each rectangular extension element, these additional equalities are necessary to extend
the linear stress distribution to the fourth node. The Equations are shown below.
13
𝜎𝑥4 = 𝜎𝑥1 + 𝜎𝑥3 − 𝜎𝑥2 (2.18)
𝜎𝑦4 = 𝜎𝑦1 + 𝜎𝑦3 − 𝜎𝑦2 (2.19)
𝜏𝑥𝑦4 = 𝜏𝑥𝑦1 + 𝜏𝑥𝑦3 − 𝜏𝑥𝑦2 (2.20)
and
𝑎2 𝑥 = 0 (2.21)
where
𝑎2 = [𝐼 − 𝐼 𝐼 − 𝐼] (2.22)
1 0 0
𝐼 = [0 1 0] (2.23)
0 0 1
The normal and shear stresses acting on a plane that is inclined at an angle to the x-axis in
Figure 2.4 are expressed as Equations (2.25) and (2.26), respectively
𝜎𝑛 = sin2 𝜃𝜎𝑥 + cos 2 𝜃𝜎𝑦 − sin2𝜃𝜏𝑥𝑦 (2.25)
1
𝜏 = 2 sin2𝜃(𝜎𝑦 − 𝜎𝑥 ) + cos2𝜃𝜏𝑥𝑦 (2.26)
Based on Figure 2.3
𝜎𝑛1 = 𝜎𝑛2 and 𝜎𝑛3 = 𝜎𝑛4 (2.27)
𝜏1 = 𝜏2 and 𝜏3 = 𝜏4 (2.28)
Then put Equations (2.25) and (2.26) into Equations (2.27) and (2.28)
𝑎3 𝑥 = 0 (2.29)
where
𝑇−𝑇 0 0
𝑎3 = [ ] (2.30)
0 0 𝑇−𝑇
sin2 𝜃 cos 2 𝜃 −sin2𝜃

𝑇=[ 1 1 ] (2.32)
− 2 sin2𝜃 sin2𝜃 cos2𝜃
2
14
Stress boundary conditions
For stress boundary conditions, the normal and shear stresses are set as (𝑞1 , 𝑡1 ) and (𝑞2 , 𝑡2 ), as
shown in Figure 2.5. Then the stresses can be expressed in Equations (2.33) and (2.34)
𝜎𝑛1 = 𝑞1 and 𝜎𝑛2 = 𝑞2 (2.33)
𝜏1 = 𝑡1 and 𝜏2 = 𝑡2 (2.34)
Then substituting the stress Equations (2.25) and (2.26) into (2.33) and (2.34)
𝑎4 𝑥 = 0 (2.35)
where
𝑇 0
𝑎4 = [ ] (2.36)
0 𝑇
𝑏4𝑇 = {𝑞1 , 𝑡1 , 𝑞2 , 𝑡2 } (2.38)
The additional constraints place due to involving a uniform applied surface traction
𝜎𝑛1 = 𝜎𝑛2 (2.39)
𝜏1 = 𝜏2 (2.40)
Then substituting Equations (2.25) and (2.26) into these Equations
𝑎5 𝑥 = 0 (2.41)
where
𝑎5 = [𝑇 − 𝑇] (2.42)
Yield conditions
For LB, the stresses at all N nodes in the element must satisfy Equation (2.44)
𝑓 (𝜎𝑖𝑗𝑙 ) ≤ 0, 𝑙 = 1, 2 ⋯ 𝑁 (2.44)
15
2.5.2 Finite element upper bound limit analysis methods
The UB is based on the kinematic theorem (Lyamin and Sloan 2002b). For the past decades,
most slope stability studies are only based on the UB (Donald and Chen 1997; Chen et al. 2003;
Donald and Chen 1997; Sloan and Kleeman 1995; Chen et al. 2003; Kim et al. 2002; Gao et
al. 2013). The initial research based on UB are related to Anderheggen and Knöpfel (1972),
Pastor and Turgeman (1976), and Bottero et al. (1980). However, similar to LB theorem,
applying them to complex geometries problems is difficult (Sloan and Kleeman 1995). In 1989,
Sloan introduced the active set algorithm (Sloan and Kleeman 1995). In 1994, Yu et al.
developed a quadratic element based on UB, which can be used to assume incompressible
velocity field without considering the grid arrangements. This process makes the UB much
more efficient compared with other methods. In addition, due to the initial UB based on linear
programming, many research based on non - linear programming have also been proposed
(Jiang 1995; Lyamin and Sloan 2002b; Liu et al. 1995; Vicente da Silva and Antao 2007; Chen
et al. 2003; Krabbenhoft et al. 2005).
For the UB，the element is based on a three node triangle and considers three stress
components (𝜎𝑥 , 𝜎𝑦 , 𝜎𝑧 ). For each node, two velocity components are considered, as shown in
Figure 2.6. The two velocities are assumed in Equations (2.45) and (2.46).
𝑢 = ∑i=3
i=1 𝑛𝑖 𝑢𝑖 (2.45)
𝑣 = ∑i=3
i=1 𝑛𝑖 𝑣𝑖 (2.46)
where
(ui, vi) are nodal velocities in the x-axis and y-axis respectively, ni represents the linear shape
functions.
For UB, the velocity field must satisfy associated flow rule. The associated flow rule based on
Equations (2.47) – (2.49).
𝜕𝑢 𝜕F
𝜀̇𝑥 = = 𝜆̇ 𝜕𝜎 (2.47)
𝜕𝑥 𝑥
𝜕 𝜕F
𝜀̇𝑦 = 𝜕 𝑣 = 𝜆̇ 𝜕𝜎 (2.48)
𝑦 𝑦
𝜕 𝜕 𝜕F
𝛾̇𝑥𝑦 = (𝜕𝑣 + 𝜕𝑢) = 𝜆̇ 𝜕𝜏 (2.49)
𝑥 𝑦 𝑥𝑦
16
where
𝜆̇ is a plastic multiplier, when 𝜆̇ ≥ 0, the tensile strains are taken as positive.
Yield condition
The stresses in the upper bound formulation must satisfy the yield condition, as shown in
Equation (2.50)
𝑓 (𝜎𝑗𝑖 ) ≤ 0 (2.50)
The stresses of element are assumed to be constant. Therefore, only one yield condition per
finite element is considered. The stress in the finite element model must satisfy the following
Equations.
𝑓 (𝜎𝑖𝑗𝑒 ) ≤ 0, 𝑒 = 1, 2 ⋯ 𝐸 (2.51)
𝑓 (𝜎) ≤ 0, 𝑗 ∈ 𝐽𝜎 (2.52)
Plastic shearing in discontinuities
𝑢 = 𝑁𝑘 𝑢𝑘 + 𝑁1 𝑢1 + 𝑁𝑚 𝑢𝑚 (2.53)
𝑣 = 𝑁𝑘 𝑣𝑘 + 𝑁1 𝑣1 + 𝑁𝑚 𝑣𝑚 (2.54)
where
𝑎𝑘 +𝑏𝑘 𝑥+𝑐𝑘 𝑦
𝑁𝑘 = (2.55)
2∆
𝑎𝑙 +𝑏𝑙 𝑥+𝑐𝑙 𝑦
𝑁𝑙 = (2.56)
2∆
𝑎𝑚 +𝑏𝑚 𝑥+𝑐𝑚 𝑦
𝑁𝑚 = (2.57)
2∆
𝑎𝑘 = 𝑥1 𝑦𝑚 − 𝑥𝑚 𝑦1 , 𝑏𝑘 = 𝑦1 − 𝑦𝑚 , 𝑐𝑘 = 𝑥𝑚 − 𝑥1 (2.58)
𝑎𝑙 = 𝑥𝑚 𝑦𝑘 − 𝑥𝑘 𝑦𝑚 , 𝑏𝑙 = 𝑦𝑚 − 𝑦𝑘 , 𝑐𝑙 = 𝑥𝑘 − 𝑥𝑚 (2.59)
𝑎𝑚 = 𝑥𝑘 𝑦𝑙 − 𝑥𝑙 𝑦𝑘 , 𝑏𝑚 = 𝑦𝑘 − 𝑦𝑙 , 𝑐𝑚 = 𝑥𝑙 − 𝑥𝑘 (2.60)
The Δ represents the triangle area and matrix B is given by Equation (2.61)
𝑏𝑘 0 𝑏𝑙 0 𝑏𝑚 0
1 1 ̅
𝐵
𝐵 = [𝐵𝑘 𝐵𝑙 𝐵𝑚 ] = 2∆ [ 0 𝑐𝑘 0 𝑐𝑙 0 𝑐𝑚 ] = [𝐵̅𝑘 𝐵̅𝑙 𝐵̅𝑚 ] = ∆ (2.61)
∆
𝑐𝑘 𝑏𝑘 𝑐𝑙 𝑏𝑙 𝑐𝑚 𝑏𝑚
17
The power dissipated in element regardless of its area can calculated as:
𝑊 = ∫∆ 𝜎𝜀̇𝑑∆= ∫∆ 𝜎𝐵𝑢 𝑑∆= 𝜎𝐵̅𝑢 (2.62)
𝐵̅𝑘 → 0 𝑎𝑛𝑑 𝐵̅𝑙 → −𝐵̅𝑚 ⟹ 𝐵̃ = [0 − 𝐵̅𝑙𝑚 𝐵̅𝑙𝑚 ]
Then
𝜀̇ = 𝐵̃𝑢 = 𝐵̅𝑙𝑚 ∆𝑢𝑙𝑚 (2.63)
Velocity boundary conditions
Two different velocity boundary conditions are used, one is to define the loading caused by a
stiff structure and the other to define the loading by a uniform normal pressure.
For loading caused by a stiff structure, the nodal velocity components (𝑢𝑖 , 𝑣𝑖 ) must satisfy the
condition as shown below.
cos𝜃 sin𝜃 𝑢𝑖 𝑢̅
[ ] {𝑣 } = { } (2.64)
− sin𝜃 cos𝜃 𝑖 𝑣̅
Then
𝑎31 𝑥1 = 𝑏3 (2.65)
where
cos𝜃 sin𝜃
𝑎31 = [ ] (2.66)
− sin𝜃 cos𝜃
𝑏3𝑇 = {𝑢̅, 𝑣̅ } (2.67)
𝑥1𝑇 = {𝑢, 𝑣} (2.68)
For loading by a uniform normal pressure, the Equations shown below.
1
𝑄 = ∫𝑠 𝑣𝑑𝑆 = 2 ∑𝑒𝑑𝑔𝑒𝑠[(𝑣1 + 𝑣2 )cos𝜃12 − (𝑢1 + 𝑢2 )sin𝜃12 ] 𝑙12 (2.69)
Q represents the rate of material flow across the boundary S.
Therefore
𝑎41 𝑥1 = 𝑏4 (2.70)
1
𝑎41 = 2 [−𝑙12 sin𝜃12 , 𝑙12 cos𝜃12 , −𝑙12 sin𝜃12 𝑙12 cos𝜃12 , ⋯ ] (2.71)
18
𝑏4 = 𝑄 (2.72)
𝑥1𝑇 = {𝑢1 , 𝑣1 , 𝑢2 , 𝑣2 , ⋯ } (2.73)
2.6 SOIL FAILURE CRITERION
In the discipline of soil mechanics, soils are assumed to be essentially frictional materials,
whose strengths depend on the effective stresses in the soil. Moreover, the soil strength will
depend on the soil undrained / short - term condition or drained / long - term condition (Duncan
et al. 2014). Therefore, in assessing the stability of soil constructions, analyses are usually
performed to check the two different situations (Tang et al. 1976; Li and Lumb 1987; Jewell
and Greenwood 1988; Powers et al. 2005).
The Mohr - Coulomb failure criterion (Coulomb 1973) is a widely used strength criterion in
geotechnical engineering applications (Bai and Wierzbicki 2010; Al-Ajmi and Zimmerman
2006). The Mohr - Coulomb failure criterion is based on two parts, one is constant cohesion
and another one is friction angle, and their relationship is shown in Equation (2.74). The
cohesion is based on the intercept of the failure envelope with the τ axis and the friction angle
based on the slope of the failure envelope (Coulomb 1973). The Mohr - Coulomb failure
criterion is a linear equation used for soil analysis (Labuz and Zang 2012), as shown in Figure
2.7, in which 1 and 3 are the major principal stress and minor principal stress, respectively.
The Mohr–Coulomb failure criterion data are based on a strength test in triaxial compression
(Bishop and Henkel 1962; Griffiths 1990). For the Mohr - Coulomb failure criterion, the Mohr
circles must be based on several different tests. The failure only depends on the major principal
stress and minor principal stress (Mohr 1900). Based on these two stresses, a series of Mohr
circles can be plotted in Figure 2.7. Thus, the Mohr - Coulomb failure envelope is the line
tangential to all the Mohr circles.
= 𝜎𝑛 𝑡𝑎𝑛 ϕc (2.74)
where
 represents the shear strength, n represents the normal stress, c is cohesion and ϕ is friction
angle
2.7 THE GENERALISED HOEK - BROWN FAILURE CRITERION
The Hoek - Brown failure criterion for rock slope analysis is a widely accepted method in the
geotechnical engineering field (Han and Leshchinsky 2004). Moreover, the Hoek - Brown
19
failure criterion is a non - linear failure criterion proposed and developed to estimate the heavily
jointed rock masses for slopes (Hoek and Brown 1997; Hammah et al. 2004). In 2002, Hoek et
al. introduced the disturbance factor (D) and developed the new relationships between D and
equations, as shown in Equations (2.75) to (2.78).
𝜎′
𝜎1′ = 𝜎3′ + 𝜎𝑐𝑖 ( 𝑚𝑏 𝜎3 + 𝑠)𝑎 (2.75)
𝑐𝑖
where
𝜎1′ represents the major effective principal stress at failure and 𝜎3′ is the minor effective
principal stress at failure, 𝜎𝑐𝑖 is the uniaxial compressive strength of the intact rock material,
mb is a reduced value of the material constant, s and a are all constants for the rock mass, the
relationships given by Equations (2.76) – (2.78).
GSI − 100
𝑚𝑏 = 𝑚𝑖 exp( ) (2.76)
28 − 14𝐷
GSI − 100
𝑠 = exp( ) (2.77)
9 − 3𝐷
1 1
𝑎= + (𝑒 −GSI/15 − 𝑒 −20/3 ) (2.78)
2 6
where
D represents the disturbance factor.
2.8 THE GEOLOGICAL STRENGTH INDEX
Rock mass classifications assess rock mass conditions for civil engineering and have drawn
the attention of many investigators (Bieniawski 1979; Bieniawski 1988; Marinos and Hoek
2000; Cai et al. 2004). Over the years, many rock mass classification systems, such as the
geomechanic classification rock mass rating by Bieniawski in 1973 and, the Q - system by
Barton et al. in 1974, have been developed based on tunnel design. However, Marinos et al.
(2005) reported that, the rock mass rating is difficult to apply in determining weak rock masses.
This problem causes a rock mass rating that cannot fit to the parameters based on the Hoek -
Brown failure criterion, particularly for weak rock masses. Therefore, the geological strength
index (GSI) was introduced by Hoek et al. (1992) and (1995). The GSI is a system used for
estimating the reduction in rock mass strength (Hoek and Brown 1997; Hoek et al. 1998). The
GSI system is now widely used for the estimation of the rock mass strength (Marinos et al.
20
2006), and it can be directly linked to engineering parameters, such as Mohr - Coulomb strength
parameters, Hoek - Brown strength parameters, or rock mass modulus (Cai et al. 2004). The
ranges of the GSI for typical rock masses are from 10 to 90 (Marinos and Hoek 2000). As a
rock - mass characterization system, the estimated GSI based on heterogeneous rock masses is
presented in Figure 2.8 (Marinos and Hoek 2001) and in Figure 2.9 for low quality rock masses
(Marinos and Hoek 2000; Hoek et al. 1998; Marinos and Hoek 2001). Since the introduction
of the GSI introduced, the Hoek - Brown failure criterion has been expanded and integrated
with such an index (Hoek and Brown 1997).
2.9 DISTURBANCE FACTOR
Based on the latest Hoek - Brown failure criterion (Hoek et al. 2002), recent investigations (Li
et al. 2011; Li et al. 2012) have shown that the D plays an important role in rock slope stability
evaluation. The range of D is between 0 and 1 (0 for undisturbed rock and 1 for highly disturbed
rock mass), which is based on the research of Hoek et al. (2002). The chosen D is based on
geotechnical experience, and the guidelines for estimating D are summarized in Figure 2.10
(Hoek et al. 2002). The two most important things considered when applying the D, are
choosing a suitable value and defining the extent of the damaged zone (Hoek 2012). For the
thickness of the damaged zone (T) design, Hoek (2012) highlighted that a constant D applied
to an entire rock mass could significantly underestimate the strength and stability of the overall
rock mass. Therefore, D should only be applied to the actual damaged zone. Based on the
research of Cheng and Liu (1993), the blast damaged zone can only surround the excavation
perimeter to a depth of approximately 2 meters. However, for large open pit mine slopes, the
blast damage also can be covering an area over 100 meters behind the slope excavated face can
also be observed (Marinos et al. 2005). In 2000, Hoek and Karzulovic suggested the
relationships between the T and slope height and used these relationships as a starting point to
refer the extent of the blast damaged zone, as shown in Table 2.6.
2.10 APPLICATION OF MOHR - COULOMB FAILURE CRITERION TO ROCK

MASS
Most rock geotechnical software are written based on the Mohr - Coulomb failure criterion (Li
et al. 2008; Qian et al. 2017). For the rock slope stability analysis, the rock mass strength has
to transfer to the cohesive strength (c ′ ) and the angle of friction (ϕ′) (Hoek and Brown 1997),
as shown in Equation (2.79). On the basis of the research carried out by Han and Leshchinsky
21
, the related determining equivalent c ′ and ϕ′ for the Mohr - Coulomb failure criterion are
′
shown in Equations (2.80) and (2.81). The value of 𝜎3𝑛 in Equations (2.80) and (2.81) is written
′
in Equation (2.82). In addition, the value of 𝜎3𝑚𝑎𝑥 is based on two individual cases, one is for
tunnels and another one is for slopes (Hoek et al. 2002; Sofianos 2003). For slope design, Hoek
′ ′
et al. suggested that 𝜎3𝑚𝑎𝑥 and 𝜎𝑐𝑚 can be estimated in Equations (2.83) and (2.84),
respectively.
= c ′ + 𝜎𝑛 𝑡𝑎𝑛 ϕ′ (2.79)
where
 is shear strength, c ′ is the cohesion, n the normal stress and ϕ′ is friction angle.
′
𝜎𝑐𝑖 [(1+2 𝑎)𝑠+(1−𝑎)𝑚𝑏 𝜎3𝑛 ′ )𝑎−1
](𝑠+ 𝑚𝑏 𝜎3𝑛
c′ = (2.80)
1+(6 𝑎 𝑚𝑏 (𝑠+𝑚𝑏 𝜎′3𝑛 )𝑎−1
(1+𝑎)(2+𝑎)√
(1+𝑎 )(2+𝑎)
𝑎−1
6 𝑎 𝑚 (𝑠+ 𝑚 𝜎 ′ )
ϕ′ = Sin−1 [ 2(1+ 𝑎)(2+𝑎)+6
𝑏 𝑏 3𝑛
𝑎 𝑚 (𝑠+ 𝑚 ′ 𝑎− 1 ] (2.81)
𝑏 𝑏 𝜎3𝑛 )
where
′
𝜎3𝑛 = 𝜎3𝑚𝑎𝑥
′
/ 𝜎𝑐𝑖 (2.82)
′ −0.91
𝜎3𝑚𝑎𝑥 𝜎′
𝑐𝑚
′ = 0.72 [ ϒH ] (2.83)
𝜎𝑐𝑚
𝑚
[𝑚𝑏 +4𝑠− 𝛼(𝑚𝑏 −8𝑠)]( 𝑏 +𝑠)𝛼−1
′
𝜎𝑐𝑚 = 𝜎𝑐𝑖 4
(2.84)
2 (1+𝛼)(2+𝛼)
where
𝜎𝑐𝑖 is the uniaxial compressive strength of the intact rock material, 𝑚𝑏 is a reduced value of the
′
material constant, s and a are constants for the rock mass, 𝜎𝑐𝑚 is the compressive strength of
the rock mass, H is the slope height, and ϒ is the slope material unit weight. In addition, the
stress range is 𝜎𝑡 < 𝜎3′ < 𝜎𝑐𝑖 /4
2.11. SLOPE STABILITY NUMBER AND CHARTS
The non - dimensional stability number is used to present the soil slope stability proposed
earlier by Lim et al. (2015). Its definition is shown in Equation (2.85) below. Another
22
dimensionless stability number based on rock parameters is defined in Equation (2.86) (Li et
al. 2008).
Equations (2.85) and (2.86) are modified based on the stability number proposed by Taylor in
1937. Soil and rock slope stability numbers are inversely proportional to the F. The stability
chart solutions can be seen as convenient tools to provide an easier way to determine the F of
a slope (Qian et al. 2014; Li et al. 2009). In relation to this, the slope stability charts have drawn
the attention of many investigators (Taylor 1937; Hoek and Bray 1981; Li et al. 2011, Qian et
al. 2014; Li et al. 2009; Michalowski 2010; Li et al. 2008; Qian et al. 2017; Li et al. 2010; Lim
et al. 2015).
𝑁2𝑐 = 𝑐𝑢1 ⁄γHF (2.85)
𝑁 = 𝜎𝑐𝑖⁄γHF (2.86)
In Equation (2.85) and Equation (2.86), γ is the unit of weight, H is the slope height, F is the
factor of safety, 𝑐𝑢1 is undrained shear strength for the filled soil, and 𝜎𝑐𝑖 is the uniaxial
compressive strength.
The major contributions of soil slope stability analysis were presented by Michalowski and his
co-worker, who following Taylor (1937), provided several sets of 2D and 3D stability charts
for various soil profiles. The effects of pore pressure seismic loadings, slope reinforcement,
and footing load on the slope stability were taken into account in their studies (Michalowski
1989; Michalowski 2002; Michalowski and Tabetha Martel 2011; Nadukuru and Michalowski
2013). Meanwhile, the literature review reveals that Kim et al. in 1999 and Loukidis et al. in
2003 proposed inhomogeneous soil slopes stability charts and cohesive frictional soil slopes
stability, respectively.
In addition, for rock slope design, only few stability charts are based on rock mass yield criteria
(Li et al. 2009). The first stability charts based on the Hoek - Brown failure criterion were
developed by Li et al. in 2008. A year later, the rock stability charts that considered seismic
force Were introduced by Li et al. (2009).
2.12 PREVIOUS STUDIES BASED ON THE PSEUDO STATIC METHOD
Seismic force has a significant effect on the stability of slopes (Ling et al. 1997). Therefore,
slope investigations considering seismic force are necessary. In recent decades, the seismic
behaviour has been the focus of many studies (Chang et al. 1984; Hongshuai et al. 2005;
23
Kramer and Smith 1997; Bray and Rathje 1998; Wilson and Keefer 1983; Chunling et al. 2004;
Shou and Wang 2003). Similarly, seismic slope stability charts have also drawn the attention
of many investigators (Li et al. 2009; Leshchinsky and San 1994; Baker et al. 2006)
The pseudo - static slope stability method (Seed 1979) is one of the most commonly used
methods in seismic slope design (Bray and Travasarou 2009). It has also been considered by
many researchers (Leshchinsky and San 1994; Chen and SAWADA 1983; Sarma 1975;
Choudhury and Nimbalkar 2005). In this method, the effect of an earthquake on a potential
sliding mass can be simplified as horizontal and vertical seismic coefficients (Li et al. 2009;
Bray and Travasarou 2009; Khazai and Sitar 2000), as shown in Figure 2.11. The horizontal
and vertical accelerations acting on a slope are considered constant as shown below:
𝑎ℎ = k h W (2.87)
𝑎𝑣 = k v W (2.88)
where
𝑎ℎ and 𝑎𝑣 are horizontal and vertical acceleration, k h and k v are horizontal and vertical seismic
coefficients, W is the potential sliding mass weight.
Li et al. reported that the horizontal seismic direction is the most recommended coefficient
during the slope stability design. The horizontal seismic coefficient (Kh) was developed for
embankments constructed with no significant strength loss during earthquake crest
accelerations of less than 0.75 g (Seed 1979). The recommended horizontal seismic values
range from 0 to 0.375 (DeWayne and Julie in 1997), as shown in Figure 2.12. The Kh is equal
to 0.1 for earthquakes with a Richter magnitude of 6.5, whereas the Kh is equal to 0.15 for
earthquakes with Richter magnitude of 8.5. Seed suggested that F ≥ 1.15 is required for
engineering design.
2.13 ARTIFICIAL NEURAL NETWORK
ANN has recently been used to estimate various geotechnical problems. Several slope stability
studies have utilized this method for slope stability prediction (Sakellariou and Ferentinou
2005; Abdalla et al. 2012; Gordan et al. 2016; Wang et al. 2005), rock slope stability analyses
(Li et al. 2016; Li et al. 2014), and slope stability analyses that consider seismic effects (Gordan
et al. 2016). Other studies have used ANN for landslide susceptibility analyses (Lu and
24
Rosenbaum 2003; Lee et al. 2004). The ANN is a mathematical model, which consists of a
network that simulates human brains (Agatonovic-Kustrin and Beresford 2000). It has the
ability to learn from a set of data, therefore, suitable in solving the uncertainty problem (Liang
and Zhang 2010; Melchiorre et al. 2008). In 2001, Shahin et al. proved that the results based on
ANN analysis may be better than those calculated from theoretical formulations. Moreover,
Sakellariou and Ferentinou (2005) showed that the F predicted by the ANN are reasonably
accurate.
The ANN can auto adjust its weights by training data (Basheer and Hajmeer 2000). Once the
training is successfully accomplished, the ANN can predict the output from a set of inputs.
Moreover, the results based on ANN can always be improved by new training (Shahin et al.
2001). The ANN has three main parts: input layer, hidden layer, and output layer (Choobbasti
et al. 2009; Li et al. 2016), as shown in Figure 2.13. In Figure 2.13, n inputs (x1, x2, x3 ⋯ xn) are
assumed and m outputs (y1, y2, y3 ⋯ ym) can be obtained. φ (.) is the activation transfer function
with M (1, 2 ⋯ l, M) linear nodes in the hidden layer, wij ( for i=1 ⋯ M and j=1 ⋯ n) are input
weights, and ij ( for i=1 ⋯ m and j=1 ⋯ M) are output weights. Therefore, the input data
vector 𝑥(𝑘) and the output data vector 𝑦(𝑘) can be expressed
𝑥(𝑘) = [ 𝑥1 (𝑘) 𝑥2 (𝑘) ⋯ 𝑥𝑛 (𝑘)]𝑇 (2.89)
𝑦(𝑘) = [ 𝑦1 (𝑘) 𝑦2 (𝑘) ⋯ 𝑦𝑚 (𝑘)]𝑇 (2.90)
The output is a function of the weights sum of the inputs, which is summarized in Equations
(2.91) and (2.92)
𝑜𝑖 = 𝜑[∑𝑛(𝑗=1) 𝑤𝑖𝑗 𝑥𝑗 (𝑘)] = 𝜑 (𝑤𝑖𝑇 𝑥(𝑘)) for i = 1, 2, ⋯ M (2.91)
𝑤𝑖 = [ 𝑤𝑖1 (𝑘) 𝑊𝑖2 (𝑘) ⋯ 𝑤𝑖𝑛 (𝑘)]𝑇 for i = 1, 2, ⋯ M (2.92)
For output of the neural network, 𝑦𝑖 (𝑘) can be expressed as:
𝑦𝑖 (𝑘) = ∑𝑀 𝑀 𝑇 𝑇
(𝑗=1) 𝛼𝑗𝑖 𝑜𝑗 = ∑(𝑗=1) 𝛼𝑗𝑖 𝜑 (𝑤𝑗 𝑥(𝑘)) = 𝝃 (𝑘)𝛼𝑖 for i = 1, 2, ⋯ m (2.93)
where
𝛼𝑖 = [𝛼1𝑖 𝛼2𝑖 , ⋯ 𝛼𝑀𝑖 ] for i = 1, 2, ⋯ m (2.94)
25
𝑇
𝝃(𝑘) = [𝜑(𝑤1𝑇 𝑥(𝑘)) 𝜑(𝑤2𝑇 𝑥(𝑘)) ⋯ 𝜑(𝑤𝑀
𝑇
𝑥(𝑘))] (2.95)
Based on Equations (2.93) to (2.95)
𝑦(𝑘) = [𝑦1 (𝑘)𝑦2 (𝑘) ⋯ 𝑦𝑚 (𝑘)]𝑇 = [𝝃𝑇 (𝑘)𝛼1 𝝃𝑇 (𝑘)𝛼2 ⋯ 𝝃𝑇 (𝑘)𝛼𝑚 ]𝑇 = [𝝃𝑇 (𝑘)𝛼 ]𝑇 (2.96)
where
𝛼 = [𝛼1 𝛼2 , ⋯ 𝛼𝑚 ] (2.97)
Beside, based on number of N training input vectors x(1), x (2) ⋯ x (N) and N of desired output
data vectors yd(1), yd(2) ⋯ yd(N).
𝜉 𝑇 (1) 𝑦 𝑇 (1)
𝐺𝑎 = [ ⋮ ] 𝛼 = [ ⋮ ] (2.98)
𝜉 𝑇 (𝑁) 𝑦 𝑇 (𝑁)
where
𝜑(𝑤1𝑇 𝑥(1)) ⋯ 𝑇
𝜑(𝑤𝑀 𝑥(1))
𝐺=[ ⋮ ⋱ ⋮ ] (2.99)
𝜑(𝑤1𝑇 𝑥(𝑁)) ⋯ 𝑇
𝜑(𝑤𝑀 𝑥(𝑁))
Where
G is the hidden layer output matrix. The input weights and the biases of the hidden layer can
be randomly assigned based on the extreme learning training machine training algorithm
(Huang et al. 2006). After the training process, the output weight matrix can then be computed
in a single iteration.
𝛼 = (𝐺 𝑇 𝐺) −1 𝐺 𝑇 𝑌𝑑 (2.100)
𝑦𝑑𝑇 𝑥(1)
𝑌𝑑 = [ ⋮ ] (2.101)
𝑦𝑑𝑇 𝑥(𝑁)
2.14 MESH ARRANGEMENT
Mesh refinement is improves numerical precision and efficiency due to the accuracy that is
directly related to the finite element mesh (Tschuchnigg et al. 2015). The finite element limit
analysis based on two different method. When the UB and LB techniques are employed, the
26
results can become increasingly reliable. However, the general mesh arrangement for a slope
with uniform rock messes may not yield satisfactory results. For steep slopes, the different
number of subdivisions can directly influence the results, as shown in Figure 2.14. Moreover,
the results obtained based on LB are less sensitive than those obtained based on UB (Loukidis
et al. 2003). The mesh may need to be redefined so that the true solutions can be bracketed
within a small range between upper bound and lower bound (Tschuchnigg et al. 2015).
However, conducting fine mesh analyses with a finer mesh can lead to a longer single case
computation time.
The study of Liu and Zhao (2012) shows that the results can be consistent if the fine mesh is
sufficient, as shown in Figure 2.15. In this research, the slope stability analysis convergence
refers to the maximum number of iterations, which is set as 3000. Formally, based on rock
slope stability analysis, particularly in seismic model, the mesh build may sometimes be unable
to converge within the assumed iteration. Therefore, for this particular case, a reliable
alternative mesh should be built based on trials and errors.
2.15 LIMITATIONS
For slope stability analysis, various methods have been investigated. However, each method
has its own limitations. For engineering design, understanding the limitations of each methods
is important. Therefore, the relative shortcomings based on each method are summarized
below:
Limit equilibrium method
The LEM requires the slope to be divided into many slices. Therefore, the potential slip surface
must be assumed before calculating the stability of slopes (Duncan 1996; Krahn 2003). The
inter column shear force also must be assumed for each slice (Bishop 2008; Janbu 1973); In
addition, this method in most case only be used for circular surfaces (Kaur and Sharma ).
Finite element method
For finite element method, the obtained F is based on strength reduction method (Griffiths and
Lane 1999). However, the present research shows that the strength reduction method is sensitive
and difficult to estimate (Cheng et al. 2007).
27
Limit analysis
The limitation for limit analysis is similarly to LEM, which requires the assumption of potential
slip surface (Gao et al. 2013). However, failure mechanisms are highly influenced by the slope
geometry and the soil properties (Gao et al. 2013), and the failure surface is difficult to assume.
Finite element limit analysis methods
For LA, the mesh arrangement influences the results, and case computation time is increased
with increased mesh refinement. In addition, the true displacement of the slope slide is difficult
to predict.
Artificial neural network
The ANN requires a large amount of input data, which can be difficult to debug. However, the
results based on probabilistic method do not have explanatory power. In other words, the entire
computation is in a black box once the data are trained (Basheer and Hajmeer 2000 ).
28
2.16 REFERENCES
Schmid A, Li A-J, Lim K and Nepal K (2016) Slope Stability Analysis Using Phase 2. Geo-
China 20162016. p. 151-7.
Lim K, Li A-J and Lyamin A (2014) Slope Stability Analysis for Filled Slopes Using Finite
Element Limit Analysis Method. Soil Behavior and Geomechanics2014. p. 615-24.
Cassidy MJ, Uzielli M and Lacasse S (2008) Probability risk assessment of landslides: A case
study at Finneidfjord. Can Geotech J2008. p. 1250-67.
Duncan JM, Brandon TL, Wright SG and Vroman N (2008) Stability of I-Walls in New Orleans
during Hurricane Katrina. Journal of Geotechnical and Geoenviromental Engineering2008. p.
681-91.
Li AJ, Cassidy MJ, Wang Y, Merifield RS and Lyamin AV (2012) Parametric Monte Carlo
studies of rock slopes based on the Hoek–Brown failure criterion. Computers and
Geotechnics2012. p. 11-8.
Lyamin AV and Sloan S (2002b) Upper bound limit analysis using linear finite elements and
non‐linear programming. International Journal for Numerical and Analytical Methods in
Geomechanics2002b. p. 181-216.
Krabbenhoft K, Lyamin AV, Hjiaj M and Sloan SW (2005) A new discontinuous upper bound
limit analysis formulation. International Journal for Numerical Methods in Engineering2005.
p. 1069-88.
Cheng Y, Lansivaara T and Wei W (2007) Two-dimensional slope stability analysis by limit
equilibrium and strength reduction methods. Computers and Geotechnics2007. p. 137-50.
Michalowski R (1995) Slope stability analysis: a kinematical approach. Geotechnique1995. p.

283-93.
Bishop AW (1955) The use of slip circle in stability analysis of slopes. Geotechnique1955. p.
7-17.
Morgenstern NR and Price VE (1965) The analysis of the stability of generalised slip surfaces.
29
Fredlund DG and Krahn J (1977) Comparison of slope stability methods of analysis. Can
Geotech J1977. p. 429-39.
Huang CC, Tsai CC and Chen YH (2002) Generalized method for three-dimensional slope
stability analysis. Journal of Geotechnical and Geoenviromental Engineering, ASCE2002. p.
836-48.
Petterson KE (1955) The early history of circular sliding surfaces. Geotechnique1955. p. 275-
96.
Bishop AW (1955) The use of the Slip Circle in the Stability Analysis of Slopes.
Géotechnique1955. p. 7-17.
Fellenius W (1936) editorêditors. Calculation of the stability of earth dams. Transactions of

the 2nd congress on large dams, Washington, DC; 1936: International Commission on Large
Dams (ICOLD) Paris.
Steward T, Sivakugan N, Shukla S and Das B (2010) Taylor’s slope stability charts revisited.
International Journal of Geomechanics2010. p. 348-52.
Zhu D, Lee C and Jiang H (2003) Generalised framework of limit equilibrium methods for
slope stability analysis. Geotechnique2003.
Chen R and Chameau J-L (1983) Three-dimensional limit equilibrium analysis of slopes.
Greco VR (1996) Efficient Monte Carlo technique for locating critical slip surface. Journal of
Geotechnical Engineering1996. p. 517-25.
Zhu D-Y (2001) A method for locating critical slip surfaces in slope stability analysis.
Canadian Geotechnical Journal2001. p. 328-37.
Cheng Y, Liang L, Chi S and Wei W (2008) Determination of the critical slip surface using
artificial fish swarms algorithm. Journal of geotechnical and geoenvironmental
engineering2008. p. 244-51.
30
Hajiazizi M and Tavana H (2013) Determining three-dimensional non-spherical critical slip

surface in earth slopes using an optimization method. Engineering Geology2013. p. 114-24.
Huang C-C and Tsai C-C (2000) New method for 3D and asymmetrical slope stability analysis.
Journal of Geotechnical and Geoenvironmental Engineering2000. p. 917-27.
Huang C-C, Tsai C-C and Chen Y-H (2002) Generalized method for three-dimensional slope
stability analysis. Journal of Geotechnical and Geoenvironmental Engineering2002. p. 836-48.
Lim K, Li A, Lyamin A and Cassidy M (2015) Slope stability analysis for fill slopes using
finite element limit analysis. 2015.
Yu H, Salgado R, Sloan S and Kim J (1998) Limit analysis versus limit equilibrium for slope
stability. Journal of Geotechnical and Geoenvironmental Engineering1998. p. 1-11.

283-93.
Donald IB and Chen Z (1997) Slope stability analysis by the upper bound approach:
fundamentals and methods. Canadian Geotechnical Journal1997. p. 853-62.
Farzaneh O and Askari F (2003) Three-dimensional analysis of nonhomogeneous slopes.

Michalowski RL and Nadukuru SS (2012) Three-dimensional limit analysis of slopes with pore
pressure. Journal of Geotechnical and Geoenvironmental Engineering2012. p. 1604-10.
Michalowski R (1989) Three-dimensional analysis of locally loaded slopes.

Chen Z, Wang X, Haberfield C, Yin J-H and Wang Y (2001a) A three-dimensional slope
stability analysis method using the upper bound theorem: Part I: theory and methods.
International Journal of Rock Mechanics and Mining Sciences2001a. p. 369-78.
Chen Z, Wang J, Wang Y, Yin J-H and Haberfield C (2001b) A three-dimensional slope
stability analysis method using the upper bound theorem Part II: numerical approaches,
applications and extensions. International Journal of Rock Mechanics and Mining
Sciences2001b. p. 379-97.
31
Michalowski RL (2010) Limit analysis and stability charts for 3D slope failures. Journal of
Geotechnical and Geoenvironmental Engineering2010. p. 583-93.
Michalowski RL and Martel T (2011) Stability charts for 3D failures of steep slopes subjected
to seismic excitation. Journal of geotechnical and geoenvironmental engineering2011. p. 183-
9.
Gao Y, Zhang F, Lei G, Li D, Wu Y and Zhang N (2012) Stability charts for 3D failures of
homogeneous slopes. Journal of Geotechnical and Geoenvironmental Engineering2012. p.
1528-38.
Gao Y, Zhang F, Lei G-H and Li D (2013) An extended limit analysis of three-dimensional
slope stability. Geotechnique2013. p. 518.
Leshchinsky B and Ambauen S (2015) Limit equilibrium and limit analysis: comparison of
benchmark slope stability problems. Journal of Geotechnical and Geoenvironmental
Engineering2015. p. 04015043.
Chen J, Yin J-H and Lee C (2004) Rigid finite element method for upper bound limit analysis
of soil slopes subjected to pore water pressure. Journal of engineering mechanics2004. p. 886-
93.
Lim K, Li A, Schmid A and Lyamin A (2016) Slope-stability assessments using finite-element

limit-analysis methods. International Journal of Geomechanics2016. p. 06016017.
Michalowski R and Drescher A (2009) Three-dimensional stability of slopes and excavations.

Courant R (1943) Variational method for solutions of problems of equilibrium and vibrations
l J J. Bul1 Am Math Soc1943. p. 1-23.
Dhatt G, Lefrançois E and Touzot G (2012) Finite element method: John Wiley & Sons; 2012.
Bathe K-J and Wilson EL (1976) Numerical methods in finite element analysis. 1976.
Hammah R (2005) editorêditors. A comparison of finite element slope stability analysis with
conventional limit-equilibrium investigation. Proceedings of the 58th Canadian Geotechnical
32
and 6th Joint IAH-CNC and CGS Groundwater Specialty Conferences–GeoSask 2005; 2005:
Citeseer.
Zienkiewicz OC and Taylor RL (2000) The finite element method: solid mechanics:
Butterworth-heinemann; 2000.
Zienkiewicz OC and Taylor RL (2005) The finite element method for solid and structural
mechanics: Butterworth-heinemann; 2005.
Potts D, Kovacevic N and Vaughan P (2009) Delayed collapse of cut slopes in stiff clay.
Selected papers on geotechnical engineering by PR Vaughan: Thomas Telford Publishing;
2009. p. 362-91.
Troncone A (2005) Numerical analysis of a landslide in soils with strain-softening behaviour.

Matsui T and San K-C (1992) Finite element slope stability analysis by shear strength reduction
technique. Soils and foundations1992. p. 59-70.
Hammouri NA, Malkawi AIH and Yamin MM (2008) Stability analysis of slopes using the
finite element method and limiting equilibrium approach. Bulletin of Engineering Geology and
the Environment2008. p. 471.
Ozbay A and Cabalar A (2015) FEM and LEM stability analyses of the fatal landslides at
Çöllolar open-cast lignite mine in Elbistan, Turkey. Landslides2015. p. 155-63.
Jianxun M, Zhisheng L, Qinge C and Zhenli X (2004) 3D FEM analysis of slope stability based
on strength reduction method [J]. Chinese Journal of Rock Mechanics and Engineering2004.
p. 005.
Zhao S-Y, Zheng Y-R and Zhang Y-F (2005) Study on slope failure criterion in strength
reduction finite element method [J]. Rock and Soil Mechanics2005.
Dawson E, Roth W and Drescher A (1999) Slope stability analysis by strength reduction.
33
Hammah R, Yacoub T, Corkum B and Curran J (2005) editorêditors. The shear strength
reduction method for the generalized Hoek-Brown criterion. Alaska Rocks 2005, The 40th US
Symposium on Rock Mechanics (USRMS); 2005: American Rock Mechanics Association.
Shangyi Z, Yingren Z and Weidong D (2003) Stability analysis on jointed rock slope by
strength reduction fem [J]. Chinese Journal of Rock Mechanics and Engineering2003. p. 020.
Sloan S (1988) Lower bound limit analysis using finite elements and linear programming.
International Journal for Numerical and Analytical Methods in Geomechanics1988. p. 61-77.
Lyamin A and Sloan S (2002b) Lower bound limit analysis using non‐linear programming.
International Journal for Numerical Methods in Engineering2002b. p. 573-611.
Sloan S (1989) Upper bound limit analysis using finite elements and linear programming.
Sloan S (2013) Geotechnical stability analysis. Géotechnique2013. p. 531-71.
Bottero A, Negre R, Pastor J and Turgeman S (1980) Finite element method and limit analysis
theory for soil mechanics problems. Computer Methods in Applied Mechanics and
Li A, Merifield R and Lyamin A (2008) Stability charts for rock slopes based on the Hoek–
Brown failure criterion. International Journal of Rock Mechanics and Mining Sciences2008. p.
689-700.
Sutcliffe D, Yu H and Sloan S (2004) Lower bound solutions for bearing capacity of jointed
rock. Computers and Geotechnics2004. p. 23-36.
Merifield RS, Lyamin AV and Sloan S (2006) Limit analysis solutions for the bearing capacity
of rock masses using the generalised Hoek–Brown criterion. International Journal of Rock
Mechanics and Mining Sciences2006. p. 920-37.
Soubra A-H (1999) Upper-bound solutions for bearing capacity of foundations. Journal of
34
Chai J and Carter JP (2009) Simulation of the progressive failure of an embankment on soft
soil. Computers and Geotechnics2009. p. 1024-38.
Huang J, Lyamin A, Griffiths D, Krabbenhoft K and Sloan S (2013) Quantitative risk

assessment of landslide by limit analysis and random fields. Computers and Geotechnics2013.
p. 60-7.
Shiau JS, Merifield R, Lyamin AV and Sloan SW (2011) Undrained Stability of Footings on
Slopes. International Journal of Geomechanics2011. p. 381-90.
Lysmer J (1970) Limit analysis of plane problems in soil mechanics. Journal of Soil Mechanics
& Foundations Div1970.
Donald IB and Chen ZY (1997) Slope stability analysis by the upper bound approach:
fundamentals and methods. Can Geotech J1997. p. 853-62.
Chen J, Yin JH and Lee CF (2003) Upper bound limit analysis of slope stability using rigid
finite elements and nonlinear programming. Can Geotech J2003. p. 742-52.
Sloan S and Kleeman P (1995) Upper bound limit analysis using discontinuous velocity fields.
Computer Methods in Applied Mechanics and Engineering1995. p. 293-314.
Chen J, Yin J-H and Lee C (2003) Upper bound limit analysis of slope stability using rigid
finite elements and nonlinear programming. Canadian Geotechnical Journal2003. p. 742-52.
Kim J, Salgado R and Lee J (2002) Stability analysis of complex soil slopes using limit
analysis. Journal of Geotechnical and Geoenvironmental Engineering2002. p. 546-57.
Gao Y, Zhang F, Lei GH, Li D, Wu Y and Zhang N (2013) Stability charts for 3D failures of
homogeneous slopes. Journal of Geotechnical and Geoenviromental Engineering2013. p.
1528-38.
Anderheggen E and Knöpfel H (1972) Finite element limit analysis using linear programming.
International Journal of Solids and Structures1972. p. 1413-31.
Pastor J and Turgeman S (1976) Mise en oeuvre numérique des méthodes de l'analyse limite
pour les matériaux de von Mises et de Coulomb standards en déformation plane. Mechanics
Research Communications1976. p. 469-74.
35
Yu H, Sloan S and Kleeman P (1994) A quadratic element for upper bound limit analysis.
Engineering Computations1994. p. 195-212.
Jiang GL (1995) Non‐linear finite element formulation of kinematic limit analysis.

International journal for numerical methods in engineering1995. p. 2775-807.
Liu Y, Cen Z and Xu B (1995) A numerical method for plastic limit analysis of 3-D structures.
International Journal of Solids and Structures1995. p. 1645-58.
Vicente da Silva M and Antao A (2007) A non‐linear programming method approach for upper
bound limit analysis. International Journal for Numerical Methods in Engineering2007. p.
1192-218.
Duncan JM, Wright SG and Brandon TL (2014) Soil strength and slope stability: John Wiley
& Sons; 2014.
Tang W, Yucemen M and Ang A-S (1976) Probability-based short term design of soil slopes.
Li K and Lumb P (1987) Probabilistic design of slopes. Canadian Geotechnical Journal1987.

p. 520-35.
Jewell R and Greenwood J (1988) Long term strength and safety in steep soil slopes reinforced
by polymer materials. Geotextiles and Geomembranes1988. p. 81-118.
Powers RF, Scott DA, Sanchez FG, Voldseth RA, Page-Dumroese D, Elioff JD, et al. (2005)
The North American long-term soil productivity experiment: findings from the first decade of
research. Forest Ecology and Management2005. p. 31-50.
Coulomb C (1973) ESSAI SUR UNE APPLICATION DES REGLES DE MAXIMIS ET

MINIMIS A QUELQUES PROBLEMES DE STATIQUE RELATIFS A L'ARCHITECTURE
(ESSAY ON MAXIMUMS AND MINIMUMS OF RULES TO SOME STATIC PROBLEMS
RELATING TO ARCHITECTURE). 1973.
Bai Y and Wierzbicki T (2010) Application of extended Mohr–Coulomb criterion to ductile

fracture. International Journal of Fracture2010. p. 1.
36
Al-Ajmi AM and Zimmerman RW (2006) Stability analysis of vertical boreholes using the
Mogi–Coulomb failure criterion. International Journal of Rock Mechanics and Mining
Sciences2006. p. 1200-11.
Labuz JF and Zang A (2012) Mohr–Coulomb failure criterion. The ISRM Suggested Methods
for Rock Characterization, Testing and Monitoring: 2007-2014: Springer; 2012. p. 227-31.
Bishop AW and Henkel DJ (1962) Measurement of soil properties in the triaxial test. 1962.
Griffiths D (1990) Failure criteria interpretation based on Mohr-Coulomb friction. Journal of

Geotechnical Engineering1990. p. 986-99.
Mohr O (1900) Welche Umstände bedingen die Elastizitätsgrenze und den Bruch eines
Materials. Zeitschrift des Vereins Deutscher Ingenieure1900. p. 1572-7.
Hoek E and Brown E (1997) Practical estimates of rock mass strength. International Journal of
Rock Mechanics and Mining Sciences1997. p. 1165-86.
Hammah RE, Curran JH, Yacoub T and Corkum B (2004) editorêditors. Stability analysis of
rock slopes using the finite element method. Proceedings of the ISRM Regional Symposium
EUROCK; 2004: Citeseer.

Bieniawski Z (1979) editorêditors. The geomechanics classification in rock engineering

applications. 4th ISRM Congress; 1979: International Society for Rock Mechanics.
Bieniawski Z (1988) The rock mass rating (RMR) system (geomechanics classification) in
engineering practice. Rock Classification Systems for Engineering Purposes: ASTM
International; 1988.
37
Marinos P and Hoek E (2000) editorêditors. GSI: a geologically friendly tool for rock mass
strength estimation. ISRM International Symposium; 2000: International Society for Rock
Mechanics.
Cai M, Kaiser P, Uno H, Tasaka Y and Minami M (2004) Estimation of rock mass deformation
modulus and strength of jointed hard rock masses using the GSI system. International Journal
of Rock Mechanics and Mining Sciences2004. p. 3-19.
Bieniawski Z (1973) Engineering classification of jointed rock masses. Civil Engineer in South
Africa1973.
Barton N, Lien R and Lunde J (1974) Engineering classification of rock masses for the design
of tunnel support. Rock mechanics1974. p. 189-236.
Marinos V, Marinos P and Hoek E (2005) The geological strength index: applications and
limitations. Bulletin of Engineering Geology and the Environment2005. p. 55-65.
Hoek E, Wood D and Shah S (1992) editorêditors. A modified Hoek–Brown failure criterion
for jointed rock masses. Rock Characterization: ISRM Symposium, Eurock'92, Chester, UK,
14–17 September 1992; 1992: Thomas Telford Publishing.
Hoek E, Kaiser PK and Bawden WF (1995) Support of underground excavations in hard rock.
1995.
Hoek E, Marinos P and Benissi M (1998) Applicability of the Geological Strength Index (GSI)
classification for very weak and sheared rock masses. The case of the Athens Schist Formation.
Bulletin of Engineering Geology and the Environment1998. p. 151-60.
Marinos P, Hoek E and Marinos V (2006) Variability of the engineering properties of rock
masses quantified by the geological strength index: the case of ophiolites with special emphasis
on tunnelling. Bulletin of Engineering Geology and the Environment2006. p. 129-42.
Marinos P and Hoek E (2000) editorêditors. GSI: a geologically friendly tool for rock mass
strength estimation. Proc GeoEng2000 Conference, Melbourne; 2000.
Marinos P and Hoek E (2001) Estimating the geotechnical properties of heterogeneous rock
masses such as flysch. Bulletin of engineering geology and the environment2001. p. 85-92.
38
Hoek E and Brown ET (1997) Practical estimates of rock mass strength. International Journal
of Rock Mechanics and Mining Sciences1997. p. 1165-86.
p. 11-8.
Hoek E (2012) Blast Damage Factor D. Technical note for RocNews2012.
Cheng Y and Liu S (1993) Power caverns of the Mintang pumped storage project, Taiwan,
Comprehensive Rock Engineering. Pergamon1993. p. 111-31.
Hoek E and Karzulovic A (2000) Rock mass properties for surface mines. Slope Stability in
Surface Mining, WA Hustrulid, MK McCarter and DJA van Zyl, Eds, Society for Mining,
Metallurgical and Exploration (SME), Littleton, CO2000. p. 59-70.

Sofianos A (2003) Tunnelling Mohr–Coulomb strength parameters for rock masses satisfying
the generalized Hoek–Brown criterion. International Journal of Rock Mechanics and Mining
Sciences2003. p. 435-40.
Lim K, Li AJ and Lyamin AV (2015) Three-dimensional slope stability assessment of two-

layered undrained clay. Computers and Geotechnics2015. p. 1-17.
Taylor DW (1937) Stability of earth slopes. J Boston Soc Civ Engrs1937. p. 197-246.
Qian Z, Li AJ, Merifield R and Lyamin AV (2014) Slope stability charts for two-layered purely
39
Li AJ, Lyamin AV and Merifield R (2009) Seismic rock slope stability charts based on limit
Taylor DW (1937) Stability of earth slopes: Boston Society of Civil Engineers Boston; 1937.
Hoek E and Bray JW (1981) Rock slope engineering. 3rd ed. London: Institute of Mining and
Metallurgy; 1981.
Li AJ, Merifield RS and Lyamin AV (2011) Effect of rock mass disturbance on the stability of
rock slopes using the Hoek-Brown failure criterion. Computers and Geotechnics2011. p. 546-
58.
Li A-J, Merifield RS and Lyamin AV (2010) Three-dimensional stability charts for slopes
based on limit analysis methods. Canadian Geotechnical Journal2010. p. 1316-34.
Lim K, Lyamin A, Cassidy M and Li A-J (2015) Three-dimensional slope stability charts for
Michalowski RL (1989) Three-dimensional analysis of locally loaded slopes.

Geoenviromental Engineering, ASCE2002. p. 351-5.
ASCE2011. p. 183-9.
Nadukuru SS and Michalowski RL (2013) Three-dimensional displacement analysis of slopes

subjected to seismic loads. Can Geotech J2013. p. 650-61.
Kim J, Salgado R and Yu HS (1999) Limit analysis of soil slopes subjected to pore- water
pressures. Journal of Geotechnical and Geoenviromental Engineering, ASCE1999. p. 49-58.
Loukidis D, Bandini P and Salgado R (2003) Stability of seismically loaded slopes using limit
analysis. Geotechnique2003. p. 463-79.
40
Ling HI, Leshchinsky D and Mohri Y (1997) Soil slopes under combined horizontal and
vertical seismic accelerations. Earthquake engineering & structural dynamics1997. p. 1231-41.
Chang C-J, Chen WF and Yao JT (1984) Seismic displacements in slopes by limit analysis.
Hongshuai L, Jingshan B and Dedong L (2005) Review on study of seismic stability analysis
of rock-soil slopes [J]. Earthquake Engineering and Engineering Vibration2005.
Kramer SL and Smith MW (1997) Modified Newmark model for seismic displacements of
compliant slopes. Journal of Geotechnical and Geoenvironmental Engineering1997. p. 635-44.
Bray JD and Rathje EM (1998) Earthquake-induced displacements of solid-waste landfills.

Wilson RC and Keefer DK (1983) Dynamic analysis of a slope failure from the 6 August 1979
Coyote Lake, California, earthquake. Bulletin of the Seismological Society of America1983.
p. 863-77.
Chunling L, Shengwen Q, Liqiang T and Fasuo Z (2004) Stability analysis of slope under
earthquake with flac~(3d)[J]. Chinese Journal of Rock Mechanics and Engineering2004. p.
014.
Shou K-J and Wang C-F (2003) Analysis of the Chiufengershan landslide triggered by the 1999
Chi-Chi earthquake in Taiwan. Engineering Geology2003. p. 237-50.
Leshchinsky D and San K-C (1994) Pseudostatic seismic stability of slopes: Design charts.
Journal of geotechnical engineering1994. p. 1514-32.
Seed HB (1979) Considerations in the earthquake-resistant design of earth and rockfill dams.
41
Bray JD and Travasarou T (2009) Pseudostatic coefficient for use in simplified seismic slope
stability evaluation. Journal of geotechnical and geoenvironmental engineering2009. p. 1336-
40.
Chen WF and SAWADA T (1983) Earthquake-induced slope failure in nonhomogeneous,

anisotropic soils. Soils and Foundations1983. p. 125-39.
Sarma SK (1975) Seismic stability of earth dams and embankments. Geotechnique1975. p.

743-61.
Choudhury D and Nimbalkar S (2005) Seismic passive resistance by pseudo-dynamic method.

GEOTECHNIQUE-LONDON-2005. p. 699.
Khazai B and Sitar N (2000) Assessment of seismic slope stability using GIS modeling.
Geographic Information Sciences2000. p. 121-8.
DeWayne H and Julie M (1997) Guidelines for Evaluating and Mitigating Seismic Hazards in
California. California: The Public Information Offices of the California Geological Survey;
1997.
Sakellariou M and Ferentinou M (2005) A study of slope stability prediction using neural
networks. Geotechnical & Geological Engineering2005. p. 419.
Abdalla JA, Attom M and Hawileh R (2012) editorêditors. Artificial neural network
prediction of factor of safety of slope stability of soils. Proceedings of the 14th International
Conference on Computing in Civil and Building Engineering; 2012.
Gordan B, Armaghani DJ, Hajihassani M and Monjezi M (2016) Prediction of seismic slope
stability through combination of particle swarm optimization and neural network. Engineering
with Computers2016. p. 85-97.
Wang H, Xu W and Xu R (2005) Slope stability evaluation using back propagation neural
networks. Engineering Geology2005. p. 302-15.
42
Li A, Khoo S, Wang Y and Lyamin A (2014) editorêditors. Application of neural network to

rock slope stability assessments. Proceedings of the 8th European Conference on Numerical
Methods in Geotechnical Engineering (NUMGE 2014); 2014.
Agatonovic-Kustrin S and Beresford R (2000) Basic concepts of artificial neural network

(ANN) modeling and its application in pharmaceutical research. Journal of pharmaceutical and
biomedical analysis2000. p. 717-27.
Liang H and Zhang H (2010) editorêditors. Identification of slope stability based on the
contrast of BP Neural Network and SVM. Computer Science and Information Technology
(ICCSIT), 2010 3rd IEEE International Conference on; 2010: IEEE.
Melchiorre C, Matteucci M, Azzoni A and Zanchi A (2008) Artificial neural networks and
cluster analysis in landslide susceptibility zonation. Geomorphology2008. p. 379-400.
Basheer I and Hajmeer M (2000) Artificial neural networks: fundamentals, computing, design,
and application. Journal of microbiological methods2000. p. 3-31.
Choobbasti A, Farrokhzad F and Barari A (2009) Prediction of slope stability using artificial
neural network (case study: Noabad, Mazandaran, Iran). Arabian journal of geosciences2009.
p. 311-9.
Li AJ, Khoo S, Lyamin AV and Wang Y (2016) Rock slope stability analyses using extreme
Huang G-B, Chen L and Siew CK (2006) Universal approximation using incremental
constructive feedforward networks with random hidden nodes. IEEE Trans Neural
Networks2006. p. 879-92.
Tschuchnigg F, Schweiger H and Sloan S (2015) Slope stability analysis by means of finite
element limit analysis and finite element strength reduction techniques. Part I: Numerical
studies considering non-associated plasticity. Computers and Geotechnics2015. p. 169-77.
43
Liu F and Zhao J (2012) Limit analysis of slope stability by rigid finite-element method and
linear programming considering rotational failure. International Journal of Geomechanics2012.
p. 827-39.
Journal of Geotechnical engineering1996. p. 577-96.
Krahn J (2003) The 2001 RM Hardy Lecture: The limits of limit equilibrium analyses.
Bishop AW (2008) The use of the slip circle in the stability analysis of slopes. The Essence of
Geotechnical Engineering: 60 years of Géotechnique: Thomas Telford Publishing; 2008. p.
223-33.
Janbu N (1973) Slope stability computations, Embankment dam engineering Casagrande. John
Wiley and Sons, NY; 1973.
Kaur A and Sharma R SLOPE STABILITY ANALYSIS TECHNIQUES: A REVIEW.
p. 387-403.
Bishop A (1955) The Use of the Slip Circle in the Stability Analysis of Slopes, Geoteohnique,
VOl. V, PPo7" l701955.
Janbu N, Bjerrum L and Kjaernsli B (1956) Soil mechanics applied to some engineering
problems. Norwegian Geotechnical Institute Publication1956.
Morgenstern N and Price VE (1965) The analysis of the stability of general slip surfaces. 1965.
Spencer E (1967) A method of analysis of the stability of embankments assuming parallel inter-
slice forces. Geotechnique1967. p. 11-26.
Fredlund D and Krahn J (1977) Comparison of slope stability methods of analysis. Canadian
Geotechnical Journal1977. p. 429-39.
Fredlund D, Krahn J and Pufahl D (1981) editorêditors. The relationship between limit
equilibrium slope stability methods. Proceedings of the International Conference on Soil
Mechanics and Foundation Engineering; 1981.
44
Leshchinsky D (1990) Slope stability analysis: generalized approach. Journal of geotechnical

Leshchinsky D and Huang C-C (1992) Generalized three-dimensional slope-stability analysis.

Journal of geotechnical engineering1992. p. 1748-64.
Zhu D, Lee C, Qian Q, Zou Z and Sun F (2001) A new procedure for computing the factor of
safety using the Morgenstern-Price method. Canadian Geotechnical Journal2001. p. 882-8.
Baker R (2003) A second look at Taylor’s stability chart. Journal of geotechnical and
Ugai K (1985) Three-dimensional stability analysis of vertical cohesive slopes. Soils and
Foundations1985. p. 41-8.
Cavoundis S (1987) On the ratio of factors of safety in slope stability analyses.

Geotechnique1987.
Hungr O (1987) An extension of Bishop's simplified method of slope stability analysis to three
dimensions. Geotechnique1987. p. 113-7.
Gens A, Hutchinson J and Cavounidis S (1988) Three-dimensional analysis of slides in

cohesive soils. Geotechnique1988. p. 1-23.
Leshchinsky D and Huang C-C (1992a) Generalized three-dimensional slope-stability analysis.

Journal of geotechnical engineering1992a. p. 1748-64.
Leshchinsky D and Huang C-c (1992b) Generalized slope stability analysis: Interpretation,
modification, and comparison. Journal of geotechnical engineering1992b. p. 1559-76.
Lam L and Fredlund D (1993) A general limit equilibrium model for three-dimensional slope
stability analysis. Canadian Geotechnical Journal1993. p. 905-19.
Stark TD and Eid HT (1998) Performance of three-dimensional slope stability methods in

practice. Journal of Geotechnical and Geoenvironmental engineering1998. p. 1049-60.
Janbu N (1973) Slope stability computations, embankment-dam engineering, vol Casagrande.

Wiley, New York; 1973.
45
and differential straining of block contacts. Canadian geotechnical journal2002. p. 799-811.
Xie M, Esaki T and Cai M (2006) GIS-based implementation of three-dimensional limit

equilibrium approach of slope stability. Journal of geotechnical and geoenvironmental
Cheng Y and Yip C (2007) Three-Dimensional asymmetrical slope stability analysis extension
of Bishop’s, Janbu’s, and Morgenstern–Price’s techniques. Journal of geotechnical and
Lane P and Griffiths D (2000) Assessment of stability of slopes under drawdown conditions.
Journal of geotechnical and geoenvironmental engineering2000. p. 443-50.
Lechman J and Griffiths D (2000) Analysis of the progression of failure of earth slopes by
finite elements. Slope Stability 20002000. p. 250-65.
Manzari MT and Nour MA (2000) Significance of soil dilatancy in slope stability analysis.
Zheng H, Liu D and Li C (2005) Slope stability analysis based on elasto-plastic finite element
method. International Journal for Numerical Methods in Engineering2005. p. 1871-88.
Wei W, Cheng Y and Li L (2009) Three-dimensional slope failure analysis by the strength
reduction and limit equilibrium methods. Computers and geotechnics2009. p. 70-80.
Michalowski RL (1997) Stability of uniformly reinforced slopes. Journal of Geotechnical and

geoenvironmental Engineering1997. p. 546-56.
Kim J, Salgado R and Yu H (1999) Limit analysis of soil slopes subjected to pore-water
pressures. Journal of Geotechnical and Geoenvironmental Engineering1999. p. 49-58.
Zhang X (1999) Slope stability analysis based on the rigid finite element method.
46
Viratjandr C and Michalowski RL (2006) Limit analysis of submerged slopes subjected to

water drawdown. Canadian Geotechnical Journal2006. p. 802-14.
Chen J, Yin J-H and Lee C (2005) A three-dimensional upper-bound approach to slope stability
analysis based on RFEM. Geotechnique2005. p. 549-56.
Li A, Merifield R and Lyamin A (2009) Limit analysis solutions for three dimensional
undrained slopes. Computers and Geotechnics2009. p. 1330-51.
47
2.17 APPENDIX
Table 2.1. Summary table for the 2D limit equilibrium method
2 D Limit Equilibrium Method

Author/Year Comments Charts
Simplified method of slices / Satisfying moment equilibrium only
Bishop 1955 No
Assumption of no inter-column shear force and slip surface
Satisfying force equilibrium
Janbu et al. 1956 No
Assumption of a correction factor for inter-column shear force
Satisfying both moment and force equilibrium
Morgenstern and
Assumption on inter-column shear force as a function and slip No
Price 1965
surface
Satisfying both moment and force equilibrium
Spencer 1967 Consider inter column shear force as a constant function and slip No
surface
Fredlund and
Compilation of existing limit equilibrium method No
Krahn 1977
Fredlund et al.
Derivation of general limit equilibrium method No
1981
No static assumption required
Leshchinsky 1990 No
Assumption of slip surface
Leshchinsky and
Generalized from Leshchinsky (1990) No
Huang 1992
No static assumption required
Leshchinsky and
Assumption of failure mechanism and Consideration of seismic No
San 1994
loading
Consideration of seismic loading both vertical and horizontal
Ling et al. 1997 No
Assumption of failure mechanism
Zhu et al. 2001 Modified Morgenstern and Price (1965) to inter-column forces No
Baker 2003 Modified Taylor (1937) to locate critical slip surface No
Krahn 2003 Discussion of limitation and consider finite element method No
Baker et al. 2006 Consideration of seismic loading Yes
48
Table 2.2. Summary table for the 3D limit equilibrium method
3D limit equilibrium method

Ugai 1985 Assumption of failure surface to be split cylinders with curved ends No
Cavoundis
On the note that 3D factor of safety is higher than 2D factor of safety No
1987
Hungr 1987 Extension of Bishop (1955) No
For purely cohesive soils and Assumption of rotational slides
Gens et al. Extension of Fellenius (1936)
Yes
1988 Supports the fact that 3D factor of safety is higher than 2D factor of
safety
Leshchinsky
and Huang Extension of Leshchinsky and Huang (1992b) No
1992a
Lam and
Extension of Fredlund and Krahn (1977) and Fredlund et al. (1981) No
Fredlund 1993
Stark and Eid
Supports the idea that 2D underestimates the factor of safety No
1998
Huang and Extension of Bishop (1955), Consideration of asymmetrical slope
No
Tsai 2000 Assumption of semi spherical failure surface and failure mechanism
Huang et al. Modified and generalized from Huang and Tsai (2000)
No
2002 Extension of Janbu (1973)
Chang 2002;
Assumption of sliding mass as block system No
Xie et al. 2006
Incorporation of Geographical Information System into slope stability
Xie et al. 2006 No
analysis to improve slope spatial properties input
Asymmetrical extension of Bishop (1955) and Janbu (1973)
Cheng and Yip
Assumption of unique sliding direction as opposed to Huang and Tsai No
2007
(2000) and Huang et al. (2002)
49
Table 2.3. Summary table for the finite element method
Finite element method

Author / Year Comments Charts
Verify the use of Strength Reduction Method
Matsui and
Comparison of the results by Fellenius (1936) to justify Strength No
San 1992
Reduction Method
Potts et al.
Study of progressive failure No
2009
No prior assumption required
Griffiths and Application of Strength Reduction Method and non-convergence criteria
No
Lane 1999 to determine failure
Non-convergence could be due to several factors other than failure
Lane and
Griffiths Slope stability analysis under draw down conditions Yes
2000
Lechman and
Griffiths Study of progressive failure with incremental gravity No
2000
Manzari and
Study the effect of soil dilatancy on slope stability analysis No
Nour 2000
Troncone
Study of progressive failure No
2005
Zheng et al. Study the effect of iteration limit and selection of Poisson ratio and
No
2005 found that factor of safety could be underestimated
Wei et al. State that the Strength Reduction Method can be sensitive to soft band
No
2009 with frictional soil and convergence criteria
50
Table 2.4. Summary table for the 2D finite element limit analysis
2D finite element limit analysis

Donald and
Upper bound theorem for randomly generated slip surface No
Chen 1997
Michalowski Upper bound theorem for uniformly reinforced slopes and assumed
Yes
1997 failure surface
Upper and lower bound theorem
Yu et al. 1998 No
Study the application of finite element limit analysis
Kim et al.
Finite element limit analysis consider with pore pressure Yes
1999
Rigid finite element limit analysis based on lower bound theorem.
Zhang 1999 Assumed circular slip surface No
Division of soil mass into slices and wedges or arbitrary polyhedral
Upper bound theorem
Michalowski
Consideration of pore pressure/seismic loading Yes
2002
Assumed collapse mechanism and failure surface
Chen et al. Upper bound theorem finite element limit analysis
No
2003 Assumed failure surface
Loukidis et al. Finite element limit analysis based on both upper and lower bound
No
2003 theorem consider seismic loading
Chen et al. Upper bound finite element limit analysis consider pore pressure
No
2004 Assumed collapse mechanism
Viratjandr and
Upper bound theorem for slope under water drawdown
Michalowski Yes
Assumed collapse mechanism and failure surface
2006
51
Table 2.5. Summary table for the 3D finite element limit analysis
3D finite element limit analysis

Author and Year Comments Charts
Upper bound theorem
Study the limit load of locally loaded slope
Michalowski 1989 Yes
Chart solution for selective parameters
Failure mass assumed to be rigid block
Extension of Donald and Chen (1997)
Chen et al. 2001a; Upper bound theorem
No
Chen et al. 2001b Assumed slip surface
Failure mass in the form of prism
Farzaneh and Askari Extension of Michalowski (1989) to consider non loaded slope
No
2003 Upper bound theorem
Upper bound theorem
Chen et al. 2005 Long computational time No
Assumed slip surface
Li et al. 2009 Finite element limit analysis Yes
Li et al. 2010 Upper bound theorem Yes
Michalowski 2010 Assumed collapse mechanism Yes
Table 2.6. Relationships between the blast damaged zone thicknesses and slope height (Hoek and
Karzulovic 2000)
Large production blast, confined and with little or no control D = 2 to 2.5 H

Production blast with no control but blasting to a free face D = 1 to 1.5 H
Production blast, confined but with some control, e.g. one or more
D = 1 to 1.2 H
buffer rows
Production blast with some control, e.g. one or more buffer rows, and
D = 0.5 to 1 H
blasting to a free face
Carefully controlled production blast with a free face D = 0.3 to 0.5 H
52
Figure 2.1. The slope model and forces based on the research of Krahn in 2003
Figure 2.2. The different types of element based on lower bound analysis
Figure 2.3. The model of stress discontinuity in lower bound analysis
53
Figure 2.4. The stresses sign convention
Figure 2.5. The stress boundary conditions
Figure 2.6. The finite element upper bound limit analysis methods
54
Figure 2.7. The linear equation for Mohr - Coulomb presented graphically
Figure 2.8. The geological strength index chart for heterogeneous rock masses
(Marinos and Hoek 2001)
55
Figure 2.9. The Geological strength index classification based on poor quality
rock masses (Marinos et al. 2005)
56
Figure 2.10. The guidelines for estimating disturbance factor (Hoek et al. 2002)
57
Figure 2.11. The seismic rock slope model based pseudo static method
Figure 2.12. The recommended horizontal seismic values (DeWayne and Julie
1997)
Figure 2.13. Artificial neural networks based on single hidden layer neural
network (Li et al. 2016)
58
(a) Fine mesh (b) Cause mesh
Figure 2.14. The mesh arrangement
Figure 2.15. The relationship between results and mesh refine ( Liu and Zhao
2012)
59
CHAPTER 3 SLOPE STABILITY CHARTS FOR TWO-LAYERED

PURELY COHESIVE SOILS BASED ON THE FINITE ELEMENT
LIMIT ANALYSIS METHODS
Z. G. Qian, A. J. Li, R.S. Merifield, A.V. Lyamin
3.1 ABSTRACT
Stability charts for soil slopes, first produced in the first half of the twentieth century, continue
to be used extensively as design tools and draw the attention of many investigators. This paper
uses finite-element upper and lower bound limit analysis to assess the short term stability of
slopes in which the slope material and subgrade foundation material have two distinctly
different undrained strengths. The stability charts are proposed, and the exact theoretical
solutions are bracketed to within 4.2% or better. In addition, results from the limit equilibrium
method (LEM) have been used for comparison. Differences of up to 20% were found between
the numerical limit analysis and LEM solutions. It also shown that the LEM sometimes leads
to errors, although it is widely used in practice for slope stability assessments.
3.2 INTRODUCTION
The stability problem of natural slopes, filled slopes (such as embankments, earth dams and
levees), or cut slopes are commonly encountered in civil engineering projects. Predicting the
slope stability is an everyday task for geotechnical engineers. The combined effects of geology,
hydrology, and soil properties generally exist as slope stability issues. Because of its practical
importance, slope stability analysis has drawn the attention of many investigators (Duncan
1996; Chang 2002; Baker et al. 2006; Michalowski and Tabetha Martel 2011).
The first set of slope stability charts were proposed by Taylor (1937) and continue to be used
extensively as simple design and verification tools (Hoek and Bray 1981; Li et al. 2011). More
sophisticated numerical modelling is still needed during design, especially for large scale
complex projects. For practicing engineers, however, performing numerical simulations is not
always warranted or feasible because of time and cost constraints. Therefore, stability charts
can be seen as convenient tools to provide an easier way to determine the factor of safety for a
slope.
60
Generally speaking, the limit equilibrium method (LEM) is the most popular approach to
evaluate slope stability problems (Cassidy et al. 2008; Duncan et al. 2008). However, it is
known that the best solutions cannot be guaranteed by using the LEM (Li et al. 2012). As
indicated by Li et al. (2009a), the factor of safety was found to be overestimated (>10%)
compared with the upper bound (UB) solutions. because the slip surface needs to be assumed
before any safety factor calculation. However, use of finite element limit analysis methods
(Lyamin and Sloan 2002a, b; Krabbenhoft et al. 2005) can avoid this drawback, as the failure
mechanism is generated automatically.
In addition, using limit theorems not only can provide a simple and useful way of analysing
the stability of geotechnical structures, but also avoids the shortcomings of the arbitrary
assumptions underpinning the LEM. In this study, the numerical UB and lower bound (LB)
limit analysis methods developed by Lyamin and Sloan (2002a, b) and Krabbenhoft et al.
(2005) are used to produce stability charts by considering two - layered purely cohesive soils.
Recently, these techniques have been applied to several two- and three-dimensional (2D and
3D) slope stability investigations in soils (Yu et al. 1998; Loukidis et al. 2003; Li et al. 2010;
2009b). It should be noted that the slopes composed by two-layered purely cohesive soils are
common features for levee projects (Duncan et al. 2008); therefore, it is worth having this
thorough study. Comprehensive chart solutions are provided as well.
3.3 PREVIOUS STUDIES
The LEM is one of the most popular methods to assess the stability of slopes (Han and
Leshchinsky 2004). Currently, LEM has been applied to both 2D (Bishop 1955; Morgenstern
and Price 1965; Fredlund and Krahn 1977) and 3D (Chang 2002; Huang et al. 2002) slope
stability analyses. In addition to Taylor’s charts (Taylor 1937), Gens et al. (1988) produced
chart solutions by considering 3D boundary effects for cohesive soils. Stability charts for
cohesive-frictional slopes were proposed by Jiang and Yamagami (2006), who accounted for
the stability of the long slopes. In addition, Koppula (1984) and Baker et al. (2006) adopted the
pseudo static (PS) method in limit equilibrium analysis and proposed 2D seismic chart
solutions for purely cohesive and cohesive frictional soil slopes.
It is known, however, that in using LEM, the potential slip surface must be assumed before
calculating the factor of safety (F) for the slope. Moreover, arbitrary assumptions need to be
made regarding forces between two slices. Because of these assumptions, the results are often
61
questioned. In particular, in some cases the slip surface must be assumed based on experience
or judgement.
The finite element method (FEM) is also a popular approach used for slope analyses. As
highlighted by Duncan (1996), FEM is an useful tool to calculate stresses, movements, pore
pressure, and other characteristics of earth masses during construction without previously
assuming the potential sliding surface. To estimate the slope stability and obtain its factor of
safety by using finite element analysis, the strength reduction method (SRM) is widely used
(Griffiths and Lane 1999; Griffiths and Marquez 2007). However, the failure load is determined
subjectively, generally based on observation of the slope displacement (Hoek et al. 2000;
Manzari and Nour 2000). In fact, FEM is rarely used to perform slope stability charts, because
it is not time effective.
In past decades, most slope stability studies have been based only on the UB method (Donald
and Chen 1997; Michalowski 2002; Chen et al. 2003; Han et al. 2012; Gao et al. 2013). Likely
because of the difficulty in manually constructing statically admissible stress fields for the LB
method. Major contributions for soil slope stability analysis were presented by Michalowski
and co-workers (Michalowski 1989; Michalowski 2002; Michalowski and Tabetha Martel
2011; Nadukuru and Michalowski 2013), who provided several sets of 2D and 3D stability
charts for various soil profiles. The effects of pore pressure seismic loadings, slope
reinforcement, and footing load on slope stability were taken into account in their studies. It
should be stated that when using the UB or LB method alone, the true solution cannot be
bracketed.
For the slope stability evaluations using both UB and LB limit analysis methods, Shiau et al.
(2011) investigated the effects of external loading on undrained slopes. Moreover, Kim et al.
(1999) and Loukidis et al. (2003) proposed sets of stability charts for nonhomogeneous soil
slopes and cohesive-frictional soil slopes subjected to pore pressure and seismic loadings
respectively. Recently, numerical UB and LB techniques (Lyamin and Sloan 2002a, b;
Krabbenhoft et al. 2005) have been applied to 3D slope (Li et al. 2009b, 2010) and trench
stability problems (Li et al. 2014).
According to the preceding discussion, a set of chart solutions for slopes having a cohesive soil
filled on another purely cohesive soil do not exist. The aim of this paper is to provide stability
charts, produced using numerical UB and LB limit analysis methods (Lyamin and Sloan 2002a,
62
b; Krabbenhoft et al. 2005), as convenient tools for assessing the safety factor for slopes that
have this feature.
3.4 PROBLEM DEFINITION
To simplify the problem, only plane strain cases are considered in this study. The illustration
of the slope stability problem investigated herein is shown in Figure 3.1. The slope geometry
analysed for the filled slopes with two purely cohesive layered soils can be seen. It should be
noted that the undrained shear strength in Region 1 (cu1) and Region 2 (cu2) are different. In
this study, for given slope height (H), slope angle () and undrained shear strength (cu1 and
cu2), the optimized solutions of the UB and LB programs can be carried out with respect to the
unit weight, . In this paper, a range of slope inclinations ( = 15° - 75°) and depth factor (d/H
= 1.5 - 5) are taken into account. It should be noted that a d/H ratio of one is meaningless, as
the case is homogeneous slope which has been investigated by Taylor (1937).
For the UB theorem, the power dissipated by any kinematically admissible velocity field can
be equated to the power dissipated by the external loads to give a rigorous UB on the true limit
load (Lyamin and Sloan 2002b). The LB theorem states that the admissible stress field must
fulfil equilibrium, the stress boundary condition, and yield conditions (Lyamin and Sloan
2002a). For 2D limit analysis modelling, the mesh generation must follow two important
guidelines: (1) the overall mesh dimensions are adequate to contain the computed stress field
(LB) or velocity/plastic field (UB); and (2) there is an adequate concentration of elements
within critical regions. The final finite element mesh arrangements (both UB and LB) were
selected only after considerable refinements were made. The typical finite element meshes and
boundary conditions of the UB and LB limit analysis are displayed in Figure 3.2 (a and b),
respectively.
To present the results as chart solutions, a dimensionless stability number (N2c) is defined in
Equation (3.1)
N2c = cu1 ⁄γHF (3.1)
which is modified based on the stability number proposed by Taylor (1937). In addition, the
chart solutions are presented by considering the ratio of cu1/cu2 that ranges from 0.2 to 5. This
covers most problems of practical interest (Merifield et al. 1999). The solutions obtained from
LEM based on Bishop’s simplified method (Bishop 1955) are used for comparison.
63
3.5 RESULTS AND DISCUSSIONS
The computed UB and LB estimates of the stability numbers N2c for the slope with a purely
cohesive soil filled on another purely cohesive soil are given in Table 3.1 - 3.5 and shown
graphically in Figure 3.3 - 3.7. These results indicate that, for practical design purposes,
sufficiently small error bounds were achieved. The stability numbers are bracketed within 8%
or better. Because the obtained unit weight for LB is smaller, a LB calculation actually gives
an UB on the stability numbers, and vice versa.
In addition, the solutions obtained from the LEM are also illustrated in Figure 3.3 - 3.7, shown
as solid lines. For most cases, the stability numbers from the LEM are obviously smaller than
those from the numerical UB and LB limit analysis methods. In general, the LEM solutions are
closer to the UB results, implying that the unsafe results could be provided by the LEM. For
some cases, differences in N2c of up to 20 % can be achieved between the LEM and LB, which
is significant. Even compared with the UB solutions, the discrepancies are still 14 %. As stated
by Yu et al. (1998), LB results can be applied to designs straightforwardly because they are
conservative solutions. This discussion implies that the LEM would not always be adequate to
deal with slope stability problems with multiple soil layers.
In Figure 3.3 - 3.7, it can be observed that N2c increases with increasing slope inclination ()
and the ratios of d/H and cu1/cu2 when cu1/cu2 ≥ 1.5. The failure surfaces can be obtained by
observing the plastic zones from the UB solutions. Based on the comprehensive observations,
the slope failure mode is of base failure when cu1/cu2 ≥ 1.5. In addition, there is no significant
shape change in the failure mechanism for different cu1/cu2 ratios. However, while cu1/cu2 ≤ 0.8,
N2c is almost a constant (flat line) for various d/H ratios. The exceptions are the slopes that
have lower slope inclination ( ≤ 30°). To explain this phenomenon, the failure surfaces are
observed more thoroughly.
Based on the observations of the slopes for  ≥ 45°, the failure surface is found to be controlled
by the soil in Region 2 which can be seen as a rigid layer. Because only the soil in Region 1
plays the significant role for N2c, it should be a constant for various d/H ratios. However, this
phenomenon does not exist for  ≤ 30°. Figure 3.8 shows the plastic zones from the UB
solutions for  = 15° and d/H = 1.5. In Figure 3.8, the failure surfaces for cu1/cu2 ≤ 0.57 only
touch the top of the soil in Region 2 which is different from those for cu1/cu2 = 0.66. As shown
in Figure 3.8 (c), the slip surface passes through both of the soils in Regions 1 and 2 for cu1/cu2
64
= 0.66. Figure 3.9 displays the failure surfaces for  = 15°, cu1/cu2 = 0.5, and various d/H ratios.
An interesting phenomenon is found that the depth of failure surface increases as d/H
increasing. The soil masses slip along two soil boundary when d/H ≤ 2. However, the failure
mechanism is of base failure for d/H ≥ 3. It should be noted that in Figure 3.9 (b) the plastic
zone appears in both the base failure mode and the interface between two soils.
The plastic zones for  = 30°, cu1/cu2 = 0.66, and various d/H ratios are shown in Figure 3.10.
A trend similar to that in Figure 3.9 is observed, in that the depth of failure surface increases
as d/H increasing. In Figure 3.10, the failure mechanism involves both of the soils in Regions
1 and 2 for d/H = 4 and 5 as demonstrated by the velocities shown in Figure 3.10 (b and c).
Figure 3.8 – 3.10 show the transitions of the slope failure mode for various , cu1/cu2, and d/H
values. Because of the change of the failure mode, N2c is not always a constant. These findings
explain why N2c is not a flat line when  ≤ 30° and cu1/cu2 ≤ 0.8.
Some of the failure mechanisms (solid lines) obtained from the LEM are also displayed in
Figure 3.8 – 3.10. Because unsafe solutions will be obtained using LEM, the circular failure
mode would not be the most critical for the slope that combines two different clays. In addition,
Figure 3.10 is a special case in which the slip surfaces from the LEM are found to be unchanged
when d/H is changed from 3.3 to 3.5. This is different from the UB results, in which the depth
of the slip surface increases with increasing d/H.
3.6 APPLICATION EXAMPLE
A filled slope design project, Case 1, has H = 5m,  = 18 kN/m3, d/H = 2, cu1 = 50 kN/m2, and
cu1/cu2 = 4.0. Based on the preceding information, cu1/H = 50 / (5×18) = 0.555 can be obtained.
Using the stability number in Table 3.1 – 3.5, F can be simply calculated as 0.555/N2c for
various slope angles. The results are shown in Table 3.6, and therefore  can be designed based
on a desired magnitude of F. It can be seen in Table 3.6 that the slope will be close to critical
state if  ≥ 60°. By comparing F from three different methods, it can be observed that F(LEM)
is slightly less sensitive when  is changed from 30° to 75°. This phenomenon is also found
when d/H ≥ 2 and cu1/cu2 ≥ 2. In addition, Table 3.6 also shows that F(LEM) > F(UB) > F(LB),
indicating that LEM can lead to a non-conservative design. Because F is calculated based on
Equation (3.1), the differences in F between LEM and numerical limit analysis methods are
also the differences in N2c, which are shown in Table 3.1 – 3.5.
65
Moreover, the chart solutions can be used to design the slope height. For example, Case 2 has
similar information to Case 1; however, the slope height is unknown and  = 30°. From Table
3.1 - 3.5, N2c is 0.495, 0.478 and 0.461 for LB, UB and LEM, respectively. The calculate ratios
of cu1/H for different slope heights are displayed in Table 3.7. Therefore, F can be obtained
based on the previous information. The results of F, which can be seen in Table 3.7, indicate
that the slope will be critical when H ≥ 6 m. Again, the user can make decisions for the slope
height based on the desired magnitude of F.
Because the ratio of cu1/cu2 is 4.5 for the above cases, it should be noted that the slope failure
mechanism pass through both soils in Region 1 and Region 2 (Figure 3.1) which has been
discussed previously. Therefore, it is interesting to investigate a case if the failure surface could
involve only the soil in Region 1. The ratio of cu1/cu2 = 0.8 and cu1 = 40 kN/m2 are assumed for
Case 3. Other required inputs are the same as Case 1. The ratio cu1/H = 40 / (5×18) = 0.444 is
thus calculated. The obtained results of F for different slope inclinations are shown in Table
3.8, where the stability numbers (N2c) are from Table 3.1 - 3.5.
Compared with the results in Table 3.6, it can be found that F (LB) for cu1/cu2 = 4.5 is smaller
than that for cu1/cu2 = 0.8, although cu1 has decreased from 50 to 40 kN/m2. In addition, the
influence of  on the safety factor is found to be less significant for Case 1. The difference in
F(LB) between  = 15° and 75° is 28%, as shown in Table 3.6. However, the discrepancy is
approximately 79% for Case 3 when  = 15° is changed to  = 75°. Based on detailed
observation, the influence of  on the safety factor is more significant if the slope failure mode
regards only the soil in Region 1. However, there is no obvious trend that can quantify the
influence of  on the safety factor for various cu1/cu2 ratios.
3.7 CONCLUSION
This paper uses finite element UB and LB limit analysis methods to analyse filled slopes
composed of two different purely cohesive soils. In addition, limit equilibrium analysis
(Bishop’s simplified method (1955)) is adopted for comparison. In this study, the stability
numbers, N2c, are presented using chart solutions. The true solutions are bracketed within 8%
or better. By observing the UB plastic zones, the majority slope failure mode is base failure
when cu1/cu2 ≥ 1.5. Toe failure is noticed for cu1/cu2 ≤ 0.8 and  ≥ 45°. This study found that the
depth of failure surface increases with increasing d/H when cu1/cu2 ≤ 0.8 and the slope has a
66
smaller inclination ( ≤ 30°). Also, the predetermined circular slip surface in LEM is not always
adequate to assess the filled slopes investigated in this study.
It should be noted that LB solutions are more valuable than UB and LEM solutions, as they
can be used in practice directly to give a safe design. The results from the LEM are generally
smaller than those from numerical limit analysis methods. Although LEM can give reasonable
solutions for most cases investigated in this study, it still can lead to an overestimation of safety
factor for slope designs; sometimes the difference can be up to 20%. Therefore, engineers
should use LEM carefully when it is applied to a slope with layered soils.
67
3.8 REFERENCES
Journal of Geotechnical Engineering, ASCE1996. p. 577-96.
stability analysis. Soils dynamics and earthquake engineering2006. p. 813-23.
ASCE2011. p. 183-9.
Taylor DW (1937) Stability of earth slopes. J Boston Soc Civ Engrs1937. p. 197-246.
Hoek E and Bray JW (1981) Rock slope engineering. 3rd ed. London: Institute of Mining and
Metallurgy; 1981.
Li AJ, Merifield RS and Lyamin AV (2011) Effect of rock mass disturbance on the stability of
rock slopes using the Hoek-Brown failure criterion. Computers and Geotechnics2011. p. 546-
58.
681-91.
Li AJ, Lyamin AV and Merifield RS (2009a) Seismic rock slope stability charts based on limit
analysis methods. Computers and Geotechnics2009a. p. 135-48.
68
Lyamin AV and Sloan SW (2002a) Lower bound limit analysis using non-linear programming.
International Journal for Numerical Methods in Engineering2002a. p. 573-611.
p. 1069-88.
Lyamin AV and Sloan SW (2002b) Upper bound limit analysis using linear finite elements and
non-linear programming. International Journal for Numerical and Analytical Methods in
Geomechanics2002b. p. 181-216.
Yu HS, Salgado R, Sloan SW and Kim JM (1998) Limit analysis versus limit equilibrium for
slope stability. Journal of Geotechnical and Geoenviromental Engineering, ASCE1998. p. 1-
11.
Li AJ, Merifield RS and Lyamin AV (2010) Three dimensional stability charts for slopes based
on limit analysis methods. Can Geotech J2010. p. 1316-34.
Li A, Merifield R and Lyamin A (2009b) Limit analysis solutions for three dimensional
undrained slopes. Computers and Geotechnics2009b. p. 1330-51.
based numerical method for analyzing stability of MSE walls. 17th ASCE Engineering
Mechanics Conference; 2004 13-16 June; University of Delaware, Newark, DE.
Bishop AW (1955) The use of slip circle in stability analysis of slopes. Geotechnique1955. p.
7-17.
Morgenstern NR and Price VE (1965) The analysis of the stability of generalised slip surfaces.
Fredlund DG and Krahn J (1977) Comparison of slope stability methods of analysis. Can
Geotech J1977. p. 429-39.
69
Huang CC, Tsai CC and Chen YH (2002) Generalized method for three-dimensional slope
stability analysis. Journal of Geotechnical and Geoenviromental Engineering, ASCE2002. p.
836-48.
Gens A, Hutchinson JN and Cavounidis S (1988) Three-dimensional analysis of slides in

Jiang J-C and Yamagami T (2006) Charts for estimating strength parameters from slips in
homogeneous slopes. Computers and Geotechnics2006. p. 294-304.
Koppula SD (1984) Pseudo-static analysis of clay slopes subjected to earthquake.

Griffiths DV and Lane PA (1999) Slope stability analysis by finite elements.

Griffiths DV and Marquez RM (2007) Three-dimensional slope stability analysis by elasto-

plastic finite elements. Geotechnique2007. p. 537-46.
Hoek E, Read J, Karzulovic A and Chenn ZY (2000) editorêditors. Rock slopes in civil and
mining engineering. Proceedings of the International Conference on Geotechnical and
Geological Engineering; 2000; Melbourne.
Manzari MT and Nour MA (2000) Significance of soil dilatance in slope stability analysis.
Journal of Geotechnical and Geoenviromental Engineering, ASCE2000. p. 75-80.
Geoenviromental Engineering, ASCE2002. p. 351-5.
Han C-Y, Wang J-H, Xia X-H and Chen J-J (2012) Limit analysis for local and overall stability
of slurry trench in cohesive soil. Internation Journal for Geomechanics2012. p.
doi:10.1061/(ASCE)GM.943-5622.0000268.
70
1528-38.
Michalowski RL (1989) Three-dimensional analysis of locally loaded slopes.

Nadukuru SS and Michalowski RL (2013) Three-dimensional displacement analysis of slopes

subjected to seismic loads. Can Geotech J2013. p. 650-61.
Shiau JS, Merifield R, Lyamin AV and Sloan SW (2011) Undrained Stability of Footings on
Slopes. International Journal of Geomechanics2011. p. 381-90.
Kim J, Salgado R and Yu HS (1999) Limit analysis of soil slopes subjected to pore- water
pressures. Journal of Geotechnical and Geoenviromental Engineering, ASCE1999. p. 49-58.
Li AJ, Merifield RS and Lyamin AV (2009b) Limit analysis solutions for three dimensional
undrained slopes. Computers and Geotechnics2009b. p. 1330-51.
Li AJ, Merifield R, Lin HD and Lyamin AV (2014) The trench stability under bentonite
pressure in purely cohesive clay. International Journal of Geomechanics2014. p. 151-7.
Merifield RS, Sloan SW and Yu HS (1999) Rigorous plasticity solutions for the bearing
capacity of two-layered clays. Geotechnique1999. p. 471-90.
Bishop A (1955) The Use of the Slip Circle in the Stability Analysis of Slopes, Geoteohnique,
VOl. V, PPo7" l701955.
71
3.9 APPENDIX
Table 3.1. Stability numbers for  = 15o
F difference between F difference between

d/H cu1/cu2 N2c (LB) N2c (UB) N2c (LEM)
LEM and LB (%) LEM and UB (%)
0.2 0.090 0.087 0.082 9.76 6.10
0.25 0.090 0.087 0.083 8.43 4.82
0.33 0.090 0.087 0.082 9.76 6.10
0.4 0.090 0.087 0.082 9.76 6.10
0.5 0.090 0.087 0.083 8.43 4.82
0.57 0.090 0.087 0.083 8.43 4.82
0.66 0.095 0.092 0.087 9.20 5.75
0.8 0.110 0.106 0.107 2.80 -0.93
1.5
1.5 0.171 0.164 0.164 4.27 0.00
2 0.204 0.195 0.185 10.27 5.41
2.5 0.231 0.220 0.209 10.53 5.26
3 0.254 0.242 0.234 8.55 3.42
3.5 0.274 0.260 0.243 12.76 7.00
4 0.291 0.276 0.258 12.79 6.98
4.5 0.307 0.290 0.271 13.28 7.01
5 0.322 0.303 0.281 14.59 7.83
0.2 0.090 0.087 0.084 7.14 3.57
0.25 0.090 0.087 0.084 7.14 3.57
0.33 0.090 0.087 0.085 5.88 2.35
0.4 0.090 0.087 0.085 5.88 2.35
0.5 0.090 0.087 0.085 5.88 2.35
0.57 0.094 0.092 0.092 2.17 0.00
0.66 0.107 0.104 0.104 2.88 0.00
0.8 0.126 0.123 0.123 2.44 0.00
2
1.5 0.208 0.203 0.204 1.96 -0.49
2 0.257 0.249 0.250 2.80 -0.40
2.5 0.298 0.289 0.290 2.76 -0.34
3 0.334 0.324 0.324 3.09 0.00
3.5 0.366 0.354 0.354 3.39 0.00
4 0.394 0.381 0.381 3.41 0.00
4.5 0.420 0.406 0.404 3.96 0.50
5 0.446 0.429 0.425 4.94 0.94
0.2 0.090 0.086 0.083 8.43 3.61
0.25 0.090 0.086 0.083 8.43 3.61
0.33 0.090 0.086 0.083 8.43 3.61
3
0.4 0.090 0.086 0.083 8.43 3.61
0.5 0.090 0.089 0.089 1.12 0.00
0.57 0.101 0.100 0.100 1.00 0.00
72
0.66 0.116 0.115 0.115 0.87 0.00

0.8 0.138 0.137 0.137 0.73 0.00
1.5 0.241 0.238 0.237 1.69 0.42
2 0.307 0.302 0.299 2.68 1.00
2.5 0.367 0.361 0.355 3.38 1.69
3 0.424 0.416 0.407 4.18 2.21
3.5 0.476 0.467 0.455 4.62 2.64
4 0.525 0.514 0.498 5.42 3.21
4.5 0.572 0.559 0.538 6.32 3.90
5 0.616 0.602 0.574 7.32 4.88
0.2 0.090 0.086 0.084 7.14 2.38
0.25 0.090 0.086 0.084 7.14 2.38
0.33 0.090 0.086 0.084 7.14 2.38
0.4 0.090 0.086 0.084 7.14 2.38
0.5 0.092 0.091 0.090 2.22 1.11
0.57 0.104 0.103 0.102 1.96 0.98
0.66 0.119 0.118 0.117 1.71 0.85
0.8 0.143 0.142 0.140 2.14 1.43
4
1.5 0.255 0.252 0.249 2.41 1.20
2 0.329 0.325 0.319 3.13 1.88
2.5 0.399 0.393 0.385 3.64 2.08
3 0.466 0.458 0.445 4.72 2.92
3.5 0.529 0.520 0.502 5.38 3.59
4 0.590 0.580 0.555 6.31 4.50
4.5 0.650 0.637 0.605 7.44 5.29
5 0.708 0.693 0.651 8.76 6.45
0.2 0.090 0.086 0.086 4.65 0.00
0.25 0.090 0.086 0.086 4.65 0.00
0.33 0.090 0.086 0.086 4.65 0.00
0.4 0.090 0.086 0.086 4.65 0.00
0.5 0.093 0.093 0.091 2.20 2.20
0.57 0.106 0.105 0.104 1.92 0.96
0.66 0.121 0.121 0.119 1.68 1.68
0.8 0.146 0.145 0.142 2.82 2.11
5
1.5 0.263 0.260 0.254 3.54 2.36
2 0.341 0.338 0.333 2.40 1.50
2.5 0.417 0.412 0.406 2.71 1.48
3 0.489 0.484 0.469 4.26 3.20
3.5 0.559 0.552 0.532 5.08 3.76
4 0.627 0.619 0.595 5.38 4.03
4.5 0.693 0.684 0.652 6.29 4.91
5 0.758 0.747 0.703 7.82 6.26
73
Table 3.2. Stability numbers for  = 30°

0.2 0.134 0.128 0.114 17.54 12.28
0.25 0.134 0.128 0.115 16.52 11.30
0.33 0.134 0.128 0.114 17.54 12.28
0.4 0.134 0.128 0.115 16.52 11.30
0.5 0.134 0.128 0.115 16.52 11.30
0.57 0.134 0.128 0.114 17.54 12.28
0.66 0.134 0.128 0.119 12.61 7.56
0.8 0.142 0.139 0.139 2.16 0.00
1.5
1.5 0.220 0.212 0.213 3.29 -0.47
2 0.261 0.251 0.252 3.57 -0.40
2.5 0.294 0.282 0.284 3.52 -0.70
3 0.322 0.308 0.309 4.21 -0.32
3.5 0.346 0.329 0.330 4.85 -0.30
4 0.366 0.348 0.352 3.98 -1.14
4.5 0.384 0.365 0.367 4.63 -0.54
5 0.402 0.380 0.376 6.91 1.06
0.2 0.134 0.127 0.114 17.54 11.40
0.25 0.134 0.127 0.114 17.54 11.40
0.33 0.134 0.127 0.114 17.54 11.40
0.4 0.134 0.127 0.114 17.54 11.40
0.5 0.134 0.127 0.114 17.54 11.40
0.57 0.134 0.127 0.114 17.54 11.40
0.66 0.134 0.127 0.120 11.67 5.83
0.8 0.147 0.144 0.142 3.52 1.41
2
1.5 0.241 0.236 0.234 2.99 0.85
2 0.298 0.289 0.287 3.83 0.70
2.5 0.346 0.336 0.331 4.53 1.51
3 0.389 0.377 0.371 4.85 1.62
3.5 0.428 0.414 0.405 5.68 2.22
4 0.463 0.447 0.435 6.44 2.76
4.5 0.495 0.478 0.461 7.38 3.69
5 0.527 0.507 0.485 8.66 4.54
0.2 0.134 0.127 0.111 20.72 14.41
0.25 0.134 0.127 0.111 20.72 14.41
0.33 0.134 0.127 0.111 20.72 14.41
0.4 0.134 0.127 0.111 20.72 14.41
3 0.5 0.134 0.127 0.111 20.72 14.41
0.57 0.134 0.127 0.111 20.72 14.41
0.66 0.134 0.127 0.122 9.84 4.10
0.8 0.150 0.148 0.145 3.45 2.07
1.5 0.260 0.256 0.250 4.00 2.40
74
2 0.330 0.324 0.315 4.76 2.86

2.5 0.395 0.388 0.374 5.61 3.74
3 0.456 0.447 0.429 6.29 4.20
3.5 0.513 0.503 0.478 7.32 5.23
4 0.569 0.556 0.523 8.80 6.31
4.5 0.622 0.608 0.565 10.09 7.61
5 0.674 0.658 0.603 11.77 9.12
0.2 0.134 0.126 0.114 17.54 10.53
0.25 0.134 0.126 0.114 17.54 10.53
0.33 0.134 0.126 0.114 17.54 10.53
0.4 0.134 0.126 0.114 17.54 10.53
0.5 0.134 0.126 0.114 17.54 10.53
0.57 0.134 0.126 0.114 17.54 10.53
0.66 0.134 0.127 0.122 9.84 4.10
0.8 0.152 0.151 0.145 4.83 4.14
4
1.5 0.268 0.265 0.256 4.69 3.52
2 0.345 0.341 0.329 4.86 3.65
2.5 0.418 0.413 0.396 5.56 4.29
3 0.487 0.482 0.459 6.10 5.01
3.5 0.554 0.548 0.517 7.16 6.00
4 0.620 0.612 0.572 8.39 6.99
4.5 0.684 0.675 0.624 9.62 8.17
5 0.747 0.736 0.672 11.16 9.52
0.2 0.134 0.126 0.116 15.52 8.62
0.25 0.134 0.126 0.116 15.52 8.62
0.33 0.134 0.126 0.116 15.52 8.62
0.4 0.134 0.126 0.116 15.52 8.62
0.5 0.134 0.126 0.115 16.52 9.57
0.57 0.134 0.126 0.116 15.52 8.62
0.66 0.134 0.127 0.122 9.84 4.10
0.8 0.153 0.152 0.146 4.79 4.11
5
1.5 0.273 0.271 0.261 4.60 3.83
2 0.354 0.351 0.338 4.73 3.85
2.5 0.431 0.427 0.411 4.87 3.89
3 0.505 0.501 0.479 5.43 4.59
3.5 0.578 0.572 0.544 6.25 5.15
4 0.648 0.642 0.605 7.11 6.12
4.5 0.718 0.710 0.663 8.30 7.09
5 0.787 0.778 0.718 9.61 8.36
75

0.2 0.167 0.162 0.152 9.87 6.58
0.25 0.167 0.162 0.152 9.87 6.58
0.33 0.167 0.162 0.150 11.33 8.00
0.4 0.167 0.162 0.152 9.87 6.58
0.5 0.167 0.162 0.152 9.87 6.58
0.57 0.167 0.162 0.151 10.60 7.28
0.66 0.167 0.162 0.151 10.60 7.28
0.8 0.167 0.162 0.152 9.87 6.58
1.5
1.5 0.237 0.231 0.226 4.87 2.21
2 0.281 0.272 0.267 5.24 1.87
2.5 0.316 0.306 0.300 5.33 2.00
3 0.346 0.333 0.328 5.49 1.52
3.5 0.371 0.356 0.347 6.92 2.59
4 0.393 0.376 0.366 7.38 2.73
4.5 0.413 0.393 0.386 6.99 1.81
5 0.435 0.409 0.399 9.02 2.51
0.2 0.167 0.159 0.151 10.60 5.30
0.25 0.167 0.159 0.151 10.60 5.30
0.33 0.167 0.159 0.150 11.33 6.00
0.4 0.167 0.159 0.150 11.33 6.00
0.5 0.167 0.159 0.150 11.33 6.00
0.57 0.167 0.159 0.151 10.60 5.30
0.66 0.167 0.159 0.150 11.33 6.00
0.8 0.167 0.159 0.152 9.87 4.61
2
1.5 0.253 0.247 0.241 4.98 2.49
2 0.310 0.303 0.294 5.44 3.06
2.5 0.360 0.351 0.343 4.96 2.33
3 0.405 0.394 0.383 5.74 2.87
3.5 0.445 0.432 0.420 5.95 2.86
4 0.481 0.466 0.448 7.37 4.02
4.5 0.517 0.498 0.480 7.71 3.75
5 0.558 0.531 0.500 11.60 6.20
0.2 0.167 0.158 0.152 9.87 3.95
0.25 0.167 0.158 0.150 11.33 5.33
0.33 0.167 0.158 0.151 10.60 4.64
0.4 0.167 0.158 0.152 9.87 3.95
3 0.5 0.167 0.158 0.151 10.60 4.64
0.57 0.167 0.158 0.151 10.60 4.64
0.66 0.167 0.158 0.152 9.87 3.95
0.8 0.167 0.159 0.152 9.87 4.61
1.5 0.267 0.263 0.252 5.95 4.37
76
2 0.338 0.333 0.320 5.63 4.06

2.5 0.404 0.397 0.378 6.88 5.03
3 0.465 0.457 0.435 6.90 5.06
3.5 0.524 0.514 0.485 8.04 5.98
4 0.580 0.568 0.532 9.02 6.77
4.5 0.635 0.621 0.572 11.01 8.57
5 0.688 0.672 0.612 12.42 9.80
0.2 0.168 0.158 0.151 11.26 4.64
0.25 0.168 0.158 0.151 11.26 4.64
0.33 0.168 0.158 0.151 11.26 4.64
0.4 0.168 0.158 0.152 10.53 3.95
0.5 0.168 0.158 0.151 11.26 4.64
0.57 0.168 0.158 0.152 10.53 3.95
0.66 0.168 0.158 0.151 11.26 4.64
0.8 0.168 0.159 0.151 11.26 5.30
4
1.5 0.274 0.271 0.258 6.20 5.04
2 0.351 0.347 0.331 6.04 4.83
2.5 0.425 0.419 0.399 6.52 5.01
3 0.495 0.488 0.462 7.14 5.63
3.5 0.562 0.554 0.521 7.87 6.33
4 0.628 0.618 0.576 9.03 7.29
4.5 0.693 0.682 0.628 10.35 8.60
5 0.757 0.744 0.676 11.98 10.06
0.2 0.168 0.156 0.150 12.00 4.00
0.25 0.168 0.156 0.151 11.26 3.31
0.33 0.168 0.156 0.150 12.00 4.00
0.4 0.168 0.156 0.150 12.00 4.00
0.5 0.168 0.156 0.151 11.26 3.31
0.57 0.168 0.156 0.151 11.26 3.31
0.66 0.168 0.156 0.150 12.00 4.00
0.8 0.168 0.158 0.152 10.53 3.95
5
1.5 0.278 0.275 0.262 6.11 4.96
2 0.359 0.355 0.340 5.59 4.41
2.5 0.437 0.432 0.412 6.07 4.85
3 0.511 0.506 0.481 6.24 5.20
3.5 0.584 0.577 0.546 6.96 5.68
4 0.654 0.647 0.607 7.74 6.59
4.5 0.724 0.716 0.666 8.71 7.51
5 0.794 0.784 0.720 10.28 8.89
77

0.2 0.194 0.189 0.178 8.99 6.18
0.25 0.194 0.189 0.178 8.99 6.18
0.33 0.194 0.189 0.177 9.60 6.78
0.4 0.194 0.189 0.177 9.60 6.78
0.5 0.194 0.189 0.178 8.99 6.18
0.57 0.194 0.189 0.177 9.60 6.78
0.66 0.194 0.189 0.177 9.60 6.78
0.8 0.194 0.189 0.177 9.60 6.78
1.5
1.5 0.249 0.242 0.229 8.73 5.68
2 0.293 0.284 0.271 8.12 4.80
2.5 0.329 0.317 0.307 7.17 3.26
3 0.360 0.345 0.341 5.57 1.17
3.5 0.386 0.369 0.365 5.75 1.10
4 0.412 0.390 0.390 5.64 0.00
4.5 0.440 0.410 0.410 7.32 0.00
5 0.472 0.433 0.430 9.77 0.70
0.2 0.194 0.186 0.175 10.86 6.29
0.25 0.194 0.186 0.173 12.14 7.51
0.33 0.194 0.186 0.173 12.14 7.51
0.4 0.194 0.186 0.173 12.14 7.51
0.5 0.194 0.186 0.176 10.23 5.68
0.57 0.194 0.186 0.176 10.23 5.68
0.66 0.194 0.186 0.176 10.23 5.68
0.8 0.194 0.186 0.173 12.14 7.51
2
1.5 0.261 0.256 0.243 7.41 5.35
2 0.319 0.312 0.299 6.69 4.35
2.5 0.370 0.361 0.348 6.32 3.74
3 0.415 0.404 0.386 7.51 4.66
3.5 0.457 0.443 0.421 8.55 5.23
4 0.497 0.479 0.453 9.71 5.74
4.5 0.542 0.517 0.480 12.92 7.71
5 0.591 0.562 0.505 17.03 11.29
0.2 0.195 0.186 0.174 12.07 6.90
0.25 0.195 0.186 0.174 12.07 6.90
0.33 0.195 0.186 0.174 12.07 6.90
0.4 0.195 0.186 0.173 12.72 7.51
3 0.5 0.195 0.186 0.174 12.07 6.90
0.57 0.195 0.186 0.174 12.07 6.90
0.66 0.195 0.186 0.174 12.07 6.90
0.8 0.195 0.186 0.174 12.07 6.90
1.5 0.273 0.269 0.254 7.48 5.91
78
2 0.345 0.339 0.322 7.14 5.28

2.5 0.411 0.403 0.380 8.16 6.05
3 0.472 0.464 0.437 8.01 6.18
3.5 0.531 0.520 0.486 9.26 7.00
4 0.587 0.575 0.531 10.55 8.29
4.5 0.643 0.629 0.572 12.41 9.97
5 0.700 0.682 0.607 15.32 12.36
0.2 0.195 0.186 0.172 13.37 8.14
0.25 0.195 0.186 0.173 12.72 7.51
0.33 0.195 0.186 0.173 12.72 7.51
0.4 0.195 0.186 0.173 12.72 7.51
0.5 0.195 0.186 0.173 12.72 7.51
0.57 0.195 0.186 0.172 13.37 8.14
0.66 0.195 0.186 0.173 12.72 7.51
0.8 0.195 0.186 0.173 12.72 7.51
4
1.5 0.278 0.275 0.260 6.92 5.77
2 0.357 0.352 0.334 6.89 5.39
2.5 0.430 0.424 0.399 7.77 6.27
3 0.500 0.493 0.465 7.53 6.02
3.5 0.568 0.559 0.521 9.02 7.29
4 0.634 0.624 0.577 9.88 8.15
4.5 0.700 0.687 0.629 11.29 9.22
5 0.764 0.750 0.678 12.68 10.62
0.2 0.196 0.184 0.177 10.73 3.95
0.25 0.196 0.184 0.177 10.73 3.95
0.33 0.196 0.184 0.177 10.73 3.95
0.4 0.196 0.184 0.177 10.73 3.95
0.5 0.196 0.184 0.176 11.36 4.55
0.57 0.196 0.184 0.177 10.73 3.95
0.66 0.196 0.184 0.176 11.36 4.55
0.8 0.196 0.185 0.176 11.36 5.11
5
1.5 0.281 0.278 0.261 7.66 6.51
2 0.363 0.359 0.340 6.76 5.59
2.5 0.442 0.436 0.413 7.02 5.57
3 0.517 0.511 0.482 7.26 6.02
3.5 0.589 0.582 0.547 7.68 6.40
4 0.660 0.652 0.608 8.55 7.24
4.5 0.730 0.721 0.666 9.61 8.26
5 0.801 0.790 0.720 11.25 9.72
79

0.2 0.224 0.218 0.204 9.80 6.86
0.25 0.224 0.218 0.204 9.80 6.86
0.33 0.224 0.218 0.203 10.34 7.39
0.4 0.224 0.218 0.203 10.34 7.39
0.5 0.224 0.218 0.204 9.80 6.86
0.57 0.224 0.218 0.202 10.89 7.92
0.66 0.224 0.218 0.201 11.44 8.46
0.8 0.224 0.218 0.203 10.34 7.39
1.5
1.5 0.261 0.256 0.246 6.10 4.07
2 0.306 0.298 0.292 4.79 2.05
2.5 0.343 0.334 0.327 4.89 2.14
3 0.375 0.364 0.359 4.46 1.39
3.5 0.406 0.393 0.383 6.01 2.61
4 0.440 0.425 0.412 6.80 3.16
4.5 0.479 0.461 0.432 10.88 6.71
5 0.523 0.500 0.479 9.19 4.38
0.2 0.225 0.217 0.202 11.39 7.43
0.25 0.225 0.217 0.201 11.94 7.96
0.33 0.225 0.217 0.199 13.07 9.05
0.4 0.225 0.217 0.201 11.94 7.96
0.5 0.225 0.217 0.201 11.94 7.96
0.57 0.225 0.217 0.201 11.94 7.96
0.66 0.225 0.217 0.200 12.50 8.50
0.8 0.225 0.217 0.202 11.39 7.43
2
1.5 0.271 0.266 0.243 11.52 9.47
2 0.330 0.324 0.299 10.37 8.36
2.5 0.382 0.374 0.346 10.40 8.09
3 0.430 0.419 0.387 11.11 8.27
3.5 0.476 0.462 0.422 12.80 9.48
4 0.529 0.508 0.450 17.56 12.89
4.5 0.587 0.560 0.482 21.78 16.18
5 0.645 0.615 0.506 27.47 21.54
0.2 0.226 0.214 0.205 10.24 4.39
0.25 0.226 0.214 0.201 12.44 6.47
0.33 0.226 0.214 0.201 12.44 6.47
0.4 0.226 0.214 0.201 12.44 6.47
3 0.5 0.226 0.214 0.201 12.44 6.47
0.57 0.226 0.214 0.201 12.44 6.47
0.66 0.226 0.214 0.201 12.44 6.47
0.8 0.226 0.214 0.201 12.44 6.47
1.5 0.279 0.274 0.255 9.41 7.45
80
2 0.352 0.345 0.322 9.32 7.14

2.5 0.419 0.410 0.381 9.97 7.61
3 0.483 0.471 0.437 10.53 7.78
3.5 0.544 0.529 0.488 11.48 8.40
4 0.607 0.586 0.532 14.10 10.15
4.5 0.678 0.649 0.575 17.91 12.87
5 0.748 0.715 0.613 22.02 16.64
0.2 0.226 0.213 0.197 14.72 8.12
0.25 0.226 0.213 0.197 14.72 8.12
0.33 0.226 0.213 0.197 14.72 8.12
0.4 0.226 0.213 0.196 15.31 8.67
0.5 0.226 0.213 0.197 14.72 8.12
0.57 0.226 0.213 0.197 14.72 8.12
0.66 0.226 0.213 0.197 14.72 8.12
0.8 0.226 0.213 0.197 14.72 8.12
4
1.5 0.283 0.279 0.260 8.85 7.31
2 0.362 0.357 0.332 9.04 7.53
2.5 0.437 0.431 0.400 9.25 7.75
3 0.508 0.501 0.464 9.48 7.97
3.5 0.577 0.568 0.521 10.75 9.02
4 0.644 0.633 0.578 11.42 9.52
4.5 0.713 0.699 0.630 13.17 10.95
5 0.788 0.768 0.678 16.22 13.27
0.2 0.229 0.210 0.201 13.93 4.48
0.25 0.229 0.210 0.201 13.93 4.48
0.33 0.229 0.210 0.200 14.50 5.00
0.4 0.229 0.210 0.201 13.93 4.48
0.5 0.229 0.210 0.201 13.93 4.48
0.57 0.229 0.210 0.200 14.50 5.00
0.66 0.229 0.211 0.200 14.50 5.50
0.8 0.229 0.211 0.201 13.93 4.98
5
1.5 0.285 0.282 0.263 8.37 7.22
2 0.368 0.364 0.340 8.24 7.06
2.5 0.447 0.443 0.413 8.23 7.26
3 0.523 0.518 0.482 8.51 7.47
3.5 0.597 0.591 0.547 9.14 8.04
4 0.669 0.661 0.608 10.03 8.72
4.5 0.740 0.731 0.667 10.94 9.60
5 0.813 0.802 0.720 12.92 11.39
81
Table 3.6. Results for Case 1 (cu1/cu2 = 0.2)
 = 15°  = 30°  = 45°  = 60°  = 75°

N2c (LB) 0.42 0.495 0.517 0.542 0.587
F (LB) 1.32 1.12 1.07 1.02 0.95
N2c (UB) 0.406 0.478 0.498 0.517 0.56
F (UB) 1.37 1.16 1.11 1.07 0.99
N2c (LEM) 0.404 0.461 0.48 0.48 0.482
F (LEM) 1.37 1.20 1.16 1.16 1.15
Table 3.7. Factors of safety for Case 2
H=3m H=4m H=5m H=6m H=7m

cu1/H 0.926 0.694 0.555 0.463 0.397
N2c (LB) = 0.495 F (LB) 1.87 1.40 1.12 0.94 0.80
N2c (UB) = 0.478 F (UB) 1.94 1.45 1.16 0.97 0.83
N2c (LEM) = 0.461 F (LEM) 2.01 1.51 1.20 1.00 0.86
Table 3.8. Results for Case 3 (cu1/cu2 = 0.8)
 = 15°  = 30°  = 45°  = 60°  = 75°

N2c (LB) 0.126 0.147 0.167 0.194 0.225
F (LB) 3.52 3.02 2.66 2.29 1.97
N2c (UB) 0.123 0.144 0.159 0.186 0.217
F (UB) 3.61 3.08 2.79 2.39 2.05
N2c (LEM) 0.123 0.142 0.152 0.173 0.217
F (LEM) 3.61 3.13 2.92 2.57 2.05
82
Figure 3.1. Problem configuration for cohesive material filled on the purely
cohesive soil
u=v=0
u=v=0
u=v=0
(a) Upper bound
n =  = 0
0

=
=
n =  = 0 
n
(b) Lower bound
Figure 3.2. Typical two dimensional finite element meshes and boundary
conditions used in limit analysis
83
0.8
cu1 / cu2 = 5
0.7 LB cu1 / cu2 = 5
cu1 / cu2 = 4.5
UB
0.6 LEM cu1 / cu2 = 4
cu1 / cu2 = 4 cu1 / cu2 = 3.5
0.5
N2c = cu1 / HF
cu1 / cu2 = 3
0.4 cu1 / cu2 = 2.5

cu1 / cu2 = 2
0.3
cu1 / cu2 = 1.5
0.2
cu1 / cu2 = 0.8
0.1 cu1 / cu2 = 0.2
0.0 d /H
1.5 2 3 4 5
Figure 3.3. The chart solutions for  = 15°
0.8
cu1 / cu2 = 5
cu1 / cu2 = 4.5

0.7 LB
UB cu1 / cu2 = 5
cu1 / cu2 = 4
0.6 LEM
cu1 / cu2 = 3.5
N2c = cu1 / HF
cu1 / cu2 = 3.5

0.5 cu1 / cu2 = 3
cu1 / cu2 = 2.5

0.4
cu1 / cu2 = 2
0.3
cu1 / cu2 = 1.5
0.2
cu1 / cu2 = 0.8
cu1 / cu2 = 0.2-0.66
0.1 d /H
1.5 2 3 4 5
84
0.8 cu1 / cu2 = 5

LB cu1 / cu2 = 4.5
0.7 UB cu1 / cu2 = 5
LEM cu1 / cu2 = 4
0.6 cu1 / cu2 = 3.5
N2c = cu1 / HF
0.5 cu1 / cu2 = 3

cu1 / cu2 = 3
cu1 / cu2 = 2.5
0.4
cu1 / cu2 = 2
0.3
cu1 / cu2 = 1.5
0.2
cu1 / cu2 = 0.2-0.8
0.1 d /H
1.5 2 3 4 5
0.8 cu1 / cu2 = 5

LB
UB cu1 / cu2 = 4.5
0.7
LEM cu1 / cu2 = 5 cu1 / cu2 = 4
0.6 cu1 / cu2 = 3.5

cu1 / cu2 = 3.5
N2c = cu1 / HF
0.5 cu1 / cu2 = 3
cu1 / cu2 = 2.5

0.4
cu1 / cu2 = 2
cu1 / cu2 = 2
0.3
cu1 / cu2 = 1.5
0.2 cu1 / cu2 = 0.2-0.8
0.1 d /H
1.5 2 3 4 5
85
0.9
LB
UB cu1 / cu2 = 5
0.8
LEM cu1 / cu2 = 4.5
0.7
cu1 / cu2 = 5 cu1 / cu2 = 4
0.6 cu1 / cu2 = 4 cu1 / cu2 = 3.5

N2c = cu1 / HF
cu1 / cu2 = 3
0.5
cu1 / cu2 = 2.5
0.4
cu1 / cu2 = 2
0.3
cu1 / cu2 = 1.5
0.2 cu1 / cu2 = 0.2-0.8
0.1 d /H
1.5 2 3 4 5
(a) cu1/cu2 = 0.5
(b) cu1/cu2 = 0.57
(c) cu1/cu2 = 0.66
Figure 3.8. The transition of UB plastic zones for  = 15 ° and d/H = 1.5
86
(a) d/H = 2
(b) d/H = 3
Figure 3.9. The transition of UB plastic zones for  = 15 ° and c u1 /c u2 = 0.5
(a) d/H = 3
(b) d/H = 4
(c) d/H = 5
Figure 3.10. The transition of UB plastic zones for  = 30 ° and c u1 /c u2 = 0.66
87
CHAPTER 4 APPLICATION OF ARTIFICIAL NEURAL NETWORK TO

INHOMOGENEOUS SOIL SLOPE STABILITY ASSESSMENTS BASED
ON FINITE ELEMENT LIMIT ANALYSIS METHODS
Z. G. Qian, A. J. Li, W. C. Chen, A.V. Lyamin
4.1 ABSTRACT
Assessment of soil slope stability is a geotechnical problem for engineering designs. This study
uses finite element upper bound (UB) and lower bound (LB) limit analysis methods (LA) to
investigate inhomogeneous soil slope stability on the basis of conventional Mohr - Coulomb
parameters. The observed differences between UB and LB solutions are within ± 2.5% or
better. The obtained stability numbers are presented as inhomogeneous soil slope stability
charts. In addition, the artificial neural network (ANN) is trained using the solutions to assess
soil slope stability. Slope stability analysis using ANN can simplify evaluations, and the
improvement leads to good accuracy.
4.2 INTRODUCTION
Slope stability problems (Bishop 1955; Griffiths and Lane 1999) are commonly considered in
civil engineering projects and have been investigated for years (Lim et al. 2015). Previous
studies have focused on safety, reliability, economics, failure mechanism, and remedial slope
measures. Slopes can be classified into many types, such as cut, natural, and filled. Problems
regarding filled slopes on two-layered undrained clay are commonly encountered in
geotechnical engineering designs (Brandon et al. 2008; Lim et al. 2015).
For civil engineers, performing numerical simulations is not always warranted or feasible due to
time and cost constraints. Over the past few years, chart solutions have drawn attention from
many investigators (Lim et al. 2015; Michalowski 2002; Qian et al. 2014; Baker et al. 2006).
These solutions are convenient tools to determine the factor of safety (Lim et al. 2015; Qian et
al. 2014). However, stability charts for filled slopes on inhomogeneous cohesive clay remain
limited. Therefore, this study aims to investigate inhomogeneous undrained soil slope stability
problems on the basis of the Mohr - Coulomb failure criterion (MC) (de Coulomb 1776; Mohr
1914). The obtained results are presented in the form of soil slope stability charts. ANN has
88
recently been applied to many geotechnical engineering investigations (Shahin et al. 2001;
Sakellariou and Ferentinou 2005; Li et al. 2016). This study adopts ANN to assess filled soil
slope stability. It must be stressed that, chart solutions are limited because parameters cannot
be fully matched during charts reading. Compared with chart reading, ANN estimates slope
stability more conveniently than chart solutions do. Moreover, the solutions based on ANN are
efficient and reliable.
4.3.1 Slope stability analysis
Limit equilibrium method (LEM), such as those proposed by Bishop (Bishop 1955) and Janbu
(1956), are widely used to evaluate slope stability problems (Cheng et al. 2007; Li et al. 2016).
LEM can provide the factor of safety (F) and is thus helpful to civil engineers when used to
evaluate slope conditions. LEM is a statistical method based on a physical theory and presented
as a translational collapse mechanism (Michalowski 1995). Therefore, the slip surface must be
assumed prior to safety factor calculation. However, Li et al. (2012) indicated that solutions
cannot be guaranteed when LEM is used. This statement was supported by the investigation of
Qian et al. (2014).
In the recent years, UB (Lyamin and Sloan 2002a; Krabbenhoft et al. 2005) and LB (Lyamin
and Sloan 2002b) in LA have become popular approaches used in slope analyses. LA is based
on two separate methods: UB that uses the theorem of kinematics (Lyamin and Sloan 2002a)
and LB that considers the static theorem (Sloan 1988; Chen et al. 2008; Lyamin and Sloan
2002b). LA method can generate failure mechanisms automatically (Krabbenhoft et al. 2005).
Therefore, the drawback of the assumed failure surface in LEM can be avoided. In the past
decades, most slope stability studies were based only on the UB method (Donald and Chen
1997; Michalowski 2002; Chen et al. 2003; Han et al. 2012; Gao et al. 2013). UB based on the
LA method alone usually cannot produce the true solution (Lim et al. 2016).
4.3.2 Stability number and chart solutions
The stability number was first proposed by Taylor (1937), who investigated uniform undrained
slopes. In the recent years, other slope stability numbers have been developed, an example is
the stability number investigated and presented as a set of stability charts based on two - layered
undrained clay slopes ( Lim et al. 2015; Qian et al. 2014), as shown in Equation (4.1). Qian et
89
al. (2014) and Lim et al. (2015a) considered the effects of cu1/cu2 ratios, as shown in Figure 4
1. The charts based on these stability numbers are inapplicable to the study of Taylor (1948).
N2c = cu1/γHF (4.1)
where
N2c is the stability number for two-layered undrained slopes, cu1 is the undrained shear strength
of the fill material, γ is the soil unit weight, H is the slope height, and F is factor of safety.
Chart solutions have been widely used due to their convenience since the proposal of the first
set of purely cohesive uniform slope stability charts by Taylor (1937). Chart solutions are
simple methods of determining slope stability (Qian et al. 2014; Li et al. 2009). Gens et al.
(1988) produced a set of 3D boundary effect stability charts on the basis of cohesive soils.
Slope stability charts have drawn the attention of many investigators (Michalowski 2002;
Michalowski and Martel 2011; Li et al. 2009; Li et al. 2010; Qian et al. 2014; Lim et al. 2015;
Lim et al. 2015). A review of previous studies reveals, that existing chart solutions may not be
applied to certain types of slopes, such as slopes with undrained material filled on
inhomogeneous cohesive clay.
4.3.3 Slope assessment using ANN
ANN has recently been employed in many investigations to solve geotechnical engineering
problems (Lee et al. 2003; Kuo et al. 2009; Cho 2009; Li et al. 2016; Sakellariou and Ferentinou
2005). ANN is a mathematical model and a network used to simulate the human brain
(Agatonovic-Kustrin and Beresford 2000). The ANN structure is composed of input, hidden,
and output layers (Cho 2009), as shown in Figure 4.2. The neurons and weight connection in
each layer directly affect the accuracy of ANN models. In Figure 4.2, n inputs (x1, x2, x3⋯ xn)
are designed, and m outputs (y1, y2, y3 ⋯ ym) can be obtained. φ (.) is the activation transfer
function with M (1, 2 ⋯ l, M) linear nodes in the hidden layer. The input weights are wij (i=1
⋯ M and j=1 ⋯ n). The output weights are ij (i=1 ⋯ m and j=1 ⋯ M). Therefore, the input
data vector x(k) and the output data vector y(k) can be expressed as Equations (4.2) and (4.3)
respectively.
𝑥(𝑘) = [𝑥1 (𝑘) 𝑥2 (𝑘) ⋯ 𝑥𝑛 (𝑘)]𝑇 (4.2)
𝑦(𝑘) = [𝑦1 (𝑘) 𝑦2 (𝑘) ⋯ 𝑦𝑚 (𝑘)]𝑇 (4.3)
90
The output of the ith hidden neuron oi can be summarized as follows:
𝑜𝑖 = 𝜑[∑𝑛(𝑗=1) 𝑤𝑖𝑗 𝑥𝑗 (𝑘)] = 𝜑 (𝑤𝑖𝑇 𝑥(𝑘)) for i = 1, 2, ⋯ M (4.4)
where
𝑤𝑖 = [𝑤𝑖1 (𝑘) 𝑊𝑖2 (𝑘) ⋯ 𝑤𝑖𝑛 (𝑘)]𝑇 for i = 1, 2, ⋯ M and 𝜑(.) is the activation function
In addition, the output of the neural network 𝑦𝑖 (𝑘) can be expressed as:
𝑦𝑖 (𝑘) = ∑𝑀 𝑀 𝑇 𝑇
(𝑗=1) 𝛼𝑗𝑖 𝑜𝑗 = ∑(𝑗=1) 𝛼𝑗𝑖 𝜑 (𝑤𝑗 𝑥(𝑘)) = 𝝃 (𝑘)𝛼𝑖 for i = 1, 2, ⋯ m (4.5)
where
𝛼𝑖 = [𝛼1𝑖 𝛼2𝑖 , ⋯ 𝛼𝑀𝑖 ] for i = 1, 2, ⋯ m
𝑇
𝝃(𝑘) = [𝜑(𝑤1𝑇 𝑥(𝑘)) 𝜑(𝑤2𝑇 𝑥(𝑘)) ⋯ 𝜑(𝑤𝑀
𝑇
𝑥(𝑘))]
Based on Equations above, the vector y(k) can be expressed as:
𝑦(𝑘) = [𝑦1 (𝑘)𝑦2 (𝑘) ⋯ 𝑦𝑚 (𝑘)]𝑇 = [𝝃𝑇 (𝑘)𝛼1 𝝃𝑇 (𝑘)𝛼2 ⋯ 𝝃𝑇 (𝑘)𝛼𝑚 ]𝑇 = [𝝃𝑇 (𝑘)𝛼 ]𝑇 (4.6)
where
𝛼 = [𝛼1 𝛼2 , ⋯ 𝛼𝑚 ]
Besides, N of training vectors x(1), x (2) ⋯ x (N) and output data vectors yd(1), yd(2) ⋯ yd(N)
𝝃𝑇 (1) 𝑦 𝑇 (1)
𝐺𝑎 = [ ⋮ ] 𝛼 = [ ⋮ ] (4.7)
𝝃𝑇 (𝑁) 𝑦 𝑇 (𝑁)
Moreover, G as hidden layer output matrix cab be expressed as:
𝜑(𝑤1𝑇 𝑥(1)) ⋯ 𝑇
𝜑(𝑤𝑀 𝑥(1))
𝐺=[ ⋮ ⋱ ⋮ ] (4.8)
𝑇 𝑇
𝜑(𝑤1 𝑥(𝑁)) ⋯ 𝜑(𝑤𝑀 𝑥(𝑁))
After training process, G can be computed in a single iteration.
𝛼 = (𝐺 𝑇 𝐺) −1 𝐺 𝑇 𝑌𝑑 (4.9)
91
𝑦𝑑𝑇 𝑥(1)
𝑌𝑑 = [ ⋮ ] (4.10)
𝑦𝑑𝑇 𝑥(N)
ANN can learn from a set of data. It has the power to predict the output from a set of inputs
once learning is successfully accomplished. However, a weakness of ANN analysis is data
extrapolation. Therefore, the selected training data should fully cover the entire range of
variables (Cho 2009).
To simplify the problem, only plane strain cases are considered in this research. A 2D
illustration of the investigated soil slope stability problem and the boundary conditions are
shown in Figure 4.3. The slope geometry analyses for the filled soil slopes with two different
layers are also displayed. The filled material and slope foundation are divided into two regions.
The undrained shear strengths in Region 1 and Region 2 differ. The cohesive strength (cu1) in
Region 1 is assumed to be a homogeneous material. cu2 is the undrained shear strength for the
surface in Region 2. The ratio of cu1/cu2 is from 0.2 to 5. In addition, Z is the depth from the
interface of Region 1 and Region 2. ρ is the rate of the linear increase in undrained shear
strength with depth Z, and ρ ranges from 0 to 5.0. Soil slope angles ( range from 15° to 75°,
and depth factors (d/H) range from 1.5 to 5. The range of slope parameter analyses is shown in
Table 4.1.
When presenting the final results using chart solutions, the dimensionless stability number
(N2ci) is defined as shown in Equation (4.11). N2ci is observed using UB and LB methods. The
obtained results, show that the differences between the two methods are less than ± 2.5%.
N2ci = cu1 / γHF (4.11)
As mentioned earlier, the ANN structure consists of input, hidden, and output layers. An
extreme learning training machine (Huang et al. 2012) is adopted in this study for the network
training, as shown in Figure 4.4. β, cu1/cu2, d/H, and  are selected as inputs of the ANN
analysis, and N2ci is considered the output. Notably, N2ci in the training section is based on the
average of UB and LB solutions.
92
4.5 RESULTS AND DISCUSSION
4.5.1 LA results for stability number and stability charts
The computed UB and LB estimates of average N2ci for slope  = 15° and d/H = 1.5 are shown
graphically in Figure 4.5 (a). These results indicate that N2ci generally increases with the
increase in the ratios of cu1/cu2 when cu1/cu2 ≥ 0.8. N2ci is inversely proportional to F. Therefore,
a large value of N2ci means that the slope is critical. N2ci is nearly a constant when cu1/cu2 ≤ 0.8.
The exceptions for  = 15° and d/H = 1.5 are only  from 0 to 1. As shown in Table 4.2, N2ci
increases with the increase in d/H when d/H ≤ 3. Moreover, N2ci is nearly a constant when d/H
≥ 3. Figure 4.5 (a) to (c) show the stability charts for the slope  = 15° and d/H from 1.5 to 3.0.
For slopes with d/H ≥ 3, the chart solutions for d/H = 3 can be applied to the slope evaluations
directly. Figure 4.5 (a) to (c) show that, N2ci increases with the decrease in  when cu1/cu2 ≥ 0.8.
The chart solutions presented in Figures 4.6 to 4.9 show various slope inclinations and d/H
ratios (β from 30º to 75º and d/H from 1.5 to 3.0). The similar trends for β are observed from
30º to 75º. N2ci is can be assume as constant when cu1/cu2 ≤ 0.8 and ≥ 45o. Figures 4.5 to 4.9
show that N2ci increases with the increase in β.
Figure 4.10 (a) and (b) show that the failure surfaces for cu1/cu2 = 0.2 and cu1/cu2 = 0.5 are only
limited in Region 1 when  = 0.25, d/H = 1.5, and = 15° - 45°. When cu1/cu2 = 0.8, the slip
surface for = 15° and 30° can pass through Regions 1 and 2, as shown in Figure 4.10 (c).
Figure 4.10 (d) shows that all slip surfaces for = 15°, 30°, and 45° pass through the soils in
Regions 1 and 2. Therefore, for a gentle slope (= 15° to 30°), the ratio of cu1/cu2 generally
exerts no influence on failure surfaces when cu1/cu2 is small (cu1/cu2 ≤ 0.5). In other words, the
slopes can be treated as homogenous slopes when cu1/cu2 ≤ 0.5. The slope   45° presents the
same phenomenon when cu1/cu2 ≤ 1.5. Qian et al. (2014) investigated two layered purely
cohesive soil slopes and also obtained a similar result.
For all slope angles, when cu1/cu2 ≥ 1.5, the depth of failure surfaces is nearly a constant when
d/H ≥ 3, as shown in Figure 4.11 (a) and (b). Figure 4.11 shows the UB plastic zones for =
15o or 60o, cu1/cu2 = 5.0,  = 1.5 and d/H = 1.5, 3 or 5.0. The influence of the d/H ratio on slope
stability is significant when d/H ≤ 3. A similar phenomenon was observed by Yu et al. (1998)
and Li et al (2009; 2010), who examined purely cohesive slopes with undrained shear strength
increasing with depth. Figures 4.5 to 4.9 show that  affects the assessment of slope stability.
This phenomenon is obvious when  is small. Table 4.3 shows N2ci for different  values. The
93
examples selected in Table 4.3 are based on = 75o, d/H = 3, and cu1/cu2 = 3. The difference in
N2ci between  = 4.0 and  = 5.0 is approximately 2.17%, which is only a slight change.
However, the difference in N2ci between  = 0.25 and  = 0.75 is 14.04%, which is significantly
larger than that for  = 4.0 and  = 5.0 (i.e., 2.17%). Figure 4.12 shows the observations of the
slopes for = 45°, d/H = 2, cu1/cu2 = 1.5, and  = 0.25, 0.75, and 1.5. The depth of the failure
surface decreases with the increase in .
4.5.2 Comparison of results with LEM
LA methods can bracket the true solutions only between LB and UB. This section compares
the results obtained from LEM and LA on the basis of the MC failure criterion. The LEM
results in this study are based on the simplified 2D Bishop method (Bishop 1955) and Janbu
method (Janbu 1973). The soil profile of the analysed slope based on LEM is divided into two
regions. Region 1 is the filled material, and Region 2 is divided into many layers. cu2 linearly
increases with Z because the slope profile is based on inhomogeneous soil slope, which
considers different values of , as shown in Figure 4.3. Figure 4.13 displays the difference in
N2ci between the two methods when β = 15o or 60o,  = 0.5, and cu0/cu1 = 5.0, and the values of
d/H are between 1.5 and 5. The results show that the N2ci values obtained from LEM (Bishop
and Janbu) are smaller than the LA results. The results obtained from LEM are close to the UB
results when the slope is gentle (β = 15o), as shown in Figure 4.13 (a). A Significant difference
in N2ci between LEM and LA is observed when β increases to 60o, as shown in Figure 4.13 (b).
Notably, this phenomenon becomes obvious when  increases. The outcome implies that the F
of the soil slope is overestimated when LEM is employed.
In addition, failure surfaces are investigated using the LEM solutions. The cases selected are
= 15° or 60°, d/H = 2,  = 0.5, and cu1/cu2 = 0.2, 0.8, or 3.0. A comparison of failure surfaces
with different values of the cu1/cu2 ratio and  is shown in Figure 4.14. These failure surfaces
are plastic zones obtained from UB, and these UB plastic zones can be considered the slope
failure mechanisms. The failure mode based on LEM (both Bishop and Janbu) and LA is
general similar. The results indicate that the failure surface based on the Janbu method is
slightly larger than that based on the Bishop method.
94
4.5.3 Comparison of results with two-layered purely cohesive soil slopes
In this section, the results obtained in this study and results of previous chart solutions based
on two - layered purely cohesive soil slopes (Qian et al. 2014) are compared. The foundation
material was considered a uniform layer ( = 0) in the study of Qian et al. (2014). Several cases
are selected and the details are provided in Table 4.4. The results in Table 4.4 show that the
consideration of  significantly affects the assessment of slope stability when the ratio of cu1/cu2
is sufficiently large (cu1/cu2 ≥ 0.8). Moreover, N2ci based on  = 0.25 is similar to N2c when d/H
≤ 1.5.
Table 4.4 shows that the difference between N2ci and N2c significantly increases when the d/H
ratio and  increase. The difference in N2ci between  = 0 and  = 0.25 is over 30% when  =
15°, d/H = 5, and cu1/cu2 = 5. Figure 4.15 presents the slopes with  = 45°, cu1/cu2 = 1.5, d/H =
1.5, and various  values. Figure 4.15 (a) shows that the slope failure mechanism passes
through the soils in Regions 1 and 2. However, the soil strength in Region 2 increases. The
failure surface is controlled by the soil in Region 2, as shown in Figure 4.15 (b) and (c). For 
= 5, the soil in Region 2 acts similar to a rigid layer, as shown in Figure 4.15 (c).
4.5.4 ANN training and validation analysis
As previously discussed, this study produces slope stability charts by adopting  = 15°, 30°,
45°, 60°, and 75°. However, manual estimation of chart solutions is required, such as  = 43°.
Thus, this study also proposes slope stability analysis based on ANN. ANN in this research
should be trained as an evaluation package based on the MC failure criterion. ANN can then
be used to assess slope stability. The results based on ANN are presented in the form of stability
numbers (NANN). Therefore, this research selects , the ratio of cu1/cu2, depth factor, and  as
the training inputs, and NANN is used as the training output. The data based on the average N2ci
of UB and LB are also randomly selected.
Recent investigations have shown that a total of 200 hidden nodes are considered in ANN to
build a continuous differentiable that can map the inputs to the outputs (Huang et al. 2006, Li
et al. 2016). The input weights are generated randomly within [- 1.1], and the sigmoid function
can be written as Equation (4.12).
1
φ(t) = 1+exp(−τ)′ (4.12)
95
The above - mentioned equation is continuously differentiable and used as the nonlinear
activation function of the hidden nodes.
The ANN training results (NANN) are compared with the actual data for the slope stability
assessment (N2ci), as shown in Figure 4.16, where NANN is well trained. The ranges for ANN
analysis include from 15° to 75°, cu1/cu2 from 0.2 to 5, depth factor from 1.5 to 5, and  from
0 to 5.0. Another 800 validation data sets (20% of the trained data) are used to verify the
performance of the trained ANN. The results of the 800 validation data sets also highly agree
with the real data (N2ci), as shown in Figure 4.17. In addition, a previous study based on two -
layered purely cohesive soil slopes (Qian et al. 2014) is also considered to obtain a complete
set of evaluation packages.
4.5.5 Application of ANN to fill slope stability assessments
This section uses a few assumed cases to verify the accuracy of the trained ANN. The case
studies consider the range of cu1/cu2 from 0.2 to 5, d/H from 1.5 to 5,  from 0 to 5.0, and β
from 15 to 75. All presented case studies are disregarded in the previous training of the network.
= 15 kN/m3 is assumed for all case studies. The assumed parameters are shown in Table 4.5.
NANN can be obtained with the ANN approach, and N2ci can be calculated using the chart
solutions. The different results between ANN and stability charts are shown in Table 4.6. The
results show that, F based on NANN is close to F based on the chart solutions. For example, the
F values calculated based on NANN in Cases 3, 5, and 8 are exactly the same as the F values
obtained through chart reading. The maximum F difference between ANN and stability charts
is only 4.4 %. The poor prediction for N2ci may be caused by chart reading. Thus, the above-
mentioned case studies show that the application of ANN to soil slope stability analysis
provides high accuracy. This approach is a convenient means of preliminary slope stability
evaluations by engineers.
4.6 CONCLUSION
This study uses UB and LB methods to assess the stability of a fill soil slope composed of an
inhomogeneous cohesive soil foundation. The true solutions between UB and LB for all the
results are bracketed within ± 2.5% or better. This study applies ANN to reduce the works of
manually estimating chart solutions. The results based on ANN training are highly accurate.
The following conclusions are obtained:
96
When cu1/cu2 ≥ 0.8 and d/H ≤ 3, N2ci increases with the increase in β, the ratios of cu1/cu2, and
the ratios of d/H. In addition, N2ci increases with the reduction in .  substantially affects the
assessment of slope stability. However, the ratio of cu1/cu2, the ratio of d/H, and  generally
exert no influence on N2ci when cu1/cu2 ≤ 0.8.
The UB plastic zones show that the failure surface is limited to a certain depth in Region 2
when  is large and cu1/cu2 ≥ 1.5. In addition, d/H may affect the failure surfaces, but the depth
of failure surfaces is generally unchanged when d/H ≥ 3 and cu1/cu2 ≥ 0.8. The slope failure
surfaces for cu1/cu2 ≤ 0.8 may only occur in the top layer of the soil slope. Therefore, the slope
can be treated as homogenous slopes.
The N2ci values obtained from LEM are smaller than those from UB and LB. Therefore, the
factor of safety for slope stability designs may be overestimated. The predetermined failure
surface in LEM may not be always adequate to assess the filled slopes investigated in the
present study.
The results obtained from the ANN training package can directly estimate N2ci accurately and
effectively. The investigations in this research demonstrate that the ANN training package is a
convenient tool for the preliminary design of slope stability.
97
4.7 REFERENCES
Bishop AW (1955) The use of the slip circle in the stability analysis of slopes.
p. 387-403.
Lim K, Li AJ and Lyamin AV (2015) Three-dimensional slope stability assessment of two-

layered undrained clay. Computers and Geotechnics2015. p. 1-17.
Brandon TL, Wright SG and Duncan JM (2008) Analysis of the Stability of I-Walls with Gaps
between the I-Wall and the Levee Fill. Journal of geotechnical and geoenvironmental
Qian Z, Li AJ, Merifield R and Lyamin AV (2014) Slope stability charts for two-layered purely
Lim K, Lyamin AV, Cassidy M and Li A-J (2015) Three-dimensional slope stability charts for
de Coulomb CA (1776) Essai sur une application des règles de maximis & minimis à quelques
problèmes de statique, relatifs à l'architecture1776.
Mohr O (1914) Die Spannungszustand einer Staumauer. Abhandlungen aus dem Gebiete der
technischen Mechanik1914. p. 284-308.
98
Sakellariou M and Ferentinou M (2005) A study of slope stability prediction using neural
networks. Geotechnical & Geological Engineering2005. p. 419.
Li AJ, Khoo S, Lyamin AV and Wang Y (2016) Rock slope stability analyses using extreme
Janbu N, Bjerrum L and Kjaernsli B (1956) Soil mechanics applied to some engineering
problems. Norwegian Geotechnical Institute Publication1956.
Cheng Y, Lansivaara T and Wei W (2007) Two-dimensional slope stability analysis by limit
equilibrium and strength reduction methods. Computers and Geotechnics2007. p. 137-50.

283-93.
Lyamin AV and Sloan S (2002a) Upper bound limit analysis using linear finite elements and
Geomechanics2002a. p. 181-216.
p. 1069-88.
Lyamin AV and Sloan S (2002b) Lower bound limit analysis using non‐linear programming.
99
Chen S, Liu Y and Cen Z (2008) Lower‐bound limit analysis by using the EFG method and
non‐linear programming. International Journal for Numerical Methods in Engineering2008. p.
391-415.
Han C-Y, Wang J-H, Xia X-H and Chen J-J (2012) Limit analysis for local and overall stability
of slurry trench in cohesive soil. Internation Journal for Geomechanics2012. p.
doi:10.1061/(ASCE)GM.943-5622.0000268.
1528-38.
Lim K, Li AJ, Schmid A and Lyamin AV (2016) Slope-stability assessments using finite-
element limit-analysis methods. International Journal of Geomechanics2016. p. 06016017.
Li AJ, Lyamin AV and Merifield R (2009) Seismic rock slope stability charts based on limit
Gens A, Hutchinson J and Cavounidis S (1988) Three-dimensional analysis of slides in

Michalowski RL and Martel T (2011) Stability charts for 3D failures of steep slopes subjected
to seismic excitation. Journal of geotechnical and geoenvironmental engineering2011. p. 183-
9.
Li A, Merifield R and Lyamin A (2009) Limit analysis solutions for three dimensional
undrained slopes. Computers and Geotechnics2009. p. 1330-51.
based on limit analysis methods. Canadian Geotechnical Journal2010. p. 1316-34.
100
Lee S, Lee S and Kim Y (2003) An approach to estimate unsaturated shear strength using
artificial neural network and hyperbolic formulation. Computers and Geotechnics2003. p. 489-
503.
Kuo Y, Jaksa M, Lyamin A and Kaggwa W (2009) ANN-based model for predicting the
bearing capacity of strip footing on multi-layered cohesive soil. Computers and
Cho SE (2009) Probabilistic stability analyses of slopes using the ANN-based response surface.
Computers and Geotechnics2009. p. 787-97.
Agatonovic-Kustrin S and Beresford R (2000) Basic concepts of artificial neural network

(ANN) modeling and its application in pharmaceutical research. Journal of pharmaceutical and
biomedical analysis2000. p. 717-27.
Huang G-B, Zhou H, Ding X and Zhang R (2012) Extreme learning machine for regression
and multiclass classification. IEEE Transactions on Systems, Man, and Cybernetics, Part B
(Cybernetics)2012. p. 513-29.

Huang G-B, Chen L and Siew CK (2006) Universal approximation using incremental
constructive feedforward networks with random hidden nodes. IEEE Trans Neural
Networks2006. p. 879-92.
101
4.8 APPENDIX
Table 4.1. Parameters investigated in this research
cu1/cu2 0.2, 0.25, 0.33, 0.4, 0.5, 0.57, 0.66, 0.8, 1.5, 2, 2.5, 3, 3.5, 4, 4.5, 5
d/H 1.5, 2, 3, 4, 5
 0, 0.25, 0.5, 0.75, 1, 1.5, 2.0, 3.0, 4.0, 5.0
β 15°, 30°, 45°, 60°, 75°
Table 4.2. The relationship between slope stability and d/H
N2ci N2ci N2ci N2ci N2ci

  cu1/cu2
d/H = 1.5 d/H = 2.0 d/H = 3.0 d/H = 4.0 d/H = 5.0
0.8 0.102 0.108 0.11 0.11 0.11
0.25 2.5 0.218 0.266 0.287 0.288 0.288
5.0 0.305 0.404 0.485 0.489 0.489
0.8 0.091 0.091 0.091 0.092 0.092
15° 0.75 2.5 0.202 0.222 0.224 0.225 0.225
5.0 0.285 0.350 0.372 0.373 0.374
0.8 0.089 0.089 0.089 0.089 0.089
1.50 2.5 0.181 0.184 0.185 0.185 0.185
5.0 0.262 0.294 0.296 0.297 0.297
Table 4.3. The relationship between slope stability and 
N2ci N2ci N2ci N2ci N2ci N2ci

 d cu1/cu2
 = 0.25  = 0.75  = 1.5  = 3.0  = 4.0  = 5.0
0.8 0.102 0.091 0.089 0.089 0.089 0.089
1.5 1.5 0.160 0.146 0.128 0.110 0.107 0.107
3 0.240 0.223 0.202 0.170 0.157 0.150
15°
0.8 0.110 0.091 0.089 0.089 0.089 0.089
3 1.5 0.189 0.151 0.128 0.110 0.107 0.106
3 0.332 0.257 0.209 0.170 0.157 0.150
0.8 0.221 0.221 0.221 0.221 0.221 0.221
1.5 1.5 0.249 0.240 0.235 0.232 0.231 0.230
3 0.356 0.335 0.310 0.285 0.276 0.270
75°
0.8 0.221 0.221 0.221 0.221 0.221 0.221
3 1.5 0.250 0.241 0.236 0.232 0.231 0.230
3 0.399 0.343 0.311 0.285 0.276 0.270
102
Table 4.4. The relationship between two layered purely cohesive soil slopes and 
The difference
N2ci N2ci N2ci N2ci
 d/H cu1/cu2 between N2c and N2ci
( = 0) ( = 0.25) ( = 0.75) ( = 1.5)
( = 0.25)
1.5 0.5 0.0885 0.0885 0.0885 0.0885 0.0 %
5 0.5 0.0930 0.0886 0.0885 0.0885 4.7 %
1.5 2 0.1995 0.1918 0.1761 0.1567 3.8 %
2 2 0.253 0.2257 0.1880 0.1570 10.8 %
15
5 2 0.3395 0.2409 0.1897 0.1582 29.0 %
1.5 5 0.3125 0.3048 0.2849 0.2620 2.5 %
2 5 0.4375 0.4039 0.3499 0.2940 7.7 %
5 5 0.7525 0.4892 0.3736 0.2973 35.0 %
1.5 0.8 0.2210 0.2210 0.2210 0.2210 0.0 %
5 0.8 0.2200 0.2210 0.2210 0.2210 0.4 %
1.5 2 0.3020 0.2905 0.2734 0.2596 3.8 %
2 2 0.3270 0.3000 0.2740 0.2600 8.3 %
75
5 2 0.3660 0.3005 0.2741 0.2601 17.9 %
1.5 5 0.5115 0.4871 0.4475 0.4053 4.8 %
2 5 0.6300 0.5721 0.4899 0.4166 9.2 %
5 5 0.8075 0.6078 0.4946 0.4165 24.7 %
Table 4.5. Summary of slope data from cases studies
Case H (m) β cu1 (kN/m2) cu2 (kN/m2) d/H 

1 22 17 30 37.5 1.5 0
2 18 28 42 28.0 2 0.25
3 35 30 38 15.2 5 0.75
4 27 45 20 11.1 4.3 0.9
5 41 32 22 12.9 1.9 1
6 15 35 30 15.0 1.5 1.5
7 19 20 28 7.0 2.8 0
8 25 45 36 12.0 5 1
9 16 15 27 6.0 4 0.25
10 20 60 40 12.3 4.3 1.5
103
Table 4.6. Summary of F calculated by using NANN and N2ci
F F F difference
Case cu1/H cu1/cu2 N2ci NANN
(Chart solution) (ANN) (%)
1 0.09 0.8 0.120 0.758 0.117 0.777 2.50
2 0.16 1.5 0.209 0.744 0.210 0.741 0.48
3 0.07 2.5 0.265 0.273 0.265 0.273 0.00
4 0.05 1.8 0.231 0.214 0.226 0.219 2.16
5 0.04 1.7 0.200 0.179 0.200 0.179 0.00
6 0.13 2.0 0.212 0.629 0.214 0.623 0.94
7 0.10 4.0 0.500 0.196 0.522 0.188 4.40
8 0.10 3.0 0.300 0.320 0.300 0.320 0.00
9 0.11 4.5 0.450 0.250 0.453 0.248 0.67
10 0.13 3.3 0.310 0.430 0.309 0.431 0.32
104
Figure 4.1. The model based on the research of Qian et al. (2014) and Lim et al.
(2015a)
Figure 4.2. ANN based on single hidden layer neural network ( Li et al. 2016)
Figure 4.3. c u2 linearly increased based on  in Region 2 for purely soil material
filled on inhomogeneous cohesive soil
Figure 4.4. The trained ANN inputs forward and inverse to the outputs
105
0.35
0.3
0.25
0.2
N2ci = cu1/HF
cu1/cu2=5.0
cu1/cu2=4.5
cu1/cu2=4.0
cu1/cu2=3.5
0.15 cu1/cu2=3.0
cu1/cu2=2.5
cu1/cu2=2.0
cu1/cu2=1.5
0.1
cu1/cu2=0.2-0.8
0.075
0 1 2 3 4 5

(a) d/H = 1.5
0.45
0.4
0.3
N2ci = cu1/HF
0.2 cu1/cu2=5.0
cu1/cu2=4.5
cu1/cu2=4.0
cu1/cu2=3.5
cu1/cu2=3.0
cu1/cu2=2.5
cu1/cu2=2.0
cu1/cu2=1.5
0.1
cu1/cu2=0.2-0.8
0.084
0 1 2 3 4 5

(b) d/H = 2.0
106
0.65
0.6
0.5
0.4
0.3
N2ci = cu1/HF
0.2 cu1/cu2=5.0
cu1/cu2=4.5
cu1/cu2=4.0
cu1/cu2=3.5
cu1/cu2=3.0
cu1/cu2=2.5
cu1/cu2=2.0
cu1/cu2=1.5
0.1
cu1/cu2=0.2-0.8
0.075
0 1 2 3 4 5

(c) d/H = 3.0
Figure 4.5. LA solutions of stability numbers for  = 15°
0.4
0.35
0.3
cu1/cu2=5.0
N2ci = cu1/HF
0.25
cu1/cu2=4.5
cu1/cu2=4.0
cu1/cu2=3.5
cu1/cu2=3.0
0.2
cu1/cu2=2.5
cu1/cu2=2.0
0.15 cu1/cu2=1.5
cu1/cu2=0.2-0.8
0.12
0 1 2 3 4 5

(a) d/H = 1.5
107
0.55
0.5
0.45
0.4
0.35
N2ci = cu1/HF
0.3
0.25 cu1/cu2=5.0
cu1/cu2=4.5
cu1/cu2=4.0
cu1/cu2=3.5
0.2 cu1/cu2=3.0
cu1/cu2=2.5
cu1/cu2=2.0
0.15 cu1/cu2=1.5
cu1/cu2=0.2-0.8
0.12
0 1 2 3 4 5

(b) d/H = 2.0
0.7
0.65
0.6
0.55
0.5
0.45
0.4
N2ci = cu1/HF
0.35
0.3
0.25 cu1/cu2=5.0
cu1/cu2=4.5
cu1/cu2=4.0
cu1/cu2=3.5
0.2 cu1/cu2=3.0
cu1/cu2=2.5
cu1/cu2=2.0
0.15 cu1/cu2=1.5
cu1/cu2=0.2-0.8
0.12
0 1 2 3 4 5

(c) d/H = 3.0
108
0.45
0.4
0.35
0.3
N2ci = cu1/HF
cu1/cu2=5.0
cu1/cu2=4.5
cu1/cu2=4.0
0.25 cu1/cu2=3.5
cu1/cu2=3.0
cu1/cu2=2.5
0.2 cu1/cu2=2.0
cu1/cu2=1.5
cu1/cu2=0.2-0.8
0.15
0 1 2 3 4 5

(a) d/H = 1.5
0.55
0.5
0.45
0.4
0.35
N2ci = cu1/HF
0.3
cu1/cu2=5.0
cu1/cu2=4.5
cu1/cu2=4.0
0.25 cu1/cu2=3.5
cu1/cu2=3.0
cu1/cu2=2.5
0.2 cu1/cu2=2.0
cu1/cu2=1.5
cu1/cu2=0.2-0.8
0.15
0 1 2 3 4 5

(b) d/H = 2.0
109
0.7
0.65
0.6
0.55
0.5
0.45
0.4
N2ci = cu1/HF
0.35
0.3
cu1/cu2=5.0
cu1/cu2=4.5
0.25 cu1/cu2=4.0
cu1/cu2=3.5
cu1/cu2=3.0
cu1/cu2=2.5
0.2 cu1/cu2=2.0
cu1/cu2=1.5
0.16 cu1/cu2=0.2-0.8
0 1 2 3 4 5

(c) d/H = 3.0
0.5
0.45
0.4
0.35
N2ci = cu1/HF
0.3 cu1/cu2=5.0
cu1/cu2=4.5
cu1/cu2=4.0
cu1/cu2=3.5
0.25 cu1/cu2=3.0
cu1/cu2=2.5
cu1/cu2=2.0
cu1/cu2=1.5
0.2
cu1/cu2=0.2-0.8
0.18
0 1 2 3 4 5

(a) d/H = 1.5
110
0.6
0.55
0.5
0.45
0.4
N2ci = cu1/HF
0.35
0.3 cu1/cu2=5.0
cu1/cu2=4.5
cu1/cu2=4.0
cu1/cu2=3.5
0.25 cu1/cu2=3.0
cu1/cu2=2.5
cu1/cu2=2.0
cu1/cu2=1.5
0.2
cu1/cu2=0.2-0.8
0.18
0 1 2 3 4 5

(b) d/H = 2.0
0.7
0.65
0.6
0.55
0.5
0.45
N2ci = cu1/HF
0.4
0.35
0.3 cu1/cu2=5.0
cu1/cu2=4.5
cu1/cu2=4.0
cu1/cu2=3.5
0.25 cu1/cu2=3.0
cu1/cu2=2.5
cu1/cu2=2.0
cu1/cu2=1.5
0.2
cu1/cu2=0.2-0.8
0.18
0 1 2 3 4 5

(c) d/H = 3.0
111
0.55
0.5
0.45
0.4
N2ci = cu1/HF
0.35
cu1/cu2=5.0
cu1/cu2=4.5
0.3 cu1/cu2=4.0
cu1/cu2=3.5
cu1/cu2=3.0
cu1/cu2=2.5
0.25
cu1/cu2=2.0
cu1/cu2=1.5
cu1/cu2=0.2-0.8
0.21
0 1 2 3 4 5

(a) d/H = 1.5
0.65
0.6
0.55
0.5
0.45
N2ci = cu1/HF
0.4
0.35
cu1/cu2=5.0
0.3 cu1/cu2=4.5
cu1/cu2=4.0
cu1/cu2=3.5
cu1/cu2=3.0
cu1/cu2=2.5
0.25
cu1/cu2=2.0
cu1/cu2=1.5
cu1/cu2=0.2-0.8
0.2
0 1 2 3 4 5

(b) d/H = 2.0
112
0.8
0.7
0.6
0.5
N2ci = cu1/HF
0.4
cu1/cu2=5.0
0.3 cu1/cu2=4.5
cu1/cu2=4.0
cu1/cu2=3.5
cu1/cu2=3.0
cu1/cu2=2.5
cu1/cu2=2.0
cu1/cu2=1.5
cu1/cu2=0.2-0.8
0.2
0 1 2 3 4 5

(c) d/H = 3.0
Figure 4.9. LA solutions of stability numbers for = 75°
(a) cu1/cu2 = 0.2
(b) cu1/cu2 = 0.5
(c) cu1/cu2 = 0.8
(d) cu1/cu2 = 1.5
Figure 4.10. Transition of UB plastic zones for  = 0.25, = 15°, 30° and 45°
and d/H = 1.5
113
(a) d/H = 1.5, 3 and 5.0 for = 15°
(b) d/H = 1.5, 3 and 5.0 for = 60°
Figure 4.11. Transition of UB plastic zones for  = 1.5 and c u1 /c u2 = 5.0
Figure 4.12. UB plastic zones for different  (0.25, 0.75 or 1.5) when = 45°,
d/H = 2 and c u1 /c u2 = 1.5
0.44 0.52
0.42 0.5
0.4 0.48
0.46
0.38
N2ci = cu1/HF
0.44
Ni=Cu0/HF
0.36
0.42
0.34
0.4
0.32 0.38
LB LB
0.3 0.36
UB UB
LEM(Bishop) 0.34 LEM(Bishop)
0.28 LEM (Janbu)
LEM (Janbu)
0.27 0.325
1.5 2 3 4 5 1.5 2 3
d/H
4 5
d/H
(a) β = 15° (b) β = 60°
Figure 4.13. Stability number difference based on different method
114
(a) = 15°, d/H = 2, cu1/cu2 = 0.2 and  = 0.5
(b) = 15°, d/H = 2, Cu0/Cu1 = 0.8 and  = 0.5
(c)= 15°, d/H = 2, cu1/cu2 = 3.0 and  = 0.5
(d) = 60°, d/H = 2, cu1/cu2 = 0.2 and  = 0.5
(e) = 60°, d/H = 2, cu1/cu2 = 0.8 and  = 0.5
(f) = 60°, d/H = 2, cu1/cu2 = 3.0 and  = 0.5
Figure 4.14. Comparisons of slope failure mechanisms between UB and the LEM
115
(a)= 45°, d/H = 1.5, cu1/cu2 = 1.5 and  = 0.25
(b)= 45°, d/H = 1.5, cu1/cu2 = 1.5 and  = 1.50
(c) = 45°, d/H = 1.5, cu1/cu2 = 1.5 and  = 5.00
Figure 4.15. Comparisons slope failure mechanisms for different 
Figure 4.16. Results of ANN slope stability trained with 3200 data set (  from 0
to 1.5)
116
Figure 4.17. Results of ANN slope stability trained validation data with 800 data
set ( from 0 to 1.5)
117
CHAPTER 5 PARAMETRIC STUDIES OF DISTURBED ROCK SLOPE

STABILITY BASED ON FINITE ELEMENT LIMIT ANALYSIS
METHODS
Z. G. Qian, A. J. Li, A.V. Lyamin, C. C. Wang
5.1 ABSTRACT
This study employs the finite element upper bound and lower bound limit analysis methods to
investigate the stability of inhomogeneous rock slopes. The differences in the stability numbers
of the upper and lower bound solutions are bracketed within ± 10.5% or better, and the stability
numbers obtained are presented in rock slope stability charts. These stability charts can provide
a convenient tool for preliminary stability designs of inhomogeneous rock slopes. Various
recommended blasting damage zones are considered, and disturbance factors are used to
represent damage levels. Results showed that rock mass disturbance could significantly
influence the evaluation of rock slope stability.
5.2 INTRODUCTION
The stability of rock slopes has long been regarded as a classic and difficult problem for
geotechnical engineers because rock masses are heterogeneous, discontinuous media
composed of rock materials and naturally occurring discontinuities, such as joints, fractures,
and bedding planes. Over the past few years, many investigators have evaluated rock slope
stability, thereby developing many new methods for meeting engineering requirements, such
as the limit equilibrium method (Bishop 1955; Chang 2002; Saade et al. 2016), the finite
element method (Yao et al. 2007; Mahtab and Goodman 1970; Hammah et al. 2004; Shen et
al. 2013) and the limit analysis methods (Lyamin and Sloan 2002; Lyamin and Sloan 2002). In
the assessment of rock slope stability, the use of rock mass parameters directly from a nonlinear
failure envelope is known to yield a highly accurate safety factor. As indicated by Merifield et
al. (2006), the Hoek-Brown failure criterion Hoek et al. (2002) is one of the most widely
accepted method for evaluating rock slope stability. This criterion features a nonlinear failure
envelope, which was first proposed for estimating the strength of heavily jointed rock masses
Hoek and Brown (1998a).
118
As design tools, slope stability charts have also drawn the attention of many investigators
(Taylor 1937; Baker et al. 2006; Li et al. 2008,2011). In essence, chart solutions can facilitate
the evaluation of slope stability and provide meaningful feedback, such as safety factors. In
1937, Taylor proposed the first set of stability charts for purely cohesive uniform soil slopes
(Taylor 1937). These chart solutions have since been extended to account for different
materials and loading types (Li et al. 2008; Huat et al. 2006; Taheri and Tani 2010; Yu et al.
1998; Michalowski 2002). However, most existing stability charts are designed exclusively for
soil slopes and are based on the conventional Mohr - Coulomb equivalent parameters.
Stability charts today rarely use rock strength parameters on the basis of a nonlinear failure
criterion. In particular, no comprehensive stability chart is available for inhomogeneous rock
slopes by considering varying rock mass disturbance in a slope. Therefore, the aim of the
present study is to investigate the influence of rock mass disturbance caused by blasting on the
assessment of rock slope stability. The obtained results are presented in the form of stability
charts which are useful tools for preliminary designs. The finite element upper bound (UB) and
lower bound (LB) limit analysis (LA) methods (Lyamin and Sloan 2002; Lyamin and Sloan
2002; Krabbenhoft et al. 2005) based on the Hoek–Brown failure criterion are employed in this
study.
5.3.1 Applicability of the generalised Hoek - Brown failure criterion
The Hoek - Brown failure criterion has been applied to a large number of projects around the
world Hoek et al. (2002). The Hoek - Brown failure criterion is one of the few nonlinear criteria
used to estimate rock mass strength and determine the nonlinear characteristics of strength
envelopes (Merifield et al. 2006; Zhao 2000). In 1974, Hoek and Bray published the first
edition devoted to rock slopes. In 1980, the original Hoek-Brown failure criterion was
developed and published (Hoek and Brown 1980b,1980). In fact, there is no suitable methods
are persented to estimating rock mass strength. In 1981, Hoek and Bray introduced a non -
linear failure criterion. In addition, the design charts were presented to extend Taylor (Taylor
1948) design charts including the influence of critical tension crack and of groundwater. In the
same book, the blasting effects on rock structure were also introduced. However, in the 80s of
the last century, the software investigating rock slopes are all based on Mohr Coulomb
criterion. Therefore, in 1983, Hoek persented the relationship between non-linear rock material
constants and the Mohr-Coulomb criterion. It can transformed the rock parameters into soil
119
parameters as software inputs (Sarma 1979). In 1992, both failure criterion for intact rock and
jointed rock mass have been developed (Hoek et al. 1992). The modified Hoek Brown criterion
considered zero tensile strength for poor rock masses. It put the research project from the just
theory of intact rock analysis into more meaningful jointed rock mass practical engineering
applications. In 1995, Hoek suggested that use original Hoek Brown criterion to analysis
quality rock masses and use modified Hoek Brown criterion for poor rock masses analysis.
Besides, Hoek et al. (1995) start to considered a new concept named Geological Strength Index
(GSI). In 1997, the Hoek - Brown failure criterion was expanded and integrated with the GSI
(Hoek and Brown 1997). The latest version of the Hoek - Brown failure criterion is expressed
as the Equations (5.1 - 5.4).
𝑎
𝜎′
𝜎1′ = 𝜎3′ + 𝜎𝑐𝑖 (𝑚𝑏 𝜎 3 + 𝑠) (5.1)
𝑐𝑖
where:
GSI−100
𝑚𝑏 = 𝑚𝑖 exp (28 − 14𝐷) (5.2)
GSI−100
𝑠 = exp ( ) (5.3)
9 − 3𝐷
1 1
𝑎= + 6 (𝑒 −GSI/15 − 𝑒 −20/3 ) (5.4)
2
ci is the uniaxial compressive strength. These equations show that mb, s and a are all dependent
on the GSI Marinos et al. 2007.
GSI was introduced to estimate rock mass strength under different geological conditions
because of the fact that Bieniawski’s rock mass rating system (Bieniawski 1976) and the Q -
system (Barton 2002) were determined to be unsuitable for extremely poor rock masses (Hoek
and Brown 1997; Hoek and Kaiser 1995; Hoek and Marinos 2007). In general, a larger GSI
indicates the good quality of rock masses. A new parameter, called the disturbance factor (D),
is introduced to deal with rock mass disturbance cited by Hoek et al. (2002), such as blasting
damage. The magnitude of D is determined on the basis of engineering experiences. The value
of D ranges from 0 for undisturbed in situ rock masses to 1 for disturbed rock mass properties
(Hoek et al. 2002). As expected, rock masses slopes subjected to good blasting achieve higher
rock mass strength compared with seriously disturbed rock mass slopes. Hoek (2012)
highlighted that a constant D applied to an entire rock mass could significantly underestimate
120
the strength and stability of the overall rock mass. Therefore, D should only be applied to actual
damaged zone.
5.3.2 Mohr - Coulomb failure criterion
The Mohr-Coulomb failure criterion is a linear equation, as shown in Figure 5.1 (Labuz and
Zang 2015). Following Hoek et al. (2002), the equivalent Mohr - Coulomb strength parameters
ϕ′ and c ′ are expressed as Equations (5.5 and 5.6). Most geotechnical software programs are
written in terms of the Mohr - Coulomb failure criterion (Li et al. 2008; Hoek 1990). Therefore,
for the analysis of rock slope stability, the equivalent values of ϕ′ and c ′ must be evaluated on
the basis of rock strength parameters, which could affect the accuracy of the results.
′ ) 𝛼−1
′ −1 6 𝛼 𝑚𝑏 (𝑠+ 𝑚𝑏 𝜎3𝑛
ϕ = sin [ ′ ) 𝛼−1 ] (5.5)
2 (1+𝛼)(2+𝛼)+6 𝛼 𝑚𝑏 (𝑠+ 𝑚𝑏 𝜎3𝑛
′ ] (𝑠+ 𝑚 𝜎 ′ ) 𝛼−1
𝜎𝑐𝑖 [ (1+2𝛼) 𝑠+(1−𝛼) 𝑚𝑏 𝜎3𝑛
c′ = 𝑏 3𝑛
(5.6)
6 𝛼 𝑚𝑏 (𝑠+𝑚𝑏 𝜎′3𝑛 )𝛼−1
(1+𝛼)(2+𝛼)√1+
(1+𝛼)(2+𝛼)
where
′ ′
𝜎3𝑛 = 𝜎3𝑚𝑎𝑥 /𝜎𝑐𝑖 (5.7)
−0.91
′ ′ 𝑐𝑚 𝜎′
𝜎3𝑚𝑎𝑥 = 0.72 𝜎𝑐𝑚 [ γH ] (5.8)
𝑚
(𝑚𝑏 +4𝑠−𝛼(𝑚𝑏 −8𝑠)) ( 𝑏 +𝑠)𝛼−1
′
𝜎𝑐𝑚 = 𝜎𝑐𝑖 2 ( 1+𝛼)( 2+𝛼)
4
(5.9)
′
H is the slope height,  is the material unit weight, and 𝜎𝑐𝑚 is the compressive strength of the
rock mass.
5.3.3 Stability number and stability charts
The results obtained in this study for inhomogeneous rock slopes are presented using a
dimensionless stability number, Ni, the definition of which is shown in Equation (5.10). The
form of Equation (5.10) is the same as that proposed by Li et al. (2008), but the results obtained
in the present study are different because of the different property inputs for the rock masses.
Ni = 𝜎𝑐𝑖 /γHF (5.10)
121
where
F is the safety factor.
The results obtained in this study are presented in the form of stability charts. Such chart
solutions are regarded as convenient tools that facilitate the process of determining safety
factors. In fact, most slope stability charts still require Mohr - Coulomb strength parameters as
inputs (Taheri and Tani 2010; Zanbak 1983). Only a few stability charts are based on rock mass
parameters (Li et al. 2008; Li et al. 2009). A review of the literature revealed that the evaluation
of rock slope stability using soil parameters fails to consider the non - linear nature of rock
mass failure envelope (Fu and Liao 2010). Therefore, this study is focussed on the assessment
of inhomogeneous rock slope stability on the basis of the Hoek-Brown failure criterion.
5.3.4 Limit analysis of rock slope stability
The limit equilibrium method (LEM) is the most well - known approach used to deal with slope
stability problems (Cassidy et al. 2008; Duncan et al. 2008). Based on physical theory
(Michalowski 1995), a slope is divided into several slices, each of which is assumed to be
purely based on the principles of statics. The failure surface must be assumed before the safety
factor is calculated. The results become inaccurate as the critical slip surface is assumed. The
literature review revealed that not many studies use Hoek - Brown material parameters in rock
slope analysis based on the LEM (Xie et al. 2006; Lin et al. 2014). For the past few years, LA
based on UB and LB have been proven to be robust and efficient in estimating the strength of
rock slopes (Loukidis et al. 2003).
LA is based on two separate methods, namely, the UB analysis that uses the theorem of
kinematics and the LB analysis that considers the static theorem. For the UB theorem, the
power dissipated by any kinematically admissible velocity field is assumed to be equal to the
power dissipated by the external loads (Lyamin and Sloan 2002a). LB analysis is based on
plasticity theory that considers stress distribution, which maximizes the intensity of a
predefined external load distribution, in this case, stress satisfies a statically equilibrium
equation (Sloan 1988; Chen et al. 2008; Lyamin and Sloan 2002). LA methods (Lyamin and
Sloan 2002a; Krabbenhoft et al. 2005) can generate the failure mechanisms automatically.
Therefore, the drawback of the assumed failure surface in the LEM can be avoided. The results
also become increasingly reliable when UB and LB are employed. As indicated by Li et al.
122
(2009), the safety factor is overestimated (> 10%) according to the comparison of UB solutions.
In the present study, both of the UB and LB are adopted.
As indicated by Hoek et al. (2002), rock mass disturbance should be considered because of its
significant influence on the evaluation of rock slope stability. In the present study, rock mass
disturbance caused by blasting is considered by using the disturbance factor D. As shown in
Figure 5.2, a rock slope is divided into two regions to account for the damage zone. Region 1
is the disturbed zone caused by surface blasting; its shape is subjectively determined on the
basis of the study of Hoek and Karzulovic (2000). It should be noted that this study consider
all rock slopes as simple slopes. Therefore, the stability charts might give lead to certain
difference for a slope with very complex geometries.
Recent investigations showed that the damage zone thickness (T) may surround the excavation
perimeter to a depth of approximately 2 meters Chen and Liu 2012 to 100 meters or more
behind the excavated slope face (Marinos et al. 2005). In fact, T is influenced by the design of
the blast, such as blasting control and scale (Hoek and Karzulovic 2000). Region 2 in Figure
5.2 is the undisturbed rock mass, for which D = 0 is adopted. According to the study of Hoek
and Karzulovic (2000), T ranges from 0.5 H to 2.5 H. This range is considered in the present
study.
According to Hoek (2012), assigning only a constant D to a rock slope would be inappropriate.
Therefore, varying disturbance factor in a slope is taken into consideration for the Region 1 in
Figure 5.2. A simple variation in the disturbance factor, which decreases linearly with increase
in depth, is adopted in this study. D0 represents the rock mass disturbance on the excavated
slope surface. In this study, D0 is assumed to be either 1.0 or 0.7. The magnitude of D0 adopted
in this study is based on the proposal of Hoek et al. (2002). Accordingly, a range of parameters
shown in Table 5.1 is analysed. It should be noted that only GSI = 50 - 90 is considered in this
study because blasting is generally unnecessary in excavating rock masses with GSI ≤ 50
(Tsiambaos and Saroglou 2010).
By using LA methods, it is expected the true solutions bracketed within a small range. The
mesh must be redefined to achieve this expectation. However, the general mesh arrangement
for a slope with uniform rock messes could not yield satisfactory results, particularly for steep
slopes (β ≥ 60°). The results of UB and LB modelling tend to be increasingly sensitive as slope
123
angle increases. In general, conducting fine mesh analyses could lead to highly accurate results.
Therefore, significant refinement is needed, but such solution increases computation time. In
this study, the mesh arrangements (UB and LB) are selected only after considerable
refinements are made.
5.5 RESULTS AND DISCUSSIONS
5.5.1 LA solutions
The obtained average stability numbers (Ni) for β = 15º are shown in Figures 5.3 where includes
a range of damage zone thickness (T), GSI, mi and D0. In Figures 5.3, Ni increases as GSI
increases and mi decreases. A comparison of different D0 values in Figure 3 indicates that Ni
increases while D0 increases. The same result can be obtained by comparing Table 5.2 and 5.3.
For the inclinations of the other slopes (β = 30º - 75º), the average chart solutions are presented
in Figures 5.4 to 5.7, which also show the trends observed for β = 15º. As expected, the
observed phenomena are similar to the findings of Li et al. (2008, 2011), who investigated
homogeneous rock slopes.
The differences in the stability numbers of the UB and LB solutions for β ≤ 60º are bracketed
within ±10% or better. However, the differences in the UB and LB results for β = 75º are
difficult to maintain within ± 10% (i.e. ± 10.5%). When β = 75º, 19.4% UB and LB solutions
show differences range from ± 10% to ± 10.5%. The mesh arrangement is observed to be
sensitive to the results obtained, particularly the larger slope inclinations (β ≥ 60º). It should
be noted that all the differences between the UB and LB solutions are bracketed within ±10.5%
in this study.
As shown in Tables 5.2 and 5.3, when β increases, the difference in Ni of GSI = 50 and GSI =
90 increases. For example (Table 5.2), Ni for GSI = 50 is approximately 4 times that for GSI =
90 when β = 15º, whereas Ni for GSI = 50 is approximately 13 times that for GSI = 90 when β
= 75º. The difference in Ni of β = 15º and β = 75º increases with the increase in D0. The same
trend is depicted in Figures 5.8 and 5.9. When GSI = 50 for D0 = 0.7 (Table 5.2), Ni for β = 75º
is approximately 50 times that for β = 15º. However, when D0 = 1.0, Ni for β = 75º (table 5.3)
is approximately 65 times that for β = 15º. This trend declines with the increase in GSI. Tables
5.2 and 5.3 also show that, for a given β, mi and T, the difference in Ni of D0 = 1.0 and 0.7
becomes significant when GSI decreases. For GSI = 50, Ni in Table 5.3 is larger than that in
Table 5.2. The difference for β = 75º is approximately 50%. However, the maximum difference
124
is reduced to only 5% while GSI = 90. This result implies that the magnitude of D0 could
significantly influence the estimations of rock slope stability.
Figure 5.10 shows that Ni decreases as T decreases. This result implies that a large damage
zone reduce the safety factor, as expected. The Figure also shows a slight change in Ni between
T = 1.0 H and T = 2.5 H. This phenomenon is more evident when the slope is steep. The
difference in Ni observed between T = 1.0 H and T = 1.5 H for 60o is 7.72% which is less than
that for 30o (i.e. 19.5%).
A comparison of failure surfaces for different T values is shown in Figures 5.11, which also
presents the UB plastic zones. These UB plastic zones can be viewed as the rock slope failure
mechanisms (Sloan 2013; Lim et al. 2016). For the slope failure mode, no significant trends
can be deduced from the results when the values of T are changed. However, the failure surface
for T = 0.5H is slightly smaller than that for T = 1.5H and T = 2.5H, particularly for the slope
crest. Figure 5.12 displays several observed UB plastic zones for different slope angles given
D0 = 1.0, GSI = 60, mi = 20 and T = 0.5 H. The results indicate that the depth of the failure
surface increases with the reduction in the slope angle. In general, when β ≥ 30°, the failure
mode is almost unchanged. This condition is referred to as toe failure.
Figure 5.13 (a) and (b) show the UB plastic zones for β = 75o, D0 = 1.0, GSI = 50, T = 1.5 H
and different mi values. The depth of the slip surfaces decreases slightly as mi increases. The
plastic zones are almost unchanged for different D0 values. Thus, it can be concluded that, for
inhomogeneous slopes, the effects of D0 on the failure surface are insignificant.
5.5.2 Comparisons of LEM solutions obtained using equivalent Mohr - Coulomb

parameters
In using the Mohr - Coulomb failure criterion to evaluate rock slope stability, estimating the
equivalent ϕ′ and c ′ for the rock slope is required. Although rock or rockfill slopes can be
evaluated by using the Mohr - Coulomb criterion (Hoek 1990), transferring the rock strength
parameters to the appropriate ϕ′ and c ′ for slope stability assessments remains questionable.
According to Li et al. (2008; 2009), using ϕ′ and c ′ overestimated the factor of safety and may
result in poor predictions of failure surface, particularly for steep slopes. Therefore, the effects
of using the equivalent ϕ′ and c ′ parameters on the assessments of inhomogeneous rock slope
stability are worth investigating.
125
In the analyses based on the Mohr - Coulomb failure criterion in the present study, the
conventional LEM is employed following the Bishop Method (Bishop 1955). The rock strength
parameters are converted to the equivalent ϕ′ and c ′ for each rock slope by using RocLab. The
ϕ′ and c ′ parameters obtained are influenced by slope height, as shown in Equation (5.8). Given
that this study is based on inhomogeneous masses, D0 decreases linearly with depth. Therefore,
ϕ′ and c ′ as inputs are considered on the basis of varying D0 magnitudes. In fact, this
consideration could have a significant effect on the assessment of rock slope stability. For
comparison purposes, several cases selected from previous chart solutions are adopted.
Figure 5.14 (a) and (b) show different failure mechanisms obtained from the LEM based on
the Mohr - Coulomb failure criterion and the Hoek - Brown failure criterion, respectively.
These failure mechanisms are compared with the UB plastic zones based on the Hoek - Brown
failure criterion. The cases considered are of T = 2.5 H, mi = 10, GSI = 50, D0 = 1.0 for β = 15o
or 75o. The safety factors assessed on the basis of the Hoek - Brown failure criterion (LA) are
equal to 1. However, the safety factors assessed on the basis of the Mohr - Coulomb failure
criterion (LEM) are 0.98 and 1.311 for different slope angles. Such resulting difference
indicates that using the equivalent Mohr - Coulomb parameters could significantly
overestimate the safety factor, particular for steep slopes β ≥ 45o. This finding is similar to that
of Li et al. (2008), who investigated homogeneous rock slope. For the gentle slope (β = 15o)
shown in Figure 5.14 (a), the difference in F of the UB and LEM results is approximately 2%.
However, different failure surfaces can be observed. In particular, an increasingly shallow slip
surface is obtained when the equivalent Mohr - Coulomb parameters are used. The failure
surface predicted could influence the slope reinforcement consideration. It should be noted that
determining the slip surface prior to the calculation of F is unnecessary when using LA method.
5.5.3 Comparisons of LEM solutions obtained using Hoek - Brown parameters
As mentioned previously, by using the UB and LB methods, the true solutions can be
bracketed. In this section, the results obtained from the LEM and LA based on the Hoek -
Brown failure criterion are compared. The LEM in this study is based on the simplified 2D
Bishop method (Bishop 1955). Figure 5.15 shows that Ni obtained from the LEM is larger than
the UB results. The results obtained from the LEM are also close to the LB result. However,
Figure 5.15 shows that a large portion of Ni obtained from the LEM is larger than the LB result.
This outcome implies that the safety factor of the rock slope is underestimated when the LEM
is employed. This phenomenon is more evident when D0 increases.
126
5.6 CASE STUDIES
5.6.1 Rock slopes in Iran
The cases considered in the present study are adopted from the study of Taheri and Tani in
2010. A few mines or civil projects in Iran are selected, and the safety factors are presented on
the basis of the CLARA code. The selected cases are only for GSI ≥ 50, as shown in Table 5.4.
It must be point out that the GSI for the Angooran case is 48, and it is assumed as 50 which
should be still a reasonable value (Hoek et al. 1998; Li et al. 2012). In the investigation of
Taheri and Tani (2010), the rock mass disturbance factor (df) was considered following the
work of Sonmez et al. (2004). As the stability chart in this study uses D on the basis of the
Hoek – Brown failure criterion, df must be converted to D according to the study of Sonmez et
al. (2004). The values of D estimated in the case studies are based on Figure 5.16 and are shown
in Table 5. Figure 16 shows the df value proposed by Taheri and Tani (2010) and the D value
proposed by Hoek et al. (2002). It must be stressed that the information regarding the damage
zone (T) is unavailable. Therefore, T = 2.5 H is adopted to assess the rock slopes. This value
is regarded as conservative. The safety factors assessed by using stability charts are shown in
Table 5.5, which also indicates that the slopes are stable. The results obtained agreed with that
presented by Taheri and Tani (2010).
5.6.2 Rock slopes presented by Douglas
Table 5.6 lists the case studies on rock slopes presented by Douglas (2002). Given that each
case may have several stable or unstable slopes (labeled as a – c and others), γ is assumed to
be 25 kN/m3 for all cases shown in Table 5.6. This assumption is the same as that made by Li
et al. (2011) The present study considers a GSI ranging from 50 to 90 because GSI < 50 is not
discussed in this paper. Table 5.7 displays the safety factors estimated by using the
inhomogeneous chart solutions on the basis of the average UB and LB results. In the study of
Douglas (2002), T and D were not considered; thus, the case studies are assessed by assuming
D0 = 0.7 and T = 2.5 H. As indicated in the results, the slopes are stable for the stable cases
(1a, 2a, 2b, 2c, 3, 4, 5a and 5b), as shown in Table 5.6 and 5.7. For the unstable cases (1b, 1c,
1d, 1e and 6), differences are observed in the estimated safety factors and real slope states.
However, these slopes are assessed by assuming D, which does not represent the real
magnitude for unstable rock slopes.
127
The case studies also reveal that the analyses of rock slope stability for inhomogeneous rock
masses are highly influenced by D0 and T. On the basis of the comprehensive observations, T
= 1.5 H (Table 5.7) is also considered in the assessment of the same cases. The safety factor
based on T = 1.5 H is larger than that based on T = 2.5 H. Thus, F can be significantly larger if
the damaged zone caused by blasting can be effectively controlled. The solutions for T = 2.5
H are also relatively close to those presented in the study of Li et al. (2011), where the
homogeneous rock slopes were investigated. This outcome indicates that the solutions of Li et
al. (2011) can be applied if T > 2.5 H. The maximum damage zone considered in this study (T
= 2.5 H), as proposed by Hoek and Karzulovic (2000), is reasonable for rock slope evaluations.
5.6.3 Slope failure in closely jointed rock mass in barite open pit mine
This case study is from Baskoyak barite open pit mine, in western Anatolia. The slope was also
investigated by Sonmez and Ulusay (1999) and Li et al. (2011; 2016). From Sonmez and
Ulusay , the unit weight and uniaxial compressive strength of the heavily broken part of the
schist are considered as 22.2 kN/m3 and 5.2 MPa respectively. In addition, mi = 7 and GSI =
16 were estimated. It should be noted that, the rock mass disturbance was considered as
inhomogeneous which is approximately similar to the previous investigations of this present
study. Moreover, T is 2.5H (H = 20m) and D linearly decreased from 0.7 to 0 in this damage
zone. Besides, the rock slope was considered as dry condition since there are no groundwater
in the excavation area. It should be pointed out that this case would not be ideal because
overburden materials are removed by excavators without any blasting. In fact, the damage zone
suggested by Hoek and Karzulovic (2000) is to account for blasting effect. However, this case
is still employed herein to investigate inhomogeneous effects of D on rock slopes in T. The
obtained failure surface can be seen in Figure 5.17. The comparison of the slip surface with
Sonmez and Ulusay (1999) and Li et al. (2011) indicated that the result by considering varying
D in T can still provide a similar failure shape. It revealed that the suggested T by Hoek and
Karzulovic (2000) would be reasonable. In addition, the failure surface in the present study is
slightly shallower than the homogeneous one from Li et al. (2011). It agrees well with the
findings in this study. However, more case studies are still needed to verified the effects of
varying D in T on rock slope stability assessments.
5.7 CONCLUSIONS
This study used UB and LB methods to estimate rock slope stability for inhomogeneous rock
masses. The stability number are presented as chart solutions on the basis of the Hoek–Brown
128
failure criterion. The stability charts of the inhomogeneous rock mass can facilitate the
evaluation of safety factors, thereby reducing the computation time. This study also
investigated the influence of damage zones caused by blasting by considering the different
levels of rock mass disturbance. The disturbance factor for slope surface is assumed to be either
1.0 or 0.7, which linearly decreases with the increase in depth. The following conclusions are
drawn from the results:
In all the case studies, the UB and LB solutions maintain the stability numbers within ± 10.5%.
Such result indicates that true solutions for the evaluation of inhomogeneous rock slope
stability should locate within this range.
The stability numbers (Ni) increase with the increase in β and D and the decrease in GSI and
mi. In addition to the significance influence of D on Ni, the blasting damage zone thickness (T)
can substantially affect the assessment of rock slope stability. A large damage zone makes a
slope increasingly critical. However, when T = 2.5 H, the safety factor for an inhomogeneous
rock slope is close to that for a homogeneous slope.
Based on the observations on the UB plastic zones, mi and T can have a slight effect on failure
mechanisms. Other Hoek-Brown strength parameters do not influence failure mechanism
obviously.
Using the LEM based on the equivalent Mohr - Coulomb parameters could significantly
overestimate the safety factor for inhomogeneous rock slopes. This phenomenon is more
evident when the slopes are steep. The failure mechanisms obtained from the LEM do not
exactly match the LA results. For rock slope assessments, engineers should use the equivalent
Mohr - Coulomb parameters carefully.
129
5.8 REFERENCES
Saade A, Abou-Jaoude G and Wartman J (2016) Regional-scale co-seismic landslide

assessment using limit equilibrium analysis. Engineering Geology2016. p. 53-64.
Yao Y-P, Zhou A-N and Lu D-C (2007) Extended transformed stress space for geomaterials
and its application. Journal of Engineering Mechanics2007. p. 1115-23.
Mahtab MA and Goodman R (1970) Three dimensional finite element analysis of jointed rock
slopes. 1970.
rock slopes using the finite element method. Proceedings of the ISRM Regional Symposium
Shen J, Karakus M and Xu C (2013) Chart-based slope stability assessment using the
Generalized Hoek–Brown criterion. International Journal of Rock Mechanics and Mining
Sciences2013. p. 210-9.
Lyamin A and Sloan S (2002) Lower bound limit analysis using non‐linear programming.
International Journal for Numerical Methods in Engineering2002. p. 573-611.
Lyamin AV and Sloan S (2002) Upper bound limit analysis using linear finite elements and
Geomechanics2002. p. 181-216.

130
Hoek E and Brown ET (1980a) Empirical strength criterion for rock masses. Journal of
Geotechnical and Geoenvironmental Engineering1980a.
689-700.
Huat BB, Ali FH and Rajoo R (2006) Stability analysis and stability chart for unsaturated
residual soil slope. American Journal of Enviromental Sciences2006. p. 154-60.
p. 1069-88.
Merifield R, Lyamin A and Sloan S (2006) Limit analysis solutions for the bearing capacity of
rock masses using the generalised Hoek–Brown criterion. International Journal of Rock
Zhao J (2000) Applicability of Mohr–Coulomb and Hoek–Brown strength criteria to the

dynamic strength of brittle rock. International Journal of Rock Mechanics and Mining
Sciences2000. p. 1115-21.
131
Hoek E and Bray J (1974) Rock Slope Engineering, 309 p. London: Institution of Mining &
Metallurgy1974.
Hoek E and Brown ET (1980b) Underground excavations in rock1980b.
Hoek E and Brown ET (1980) Empirical strength criterion for rock masses. Journal of
Geotechnical and Geoenvironmental Engineering1980.
Taylor DW (1948) Fundamentals of soil mechanics. Soil Science1948. p. 161.
Hoek E (1983) Strength of jointed rock masses. Geotechnique1983. p. 187-223.
Sarma SK (1979) Stability analysis of embankments and slopes. Journal of the Geotechnical
Engineering Division1979. p. 1511-24.
Hoek E, Wood D and Shah S (1992) editorêditors. A modified Hoek-Brown failure criterion
for jointed rock masses. Proc Int Conf Eurock; 1992.
Hoek E, Kaiser P and Bawden W (1995) Support of underground excavations in hard rock AA
Balkema Rotterdam 215. 1995.
Marinos P, Marinos V and Hoek E (2007) Geological Strength Index (GSI). A characterization
tool for assessing engineering properties for rock masses. Underground works under special
conditions Taylor and Francis, Lisbon2007. p. 13-21.
Bieniawski ZT (1976) Exploration for rock engineering: proceedings of the Symposium on

Exploration for Rock Engineering, Johannesburg, 1-5 November 1976: AA Balkema; 1976.
Barton N (2002) Some new Q-value correlations to assist in site characterisation and tunnel
design. International journal of rock mechanics and mining sciences2002. p. 185-216.
Hoek E and Kaiser P (1995) Bawden. WF (1995). Support of underground excavations in hard
rock1995.
Hoek E and Marinos P (2007) A brief history of the development of the Hoek-Brown failure
criterion. Soils and rocks2007. p. 1-8.
132
Hoek E (1990) editorêditors. Estimating Mohr-Coulomb friction and cohesion values from
the Hoek-Brown failure criterion. International Journal of Rock Mechanics and Mining
Sciences & Geomechanics Abstracts; 1990: Elsevier.
681-91.

283-93.
Xie M-W, Cai M-F and Esaki T (2006) GIS-based three-dimensional slope stability limit
equilibrium method and application. Yantu Lixue(Rock and Soil Mechanics)2006. p. 117-22.
Lin H, Zhong W, Xiong W and Tang W (2014) Slope stability analysis using limit equilibrium
method in nonlinear criterion. The Scientific World Journal2014.
133
391-415.
Lyamin AV and Sloan S (2002) Lower bound limit analysis using non‐linear programming.
International Journal for Numerical Methods in Engineering2002. p. 573-611.
Lyamin AV and Sloan SW (2002a) Lower bound limit analysis using non-linear programming.
International Journal for Numerical Methods in Engineering2002a. p. 573-611.
p. 1069-88.
Chen W-F and Liu X (2012) Limit analysis in soil mechanics: Elsevier; 2012.
Tsiambaos G and Saroglou H (2010) Excavatability assessment of rock masses using the
Geological Strength Index (GSI). Bulletin of engineering geology and the environment2010.
p. 13-27.
Sloan S (2013) Geotechnical stability analysis. Géotechnique2013. p. 531-71.
Lim K, Li A, Schmid A and Lyamin A (2016) Slope-Stability Assessments Using Finite-

Element Limit-Analysis Methods. International Journal of Geomechanics2016. p. 06016017.
134
p. 11-8.
Sonmez H, Gokceoglu C and Ulusay R (2004) Indirect determination of the modulus of

deformation of rock masses based on the GSI system. International journal of rock mechanics
and mining sciences2004. p. 849-57.
Douglas K (2002) The shear strength of rock masses. In School of civil and environmental
engineering Sydney, Australia: The University of New South Wales; 20022002.
Sonmez H and Ulusay R (1999) Modifications to the geological strength index (GSI) and their
applicability to stability of slopes. International Journal of Rock Mechanics and Mining
Sciences1999. p. 743-60.
135
5.9 APPENDIX
Table 5.1. Parameters investigated in this study
GSI 50, 60, 70, 80, 90

mi 5, 10, 15, 20, 25, 30, 35
D0 0.7, 1.0
β 15°, 30°, 45°, 60°, 75°
T 0.5H, 1.0H, 1.5H, 2.0H, 2.5H
Table 5.2. Average stability numbers for T = 0.5 H, mi = 5 and D0 = 0.7
GSI β = 15º β = 30º β = 45º β = 60º β = 75º

50 0.163 0.623 1.712 4.122 8.531
60 0.113 0.391 1.011 2.270 4.418
70 0.078 0.250 0.606 1.270 2.295
80 0.054 0.161 0.366 0.714 1.210
90 0.037 0.104 0.220 0.402 0.656
Table 5.3. Average stability numbers for T = 0.5 H, mi = 5 and D0 = 1.0
GSI β = 15º β = 30º β = 45º β = 60º β = 75º

50 0.177 0.854 2.549 6.181 12.146
60 0.116 0.487 1.353 3.067 5.750
70 0.080 0.288 0.739 1.566 2.787
80 0.055 0.174 0.412 0.811 1.381
90 0.038 0.107 0.232 0.427 0.692
136
Table 5.4. The rock type, GSI, mi and other relevant information for Iran slope case studies
Intact rock properties Final designed geometry

Mine Case GSI Stable
γ (kN/m3) σci (MPa) mi df H (m) β
Jajarm 58 27 172 9 0.85 135 65 Yes
Angooran 48 30 104 7 0.95 170 55 Yes
Table 5.5. Assessment of the safety factor by using LA charts solution
Final assumed
Numerical assessment
D estimated based on Figure geometry
Mine case
16 Safety
Ground water GSI Stable
factor
Jajarm 0.9 None 58 10.72 Yes
Angooran 0.75 None 50 5.22 Yes
Table 5.6. Summary of slope data from cases studies base on rock parameters
Geology Mine case H (m) β GSI mi Stable

Saprolite / Basalt 1a 70 49 40–50 25 Yes
Saprolite / Basalt 1b 41 50 40–50 25 No
Saprolite / Basalt 1c 41 55 40–50 25 No
Saprolite / Basalt 1d 46 49 40–50 25 No
Saprolite / Basalt 1e 57 50 40–50 25 No
Saprolite / Basalt 2b 60 48 50–60 25 Yes
Saprolite / Basalt 2c 60 52 50–60 25 Yes
Mudstone / Siltstone 3 38 39 50–60 7 Yes
Breccia 4 200 65 70–80 19 Yes
Siltstone 5a 157 48 60–70 7 Yes
Siltstone 5b 60 53 60–70 7 Yes
Siltstone 6 110 48 40–50 7 No
137
Table 5.7. The factor of safety analysis by using different chart solutions
Inhomogeneous Inhomogeneous
Homogeneous
Min σci solutions solutions
σci solutions
(T = 2.5H, D0 = (T = 1.5H, D0 =
e Stable
(MPa) γH (D0 = 0.7)
0.7) 0.7)
case
GSI F F F
1.7 40 1.07 - -
1a 3 Yes
1 50 2.06 2.29 2.35
2.9 40 1.90 - -
1b 3 Yes
3 50 3.80 3.85 3.90
2.9 40 1.33 - -
1c 3 Yes
3 50 2.41 2.66 2.93
2.6 40 1.76 - -
1d 3 Yes
1 50 3.51 3.48 3.57
2.1 40 1.37 - -
1e 3 Yes
1 50 2.74 2.77 2.81
3.4 50 4.31 4.54 4.6
2a 5 Yes
5 60 7.30 8.21 8.62
3.3 50 4.56 4.63 4.76
2b 5 Yes
3 60 8.12 8.77 9.01
3.3 50 3.70 4.17 4.27
2c 5 Yes
3 60 5.55 6.41 7.94
5.2 50 4.04 4.10 4.38
3 5 Yes
6 60 6.49 6.42 6.58
70 35.3 37.5 38.46
4 150 30 Yes
80 60.0 61.22 62.50
8.8 60 5.18 7.38 7.70
5a 23 Yes
6 70 8.02 12.48 12.65
15. 60 10.9 10.92 11.76
5b 23 Yes
3 70 16.6 16.83 20.24
40 1.82 - -
6 25 9.1 Yes
50 3.95 4.14 4.33
138
Figure 5.1. Comparison between the Hoek - Brown failure criterion and
equivalent Mohr - Coulomb failure envelope.
Figure 5.2. D linearly decreased with depth in Region 1 for rock slope model.
139
ROCK SLOPE STABILITY CHARTS

0.2 0.3
T = 0.5 H T = 1.0 H
0.1
0.1
0.01
0.01
0.005 0.005
5 10 15 20 25 30 35 5 10 15 20 25 30 35
0.4 0.5
T = 1.5 H T = 2.0 H
0.1
0.1
Ni = ciHF
0.01 0.01
0.005 0.005
0.5 5 10 15 20 25 30 35 5 10 15 20 25 30 35
T = 2.5 H
GSI 50 (D0 = 1.0)

0.1 GSI 60 (D0 = 1.0)
GSI 70 (D0 = 1.0)
GSI 80 (D0 = 1.0)
GSI 90 (D0 = 1.0)
GSI 50 (D0 = 0.7)
GSI 60 (D0 = 0.7)
GSI 70 (D0 = 0.7)
0.01 GSI 80 (D0 = 0.7)
GSI 90 (D0 = 0.7)
0.005
5 10 15 20 25 30 35
mi
Figure 5.3. Limit analysis solutions of stability numbers for inhomogeneous

rock slopes β = 15 o
140
1 2
T = 0.5 H T = 1.0 H
1
0.1
0.1
0.01 0.01
5 10 15 20 25 30 35 5 10 15 20 25 30 35
2 2
1 T = 1.5 H 1 T = 2.0 H
0.1 0.1
0.01 0.01
5 10 15 20 25 30 35 5 10 15 20 25 30 35
2
1 T = 2.5 H
GSI 50 (D0 = 1.0)

GSI 60 (D0 = 1.0)
GSI 70 (D0 = 1.0)
GSI 80 (D0 = 1.0)
0.1 GSI 90 (D0 = 1.0)
GSI 50 (D0 = 0.7)
GSI 60 (D0 = 0.7)
GSI 70 (D0 = 0.7)
GSI 80 (D0 = 0.7)
GSI 90 (D0 = 0.7)
0.01
5 10 15 20 25 30 35
mi

141
3 4
1
1
0.1 0.1
0.03 0.03
5 10 15 20 25 30 35 55 10 15 20 25 30 35
5
T = 1.5 H T = 2.0 H
1 1
0.1 0.1
0.03 0.03
5 10 15 20 25 30 35 5 10 15 20 25 30 35
5
T = 2.5 H
GSI 50 (D0 = 1.0)

1 GSI 60 (D0 = 1.0)
GSI 70 (D0 = 1.0)
GSI 80 (D0 = 1.0)
GSI 90 (D0 = 1.0)
GSI 50 (D0 = 0.7)
GSI 60 (D0 = 0.7)
0.1 GSI 70 (D0 = 0.7)
GSI 80 (D0 = 0.7)
GSI 90 (D0 = 0.7)
0.03
5 10 15 20 25 30 35
mi

142
7 9
T = 0.5 H T = 1.0 H
1
1
0.1 0.1
95 10 15 20 25 30 35 5
10
10 15 20 25 30 35
T = 1.5 H T = 2.0 H
1 1
0.1 0.1
10 5 10 15 20 25 30 35 5 10 15 20 25 30 35
T = 2.5 H
GSI 50 (D0 = 1.0)

GSI 60 (D0 = 1.0)
GSI 70 (D0 = 1.0)
GSI 80 (D0 = 1.0)
1 GSI 90 (D0 = 1.0)
GSI 50 (D0 = 0.7)
GSI 60 (D0 = 0.7)
GSI 70 (D0 = 0.7)
GSI 80 (D0 = 0.7)
GSI 90 (D0 = 0.7)
0.1
5 10 15 20 25 30 35
mi

143
20 20
T = 0.5 H T = 1.0 H
10 10
1 1
0.3 0.3
5 10 15 20 25 30 35 5 10 15 20 25 30 35
20 20
T = 1.5 H T = 2.0 H
10 10
1 1
0.3 0.3
20 5 10 15 20 25 30 35 5 10 15 20 25 30 35
T = 2.5 H
10
GSI 50 (D0 = 1.0)
GSI 60 (D0 = 1.0)
GSI 70 (D0 = 1.0)
GSI 80 (D0 = 1.0)
GSI 90 (D0 = 1.0)
GSI 50 (D0 = 0.7)
1 GSI 60 (D0 = 0.7)
GSI 70 (D0 = 0.7)
GSI 80 (D0 = 0.7)
GSI 90 (D0 = 0.7)
0.3
5 10 15 20 25 30 35
mi

144
200 o 200
= 15
o
100 100 = 15
o
= 30
o
= 30
o
= 45
o
= 45
o
= 60
o
10 10 = 60
o
= 75
o
= 75
1 1
0.1 0.1
0.02 0.02
5 10 15 20 25 30 35 5 10 15 20 25 30 35
mi mi
(a) D0 = 0.7 (b) D0 = 1.0
Figure 5.8. Stability numbers for GSI = 60, T = 2 H and different β
20 o 20
= 15 = 15
o
10 o
10
= 30 = 30
o
o
= 45 = 45
o
o
= 60 = 60
o
1 1
o
= 75 = 75
o
Ni = ciHF
0.1 0.1
0.01 0.01
0.004 0.004
5 10 15
5 10 15
mi 20 25 30 35
mi20 25 30 35
(a) D0 = 0.7 (b) D0 = 1.0

Figure 5.9. Stability numbers for GSI = 90, T = 0.5 H and different β
10 2
T = 0.5 H
8 T = 0.5 H
T = 1.0 H T = 1.0 H
6 T = 1.5 H 1 T = 1.5 H
T = 2.0 H T = 2.0 H
Ni = ciBHF
T = 2.5 H T = 2.5 H
4
1.2 0.1
5 10 15 20 25 30 35 5 10 15 20 25 30 35
mi mi
(a) β = 60o (b) β = 30o
Figure 5.10. Stability numbers for β = 60 o or 30 o , D 0 = 1.0, GSI = 50 and
different T
145
(a) T = 0.5 H
(b) T = 1.5 H
(c) T = 2.5 H
Figures 5.11. Failure surfaces for β = 60 o , GSI = 70, D 0 = 1.0, m i = 10 and

different T
146
(a) β = 15o
(b) β = 30o
(c) β = 45o
(d) β = 60o
(e) β = 75o
Figures 5.12. Failure surface for GSI = 60, D 0 = 1.0, m i = 20, T = 0.5 H and
different β
147
(a) mi = 5
(b) mi = 35
Figure 5.13. Comparison of the upper bound plastic zones for β = 75 o , GSI = 50,
D 0 = 1.0 and T = 1.5 H
(a) Failure surface based on Mohr-Coulomb method when β = 15o or 75o
(b) Failure surface based on Hoek-Brown when β = 15o or 75o
Figure 5.14. Failure surface comparisons between the LEM and upper bound
results for T = 2.5 H, m i = 10, GSI = 50, D 0 = 1.0
148
20
LEM
10 UB
Ni = ciHF LB
= 75
o
1 o
= 60
o
= 45
0.1
= 30
o
= 15
o
0.02 mi
5 10 15 20 25 30 35
(a) D0 = 0.7
50
LEM
UB
LB
10
= 75
o
o
= 60
1
= 45
o
= 30
o
0.1
= 15
o
0.03 mi
5 10 15 20 25 30 35
(b) D0 = 1.0
Figure 5.15. Comparisons of the upper bound, lower bound and LEM solutions
for T = 1.5 H, GSI = 50 and different β
149
Figure 5.16. Disturbance factors relation between Sonmez et al. (2004) and
Hoek et al. (2002) adapted from Sonmez et al. (2004)
Figure 5.17. Comparisons of the slip area present by D linearly decreased with T
(band), plastic zones given by Li et al (between two red lines) and Sonmez and
Ulusay (black line)
150
CHAPTER 6 SEISMIC DISTURBED ROCK SLOPE STABILITY

ASSESSMENTS BASED ON FINITE ELEMENT LOWER BOUND
LIMIT ANALYSIS METHOD
Z. G. Qian, A. J. Li, W. C. Chen, A. V. Lyamin
6.1 ABSTRACT
Seismic effects are the most commonly considered factors in rock slope safety design. This
study employs the finite element lower bound limit analysis method to investigate the seismic
stability of inhomogeneous rock slopes on the basis of Hoek - Brown failure criterion.
Moreover, this research investigates different earthquake magnitudes by considering various
seismic coefficients. Results are presented as seismic rock slope stability charts. The chart
solutions can provide a reasonable tool for the preliminary evaluations of the safety factors that
affect seismic rock slope stability. In addition, various recommended blasting damage zones
are considered in this research. The blasting damage zone thickness considers from 0.5 to 2.5
times of the slope height. The rock mass disturbance considered as 0.7 or 1.0 and linearly
decreases in the blasting damage zone. Results show that earthquake and rock mass disturbance
may have a significant influence on the evaluation of rock slope safety factors. Moreover, this
research also considers the limit equilibrium method and Mohr - Coulomb equivalent strength
parameters to analyse the rock slope stability for comparison.
6.2 INTRODUCTION
The effects of earthquakes are often an engineering challenge in the assessment of rock slope
stability. Current investigations have shown that slope failure can occur during earthquakes
(Chang et al. 1984, Li et al. 2009). Therefore, the impact of earthquakes must be considered in
rock slope analyses. Over the past few years, many investigators have evaluated seismic rock
slope stability (Yang et al. 2004, Li et al. 2007, Li et al. 2009), however, no widely accepted
method has been established to deal with seismic designs for inhomogeneous rock slopes.
Merifield et al. 2006 indicated that the Hoek - Brown failure criterion (Hoek et al. 2002) is one
of the most widely accepted failure criteria for evaluating rock slope stability. The Hoek -
Brown failure criterion is a non - linear failure criterion developed to estimate heavily jointed
151
rock masses (Hoek and Brown 1980b, Hoek 1988). A review of literature revealed that
assessment of rock slope stability using rock mass parameters from a nonlinear failure envelope
can directly influence the accuracy of the obtained factor of safety (Li et al. 2008.)
Taylor proposed the first set of stability charts for purely cohesive uniform soil slopes. Stability
chart solutions as design tools have drawn the attention of many investigators (Michalowski
2002, Baker et al. 2006, Michalowski 2010, Li et al. 2010; Qian et al. 2014). However, most
existing stability charts are designed only for soil slopes and based on Mohr - Coulomb
equivalent parameters (Li et al. 2008; Qian et al. 2017). Only a few chart solutions based on
the Hoek - Brown failure criterion have been presented recently (Li et al. 2008, Li et al. 2009,
Li et al. 2011, Shen et al. 2013, Jiang et al. 2016). It should be noted that a seismic rock slope
stability chart that considers inhomogeneous rock masses is lacking. Therefore, this research
investigates the seismic stability of inhomogeneous disturbed rock slope by using the Hoek -
Brown failure criterion. The obtained results are presented as chart solutions for convenience.
6.3.1 The Generalised Hoek - Brown failure criterion
The Hoek - Brown failure criterion is a widely accepted nonlinear criterion that estimates rock
mass strength (Hoek et al. 2002). The first Hoek - Brown failure criterion was developed and
published in 1980 (Hoek and Brown 1980b). The Hoek - Brown failure criterion is defined by
Equations (6.1 – 6.4) based on Hoek et al. 2002. Rock mass strength can be affected by different
geological conditions, however, Bieniawski’s rock mass rating system (Bieniawski 1976) and
the Q-system (Barton 2002) are unsuitable for poor quality rock mass analysis (Hoek and
Kaiser 1995; Hoek and Marinos 2007; Hoek and Brown 1997). Thus, the Hoek - Brown failure
criterion introduced the Geological Strength Index (GSI) (Hoek and Brown 1997; Hoek 1994).
GSI is used to estimate the reduction in rock mass strength (Hoek et al. 1998). A considerably
large GSI indicates good quality of rock masses. Moreover, Hoek et al. 2002 introduced the
disturbance factor (D) to deal with rock mass disturbance. The value of D is determined based
on engineering experiences. The range of D is from 0 (for undisturbed in situ rock masses) to
1 (for disturbed rock mass properties) (Hoek et al. 2002). Hoek 2012 stressed that D should
only be applied to the actual damaged zone; otherwise, it could significantly underestimate the
strength and stability of the overall rock mass.
152
𝑎
𝜎′
𝜎1′ = 𝜎3′ + 𝜎𝑐𝑖 (𝑚𝑏 𝜎 3 + 𝑠) (6.1)
𝑐𝑖
where
𝜎1′ is the major effective principal stress at failure, 𝜎3′ is the minor effective principal stress at
failure, 𝜎𝑐𝑖 is the uniaxial compressive strength, 𝑚𝑏 is the reduced value of the material
constant, and s and a are constants for the rock mass.
GSI−100
𝑚𝑏 = 𝑚𝑖 exp ( ) (6.2)
28 − 14𝐷
GSI−100
𝑠 = exp ( ) (6.3)
9 − 3𝐷
1 1
𝑎= + 6 (𝑒 −GSI/15 − 𝑒 −20/3 ) (6.4)
2
The preceding equations show that 𝑚𝑏 , s and a depend on GSI (Marinos et al. 2007).
6.3.2 Mohr - Coulomb failure criterion
The Mohr - Coulomb failure criterion is a linear equation (Labuz and Zang 2015), as shown in
Figure 6.1. Most geotechnical software programs at present are still written in terms of the
Mohr - Coulomb failure criterion (Hoek 1990, Yang and Yin 2004, Li et al. 2008). Rock related
strength has nonlinear properties. Therefore, rock slope analysis is based on a linear equation,
thereby possibly affecting the accuracy of the stability analysis. Hoek et al. 2002 introduced
equivalent Mohr - Coulomb strength parameters friction angle ϕ′ and cohesive strength c ′ to
provide the required input parameters for rock slope analysis, Equations shown in equations.
(6.5 and 6.6).
′ ) 𝑎−1
6 𝑎 𝑚𝑏 (𝑠+ 𝑚𝑏 𝜎3𝑛
ϕ′ = sin−1 [ ′ ) 𝑎−1 ] (6.5)
2 (1+𝑎)(2+𝑎)+6 𝑎 𝑚𝑏 (𝑠+ 𝑚𝑏 𝜎3𝑛
′ ] (𝑠+ 𝑚 𝜎 ′ ) 𝑎−1
𝜎𝑐𝑖 [ (1+2𝑎) 𝑠+(1−𝑎) 𝑚𝑏 𝜎3𝑛
c′ = 𝑏 3𝑛
(6.6)
6 𝑎 𝑚𝑏 (𝑠+𝑚𝑏 𝜎′3𝑛 )𝑎−1
(1+𝑎)(2+𝑎)√1+
(1+𝑎)(2+𝑎)
where
′ ′
𝜎3𝑛 = 𝜎3𝑚𝑎𝑥 /𝜎𝑐𝑖 (6.7)
−0.91
′ ′ 𝑐𝑚 𝜎′
𝜎3𝑚𝑎𝑥 = 0.72 𝜎𝑐𝑚 [ γH ] (6.8)
153
𝑚
(𝑚𝑏 +4𝑠−𝑎(𝑚𝑏 −8𝑠)) ( 𝑏 +𝑠)𝑎−1
′
𝜎𝑐𝑚 = 𝜎𝑐𝑖 4
(6.9)
2 ( 1+𝑎)( 2+𝑎)
′
Where 𝜎𝑐𝑚 is the compressive strength of the rock mass, H is the slope height,  is the material
unit weight, and the stress range 𝜎𝑡 < 𝜎3′ < 𝜎𝑐𝑖 /4.
6.3.3 Seismic coefficient
The pseudo - static (PS) approach is commonly used in seismic slope designs (Baker et al.,
2006, Bray and Travasarou 2009, Li et al. 2009). Terzaghi 1950 suggested seismic coefficients
k = 0.1 for severe earthquakes and k = 0.25 for violent - destructive earthquakes (Baker et al.
2006). Newmark applied and extended the PS method by considering a critical seismic
coefficient (kc) to evaluate the ground movement induced by an earthquake. Newmark’s
displacement method (Newmark 1965) is based on the assumption that the sliding resistance
of a slope block determines the displacements of slopes due to seismic force (Miles and Ho
1999). This approach has been accepted and extensively used to study earthquake triggered
landslides and rockslides.
The effects of an earthquake can be simplified as either horizontal or vertical seismic

coefficients (Kh and Kv) by using the PS approach. Most current investigations based on the
PS approach only consider Kh. Seed 1979 recommended Kh = 0.1 and Kh = 0.15 for earthquakes
of 6.5 and 8.5 Richter’s magnitude respectively. The acceptable design factor of safety (F) must
be above 1.15 (Seed 1979). Besides, the PS approach using Kh equals to one - half the bedrock
acceleration (Hynes-Griffin and Franklin 1984). Minimum F equals 1. Figure 6.2 summarizes
various Kh and F values based on the California Division of Mines and Geology (DeWayne
and Julie 1997), and shows that the range of Kh is from 0 to 0.375.
6.3.4 Rock slope stability number and stability charts
The results of this study on inhomogeneous rock slopes are presented using a dimensionless
stability number (Ni), as shown in Equation (6.10). The definition of Equation (6.10) is similar
to that proposed by Li et al. 2008, but the results are different due to the different rock mass
properties. The rock slope stability number can be used to present stability charts. Slope
stability charts are convenient tools that directly determine slope stability. A review of
literature revealed that most chart solutions require soil parameters as inputs (Zanbak 1983,
Han and Leshchinsky 2004; Taheri and Tani 2010). However, evaluation of rock slope stability
based on soil parameters does not fit the non-linear nature of the rock mass failure envelope,
154
and thus directly affects the accuracy of the stability analysis (Fu and Liao 2010, Li et al. 2011,
Li et al. 2016).
Ni = 𝜎𝑐𝑖 /γHF (6.10)
where
σci is the intact uniaxial compressive strength.
6.3.5 Analysis of rock slope stability
For the past few years, numerous investigators have evaluated rock slope stability to comply
with geotechnical engineering requirements, such as limit equilibrium method (LEM) (Bishop
1955; Chang 2002), finite element method (Mahtab and Goodman 1970, Shen et al. 2013;
Hammah et al. 2004), and finite element limit analysis methods (LA) (Lyamin and Sloan
2002a, Lyamin and Sloan 2002b). LEM is a popular approach that deals with slope stability
problems (Cassidy et al. 2008; Duncan et al. 2008). This approach is based on physical theory
(Michalowski 1995). In LEM, a rock slope is divided into several slices/blocks. Each
slice/block is assumed to be based on principles of statics. However, LEM can affect the
accuracy of the result because the failure surface must be assumed before the safety factor is
calculated (Qian et al. 2014).
LA is based on two separate methods and has been proven to be robust and efficient in
estimating the strength of rock slopes (Loukidis et al. 2003). Upper bound (UB) is based on the
theorem of kinematics (Lyamin and Sloan 2002a), and lower bound (LB) limit analysis uses
the static theorem (Sloan 1988, Lyamin and Sloan 2002b, Chen et al. 2008). LB solutions are
generally conservative and beneficial in practice (Yu et al. 1998). Therefore, the present study
only adopts LB finite elements (Sloan 1988, Lyamin and Sloan 2002b) to analyse rock slopes.
The critical failure mechanism can be determined automatically without assuming its shape
from the LB solutions.
The stability of the analysed inhomogeneous rock slope model is divided into two regions, as
shown in Figure 6.3. According to of the study of Hoek and Karzulovic 2000, Region 1 in
Figure 6.3 is the disturbed zone caused by surface blasting. The damage zone thickness (T)
ranges considered from 0.5 H to 2.5 H based on recent investigations; that is, T may surround
155
the excavation perimeter from a depth of approximately 2 m (Chen and Liu 2012) to over 100
m behind the excavated slope face (Marinos et al. 2005).
Hoek et al. 2002 indicated that D significantly influences the evaluation of rock slope stability
(Li et al. 2011). Hoek 2012 explained that assigning only a constant D to a rock slope is
inappropriate. Therefore, varying D in a slope is considered in Region 1, where D on the
excavated slope surface is represented as D0, which is assumed to be either 1.0 or 0.7 based on
the proposal of Hoek et al. . Thereafter, this value decreases linearly to 0 with the increase in
depth in T. Meanwhile, Region 2 is the undisturbed rock mass zone, where D = 0 is adopted.
The range of the slope parameters in Table 6.1 is analysed. This study only considers GSI from
50 to 90 because blasting is generally unnecessary in excavating rock masses with GSI ≤ 50
(Tsiambaos and Saroglou 2010). In addition, only Kh is considered for the PS analyses, which
is similar to most practical designs. In this study, the values of Kh are considered from 0.1 to
0.3.
6.5. RESULTS AND DISCUSSIONS
6.5.1 LB solutions for seismic stability number and stability charts
This research investigates the slope stability of seismic inhomogeneous rocks by using the LB
method based on the Hoek - Brown failure criterion. Ni is obtained by considering a range of
rock parameters, which include different magnitudes of Kh (from 0.1 to 0.3). Figure 6.4 shows
Ni for the different GSI, D0, and intact rock yield parameter (mi) for the slope angle of β = 15o
and Kh = 0.1. The obtained Ni increases as GSI and mi decreases. Moreover, Figure 6.4 shows
that D0 influences the estimations of slope stability. Comparison of the different D0 values
indicates that Ni increases as D0 increases. Figures 6.4 to 6.6 show that Ni increases as Kh
increases. Table 6.2 also shows the same trend. For example, in Table 6.2, the Ni of GSI = 50
and D0 = 0.7 for β = 15º increases from 0.375 when Kh = 0.1 to 1.251 when Kh = 0.3. Ni is
inversely proportional to the factor of safety. Therefore, a large Ni means the slope is highly
dangerous.
Slope angle also influences the factor of safety. The results indicate that Ni increases as β
increases. Table 6.2 shows that the difference in Ni between GSI = 50 and GSI = 90 increases
when β increases. When β = 15º, Kh = 0.1 and D0 = 0.7, Ni for GSI = 50 is approximately 4.41
times that for GSI = 90. When β is increased to 75º, Ni for GSI = 50 is approximately 14.20
times that for GSI = 90. In addition, the Ni difference between β = 15o and β = 75o increases
156
while D0 increases. When D0 = 0.7, GSI = 50 and Kh = 0.2, Ni for β = 75o is approximately
28.55 times that for β = 15o. However, Ni for β = 75º is increased to approximately 43.50 times
that for β = 15o when D0 = 1.0. The chart solutions are presented in Figures 6.7 to 6.18, where
the similar trends as those for β = 15o are observed.
Figure 6.19 shows that a large damage zone can reduce the slope factor of safety. Figure 6.19
shows that the difference in Ni between T = 1.0 H and T = 2.5 H increases with a decrease in
β. The maximum difference in Ni observed between T = 1.0 H and T = 2.5 H for β = 60o is
approximately 17%, which is less than the maximum for β = 30o (over 38%). Furthermore,
Figure 6.20 presents the LB plastic points, which can be viewed as the slope failure surfaces.
Therefore, a comparison of failure surfaces for the different T and Kh values is shown in Figure
6.20. The results show that the failure surface for T = 0.5 H is slightly smaller than those for T
= 1.5 H and T = 2.5 H, particularly for the slope crest. The results presented in Figure 6.20
showed only slight differences when the values of Kh are changed. Figure 6.21 presents several
observed LB plastic points for D0 = 0.7 or 1.0, GSI = 60, mi = 20, T = 0.5 H, Kh = 0.2, and
different slope angles. Accordingly, the failure mode is nearly unchanged when β ≥ 30°. Figure
6.21, also indicates that the effects of D0 on the failure surface are insignificant for the analysis
of the seismic inhomogeneous rock slopes.
6.5.2 Comparison of results with LEM using the Mohr - Coulomb Failure Criterion
As mentioned previously, most geotechnical software programs still use soil parameters based
on the Mohr - Coulomb failure criterion. Therefore, analyses of rock slope stability should
convert the rock strength parameters to soil parameters ϕ' and c'. This research evaluates the
slope stability of seismic rocks by using Mohr - Coulomb parameters based on LEM for
comparison.
The rock strength parameters are converted to ϕ' and c' by using RocLab software. The Mohr -
Coulomb parameters are obtained based on the rock strength parameters, which rely on slope
height. Figure 6.22 shows that Region 1 is divided into 10 layers. D0 linearly decreases in each
layer because the slope model is based on inhomogeneous rock slope cases. Soil parameters
are considered inputs based on different D0 magnitudes. Consequently, ϕ' and c' increase
because D0 linearly decreases in each layer. This consideration could exert a significant effect
on the assessments of the rock slope stability. The simulation of the seismic effects in LEM
only considers the horizontal seismic coefficient. The results show that converting the rock
157
strength parameters to ϕ' and c' as a solution to seismic rock slope problems may affect the
accuracy of the result, particularly for steep slopes.
Table 6.3 shows that F in these comparison cases based on the Hoek - Brown failure criterion
Bishop method (Bishop 1955) and Janbu method (Janbu 1973) is equal to 1. Table 6.3 shows
that the difference between Mohr - Coulomb and Hoek - Brown is approximately 2.8% (Bishop
method) and 5.6% (Janbu method) when β = 15o and Kh = 0.1. The difference is approximately
0.9% (Bishop method) and 2.5% (Janbu method) respectively, for the case that assumes that
Kh increased to 0.3 for β = 15o. Table 6.3 shows that F obtained using the Hoek - Brown failure
criterion can be significantly different from that obtained using the Mohr - Coulomb failure
criterion when β increases (β = 75o). The difference is approximately 47.4% (Bishop method)
and 58.0% (Janbu method) for Kh = 0.1. When Kh = 0.3, the difference is approximately 64.1%
(Bishop method) and 67.9% (Janbu method). Therefore, F obtained based on ϕ' and c' may
achieve similar results to those obtained based on the Hoek - Brown failure criterion for gentle
slopes. However, the differences in F between the Mohr - Coulomb and Hoek - Brown criteria
increase as β increases. This phenomenon was also noted by Li et al. 2008, who investigated
only homogeneous rock slopes using the Bishop method.
Figure 6.23 shows the different failure surfaces. It can be seen that the poor prediction for slope
failure surface may be caused by the Mohr - Coulomb parameters. A shallow slip surface is
obtained based on the Mohr - Coulomb parameters when β = 15° (Bishop method and Janbu
method), as shown in Figure 6.23 (a) to (b). A similar slip surface is observed when β = 30°,
as shown in Figure 6.23 (c) to (d). For steep slopes, Figure 6.23 (e) and (f) show that the failure
surfaces based on the Hoek - Brown failure criterion are shallower and closer to the slope
surface than those based on Mohr - Coulomb parameters. Therefore, steep slopes based on the
Mohr - Coulomb parameters may lead to the high estimation of F and induce a non-
conservative design.
6.5.3 Comparison of the difference between LA and LEM based on the Hoek - Brown
failure criterion
This study compares the differences in failure surface between LEM and LB on the basis of
the Hoek - Brown failure criterion. Figure 6.24 (a) to (e) present the different failure
mechanisms obtained from the numerical LB plastic points and LEM for various β. The LEM
results are based on the simplified method of either Bishop or Janbu. The results show that
most failure surfaces of the Bishop method are similar to those of LB. Failure surfaces based
158
on the Janbu method are slightly larger than those from LB and Bishop method where slopes
are gentle. In addition, Figure 6.24 (e) shows that the depth of the slip surface from LEM is
larger than that from LB for β = 75o.
Table 6.4 presents case studies based on Kh = 0.2, T = 0.5 H, D0 = 0.7, mi = 10 and GSI = 50.
The result shows that Ni from LEM approximates the results of LB when β ≤ 60o. The
difference in Ni between LB and LEM can be significant when β = 75 o, Table 6.4 shows that
the difference can reach up to 14.675% (Bishop method) and 62.559% (Janbu method). The
result shows that the predetermined failure surface that was assumed prior to the analysis using
LEM may influence the accuracy of the results when the slope is steep, particularly for the
simplified Janbu method. Therefore, engineers should be cautious when using LEM in rock
slope seismic designs.
6.5.4. Seismic effects based on the analysis of the stability of homogeneous rock slopes
Table 6.5 shows several cases selected from previous seismic chart solutions. The cases are
considered for D0 = 0.7, T = 0.5 H or 2.5 H and differences in GSI under different β (from 30o
to 75o). The values of Kh from 0.1 to 0.3 are investigated. The comprehensive observations in
Table 6 show that Ni for homogeneous rock slopes is larger than those for inhomogeneous case
studies. In addition, Ni based on T = 2.5 H considerably approximates the homogeneous results,
even for the steep slopes. This finding implies that F can be larger if the damaged zone can be
controlled better. Qian et al. 2017 highlighted that, the T proposed by Hoek and Karzulovic
(2000) is reliable, thus, T = 2.5 H can be considered the maximum. Table 6.6 shows that when
T > 2.5 H, the result based on the inhomogeneous rock slopes is similar to the homogeneous
results. Therefore, homogeneous rock slopes chart solutions would be directly applied to rock
slope evaluations.
6.5.5 Case studies 1
The case studies for open pit mines are based on the investigations of Douglas (2002). Table
6.7 shows the slope parameters. Moreover, γ is assumed to be 25 kN/m3 and T = 2.5 H for all
case studies. The same assumption was also made by Li et al. (2011). The present research only
considers the range of GSI from 50 to 90. Therefore, cases with GSI < 50 are not discussed in
this study. The magnitude of T and D was not mentioned in the research of Douglas (2002).
Thus, the case studies are assessed by assuming that D0 = 0.7 and T = 2.5 H. The values of Kh
are considered to be 0.1 or 0.3.
159
Non-seismic inhomogeneous chart solutions (Qian et al. 2017) based on average UB and LB
solutions are considered for comparison. As expected, the F results obtained using the
presented chart solutions are generally smaller than those based on non-seismic rock slopes, as
shown in Table 6.8. Table 6.8 indicates that slope stability is highly influenced by seismic
force. F values based on Kh = 0 are significantly larger than those based on Kh = 0.1 or 0.3. The
difference between Kh = 0 and Kh = 0.1 is approximately 40%. F may decrease by over 50%
when Kh increases from 0.1 to 0.3. Therefore, rock slope seismic designs using incorrect chart
solutions would present significant errors, particularly for large Kh values.
6.5.6 Case studies 2
The case study 2 is Daguangbao landslide. This slope failure is induced by the Wenchuan
earthquake on 12 May 2008 in Sichuan Province, China (Zhang et al. 2015). The rock slope
model was considered as three materials basis on the suggestion of Zhang et al. (2015), as
shown in Figure 6.26. Table 6.9 shows all parameters considered in the Daguangbao landslide
(Zhang et al. 2015). For the case study, different T comparisons have been made. T is
considered as 0.5 H, 1.5 H or 2.5 H. As presented by Zhang et al. (2015), D0 at the slope surface
is 1.0. This study account for that D0 linearly decreased in T. In addition, the rock slope model
estimated as dry condition. Kh was considered from 0 to 0.3.
Table 6.10, which also shows that F can be significantly larger if the damaged zone caused by
blasting can be effectively controlled. The magnitudes of F > 1 for all cases where Kh = 0, and
therefore all cases can be considered stable. It can be found that F < 1 for all cases when Kh =
0.3, showing that all slope cases are under an unstable situation, is the same as that proposed
by Zhang et al. (2015). However it should be stressed that F magnitudes between this study
and Zhang et al. (2015) cannot be compared directly due to the definition of F being different.
The comparison of the slip surface with Zhang et al. (2015) indicated that the result by
considering T can still provide a similar failure shape, as can be seen in Figure 6.27. It should
be noted that the failure surface for T = 0.5 H and 1.5 H are significantly smaller than that for
T = 2.5 H and the reported failure surface by Zhang et al. , particularly for the slope crest.
Therefore, T value is 0.5 H and 1.5 H may not suit well for this case study. The failure surface
for T = 2.5 H in the present study is slightly shallower than the homogeneous one. It should be
noted that the failure surface for T = 2.5 H is found to be more reasonable when comparing
with reported failure surface by Zhang et al. (2015). It implied that the maximum
recommendation (T = 2.5 H) by Hoek and Karzulovic (2000) is reasonable for rock slope
160
evaluations. Therefore, it means that consideration of a reasonable T value is important,

particularly for capturing failure surfaces.
6.6 CONCLUSION
This research is uses the finite element lower bound limit analysis method to estimate the
stability of seismic rock slope for inhomogeneous rock masses. The horizontal seismic
coefficients are based on the pseudo - static method, and the values are considered to range
from 0.1 to 0.3. The slope model simulations consider the varied thickness of the damage zone.
The disturbance factor assigned to the slope surface is either 1.0 or 0.7 and is assumed to be
linearly decreasing with slope depth. The slope stability number is presented as chart solutions
using the Hoek - Brown failure criterion. The seismic inhomogeneous rock slope stability
charts presented in this research can simplify the evaluation of the safety factor. The following
conclusions are obtained from the results.
This research presents a new seismic slope stability charts based on Kh. The results show that
Ni significantly increases as Kh increases, thereby reducing the slope factor of safety. However,
the failure mechanism remains constant. The factor of safety may be overestimated if Kh is not
considered in the seismic rock slope analysis.
Ni increases as D0 and β increase, as well as decreases as GSI and mi increase. Moreover, T

can significantly affect the assessments of rock slope stability. The results approximate those
for a homogeneous slope when T ≥ 2.5 H.
Using the equivalent Mohr - Coulomb strength parameters may significantly influence the rock
slope stability evaluations. This phenomenon is significant for steep slopes. The failure
mechanisms obtained based on LEM do not precisely match the LB results. Therefore, the
assumed slip surface for the assessment of seismic rock slope in LEM may be occasionally
inadequate to assess the stability of rock slopes.
This study compares the difference in Ni between LEM and LB based on the Hoek - Brown
failure criterion. The results show that Ni using LEM approximates the results of LB when β ≤
60o. However, the difference in Ni between LB and LEM tends to be significant when β  75o.
161
6.7 REFERENCES
Chang C-J, Chen WF and Yao JT (1984) Seismic displacements in slopes by limit analysis.
Yang X-L, Li L and Yin J-H (2004) Seismic and static stability analysis for rock slopes by a
kinematical approach. Geotechnique2004. p. 543-9.
Li H, Xiao K and Liu Y (2007) Factor of safety analysis of bedding rock slope under seismic
load. Chin J Rock Mech Eng2007. p. 2385-94.

Hoek E and Brown ET (1980b) Underground excavations in rock1980b.
Hoek E (1988) editorêditors. The Hoek-Brown failure criterion-a 1988 update. Proc 15th
Canadian Rock Mech Symp; 1988: Toronto, Dept. Civil Engineering, University of Toronto.
689-700.
162
Michalowski RL (2010) Limit analysis and stability charts for 3D slope failures. Journal of
based on limit analysis methods. Canadian geotechnical journal2010. p. 1316-34.
Shen J, Karakus M and Xu C (2013) Chart-based slope stability assessment using the
Generalized Hoek–Brown criterion. International Journal of Rock Mechanics and Mining
Sciences2013. p. 210-9.
Jiang X-y, Cui P and Liu C-z (2016) A chart-based seismic stability analysis method for rock
slopes using Hoek-Brown failure criterion. Engineering Geology2016. p. 196-208.
Bieniawski ZT (1976) Exploration for rock engineering: proceedings of the Symposium on

Exploration for Rock Engineering, Johannesburg, 1-5 November 1976: AA Balkema; 1976.
Barton N (2002) Some new Q-value correlations to assist in site characterisation and tunnel
design. International journal of rock mechanics and mining sciences2002. p. 185-216.
Hoek E and Kaiser P (1995) Bawden. WF (1995). Support of underground excavations in hard
rock1995.
Hoek E and Marinos P (2007) A brief history of the development of the Hoek-Brown failure
criterion. Soils and rocks2007. p. 1-8.
Hoek E (1994) Strength of rock and rock masses. ISRM News Journal1994. p. 4-16.
163
Marinos P, Marinos V and Hoek E (2007) Geological Strength Index (GSI). A characterization
tool for assessing engineering properties for rock masses. Underground works under special
conditions Taylor and Francis, Lisbon2007. p. 13-21.
Hoek E (1990) editorêditors. Estimating Mohr-Coulomb friction and cohesion values from
the Hoek-Brown failure criterion. International Journal of Rock Mechanics and Mining
Sciences & Geomechanics Abstracts; 1990: Elsevier.
Yang X-L and Yin J-H (2004) Slope stability analysis with nonlinear failure criterion. Journal
of Engineering Mechanics2004. p. 267-73.
Bray JD and Travasarou T (2009) Pseudostatic coefficient for use in simplified seismic slope
stability evaluation. Journal of geotechnical and geoenvironmental engineering2009. p. 1336-
40.
Terzaghi K (1950) Mechanics of Landslides' in Applications of Geological Engineering

Practice. Bull Geol Soc Amer1950. p. 83.
Newmark N (1965) Effects of Earthquakes on Dams and Embankments, Geotecnique. XV;

1965.
Miles S and Ho C (1999) Rigorous landslide hazard zonation using Newmark's method and
stochastic ground motion simulation. Soil Dynamics and Earthquake Engineering1999. p. 305-
23.
Seed HB (1979) Considerations in the earthquake-resistant design of earth and rockfill dams.
164
Hynes-Griffin ME and Franklin AG (1984) Rationalizing the seismic coefficient method.

DTIC Document; 1984.
DeWayne H and Julie M (1997) Guidelines for Evaluating and Mitigating Seismic Hazards in
California. California: The Public Information Offices of the California Geological Survey;
1997.
Mahtab MA and Goodman R (1970) Three dimensional finite element analysis of jointed rock
slopes. 1970.
rock slopes using the finite element method. Proceedings of the ISRM regional symposium
165
Lyamin A and Sloan S (2002b) Lower bound limit analysis using non‐linear programming.
681-91.

283-93.
391-415.
Chen W-F and Liu X (2012) Limit analysis in soil mechanics: Elsevier; 2012.
166
Tsiambaos G and Saroglou H (2010) Excavatability assessment of rock masses using the
Geological Strength Index (GSI). Bulletin of engineering geology and the environment2010.
p. 13-27.

Douglas K (2002) The shear strength of rock masses. In School of civil and environmental
engineering Sydney, Australia: The University of New South Wales; 20022002.
Zhang Y, Zhang J, Chen G, Zheng L and Li Y (2015) Effects of vertical seismic force on
initiation of the Daguangbao landslide induced by the 2008 Wenchuan earthquake. Soil
dynamics and earthquake engineering2015. p. 91-102.
Sonmez H and Ulusay R (1999) Modifications to the geological strength index (GSI) and their
applicability to stability of slopes. International Journal of Rock Mechanics and Mining
Sciences1999. p. 743-60.
167
6.8 APPENDIX
Table 6.1. Parameters investigated in this study
GSI 50, 60, 70, 80, 90
mi 5, 10, 15, 20, 25, 30, 35
D0 0.7, 1.0
β 15°, 30°, 45°, 60°, 75°
T 0.5H, 1.0H, 1.5H, 2.0H, 2.5H
Kh 0.1, 0.2, 0.3
Table 6.2. Stability numbers (Ni) for T = 0.5H, mi = 5, D0 = 0.7 or 1.0 under a range of Kh.
D0 = 0.7 D0 = 1.0
GSI Kh
15º 30º 45º 60º 75º 15º 30º 45º 60º 75º
0.1 0.375 1.067 2.680 6.136 12.271 0.393 1.463 3.959 8.944 20.173
50 0.2 0.719 1.702 3.975 8.408 20.524 0.745 2.339 5.786 11.983 32.408
0.3 1.251 2.622 5.714 11.201 39.968 1.294 3.599 8.165 15.601 56.054
0.1 0.258 0.662 1.549 3.275 6.168 0.265 0.824 2.059 4.341 9.019
60 0.2 0.492 1.041 2.244 4.391 9.538 0.505 1.297 2.949 5.720 14.444
0.3 0.850 1.575 3.151 5.739 19.181 0.873 1.961 4.086 7.358 23.707
0.1 0.178 0.417 0.908 1.780 3.216 0.182 0.481 1.102 2.164 3.774
70 0.2 0.338 0.647 1.287 2.336 4.883 0.345 0.745 1.548 2.808 5.694
0.3 0.580 0.963 1.767 3.002 8.818 0.592 1.107 2.106 3.567 10.053
0.1 0.123 0.265 0.535 0.975 1.644 0.125 0.287 0.600 1.099 1.862
80 0.2 0.232 0.406 0.742 1.258 2.276 0.235 0.439 0.829 1.408 2.942
0.3 0.395 0.595 1.000 1.594 3.831 0.401 0.641 1.109 1.770 4.892
0.1 0.085 0.169 0.315 0.538 0.864 0.085 0.175 0.332 0.568 0.909
90 0.2 0.159 0.256 0.430 0.684 1.149 0.160 0.264 0.451 0.720 1.378
0.3 0.268 0.370 0.570 0.856 1.935 0.270 0.381 0.596 0.897 2.129
168
Table 6.3. Compared difference in F between Hoek - Brown (HB) and Mohr - Coulomb (MC) failure
criteria
Case Method Kh GSI mi D0 T F (MC) F (HB) Different

Bishop 1.028 1 2.80%
0.1
β = 15o Janbu 1.056 1 5.60%
Bishop 1.009 1 0.90%
0.3
Janbu 1.025 1 2.50%
Bishop 0.972 1 2.80%
0.1
β = 30 o Janbu 1.028 1 2.80%
80 25 0.7 1.5
Bishop 1.009 1 0.90%
0.3
Janbu 0.998 1 0.20%
Bishop 0.1 1.474 1 47.40%
β = 75o Janbu 1.58 1 58.00%
Bishop 0.3 1.641 1 64.10%
Janbu 1.679 1 67.9
Table 6.4. Ni comparisons by using LA and LEM
Ni Ni Ni Difference between Difference between Janbu

β
(Bishop) (Janbu) (LB) Bishop (%) (%)
15 0.351 0.393 0.356 1.404 10.393
30 0.890 0.947 0.884 0.679 7.127
45 2.367 2.539 2.348 0.809 8.135
60 6.659 7.223 6.746 1.290 7.071
75 17.420 7.644 20.416 14.675 62.559
Table 6.5. Seismic slope data selected based on different Kh magnitudes
Case β D GSI mi
1 30 0.7 60 5
2 45 0.7 90 35
3 60 0.7 50 35
4 75 0.7 80 25
169
Table 6.6. The factor of safety analysis by using different chart solutions based on D0 = 0.7
Inhomogeneous Ni Inhomogeneous Ni
Case Kh Homogeneous Ni
(T = 0.5H) (T = 2.5H)
0.1 1.091 0.662 0.981
1 0.2 1.709 1.041 1.539
0.3 2.575 1.575 2.323
0.1 0.072 0.064 0.071
2 0.2 0.114 0.101 0.112
0.3 0.181 0.162 0.177
0.1 2.084 1.796 1.964
3 0.2 3.883 3.117 3.695
0.3 7.592 6.307 7.286
0.1 1.454 1.335 1.403
4 0.2 2.322 2.128 2.188
0.3 5.520 3.409 5.419
Table 6.7. Summary of slope data from cases studies base on rock parameters
Geology Mine case H (m) β GSI mi Stable

Saprolite / Basalt 1b 41 50 40–50 25 No
Saprolite / Basalt 1c 41 55 40–50 25 No
Saprolite / Basalt 1d 46 49 40–50 25 No
Saprolite / Basalt 1e 57 50 40–50 25 No
Saprolite / Basalt 2b 60 48 50–60 25 Yes
Saprolite / Basalt 2c 60 52 50–60 25 Yes
Mudstone / Siltstone 3 38 39 50–60 7 Yes
Breccia 4 200 65 70–80 19 Yes
Siltstone 5a 157 48 60–70 7 Yes
Siltstone 5b 60 53 60–70 7 Yes
Siltstone 6 110 48 40–50 7 No
170
Table 6.8. The factor of safety analysis by using different chart solutions based on D0 = 0.7
Present solutions Inhomogeneous solutions

σci (Inhomogeneous) (by Qian et al. (2017))
Case GSI
γH (Kh = 0.1) (Kh = 0.3) (Kh = 0)
F
40 ̶ ̶ ̶
1a 2
50 1.43 0.46 2.29
40 ̶ ̶ ̶
1b 3
50 2.25 0.77 3.85
40 ̶ ̶ ̶
1c 3
50 1.47 0.47 2.66
40 ̶ ̶ ̶
1d 3
50 2.18 0.71 3.48
40 ̶ ̶ ̶
1e 2
50 1.62 0.56 2.77
50 2.65 0.91 4.54
2a 3
60 4.86 1.82 8.21
50 3.03 0.93 4.63
2b 3
60 4.90 2.02 8.77
50 1.85 0.56 4.17
2c 3
60 4.63 1.59 6.41
50 2.39 0.96 4.10
3 5
60 4.05 1.95 6.42
70 19.35 7.14 37.50
4 30
80 40.00 13.33 61.22
60 4.66 1.97 7.38
5a 8.86
70 7.38 4.22 12.48
60 7.29 3.26 10.92
5b 15.3
70 11.33 6.80 16.83
40 ̶ ̶ ̶
6 9.1
50 2.40 1.17 4.14
171
Table 6.9. Parameters investigated in Daguangbao landslide based on Zhang et al. (2015)
Region σci (MPa) GSI mi  D

1 43.8 40 12 27 1
2 87.2 40 9 27 1
3 87.2 70 7 26 1
Table 6.10. The factor of safety analysis by using different condition
Case study T σci/γH N F Stable

Homogeneous solutions (Kh = 0) 2.39 2.10 1.14 Yes
Homogeneous solutions (Kh = 0.1) 2.39 3.86 0.62 No
0
0.5 H 2.39 0.43 5.55 Yes
Inhomogeneous solutions (Kh = 0) 1.5 H 2.39 0.93 2.57 Yes
2.5 H 2.39 1.19 2.01 Yes
0.5 H 2.39 0.80 2.98 Yes
Inhomogeneous solutions (Kh = 0.1) 1.5 H 2.39 1.72 1.39 Yes
2.5 H 2.39 2.19 1.09 Yes
0.5 H 2.39 1.40 1.71 Yes
Inhomogeneous solutions (Kh = 0.2) 1.5 H 2.39 2.97 0.80 No
2.5 H 2.39 3.78 0.63 No
0.5 H 2.39 2.34 1.02 Yes
Inhomogeneous solutions (Kh = 0.3) 1.5 H 2.39 4.89 0.49 No
2.5 H 2.39 6.18 0.39 No
172
Figure 6.1. Comparison between the Hoek–Brown failure criterion and

equivalent Mohr – Coulomb failure envelope
Figure 6.2. Design K h for pseudo-static analysis
Figure 6.3. D 0 linearly decreased with depth in Region 1 for seismic rock slope
model
173
0.5 0.8
T = 0.5 H T = 1.0 H
0.1
0.1
0.01 0.01
15 10 15 20 25 30 35 1 5 10 15 20 25 30 35
T = 1.5 H T = 2.0 H
0.1 0.1
0.01 0.01
2 5 10 15 20 25 30 35 5 10 15 20 25 30 35
1 T = 2.5 H
GSI 50 (D0 = 1.0)
GSI 60 (D0 = 1.0)
GSI 70 (D0 = 1.0)
GSI 80 (D0 = 1.0)
GSI 90 (D0 = 1.0)
GSI 50 (D0 = 0.7)
0.1
GSI 60 (D0 = 0.7)
GSI 70 (D0 = 0.7)
GSI 80 (D0 = 0.7)
GSI 90 (D0 = 0.7)
0.01 mi
5 10 15 20 25 30 35
Figure 6.4. Limit analysis solutions of stability numbers for rock slopes β = 15 o ,
K h = 0.1
174
1 2
T = 0.5 H T = 1.0 H
1
0.1
0.1
0.02 0.02
5 10 15 20 25 30 35
25 10 15 20 25 30 35 2
T = 1.5 H T = 2.0 H
1 1
0.1 0.1
0.02 0.02
5 10 15 20 25 30 35
3 5 10 15 20 25 30 35
T = 2.5 H
1 GSI 50 (D0 = 1.0)
GSI 60 (D0 = 1.0)
GSI 70 (D0 = 1.0)
GSI 80 (D0 = 1.0)
GSI 90 (D0 = 1.0)
GSI 50 (D0 = 0.7)
GSI 60 (D0 = 0.7)
0.1 GSI 70 (D0 = 0.7)
GSI 80 (D0 = 0.7)
GSI 90 (D0 = 0.7)
0.02 mi
5 10 15 20 25 30 35
K h = 0.2
175
2 3
T = 0.5 H T = 1.0 H
1
1
0.1 0.1
0.03 0.03
35 10 15 20 25 30 35 4 5 10 15 20 25 30 35
T = 1.5 H T = 2.0 H
1
1
0.1 0.1
0.03 0.03
4 5 10 15 20 25 30 35 5 10 15 20 25 30 35
T = 2.5 H
GSI 50 (D0 = 1.0)

1 GSI 60 (D0 = 1.0)
GSI 70 (D0 = 1.0)
GSI 80 (D0 = 1.0)
GSI 90 (D0 = 1.0)
GSI 50 (D0 = 0.7)
GSI 60 (D0 = 0.7)
GSI 70 (D0 = 0.7)
0.1 GSI 80 (D0 = 0.7)
GSI 90 (D0 = 0.7)
0.03 mi
5 10 15 20 25 30 35
K h = 0.3
176
2 3
T = 0.5 H T = 1.0 H
1
1
0.1
0.1
0.02 0.02
3 5 10 15 20 25 30 35 3 5 10 15 20 25 30 35
T = 1.5 H T = 2.0 H
1 1
0.1 0.1
0.03 0.03
5 10 15 20 25 30 35
45 10 15 20 25 30 35
T = 2.5 H
GSI 50 (D0 = 1.0)
1 GSI 60 (D0 = 1.0)
GSI 70 (D0 = 1.0)
GSI 80 (D0 = 1.0)
GSI 90 (D0 = 1.0)
GSI 50 (D0 = 0.7)
GSI 60 (D0 = 0.7)
GSI 70 (D0 = 0.7)
GSI 80 (D0 = 0.7)
0.1
GSI 90 (D0 = 0.7)
0.03 mi
5 10 15 20 25 30 35
Figure 6.7. Limit analysis solutions of stability numbe rs for rock slopes β = 30 o ,
K h = 0.1
177
3 4
T = 0.5 H T = 1.0 H
1
1
0.1 0.1
0.04 0.04
5 5 10 15 20 25 30 35 5 5 10 15 20 25 30 35
T = 1.5 H T = 2.0 H
1 1
0.1 0.1
0.04 0.04
5 5 10 15 20 25 30 35 5 10 15 20 25 30 35
T = 2.5 H
GSI 50 (D0 = 1.0)

GSI 60 (D0 = 1.0)
1
GSI 70 (D0 = 1.0)
GSI 80 (D0 = 1.0)
GSI 90 (D0 = 1.0)
GSI 50 (D0 = 0.7)
GSI 60 (D0 = 0.7)
GSI 70 (D0 = 0.7)
GSI 80 (D0 = 0.7)
0.1 GSI 90 (D0 = 0.7)
0.04 mi
5 10 15 20 25 30 35
Figure 6.8. Limit analysis solutions of stability numbers for rock slopes β =30 o ,
K h = 0.2
178
4 6
T = 0.5 H T = 1.0 H
1
1
0.1 0.1
0.06 0.06
5 10 15 20 25 30 35 5 10 15 20 25 30 35
7 8
T = 1.5 H T = 2.0 H
1 1
0.1 0.1
0.07 0.07
5 10 15 20 25 30 35
8 5 10 15 20 25 30 35
T = 2.5 H
GSI 50 (D0 = 1.0)

GSI 60 (D0 = 1.0)
GSI 70 (D0 = 1.0)
1 GSI 80 (D0 = 1.0)
GSI 90 (D0 = 1.0)
GSI 50 (D0 = 0.7)
GSI 60 (D0 = 0.7)
GSI 70 (D0 = 0.7)
GSI 80 (D0 = 0.7)
GSI 90 (D0 = 0.7)
0.1
0.07 mi
5 10 15 20 25 30 35
K h = 0.3
179
5 6
T = 0.5 H T = 1.0 H
1
1
0.1 0.1
0.05 0.05
75 10 15 20 25 30 35 8 5 10 15 20 25 30 35
T = 1.5 H T = 2.0 H
1 1
0.1 0.1
0.06 0.06
8 5 10 15 20 25 30 35 5 10 15 20 25 30 35
T = 2.5 H
GSI 50 (D0 = 1.0)
GSI 60 (D0 = 1.0)
GSI 70 (D0 = 1.0)
1 GSI 80 (D0 = 1.0)
GSI 90 (D0 = 1.0)
GSI 50 (D0 = 0.7)
GSI 60 (D0 = 0.7)
GSI 70 (D0 = 0.7)
GSI 80 (D0 = 0.7)
GSI 90 (D0 = 0.7)
0.1
0.06 mi
5 10 15 20 25 30 35
Figure 6.10. Limit analysis solutions of stability numbers for rock slopes β =
45 o , K h = 0.1
180
7 9
T = 0.5 H T = 1.0 H
1
1
0.1 0.1
10 5 10 15 20 25 30 35 5 10 15 20 25 30 35
10
T = 1.5 H T = 2.0 H
1 1
0.1 0.1
5 10 15 20 25 30 35 5 10 15 20 25 30 35
10
T = 2.5 H
GSI 50 (D0 = 1.0)
GSI 60 (D0 = 1.0)
GSI 70 (D0 = 1.0)
GSI 80 (D0 = 1.0)
GSI 90 (D0 = 1.0)
1 GSI 50 (D0 = 0.7)
GSI 60 (D0 = 0.7)
GSI 70 (D0 = 0.7)
GSI 80 (D0 = 0.7)
GSI 90 (D0 = 0.7)
0.1 mi
5 10 15 20 25 30 35
45 o , K h = 0.2
181
10 20
T = 0.5 H T = 1.0 H
10
1
1
0.15 0.15
20 5 10 15 20 25 30 35 20 5 10 15 20 25 30 35
T = 1.5 H T = 2.0 H
10 10
1 1
0.15 0.15
20 5 10 15 20 25 30 35 5 10 15 20 25 30 35
10 T = 2.5 H
GSI 50 (D0 = 1.0)
GSI 60 (D0 = 1.0)
GSI 70 (D0 = 1.0)
GSI 80 (D0 = 1.0)
GSI 90 (D0 = 1.0)
GSI 50 (D0 = 0.7)
1 GSI 60 (D0 = 0.7)
GSI 70 (D0 = 0.7)
GSI 80 (D0 = 0.7)
GSI 90 (D0 = 0.7)
0.15 mi
5 10 15 20 25 30 35
45 o , K h = 0.3
182
10 20
T = 0.5 H T = 1.0 H
10
1
1
0.15 0.15
20 5 10 15 20 25 30 35 20 5 10 15 20 25 30 35
T = 1.5 H
10 10 T = 2.0 H
1 1
0.15 0.15
5 10 15 20 25 30 35 5 10 15 20 25 30 35
20
10 T = 2.5 H
GSI 50 (D0 = 1.0)

GSI 60 (D0 = 1.0)
GSI 70 (D0 = 1.0)
GSI 80 (D0 = 1.0)
GSI 90 (D0 = 1.0)
1 GSI 50 (D0 = 0.7)
GSI 60 (D0 = 0.7)
GSI 70 (D0 = 0.7)
GSI 80 (D0 = 0.7)
GSI 90 (D0 = 0.7)
0.15 mi
5 10 15 20 25 30 35
Figure 6.13. Limit analysis solutions of sta bility numbers for rock slopes β =
60 o , K h = 0.1
183
20 20
T = 0.5 H T = 1.0 H
10 10
1 1
0.25 0.25
20 5 10 15 20 25 30 35 20 5 10 15 20 25 30 35
T = 1.5 H T = 2.0 H
10 10
1 1
0.25 0.25
20 5 10 15 20 25 30 35 5 10 15 20 25 30 35
T = 2.5 H
10
GSI 50 (D0 = 1.0)
GSI 60 (D0 = 1.0)
GSI 70 (D0 = 1.0)
GSI 80 (D0 = 1.0)
GSI 90 (D0 = 1.0)
GSI 50 (D0 = 0.7)
GSI 60 (D0 = 0.7)
1 GSI 70 (D0 = 0.7)
GSI 80 (D0 = 0.7)
GSI 90 (D0 = 0.7)
0.25 mi
5 10 15 20 25 30 35
60 o , K h = 0.2
184
20 30
T = 0.5 H T = 1.0 H
10
10
1 1
0.4 0.4
5 10 15 20 25 30 35 5 10 15 20 25 30 35
30 30
T = 1.5 H T = 2.0 H
10 10
1 1
0.4 0.4
30 5 10 15 20 25 30 35 5 10 15 20 25 30 35
T = 2.5 H
10 GSI 50 (D0 = 1.0)

GSI 60 (D0 = 1.0)
GSI 70 (D0 = 1.0)
GSI 80 (D0 = 1.0)
GSI 90 (D0 = 1.0)
GSI 50 (D0 = 0.7)
GSI 60 (D0 = 0.7)
GSI 70 (D0 = 0.7)
GSI 80 (D0 = 0.7)
1
GSI 90 (D0 = 0.7)
0.4 mi
5 10 15 20 25 30 35
60 o , K h = 0.3
185
30 30
T = 0.5 H T = 1.0 H
10 10
1 1
0.4 0.4
30 5 10 15 20 25 30 35
30
5 10 15 20 25 30 35
T = 1.5 H T = 2.0 H
10 10
1 1
0.4 0.4
30 5 10 15 20 25 30 35 5 10 15 20 25 30 35
T = 2.5 H
GSI 50 (D0 = 1.0)

10 GSI 60 (D0 = 1.0)
GSI 70 (D0 = 1.0)
GSI 80 (D0 = 1.0)
GSI 90 (D0 = 1.0)
GSI 50 (D0 = 0.7)
GSI 60 (D0 = 0.7)
GSI 70 (D0 = 0.7)
GSI 80 (D0 = 0.7)
GSI 90 (D0 = 0.7)
1
0.4 mi
5 10 15 20 25 30 35
75 o , K h = 0.1
186
50 50
T = 0.5 H T = 1.0 H
10 10
1 1
0.8 0.8
50 5 10 15 20 25 30 35 50 5 10 15 20 25 30 35
T = 1.5 H T = 2.0 H
10 10
1 1
0.8 0.8
70 5 10 15 20 25 30 35 5 10 15 20 25 30 35
T = 2.5 H
GSI 50 (D0 = 1.0)

GSI 60 (D0 = 1.0)
GSI 70 (D0 = 1.0)
10 GSI 80 (D0 = 1.0)
GSI 90 (D0 = 1.0)
GSI 50 (D0 = 0.7)
GSI 60 (D0 = 0.7)
GSI 70 (D0 = 0.7)
GSI 80 (D0 = 0.7)
GSI 90 (D0 = 0.7)
1
0.5 mi
5 10 15 20 25 30 35
75 o , K h = 0.2
187
100 100
T = 0.5 H T = 1.0 H
10 10
1 1
5 10 15 20 25 30 35 5 10 15 20 25 30 35
100 100
T = 2.0 H
T = 1.5 H
10 10
N =  HF
i ci
1 1
100 5 10 15 20 25 30 35 5 10 15 20 25 30 35
T = 2.5 H
GSI 50 (D0 = 1.0)

GSI 60 (D0 = 1.0)
GSI 70 (D0 = 1.0)
GSI 80 (D0 = 1.0)
10 GSI 90 (D0 = 1.0)
GSI 50 (D0 = 0.7)
GSI 60 (D0 = 0.7)
GSI 70 (D0 = 0.7)
GSI 80 (D0 = 0.7)
GSI 90 (D0 = 0.7)
1
5 10 15 20 25 30 35 mi
75 o , K h = 0.3
188
2 9
T = 2.5H (D0 = 1.0) T = 2.5H (D0 = 1.0)
T = 2.0H (D0 = 1.0) T = 2.0H (D0 = 1.0)
T = 1.5H (D0 = 1.0) T = 1.5H (D0 = 1.0)
T = 1.0H (D0 = 1.0) T = 1.0H (D0 = 1.0)
1 T = 0.5H (D0 = 1.0) T = 0.5H (D0 = 1.0)
T = 2.5H (D0 = 0.7) T = 2.5H (D0 = 0.7)
T = 2.0H (D0 = 0.7) T = 2.0H (D0 = 0.7)
T = 1.5H (D0 = 0.7)
T = 1.5H (D0 = 0.7)

T = 1.0H (D0 = 0.7) T = 1.0H (D0 = 0.7)
T = 0.5H (D0 = 0.7) T = 0.5H (D0 = 0.7)
1
0.1 0.8 mi
0.09 mi
5 10 15 20 25 30 35 5 10 15 20 25 30 35
(a) β = 30 o
(b) β = 60 o
Figure 6.19. N i for GSI = 60, D 0 = 0.7 or 1.0, k h = 0.1 with different T
(a) T = 0.5 H
(b) T = 1.5 H
(c) T = 2.5 H
Figure 6.20. Failure surfaces for β = 45 o , GSI = 70, D 0 = 1.0, m i = 10 and

different T
189
(a) β = 15o
(b) β = 30o
(c) β = 45o
(d) β = 60o
(e) β = 75o
Figure 6.21. The LB failure surface based on plastic points for different slope
angles
Figure 6.22. Assumed 2D seismic rock slopes model based on LEM
190
(a) β = 15o, Kh = 0.1
(b) β = 15o, Kh = 0.3
(c) β = 30o, Kh = 0.1
(d) β = 30o, Kh = 0.3
(e) β = 75o, Kh = 0.1
(f) β = 75o, Kh = 0.3
Figure 6.23. The failure surfaces difference between MC and HB
191
(a) β = 15o
(b) β = 30o
(c) β = 45o
(d) β = 60o
(e) β = 75o
Figure 6.24. Comparison seismic failure surface when T = 0.5, GSI = 50, m i =
10 and D 0 = 0.7
192
(a) β = 30o, Kh = 0.2
(b) β = 45o, Kh = 0.3
(c) β = 60o, Kh = 0.1
(d) β = 75o, Kh = 0.3

inhomogeneous and homogeneous rock slope
Figure 6.26. Problem configuration for Daguangbao landslide analysis
193

inhomogeneous and homogeneous rock slope when K h = 0.2 or 0.3
194
CHAPTER 7 CONCLUDING REMARKS
This thesis applies the finite element UB and LB limit analysis methods to evaluate
inhomogeneous soil and rock slope stability problems based on the 2D model. The slope
stability numbers can provide convenient solutions, which are the core and final results of this
thesis. These stability numbers can also be presented through slope stability charts, which can
help improve current slope designs. The chart solutions can make slope safety assessments
more direct and accurate than the conventional LEM. In addition, ANN technique is used to
assess soil slope stability. Based on ANN, the results of the soil slope are efficient and reliable.
The investigations can be summarized into four research sections, the following topics:
Soil slope stability analysis:
1) Fill slopes stability analysis by consider two - layered purely cohesive soils (Chapter 3)
2) Application of artificial neural network to inhomogeneous soil slope stability assessments

based on finite element limit analysis methods (Chapter 4)
Rock slope stability analysis:
3) Parametric studies of disturbed rock slope stability based on finite element upper bound
and lower bound limit analysis methods (Chapter 5)
4) Seismic disturbed rock slope stability assessments based on finite element lower bound limit
analysis method (Chapter 6)
7.1 GENERAL OVERVIEW
Fill slopes are commonly encountered construction problems in soil design. This thesis
estimates stability numbers on the basis of the two - layer slope model. Chapters 3 and 4 discuss
the soil slope stability analysis based on the Mohr - Coulomb failure criterion. The thesis
presented in these two chapters considers various soil parameters, such as , d/H, and cu1/cu2.
Chapter 3 employs soil stability charts by considering two - layered purely cohesive soils, and
Chapter 4 considers that cohesive strength for foundation materials increases linearly with
depth. Furthermore, Chapters 3 and 4 propose a stability analysis based on ANN. The data in
these two chapters are trained as an evaluation package based on the Mohr - Coulomb failure
criterion. The ANN training results are highly accurate.
195
The slope stability charts presented in Chapters 5 and 6 are based on rock slope stability
analyses. The thesis considers the non - linear nature of rock mass failure envelopes. In
Chapters 5 and 6, the rock slope stability assessments are based on the Hoek - Brown failure
criterion. Rock slope parameters, such as T, D, mi, and GSI, are taken into account in the two
Chapters 5 and 6. Chapter 5 investigates the influence of rock mass disturbance on rock slope
stability assessments. The results of this chapter provide a reasonable tool for preliminary rock
safety factor evaluations. Seismic rock slope stability analysis is a common and difficult
problem in geotechnical engineering. Chapter 6 addressed this issue by considering different
horizontal seismic coefficients. The results are presented as seismic rock stability charts.
7.2 FILL SLOPES COMPOSED OF TWO - LAYERED PURELY COHESIVE SOILS
Soil slope stability analysis based on filled slopes with two - layered purely cohesive soils is
shown in Chapter 3. The ratio of undrained shear strength between the filled materials and the
bottom layer is investigated. The range of the cu1/cu2 ratio is considered to be from 0.2 to 5.
According to the details on Chapter 3, when cu1/cu2 ≤ 0.8, the stability number is almost a
constant. Toe failure is noted for filled slopes that are considerably weaker than the foundation
(cu1/cu2 ≤ 0.8) and where β is larger than 45o. The depth of failure surface increases with
increasing bottom layer depth when cu1/cu2 ≤ 0.8 and β is smaller than 30o. Fill materials may
also highly effect the failure mechanisms when cu1/cu2 ≥ 1.5. In addition, the F of the slope
decreases with an increase in cu1/cu2 ratio when cu1/cu2 ≥ 1.5.
The research also found that safety factor decreases with increasing β, the depth factor, and
ratio of cu1/cu2 when cu1/cu2 ≥ 1.5. The results from the LEM are generally smaller than those
from the numerical UB and LB limit analysis methods.
7.3 APPLICATION OF ARTIFICIAL NEURAL NETWORK TO INHOMOGENEOUS

SOIL SLOPE STABILITY ASSESSMENTS
Similar to the previous research presented in Chapter 3, Chapter 4 in this thesis is discusses the
slope on the basis of the different soil cohesive properties. The ratio of undrained shear strength
between the top and bottom layers is considered to be from 0.2 to 5. A ratio of increment of
the cohesive strength in the bottom layer is considered to be from 0 to 5. The β is from 15° to
75° and d/H is from 1.5 to 5. As results, the stability number increases with the increase in β,
cu1/cu2, and the ratios of the depth factors when cu1/cu2 ≥ 0.8 and d/H ≤ 3. Moreover, the results
196
also reveal that the stability number increases with the reduction in the ratio of increment of
the cohesive strength in the bottom layer.
Stability charts are also used to present the results. The ANN adopted is based on chart
solutions, which shows that ANN once trained the data successfully. The results can be
obtained quickly and accurately as ANN training outputs. The research in this thesis
demonstrated that the ANN training package is a convenient tool for carrying out the
preliminary 2D design of filled soil slope stability.
7.4 THE DISTURBED ROCK SLOPE STABILITY ANALYSIS BASED ON FINITE

ELEMENT LIMIT ANALYSIS METHODS
Chapter 5 discusses the estimated slope stability of inhomogeneous rock masses on the basis
of the Hoek - Brown failure criterion. The magnitude of the D is considered to be either 1.0 or
0.7, which linearly decreases in the T. The T is from half of the slope height to 2.5 times the
slope height. The GSI is from 50 to 90. In addition, the differences in the stability numbers of
the UB and LB solutions are bracketed within ± 10.5%. The results show that the T, GSI, and
D can significantly influence stability assessment. The F for an inhomogeneous rock slope
increases with an increasing GSI and a decreasing D but, decreases with an increasing β.
The F of rock slope decreases with increasing T, when the T is deep enough (T = 2.5 H), and
the F for an inhomogeneous rock slope is close to that for a homogeneous slope. Therefore, the
slope stability analysis can be considered for a homogeneous rock mass slope and the
homogeneous rock slope stability chart can be used.
7.5 THE DISTURBED ROCK SLOPE STABILITY CONSIDER DIFFERENT

SEISMIC COEFFICIENT EFFECTS
Chapter 6, considers the influence of the horizontal seismic coefficients on the rock slope
stability assessment using a pseudo - static method. The T is considered to be from 0.5 to 2.5
times that of the slope height, and D is considered to be 0.7 or 1.0, which linearly decreases in
T. The results provide a reasonable and acceptable range for estimating the rock slope F based
on the Hoek - Brown failure criterion.
In addition, in Chapter 6 only account the horizontal seismic coefficients and the slope
simulations are based on details in Chapter 5. The range of the horizontal seismic coefficients
is from 0.1 to 0.3. The result shows that that when the seismic coefficient increases, seismic
197
force can have a significant influence on rock slope stability, the F of the rock slopes can
decrease heavily. F may decrease by more than 100 % when the seismic force coefficient
increases by 0.1.
Stability number increases with increasing D and β, and decreases with increasing GSI and mi.
Moreover, T can greatly affect the rock slope stability assessments. F for rock slope stability
decreases with increasing T, and the results are close to those from the homogeneous slope
when T ≥ 2.5H.
7.6 SUMMARY
For practicing engineers, performing numerical simulations or laboratory tests may not always
be feasible because of time and cost constraints. Therefore, stability chart solutions can be
convenient tools for determining slope stability.
The Mohr - Coulomb failure criterion is one of the most frequently used in soil slope design.
Therefore, to analyse soil slope stability, this research applies finite element UB and LB limit
analysis to soil slope assessments on the basis of this failure criterion. Furthermore, many slope
designs are based on LEM. The use of LEM, however cannot be guaranteed the best solutions
because the slip surface must be assumed. However, the finite element UB and LB limit
analysis methods can avoid such assumption.
Slope failure surface can be influenced by numerous factors, including the manual estimation
of chart solutions. Therefore, a stability analysis based on ANN is conducted in this research
to provide a convenient method for predicting soil slopes stability.
In rock slope design, sometimes the rock mass strength has to transform to the cohesive
strength and the angle of friction because most rock geotechnical software are still based on
the Mohr - Coulomb failure criterion. Thus this study is focused on assessing rock slope
stability on the basis of the Hoek - Brown failure criterion, which is a non - linear failure
criterion developed for estimating the rock masses for slopes. In this study, the rock slope
stability charts are based on the non - linear Hoek - Brown failure criterion as the charts use
rock parameters directly. The chart solutions based on rock parameters have distinct
improvements in performance on both algorithm stability and accuracy.
Seismic force plays a crucial role in slope stability assessments as well. The ranges of different
horizontal seismic coefficient between 0.1 and 0.3 are considered in rock slope stability studies.
198
7.7 RECOMMENDATIONS FOR FUTURE WORK
Pore pressure effects
Pore pressure is important in the design of slope stability. In soil slope design, soil shear
strength decreases as pore pressure increases. However, many rock slopes also show decreasing
strength when the pore pressures is considered. Therefore, considering the significance
influence of pore pressure effects within fill soil or rock slope stability should be investigated
in the future.
Seismic loading for soil slope stability analysis
The analysis of seismic rock slope stability is a common and difficult problem in geotechnical
engineering because the impact of earthquakes must be considered. As the analysis of soil slope
stability in this research does not consider seismic force. In fact, the soil slope stability due to
the seismic load also can be presented in the form of stability charts, such effects are worth
investigating using dynamic software analysis in future work.
Adoption of artificial neural network in slope stability analysis
Although one study has utilized ANN analysis to investigate the stability of soil slopes, more
complex data analysis is still necessary to improve stability probabilistic analysis under
different soil slopes or rock mass, and considering seismic force effects and pore water
pressure. In addition, ANN analysis based on rock parameters (Chapters 5 and 6) is not
discussed in the thesis. Moreover, in ANN, the various multi - layered slope parameters can
considered as inputs, and the output can be used to predict slope stability. Determining how to
improve ANN performance is a valuable research topic.
Multi - layered slopes (Fill slope considering different materials)
The current research only considered a two - layered soil slope or rock slopes. The results based
on multi - layered slopes are not be employed in chart solutions. Therefore, more complex
analysis is worthy to investigating, for example, the frictional fill materials based on cohesive-
frictional materials.
199
Vertical seismic coefficient
Thus far, the research in this thesis is only considered horizontal seismic coefficient effects.
However, the vertical seismic force can be seen as an additional loading on a slope. In real
designs, regardless of the situation; vertical seismic force always influences slope stability and
a result in slope stability reductions.
200

Slope Stability Assessments Considering Material Inhomogeneity

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Slope Stability Assessments Considering Material Inhomogeneity

Uploaded by

Copyright:

Available Formats

Slope Stability Assessments Considering Material Inhomogeneity

Submitted in fulfilment of the requirements for

Doctor of Philosophy (Engineering)

Signature Redacted by Library

Signature Redacted by Library

Signature Redacted by Library

Signature Redacted by Library

Signature Redacted by Library

Signature Redacted by Library

Signature Redacted by Library

Signature Redacted by Library

I am deeply indebted to my co-supervisors, Professor Kanagaratnam Baskaran and Dr. Sui

Publications based on this thesis are as follows:

1. Paper 1 (Chapter 3):

2. Paper 2 (Chapter 4):

3. Paper 3 (Chapter 5):

4. Paper 4 (Chapter 6):

LIST OF PUBLICATIONS ...................................................................................................ii

ABSTRACT ......................................................................................................................... iii

CHAPTER 1 INTRODUCTION ........................................................................................... 1

1.1 INTRODUCTION ....................................................................................................... 1

1.2 RESEARCH MOTIVATIONS AND OBJECTIVES ................................................. 1

1.2.1 Stability charts for civil engineering design ........................................................ 1

1.2.3 Application of blasting damage zone .................................................................. 2

1.2.4 Seismic rock slope stability ................................................................................. 3

1.2.5 Slope stability analysis by artificial neural network ............................................ 3

1.3 THESIS OUTLINE ..................................................................................................... 3

1.4 REFERENCES ............................................................................................................ 5

CHAPTER 2 LITERATURE REVIEW ................................................................................ 9

2.1 INTRODUCTION ....................................................................................................... 9

2.2 LIMIT EQUILIBRIUM METHOD ............................................................................ 9

2.3 LIMIT ANALYSIS ................................................................................................... 10

2.4 FINITE ELEMENT METHOD................................................................................. 11

2.5 FINITE ELEMENT LIMIT ANALYSIS METHODS ............................................. 11

2.5.1 Finite element lower bound limit analysis methods .......................................... 12

2.5.2 Finite element upper bound limit analysis methods .......................................... 16

2.6 SOIL FAILURE CRITERION .................................................................................. 19

2.7 THE GENERALISED HOEK - BROWN FAILURE CRITERION ........................ 19

2.8 THE GEOLOGICAL STRENGTH INDEX ............................................................. 20

2.9 DISTURBANCE FACTOR ...................................................................................... 21

2.10 APPLICATION OF MOHR - COULOMB FAILURE CRITERION TO ROCK

2.11. SLOPE STABILITY NUMBER AND CHARTS.................................................. 22

2.12 PREVIOUS STUDIES BASED ON THE PSEUDO STATIC METHOD ............. 23

2.13 ARTIFICIAL NEURAL NETWORK ..................................................................... 24

2.14 MESH ARRANGEMENT ...................................................................................... 26

2.15 LIMITATIONS ....................................................................................................... 27

2.16 REFERENCES ........................................................................................................ 29

2.17 APPENDIX ............................................................................................................. 48

CHAPTER 3 SLOPE STABILITY CHARTS FOR TWO-LAYERED PURELY

3.1 ABSTRACT .............................................................................................................. 60

3.2 INTRODUCTION ..................................................................................................... 60

3.3 PREVIOUS STUDIES .............................................................................................. 61

3.4 PROBLEM DEFINITION ........................................................................................ 63

3.5 RESULTS AND DISCUSSIONS ............................................................................. 64

3.6 APPLICATION EXAMPLE ..................................................................................... 65

3.7 CONCLUSION ......................................................................................................... 66

3.8 REFERENCES .......................................................................................................... 68

3.9 APPENDIX ............................................................................................................... 72

CHAPTER 4 APPLICATION OF ARTIFICIAL NEURAL NETWORK TO

4.1 ABSTRACT .............................................................................................................. 88

4.2 INTRODUCTION ..................................................................................................... 88

4.3 PREVIOUS STUDIES .............................................................................................. 89

4.3.1 Slope stability analysis ...................................................................................... 89

4.3.2 Stability number and chart solutions ................................................................. 89