Académique Documents
Professionnel Documents
Culture Documents
discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/49909741
CITATIONS READS
4 185
1 author:
Kok Shien Ng
Universiti Teknologi Mara (Pulau Pinang)
22 PUBLICATIONS 45 CITATIONS
SEE PROFILE
Some of the authors of this publication are also working on these related projects:
All content following this page was uploaded by Kok Shien Ng on 21 November 2014.
NG KOK SHIEN
A project report submitted in partial fulfillment of the requirement for the award of
the degree of Master of Engineering (Civil Engineering – Geotechnics)
MARCH 2005
“I declare that I have read through this project report and to my opinion
this project is adequate in term of scope and quality for the purpose of
awarding the degree of Master of Engineering (Civil-Geotechnics).”
Signature : …………………………….
Name of Supervisor : DR. NURLY GOFAR
Date : March 2005
ii
“I declare that this project report is my own work except for the
quotations and summarizes which I have explained the source.”
Signature : ....................................................
Name : NG KOK SHIEN
Date : March 2005
iii
ADKNOWLEDGEMENTS
ABSTRACT
Slope stability is one of the geotechnical subject most dominated by uncertainty. The
current practice in slope stability analysis is based on the limit equilibrium and not
directly addresses uncertainty. The objective of this study is to integrate probabilistic
approach as a rational means to incorporate uncertainty in the slope stability analysis.
The study was made through a hypothetical problem which includes a sensitivity
analysis, and two real cases. The methodology is based on Monte Carlo simulation
integrated in commercially available computer program SLOPE/W. The output of
the analysis is presented as the probability of failure as a measure of the likelihood of
the slope failure. Results of this study have verified that the probability of failure is a
better measure of slope stability as compared to the factor of safety because it
provides a range of value rather than a single value.
v
ABSTRAK
Kestabilan cerun merupakan salah satu subjek geoteknik yang dibelenggui oleh
ketidakpastian. Penggunaan umum masa kini dalam taksiran kestabilan cerun adalah
berdasarkan had keseimbangan dan ia tidak dapat menangani unsur-unsur
ketidakpastian. Objektif kajian ini adalah untuk mengintegrasikan kaedah
kebarangkalian dalam analysis cerun supaya unsur ketidakpastian dapat dinilai.
Kajian ini dijalankan melalui satu masalah andaian dan dua kajian kes. Metodologi
kajian ini melibatkan penggunaan Simulasi Monte Carlo yang sediada dalam perisian
SLOPE/W. Hasil analisis disampaikan dalam bentuk kebarangkalian gagal yang
boleh dijadikan sebagai ukuran kemungkinan gagal sesuatu cerun. Hasil kajian ini
membuktikan kebarangkalian gagal merupakan kaedah yang lebih bagus berbanding
dengan faktor keselamatan kerana ia membekalkan satu julat nilai dan bukannya
satu nilai sahaja.
vi
TABLE OF CONTENTS
ACKNOWLEDGEMENTS iii
ABSTRAC iv
ABSTRAK v
CONTENTS vi
LIST OF TABLE ix
LIST OF FIGURE x
LIST OF SYMBOL xiii
1 INTRODUCTION 1
1.0 Overview 1
1.1 Problem Statement 2
1.2 Objective 3
1.3 Scope 3
2 LITERATURE REVIEW 4
2.1 Introduction 4
2.2 Conventional Slope Stability Analysis 5
2.2.1 Fellenius (Swedish) Method 8
2.2.2 The Bishop (Routine) Method 8
2.2.3 General Limit Equilibrium (GLE) Method 9
vii
3 METHODOLOGY 35
3.1 Introduction 35
3.2 Simulation Process 35
3.2.1 Determination of Critical Slip Surface 37
3.2.2 Monte Carlo Simulation 39
3.2.2.1 Optimum Number of Trial 39
viii
5 CONCLUSION 75
5.1 Summary 75
5.2 Conclusion 76
5.3 Recommendations 77
REFERENCES 78
ix
LIST OF TABLE
LIST OF FIGURE
4.6 PDF and CDF of the factor of safety for case 1 of no water 48
table and c-φ uncorrelated.
4.7 PDF and CDF of the factor of safety for case 1 with water 49
table and c-φ uncorrelated.
4.8 PDF and CDF of the factor of safety for case 2 of no water 50
table and c-φ uncorrelated.
4.9 PDF and CDF of the factor of safety for case 2 with water 51
table and c-φ uncorrelated.
4.10 PDF and CDF of the factor of safety for case 3 of no water 52
table and c-φ uncorrelated.
4.11 PDF and CDF of the factor of safety for case 3 with water 53
table and c-φ uncorrelated.
4.12 PDF and CDF of the factor of safety for case 1 of no water 55
table and c-φ correlated.
4.13 PDF and CDF of the factor of safety for case 1 with water 56
table and c-φ correlated.
4.14 PDF and CDF of the factor of safety for case 2 of no water 57
table and c-φ correlated.
4.15 PDF and CDF of the factor of safety for case 2 with water 58
table and c-φ correlated.
4.16 PDF and CDF of the factor of safety for case 2 of no water 59
table and c-φ correlated.
4.17 PDF and CDF of the factor of safety for case 2 with water 60
table and c-φ correlated.
4.24 PDF of the factor of safety for specified slip surface slope: 71
a) Peak strength and b) Residual strength
LIST OF SYMBOL
CHAPTER 1
INTRODUCTION
1.0 Overview
Nowadays, most of the slope analysis and design are based on deterministic
approaches. The shear strength, slope geometry, external load and pore water
pressures are assigned as specific unvarying values and as a result, single factor of
safety is then determined. However, the single deterministic safety factor of a slope
is often not enough to analyze the slope stability due to uncertainty in input
parameters.
Most slope analysis and design is based on deterministic approach i.e a set of
single valued design parameter are adopted and a set of single valued factor of safety
(FOS) is determined. Usually the FOS is selected in view of the understanding and
knowledge of the material parameters, the problem geometry, the method of analysis
and the consequences of failure. This results in different FOS obtained by different
designers. This inherent variability characteristic dictates that slope stability problem
is a probabilistic problem rather than deterministic problem. Furthermore, the FOS
approach cannot quantify the probability of failure or level of risk associated with a
particular design situation.
Given the appeal of probabilistic slope stability analysis (PSSA) and the
advanced state-of-practice of reliability techniques, there is a need to facilitate the
adoption of PSSA concept in slope design practice among geotechnical engineer.
3
1.3 Objective
1.4 Scope
CHAPTER 2
LITERATURE REVIEW
2.1 Introduction
Stability analyses are routinely performed in order to assess the safe and
functional design of an excavated slope (e.g. open pit mining, road cuts, etc.), fill
slope (e.g. embankment, earth dam), and/or the equilibrium conditions of a natural
slope. The analysis technique chosen depends on both site conditions and the
potential mode of failure, with careful consideration being given to the varying
strengths, weaknesses and limitations inherent in the methodology being used in the
analysis.
uncertainty. This approach does not necessarily yield sound economic designs. Nor
does it explicitly give a finite indication of the safety of the design. Probabilistic
slope analysis, on the other hand, explicitly accounts for uncertainty. The output of
the probabilistic analysis, in terms of failure probability or reliability index, is a
measure of the reliability of the design. Probabilistic analysis provide greater insight
into design reliability, thus, enhancing the engineering judgment and improving the
decision making process. So, the clarity, simplicity and cost/time effectiveness are
essential elements in order to effectively convey and communicate a probabilistic
methodology to practicing engineers.
Slope stability analysis is usually done using limit equilibrium methods. Most
limit equilibrium method assumes the validity of Coulomb’s failure criterion along
an assumed failure surface. A free body of the slope is considered to act upon by
known and unknown forces. Shear stress induced on the assumed failure surface by
the body and external forces is compared with the available shear strength of the
material.
b
Xi
Xi-1
W
Ei-1 Ei
Ti
α Ni-U
It can be seen that there is many unknown forces involve the equation of
equilibrium for a slice. However if pore water pressure and the inter-slice forces E
and X are assumed to be equal to zero, then the factor of safety can be evaluated
based only on the weight of the slice which can be resolved into two components T
and N.
7
(2.1)
The component T which is parallel to the base of the slice tends to cause
sliding. Taking moment about O, the sum of the moments of the shear forces T on
the failure arc AC must equal to the moment due to the weight of the soil mass
ABCD.
(2.2)
.
(2.3)
(2.4)
8
In this method, the horizontal and vertical forces are assumed to be equal and
opposite and cancel each other out, i.e. Ei = Ei-1 and Xi = Xi-1. It is now only
necessary to resolve the forces acting on the base of the slice, so that:
1¶ :FRV. - ul
KEFRV. - XE VHF. (2.5)
Hence,
(2.6)
The number of slices taken should not be less than five, and larger number
would yield better estimate FOF. However, this method tends to give a value for
FOS which may be as much as 50 percent on the low side (Whitlow, R 2001).
Bishop assumed that the resultant of inter-slice forces (Xi – Xi-1) is equal to
zero, but the sum of the horizontal inter-slice force (Ei- Ei-1) is not zero.
(2.7)
(2.8)
9
(2.9)
(2.10)
(2.11)
3. The summation of moments about a common point for all slices. The
equation can be rearranged and solved for the moment equilibrium factor
of safety, Fm.
(2.12)
(2.13)
where
Even with the above equations of static, the analysis is still indeterminate,
and a further assumption is made regarding the direction of the resultant interslice
forces. The direction is assumed to be described by a interslice force function. The
factors of safety can now be computed based on moment equilibrium (Fm) and force
equilibrium (Ff). These factors of safety may vary depending on the percentage
d in the computation.
X = Eλf(x) (2.14)
where f (x) is a function, . is the percentage (in decimal form) of the function used, E
is the interslice normal force, and X is the interslice shear force.
A very important point here is that the slice base normal is dependant on the
interslice shear forces XR and XL on either side of a slice. The slice base normal is
consequently different for the various methods depending on how each method deals
with the interslice shear forces.
Besides the Fellenius, Bishop and GLE method described above, there are
many other methods of slices have been developed by other researchers. The
differences between various approaches stems from (a) the equilibrium conditions
that are satisfied and (b) the assumptions that makes the problem determinate as
shown in Table 2.1.
12
The sources of variability in soil parameters are illustrated in Figure 2.2. The
parameter uncertainty is divided into data scatter and systematic error. Data scatter is
the dispersion of measurement around the mean. It is emanated from the spatial
variation of soil properties and random testing error. Spatial variability is the true
variation of soil properties from one point to another. It is inherent to the soil and
cannot be reduced. It must be considered in any analysis of uncertainty. Random
testing errors arise from factors related to the measuring process such as operator
14
error or a faulty device. Random errors should be removed from measurement prior
to analysis.
Systematic error results in the mean of the measured property varying from
the mean of the desired property. Statistical error is the uncertainty in the estimated
mean due to limited sample size. While measurement bias occurs when measured
property is consistently overestimated or underestimated at all locations.
8QFHUWDLQW\ LQ
6RLO 3URSHUWLHV
Figure 2.2 Uncertainty in Soil Properties (Source: Christian, Ladd, Beacher, 1994)
The gap between the theory adopted in prediction models and reality is called
model uncertainty. Analytical models, particular in engineering are usually
characterized by simplifying assumptions and approximations. Model uncertainty is
probably the major source of uncertainty in geotechnical engineering (Wu et al.,
1987; Morgenstern, 1995; Whitman, 1996). Comparing model predictions with
observed performance or predictions of more rigorous and comprehensive model is
probably the most direct and reliable approach to quantify model uncertainty.
However the data needed for direct comparison with observed performance are
seldom available in practice.
15
a) Mean value and expected value, . Indicates the center of gravity of a probability
distribution. A typical application would be the analysis of a set of results x1, x2,
……xn laboratory test. Assuming that there is n individual test values xi, the mean x
is given by:
(2.15)
b) Variance, s2. The variance of a random variable x is the expected value of the
squared difference between the random variable and its mean value. Where actual
data are available, subtracting each value from the mean, squaring the result, and
determining the average of these values can calculate the variance of the data:
(2.16)
To obtain an unbiased estimate of the population working from a finite sample, the n
is replaced by n – 1 instead of using n.
d) Coefficient of variation, COV. Is the ratio of the standard deviation to the mean,
i.e. COV = SD/ . COV can provide a convenient dimensionless expression of the
uncertainty inherent in a random variable.
Since three decades ago, many probabilistic methods have been devise for
slopes analysis. However they can be grouped to three categories: analytical
methods, approximate methods or Monte Carlo simulation
Since most of approximate methods are modified version from First Order
Second Moment Reliability Method and Point Estimate Method, only they will be
covered here. Both approaches require knowledge of the mean and variance of all
input variables as well as the performance function g(x1, x2,…xn) that define safety
factor (e.g. Bishop equation), where x1, x2,…xn are input variables (e.g. soil
properties)
One of the great advantages of the FOSM method is that it reveals the relative
contribution of each variable to the overall uncertainty in a clear and easily tabulated
manner. Due to its simplicity, FOSM is commonly used in probabilistic slope
stability analysis (Cornell, 1972; Yucemen & Tang, 1975; Alonso, 1976; Tang et.al.
1976; Anderson, et.al., 1984; Li and White, 1987; Christian et.al., 1994).
Many attempts have been done to analyze slope stability problem using
Monte Carlo simulation (Kim, et al., 1978; Major et.al., 1978; Tobutt, 1982; Priest &
Brown, 1983; Nguyen & Chowdhury, 1984). Monte Carlo Simulation uses randomly
generated points to cover the range of the values that enter into a calculation. The
computation of probabilities by Monte Carlo simulation is a procedure commonly
adopted to solve problems that are not readily solved by analytical. The procedure is
as follow:
low mean and a high standard deviation (e.g., effective cohesion) could imply
negative values associated with the low probability tail of the distribution which is
not admissible for most parameters.
R + Pf= 1
R = 1 - Pf
Pf = 1 - R
Figure 2.7 shows 90% confidence interval (confidence level = 90%). It is the
range that encloses middle 90% of the likelihood, and thus excluded the lowest 5%
and the highest 5% of the possible value of x.
United States Corps of Engineer (U.S. Army, 1992) proposed that reliability
analysis should includes the following steps:
c. The expected value and standard deviation of the performance function are
calculated. In concept, this involves integrating the performance function over
the probability density functions of the random variables. In practice,
approximate values are obtained using the expected value, standard deviation,
and correlation coefficients of the random variables in the Taylor's series method
or the point estimate method.
d. The reliability index is calculated from the expected and standard deviation of the
performance function. The reliability index is a measure of the distance between
the expected value of ln(C/D) or ln(FS) and the limit state.
Lacasse and Nadim (1996) demonstrated the above concept in slope stability
problem as shown in Figure 2.8.
31
The reliability index (β) is defined in term of the mean (E) and the standard
deviation (SD) of the trial factor of safety as (Christian, Ladd and Baecher, 1994):
(2.17)
33
The reliability index describes the stability of slope by the number of standard
deviations separating the mean factor of safety from its defined failure value of 1.0.
It can be also defined as a way of normalizing the factor of safety with respect to its
uncertainty.
When the shape of the probability distribution is known, the reliability index
can be related to the probability of failure. Figure 2.9 illustrates the relationship of
the reliability index to the probability of failure for a normal distributed factor of
safety. The target reliability values shown in Table 2.4 proposed by US Army (1999)
can also be used for comparison.
Figure 2.9 Probability of failure as a function of the reliability index for a normal
distributed factor of safety (Source: Christian, Ladd and Baecher,
1994)
34
CHAPTER 3
METHODOLOGY
3.1 Introduction
Probabilistic slope design comprises of two main stages. First stage is the
statistical characterization of uncertainty while second stage is probabilistic slope
stability analysis for estimating the reliability criterion. Probabilistic slope stability
analysis allows for the consideration of variability in the input parameter, and it
quantifies the probability of failure of a slope. Commercial software SLOPE/W can
perform probabilistic slope stability analysis using the Monte Carlo method. The
following sections detail the concepts and procedures of the proposed approach using
SLOPE/W as analytical tool.
on the mean value of the input parameter using Bishop methods. Probabilistic
analysis is then performed on the critical slip surface, taking into consideration the
variability of the input parameters. The variability of the input parameters is assumed
to be normally distributed with user-specified mean values and standard deviations.
Definition
of the problem
Slope geometry
Soil properties
Distribution parameters
(mean & st dev.)
Deterministic analysis
(mean value of soil
parameters)
Analysis of
output Case study
Conclusions
- Soil properties including material unit weight, cohesion, and frictional angles
- Line loads
- Piezometric lines
- ru, coefficients and B-bar parameter
- The horizontal and vertical seismic coefficients
As described in part 2.4.3, the shear strength parameter c and φ are negatively
correlated while the unit weight is not correlated to other parameter. Thus this study
will consider uncorrelated values and correlated c and φ values with correlation
coefficient of –0.5
The selection of the location and radius of critical slip surface is the most
difficult part of the slope stability analysis. It is not only depend on the geometry of
the slope being analyzed but also on the strength parameters. In deterministic
analysis of slope, mean values of input parameter are always used and this will yield
on a particular failure surface. Many researchers (Tobutt and Richards, 1979;
Chowdhury and Tang, 1987; Hassan and Wolff, 1999) indicated that he deterministic
critical surface is not always the most critical surface in a probabilistic analysis. In
probabilistic analysis, the search for critical slip circle should be done for each set of
randomly generated inputs. However this involves a great computational effort and
significant computer time.
38
In SLOPE/W (2001), the use of a probabilistic analysis will not affect the
deterministic solution. SLOPE/W computes the factor of safety of all slip surfaces
first and determines the critical slip surface as if no probabilistic analysis is chosen.
The probabilistic analysis is then performed on the deterministic critical slip surface.
The factor of safety presented on the SOLVE main window during the probabilistic
analysis is the deterministic minimum factor of safety of all computed slip surface;
however when the analysis is completed, the factor of safety presented on the
SOLVE main window is the mean factor of safety of all Monte Carlo trials.
Fundamental to the Monte Carlo method are the randomly generated input
parameters that are fed to a deterministic model. In SLOPE/W, this is done using a
random number generation function. To ensure that a new set of random numbers is
generated every time SLOPE/W is executed, the random number function is seeded
with the current time of the computer clock.
The random numbers generated from the function are uniformly distributed
with value between 0 and 1.0. In order to use the uniformly generated random
number in the calculations of the normally distributed input parameter, it is necessary
to transform the uniform random number to a normally distributed random number.
The ‘normalization’ process is done using the following transformation equation as
suggested by Box and Muller(1958):
(3.1)
Where:
N = normalized random number
R1 = uniform random number 1
R2 = uniform random number 2
39
In the Monte Carlo simulation analysis, the main output is the probability
density function (PDF) of the factor of safety. Mean and standard deviation can be
inferred from the PDF. In reliability analysis, we are concerned about how well the
slope performs. This can be done through interpretation of the PDF where
probability of failure is the probability of the safety factor less than one and compare
it with reliability criterion (Table 2.3 and Table 2.4).
A simple slope with gradient 1v: 1h and height 6m is modeled with two cases:
i.e. water table at a specified location (Figure 3.2) and no water table. The
importance of having cohesion and friction angle correlated and uncorrelated is
studied as well. Besides, the sensitivity study is also conducted to find out the
importance of input variables toward the analysis result.
19
18
17
16
Description: Hypothetical Slope
15
File Name: Hypothetical slope-with wt.slz
14
Analysis Method: Bishop (with Ordinary & Janbu)
13
121
Elevation (m)
11
10
9
8
7 Description: Soil c-phi
6 Soil Model: Mohr-Coulomb
5
Unit Weight: 18
4
Cohesion: 15
3
Phi: 25
2
1
02
0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32
Distance (m)
Two cases of slope failure are studied to support for the result of study using
hypothetical model. First case study is the slope failure occurred next to proposed
Institute of Fundamental Sciences Ibnu Sina in UTM, Skudai in 2001. The gradient
of the original slope is about 1h: 1.4v to 1h: 1.5v. The second case study is Muar
Trial Embankment (1986) constructed to failure at the height of 5.4m in 100 days.
42
CHAPTER 4
19
18
17
16 Description: Hypothetical Slope
15
File Name: Hypothetical slope-with wt.slz
14
Analysis Method: Bishop (with Ordinary & Janbu)
13
121
Elevation (m)
11
10
9
8
7 Description: Soil c-phi
6 Soil Model: Mohr-Coulomb
5
Unit Weight: 18
4
Cohesion: 15
3
Phi: 25
2
1
02
0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32
Distance (m)
Figure 4.1 Slope geometry and soil properties of the hypothetical problem
43
The minimum factor of safety for the critical slip surface is 1.69 for cases of
no water table and 1.59 for cases with water table. The critical slip surface is as
illustrated in Figure 4.2 for problem without water table and in Figure 4.3 for
problem with water table.
19
18
17
16 Description: Hypothetical Slope
15
File Name: Hypothetical slope-base case.slz
14
Analysis Method: Bishop
13
121
Elevation (m)
11
10
9
8 Description: Soil c-phi
7 Soil Model: Mohr-Coulomb
6 Unit Weight: 18
5
Cohesion: 15
4
Phi: 25
3
2
1
02
0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32
Distance (m)
Figure 4.2 Critical slip surface for slope without water table
44
19
18
17
16 Description: Hypothetical Slope
15
File Name: Hypothetical slope-with wt.slz
14
Analysis Method: Bishop
13
12
Elevation (m)
11
10
9
8
7 Description: Soil c-phi
6 Soil Model: Mohr-Coulomb
5 Unit Weight: 18
4
Cohesion: 15
3
Phi: 25
2
1
0
0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32
Distance (m)
Figure 4.3 Critical slip surface for slope with water table
For all cases with water table, the optimum number of trial required for 4
input variables i.e. friction angle, cohesion unit weight and peizometric line is 20000
as depicted in Figure 4.5. This is obtained through case 1-with water table and c- φ
uncorrelated.
45
3
2.8
2.6
2.4
2.2
2.06
2
1.8
1.6
Figure 4.4 Optimum number of N=15000 in case 1-no water table and c-φ uncorrelated.
46
14
12
10
8.38
8
2
0 5000 10000 15000 20000 25000 30000 35000 40000 45000
Number of Trial (N)
Figure 4.5 Optimum number of N=20000 in case 1-withwater table and c- φ uncorrelated.
47
4.1.1 c- φ Uncorrelated
The result obtained from analysis of case 1-3 where no correlation is assigned
between cohesion and angle of friction is shown in Table 4.2. In general, the mean
factor of safety increase as compared to the minimum factor of safety obtained from
deterministic analysis. For each case, mean factor of safety for problem with water
table is less than the problem without water table. The probability of failure increase
or the reliability index decrease when the standard deviation of soil parameters
increases. Probability density function (PDF) and cumulative distribution function
(CDF) of the factor of safety for these three cases is provided in Figure 4.6 to Figure
4.11
Table 4.2 Result of probabilistic analysis for case 1 to 3 and c-φ uncorrelated.
Case 1 Case 2 Case 3
No w.t With w.t No w.t With w.t No w.t Now.t
Mean FOS 1.70 1.53 1.72 1.56 1.78 1.62
Reliability Index 2.04 1.38 1.05 0.83 0.76 0.64
P (Failure) (%) 2.06 8.38 14.56 20.25 22.30 25.95
Standard Dev. 0.35 0.39 0.69 0.67 1.03 0.96
The above result shows that even though the mean factor safety is high (i.e
FOS>1.5) it does not mean that the slope is safe because of high probability of
failure or low reliability index. Besides, there is no direct relationship between factor
of safety and probability of failure. In other words, a slope with a higher factor of
safety may not be more stable than a slope with a lower factor of safety as we can
compare in case 1 and 3 for problem of no water table, where case 3 with factor of
safety of 1.78 and standard deviation of 1.02 have a much higher probability of
failure than case 1 with factor of safety of 1.70 and a standard deviation of 0.345.
48
80
10
60
Frequency (%)
Probability (%)
40
5
20
0 0
0.425 0.695 0.965 1.235 1.505 1.775 2.045 2.315 2.585 2.855 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5
Figure 4.6 PDF and CDF of the factor of safety for case 1 of no water table and c-φ uncorrelated.
49
20 80
15 60
Frequency (%)
Probability (%)
10 40
5 20
0 0
-1.3 -0.7 -0.1 0.5 1.1 1.7 2.3 2.9 3.5 4.1 0.000 0.438 0.875 1.313 1.750 2.188 2.625 3.063 3.500
Figure 4.7 PDF and CDF of the factor of safety for case 1 with water table and c-φ uncorrelated
50
80
10
60
Frequency (%)
Probability (%)
40
5
20
0 0
-1.0 -0.4 0.2 0.8 1.4 2.0 2.6 3.2 3.8 4.4 0.000 0.833 1.667 2.500 3.333 4.167 5.000
Figure 4.8 PDF and CDF of the factor of safety for case 2 of no water table and c-φ uncorrelated
51
80
15
60
10
Frequency (%)
Probability (%)
40
5
20
0 0
-2.85 -1.95 -1.05 -0.15 0.75 1.65 2.55 3.45 4.35 5.25 0.000 0.714 1.429 2.143 2.857 3.571 4.286 5.000
Figure 4.9 PDF and CDF of the factor of safety for case 2 with water table and c-φ uncorrelated
52
20 80
15 60
Frequency (%)
Probability (%)
10 40
5 20
0 0
-6.9 -5.1 -3.3 -1.5 0.3 2.1 3.9 5.7 7.5 9.3 0.0 1.2 2.4 3.6 4.8 6.0 7.2 8.4 9.6 10.8 12.0
Figure 4.10 PDF and CDF of the factor of safety for case 3 of no water table and c-φ uncorrelated
53
80
10
60
Frequency (%)
Probability (%)
40
5
20
0 0
-2.55 -1.65 -0.75 0.15 1.05 1.95 2.85 3.75 4.65 5.55 0.000 0.667 1.333 2.000 2.667 3.333 4.000 4.667 5.333 6.000
Figure 4.11 PDF and CDF of the factor of safety for case 3 with water table and c-φ uncorrelated
54
4.1.2 c- φ Correlated
Tables 4.3 show the result of analysis where c and φ are negatively correlated
(case 1-3). The effect of correlation between c-φ parameters on the probability
density function of factor of safety is demonstrated in Figure 4.12 to Figure 4.17.
Figure 4.18 also shows the comparisons in probability of failure for all cases. The
mean factor of safety obtained for these cases are higher than the minimum factor of
safety obtained from deterministic analysis.
80
10
60
Probability (%)
Frequency (%)
40
5
20
0 0
0.59 0.83 1.07 1.31 1.55 1.79 2.03 2.27 2.51 2.75 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0 2.2 2.4 2.6 2.8
Figure 4.12 PDF and CDF of the factor of safety for case 1 of no water table and c-φ correlated.
56
20 80
15 60
Probability (%)
Frequency (%)
10 40
5 20
0 0
-1.3 -0.7 -0.1 0.5 1.1 1.7 2.3 2.9 3.5 4.1 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5
Figure 4.13 PDF and CDF of the factor of safety for case 1 with water table and c-φ correlated.
57
80
15
60
10
Probability (%)
Frequency (%)
40
5
20
0 0
-1.1 -0.5 0.1 0.7 1.3 1.9 2.5 3.1 3.7 4.3 0.000 0.409 0.818 1.227 1.636 2.045 2.455 2.864 3.273 3.682 4.091 4.500
Figure 4.14 PDF and CDF of the factor of safety for case 2 of no water table and c-φ correlated.
58
80
15
60
10
Probability (%)
Frequency (%)
40
5
20
0 0
-1.2 -0.6 0.0 0.6 1.2 1.8 2.4 3.0 3.6 4.2 0.0 0.4 0.8 1.2 1.6 2.0 2.4 2.8 3.2 3.6 4.0
Figure 4.15 PDF and CDF of the factor of safety for case 2 with water table and c-φ correlated.
59
20 80
15 60
Probability (%)
Frequency (%)
10 40
5 20
0 0
-3.8 -2.6 -1.4 -0.2 1.0 2.2 3.4 4.6 5.8 7.0 0.000 0.636 1.273 1.909 2.545 3.182 3.818 4.455 5.091 5.727 6.364 7.000
Figure 4.16 PDF and CDF of the factor of safety for case 3 of no water table and c-φ correlated.
60
80
15
60
10
Probability (%)
Frequency (%)
40
5
20
0 0
-2.75 -1.85 -0.95 -0.05 0.85 1.75 2.65 3.55 4.45 5.35 0.000 0.714 1.429 2.143 2.857 3.571 4.286 5.000
Figure 4.17 PDF and CDF of the factor of safety for case 7 with water table and c-φ correlated.
61
With w.t
Case 3
No w.t
With w.t
Case
Case 2
No w.t
With w.t
Case 1
No w.t
0 5 10 15 20 25 30
Probability of Failure (%)
SD=3
SD=2
Unit
Weight
SD=1
SD=5
SD=4
SD=3
SD=2
Friction Angle
SD=1
Parameter
SD=5
SD=4
SD=3
SD=2
Cohesion
SD=1
Figure 4.19 Sensitivity of FOS to variability in soil parameter for c-φ uncorrelated.
64
SD=3
SD=2
Unit
Weight
SD=1
SD=5
SD=4
SD=3
SD=2
Friction Angle
SD=1
Parameter
SD=5
SD=4
SD=3
SD=2
Cohesion
SD=1
0 1 2 3 4 5 6 7 8
Probability of Failure (%)
Figure 4.20 Sensitivity of probability of failure to variability in soil parameter for c-φ uncorrelated.
65
Table 4.5 shows the result of analysis where cohesion and friction angle are
negatively correlated. Figure 4.21 shows the mean factor of safety does not change
as standard deviations for soil parameters are varied. By contrast, the probability of
failure increases steadily when the standard deviation of soil parameter (Figure 4.22).
In can also be concluded that the angle of friction and unit weight has a relatively
small influence on the probability distribution of a particular factor of safety
compare to cohesion variable.
Table 4.5 Sensitivity result for cases with c- φ correlation coefficient of –0.5
Parameter Without water table With water table
Mean FOS Pf Mean FOS Pf
Cohesion (mean = 15)
SD=1 1.6951 0 1.5325 1.03462
SD=2 1.6963 0 1.535 1.55112
Case 4
SD=3
SD=2
Unit
Weight
SD=1
SD=5
SD=4
SD=3
SD=2
Friction Angle
SD=1
Parameter
SD=5
SD=4
SD=3
SD=2
Cohesion
SD=1
Figure 4.21 Sensitivity result for soil parameter to factor of safety for cases with c- φ correlation coefficient of –0.5
67
SD=3
SD=2
Unit
Weight
SD=1
SD=5
SD=4
SD=3
SD=2
Friction Angle
SD=1
Parameter
SD=5
SD=4
SD=3
SD=2
Cohesion
SD=1
0 1 2 3 4 5 6 7 8
Probability of Failure (%)
Figure 4.22 Sensitivity result for soil parameter to probability of failure for cases with c- φ correlation coefficient of –0.5
68
4.1.4 Discussion
The optimum trial number for 3 input variables: cohesion, angle of friction,
and unit weight is 15000 and for 4 input variables: cohesion, angle of friction, unit
weight and peizometric line is 20000. This means that the more variables considered
for analysis the higher the number of iteration required for analysis.
The slope with high mean factor of safety does not mean it is safe because it
may have high probability of failure as well. So, there is no direct correlation
between the Probability of failure and the factor of safety.
Sensitivity study on case 4 to 6 has shown clearly that among the soil
variables considered in the analysis, the variable of cohesion has the most influence
on the resulting probability of failure. However, there is not much difference in mean
factor of safety throughout the cases.
The case studies discussed in item 4.2 and 4.3 showed the importance of
probabilistic analysis in failed slope because it takes into account the variability in
soil parameters.
69
70
50
Elevation (m)
40
30
20 Failure plane
10
0
0 10 20 30 40 50 60 70 80 90 100 110
Distance (m)
Table 4.6 Soil data of peak strength and residual strength parameter obtained
from direct shear test.
Parameter Peak Residual
C (Kpa)= 15.7 5.9
SD 10 6
Phi (Degree) = 24 20
SD = 3 1
Firstly, The slip surface is specified (see Figure 4.23). Then, the search for
optimum number of trial is carried out. The optimum number of trial used for Monte
Carlo simulation for this case is 25,000. The probabilistic analysis results are
presented in Table 4.7, while the PDF for the cases are shown in Figure 4.24. It can
be seen that the slope is safe when peak strength is mobilized along the failure surface
(mean FOS = 1.56 and Pf = 0.59%). On the other hand, if the soil is disturbed and
reached its residual strength, then the mean factor of safety decreases to 1.08, and the
probability of failure has increased to 25%. Thus, one of the cause of slope failure is
the reduction in shear strength.
Table 4.7 Result using data from direct shear box test.
Peak Residual
Mean F of S 1.556 1.081
Reliability Index 2.514 0.677
P (Failure) (%) 0.594 24.892
Standard Dev. 0.221 0.120
71
10 10
Frequency (%)
Frequency (%)
5 5
0 0
0.545 0.755 0.965 1.175 1.385 1.595 1.805 2.015 2.225 2.435 0.52 0.64 0.76 0.88 1.00 1.12 1.24 1.36 1.48 1.60
(a) (b)
Figure 4.24 PDF of the factor of safety for specified slip surface slope: a) Peak
strength and b) Residual strength
4.3.1 Introduction
Figure 4.25 shows the critical slip surface obtained from stability analysis,
which is well matched with the actual geometry of failure. The minimum factor of
safety obtained for this failure surface is 1.05.
34
32
Description: Muar Trial Embankment 4.7m Fill
30
File Name: Muar Trial emabankment-4.7.slz
28
Analysis Method: Bishop
26
24
22
20
Embankment Fill
Elevation (m)
18
16
Weathered Clay Crust
14
12
Very Soft Silty Clay
10
4
Soft Silty Clay
0
0 10 20 30 40 50
Distance (m)
Table 4.8 Statistical Parameters for Input Variable of undrained shear strength
Bulk unit
Bjerrum Corrected Su Mean, E
weight SD
factor (kPa) (kPa)
(kN/m3)
20.6
Embankment fill - - 60 12
SD =3
Weathered clay crust
(Su =10-40 kPa)
15.5 0.75 7.5-30.0 18.75 6.538
Very soft silty clay
14 0.8 8.0-12.8 10.4 1.429
(Su = 10-16kPa)
Soft silty clay
(Su = 16-36 kPa)
16 9.5 15.2-34.2 24.7 5.528
The probability analysis was conducted by Monte Carlo trial simulations with
90,000 iterations. This shows that the low mean factor of safety of 1.05 with standard
deviation of 0.21 was obtained when the height of embankment reached 4.7 m. The
reliability index for this case is 0.234 and the probability of failure is as high as
40.71%. Figure 4.26 shows the PDF and CDF for this case.
Result
Mean F of S 1.0481
Reliability Index 0.234
P (Failure) (%) 40.7090
Standard Dev. 0.205
74
80
30
60
20
Probability (%)
Frequency (%)
40
10
20
0 0
-1.8 -1.2 -0.6 0.0 0.6 1.2 1.8 2.4 3.0 3.6 0.000 0.375 0.750 1.125 1.500 1.875 2.250 2.625 3.000
Figure 4.26 PDF and CDF of the factor of safety, Muar Trial Embankment
75
CHAPTER 5
CONCLUSION
5.1 Summary
The study carried out herein indicates that reliability theory can be used as
alternative and much more consistent approach to the deterministic analysis.
Combining both approach rather than replacing deterministic analysis with
probabilistic analysis provide greater insight to design reliability and enhance
decision-making process.
critical slip surface. Bishop analysis method is adopted in the study. To demonstrate
the capabilities of this methodology, a hypothetical problem and two case studies are
evaluated with the Monte Carlo simulation method.
5.2 Conclusion
The following conclusions can be drawn from the result of the analytical
study on reliability analysis of slope stability i.e:
5.3 Recommendations
1. The main draw back in the use of reliability analysis is the scarcity of suitable
data, quantitative information about how closely these data represent field
conditions and determination of the correlation coefficient between the
random variables representing soil parameter.
2. In SLOPE/W, the critical slip surface is first determined based on the mean
value of the input parameters in deterministic analysis. Probabilistic analysis
is then performed on the critical slip surface. However in reality, the critical
slip surface from deterministic analysis may not be the critical slip surface for
probabilistic analysis.
REFERENCES
Anderson, L.R., Sharp, K.D., Bowles, D.S. and Canfield, R.V.(1984). Application of
Method of Probabilistic Characterization of Soil Properties. Proceeding of ASCE
Symposium on Probabilistic Characterization of Soil Properties- Bridge between
Theory and Practice. Atlanta, USA, May 17. pp 90-105.
Baecher, B.T. and Chritian J.T. (2003). Reliability and Statistics in Geotechnical
Engineering. England, John Wiley & Sons.
Bergado, D.T., Patron, B.C, Youyongwatana,. W., Chai, J.C. and Yudhbir (1994).
Reliability-based Analysis of Embankment on Soft Bangkok Clay. Journal of
Structural Safety. Vol.(13): 247-266.
Box, G.E.P. and Muller, M.E. 1(958). A note on the generation of random normal
deviates. The Annals of Mathematical Statistics, American Statistical Association,
USA., Vol. 29: 610-613.
79
Bromhead, E.N. (1994). The Stability of Slopes. 2nd Edition. Blackie Academic &
professional, Glasgow.
Cornell, C.A. (1972). First Order Uncertainty Analysis of Soil Deformation and
Stability. Proceeding of 1st International Conference on Application of Statistic and
Probability to Soil and Structural Engineering, Hong Kong. September 13-16, 1971,
pp 129-144.
Chowdhury, R.N and Tang, W.H. (1987). Comparison of Risk Model for Slopes.
Reliability and Risk Analysis in Civil Engineering 2: Proceeding of 5th International
Conference on Application of Statistic and Probability in Soil and Structural
Engineering. Vancouver, Canada, May 25-29. Vol. 2: 863-869.
&KULVWLDQ, J.T, Ladd, C., & Baecher, G.B. (1994). Reliability Applied to Slope
Stability Analysis. Journal of Geotechnical and Engineering Division, ASCE. Vol.
120: 2180-2207.
80
Fredlund, D.G., Krahn, J., and Pufahl, D.E. (1981). The Relationship Between Limit
Equilibrium Slope stability Methods. In Proceedings of the 10th Conference of the
International Society for Soil Mechanics and Foundation Engineering (ISSMFE),
Stockholm, Sweden. Vol. 3: 409–416.
Fredlund, D.G. (1984). Analytical Methods for Slope Stability Analysis. 4th
International Symposium on Landslides. Vol.1: 229-250.
Grivas, D.A. (1981). How Reliable are The Present Slope Prediction Method?
Proceeding of The Tenth international Conference of The Soil Mechanic and
Foundation Engineering, Stockholm, Sweden. Vol. 3 : 427-430.
Gue, C.S (2003). Slope stability : A Case Study of Slope Failure at Universiti
Teknologi Malaysia, Skudai. Universit Teknologi Malaysia: Projek Sarjana Muda.
Harr, M.E. (1987). Reliability Based Design in Civil Engineering. McGraw-Hill Inc,
USA.
Hassan, A. and Wolff, T. (1999). Search Algorithm for Minimun Relaibility Index of
Earth Slope. Journal of Geotechnical and Geoenvironmental Engineering, ASCE.
125 (4): 301-308.
Honjo, Y. and Kuroda, K. (1991). A New Look on Fluctuating Geotechnical Data for
Reliability Design. Soil and Foundation. Vol. 31(1): 110-120
81
Kim, H.S , Major, G. and Ross-Brown, D.M. (1978). Application of Monte Carlo
technique to Slope Stability Analyses. Proceeding of 19th U.S Symposium on Rock
Mechanic, Nevada, USA. pp 28-39.
Liew, S. S. (2004). Slope Failures in Tropical Residual Soils. Tropical Residual Soils
Engineering (TRSE). 6 – 7 July 2004.
Li, K.S. (1992a). A Point Estimate Method for Calculating The Reliability Index of
Slope. Proceeding of the 6th Australian-New Zealand Conference on Geomechanics,
Christchurch. pp 448-451.
Li, K.S. and Lumb, P. (1987) Probabilistic Design of Slope. Canadian Geotechnical
Journal. 24: 520-535.
82
Li, K.S. and White, W. (1987). Reliability Index of Slope. Reliability and Risk
Analysis in Civil Engineering: Proceeding of 5th International Conference on
Application of Statistic and Probability to Soil and Structural Engineering,
Vancouver, Canada. May 25-29. Vol. 2: 755-762.
Major, G., Ross-Brown, D.M, and Kim, H.S. (1978). A General Probabilistic
Analysis for Three Dimensional Wedge Failure. Proceeding of 19th U.S Symposium
on Rock Mechanic, Nevada, USA. pp 45-56.
Marek, J.N. and Savely, J.P. (1978). Probabilistic Analysis on The Plane Shear
Failure Mode. Proceeding of The 9th US Symposium on The Rock Mechanic, Nevada,
USA. pp 40-44
McGuffey, V. Grica, D., Lori, J., and Kyfor, Z. (1982). Conventional and
Probabilistic Embankment Design. Journal of Geotechnical Engineering Division,
ASCE. Vol. 108 (10): 1246-1254
Nguyen, V.U. and Chowdhury R.N. (1984). Probabilistic of A Spoil Pile Stability in
Strip Coal Mines-Two Technique Compare. International Journal of Rock
Mechanic, Mining Science and Geomechanic. Vol. 21 (6): 303-312.
Polous, H.H., Lee, Y.C., and Small, J.C. (1990). Predicted and observed behaviour
of a test embankment on Malaysian soft clays. Research Report No. R620, University
of Sydney
Priest. D.S, and Brown, E.T (1983). Probabilistic Stability Analyses of Variable
Rock Slope. Transaction of the Institute of Mining and Metallurgy (Section A:
Mining Industry), 92: A1-A12
Ramalho-Ortigao, J.A ., Mauro, L.g., and Wiley, A.L (1984). Embankment Failure
on Clay near Rio De Janeiro. Journal of Geotechnical Engineering, ASCE. Vol.109
(11): 1460-1479.
Hassan E.R. (2001). Probabilistic Analyses of Land Slide Hazards and Risk:
Bridging Theory and Risk. Edmontum Alberta: Thesis of Doctor of Philosophy
84
SLOPE/W. 2001. A software package for slope stability analysis. Ver. 5.14. GEO-
SLOPE International, Calgary, Alta.
Sowers, G.S. (1991). The Human Factor in Failure. Civil Engineering, ASCE. March.
pp72-73.
Tang, W. , Yucemen, M., & Ang, A. (1976). Probability-Based Short Term Design
of Soil Slopes. Canadian Geotechnical Journal. Vol. 13: 201-215.
Thornton, S., (1994). Probability Calculation for Slope Stability. Computer Methods
and Advances in Geomechanics. 2505-2509.
Tobutt, D. (1982). Monte Carlo Simulation Methods for Slope Stability. Computers
& Geosciences. Vol. 8: 199-208.
85
Tobutt, D.C & Richards E.A (1979). The Reliability of Earth Slopes. International
Journal for Numerical and Analytical Methods in Geomechanic. Vol. 3: 323-354.
Whitlow, R. (2001). Basic Soil Mechanics. Pearson Education Ltd. England. Fourth
edition. Pearson Education Ltd, England.
Wolff, T. F and Harr, M.E. (1987). Slope Design for Dam. Reliability and Risk
Analysis in Civil Engineering 2: Proceeding of 5th International Conference on
Application of Statistic and Probability to Soil and Structural Engineering,
Vancouver, Canada. May 25-29. Vol. 2: 725-732.
Geotechnical Special Publication, No. 58, ASCE. NY, USA. August 31- July 3. Vol.
2: 419-422.
Wu, T.H., Lee, I.M., Potter, J.C. and Kjekstad, O. (1987). Uncertainties in
Evaluation of Strength of Marine Sand. Journal of Geotechnical Engineering, ASCE.
113 (7): 719-738.
Wu, T.H., and Tang, W.H. (1970).Safety Analysis of Slope. Journal of Soil
Mechanics and Foundation Engineering Division, ASCE. 96,SM2, 1970, pp. 609-
630.
Yucemen, M.S. and Tang, W.H. (1975). Long Term Stability of Slopes, A
Reliability Approach. Proceeding of 2nd International Conference on Application of
Statistics and Probability in Soil and Structural Engineering, Aachen, Germany.
September 15-18, pp 215-230.