You are on page 1of 12

This article was downloaded by: [University of Auckland Library]

On: 13 October 2014, At: 15:06


Publisher: Taylor & Francis
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House,
37-41 Mortimer Street, London W1T 3JH, UK

Geomechanics and Geoengineering: An International


Journal
Publication details, including instructions for authors and subscription information:
http://www.tandfonline.com/loi/tgeo20

Finite element reliability analysis of reinforced


retaining walls
a a a
Sajna Sayed , G.R. Dodagoudar & K. Rajagopal
a
Department of Civil Engineering , Indian Institute of Technology Madras , Chennai, 600
036, India
Published online: 31 Aug 2010.

To cite this article: Sajna Sayed , G.R. Dodagoudar & K. Rajagopal (2010) Finite element reliability analysis of reinforced
retaining walls, Geomechanics and Geoengineering: An International Journal, 5:3, 187-197, DOI: 10.1080/17486020903576788

To link to this article: http://dx.doi.org/10.1080/17486020903576788

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) contained
in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no
representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of the
Content. Any opinions and views expressed in this publication are the opinions and views of the authors, and
are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon and
should be independently verified with primary sources of information. Taylor and Francis shall not be liable for
any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoever
or howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use of
the Content.

This article may be used for research, teaching, and private study purposes. Any substantial or systematic
reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any
form to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http://
www.tandfonline.com/page/terms-and-conditions
Geomechanics and Geoengineering: An International Journal
Vol. 5, No. 3, September 2010, 187–197

Finite element reliability analysis of reinforced retaining walls


Sajna Sayed, G.R. Dodagoudar* and K. Rajagopal
Department of Civil Engineering, Indian Institute of Technology Madras, Chennai – 600 036, India

(Received 23 September 2008; final version received 5 November 2009)

This paper presents the reliability analysis of reinforced retaining wall using finite element method. Response surface approach is used to approximate
the performance function and a first-order reliability method (FORM) is used to evaluate the reliability index. In the analysis, displacement response of
Downloaded by [University of Auckland Library] at 15:07 13 October 2014

the reinforced retaining wall is considered as performance function and the corresponding reliability index is evaluated with the aid of a spreadsheet.
Uncertainties associated with the soil and reinforcement properties are explicitly taken into account in the analysis. A parametric sensitivity analysis
has been performed to bring out the effect of important uncertain parameters by evaluating the sensitivity of the reliability index with respect to each of
the uncertain parameters. Results of the response surface method coupled with finite element analysis show the ease and successful implementation of
the reliability analysis procedure for the reinforced retaining walls.

Keywords: reinforced retaining wall; performance function; FORM; reliability index

1. Introduction geomechanical properties. It is the purpose of geotechnical


reliability-based design to produce an engineered system whose
Geotechnical engineers are increasingly being asked to tackle failure would be an event of very low probability. Reliability ana-
many types of non-traditional problems for which there are little lysis thus aims at computing the probability of failure of a system
or no experience to provide guidance. Examples include the with respect to a prescribed failure criterion by accounting for
burgeoning involvement of geotechnical engineers in geoenvir- uncertainties arising in the model description or the environment.
onmental engineering and the use of geosynthetics in many An excellent overview of the development of reliability-based
geotechnical applications such as reinforced retaining walls. design as it applies to geotechnical engineering is provided by
The present design methods of reinforced retaining wall lack Kulhawy and Phoon (2002). Christian et al. (1994), Low and
the logical basis of describing uncertainty. It is to be noted that Tang (1997), Low (2005), Sayed et al. (2008), Bathurst et al.
uncertainties in geotechnical engineering analysis and design are (2008) and others have also described excellent examples of use
unavoidable and numerous practical advantages are realizable if of probabilistic reliability analysis in geotechnical engineering.
uncertainties and associated risk can be quantified (Whitman Reliability analysis of retaining walls has been proved to be
1984, Low and Tang 1997, Duncan 2000, Christian and rational which considers uncertainties due to variability of soil
Baecher 2003). Most geotechnical analyses are generally treated properties and models (Hoeg and Muruka 1974, Smith 1985,
as deterministic. The results of such deterministic analyses are a Duncan 2000). Recent research studies on reinforced retaining
first-order approximation to the mean response; but may easily walls include consideration of uncertainties in reinforcement
miss the true failure mechanism, particularly where failure properties (Chun et al. 2004, Sayed et al. 2008), engineering
surfaces follow the weakest path through the soil. Probabilistic design optimization (Yuan et al. 2003) and application of Monte
methods are tools that can effectively supplement traditional Carlo simulation to evaluate the Hasofer–Lind reliability index
methods of geotechnical designs, providing better insights in (Chalermyanont and Benson 2004). Reliability charts were
design evaluation by taking into account the uncertainties. developed to select geosynthetic materials to form a liner that
The values of the soil properties used in geotechnical engineering will satisfy the minimum requirement of safety against sliding
involve a significant level of uncertainty arising from several (Sia and Dixon 2008). However, the conventional reliability
sources such as inherent random heterogeneity, measurement analysis is unable to represent a structure as realistically as
errors, statistical errors and uncertainty in transforming the possible which is performed by the finite element method
index soil properties obtained from soil tests into desired (FEM). There have been several studies on the lateral wall
deformation of reinforced retaining wall due to foundation yield-
ing (Yoo 2004, Yoo and Song 2006) and due to the influence of
*Corresponding author. Email: goudar@iitm.ac.in

ISSN 1748-6025 print=ISSN 1748-6033 online


# 2010 Taylor & Francis
DOI: 10.1080=17486020903576788
http:==www.informaworld.com
188 S. Sayed et al.

creep of reinforcement (Kazimierowicz-Frankowska 2005) offers a rational way to evaluate the stability of reinforced
which do not consider uncertainties in soil properties. When retaining wall accounting for uncertainties associated with soil
the performance function is not available as an explicit, closed- and reinforcement properties, environmental conditions and
form function of the input variables (such as displacement of the model errors of the analytical technique adopted.
reinforced retaining wall), approximate analytical methods like Probability of failure Pf and reliability index b are related by
first-order reliability method (FORM) or second-order reliability Pf ¼ F(-b) where F is the cumulative normal distribution
method (SORM) are difficult to apply (Thurner and Schweiger function (CDF) corresponding to the probability of failure.
2000). In such cases, one has to resort to alternate methods of Generally the structure is considered to be safe when b is
reliability analysis. Monte Carlo simulation is a computational greater than 3. A number of approaches have been proposed
tool to solve the reliability problem but is computationally to calculate b like first-order reliability method (FORM) etc.
expensive. In such situations, the response surface method pro- Reliability based design of bridge and building structural com-
vides a powerful tool for estimating the failure probability of the ponents has been conceived to produce a probability of failure
mechanical system (Bucher and Bourgund 1990, Gupta and of about 1 in 5000 (i.e. b ¼ 3.54). However, Withiam et al.
Manohar 2004). Some advanced studies using response surface (1998) have noted that the probability of failure for foundations
method to solve reliability problems in structural and geotechni- as approximately 1 in 1000 based on past geotechnical design
cal engineering have been attempted in the literature (Yao and practice. For reinforced retaining walls, the multiple layers of
Downloaded by [University of Auckland Library] at 15:07 13 October 2014

Wen 1996, Tandjiria et al. 2000, Goh and Kulhawy 2005, horizontal reinforcement result in highly strength-redundant
Kaymaz and McMahon 2005, Babu and Srivastava 2007). systems, and the failure or overstress of a single reinforcement
In light of these, it is desirable to evaluate the stability of a layer or strip will not result in failure of the wall. Hence, the
reinforced retaining wall combining the finite element method internal stability of the reinforced wall as a system has much
and reliability analysis. In this paper, the spreadsheet method greater reliability and much lower probability of failure than an
along with response surface approach of reliability analysis is individual reinforcement layer or element. Furthermore, the
applied to the reinforced retaining walls considering serviceabil- reinforced soil is a deformable medium that assists to redistri-
ity limit state function. In the paper, first the fundamental con- bute loads through the composite structure (Allen et al. 2001).
cepts of reliability are explained. Later, the paper addresses the The general probabilistic problem involves the propagation
details of finite element analysis and then illustrates the reliability of input uncertainties and model uncertainties through a com-
analysis procedure based on the FEM with the help of a numer- putational model to arrive at a random output vector. Ideally,
ical example. A parametric sensitivity analysis is presented to the full finite-dimensional distribution function of the random
bring out the effect of uncertainties in the soil parameters and output factor is desired, although partial solutions such as
soil-reinforcement interaction angle on the value of the minimum second-moment characterizations and probabilities of failure
reliability index. Final section of the paper discusses the relia- may be sufficient in some applications. The key feature of the
bility evaluation of the actual case study of the reinforced retain- probabilistic approach is the definition of a so called perfor-
ing wall which has failed and well documented in the literature. mance function G(X) which can be written in general form as

GðXÞ ¼ RðXÞ  SðXÞ ð1Þ


2. Probability of failure and reliability index where R(X) is the ‘resistance’, S(X) is the ‘force’, and X is the
collection of random input parameters. For G(X) < 0 failure is
The factor of safety calculated from deterministic analysis is not implied, while G(X) > 0 means stable behaviour. The boundary
a consistent measure of risk because of the subjective conserva- defined by G(X) ¼ 0 separating the stable and unstable state is
tism involved. The past experience shows that apparently con- called the limit state boundary. Determination or definition of
servative designs are not always safe against failure (Hoeg and the resistance R(X) is generally not straightforward, especially
Muruka 1974, Duncan 2000). Analyses based on probability when R(X) is represented by a threshold value or serviceability
theory allow uncertainty to be quantified and incorporated states or when different failure modes are possible. The prob-
rationally into the geotechnical design process. The end point ability of failure Pf is defined as
of characterizing uncertainties in the design input parameters is
to evaluate their impact on the performance of a design. Pf ¼ Z
P½Gð X Þ <Z0
Reliability analysis focuses on the most important aspect of
performance, namely the probability of failure. This probability ¼ ::: f X ðx1 ; x2 ; : : :; xn Þ dx1 dx2 : : : dxn ð2Þ
of failure clearly depends on both parametric and model uncer- G ðxÞ < 0
tainties. The probability of failure is a more consistent and
complete measure of safety because it is invariant to all mechani- where fX(x) is the joint probability density function of the basic
cally equivalent definitions of safety and it incorporates addi- variable vector, X. This equation is hard to evaluate, because it
tional uncertainty information. The failure of a structure is its is difficult to identify the joint density function, f(X), and to
inability to perform its intended function adequately. In reliabil- perform the integration of f(X) over the entire multidimensional
ity literature, the measure of success is called reliability and it is failure domain, G(X)  0. Indeed, the failure domain is often
denoted by reliability index b. Probabilistic stability analysis defined by means of response quantities (e.g. displacements,
Geomechanics and Geoengineering: An International Journal 189

stresses, strains, etc.), which are computed by means of com- Low 2006) demonstrated that the SOLVER function in EXCEL
puter programs in practical applications, meaning that the fail- can be easily implemented to calculate the first-order reliability
ure domain is implicitly defined as a function of X. Thus index for a range of practical problems. EXCEL is the spread-
numerical methods have to be employed. A good approxima- sheet program developed by Microsoft. The FORM can also be
tion to Equation (2) is to combine the response surface method implemented easily within MATLAB as well. The spreadsheet
with the first-order reliability method. reliability procedure can be applied to stand-alone numerical
packages via the established response surface method.

3. Response surface method


4. Reinforced retaining wall
Response surface methods are commonly used to approximate
the mechanical response of the structure. The response surface One of the greatest advantages of the reinforced retaining wall
method (RSM) was first developed to represent the relationship is their flexibility and capability to absorb deformations due to
between the input and output of a physical experiment by a poor subsoil conditions in the foundations. It was observed at
simple mathematical expression. Quadratic polynomials are many sites that these structures have demonstrated a higher
Downloaded by [University of Auckland Library] at 15:07 13 October 2014

shown to be suitable for localized approximation of geotechni- resistance to seismic loading than the rigid concrete structures.
cal systems (Low and Tang 1997, Xu and Low 2006). The large Various analyses and design methods have been proposed for
part of the computational cost lies in the evaluation of the reasonable design of reinforced retaining wall. Given the avail-
polynomial coefficients. Then the failure probability can be ability of different methods and research in the last decade,
simply evaluated by using the response surface which is an general agreement has reached that a complete design approach
analytical expression, instead of the mechanical model itself should consist of the following (FHWA 2001): (i) Working
(e.g. complex finite element model). A direct coupling between stress analyses, (ii) Limit equilibrium analyses, and (iii)
reliability analysis and finite element computations is possible Deformation evaluations.
through the response surface method. The basic idea of the
RSM is to approximate the exact limit state function G(x), 4.1 Deformation evaluations
which is usually known only through an algorithmic procedure,
by a polynomial function G0 (x) (Bucher and Bourgund 1990): A deformation response analysis allows for an evaluation of the
anticipated performance of the structure with respect to hori-
0
X
r X
r
zontal and vertical displacements. Horizontal deformation ana-
G ðXÞ ¼ l þ mi X i þ ni Xi2 ð3Þ
i¼1 i¼1
lyses are the most difficult and least certain of the performed
analysis (FHWA 2001). In many cases they are done approxi-
where Xi, i ¼ 1, . . ., r ¼ random variables and the parameters l, mately or it is simply assumed that the usual factors of safety
mi and ni ¼ coefficients that need to be determined. A limited against external or internal stability failure will ensure that the
number of evaluations of the limit state function (i.e. number of deformations will be within tolerable limits. A rough estimate
finite element runs) are required to build the response surface. of probable lateral displacements of simple structures that may
Early applications of this method to the analysis of slope stabi- occur during construction can be made based on the reinforce-
lity can be found in Wong (1985). The author selected 2N as the ment length to wall-height ratio and reinforcement extensibility
number of fitting points to build the surface where N is the as shown in Figure 1. The figure indicates that increasing the
number of random variables. In order to reduce the number of reinforcement length-to-wall height ratio from its theoretical
fitting points in case when N is large, Bucher and Bourgund value of 0.5 to 0.7H, decreases the deformation by 50% where
(1990) proposed a simplified quadratic expression which is H is the height of reinforced retaining wall. In general, a typical
defined by only (2N þ 1) coefficients. The reliability analysis practice followed in seismically active areas is to design walls
can be performed by means of the analytical expression given in for reduced seismic pressure corresponding to 50 to 100 mm of
Equation (3) instead of the true limit state function using displacement (FHWA 2001).
FORM. The FORM provides a practical scheme of computing
small probabilities of failure at high dimensional space spanned
by the random variables in the problem (Hasofer and Lind 5. Numerical example
1974, Haldar and Mahadevan 2000, Melchers 1999). The
FORM is capable of handling any nonlinear performance func- A reinforced retaining wall of height 6 m has been considered
tion and any combination of correlated non-normal random for the reliability analysis using the RSM. The properties of the
variables. With reference to time-invariant reliability calcula- reinforced soil and the backfill soil are assumed as: c ¼ 0, f ¼
tion, Rackwitz (2001) observed that: ‘For 90% of all applica- 30 and  ¼ 16 kN/m3. The surface of the backfill was assumed
tions this simple first-order theory fulfills all practical needs. Its to be horizontal and carries a permanent uniform surcharge of
numerical accuracy is usually more than sufficient’. At present, 20 kPa. The soil at the site is assumed to be compacted in layers
no solution methods exist that is of comparable simplicity to of 250 mm thickness. The long term allowable design strength
FORM. Low and co-workers (e.g. Low and Tang 2004, Xu and of the reinforcement layers was assumed to be 50 kN/m and
190 S. Sayed et al.

3 interaction angle () is 23 (Figure 2). Seven layers of reinfor-


δmax = δR.H/250 (Inextensible) cement are obtained from the conventional stability analysis
δmax = δR.H/75 (Extensible)
where δmax = Maximum displacement (FHWA 2001) spaced in multiples of compact layer thickness.
in units of H
Relative displacement, δR

H = Height of wall, m
2 δR = Empirically derived relative
displacement 5.1 Finite element analysis of reinforced retaining wall

The FEM was used to estimate the actual displacements the


retaining wall may experience after surcharge loading and this
1 can be compared with the allowable displacement for the wall,
resulting in a safety margin on displacements. In this study,
finite element analysis has been carried out using the finite
element program ‘GEOFEM’ developed at IIT Madras,
Chennai (Rajagopal 1998). The finite element mesh used for
0 the numerical simulation is shown in Figure 3.
0.0 0.5 1.0 1.5
The finite element mesh used for the reinforced retaining
Downloaded by [University of Auckland Library] at 15:07 13 October 2014

L/H
wall consists of 2200 nodal points, 178 interface elements, 576
Figure 1. Empirical curve for estimating probable anticipated lateral eight-noded quadrilateral elements, 11 nodal link elements and
displacement during construction for MSE walls (after FHWA 2001).
77 bar elements. Both the soil and panel elements were mod-
elled using 8 noded quadrilateral elements. The facing panel
material was assumed to be linear elastic. The soil forming the
backfill as well as the infill was simulated by elasto-plastic
qo = 20 kPa Mohr–Coulomb model.
The Mohr–Coulomb yield function F, and the plastic potential
Q, are defined in terms of internal friction angle f, the angle of
dilatancy c, and the cohesion c as given in Equations (4) and (5):
. . . . .  
Reinforced fill Backfill J1 pffiffiffiffi 1
F ¼ sin fþ J2 cos   pffiffiffi sin  sin f  c cos f ð4Þ
γr = 16 kN/m3 γb = 16 kN/m3 3 3
 
φr = 30° φb = 30° J1 pffiffiffiffi 1
Q ¼ sin cþ J2 cos   p ffiffi
ffi sin  sin c ð5Þ
6m 3 3
. . . . .
Geogrid where
. . . . . J1 ¼
1
s x þ sy þ s z

ð6aÞ
3
. . . . .    12
pffiffiffiffiffi
. . L =.5 m . . J2 ¼
1 2
2
sx þ s2y þ s2z þ s2xy ð6bÞ
. . . .
J3 ¼ sx sy sz  sz s2xy ð6cÞ
Figure 2. Reinforced retaining wall: numerical example.

Props present during Reinforcement


construction phase (7 layers)

Full height facing Smooth, rigid surface


panel 180 mm thick
and 6 m high
6m

Rough, rigid surface


5m
15 m
Figure 3. Details of finite element mesh for reinforced retaining wall.
Geomechanics and Geoengineering: An International Journal 191

sx ¼ sx  J1 ; sy ¼ sy  J1 ; sz ¼ sz  J1 ð6dÞ 5.2 Reliability analysis procedure via RSM


2 3
pffiffiffi Programs can be written in spreadsheet to handle explicit limit
1 6 3 3  J3  7
y ¼ sin1 4 5 ð6eÞ state function. However, there are situations where serviceabil-
3 2 J 32 ity limit states can only be evaluated using finite element or
2
finite difference programs. In these circumstances, reliability
Non-associated type flow rule is resulted if the two angles f analysis can still be performed, provided one first obtains a
and c are not equal to each other. The elastic constitutive response surface function via the established response surface
matrix (D) is first formulated based on the current tangent methodology which closely approximates the outcome of the
modulus value and Poisson’s ratio. During the plastic flow, finite element or finite difference programs. Once the closed-
correction is applied to this matrix to obtain the elasto-plastic form response functions have been obtained, performing relia-
constitutive matrix (Dep). bility analysis for a target reliability index is straightforward.
The reinforcement was modelled using 3-noded bar ele- Xu and Low (2006) illustrate the finite element reliability
ments, which are compatible with the 8-noded quadrilateral analysis of embankments on soft ground via the response sur-
elements used for modelling the soil. The interfaces between face methodology.
the soil and the panel wall were modelled using 6-noded joint The second-order polynomial function given by Equation (3),
Downloaded by [University of Auckland Library] at 15:07 13 October 2014

elements. The analysis was carried out in two stages. In the first without interaction terms, suggested by Bucher and Bourgund
stage, the self weight of the soil was applied. The wall panels (1990), is adopted in this study. By knowing the values at sam-
were restrained in the lateral direction using external props pling points selected for each variable, the value of the function at
simulated using 2-noded bar elements. In the second stage, any point in the design space can be estimated by fitting a
the uniform surcharge pressure was applied in increments of polynomial of order two, and only 2N þ 1 sampling points are
0.2 kPa per load step. The maximum number of iterations per required to form the function for problems involving N variables.
load step was limited to 25. The properties of the elements used The procedure followed for the finite element reliability analysis
in the finite element analysis are given in Tables 1–3. is summarized in the following (Xu and Low 2006):

(1) The random variables considered in this study are friction


angle (f), unit weight () and interaction angle between
Table 1. Properties of continuum elements the reinforcement and the soil (). These variables are the
Properties Panel elements Soil elements
most important parameters that affect the stability of the
reinforced retaining wall. The effects of these random
Elastic modulus (kN/m2) 2.1  107 25,000 variables on the stability of the reinforced retaining wall
Poisson’s ratio 0.1 0.3
Cohesion (kN/m2) – 0.0
have been reported by Chalermyanont and Benson (2004)
Unit weight (kN/m3) 24.0 16.0 and Chun et al. (2004). The basic statistics of these ran-
Friction angle ( ) – 30.0 dom variables are listed in Table 4.
(2) The values of the random variables at the chosen sampling
points are determined. For each random variable, its mean
value, mi, and two other values, mi  msi are sampled,
Table 2. Properties of interface elements where m is an arbitrary number. In the present analysis,
the value of m was selected as 1. Consequently, for stabi-
Properties Soil/reinforcement interface Panel/soil interface lity problems involving N random variables, the number
Initial tangential 1.0  107 1.0  107 of sampling points required to form the response surface
stiffness (kN/m2) for the performance function is 2N þ 1.
Initial normal stiffness 1.0  107 1.0  107
(kN/m2)
Residual tangential 1.0  107 2.5  102
stiffness (kN/m2)
Residual normal stiffness 1.0  107 1.0  107 Table 4. Mean m, standard deviation s, coefficient of variation cov and
(kN/m2) distribution of random variables
Cohesive strength (kN/m2) 0.0 0.0
Interface friction angle ( ) 30.0 23.0 Random
variable m COV s Distribution Sources of data
o
f( ) 30 20% 6 Normal Low and Tang (1997), Phoon and
Kulhawy (1999), Hoeg and
Table 3. Properties of bar elements Muruka (1974)
 (kN/m3) 16 10% 1.6 Normal Low and Tang (1997), Phoon and
Properties Reinforcement elements Prop elements Kulhawy (1999), Hoeg and
Muruka (1974)
Elastic modulus (kN/m2) 500.0 5000.0 T (kN/m) 50 10% 5 Normal Low and Tang (1997)
Tensile strength (kN/m) 50.0 –  (o) 23 10% 2.3 Normal Chun et al.(2004)
192 S. Sayed et al.

(3) The chosen values of the random variables are then used built-in SOLVER optimization tool to minimize b with the
in the finite-element program GEOFEM to give the values constraint that G(X) ¼ 0, and by automatically changing the
of the deformation. The calculated deformation and the values of the random variables, Xi. Based on the value of
corresponding input variables are used to construct a reliability index, the failure probability can be approximately
response surface G(X) which represents the performance determined. In the present study, the spreadsheet reliability
function for the reinforced retaining wall considered in the evaluation approach is illustrated for the reinforced retaining
analysis. Note that the obtained response surface will pass wall using numerical example and a case study.
through all the sampled points, since the number of unde-
termined coefficients in the response surface function is
the same as the number of equations. After the approx- 5.3 Results and discussion
imate performance function G(x) is obtained, the Based on the procedure described in the previous section, the
Hasofer–Lind reliability index, b or failure probability is sampling points and the resulting deformation of the reinforced
then determined by using the FORM. The matrix formula- retaining wall are summarized in Table 5. The performance
tion of the Hasofer–Lind index, bHL is function is formed by finding the difference between the dis-
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi placement evaluated by finite element analysis and the max-
Downloaded by [University of Auckland Library] at 15:07 13 October 2014

bHL ¼ minx 2 F ðX  M ÞT C1 ðX  MÞ ð7Þ imum value of the limiting displacement (lim) chosen based on
the serviceability requirements.
where X, vector of random variables; M, vector of the mean The reliability index is calculated using the ellipsoid method
values of random variables; C, covariance matrix of random based on the FORM. Figure 5 shows the spreadsheet setup of
variables; and F, failure region. According to Equation (7), the the ellipsoid method for calculating the reliability index for the
Hasofer–Lind index, bHL can be regarded as the minimum performance function evaluated from FEM results. It can be
distance in units of directional standard deviation from the seen from Figure 5 that when lim is 66 mm, the corresponding
mean value point of the random variables to the boundary of reliability index obtained is 0.687. Reliability indices corre-
the limit state surface. Low and Tang (1997) used the following sponding to different values of lim can be obtained by repeat-
equivalent form of Equation (7) in order to show the correlation edly evaluating b at closely spaced increments of lim. The plot
structure more explicitly: of b versus lim for the numerical example is shown in Figure 6.
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
   ffi
It can also be noted that the reliability index decreases as the
xi  Ei T 1 xi  Ei coefficient of variation of friction angle of the soil increases.
b ¼ minx 2 F ½ R ð8Þ The relationship between lim and the probability of failure,
si si
P( > lim) is shown in Figure 7. This plot allows the designer to
where R, correlation matrix; Ei, vector of mean values of ran- evaluate the probability of exceeding lim. For example, if the
dom variable xi; and si, vector of standard deviation values of maximum allowable displacement (lim) is 90 mm, Figure 7
random variable xi. The reliability index and the corresponding depicts the probability of wall deflection exceeding lim
design point, xD, are calculated based on the limit state function, approximately is 40%. If the lim is 110 mm, then the probability
G(X) ¼ 0 (Figure 4). This can be achieved by using EXCEL’s of wall deflection exceeding lim is approximately 8%. However,
it should be emphasized that the plots in Figures 6 and 7 are site
and problem specific and applicable only for the example con-
sidered in the study.
As discussed in the previous section, the sampling points
were chosen at mi  msi in the response surface method. In
order to investigate the effect of the parameter m on the relia-
bility index, the probabilistic stability analyses based on the
FEM for the reinforced retaining wall were repeated using

Table 5. Input sampling points for approximation of performance function


based on the FEM

Sample Friction Unit weight,  Interaction Displacement,


point angle, f ( ) (kN/m3) angle,  ( ) (mm)

1 30 16 23 53.6
2 36 16 23 41.3
3 24 16 23 72.2
4 30 17.6 23 57.3
5 30 14.4 23 49.0
6 30 16 25.3 53.4
7 30 16 20.7 53.8
Figure 4. Design point, mean-value point and reliability index in plane.
Geomechanics and Geoengineering: An International Journal 193

((x-m)/σ) 1

x* Mean(m) Std. Dev(σ) vn

Reliability index (β)


φ 25.95 30 6 –0.67 0.8
γ 16.22 16 1.6 0.14
δ 22.98 23 2.3 –0.01

Correlation matrix, “crmat”


0.6
1 0 0
0 1 0 β 0.687
0 0 1
G(x) 0
0.4
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6
Value of the parameter (m)
Array formula: Ctrl + Shift, then Enter
=SQRT(MMULT(TRANSPOSE(vn),MMULT(MINVERSE(crm Figure 8. Effect of parameter m on reliability index for reinforced retaining wall.
at),vn)))
Downloaded by [University of Auckland Library] at 15:07 13 October 2014

different values of m, which ranges from 0.2 to 1.5. Figure 8


Figure 5. Spreadsheet setup of ellipsoid method for reliability analysis of
reinforced retaining wall with performance function G(x) evaluated from FEM shows the variation of the reliability index with the parameter
results. m. It can be seen that the reliability index is insensitive to the
changes in parameter m when all the random variables consid-
ered together in the reliability analysis. For highly nonlinear
8
limit state functions, the method is extremely sensitive to the
7 algorithm parameters, especially the choice of parameter m. In
COV (φ) = 0.1 COV (φ) = 0.2
the present study, the fitted limit state function is not highly
6 nonlinear. Therefore, the probability of failure evaluated for the
Reliability index (β)

reinforced retaining wall is not underestimated.


5

4 5.4 Parametric sensitivity analysis


3 Probabilistic studies may be carried out by treating some of the
2
key soil properties as random variables. The soil parameters
that do not cause any significant variation in the analyses may
1 be treated deterministically to reduce the complexity of the
problem. Based on the parametric studies, engineers could
0 further refine their design and construction requirements to
0 50 100 150 200
δlim (mm)
minimize the project cost. Spearman rank correlation coeffi-
cient between the reliability index and the input variables can
Figure 6. Variation of reliability index with limiting lateral wall deflection. be calculated by performing parametric sensitivity analysis
(Figure 9). Spearman’s rank correlation coefficient assesses

0.05
0.045
0.04
Probability of failure (Pf)

0.035 –0.19

0.03
δ
Input variable

0.025 γ
–0.14
0.02 φ

0.015
0.01 0.71

0.005
0
0 50 100 150 200
δlim (mm) –0.3 –0.2 –0.1 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
Spearman rank correlation coefficient
Figure 7. Probability of exceeding the limiting lateral wall deflection for
reinforced retaining wall. Figure 9. Results of sensitivity analysis for reinforced retaining wall.
194 S. Sayed et al.

how well an arbitrary monotonic function could describe the segmental retaining wall. In July 2003, a 7.4-m high segmental
relationship between two variables, without making any retaining wall (SRW) constructed in a land development site in
assumptions about the frequency distribution of the variables. South Korea for landscaping and earthwork collapsed and
The calculation of Spearman’s rank correlation coefficient details of this wall have been reported by Yoo and Jung
(r) involves two sets of data Xi and Yi converted to rankings xi (2006). The wall was constructed for a newly developed factory
and yi before calculating the coefficient (Equation 9). The raw complex in Chung-Nam Province, approximately 200 km south
scores are converted to ranks, and the differences di between the of Seoul, South Korea. The landscaping and earthwork for the
ranks of each observation on the two variables are calculated. factory complex required the construction of several geosyn-
P thetic reinforced segmental retaining walls. At one location, a
6 di2 150-m long retaining wall was required, ranging in height from
r¼1 ð9Þ
nðn2  1Þ 1 to 7.4 m, to retain an embankment for use as an approach road
to the factory complex. The wall was situated on a slightly
where di ¼ xi - yi is the difference between the ranks of
sloping ground, immediately next to a 2-m wide waterway
corresponding values Xi and Yi; and n is the number of values
located approximately 4 m away from the wall face.
in each data set (same for both sets). If the value of r is closer is
The evidences from the site observations led to a conclusion
to þ1 or -1, the likely correlation will be the stronger one. A
that the wall failure was somewhat closely related to the global
Downloaded by [University of Auckland Library] at 15:07 13 October 2014

perfect positive correlation is þ1 and a perfect negative corre-


slope and external instabilities. A further field investigation
lation is -1. The Spearman correlation coefficients could be
revealed that the failure pattern followed the circular type and
used as measures of the relative contribution of each input
was similar to the calculated one from the limit equilibrium-
variable to the uncertainty in the deformation of the reinforced
based slope stability analysis, passing the sloping toe and the
retaining wall. This contribution comprises two elements, the
retained zone as depicted in Figure 10. After the failure of the
degree of uncertainty of the input parameter and the sensitivity
wall, field investigation identified a number of causes that
of the deformation to changes in that parameter.
contributed to the wall failure. It was reported that the stability
It is noted from the results of the sensitivity analyses that the
analyses based on the current design approaches indicated that
friction angle of soil has significant influence on the deforma-
the wall had not been adequately designed to meet the stability
tion of the reinforced retaining wall. Based on the present study,
requirements specified by the current design approaches of
the friction angle of soil is positively correlated whereas the
FHWA (2001) recommendations, due possibly to improper
other two parameters, i.e., unit weight and interaction angle are
modelling of the wall system including the topology.
negatively correlated with the displacement of the reinforced
In fact, both the external and the global factor of safety values
retaining wall. This information has to be taken to advantage
were well below their required minimum values. Secondly, a
while proportioning the retaining wall along with the placement
low-quality soil, with a significant percentage of fines, avail-
of the reinforcement layers in the wall.
able at the site was used as the select fill, presumably assigning
an internal friction angle to the soil that is considerably greater
6. Application to case study than the actual value (Yoo and Jung 2006). It was also noted
that the rainfall together with the construction deficiency of not
The procedures given in FHWA (2001) and Sajna et al. (2008) waterproofing the surface of the reinforced zone was also
for internal and external stability analyses of reinforced retain- responsible for the wall failure as the rainwater infiltration
ing wall are made use of in the back analysis of a failed might have decreased the shear strength of the select fill.

Figure 10. Wall geometry and trace of failure surface obtained from limit equilibrium analysis (after Yoo and Jung 2006).
Geomechanics and Geoengineering: An International Journal 195

Table 6. Mean m, standard deviation s, COV and distribution of random variables

Standard Coefficient
Random variable Mean value, m deviation, s of variation Distribution Source of data

f( ) 22 4.4 20% Normal Low and Tang (1997), Phoon and Kulhawy (1999),
Hoeg and Muruka (1974)
 (kN/m3) 19 1.9 10% Normal Low and Tang (1997), Phoon and Kulhawy (1999),
Hoeg and Muruka (1974)
T (kN/m) 60 6 10% Normal Low and Tang (1997)
 ( ) 15 1.5 10% Normal Chun et al.(2004)

In this study, an attempt has been made to evaluate the effect response surface method as discussed in the previous sections
of variability of soil properties, soil reinforcement interaction was used. The sampling points considered for the reliability
angle and tensile strength of reinforcement on the overall sta- analysis and the deformation evaluated using the FEM at the
bility of the SRW. The mean and other distribution properties of corresponding points are given in Table 9. The spreadsheet
Downloaded by [University of Auckland Library] at 15:07 13 October 2014

the selected random variables are presented in Table 6. Taylor setup for the evaluation of reliability index for the SRW is
series method (Duncan 2000) is used to evaluate the reliability shown in Figure 11 and the value of reliability index is also
index for external and internal stability of the wall and results given. It is noted that the reliability index obtained was -3.21,
are given in Tables 7 and 8. In the reliability analysis of SRW which supports the fact that the reinforced segmental retaining
the effect of cohesion has not been included, although the select wall was not designed properly and also the variability of soil
fill was having a little apparent cohesion. The coefficient of properties and soil reinforcement interaction angle were not
variation (COV) of shear strength of soil varies from 20 to 55% accounted for during design stage. All these might have con-
and COV of water content from 8 to 30% and permeability tributed to the failure of the segmental retaining wall. Based on
around 100% (Phoon 2004). In case these variations could have the present study, it can be concluded that the reliability analy-
been included in the reliability analysis of the SRW, it is sis is the only methodology ensuring self-consistency in both
possible to show that the reliability index would have further physical and reliability terms. Therefore, the reliability ana-
reduced. lyses have to be used in the back analysis of the failed case
In order to evaluate the reliability index with respect to the histories in the geotechnical engineering.
deformation of the reinforced segmental retaining wall, the It can be inferred from Table 7 that the probability of failure
is less and the wall is safe against overturning as stated by
deterministic analysis (Yoo and Jung 2006). In the case of
Table 7. Factors of safety and reliability index for different external modes of sliding failure, a negative reliability index is obtained which
failure means a very high probability of failure (greater than 0.5)
indicating a certain failure event. It is to be noted that whenever
Mode of failure Factor of safety Reliability index (b)
the reliability index becomes zero, the normal probability of
External sliding 1.03 -2.35 failure will become equal to 0.5. It is noted from Table 8 that the
Overturning 2.57 3.70 reliability index values are in line with the factor of safety
values in the case of pullout failure except for two reinforce-
ment layers. For the layers located at a depth of 1.6 and 2.2 m
from the top of the SRW, the factors of safety are greater than
Table 8. Factors of safety and reliability index for internal modes of failure
the required values, but the reliability index values are less than
Reliability index (b) Factor of safety the recommended values.
Depth of reinforcement layer Rupture Pullout Rupture Pullout
from top of retaining wall (m) case case case case
Table 9. Input sampling points for approximation of performance function
0.4 6.93 -0.49 9.44 2.37 based on the FEM
1.0 4.62 0.37 3.86 1.29
1.6 3.88 1.27 2.41 2.06 Sample Friction Unit weight,  Interaction Displacement,
2.2 3.11 2.19 1.75 2.83 point angle, f ( ) (kN/m3) angle,  ( ) (mm)
2.8 2.31 3.04 1.38 3.59
3.4 1.52 3.79 1.13 4.36 1 22 19 15 873
4.0 0.77 4.39 0.96 5.13 2 26.4 19 15 657
4.6 0.08 4.84 0.84 5.89 3 17.6 19 15 1150
5.2 -0.49 5.13 0.74 6.66 4 22 20.9 15 880
5.8 -0.94 5.26 0.66 7.43 5 22 17.1 15 872
6.4 -1.24 5.21 0.60 8.19 6 22 19 16.5 873
7 -1.97 5.44 0.47 7.63 7 22 19 13.5 878
196 S. Sayed et al.

((x-m)/σ) surface method. Even when a comprehensive reliability analy-


sis cannot be made, sensitivity analysis can be used to identify
the random variables that contribute most to the failure prob-
x* Mean(m) Std. Dev(σ) vn
ability. In case a high accuracy is needed, the results obtained
φ 36.15 22 4.4 3.21
from the sensitivity analysis can be improved by Monte Carlo
γ 18.89 19 1.9 −0.05
δ 15.05 15 1.5 0.03 simulations. Based on the application of the finite element
reliability analysis to the failed segmental retaining wall pro-
Correlation matrix, “crmat”
blem, it is noted that if the reliability analysis could have been
1 0 0 performed before the construction of the retaining wall, the
0 1 0 β –3.21 collapse of the wall could have been avoided. This is possible
0 0 1
G(x) 0 by changing the design based on the results of the reliability
analysis. The reliability analysis is more rational in the sense
that it not only explicitly considers the uncertainties but also
Array formula: Ctrl + Shift, then Enter aids in decision making by providing alternate designs with
=SQRT(MMULT(TRANSPOSE(vn),MMULT(MINVERSE( consistent measure of risk.
crmat),vn))) Moreover, it is to be noted that the response surface method
Downloaded by [University of Auckland Library] at 15:07 13 October 2014

Figure 11. Spreadsheet setup of ellipsoid method for reliability analysis of gives an approximate closed-form expression, based on the
reinforced retaining wall with performance function G(x) based on the FEM. values selected. The approximation could be inadequate for
highly nonlinear performance functions. Also, as a general
rule, a closed-form expression developed using regression ana-
7. Summary and conclusions lysis is valid only within the range of the values considered for
the random variables; extrapolation beyond the range may not
Uncertainties are inherent in engineering problems and the be accurate. In the response surface approach, the number of
scatter of geotechnical parameters from their nominal ideal deterministic analyses required to construct an approximate
values is unavoidable. The response of geotechnical systems closed-form expression may be quite large for problems with
can sometimes be very sensitive to uncertainties encountered in a large number of random variables, thus making this method
the material properties, manufacturing conditions, external time-consuming. Even if a problem has a small number of
loading conditions and analytical and/or numerical modelling. random variables, it should be kept in mind that the reliability
Probabilistic-based formulations of safety evaluation of geo- estimate using the response surface approach is only as accurate
technical problems have been developed to account for uncer- as the closed-form approximation to the performance function.
tainties through stochastic simulations and probabilistic It is opined that the present study helped to overcome a lan-
analysis. Reliability-based design is the only engineering meth- guage barrier that hampers wider adoption of the more consis-
odology currently available which can ensure self-consistency tent Hasofer–Lind reliability index. The methodology of
in both physical and reliability terms. It is also compatible with reliability analysis is well established and it would be very
the theoretical basis underlying other disciplines such as struc- useful in evaluating the safety of new and emerging geotechni-
tural design. The serviceability performance of the retaining cal problems.
wall systems is of considerable importance because of the
concerns of damage to the nearby infrastructures from exces-
sive ground movements.
The main purpose of the paper was to evaluate the probability References
of failure of the reinforced retaining wall by performing defor- Allen, T.M., Christopher, B.R., Elias, V., and DiMaggio, J., 2001.
mation analysis due to various uncertainties in soil and soil- Development of the simplified method for internal stability of
reinforcement interaction parameters. Once the spreadsheet is mechanically stabilized earth (MSE) walls. Washington State
set up, reliability indices corresponding to different values of Department of Transportation, Report WA-RD 513.1, 108 pp.
lim can be obtained by repeatedly evaluating b at closed spaced Babu, G.L.S., and Srivastava, A., 2007. Reliability analysis of allow-
increments of lim. The plots obtained by such an analysis can able pressure on shallow foundation using response surface
assist in evaluating the serviceability performance of the rein- method. Computers and Geotechnics, 34, 187–194.
forced retaining walls when uncertainties of the variables Bathurst, R.J., Allen, T.M., and Nowak, A.S., 2008. Calibration con-
cepts for load and resistance factor design (LRFD) of reinforced
involved in the analysis are explicitly considered. The spread-
soil walls. Canadian Geotechnical Journal, 45, 1377–1392.
sheet approach is simple and intuitive because it works in the
Bucher, C.G., and Bourgund, U., 1990. A fast and efficient response
original space of the variables. It does not involve the orthogo- surface approach for structural reliability problems. Structural
nal transformation of the correlation matrix and iterative Safety, 7, 57–66.
numerical partial derivatives are done automatically on spread- Chalermyanont, T., and Benson, C.H., 2004. Reliability-based design
sheet objects which may be implicit or contain codes. The for internal stability of mechanically stabilized earth walls.
spreadsheet-based reliability approach presented in this study Journal of Geotechnical and Geoenvironmental Engineering,
can operate on stand-alone numerical packages via the response ASCE, 130(2), 163–173.
Geomechanics and Geoengineering: An International Journal 197

Christian, J.T., and Baecher, G.B., 2004. Reliability and Statistics Phoon, K.K., 2004. Towards reliability-based design for geotechnical
in Geotechnical Engineering. New York: John Wiley and engineering, Special Lecture for Korean Geotechnical Society,
Sons. Seoul, 9, July 2004, 1–23.
Christian, J.T., Ladd, C.C., and Baecher, G.B., 1994. Reliability and Phoon, K.K., and Kulhawy, F.H., 1999. Characterization of geotechni-
probability in stability analysis. Journal of Geotechnical cal variability. Canadian Geotechnical Journal, 36(4), 612–624.
Engineering, ASCE, 120, 1071–1111. Rackwitz, R., 2001. Reliability analysis – a review and some perspec-
Chun, B.S., Kim, K.M., and Min, D.K., 2004. A study on reliability tives. Structural Safety, 23(4), 365–395.
analysis for reinforced earth retaining walls. Third Asian Rajagopal, K., 1998. Users manual for geotechnical finite element
Geotechnical Conference on Geosynthetics, 21–24 June 2004, modelling GEOFEM, Department of Civil Engineering, Indian
Korea. Seoul, 248–254. Institute of Technology Madras, Chennai, India, 62 pp.
Duncan, J.M., 2000. Factors of safety and reliability in geotechnical Sajna Sayed, Dodagoudar, G.R., and Rajagopal, K., 2008. Reliability
engineering. Journal of Geotechnical and Geoenvironmental analysis of reinforced soil walls. Indian Geotechnical Journal,
Engineering, ASCE, 126(4), 307–316. 38(1), 47–65.
FHWA, 2001. Mechanically stabilized earth walls and reinforced soil Sayed, S., Dodagoudar, G.R., and Rajagopal, K., 2008. Reliability
slopes design and construction guidelines. National Highway analysis of reinforced soil walls under static and seismic forces.
Institute, Federal Highway Administration, US Department of Geosynthetics International, 15(4), 246–257.
transportation, Washington, DC. Sia, A.H.I., and Dixon, N., 2008. Deterministic and reliability-based
Downloaded by [University of Auckland Library] at 15:07 13 October 2014

Goh, A.T.C., and Kulhawy, F.H., 2005. Reliability assessment of design: veneer cover soil stability. Geosynthetics International,
serviceability performance of braced retaining walls using a 15(1), 1–13.
neural network approach. International Journal for Numerical Smith, G.N., 1985. The use of probability theory to assess the safety of
and Analytical Methods in Geomechanics, 29, 627–642. propped embedded cantilever retaining walls. Geotechnique,
Gupta, S., and Manohar, C.S., 2004. An improved response surface 35(4), 451–460.
method for the determination of failure probability and impor- Tandjiria, V., Teh, C.I., and Low, B.K., 2000. Reliability analysis of
tance measures. Structural Safety, 26, 123–139. laterally loaded piles using response surface methods. Structural
Gupta, S., and Manohar, C.S., 2004. Improved response surface Safety, 22, 335–355.
method for time-variant reliability analysis of nonlinear random Thurner, R., and Schweiger, H.F., 2000. Reliability analysis for geotechni-
structures under non-stationary excitations. Nonlinear Dynamics, cal problems via finite elements-a practical application. Proc. GeoEng
36, 267–280. 2000 and International Conference on Geotechnical and Geological
Haldar, A., and Mahadevan, S., 2000. Probability, reliability, and Engineering, 19–24 November 2000 Australia. Melbourne.
statistical methods in engineering design. New York: John Whitman, R.V., 1984. Evaluating calculated risk in geotechnical engi-
Wiley and Sons. neering. Journal of the Geotechnical Engineering, ASCE, 110(2),
Hasofer, A.M., and Lind, N.C., 1974. Exact and invariant second 145–188.
moment code format. Journal of Engineering Mechanics, Withiam, J.L., Voytko, E.P., Barker, R.M., Duncan, J.M., Kelly, B.C.,
ASCE, 100(1), 111–121. Musser, S.C., and Elias, V., 1998. Load and resistance factor
Hoeg, K., and Muruka, R.P., 1974. Probabilistic analysis and design of design (LRFD) for highway bridge substructure’, FHWA HI-98-
a retaining wall. Journal of the Geotechnical Engineering 032, Federal Highway Administration, Washington, DC, USA.
Division, ASCE, 100(3), 349–365. Wong, F.S., 1985. Slope reliability and response surface method.
Kaymaz, I., and McMahon, C.A., 2005. A response surface method Journal of the Geotechnical Engineering, ASCE, 111, 32–53.
based on weighted regression for structural reliability analysis. Xu, B., and Low, B.K., 2006. Probabilistic stability analyses of
Probabilistic Engineering Mechanics, 20, 11–17. embankments based on finite – element method. Journal of the
Kazimierowicz-Frankowska, K., 2005. A case study of a geosynthetic Geotechnical and Geoenvironmental Engineering Division,
reinforced wall with wrap-around facing. Geotextiles and ASCE, 132(11), 1444–1454.
Geomembranes, 23, 107–115. Yao, T.H.J., and Wen, Y.K., 1996. Response surface method for time-
Kulhawy, F.H., and Phoon, K.-K., 2002. Observations on geotechnical variant reliability analysis. Journal of Structural Engineering,
reliability-based design development in north america. Foundation ASCE, 122(2), 193–201.
design codes and soil investigation in view of international harmo- Yoo, C., 2004. Performance of a 6-year-old geosynthetic reinforced
nization and performance based design. In: Y. Honjo, O. Kusakabe, segmental retaining wall. Geotextiles and Geomembranes, 22,
K. Matsui, M. Kouda and G. Pokharel, eds. Balkema, Lisse- 377–397.
Netherlands, April 2002, pp. 31–48. Yoo, C., and Jung, H.-Y., 2006. Case history of geosynthetic reinforced
Low, B.K., 2005. Reliability-based design applied to retaining walls. segmental retaining wall failure. Journal of Geotechnical and
Geotechnique, 55, 63–75. Geoenviornmental Engineering, ASCE, 132(12), 1538–1548.
Low, B.K., and Tang, W.H., 1997. Reliability analysis of reinforced Yoo, C., and Song, A.R., 2006. Effect of foundation yielding on
embankments on soft ground. Canadian Geotechnical Journal, performance of two-tier geosynthetic-reinforced segmental
34(5), 672–685. retaining walls: a numerical investigation. Geosynthetics interna-
Low, B.K., and Tang, W.H., 2004. Reliability analysis using object- tional, 13(5), 181–194.
oriented constrained optimization. Structural Safety, 26(1), Yuan, J.X., Yang, Y., Tham, L.G., Lee, P.K.K., and Tsai, Y., 2003.
69–89. New approach to limit equilibrium and reliability analysis of soil
Melchers, R.E., 1999. Structural reliability: analysis and prediction. nailed walls. International Journal of Geomechanics, ASCE,
Chichester: John Wiley and Sons. 3(3), 145–151.

You might also like