You are on page 1of 7

Comparison between Numeric

Ch
1

2
Department o Ci!il En"ine
E$
%bstract
Uncertainty in soil properties may lead
slope stability analysis (e.g. Heterogeneous
natural slopes are difficult to evaluate
strength. This is a main cause of wrong cal
safety and wrong design to protect
probabilistic study is carried out to asses
uncertainty of soil properties on stability of s
describes the Random-finite element (R
stability ris" assessment compares between
limit e#uilibrium method to analy$e slope
Furthermore this paper compares
element method and random-finite elemen
result of RFE analyses with monte carlo
describe by a range of parametric varia
method is used in con%unction with monte
to determine probability of failure (&
f
! and it
probability distributions.
Ke&words' 'lope 'tability (nalysis) Random
method) *umerical ethod.
1( )ntroduction
'lope stability analysis is one of the
geotechnical engineering.The earliest studi
the +,-.s (e.g.) atsuo and /uroda) +,-0
Tang et al.) +,-21 3anmarc"e) +,--! and
steadily (e.g.) 45(ndrea and 'angrey) +,67
+,6-1 9hitman) 7...1 9olff) +,,21 8acasse)
et al.) +,,01 :hristian) +,,,1 8acasse and
Hassan and 9olff1 4uncan) 7...1 '$yna"iew
E;-Ramly et al.) 7..71<riffiths and Fenton) 7
al.) 7..6!.
This paper focus on review slope stab
using limit e#uilibrium methods and num
Especially random finite element method
rigorous method of probabilistic geotechn
which finite-element methods are combine
field generation techni#ues and fully acco
correlation and averaging. RFE method
slope stability ris" assessment tool that doe
prior assumption relating to the shape or
failure mechanism. (Griffiths and Fenton, 200
1*
th
National Con!e
*$1+Ma&2+1,-he Empress .o
merical and Limit Equilibrium Methods or /lope
Chollada Kanjanakul
10
-anan Chub$uppakarn
2

"ineerin"# 1acult& o En"ineerin"# 2rince o /on"kla 3ni!ersit
$mail'
1
chollada$ka4hotmail(com#
2
tanan22*54psu(ac(th
=
lead to problems in
ous in soil mass of
te the e>act shear
g calculate factor of
ection slope!.Thus)
ssess the effect of
of slope.This paper
t (RFE! to slope
een numerical and
stability.
es between finite
ment method. The
carlo simulation are
variations. (naly$ed
nte carlo simulation
nd its corresponding
dom-finite element
the oldest tas"s in
tudies appeared in
+,-01 (lonso) +,-21
nd have continued
+,671 8i and 8umb)
asse) +,,01 :hristian
and *adim) +,,01
"iewic$ et al.) 7..71
n) 7..-1 <riffiths et
stability analysis by
numerical methods.
od (RFE!) a more
echnical analysis in
bined with random-
accounts for spatial
is also powerful
t does not re#uire a
or location of the
, 2004!
2( /lope stabilit& anal&sis t
(nalysis method of slope stab
ma%or parts as shown in figure+.(
and *umerical methods!
2(1 Limit equilibrium metho
8imit e#uilibrium methods f
can be separated in two parts.Th
and moment of e#uilibrium) The
factor of safety) 9hich is de
e#uilibrium and another one is
analysis.The result of these meth
9hich is depend on random vari
we concern.

Slope
StabilityAna
Limit Equilibrium
Methods
Satisfies force &
moment of equilibrium
Probabilistic Slope
Stability Analysis
Force
- Force equilibrium methods
(e..Lo!e and "arafiath #$%&'
(.S. Army )orps of Enineers
#$*&+
Moment
- ,rdinary Method of Slices
(Fellenius #$-*+
Force+Moment
- .ishop/s Modified Method
#$00
- 1anbu2s 3enerali4ed Procedure
of slices(1anbu #$%5+
- Morenstern & price2s Method
#$%0
- Spencer #$%*

Fig. + 'lope 'tability (

2(1(1 /atisies orce and mo
8imit e#uilibrium method is
method to discuss about balan
force.'lope will have a balance w
of soil mass! is e#ual e>ternal
(driving force! and balance of so
increase of slope angle becau
influence on the reduction of fa
of F.'. is less than +.. the slope s
n!ention on Ci!il En"ineerin"
ss .otel# Chian" Mai# -hailand
lope /tabilit& %nal&sis
ersit&# /on"kla 6+11+
sis theories
stability can be divided in two
re+.(8imit e#uilibrium methods
ethods
ds for slope stability analysis
ts.The first one is satisfies force
The result of these method is
depend on mean of force
e is probabilistic slope stability
method is probability of failure)
variable of soil parameter that
tyAnalysis
6umerical
Methods
Finite 7ifference
methods (F7M+
Finite Element
Methods (FEM+
8andom Finite Element
Method (8FEM+
7iscrete Element
methods (7EM+
.oundary Element
methods (.EM+
7iscrete Fracture
6et!or9 methods
(7F6+
Meshless Local Petro:-
3aler9in
Method(MLP3+
Lodalen Slide
lity (nalysis ethod
moment o equilibrium
d is satisfied a stability analysis
alance of resisting and driving
ce when resisting force (weight
rnal force acting on soil mass
of soil slope is depend on the
ecause it has a significantly
of factor of safety. ;f the value
pe stability is unstable.
2
Table + describe forces e#uilibrium methods (e.g.) 8owe
and "arafiath +,2.1 U.'.(rmy :orps of Engineers) +,-. !.This
applies to any shape of slip surfaces and satisfies both
hori$ontal and vertical force e#uilibrium.However this
method is not consider moment e#uilibrium.(nother
method for slope stability analysis) for e>ample bishop5s
modified method(?ishop) +,@@! doesn5t satisfies hori$ontal
force e#uilibrium and applies only for circular slip surfaces.

Table + :haracteristics of 'lope 'tability (nalysis (4uncan
and 9right) +,6.!
Method Characteristics
'lope 'tability :harts
(Aanbu +,26 1 4uncan
et al) +,6-!
(ccurate enough for many purposes
Faster than detailed computer ana-
lyses
=rdinary ethod of
'lices (Fellenius) +,7-!
+. =nly for circular slip surfaces
7. 'atisfies moment e#uilibrium
B.4oes not satisfy hori$ontal or verti-
cal force e#uilibrium
?ishop5s odified
ethod (?ishop) +,@@!
+. =nly for circular slip surfaces
7. 'atisfies moment e#uilibrium
B.'atisfies vertical force e#uilibrium
0.4oesn5t satisfy hori$ontal force
e#uilibrium
Forces e#uilibrium
method (e.g. 8owe and
/arafiath +,2.1U.'.(rmy
corps of Engineer)+,-.!
+. (ny shape of slip surfaces
7. 4on5t satisfy moment e#uilibrium
B.'atisfies both hori$ontal and verti-
cal force e#uilibrium
Aanbu5s <enerali$ed
&rocedure of 'lices
(Aanbu) +,26!
+. (ny shape of slip surfaces
7.'atisfies all condition of e#uilibrium
B.&ermits side force locations to be
varied
0.ore fre#uent numerical problems
than some other methods
orgenstern and prices
ethod(orgenstern C
&rice) +,2@!
+. (ny shape of slip surfaces
7.'atisfies all condition of e#uilibrium
B.&ermits side force orientations to
be varied
'pencer5s ethod
('pencer) +,2-!
+. (ny shape of slip surfaces
7.'atisfies all conditions of e#uili
brium
B. &ermits side force locations to be
varied

2(1(2 2robabilistic slope stabilit& anal&sis
This is one of slope stability analysis method was
developed for solve uncertainty in soil properties.
&robabilistic analyses are normally used to evaluate
statistical distribution of a performance function. ?ased on
statistical characteristics of input variables) &rocess of this
method for analy$e factor of safety come from distribution
of random variable by monte carlo simulation. Then)
probability density function (pdf! leads to measure
probability of failure (&
f
!.
4egroot (+,,2! said that engineers can be ad%ust for best
estimate and a measure of uncertainty in the best estimate
by
a.) Mean (
x
) ; is a statistical measure of
normal distribution and measure of Unimodal &attern
data.That data collected from a population with a constant
standard deviation.

x
=
1
N
X

N
=1
(+!

b.! 'tandard 4eviation (
>
! is the s#uare root of the
variance which is used to e>plain the deviation of
population.

x
= s = _
1
N-1
(X


x
)
2 N
=1
(7!

c.! :oefficient of variation (:=3! 1 :onsideration ratio of
standard mean for compare standard deviation data more
than two value.
C0I
x
=

x

x
(B!

Random variable method in soil parameter that can be
First =rder 'econd oment ethod) &oint Estimate ethod
and onte :arlo 'imulation.

2(1(2(1 1irst order second moment method
718/M9
F=' method is suitable for function that have more
than or e#ual two variable parameter that used first part
from Taylor5s series appro>imation for estimate factor of
safety (F.'.!.
e
n
X X X X g S F + = ) .., ,.........
3
,
2
,
1
( . . (0!
9hen
g (D
;
! is Factor safety function
e is odel Error
E#uation @)2 shows coefficient of variation (:=3! that
consideration ratio between mean and standard
deviation of random variable.

| | | |
2
u
s

u
s
cov
2
u
s
F.S.
F.S. V
(

= (@!
3
| |
2
i
x
2
j
x
g
F.S. V
F.S.
2

= =
(
(

(2!
9hen

i
x is 3ariable parameter
e is odel Error

( ) ( ) 1
U
S g 1
U
S g F.S. + = (-!

From e#uation 0 lead to find variable parameter of F.'. and
&robability density function (pdf! from vertical a>is (&hoon
and /ulhawy) +,,,! because of undrained shear strength
('
u
! and :haracteristic of distribution of F.'. as shown in Fig
7 after that find &robability of failure that F.'. < +


Fig. 7 &robability of failure (&
f
!
2(1(2(2 2oint estimate method 72EM9
9hen function has variable parameter more than two
(E#. 0!) the method of stability analysis is therefore
complicated. Rosenblueth (+,-@! introduced point estimate
method) &E approach) based on Random 3ariable (ean)
3ariance and '"ewness! to understand similarity of
probabilistic distribution and load distribution.For e>ample
rigid beam that is applied vertical load action in beam as
shown in Fig B
Harr. (+,6-! compare radius of gyration that shows
distribution from considerate point. ;t means standard
deviation as shown in fig. B. 9hen beam was acted by load)
reaction has occurred. Hence) p
-
force has appeared at > E
>
-
) and at > E >
F
for p
F
force. This approach was called as
GTwo-point estimatesH of distribution of function f(>! .They
have relation about variable as shown in e#uation 6.


(
(
(
(
(

\
|
+
=
+
2
2

1
1
1 1
2
1
p
(6!

Fig. B ?eam with 3ertical 8oad ((fter Harr +,6-!


+
=

p 1 p (,!

+

+ =
+
p
p
x

x
x (+.!

+
+ =

p
p
x

x
x (++!
9hen
p
-
E The resultant force at > E >
-
from beam.
p
F
E The resultant force at > E >
F
from beam.


E '"ewness
2(1(2(, Monte carlo simulation
onte :arlo 'imulation or implementation is a
statistical method.(rising from implementation of the
parameter values is selected for the analysis or calculation
repeated several times.
Finally replied in a probability distribution.Until it get
the results by calculating the factor of safety repeat many
times and large enough to have the distribution of F.'.
(&robability density function!.Then &
f
can be calculated
from the possibility that F.'. I + .
2(2 Numerical method
*umerical modelling starts by dividing the slope into a
finite number of $ones or elements. Forces and strains are
then calculated for each element using the appropriate
constitutive laws for the materials in the slope.The most
common numerical analysis methods available are Finite
4ifference methods (F4!) Finite Element ethods (FE!)
4iscrete Element methods (4E!) ?oundary Element
methods (?E!) and 4iscrete Fracture *etwor" methods
(4F*!.
pdf
F.S.

F.S.
F.S. =1
P
f
x
x
f(x)
f(x)
a
b
A
A
B
B
E [X] 1
E [X]
X
-
p
-
X
+
p
+
4
The difference of numerical method compared with
limit e#uilibrium methods is constitutive model and shear
strength reduction techni#ue has been investigated.
2(2(1 Constituti!e model ormulation
The mechanical behavior has been described on
numerous models such as
2(2(1(1 /hear stren"th equilibrium
The mechanical constitutive formulation is based on
?ishop5s effective stress and a linear elastic) perfectly
plastic ohr-:oulomb soil model.The failure criterion
e>tended for unsaturated conditions is

' )tan
w
u
a
(u ' )tan
a
u ( c'
!
" + + =
(+7!

9hen

f
E 'oil shear strength
:5 E The effective cohesion
E The effective angle of friction
E The total stress
U
a
E The pore air pressure
U
w
E The pore water pressure

2(2(1(2 .&draulic conducti!it& unction
Hydraulic behavior describes the hydraulic conductivity
suction level is
(s)
#
$
sat
$ $(s) = (+B!

9here the relative coefficient of permeability "
r
(s! has
been modeled using the function proposed by 3an
<enuchten (+,6.!


2
2n
1 n
n
s 1
n
n 1
n
s 1
1 n
s 1
sat
k k s

=
(

(
(
(

(+0!

and where "
sat
is the saturated hydraulic conductivity
.
(&.
(rnold and .(. Hic"s)7.++!
7.7.+(B /hear /tren"th :eduction -echnique
The principal of shear strength reduction techni#ue in
finite element analysis is to simultaneously reduce c and
tan in small increments until failure occurs in the
numerical analysis.;f shear strength parameters at failure are
c
r
and

r
.The factor of safety (F.'.! can be defined as (;ndra
and Helmut F.) 7.++!
F. S. =
tun
tun
r
=
c
c
r
(+@!
2(2(2 -he :andom 1inite Element Method
the Random Finite Element ethod (RFE! that
combines random field theory with deterministic finite
element analysis was developed by <riffiths and Fenton in
the early +,,.5s and has been applied to a wide range of
geotechnical applications. ;n a stability analysis) input to
RFE is provided in the form of the mean) standard
deviation and spatial correlation length of the soil strength
parameters at the point level) which may consist of several
layers with different statistical input parameters. ;n the
absence of site specific information) there is an increasing
number of publications presenting typical ranges for
thestandard deviation of familiar soil properties) e.g. 8ee et
al.
;n RFE) local averaging is fully accounted for at the
element level indicating that the mean and standard
deviation of the soil properties are statistically consistent
with the mesh density. 'ince the finite element method of
slope stability allows mechanisms to see" out the most
critical path through the soil) the methodoffers great
promise for more realistic reliability assessment of slopes
and other geotechnical applications. (<riffiths and Fenton
7...!
2(, Comparison between Numerical and
limit Equilibrium Methods
The difference between *umerical and 8imit E#uilibrium
ethods is
(+! 8imit e#uilibrium methods %ust give an estimate of factor
of safety with no information on deformation of the slope
but in numerical analysis) the failure surface can evolve
during the calculation in a way that is representative of the
natural evolution of the physical failure plane in the slope)
(9yllie and ah) 7..0!.
(7! :undall (7..7! compared the characteristics of numerical
solutions and limit e#uilibrium methods in solving for the
factor of safety of slopes and concluded that continuum
mechanics-based numerical methods have the following
advantagesJ
7.+ *o pre-defined slip surface is needed.
7.7 The slip surface can be of any shape.
7.B ultiple failure surfaces are possible.
7.0 *o statical assumptions are needed.
7.@'tructures (such as footings) tunnels) etc.! and structu
rals elements (such as beams) cables) etc.! and interface
can be included without concern about compatibility.
7.2/inematics is satisfied.
5
;t is important to recogni$e that the limit e#uilibrium
solution only identifies the onset of failure) whereas the
numerical solution includes the effect of stress
redistribution and progressive failure after failure has been
initiated. The resulting factor of safety allows for this
wea"ening effect) (;tasca :onsulting <roup) 7..@!. Hoe" et al
(7...! stated that numerical models can also be used to
determine the factor of safety of a slope in which a number
of failure mechanisms can e>ist simultaneously or where
the mechanism of failure may change as progressive failure
occurs. (Henry) 7.+.!.
For slopes) the factor of safety often is defined as the
ratio of the actual shear strength to the minimum shear
strength re#uired to prevent failure) (4awson et al) +,,,!. (
logical way to compute the factor of safety with a finite
element is to reduce the shear strength until GcollapseH
occurs in the model. The factor of safety is then the ratio of
the roc"5s actual strength to the reduced shear strength at
failure. This method is called the shear strength reduction
method) (4awson et al) +,,,) <riffith and 8ane) +,,,)
Hammah et al) 7..0!.
2(5 Comparison between 1inite Element
Method and :andom 1inite Element Method
<riffiths(7..6! compared the characteristics of finite
element method and random finite element method by
the results of nonlinear RFE analyses with monte carlo
simulations. the results are described) based on a range of
parametric variations of that called this value in statically
as the spatial correlation length as shown in Fig 0.


Fig. 0 'lope test problem (<riffiths and Fenton 7...!

;n this study) the slope inclination and dimensions given
by E72.2) H and 4E7 and the saturated unit weight of the
soil
sut
are held constant)while the undrained shear
strength c
u
is assume to be a random variable.;n the
interests of generality)the undrained shear strength will be
e>pressed in dimensionless from c)where

c =
c
u

sat
H
(+2!
2(5(1 -he /patial correlation len"th 79
The 'patial correlation length describes the distance
over which the spatially random values will tend to be
significantly correlated in the underlying guassian field.The
spatial correlation length can be estimated from a set of
shear strength data ta"en over some spatial region simply
by performing the statistical analyses on the log data.;n
practice) however)
Inc
is purposes)
c
and
Inc
are
interchangeable given their inherent uncertainty in the first
place not much in the current study)the spatial correlation
length has been nondimensionali$ed by dividing it by the
height of the emban"ment H and will be e>pressed in the
form

H
lnc

= (+-!

different in magnitude from the correlation length in
real space)and)for most ;t has been suggested (see) e.g. 8ee
et al.) +,6B1 /ulhawy et al.) +,,+! that typical v
c
values
for undrained shear strength lie in the range ..+-..@.The
spatial correlation length) however) is less well documented
and may well e>hibit anisotropy) especially since soils are
typically hori$ontally layered.9hile the advanced analysis
tools used later in this study have the capability of
modeling an anisotropic spatial correlation field) the spatial
correlation) when considered) will be assumed to be
isotropic.
the result is shown in Figure @ and 2 and consists of
undrained clay) with shear strength parameters
u
E . and
c
u
.


Fig @. :omparison of the probabilities of failure predicted by
RFE and by Finite element local geometric averaging
alone (<riffiths and Fenton 7...!

6
From Fig. @ the probability of failure (&
f
! from random
finite element method is increase at lower coefficient of
variation (v
c
! when spatial correlation length becomes large)
the probability of failure is conservative because a large
value of spatial correlation length ( ! will imply smoothy
varying field) while a small value will imply a raged field.'o)
spatial correlation length has a significantly influence on
RFE results compared with FE and Fig.2 shows the result
is shown that spatial correlation length only starts to have a
significantly influence on the Fs vs. &f relationship using
finite-element when the correlation length becomed small (
II+ !. (<riffiths and Fenton 7...!


Fig 2. &robability of failure versus Fs (?ased on mean! using
Finite-element (<riffiths and Fenton 7...!
2(; /ummar&
The difference between *umerical and limit e#uilibrium
methods is numerical implement is dividing the slope into a
finite number of $ones or elements and characteristics of
solutions is representative of the natural evolution of the
physical failure plane in the slope because numerical
satisfies both of stress and strain ( /inematics ! but limit
e#uilibrium methods satisfies only stress.
Random finite element method solutions always gives a
higher probability of failure and single random variable
approach becomes vary over a wider range of coefficient of
variation than when using finite element local averaging
alone. This is caused by strength distribution find the
wea"est point in the slope element great and intensive
more than finite element method and spatial correlation
length has a significantly influence on RFE results because
spatial correlation length is using in random variable
aapproach for find probability of failure %ust only RFE
method.
2(< Notice
;n thesis) The author selected unsaturated slope
stability analysis into application form the Random Finite
Element ethod (RFE!) it uses elastoplasticity in a finite-
element model combined with random field theory in a
onte-:arlo framewor".9hich implement that shown in fig
-.


Fig -. the Random Finite Element ethod (RFE!

;n this study) The random variables characteri$ed
statistically by lognormal distributions. ;t has three
parameters the mean) the standard deviation and the
spatial correlation length. The variability can conveniently
be e>pressed by the dimensionless coefficient of variation.(
benefit of RFE is that the shape and location of the failure
surface is not determined a priori and the algorithm is able
to see" out the most critical path through the
heterogeneous soil mass.

:eerences
K+L (lonso) E.E. GRis" analysis of slope and its application to
slope in :anadian sensitive clays.H <eotechni#ue) 72) 0@B-
0-7.)+,-2.
K7L :hristian) A.T. and ?aecher) <.?. G&oint-estimate method as
numerical #uadrature.H (':E. A.<eotech <eoenv. Eng.)
+7@(,!) --,--62)+,,,.
KBL :hristian) A.T.)8add) :.: and ?aecher) <.?. GRealiability
applied to slope stability analysis.H (':E. A.<eotech Eng.)
+7.(+7!) 7+6.-77.-)+,,0.
K0L 45(ndrea) R.(. and 'angrey) 4.(. G'afety factors for
probabilistic slope designH (':E A.<eotech. Eng.) +.6(<T,!)
++.+-+++6)+,67.
K@L 4egroot 4.A. G(naly$ing 'patial 3ariability of ;n 'itu 'oil
&roperties.H &roceedings of Uncertainty ,2)<eotechnical
'pecial &ublication *o. @6. (':E. +J7+.-7B6)+,,2.
7
K2L 4uncan) A.. GFactor of safety and reliability in geotechnical
engineering).H (':E. A.<eotech. <eoenv. Eng.) +72(0!) B.--
B+2) 7....
K-L 4.3. <riffiths) Ainsong Huang) <ordon (. Fenton. Ris"
(ssessment in <eotechnical Engineering. Aohn wiley C'ons)
;nc.)7..6) pp.B6+-0...
K6L 4.3. <riffiths) Ainsong Huang) <ordon (. Fenton. Ris"
(ssessment in <eotechnical Engineering. Aohn wiley C'ons)
;nc.)7..6) pp.B6+-0...
K,L E.;.-Ramly) H.) orgenstern) *.R.) and :ruden) 4..
G&robabilistic slope stability analysis for practiceH :an.
<eotech. A.)B,) 22@-26B.) 7..7.
K+.L Fenton) <. (.) and <riffiths) 4. 3. GRis" (ssessment in
<eotechnical Engineering.H Aohn 9iley C 'ons) Hobo"en)
*ew Aersey)7..6.
K++L <riffiths) 4.3. and Fenton <.(. G&robabilistic ethods in
<eotechnical EngineeringH. 'pringer) 9ien) *ew Mor") :;'
:ourses and 8ectures *o. 0,+) ;nternational :entre for
echanical 'ciences.)7..-
K+7L Hassan) (.. and 9olff) T.F. GEffect of deterministic and
&robabilistic models on slope reliability inde>.)H in slope
stability 7...) geotechnical special publication *o.+.+)
(merican society of civil engineers) *ew Mor") pp. +,0-7.6.)
7....
K+BL H. T. :hiwaye. ( :omparison of The 8imit e#uilibrium and
*umerical odelling (pproaches to ris" analysis for open pit
mine slopes. .(. thesis) University of the 9itwatersrand
Aohannesburg) 7.+..
K+0L ;ndra *oer Hamdhan and Helmut F.G'lope stability analysis
of unsaturated soil with fully coupled flow-deformation
analysis.H ;(< publication 'al$burg) (ustria) 7.++.
K+@L A.. 4uncan. State of The Art:Limit Equii!rium and Finite
Eement Ana"sis of Sopes. A. <eotech. Eng. (m. 'oc. :iv.
Eng.) +,,2) +77(-!)pp. @---@,2.
K+2L 8acasse) '. GReliability and &robabilistic methodsH in &roc
+B
th
;nt. :onf. on soil mechanics foundation engineering.)
pp.77@-77-.) +,,0.
K+-L 8acasse) '. and *adim) F. GUncertainties in :haracterising
soil properties.H ;n (':E Uncertainties5,2 :onference
&roceedings) :.H. ?enson) Ed.) adison.) 9.;.) pp0,--@.) +,,2.
K+6L 8ee) ;./.) 9hite) 9.) and ;ngles.) =. <.G<eotechnical
Engineering.H &itman) 8ondon odelling of 'tability and Ris"
of <eotechnical 'ystems in Highly 3ariable 'oils) +,6B.
K+,L 8i) /.'.) and 8umb) &. G&robabilistic design of slopes.H :an.
<eotech. A.) 70) @7.-@B+) +,6-.
K7.L atsuo) .) and /uroda) /. G&robabilistic approach to the
design of emban"ments.H 'oil Found) +0(+!) +-+-) +,-0.
K7+L &. (rnold and .(. Hic"s G'tochastic modeling of
unsaturated slope stability.H Unsaturated 'oils-(lonso C
<ens(eds!)Taylor C Francis <roup)8ondon) 7.++.
K77L '$yna"iewic$) T.) <riffiths) 4.3. and Fenton) <.(. G(
&robabilistic investigation of c5) C5 slope stability.H ;n &roc.
2
th
;nt. :ong. *umerical method in engineering and scientific
applications.) :;E*;:'5.7) 'ociedad 3ene$olana de
e5todos *ume5ricos en ;ngenier5ia) pp.7@-B2.) 7..7.
K7BL Tang) 9.H.) Muceman) .'.) and (ng) (.H.'. G&robability
based short-term design of slopes.H :an.<eotech. A.+B) 7.+-
7+@) +,-2.
K70L 3anmarc"e) E.H. G&robabilistic modeling of soil &ro-files.H
(':E A.<eotech) Eng.) +.B(<T ++!. +77--+702)+,--.
K7@L 9hitman) R.3. G=rgani$ing and evaluating uncertainty in
geotechnical engineering.H (':E. A.<eotech) <eoenv. Eng.)
+72(-!) @6B-@,) 7....
K72L 9olff) T.F. G&robabilistic slope stability in theory=rgani$ing
and evaluating uncertainty in geotechnical engineering.H
(':E. A.<eotech) <eoenv. Eng.) +72(-!) @6B-@,) 7....

You might also like