Вы находитесь на странице: 1из 17

Proceedings of the International Symposium on Slope stability Engineering-IS-

shikoku'99,Invited
Keynote
paper,
Matsuyama, Shikoku, Japan. pp. 3 I 47-
November 8-l l, 1999
Slop Stability Engineeing, Yagi, Yamagani&Jiango 1999 Batt<ema, Ronedam, |SBN
gO
sslg O7g s
Using limit equilibrium concepts in finite element slope stability analysis
D.G. Fredlund & R. E.G. Scoular
University of Saslatchewan, Saskatoon, Sask., Canada
ABSTRACT: This paper reviews thc development of finite element
slope stability analyses and proposes
that such a method can form a practical procedure for solving slope stability problems. Several stope stabitity
methods have been proposed that make use of the finite element merhods; tir"i" "r" summarized in-this paper.
The proposed finite.e-lement method is in a fonn th3l-g* be conveniently
used in engineering practici. The
procedure lends itself to
Present
day numerical modelling techniques.
'Ihe
method tri Ueen ,ipA"t"a to take
advantage ofrecent advances in computer technology and algorithms.
The combination of a finite element strels analysi: with-a limit equilibrium analysis provides greater cer-
tainty and flexibility regarding the intemal distribution of stesses within the soil'mass. The nirmal forc"
a-long any selected slip surface can be calculated from the sfiess distribution that has been calculateJ *i.rg "
linear and non-linear stess analysis. The overall factor of safety for a slope, when the n"it. il"m"nl method
is used, can be defined as the available shear strcngth of the soil dividedbythe resisting slear st roeth. Th.
overall factor of safety is a combination of the local factors of safety within the.fop". ile.etuttine-ou"ott
factor of safety retains the basic assumptions inhercnt to the limii equilibrium de'finition oi tt "
-L.to,
of
safety.
The local facton ofsafe-ty att an exprcssion ofthe stability ofthe soil mass at each point along the slip sur-
face' The overall factor of-safety computed using the finite element method shows good "gt""fr"ot with the
facton of safety computed using any one of several limit equilibrium methods. fnl nnitJ element method
provides additional information-regarding the potential performance of a slope; information not available
when using traditional limit equilibrium methods. The results indicate that it is'important to use the effective
shear stength characterization of the soil when performing
the slope stability anaiysis. The computed factor
of safety obtained when using "
_toq
she{ shength characterizatio; of the soil, rr"y oot agree wiih the factor
of safety computed when using the finite element stress analysis method.
-
K9f wgrds:qlope stability analysis, finite elemnt, enhanced method direct method strength metho4 stess
level method, factor ofsafety, Iocal factor ofsafety.
I INTRODUCTION
Limit equilibrium methods of analysis have proven
to be a widely used and successful method for the
assessment of the stability of a slope. Limit equilib
rium methods sum forces and moments related to an
assumed slip surface passed through a soil mass
(Fredhmd
and Krabn, 1975; Fredhmd et al, l98l).
However, these methods do not utilize the stress ver-
sus stain characteristics of the soils involved. It is
well known, and intuitively rmderstood that the sta-
bility of a slope should be influenced by the stress
versus strain characteristics of a soil (Kondner
1963). A finite element analysis utilizes a stress ver-
sus stain model for the soils involved to calculate
the stresses in the soil mass. These sEesses can sub-
sequently be used to compute a factor ofsafety (Fig.
1). The complete stress state from the finite
element
analysis can be
"imported"
into a limit equilibrium
analysis where the normal stress and the shear stress
are computed corresponding to any selected slip sru-
face.
The objective of this paper is to demonsEate a
procedure for combining a finite element stress
analysis on a slope with the concepts of a limiting
equilibrium method of a'alysis. The final method is
called a
"finite
element method of slope stability
analysis" and the results are compapd to results ob-
tained when using conventional limit equilibrium
method of analysis.
31
Finite Element Analysis for Stresses
Limit Equilibrium Analysis
Figue I.Illustration showing sresses that arc
"rmported"
from a finite element analysis into a limit equilibriurn malysis.
2 BACKGROI.'ND
Bishop (1952) noted that the stresses from a limit
equilibrium method of analysis did not agree with
the achral stresses within an earth stnrcture. Other
researchers have confirmed this observation both
with experimental evidence and with numerical
modelling. Ia Rochelle (1960) estimated the stness
conditions in sGep slopes using photoelastic tests on
gelatine models. The results showed that stresses
along a slip surface were over-stressed in the lower
portion of the slip circle. Brown and King (1966)
produced critical slip surfaces from a finite element
strcss analysis of slopes using a linear elastic soil
model. The critical slip surfaces were produced by
using the angle of obliqwty, 0, along the slip surface
(i.e., dequal to (45o +
Q/a)).
Each critical slip sur-
face represented a close approximation to an essen-
tially circular shaped slip surface.
Clough and Woodward (1967) trndertook a study
to evaluate the effect of incremental loading with
single step loading as it related to strcsses and de-
formations. It was concluded that: l) stresses and de-
formations in an embankment obtained from a direct
application of the gravitational body forces on the
complete structtrre were not completely accurate,
and 2) changing Poisson's ratio interferes with the
relationship between sEesses and displacements, re-
quiring a new analysis for each case. It was con-
cluded that
"meaningfttl
stability analysis can be
made only
d
the stress distribution within the struc-
ture can be predicted reliably."
Kulhawy (1969) developed a compder program
to obtain an independent assessment of the nonnal
and shear stress distribution along an assumed slip
surface. The normal and shear shesses from an elas
tic analysis were used to calculate an overall facto
of safety. The formulation of Kulhawy (1969) wa
classified as an
"Enhanced
Limit Strength Method".
A number of finite element slope stability meth
ods have been proposed and the methods can b,
categorized as
"enhanced
limit methods,' or
"direc
methods", as shown in Figure 2.
Wright (1969) compared the factors of safety cal
culated using the
"enhanced
limit strength" methor
with factors of safety calculated using Bishop's Sim
plified method (1952). A slip surface was selecter
for comparative purposes that had a factor ofsafet
of 1.0 when using the Bishop's Simplified method. i
was concluded that the factors of safety determiner
by the
"enhanced
limit strength" method (Kulhawf
1969) were approximately 3% higher than those de
termined applyrng Bishop's Simplified method
Wright et al. (1973), using the
"enhanced
limir
sEength" method, showed that: l) along one third or
the slip surface, the local factors of safety are lesl
than the overall factor of safety, 2) the factors or
safety calculated by the finite element method using
linear elastic material properties ranged from 0% tr
4.5% higher than those calculated using Bishop's
Simplified method, and 3) the factors of safety cll-
culated by the finite element method using non-
linear elastic material properties increased witt.
Poisson's ratio and arc 2%o to 8% higher than those
calculated using the Bishop's Simplified method.
Rcsindiz (1974) ageed with the concept of using
the finite element method to calculate the stability ol
a slope; however, disagreed with points No. 2 and
No.3 of the results of Wright et al. (1973) because
the factor of safety differences were too small
Res6ndiz had developed a finite element method oi
32
Finile Element Slope Stability Methods
Direct melhods
(finite element analysis only)
Enhanced limit methods
(finile element analysis with
a limit equilibrium analysis)
Load increase
to failure
Strength decrease
to failure
Definition of Faclor of
Strength
Kulhaury 1969
Stress Level
Zenkiewiaet al 1975
Streng[h & Stress Level
Adikari and Commins 1985
. _I
( c' + dt a$' ) M
n-
zr at
Figure 2. Finite element approaches proposed in comprrting the factor ofsafety in a slope stability analysis.
slope stability analysis defined as an
"enhanced
limit
stress-level" method" n 1972 (Fig. 2). This method
used the maximum principal
stress difference of the
soil at failure to define the factor of safety. Analyses
made using non-linear stress versus strain relation-
ships led to factors of safety which in all cases were
higher (i.e., differences as large as 3Ao/o) than con-
ventional facton of safety (e.g., Ordinary method or
Bishop's Simplified method).
Zier/rriewicz et al. (1975) also proposed a finite
element method of analysis to compute the factor of
safety by using the principal stess difference in the
soil at failure to define the factor of safety. The
method is an
"enhanced
limit stress
-
level method"
(Fig. 2). Both the Resdndiz (1972) aurrd, Zienktewicz
et al. (1975) formulations are classified as
"enhanced
limit stress-level" methods.
Naylor (1982) estdblished two tJpes of finite ele-
ment slope stability methods, a
"direct,'
and an
"en-
hanced limit" method of analysis. The direct method
used a finite element nodal formulation to define the
slip surface and the factor of safety directly from the
analysis.
The proposed
"direct"
slope stability
method defined the factor of safety either as the in-
creased load necessary to cause failure, or as the re-
ciprocal of the reduction in the strength properties
required in order to achieve failure. These methods
have also been studied by Martins et al. (l9gl) and
Tan and Donald (1985).
The
"enhanced
limit" slope stability methods are
based on stresses calculated using a finite element
analysis and combined with a limit equilibrium noe
ofanalysis along a prescribed slip surface, to d"fin"
the factor ofsafety. The prescribed slip surface is the
one defined by the lowest factor of safety and is
found using a trial and error procedure. The stresses
along the slip surface are computed using a finite
element analysis and can either be usea in "
"strength"
method or a
"stress-level"
method. Farias
and Naylor (1996) stated that when using the
"di-
rect" finite element method it is,
,'not
easv to obtain
a safety
factor
accttrate to within the'confidence
Iimits achievable by linit equilibrium methods". The
authors noted that: l) a
fine
mesh is required,Z) a
code capable of giving reliable results with ihe
Mohr Coulomb elasto-plastic model
for
toading
states close tofailure is needed, and 3) il is tuually
necessary to carry out a set of analyses with c, and
tanQ'progressively reduced by a
factor
which will
become the safety
factor
when
failure
is eventually
reached.'Enhanced limit" methods require only one
finite element analysis to calculate factors ofsafety
for a slope with various combinations of c, and
tan6'.
33
Adilori and Cumrnins (1985) produccd a finite
element method that combine the
"sfength"
and the
nstress-level"
methods as defined by Kulhawy
(1969) and ZierJriewicz et al. (1975), rEspectively
(Fig. 2). The Adikari and Cummins (1985) method
prodtrccd facton of safety that were between the
values obtained when applyng the Kulhawy (1969)
and the Ziern/ri*vicz et al. (1975) methods. It was
noted that for near-failure conditions (i.e., as defined
by Bishop's Sirnplified method, 1955), the value of
the factor of safety calculated by the Adikari and
Cummins (1985) method approached 1.0, while the
value ofthe factor ofsafety calculated by the Zien-
kiewicz et al. (1975) method remained high. The
factor of safety by the Kulhawy (1969) method also
approached unity with the factor ofsafety being de-
pendent on the percentage of the sftngth mobiliza-
tion in the component materials. The main differ-
ence in results appean related to using the stresses
on the principal plane (Zienkiewicz
et al. 1975)
rather than on the plane. By definition, failure does
not occur on the plane of principal stress and there-
forc, the Zier*iewicz et al. (1975) method (or any
stressJevel method) is computing a factor of safety
that must be higher than the factors of safety pro-
duced by a
"strength"
method.
Duncan et al. (1996) provided a summary of the
limit equilibrium and finite element methods that
have been proposed for slope stability analyses.
3 SUGCESTED STTIDY FOR COMPARISON
BETWEEN TTIE FINITE ELEMENT AND THE
LMIT EQUILIBRJI.'M METHODS OF SLOPE
STABILITY ANALYSIS
The frnite element slope stability method proposed
in this paper is of the
"enhanced
limit strength" tlpe
(Scoular, 1997). The finite element method uses the
Kulhawy (1969) definition for the factor of safety
combined with a finite element stress analysis of the
slope. Stress analyses werc done using Poisson's ra-
tios equal to 0.33 and 0.48. For each stress analysis,
the cohesion and the angle of internal friction of the
soil were altered as the stability of the slope was
computed The selected values for cohesion, c', were
10, 20 and 40 kPa, and for the angle of intemal fric-
tion,
Q',
were 10, 20 and 30 degrees.
The finite clement slope stability method pro.
duces an overall factor ofsafety that is an expression
of the sability of the slope based on the calculated
stresses within the slope. Slope stability problems
solved using the finite element method have two im-
portant distinctions from limit equilibrium methods.
First, the finite element slope stability equation is
determinate; therefore, no further assumptions arc
rcguircd to complete the calculations. Second, the
factor of safety equation is linear, because the nor-
mal stress at the base'of a slice is known. On the
other hand, limit equilibrium methods, sarting with
Bishop's Simplified method (1955), have.uscd an
estimated factor of safety when computing the nor-
mal force at the base ofa slice. The final factor of
safety is found through an ilerative procss. The fi-
nite element method factor of safety is defined using
the normal and shear stresses computed using a fi-
nite element analysis.
Finite element numerical stress analyses have
been available for many years. The finite element
method however, has not bccome popular for slope
stability studies due to intense computational rc-
quirements and difticulties in assessing the stress
versus strain characteristics ofthe soils. In addition,
inexpensive and easy to use limit equilibrium meth-
ods have provided factors of safety that appear to
reprsent failure conditions in the field in most
situations. Microcomputers now have sufficient
computational capacity to perform combined stress
and limit equilibrium analyses. As a result, it is an-
ticipated that the latter procedure will become more
cornmon in engineering practice.
3.1 Procedure usedfor thefinite element analysis
The enhanced limit (strength) finite element method
proposed by Kulhawy (1969) was selected as the
most appropriate method for slope stability analysis.
The finite element stress-deformation software.
Sigma/W (a proprietary product of Geo-Slope lnter-
national Ltd., Calgary, Albena, Canada), was modi-
fied to utilizes a search algorithm in order to assign
and transfer calculated finite element stresses to a
designed point on the slip surface (Bathe, 1982;
Krahn et al., 1996). The calculated finite element
calculated stresses are used to calculate the normal
and shear stresses on the slip surface. The latter
strcsses are used to calculate local facton of safety
at the center ofthe base ofeach slice as well as the
overall factor ofsafety for the entire slip surface.
3.2 DeJinition offacor of safety
The overall factor ofsafety is defined in accordance
with the finite element slope stability method dc-
scribed by Kulhawy (1969), and expressed as the ra-
tio of the sum of the incremental resisting shear
strcngths, .S7, to the sum of the mobilized shear
forces, Sn', along the slip surface.
- ES.
Frrr"r
=F.
(l )
L " f r
The resisting force for each slice is calculated in
terms of the shear strength, a, at the center of a slice
multiplied by the base length of the slice,
f.
T}rc
available resisting shear strength for a satu-
rated/rursaturated soil (Fredlund
and Rahardjo, 1993)
can be written as:
34
t
s
t
I
I
N1 ( - 1 , 1 ) Nr ( 1, 1)
Global Coordinates (x,y)
x
-
Coordinate
Figurc 3. Definition ofthe global and local coordinates for a rectangular finite element.
tr
o
o
E
o
o
;
g, =sp={c' +(oo -u"
)tanl ' +(u"
-u*)tar/b
1p
(2)
The mobilized shear force, S,,, for each slice is
calculated as the mobilized shear stress, 2., at the
center of a slice multiplied by the base lenglh,
B.
S^
=
t ^f ( 3)
The local factor ofsafety is defined as the ratio of
the resisting shear force, ,S7, at a point along the slip
surface divided by the mobilized shear force, .Sr, at
the same point,
S, r B
t *
=L
=; 7
The resisting shear force, .S7, and the mobilized
shear force, S,,, tr both calculated using the
stresses computed in the finite element analysis. The
normal stess, on, frd shear stress, rv1; @r! b
'im-
ported"
as known values to the limit equilibrium
analysis and the definition of both the overall and
local factor ofsafety equations are linear.
3.3 Element identiJication corresponding to the
base of a slice
Each element
must be checked to confirm that thc
ccnter
of the base of the slice is located within the
elcment
under consideration. Then the stresses cal-
culated
by the finite element analysis can be
"im-
ported"
into the stability analysis. Oncc the element
cmbracing
the center of a portion along the slip sur-
tace ts located,
stress values from the Gauss points
of the element can be transferrcd to the nodes bf the
clcment
and consequently to the center ofthe base.
t ne procedure
is in accordance with the mcthod de-
scribcd
by Bathe (19g2).
A common set of coordinates is used to identi$
the center of a slice along a slip surface with respect
to the surrounding finite element. The global coordi-
nates for the center of the base are calculated in or-
der to determine the location of the base center
within the slope, and to determine which element is
associated with the center ofthe base. The local co-
ordinates ofthe center ofthe base are then calculatcd
within the element that encompasses the center of
the base (Fig. 3).
The global coordinates for the center ofthe base
of a slice are related to the global coordinates of the
finite element nodal poins through use of the shape
fimctions.
x =<N>{ X} ( 5)
Where x
=
global x coordinates for the center
ofthe base ofa slice;y
=
globaly coordinates for the
center ofthe base ofa slice;
{X}
=
global x coordi-
nates for the element nodal points;
{Y}
=
global y
coordinates for the element nodal points; and
<y'{> :
matrix of shape firnctions.
The shape functions <ly'>
are defined in terms of
.
the local coordinates (e s). Since the global coordi-
nates for the center of the base of a slice and the
nodes are known, the local coordinates can be ob-
tained by solving Equations (5) and (6), simultane-
ously. The shape finctions for a rectangular finite
element with fou nodes are as follows (Bathe 1982):
N,
=11t +r 1t +s ;
N ,
=
i ( l
- r X ,
+ s )
(4)
(7)
Local Coordinetes (r,s)
Finite Element
x
-
Coordinate
35
(8)
wher / and s
=
local coordinates within the element.
The local coordinates vary between
-l
and +l (Fig.
3). A knowledge of the local coordinates is crucial to
identifuing the element overlapping the center of the
base of a slice. By definition, an element surrounds
the center of the base of a slice if the followine con-
ditions are met:
For a triangular element,
( 0<r >l ) and( 0<s >l ) ( l l )
For a rectangular element,
( - l <r ) l ) and( - l
<s>l ) ( 12)
The center of the base is outside an element if the
local coordinates are not within the above specified
ranges. The search continues until an element is
found that satisfies these conditions.
3.4 Transfer of element stresses to the center of the
base ofa slice
Calculated shesses are stored within the computer
software relative to the Gauss points of an element.
SEesses must be transferred from the Gauss points of
an element to the nodes of the element and then to
the center ofthe base ofa slice.
The local coordinates of a point within a finite
element are defined in relationship to the global co-
ordinates at the nodes of the element by using the
shape functions, as per Equations (5) and (6):
x =
1 N r
N 2 N 3 N 4
>
(14)
where r and y
=
global coordinate positions within
the element that are known as the center of base of a
slice (Fig. 4); X and
y:
global coordinate at the ele-
ment nodes; and lf7, N2, N j and N4
=
the shape firnc-
tions defined in Equations 7 to 10.
The stresses from a finite element analysis are
stored at the Gauss points. The shape firnctions can
be used to describe the change of a variable within an
element in terms of nodal values. The finite element
slope stability calculations require that stresses at the
center of the base for each slice be within an ele-
ment. This is achieved using the following proce-
dure:
{ o} , =<N>{ F}
( 15)
wherc aln
=
stresses at the element node; <|y'> =
matrix of the shape functions; md
{F}
=
stress values
at the Gauss points.
The local Gauss point integration coordinates arc
(0.577,0.577), however, when the local Gauss point
integration coordinates are projected outward to the
element nodes, the local coordinates bccome
(1.7320, 1.7320)
Gig.
5). This projection is carried
out for each element and the values for the stresscs
from each contributing element are averaged at each
node. Accordingly, the values of or, o, and to c,an
be computed at each node of the finite element mesh.
The nodal stresses, ar, ay, and lryy, of an element are
transferred to the center ofthe base ofa slice along
the slip surface.
{ o} =<
N >{ o} , ( 16)
where
/ol
=
stresses at the center of the base of a
slice.
The stresses, a,, ay, and rry, can now be computed
at the center ofthe base for each slice.
3.5 The normal and shear stresses at the center ofa
slice
Once the stresses, o\, oy rnd r- are known at the
center of the base for each slice, the normal stress,
on, and the mobilized shear stess, r,
,
can be cd-
culated using Equations (17) and (18), respectively
(Higdon etaL 1976):
6r * ou or ' ouc os 20
On
=- - - - - - +t r yst nLf l
(r7)
o-
-o..
sin20
rm
=
tocos2Q (18)
where o,= total stress in the x-direction at the centcr
ofthe base;o"
=
total stress in the y-direction at the
center ofthe base; r,n
=
shear stress in the x- and y-
direction at the center of the base; and d
=
angle
measured from the positive x-axis to the line of ap.
plication ofthe nonnal stress.
The above steps provide the necessary information
required to calculate the stability of a slope using the
finite element stresses. The calculated values for the
normal stress, o,r, and the mobilized shear stress, r-,
at the center of the base of a slice are entered into
Equations (2) and (3) to give the resisting shear force
N ,
=
i ( l
- r [ l - s )
N.
:
i
( I + r X I
-
s )
(e)
(10)
I x , )
t ; l
( - )
I
xt
I
t - l
l xt )
l Y, )
t - t
l Y, l
{ - }
t Y , l
t - l
l Y. )
(13)
36
/
Fictitious slice defined with
the Limit Equilibrium analYsis
x_____+.
x-Coordinate
Figure 4. Location ofthe center ofthe base along the slip surface within a particular finite element.
u.73m,1.73n1
t1.7320,1.7320)
tr.#o.sro o.s2,6.5r4
(-0.srt, 4.574
p.1n, 4.iln
(-1.7320, -1.73n) (1.7320, -r.7320)
Figure 5. Gauss point projections to the nodes ofa finite clcnent.
(strength)
and the mobilized shear force (actuating
shear), respectively.
The local factor ofsafety is computed as the ratio
of the resisting shear force to the mobilized shear
force. The overall factor ofsafety is the sum ofthe
shear force resistance values divided by the sum of
the actuating shear forces along the slip surface.
t
ID
o
o
o
I
O
EementNodes
+ Eement C'auss Pcints
4 PARAMETRIC STI.JDIES ON A SIMPLE 2:I
SLOPE
A slope at 2 horizontal to I vertical is analyzed for 4
conditions (Scoular, 1997). The first case is a free-
standing slope with zero pore-water pressures and
the slope is referred to as a dry slope
@ig.
6). The
second case is a free-standing slope with a pie-
zometric line at thrce quaders of the slope height,
and the slope is referred to as a wet slope (Fig. 6).
(x, y) known
(r, s) unknown
- Center of the base of a slice (x, y)
x
-
Cmrdinate
37
Piezometric Line
I
t l
E
o
o
E
o
o
I
u 4u
60
EO 1OO
x
-
Coordinate (m)
Figure 6. Serected 2:l free-standing slopc with a piczomctic
rine exiting at the toe ofthe srope.
Pr titd
case is a slope partially
submerged
in wa-
r-er
Tm
zero pore_water pressures
in the slope
(re_
{e1-ed
to as dry) (Fig.
D.
The fourth "*" ir'" o*_
ltally
submerged slope with a piezometric
line at one
halt of the.slope height (refened
to as wet) (Fig.7).
'lhe
partially submerged slope is covercd with w1rcr
to one halfofthe slope height, providing
support for
me stope and tncreasing the factors of safety. The
cohesion of the soil was varied from l0 to 4O kp"
and the angle of internal friction was varied from l0
to 30 degrees for each slope t1pe.
4.1 Limit equilibrium analysis
The limit equilibrium analyses are performed
using
9" F :tt_Limit
Equilibrium
methoa (CG),
(Fredlund
& kahn 1977) which provides
" ,orn_
bined moment and force equilibrium
solution.
L
empirical
finite element interslice force function,
911e9
on an independent stess analysis (Fan
et al.
1986) was ued. The General Limit Equilibrium
pethod along with a finite element intersiice force
fimction provides a method of comparison
betureen
the finite element based analysis "nd th" limit equi_
librium analysis.
4.2 Finite element stress analysis
The finite element shess analysis was performed
by
"switching-on"
gravity for the fr"e-rtaodine
,LD,
and for the partially submerged slope. The toaa Lf
the water and the lateral support it provides
to the
sl.ope is simulated.by point loads equal to the weight
ot.water on the slope. The analyses are performld
usrng
potsson's
ratios of 0.33 and 0.4g, and a
Io*g't
modulus equal to 20,000 and ZOO,TiOO tpa.
lne results showed that the stesses change with a
"l"rg-g poisson's
ratio, but ar" "orrit*t fo,
changes in the Young's modulus. fhis observation is
consrstent with the observations of Matos (19g2).
s_RESI.JLTS
OF TI{E FIMTE ELEMENT SLOPE
STABILITYMETHOD
Tl"
l*ql factors of safety differs along the overall
slip surface (Fig. 8). Ioial factors oi-r"f"r" r"r"
*Tpl1t9d for a2:l (dry) slope with a cohesioi "quaf
ro 4u t(pa an<l an angle of intemal
friction equai to
30 degrees. While thi tocal facton-oi;;fira
fong
the slip surface, the overall n rit.- lf"rn"*
facton of safeg fall within the .-g" "i t " iir,rit
equilibrium factors of safety. The ?itr";;"-;_
tween the local factors of safety for
poisson;s
Ltio,
of 0.33 and 0.48, calculated using the fi"il;t";;
l1e9 f.*n-p*
in Figure.8.
m" r""toroir"f"ty
TTputed
by rhe limit equilibrium
method and the
finite element."Fd
appear to U" u.ry,i.ifrr.
fi"
resuts appear to be within-_the
limits of unccrtainty
associatcd with slope stability
calculations.
ilc i_
nite elemcnt method incorporates
the stress-;train
characteristics of the soil when computins
th;;h;
strength and actuating shear force oitne-rol io G
calculation ofthe factor ofsafety (Fig.
9).
38
Crest
\ l
\-
---
Water l
I
I r oe
60
x
-
Coordinate (m)
Figure 7. Selected 2: I partially submerged slope with a horizontal piezometric line at mid-slope.
80
E
o
G
E 4 0
o
o
I
1(x) 20
The factor of safety results computed using the
finite element method (i.e., F3 corresponding to a
Poisson's ratio of 0.33, F4 conesponding to a Pois-
son's ratio of 0.48) are compared to the factors of
safety computed using the limit equilibrium method
(GLE) and are shown in Tables I and 2. To assess
the variations in the factor of safety by each method
ofanalysis, the results are grouped according to co-
hesion and angle of intemal friction. The factors of
safety grouped according to cohesion, c', are plotted
versus the stability number,
l(7lftan{')/cf,
(Janbu'
1954). The factors of safety grouped according to
the angle of intemal friction,
/',
are plotted versus
the stability coefficient, (c/7II) (Taylor, 1937),
where
7
is the unit weight of the soil, I/ is the height
of the slope,
l'is
the angle of intemal friction, and c'
is the cohesion.
The factors of safety are grouped according to the
soil parameters and plotted versus the stability num-
ber and the stability coefficient. The greatest differ-
ence in factors ofsafety is noticed at high angles of
intemal friction, at low values of cohesion and at the
maximum values of Poisson's ratio.
The factors of safety for the (dry) free-standing
slope, when grouped according to cohesion and
plotted venus the stability number (Fig. l0) show a
S
loo
ft
zso
h
$
zoo
.Eo
rso
O
E
100
E
F s o
oo
X o
20 30 40 50
x+oordinate
(m)
Figure 8. Prescntation ofthe local and global faaors ofsafety for a 2:l dry slope.
Poi ssonRati o,P=0.33
39
Table l. 2:l fiee-standing slope
Soil
c '
kPa
20 l 0
l 0 20
40 l 0
20 20
l 0 30
40 20
20 30
Lt 3l
1.260
t . 370
1. 6t 5
1. 794
t.892
o.677
o.782
r.021
t.102
I . J J )
Wet
F3
o.634
0. 745
0.930
t . 077
1.260
0.647
0.7s5
0.953
Parameters
0'
degree
GLE
Finite
element
Dry
F3
tt= 0.33
F4
t t =0. 48 l i ni re p=0. 33 p=0. 48
element
0.882 0.867 o.874
1. 125
1.230
1. 352
1. 639
1. 765
1. 884
l . l 5 l
1.239
t.696
1. 775
1. 918
1.239
0.995
1. 368 l n?r .368
0.969 0:988
>'
, O
rA
o
qr
x-coordinate
Figurc 9. Shear strength and shear fotce for a 2:l dry slope calculated tsing the finire element method.
Figure 10. Facors ofsafcty versrs stability nurnbcr for a 2:l dry slope as a fimction ofcohesion.
2. 5
2. O
.
o
3
1. 5
o
t
r . o
6
&
0. 5
0. 0
Ical F,
1P
=
6.33;
Global Facton ofSafcy
Bi$op 2.3@
Janbu 2.173
GLE (F.8. firnctiar) 2.356
4
(P
=0.33)
23ap
F, (r
=
O.+e; 2.339
frinary 2226
, mbuMel ho4Ft =2. 173
+ FI ( GLE)
+ Fr ( F =
0. 33)
-
Fs (P =
0.48)
40
Ta,ble 2. 2:l partially submerged slope
Soil
c
kPa
Parameters Dry
6' GLE
degree Finiteelement
firnction
Wet
F3 F4 GLE
p= 0.33 p= 0.48 Finite
element
fimction
F3 F4
p= 0. 33 p= 0. 48
l 0 t 0 0. E45
20 l 0 1. t 49
l 0 20 1. 344
0.843 0.827 0.649
1. 1 t 5 1. 085 0. 886
1.425 1.422 1.050
0.635 0.64r
0.874 0.880
1.046 1.06E
t . 3t 4 1. 343
1. 296 1. 316
t . 505 r. 530
1. 774 t . 795
1. 763 1. 786
20 20 1. 618
1. 586 1. s75 l . 3l E
40 l0 t.721
l 0 30 1. 865
40 20 2.297
20 30 2.337
1. 722 1. 691 I at',
2.081 n.s.&* 1.482
2.3E5 2.368 1.E00
2.268 2.204 1.783
qO
30 I.OOA
2.970 2.899 2.303 2.260 2.274
rn.s.a.:
no solution achieved
0.00 0.02 0.04 0.06 0.08 0.10 0. 12
Stability Coeficent,
[d(
yH)l
t;igure I l. Factor ofsafety versus stability coefficient for a 2:l dry slope as firnaion ofangle ofintemat friction.
Stability Number,
[(
yHtang')/d]
ligurc 12. Factor ofsafety versus stability number as a imction ofcohesion for a2'l slope with the piczomeric line at % ofthe
.l op hci ght.
2.5
)
1. 5
(,
(o
a
: 1. 0
o
E
o
LL
0.5
0.0
2.0
1. 6
b
*
1 2
an
o
E o.a
6
l!
0.4
25 20 15 10
41
2. 0
1. 6
.)
o
g 1 . 2
o

o. a
o
lr
0. 00
slight divergence in the factors of safety when the
cohesion approaches l0 kPa and the angle of inter-
nal friction approach 30 degrees.
The factors of safety by the finite element
method, with a high Poisson's ratio, is greater than
the General Limit Equilibrium solution. The slight
divergence is evident when the factors ofsafety are
grouped according to the angle of intemal friction
and plotted versus the stability coefficient (Fig. I l).
It is also evident that at high values of cohesion,
(i.e., c'equal to 40 kPa), The factors of safety com-
puted when using the General Limit Equilibrium
method are greater than those from the finite ele-
ment methods with either Poisson's ratio value.
The factors of safety for the (wet) free-standing
slope with a piezometric line at three quarters of the
slope height, are grotrped according to the cohesion
and plotted versus the stability number (Fig. l2).
The results show a slight divergence bet*,een the fi-
nite element factors of safety and the General Limit
Equilibrium facton of safety when the cohesion is
40 ud20 kPa. The difference between the factors of
safety by both methods is constant at all values of
cohesion until the angle of intemal friction becomes
ggual to_30 degrees and cohesion becomes equal to
l 0kPa(Fi g. l 3).
The grouping ofthe factors ofsafety accordins to
the angle of intemal friction, ploned venus the ita-
bility cocfficient (Fig. l5), shows the sarne pattem
as for the (dry) free-standing slope (Fig.l0). The dif-
ferences in the results ar more pronounced
as the
cohesion become less than l0 kPa.
The factors of safety for the partiallv
submersed
slope with a piezometric line at one haliof the slipe
h.eight were grouped by cohesion and plotted vetsus
the stability number (Fig. 16). The results show
close agreement between the General Limit Equilib
rium method and the finifs
sts6sal mettrod. The
0. 02
o. ' t 2 0. 10
Stabitity Coefficent,
Ic'
t(rH ll
Figure 13. Factor of safety vems stability cocfEcient as a filrction of the angle of intemal friction for a 2:l slope with the pie-
zometric line at % ofthe slope height.
0 5 1 0 1 5 2 0 2 5
Stability Numbcr,
[(tH
tan0,)/c,l
Figure 14. Factor of safety vctsus stability numbcr as a fimction of cohesion for a 2: I dry stope /, submerged with water.
3. 5
3. 0
b
2-5
o
3 2. 0
o
E
r . s
o
G
q
i . o
0. 5
0. 0
42
.>
2.5
.g
3 2. 0
o
b i . 5
tt
o
*
1. 0
2.5
2.0
8 r r
6
' v
U)
t
1. 0
6
L
0.00
0.04 0.06 0.08
Stabiflty Coefficent,
le
l1H )l
Figure I 5. Factor of safety versus stability coeffcient as a fimction of internal friction for a 2: I dry slope % submerged in water.
Stability Number,
I lHtan{')/c'l
Figure 16. Factor on safety versus stability numbr as a firnction ofcohesion for a 2:t slope halfsubmerged with a horizontal pie-
zometric line.
0. 12 0. 10
0.5
same pattem of divergence is evident as was shown
for the dry soil slope which is partially submerged
(Fig. 1a). However, the divergence is not quite as
extensive. The same conrments apply to the factor of
safety versus the stability coefficient as shown in
Figure 17.
Plotting the factors of safety for the various slope
conditions, (i.e., dry free-standing, wet free-standing
and dry partially submerged), versus stabilify num-
ber on Figure 18, shows the ranking of slopes by
factors of safety. The factor of safety can be esti-
mated for a slope that is similar to one of these cases
by calculating the stability number and selecting the
appropriate value of cohesion and angle of intemal
friction.
Both the General Limit Equilibrium method and the
finite element method of slope stability produce
factors of safety that are in close agreement. The ad-
vantage of the finite
element method is that the
stress-stain characteristics ofthe soil are used to de-
termine the stress state in the slope. If the limit equi-
librium and finite element factors of safety are
similar for a simple slope than results from the two
methods can be interpreted in similar manners. This
shrdy then sets the stage for using the finite element
method for situations where the limit equilibrium
methods is known to not yield satisfactory results.
The finite element method also produces graphs of
the local factors of safety that can be combined with
the shear strenglh-actuating shear force plots to help
explain the best support mechanism for the slope.
The close agreement between the factors of safety
when using the limit equilibrium method or the fi-
nite element method, has historically favored the use
of limit equilibrium methods. Examination of the
critical slip surfaces reveals that while the factors of
safety values are close, the location of the critical
slip surfaces may be different.
+
FI (GLE)
+
Fr ( P
=0' 33)
-
F! (p =
0.48)
F3 (GLE)
Fr (P =
0'3!l)
Fc (F =
0.48)
43
i
o
- q 1. 5
o
t
r . o
6
lt
0. 5
0. 0
0.04 0. 00
0. 08 0. 10 o. 12
3. 0
2.5
2.0
1. 5
1. 0
g
o
U)
o
o
()
o
lJ-
Stability Coefficent
, Ic.t( rH )l
Figure 17' Factor ofsafety versus sability coefficient
as a fimction
ofthc angle ofintemal friction for a 2:l slope halfsubmerge.
with a horizontal piezometric line.
Stability Number,
[(
y H tang,
ltc,l
Figure 18' Factor of safety vcrsrs sability nunber as a ftrndion of cohesion fot a 2:l slope, evaluated for dry, piezometic
and
submerged conditions.
6 ANALYSIS FOR TI{E LOCATION
OF TTIE
CRITICAL SLIP SI.JRFACE
The location of the critical circle changes depend.ing
on the sitrution
Fiog
qotfr"{. The biggesrchange
in location of critical slip surface was-experiencJd
for the (wet) free-standing sloae (Figs. 19 and 20)
and the (wet) supported slope (Figs. 2l and,22\.
In general, the finite element method slip surfaces
go deeper than the limit equilibrium slip surfaces for
the (wet) free-standing slope. The partially
sub.
merged slopes show that the limit equilibrium
slip
surfaces go deeper than the finite element method
slip surfaces. For the free- standing slope, the frnite
element method with a Poisson's ratio equal to 0.4g
showed the deepest slip surface. For the partialll.
ylmerged slope, the finite element method with a
Poisson's ratio equal to 0.4g, showed " .o*ia"r"Uty
shallower slip surface.
7 CONCLUSION
The finite element method of slope stabiliw is a vi-
able method of analysis that is now availabie for en_
gingenng
.practice.
The use of the finite "lem.nt
method yelds morc detailed information
on tt.
stress strte in the soil than is available from .onln"n-
tional limit equilibrium methods.
This information
can assists.engineers in the design of slopes and
slope retaining structures.
--r---
F" (GLE)
+
Fr ( t t =
0. 33)
--
Fr (F =
0.4E)
3/4 Piezometnc, C
=
10 kpa
4
Cohesion =
40 kPa,
Method
GLE (EE. Function)
F" (p =
0.3t)
Fg (p =
0.48)
0'=
30", Piezometric
litr- 3t4 of the ralay up the slope
X
y
R FactorofSafety
58.50
56.00 37.88
,t.741
57.50 49.50 34.69 1.627
57.50
53.00 37.83 1.661
Fs (p =
0.33)
Fs (p
=
0.48)
6
-?
50
.g
P
? ? 4 0
d
i 3 0
E
-?
40
o
6
.g
p
6 3 0
o
o
I
10 20 30 40 50 60 70 80 90 loo
.t10
x
-
Coordinate (m)
Figure 19. Location of the critical slip surface for a slope with a piczometric line where the soil properties are c,
=
40Wa and ('=
300.
10 20 30 40 50 60 70 80 90 1oo 110
x
-
Coordi nate (m)
Figure 20' Location of the critical slip surface for a slope with a piezometric
line where the factors of safety are closest to I .0.
The value of Poisson's ratio can affect the calcu- the finite element method to slope stability prob,
lation of the factor
of safety as well as the location lems, a u"tt"t *a"ot"nding is required regarding
of the slip surface.
with an
jncreasing
application of oe etrect oiFoi.*nt otio and the overall deforma-
tion model on the stability of slopes.
45
Coh.esion =
40 kPa,
0,= 3Oo,
piezometric
line 3/4 of lhe way up lhe slope
Met hod x
y
R Fj ct orof s; f ei y
GLE (F. E. Funct i on) 63. 50 S9. OO 39. 56 1. 102
F, (p =
0. 33)
63. 00 S9. OO 41. 54 1. 076
F. (p =
0. 48) 61. 50 S9. SO 42. gg 1. 100
Fs ( F =
0. 48)
Fs (rr =
0. 33)
GLE (F. E. t unct i on)
Cohesion
=
40 kPa,
0'=
30o, half submeqed slope
Method /\ Y R FaciorofSafety
GLE (F.E. Function) 58.fi) 58.50 40.20 2.303
F, (p =
0.33) 52.50 50.50 31.76 2.259
F. (p =
0.48) 51.50 51.50 30.94 2.273
E
o
G
c
E
o
o
()
I
10 20 30 40 50 60 70 80 90 100 110
x
-
Coordinate (m)
Figure 2l ' Location ofthe critical slip surface for a halfsubmerged slope where the soit properties are c,= 40kpa an d,
O'=
31,.
. 10
20 30 40 50 60 70 80 90 100 11 o
x
- Coor di nat e
( m)
Figute 22. Location of the critical slip srrface for a submerged slope where the factors of safety are closesr to 1.0.
The finite element stress.analy-sis provides
input boundary conditions are being used and that a rca-
information for the calculation of the stability of a sonable stressdeformation riodel is beine used.
slope. Further research must be undertaken on the With this arisurianc, soil structures can be dner de-
stress analysis in order to ensure that the proper
sigpedtoaccountforavarityofstressconditions.
E
o
r0
.c
E
o
o
o
I
Cohesi on
=
40 kPa,
0, =
30o,
Met hod X
GLE ( F. E. Funct i on) 62. 50
hal f submerged st op-
F, ( p =
0. 33) 54. 00
F. ( p =
0. 48) 53. 00
Y R Fact or of Saf et y
63. 50 44. 11 1. 050
5 1 . 5 0 3 1 . 9 7 1 . 0 4 6
56. 50 34. 69- 1. 068
GLE ( F. E. f unct i on)
Fs ( t r =
0. 48)
Fr ( P =
0. 33)
46
ACKNOWLEDGEMENTS
Thc authors want to acknowledge the initial dis-
cussions regarding the potential for using a finite
clemcnt slope stability method rhat were held with
hofl Wong lki Sin of Nanyang Technological
Uni-
venity, Singapore. These discussions formed the ba-
sis for the study of this topic. The assistance of Dr.
Fangshcng Shuai, Ms. Noshin Tadenadeh and Ms.
Brigine Boldt-Lrppin in assembling this manuscript
is also acknowledged. The authors are also gnateful
to Gco-Slope lntemational, Calgary, for the modifi-
cations made to their software in order thst this
study could be readily performed.
REFERENCES
Adikari. G.S.N. & P.J. Cunmins I9E5. An ctfcctive stcss
slopc sability ana.lysis mcthod for dams.
proc.
llth Int.
Conl Soil Mech. Foand. Eng.2:713-718.
Bathc, KJ. 19E2. Finite clemcnt procedures in cnginecring
analysis: 200-233. hentice..Hall.
Bisbo,p, A.W. 1952. The stability of earth dams.
ph.D.
Thesis.
Univcrsity ofLondon
Bishop, A.W. 1955. The use-of the slip circle in the subility
uulysis of slopes. Ge otechnique 5(l
):
7
-1
7.
Bror.,g C. B. & I. P. King l!)62. Automatic embanfrmerr
amlysis cguilibrium and instabiliry conditions. Journal of
Soil Mechanics and Foandation Division, ISCE-93 (SM4i
2W-2r9
Duncan J.Ir,t 1996. Srarc-of-rhc.art: Srabitity and deformation
y!y_1is.__J_ournal of Georechnical Engineering, ASCE
122(7): 577-597.
Fao, K, D.G. Frdlund & G.W. Wilson 19g6. An intcrslicc
force fimction for limit equilibrium slope stabiliry "rof"rir.
Canadian Geotechnical Journal 23e\ig7 -Z%.'
Farias. M.M. & D.J. Naylor 1996. SaGi analysis using linitc
clements" Infogeo 94 Sa6
paulo.
Brazil.
Frcdlund. D.G. A J. Knhn 1977. Comparison of slope stability
methods of analysis. Canadion Giotechnique iq!),
qZS-
439-
Frcdlun4 D.G., J. Krahn & D.E.
pu&hl
198t. Thc rclationship
bctuccn timir cquilibriumslope sability methods. ir"". i7
I cnth
-Intemationol
Conference on Soil Mechonics ontt
_
t.oundatiorls
Engineerizg, Stockholm, Swedcn 3: 4@416.
t'tedfun4
D.G. & H. Ratrardjo 1993. Soit mechotis for un-
_ ..
sanraed soils. New
york:
John Witey & Sons.
Higdon a.. E.H. Ohlscn, W.B. Stiles , l'4,. Wecse a W.F. Ri_
W-1976. Mechanics of Materials. New
york:
John Wiley
& Sons.
Janbu. N. t954. Stabiliry analysis ofslopes with dimensionless
_-
paraneters.
Harvard Soil Mcchanics Sedcs (46).
Kondncr, RL.
-
I 963. HyAerbolic soess-strain refrnse: cohe-
sive soils Journal ofthe Soit Mechania and'Foundariorc
Division.
ASCE E9(SM!): I t5-t43.
Kratr\ J., L. Lrm & D.G. Frcdlund 1996. The use of finire
elunant computed porc-natcr pnessurcs
in a slope stability
arufysis. Landslides, Senneset (editor)
2: l27j:l2gl. Rot-
terdam: Balkema.
Kulhawy. F.H. 19,69. Finite element analysis of the behavior of
embankments. phD.
Thesis, the Universiry of Califomia. at
Berkley. Cslifomia. U,S.A.
La Rochellc. P. I 960. The short term sability of slopes in Lon-
don clay. Ph.D. Thesis. University oflondon, London UK.
t"cshchinsky, D. 1990. Slope sabitity analysis: Gcncralizod
l9gq9acy
L*,nal of Geotechnicot Engineering.
ASCE
I l 6(5): E5l -867.
Marrins, J.8., EB. Rcis & A.C. Maros I9gl. New mcthods of
arulysis for srability of slopcs.
proc.
lhth Int. ('onf
Soil
Mech. Found. Eng. 3: 463467.
Matos. A,C. 1982, The numcrical influcncc of thc
poisson
ntio
on the safay factor. Proceed@ of the 4th laennatiotul
Conference on Numerical Methds in Geo-Mechonis
l:
207-2tt.
Morgcnsrern. N.R. & V.E. Price t965. The analysis of rhe sa-
bifity ofgeneral slip surfaces. Geotechnique l5(l
): 79-93.
Nayfor, D.J. 1982. Finite elemcnts and slope srabiliry. Numeri-
col Methods in Geomechanics. D. Rcidcl
publisiring
Corn-
pany.
Rescndia D. 1972. Accuracy of cmbankmcnt deformations.
Proceedings, ASCE Speciolty Conference on
perlormonce
of Earth and Earth-Supported Structures,
purdul
Univer-
siry. West Lafayene. Indiana t2-14 June. l(part l):
gt7_
836.
Resendiz D. 1974. Accuacy of equitibrium slope sability
anzlysis. Journal ofthe Soil Mechaniq and Foundotion Di
virjoz. ASCE 100(GT8): 967-970.
Scoular. RE.G. 1997. Lirnit cquilibrium slope sability analysis
uing a suess analysis. M.Sc. Thesis, Ilniversii. of ial.-
lcot chewan, Sas kat oon, C on ad a.
Ts& C.P. & I.B. Donald 19E5. Finirc elananr calculation of
dam $abifity. Proc. I lth Int. Conf. Soil Mech. Found. Eng.,
San Francisco. Califomia 4: 2CAlr-2O44.
Taylor, D.W. 1937. Srabiliry of canh slopcs. Journal of rhe
Boston Sociery ofCivil Engineers )Q(lV(3): 337-3g6.
Wrigfit, S. G. 1969. A study of slope srability and the
undrained shear strength ofclay shalcs. Ph. D. thesis, tJni-
versity ofCalifornia ar Berrlley.
Wright,.S.G., F. Kulhany, & J.M. Duncan t973. Accuracy of
equrlrbnum sfope stabiliry analysis. ASCE. Journat of-Soil
Me c h anics Fou nd at ion Divis ion 99(SMl 0): 7
g3_79
l.
-
Zienkiewicz,O.C.,
C. Humpheson A nW. tewis f Sii. Asso_
crated and non-associatcd visco_ptasticity
and plasticitv
in
soil Mechanics. Gdotechnique ZSiq: All<Sg.
47