Beruflich Dokumente
Kultur Dokumente
Faculty of Cioil Engineering, Dept of Geotechnology, Technion-Israel Institute of Technology, Ha f a 32 000, Israel
SUMMARY
The current state of art for limit equilibrium analysis of slope stability problems lacks a satisfactory
procedure for stability evaluation under general, rapid (undrained) loading conditions. Some procedures are
available for the analysis of rapid drawdown, but these suffer from several shortcomings and, furthermore,
are not applicable to other types of rapid loading. An approach is presented which overcomes these
limitations. The approach integrates four components-establishment of soil behaviour on the basis of
laboratory testing, estimation of steady-state conditions in the slope using a boundary value analysis,
estimation of distribution of undrained strength in the slope using undrained stress paths, and identification
of the critical slip surface followed by calculation of its factor of safety. The approach is illustrated through
its application to the stability analysis of an earth dam under rapid drawdown and earthquake conditions.
1. INTRODUCTION
The purpose of this paper is to describe a methodology for analysis of slope stability problems
under undrained (rapid) loading conditions. We will use the terminology rapid and undrained
loading conditions as synonymous. Note, however, that this terminology does not imply the
inclusion of inertial effects in the analysis. The types of problems belonging to this class include
(among others) the pseudo-static evaluation of the stability of slopes subjected to earthquakes,
rapid drawdown of water reservoirs, rapid excavation or build-up of embankments and, in
general, any rapid application of loads to an existing earth structure. The motivation for treating
this subject was the realization that existing texts dealing with slope stability analysis' either -'
avoid it completely, or, at best, limit the presentation to the Bishop's6 treatment of rapid
drawdown of water reservoirs.
These problems possess the following characteristics:
(1) There exists an initial state in which the system is at equilibrium or at a steady state. In this
state every point in the soil has consolidated under an anisotropic state of stress, generally
under different principal effective stress ratios at each point.*
* Associate Professor.
Post Doctoral Fellow.
*The proposed procedure is applicable also to situations in which the initial state does not correspond to the end of
consolidation stage, provided only that the distribution of effective stresses at the time of the rapid loading is known.
However, for the purpose of the presentation, we will assume that at the initial state the process of consolidation has been
completed.
(2) Additional loads are applied rapidly (i.e. without drainage) to the system. The additional
loads induce different shear and normal stresses at every point, which in turn results in the
development of a spatial distribution of excess pore water pressure.
The difficulties in the analysis of this type of problem using conventional procedures can be
summarized as follows:
(1) In order to perform an effective stress type of analysis it is necessary to estimate the
distribution of excess pore water pressure generated during the stage of undrained loading. The
magnitude of the predicted pore water pressure which is generated by a non-hydrostatic state of
stress depends on the soils constitutive behaviour. It should be realized, however, that in most
engineering projects, there is insufficient experimental data for the formulation of a reliable and
well-substantiated constitutive model. Consequently, the reliability of the predicted pore water
pressure is rather low, resulting in a low reliability of the stability analysis. As an example of this
situation, let us consider some consequences of Bishops6 approach to the analysis of the rapid
drawdown case.
In Bishops method, the pore pressures which develop during rapid drawdown are estimated
using Skemptons pore pressure coefficients A and B. It can be verified that the procedure results
in the conclusion that, for the case of a clay embankment for which the B values of the saturated
soils are equal to 1 (reasonable for saturated soils), the pore pressure change 6 u ( x , y) at any point
below the upstream slope surface is simply given by
where yw is the unit weight of water, and Az is the change in water level as a result of drawdown.
The procedure suggests, therefore, that the pore pressure which develops during rapid drawdown
depends only on the geometry of the dam. In particular, it is in no way dependent on the nature of
the soil; i.e. the pore pressure which develops in dams of the same geometry, but made up of vastly
different soils, would be expected to be the same. This conclusion appears unacceptable,
contradicting a vast amount of experimental evidence which clearly shows that the pore pressure
generated during undrained loading depends on both the type of soil and its stress history (mainly
the overconsolidation ratio). This result illustrates the inherent difficulty in the application of
effective stress type of analysis to undrained stability problems.
( 2 ) An attempt to perform a total stress type of analysis using total strength parameters runs
into the difficulty that strength envelopes expressed in terms of total stresses are not unique. In
particular, the total strength envelope depends on the principal effective stresses at the end of
consolidation. Since each point in the slope may be consolidated under different principal
effective stresses, it is clearly not practical to establish, experimentally, a large number of total
strength envelopes which would correspond to the different end of consolidation stress conditions
in the field. This difficulty was recognized by Bishop and Bjerrum* causing them to conclude that
the use of the results of consolidated-undrained triaxial tests, expressed in terms of 4cucan be
justified in only few practical applications.
(3) The more commonly applied approach to total stress analysis, known as the d, = 0 analysis,
requires knowledge of the undrained strength of the soil throughout the slope. Use of the
approach is complicated by the fact that, as stated above, this strength depends on the
consolidation, or steady-state conditions, which vary throughout the slope. Consequently,
practical applications of this approach have usually been limited to cases of undrained loading in
which a constant average soil strength could be assumed throughout the soi!
Unfortunately, this type of assumption is rarely justified.
SLOPE STABILITY ANALYSIS FOR UNDRAINED LOADING CONDITIONS 17
The present paper presents a framework for performance of undrained stability analysis using a
total stress approach of the 4 = 0 type. As pointed out by Taylor,' this name for the analysis
method is a misnomer, since soils usually possess internal friction. The approach will be
henceforth referred to in this paper as the 'direct' approach, since strength values are specified
directly, rather than through strength parameters and normal stresses. Although frameworks of
this kind have been utilized previously, the approach presented here is believed to be more
consistent, overcoming some shortcomings of other methods, as discussed in the paper.
physical interpretation of the obtained results. For that reason it is convenient to review the
general structure of limit equilibrium slope stability analysis. This type of analysis is defined by
the following components:
(1) A failure mechanism, describing the shape of a potential (possible) sliding body, is
postulated. This body is bounded by the given soil surface and a member of a family of potential
slip surfaces y ( x ) .The nature of this family depends on the problem under consideration, and the
specific procedure which is utilized. In the more general formulations, the family of potential slip
surfaces is taken as a general polygon connecting the boundaries of the system under considera-
tion.
(2) It is assumed that the shear stress z [ y ( x ) ] along a potential slip surface, satisfies a limiting
relation of the form
tCY(x)l = SrnCY(X)l = SCy(x)l/F (2)
where S [ y ( x ) ] is the strength distribution along y ( x ) , F is a factor of safety with respect to
strength, and S,[y(x)] is the mobilized strength along y(x).
The relation z [ y ( x ) ] = S,[y(x)] implies that the procedure deals with the state of failure of a
fictitious material which is characterized by the reduced (mobilized) strength distribution
S,[y(x)] = S [ y ( x ) ] / F .The notion of fictitious materials, which was introduced by Baker and
Garber,14 may, at first sight, appear contrived; note, however, that such an artificial element is
implied in every limit equilibrium analysis. Consider, for example, the bearing capacity problem;
in that problem the procedure adjusts the value of an external load so as to ensure a state of
failure for each slip surface y(x).This adjusted load is entirely artificial, bearing no relation to the
actual footing load. In slope stability analysis, loads are taken at their actual values, and the
strength is adjusted in order to consider a system at a state of failure. The two approaches are
essentially equivalent, merely applying the adjustment to opposite sides of equation (2). This view
of limit equilibrium slope stability analysis shows clearly that t [ y ( x ) ] is not the shear stress
actually acting on the slip surface; it is the stress which would act there if the material had the
strength S,[y(x)] and the system was at a state of imminent failure. The distinction is clearer if
one again considers the bearing capacity problem. In that case, one would not confuse the shear
stress distribution induced by the actual footing load with that generated by the fictitious load
causing failure.
(3) The general form of limit equilibrium formulations is obtained from a combination of the
following elements:
(a) Equilibrium conditions for the potential sliding body.
(b) The limiting condition along the slip surface [equation (2)].
(c) Certain assumptions which characterize different limit equilibrium procedures. These
assumptions are usually expressed in terms of directions or location of the line of action of
the forces acting between the slices.
Combining these elements results in the establishment of two functions:
(i) A normal stress function u = u [ y ( x ) ] acting on the potential slip surface y(x). It is
emphasized again that u [ y ( x ) ] ,like z [ y ( x ) ] , does not represent actual stresses; these are
fictitious stresses acting in the fictitious material defined by the mobilized strength
distribution S,[ y ( x ) ] = S [ y ( x ) ] / F .The fictitious nature of limit equilibrium stress distribu-
tions should be evident from the observation that they depend on the factor of safety
associated with the particular slip surface under consideration.
(ii) A criterion function F = P[y(x),Data] which associates a factor of safety with each
SLOPE STABILITY ANALYSIS FOR UNDRAINED LOADING CONDITIONS 19
potential slip surface y(x). The term Data represents all the given quantities which
characterize a particular slope stability problem.
(4)The solution of a limit equilibrium problem is the process of identifying the critical slip
surface y,(x) which realizes the minimum value F , of the criterion function F[y(x), Data]. This
step can be written formally as
F, = min(F[y(x), Data]} = F[y,(x), Data] (3)
Y (x)
Equation (3) illustrates the fact that all limit equilibrium procedures lead to minimization
problems. It is important to realize that the value of F for an arbitrary slip surface is essentially
meaningless; only the minimal factor of safety F , provides a measure of the system (slope)
stability.
The above procedure is common to all types of limit equilibrium analysis. In what follows, we
discuss some of the implications resulting from this general framework.
where &(x, y) and &(x, y) are the spatial distributions of effective mobilized cohesion and angle
of internal friction, respectively, and u = u(x, y) represents the distribution of pore water pressure.
The mobilized parameters are defined as
,:c = c'/F (5)
4; = arctan[tan(+') /F] (6)
Figure 1 shows mobilized effective strength envelopes for a number of F values. On the basis of
this figure, it is clear that the following relation is valid:
c:, cot(@,) = ct cot(&) (7)
Considering, for each potential slip surface analysed, c:, and &, to be the strength parameters of
an artificial soil, it is seen that the system of fictitious materials considered in the slope stability
problem has a common value of {c~cot(&,)}which does not depend on the factor safety F.
Recall that limit equilibrium analysis of slope stability problems deals with a state of failure of
the fictitious material characterized by the strength parameters c:, and &. Therefore, for each
potential slip surface y(x), the procedure uniquely defines the complete state of stress at every
point along y(x), as shown in Figure 2. In particular, both the principal stresses and the principal
directions can be determined as shown in that figure. Note that the validity of this construction
depends on the notion of fictitious materials which are at a state of failure. Without this concept,
z [ y ( x ) ] and o[y(x)] do not define a unique state of stress, and it is impossible to draw the Mohr
circle. The possibility of determining the principal directions is a consequence of the fact that at
failure the angle 0, between y(x) and the major principal plane must, at every point, be 0 = (45
+ &/2)". In conventional limit equilibrium analysis of slope stability problems, the principal
+
directions are generally unknown. Consequently, the relation 0 = (45 &/2)" does not intro-
duce restrictions on the geometrical nature of the functions y(x), and it normally plays no part in
the analysis.
20 R. BAKER, S. F R Y D M A N A N D M. TALESNICK
c;cot $= c * cot @
Principal Plane
it is clear that S, represents the shear stress acting on a potential slip surface. On the other hand,
as was pointed out in the discussion of the effective stress approach, potential slip surfaces are
inclined at an angle of 8 = (45 + 432)"to the major principal plane. Consequently, the strength
function S,(X, y) is equal to the shear stress acting (in the fictitious material characterized by c;
and &), on planes which are inclined at (45 + &/2)" to the major principal plane. This condition
must be considered when making direct strength assignments on the basis of experimental data.
In effective stress analyses, this feature is automatically accounted for, and the principal
directions, if required, may be determined after completion of the analysis.
Note that in the effective stress formulation, the strength depends on the normal stress function
o[y(x)], which is established automatically by the procedure. In fact, each type of limit
equilibrium procedure will result in a different normal stress distribution and, therefore, will
utilize a different strength along the slip surface. Thus, the utilized strength depends on the
approximations (static assumptions) used in the derivation of the calculation procedure. Such
situation is obviously undesirable. In the direct formulation, on the other hand, the strength is
specified, and it does not depend on the nature of the calculation scheme. The limit equilibrium
procedure still establishes the normal stress distribution a[y(x)], but uses this distribution only in
the equilibrium equations, and not for the strength determination. Thus, this formulation makes
'fewer demands' on the formal limit equilibrium approach, placing more 'responsibility' on the
analyst. From this point of view, the direct formulation may be preferable to the effective stress
approach. This is particularly true for simplified procedures such as Fellenius or Simplified
Bishop, since the normal stress function implied by such procedures cannot be considered
reliable.
close to failure (see, for example, the case presented in the section dealing with applications).
Stress distribution calculated by a limit equilibrium procedure provides a reasonable approxima-
tion to the state of stress before the rapid loading only if the steady-state factor of safety
associated with the surface under consideration is close to unity. In general, however, such a
procedure is not consistent.
The inconsistent nature of this approach can be demonstrated by considering the two potential
slip surfaces passing along a common sector AB shown in Figure 3. Recall that the limit
equilibrium stress distributions, a[y(x)] and z[y(x)], obtained along any particular slip surface,
are functions of the factor of safety associated with that surface. As the factors of safety for the two
surfaces are, in general, different, the stresses calculated along AB
will be different for the two
surfaces. These stresses represent the condition existing at the same place but in two different,
fictitious, materials. Using these stresses for the purpose of estimating undrained strength results
with two different estimations of the undrained strength at the same point, i.e. the strength at any
particular point is not even uniquely defined.
As a result of this inconsistency, it is not possible to construct a unique spatial distribution of
undrained strength in the profile. Consequently, it is virtually impossible to solve the general limit
equilibrium minimization problem [equation (3)], and identify the critical slip surface which
yields the minimum value F , of the safety factor. Thus, at best, Lowe and Karafiaths procedure
may be utilized for the purpose of analysing an existing slope failure [when the function y(x) is
presumably knownJ, but not as a design tool.
Nevertheless, Lowe and Karafiaths basic idea of relating the undrained strength to the stress
condition before drawdown should be considered as a major contribution towards the develop-
ment of a rational analysis of undrained stability, and it is utilized in the present work.
4. UNDRAINED STRENGTH
The proposed procedure for the estimation of undrained strength is based on the fundamental
work of Rendulic15 and Henkel.I6 A typical example of Henkels results, which is based on a large
series of drained and consolidated-undrained triaxial test on Weald clay, is shown in Figure 4.
The drained tests were carried out with increasing and decreasing axial stress, increasing and
decreasing radial stress and constant mean normal stress. Most undrained tests were performed
following isotropic consolidation, but some followed anisotropic consolidation. The co-ordinates
F,f F,
180
-
-
... ..
water content contours rrom
drained tests /
160 -
120 -
0; ( p s i )
-
100
80
-
end o f consolidation
a; J2 (psi)
in the figure are a:-the axial effective stress and J 2 4 , where a: is the radial effective stress.
Dashed lines represent contours of equal water content obtained during drained tests, while the
solid lines represent effective stress paths in undrained tests with pore pressure measurement.
The experimental results shown in Figure 4 lead to the following observations:
(1) Contours of equal water content obtained during drained loading are consistent with the
effective stress paths during undrained tests (which are, by definition, constant-water-
content tests). Consequently, it is possible to consider both the solid and dashed lines in
Figure 4 as effective stress paths in undrained loading. Since the drained tests were
performed with both increasing and decreasing principal stresses, the identity of the two
types of paths implies that undrained paths are also valid for both loading and unloading.
(2) The system of undrained stress paths shown in Figure 4 is unique in the sense that if the
starting point of undrained shearing is on a particular stress path, then the test will follow
this stress path to failure regardless of how the initial point was reached. For example, if the
point A, in Figure 4, represents the effective stress conditions at the end of anisotropic
consolidation, then the undrained test will follow the path m,
where B represents the stress
conditions at failure. Consequently, the state of effective stresses at failure is a unique
function of the stress conditions at the end of the consolidation stage, and a system of
experimental undrained effective stress paths provides a graphical representation of this
function. More generally, it is indicated that if the effective stress state is known at the onset,
or at any stage of the undrained loading (not necessarily at the end of consolidation), then
the stress path up to failure is uniquely established.
24 R. BAKER. S. F R Y D M A N A N D M. TALESNICK
Henkell7 used this concept of the uniqueness of undrained stress paths to discuss the behaviour
of excavations in normally and overconsolidated clays, and it has since proven useful for studying
other undrained, field problems. The uniqueness implied by Henkels results is. however, limited
by two factors.
(1) All the tests were run in the triaxial apparatus with fixed principal directions in both the
drained and the undrained stages of the tests. Thus, the uniqueness of undrained effective
stress paths has not been demonstrated for loading conditions involving rotation of
principal directions. We will ignore this limitation, assuming uniqueness under general
loading conditions, essentially due to a lack of a better alternative.
(2) Henkels tests were all carried out on samples consolidated to the same preconsolidation
pressure. The uniqueness of the undrained effective stress paths shown in Figure 4 is,
consequentiy, valid only for a given value of the preconsolidation pressure pc. In order to
overcome this limitation, we use the that if all stress components are
normalized by the value of pc, a unique set of undrained effective stress paths results. The
normalized stress paths are independent of the preconsolidation pressure, depending only
on the over consolidation ratio. Figure 5 provides an example of this unique normalized
behaviour, based on simple shear measurements.20
As a result, at every point where the effective stresses and preconsolidation pressure
([oi,, o;,,, zxYlcand pc), are known, it is possible to utilize a family of normalized undrained
effective stress paths for the purpose of estimating the stress conditions at failure at that point.
- Legend
Y)
:8 0
i
n
.. 40
v l 0
0 100 200 30 0
Verlicol Stress Uy (kN/m
6Y
1
.
P
m 0.5
L)
b ) Normalized representation.
G O 4-
L
0
z
Normalized Vertical Stress Uy /CC
Figure 5. Normalized stress paths. (a) simple shear results; (b) normalized representation
SLOPE STABILITY ANALYSIS FOR UNDRAINED LOADING CONDITIONS 25
Repeating such a procedure at various points in the profile makes it possible to establish the
spatial distribution of undrained strength S(x, y) . It is noted that if the preconsolidation pressure
does not vary from point to point, then normalization is unnecessary, and the actual undrained
stress paths may be used.
The co-ordinates of the stress paths in Figures 4 and 5 are not convenient for the purpose of
stability analysis. For the present purpose, it is convenient to introduce the following alternative
representation of undrained stress paths.
Consider the system of effective stress Mohr circles shown in Figure 6(a). These circles represent
different stages during undrained shearing of a single sample in a conventional consol-
F= 1
Exuerimental results. I
II
0 A
F=l
Pole
0 0'
0 A 0'
Figure 6 . Modified representation of undrained effective stress paths: (a) experimental results; (b) the basic construction;
(c) stress paths for limiting equilibrium analysis
26 R. BAKER, S. F R Y D M A N A N D M. TALESNICK
stress path through T between the experimental paths ABC and ABC. Practically, the
point F is determined from the requirement that
BT - CF
TB FC
where notation of the type TB refers to the distance between the points T and B. This
procedure represents essentially a radial interpolation between the experimentally deter-
mined stress paths. Note that, by definition, the point T cannot fall to the right of the last
SLOPE STABILITY ANALYSIS FOR UNDRAINED LOADING CONDITIONS 27
to/
J
/--
A
stress path which corresponds to a normally consolidated state. If the point T is to the left of
the first experimental stress path, then equation (8) is replaced by
BT - CF
TO FO (9)
about the type, magnitude and spatial distribution of constitutive parameters. Experience has
shown, however, that in problems involving a low level of boundary displacement restraints (such
as the slope stability problem), the solution for stresses is relatively insensitive to the constitutive
model employed. As a result, even a crude specification of constitutive parameters is probably
sufficient for the purpose of estimating the stress distribution at steady state. This is definitely true
for vertical stresses which are almost entirely independent of the constitutive model, depending
mainly on the geometry of the problem. On the other hand, in this type of problem, displacements
obtained from the solution of boundary value problems are less reliable in the sense that they are
sensitive to the details of the constitutive model.*
Pore pressures developed in undrained loading are calculated from the requirement of zero
volume change. Hence, if the calculated displacements are unreliable, the estimation of excess
pore water pressure and undrained strength may also be questionable. In summary, it may be
concluded that estimation of total stresses at steady state from the solution of boundary value
problems is probably reasonable, but effective stress analysis of undrained loading should be
avoided except when high-quality constitutive information is available.
*The situation is reversed in problems involving a high level of geometrical restraint. In that case the solution for the
displacements is stable, and the stresses are sensitive to the details of the constitutive model employed.
SLOPE STABILITY ANALYSIS FOR UNDRAINED LOADING CONDITIONS 29
Distibutions of
stre'is paths and distibutution of
preconsolidationpressure c', 6 ,
Y,. E,v. k
Solution of Characterization
boundary of the steady
value problems
Eflective stress
distribiition at
steady state
1
Critical slip surface and minimal factor of safety
for t h e rapid loading problem
The individual elements of this proposed procedure are not new. The present combination of
these elements is, however, considered optimal in the following sense:
(a) Boundary value problems are used for the estimation of total stresses and pore water
pressures at steady state. These variables are the least sensitive to the details of constitutive
modelling, yielding, therefore, reasonably reliable results.
(b) The set of experimental effective undrained stress paths constitutes an explicit representa-
tion of the relevant constitutive nature of the material. The proposed procedure uses this
constitutive information directly, without the corruption which results from incorporation
of such information into the framework of some formal continuum theory.
(c) The limit equilibrium procedure is used in order to obtain a single global measure of the
system's stability.
Each one of these elements is used for the purpose for which it is most suitable, so the combined
procedure may be considered optimal.
Some general characteristics of the proposed method of analysis are discussed with reference to
Figure 9. The upper diagram in this figure illustrates a schematic slope stability problem; the
30 R. BAKER, S. F R Y D M A N A N D M. TALESNICK
0
I 'I I I I L
0'
curve y = y(x) is an arbitrary slip surface; a and b are two arbitrary points on y(x). The lines
ycss(x)and ycud(x)are the critical surfaces for the steady state and undrained loading, respectively.
In general, these two surfaces will be different. The lower diagram shows the states of stress at a
and b. The proposed procedure consists of the following elements:
Steady-state effective stresses along y ( x ) are established by solving the appropriate bound-
ary value problems for total stress and pore water pressure. The state of effective stress at a
and b are represented by the two solid Mohr circles. Using the known point 0,it is possible
to construct the two mobilized envelopes labelled F , and F , (the numerical values Fa and F ,
are the local factors of safety implied by the solution of boundary value problems). The
shaded zone in the figure represents a possible range of local safety factors along the surface
y(x). Using a limit equilibrium procedure in order to estimate the steady-state factor of
safety along y(x) would probably result with some 'average' mobilized envelope F,, (shown
dashed in the figure).
T, and T, are the points of tangency between the mobilized envelopes and the Mohr
circles. These points represent the effective stresses acting at a and b during steady state.
Based on the construction in Figure 2(b), it is clear that T, and T, represent stresses acting
SLOPE STABILITY ANALYSIS FOR UNDRAINED LOADING CONDITIONS 31
on planes inclined at (45 + 4J2) to the major principal plane. Consequently, these points
(T, and Tb)are located on undrained effective stress paths of the type described earlier, and
they may be used as the starting points for the radial interpolation procedure utilized by
STRNGH.
(2) Applying the routine STRNGH, it is possible to trace out the two undrained effective stress
paths which start at T, and T, and terminate on the effective strength envelope at the points
S, and S,. The vertical co-ordinates of the points S, and S, are the undrained strengths at
the points a and b on the slip surface y(x) . Performing a similar construction for all points
along y(x) results with the function S[y(x)] describing the distribution of undrained
strength along the potential slip surface under consideration.
(3) Using the function S[(y(x)] in a limit equilibrium calculation results with the undrained
factor of safety associated with y(x), {F,d[y(x)]}. This factor of safety defines a new
mobilized strength envelope as shown in Figure 9. The intersection of this envelope with the
stress paths defines the points T, and T,, which represent the effective stresses acting (at the
geometrical points a and b), immediately following the rapid loading. The curves T,T, and
T,Tb along the undrained effective stress paths, represent the evolution* of effective
stresses at the geometrical points a and b due to the application of the undrained loading.
This construction illustrates clearly that the type of stress paths introduced in the previous
sections is the natural representation for the purpose of limit equilibrium analysis of slope
stability problems. These stress paths have the advantage that they are expressed in terms of
the same stress components which are used in the calculations.
(4) For convenience, the above discussion was carried out for an arbitrary potential slip surface
y(x). A similar situation holds for every slip surface, including the surface ycud(x)which
realizes the minimum factor of safety for the undrained case {Fsud= Fud[Ycud(X)]}. In the
proposed procedure, the routine STRNGH is used for the purpose of generating an
undrained strength distribution in the entire profile, S(x, y ) (not only S[y(x)]). This strength
distribution is then used as input to a limit equilibrium procedure for the purpose of
identifying the critical pair [Fsud,Ycud(X)].
(5) The horizontal co-ordinates of the points T: and TL are the normal effective stresses (T: and
ob immediately following the rapid loading. In general, these are fictitious quantities acting
in the artificial material characterized by the strength envelope associated with F,d[y(x)].
Only if Fud[y(x)] is close to unity, is it reasonable to consider these quantities as actual
stresses. In this case further interpretation is possible. Recall that limit equilibrium
procedures automatically yield the normal stress function (T = o[y(x)]; hence, caand c b are
also available. The difference (T, - a; is the pore pressure at a immediately following the
rapid loading. Moreover, since the steady-state value of the pore pressure is known, it is
possible to estimate the excess pore water pressure generated by the rapid loading. In
addition, using the point Ti it is possible to draw the dashed Mohr circle and establish the
effective principal stresses and their directions, as is done in Figure 9. Thus, the procedure
also yields the change in principle stress ratio, and rotation of effective stresses (along the
slip surface under consideration), due to the rapid loading. Note that the dashed circles,
which represent a limit equilibrium state of stress, are tangent to the same envelope
(Fud[y(x)]), while the solid Mohr circles, which are established on the basis of solving
boundary value problems, do not share a common envelope.
*The curves T,Ta and T,T, represents evolution in the stress space; the transition is instantaneous in the idealized
undrained loading problem.
32 R. BAKER, S. F R Y D M A N AND M. TALESNICK
6. APPLICATIONS
In the present section we present a complete stability analysis of the upstream slope of a water
reservoir in Kibutz Givat Brener, situated in central Israel. A cross-section through the reservoir
embankment is shown in Figure 10(a).The height of the embankment is 1I m, supporting 10 m of
water. Crest width is 5 m. The downstream slope has an inclination of 1 : 3 and the upstream slope
an inclination of 1 : 4. No drainage features are incorporated in the body of the embankment. This
design characterizes a large number of small water reservoirs in Israel.
The soil profile at the site consists of heavy clay to a great depth and the regional water table is
deep. The embankment is constructed from the local material and it is built partly in excavation,
and partly as a compacted embankment, as shown in Figure 10(a).
Stability analysis was performed according to the procedure described previously, as follows:
c(kPa) (6)
Compacted clay 6.5 40
~ ~
Natural clay 0 32
An average value of shear modulus for both compacted and natural clay was found to be
5800 kPa. The following parameters were also adopted for both the natural and compacted clays:
Regionrl Ground W a l c r T a b l e
- so
Figure 10. Givat Brener water reservoir: (a) the problem; (b) the finite difference mesh-general; (c) closeup view of the
finite difference mesh
This powerful procedure has the advantage of running on standard PC compatible micro-
computers, making the proposed analysis feasible to a large community of geotechnical engineers
who do not have convenient access to a main-frame computer. The utilized finite difference mesh,
consisting of 764 elements, is shown in Figure tO(b); an enlarged view of the mesh within the
embankment is shown in Figure 1O(c).
34 R. BAKER, S. FRYDMAN AND M. TALESNICK
80 I I
a ) Natural Clay
c -0
0 20 40 60 80 100 1
I:rrective Normal S t r e s s . 0 ' , ( k P a )
80
,-.6 0
rd
a
v
x
c'
$40
z
w
L.
rd
2
v)
20
0 20 40 60 80 100 1
Effective Normal Stress, cr' , ( k P a )
Figure 11. Experimental undrained effective stress paths: (a) compacted material; (b) natural material
6.2.1. ( i ) . The steady-statepow regime. The flow regime as determined by FLAC is shown in
Figure 12(a).This figure shows a system of flow vectors superimposed on equipotential lines. The
following comments are relevant with respect to this figure:
(a) The plot is presented in 'transformed co-ordinates' in order to make the flow vectors appear
perpendicular to equipotential lines (effect of anisotropic permeability).
(b) The zero head line is at the regional water table, located at the bottom of the mesh as shown
in Figure 10(b) [not included in Figure 12(a)].
(c) The left end points of the equipotential lines define the boundary of the saturated region
(phreatic surface). The small dots located to the left of this region represent unsaturated flow
vectors.
The flow regime consists essentially of vertical downward flow towards the regional water
table. This flow regime is a consequence of the fact that the permeabilities of the natural and
SLOPE STABILITY ANALYSIS FOR UNDRAINED LOADING CONDITIONS 35
a) Flow vectors
compacted clays are approximately equal. Figure 12(b) shows the associated steady-state pore
pressure distribution (in kPa). Zero pore pressure was assigned to points located above the
phreatic surface, i.e. suctions were not considered in the analysis. This flow regime closely
resembles Vedernikov's solution for flow out of a The characteristic feature of this
solution is the steep phreatic surface and relatively low value of steady-state pore water pressure.
This situation results with large steady-state effective stresses and, therefore, high strength values.
36 R. BAKER, S. FRYDMAN A N D M. TALESNICK
Note that this flow regime is significantly different from the type shown in most texts, 2 5 ~ h i ~ h
is relevant to the case of an embankment on an impervious base. The flow regime correspond-
ing to an impervious base, which is used in most reservoir stability analyses, is probably not
relevant to many actual problems, and its use leads to an extremely conservative estimation of
slope stability.
6.2.2. (ii). Eflectiue stresses at steady state. Effective stresses at steady state were estimated
using a linear-elastic-perfectly-plastic constitutive model, and the pore pressure distribution
shown in Figure 12(b).The calculations were done with the FLAC program, taking account of the
effect of pore pressure on distribution of total stresses. Steady-state effective stresses (in kPa), are
shown in Figures 13(a)-13(c). In order to improve clarity, these figures are drawn with different
vertical and horizontal scales.
The distribution of effective vertical stresses reflect mainly the geometry of the problem. The
distribution of horizontal stresses shows a slight effect of the difference between the strength of the
compacted and natural clays. For later reference, we draw attention to the anti-symmetrical
nature of the shear stress distribution which is a consequence of the approximate symmetry of the
problem.
a) Vertical stresses
b ) Horizontal stresses
c ) Shear stresses
Imrlrr.)
Figure 13. Effective stresses at steady state: (a) vertical stresses. (b) horizontal stresses. (c) shear stresses
strength parameters presented above yielded a critical factor of safety of Fsemp = 2-41.This factor
of safety is realized on a very shallow critical slip surface ycemp(x)
shown in Figure 15. Such a result
is to be expected in a c = 0 material.28
6.4.2. Steady state. Steady-state conditions usually do not control the stability of the upstream
slope. Nevertheless, we analysed this case mainly for comparison purposes, but also as a
38 R. BAKER, S. FRYDMAN A N D M. TALESNICK
phreatic surface
\ / compacted clay embankment
-
natural clay subsoil
-
0 5
scale (rn)
10
strength contours in kPa
0 10 20 30 40 so 60 x(m)
b
I I I I I I
10
i(m1
20
preliminary stage for the application of Lowe and Karafiath's approach. Effective stress analysis
utilizing the pore pressure distribution shown in Figure 12(b)yielded F,,, = 6.98 which is realized
on the critical slip surface ycss(x)as shown in Figure 15. The huge increase in the safety factor is
the result of the support provided by the body of water inside the reservoir, and the low values of
pore water pressure associated with the downward flow mechanism.
6.4.3. Stability under rapid drawdown conditions. This case represents a problem of undrained
stability, which is the main concern of the present work. Consequently, we utilize this case for the
purpose of comparison between the proposed (direct) approach, and the procedures of Bishop
and Lowe and Karafiath.
SLOPE STABILITY ANALYSIS FOR UNDRAINED LOADING CONDITIONS 39
I I
AU-0 1I AU=y$&ZlX) 1 Au=Azm,,vw
I I
I I
1 I
Phreatic surface
Figure 16. Rapid drawdown pore pressure distribution implied by Bishops procedure
6.4.3.1. The direct approach. In this case the strength distribution shown in Figure 14 was used as
the data for SSOPT, and the problem was analysed for the conditions of empty reservoir using
zero values for pore pressures (total stress analysis). The analysis yielded the critical slip surface
ycda(x)as shown in Figure 15 with the associated minimum factor of safety of Fsda= 2.18.
6.4.3.2. Bishops drawdown analysis. Use of the steady-state pore pressure distribution shown in
Figure 12(b), and Bishops procedure for the estimation of the change in pressure due to
drawdown [equation (l)] results with the pore pressure distribution as shown in Figure 16. It is
seen that Bishops procedure implies, in this case, significant negative pore pressures in the zone,
where the flow is predominantly vertical. In some regions in the vicinity of the regional water
table, the undrained pore water pressure reaches a value of - 100 kPa. Such large negative pore
water pressures, in a saturated zone, do not appear credible, and they are clearly a by-product of
Bishops assumptions. It is realized that Bishop probably never intended his procedure to be
applied to the case in which the steady-state flow is nearly vertical.
An effective stress analysis of rapid drawdown stability using the pore pressure distribution
shown in Figure 16 yielded the critical slip surface ycbdd(x)(Figure 15), with an associated factor of
safety of F s b d d = 2.81. It should be emphasized that Bishops procedure was used only for the
estimation of the drawdown pore pressures; stability calculations were done using SSOPT which
utilizes the Spencer method of analysis. The result Fsbdd = 2.81 is significantly higher than the
value of 2.18 obtained by the direct procedure. The difference is largely due to the fact that the high
negative pore water pressure resulting from Bishops procedure shifted the critical slip surface
upwards, to a zone with smaller driving forces, resulting with higher factor of safety. Bishops rapid
drawdown analysis is generally considered to be conservative; obviously, this is not the case for the
present problem.
40 R. BAKER, S. FRYDMAN A N D M. TALESNICK
6.4.3.3.Low and Karafiaths approach. Previous discussion of this approach showed that it does
not lead to a unique spatial distribution of undrained strength, necessitating two LE analyses for
each potential slip surface. This characteristic of the procedure makes it virtually impossible to
identify the critical slip surface and its associated minimum factor of safety. Consequently, we
chose to calculate Low and Karafiaths factor of safety for the three surfaces ycss(x),ycbdd(x) and
ycda(x)shown in Figure 15; these surfaces represent the critical conditions for steady state, Bishop
drawdown analysis and the present direct approach, respectively. It should be pointed out that
there is no reason to assume that one of these surfaces represents the critical conditions for Low
and Karafiaths procedure. Application of Low and Karafiaths procedure makes it necessary to
perform, on each one of these surfaces, a preliminary stability analysis of the steady-state case.
The results of these calculations are summarized in Table I. In this table, F,, represents the rapid
drawdown factor of safety calculated according to the Low and Karafiath procedure. Numbers
with an asterisk () * are critical factors of safety for the appropriate conditions.
Figure 17 shows a comparison between some results of Low and Karafiaths procedure and the
direct approach. The comparison is shown for the critical surface Ycda(x) which yielded the
smallest factor of safety for the rapid drawdown case [Figure 17(a)]. Figures 17(b)and 17(c) show
the normal and shear stresses at steady state. The solid lines in these figure represent results
obtained by FLAC, while the dashed lines are the results of the preliminary limit equilibrium
analysis required by the Low and Karafiath procedure (SSOPT results).
The normal stress functions a[y(x)] obtained by these two procedures are fairly similar,
reflecting, mainly, the distribution of vertical stresses, which are relatively insensitive to the
constitutive nature of the material. However, the shear stress functions, z[y(x)] implied by limit
equilibrium and continuum analyses differ significantly. Note, in particular, the negative shear
stresses obtained in the FLAC solution; these negative values are the result of the anti-
symmetrical nature of the shear stress distribution shown in Figure 13(c). Limiting equilibrium
procedures, which deal with the state of stress in a fictitious material on the verge of failure,
cannot yield both positive and negative shear stresses on the same slip surface. Thus, the
difference between the two shear stress functions is qualitative as well as quantitative. The big
difference between the two solutions for t[y(x)] is to be expected in view of the high value (9.54)of
the steady-state factor of safety (see Table I). Figure l%(c)provides a clear illustration that stresses
obtained from limit equilibrium calculations are fictitious quantities which should not be
considered as stresses acting in the actual system.
Both Lowe and Karafiaths procedure, and the present direct approach, utilize a set of
undrained stress paths in order to establish the undrained strength for given values of shear and
normal stresses at steady state. The two procedures use, however, different estimates for the
steady-state effective stresses, resulting, therefore, with different estimates of undrained strength.
The extent of this difference is shown in Figure 17(d). In this particular example, the strength
values which are based on the solution of boundary value problems are smaller than those
associated with the preliminary limit equilibrium analysis, resulting with a lower drawdown
0 10 20 30 40 30 60
10
=.- ( a ) The s l i p surface yc_o,
a
(x)
C'lY (x iI
k Pa
I00
0 LL I
10
I
20
I
30 40
I I
50 60
I bx(m)
_.--
_.___._____ -.-.
11'
,/'
%
% ...,. (c)Distrihution o f steady state
,' '\
'.., shear stress along y
-
I' cd$')'
P
I I I 1 I I x (m)
10 20 30 40 50 60
( d ) Distribution of undrained
strength along Y ( x )
cdr
I I I
0 10 20 30 $0 50 60
results from FLAC
-- -- --- results f r o m SSOPT
Figure 17. Comparison between the direct approach and Lowe and Karafiath procedure: (a) The slip surface ycda(x);
(b) distributions of steady-state normal stress along ycda(x); (c) distributions of steady-state shear along ycda(x);
(d) distributions of undrained strength along yCda(x)
factor of safety (2.18 vs. 2.82) as shown in Table I. Figures 17(c)and 17(d) substantiate the basic
objection to Lowe and Karafiath's proposal of using a preliminary limit equilibrium analysis for
the purpose of strength estimation.
6.4.4. Pseudo-static earthquake stability analysis. The strength distribution shown in Figure 14 is
relevant to a general undrained loading. Consequently, it is possible to utilize this undrained
42 R. BAKER, S. FRYDMAN AND M. TALESNICK
strength distribution for pseudo-static earthquake analysis of the same slope, on condition that
the cyclic, earthquake loading does not cause a significant decrease in the soil strength. The
analysis was done for a horizontal earthquake coefficient of 0.1, assuming the earthquake force to
act at 60 per cent of the height above the base of each slice. The case of a full reservoir was
analysed, i.e. water forces on submerged parts of the slope were included in the analysis. Since the
analysis was done with direct strength specification, zero pore water pressures were used. The
analysis yielded the critical slip surface ycerq(x)as shown in Figure 15 with the associated
minimum factor of safety of Fserq = 2-32.
The fact that the same strength distribution (Figure 14) is relevant for the analysis of both
earthquake and drawdown cases (and any number of other undrained rapid loading conditions),
clearly illustrates the general applicability of the proposed approach.
REFERENCES
1. R. R. Chugaev, Stability Analysis qf Earth Slopes (Translated from Russian by A. Baruch), Israel Program for
Scientific Translations 1964.
2. Y. H. Huang, Stability Analysis of Earth Slopes, Van-Nostrand, New York 1983.
3. D. Brunsden and D. B. Prior (eds) Slope Instability, Wiley, New York, 1984.
4. E. N. Bromhead, The Stability ofSlopes, Surrey University Press, 1986.
5. R. N. Chowdhury, Slope Analysis, Elsevier, Amsterdam, 1978.
6. A. W. Bishop, The use of pore-pressure coefficients in practice, Geotechnique, 1954,4, 148-152.
7. A. W. Skempton, The pore pressure coefficients A and B, Geotechnique, 1954, 4, 143-147.
8. A. W. Bishop and L. Bjerrum, The relevance of the triaxial test to the solution of stability problems, Proc. Research
ConJ on Shear Strength of Cohesive Soils, Boulder, Colorado, 1960, 437-501.
9. D. W. Taylor, Fundamentals ofsoil Mechanics, Wiley, New York, 1948.
10. A. W. Skempton, The 4 = 0 analysis of stability and its theoretical basis, Proc. 2 Int. Con$ on Soil Mech. and Found.
Eng., 1948, 1, 72-78.
11. J. Lowe and L. Karafiath, Stability of earth dams upon drawdown, Proc. 1 Panamerican Con$ on Soil Mech. and
Found. Eng. Mexico, 1959, 2, 537-545.
SLOPE STABILITY ANALYSIS FOR UNDRAINED LOADING CONDITIONS 43
12. J. Lowe and L. Karafiath, Effect of anisotripic consolidation on the undrained shear strength of compacted clays,
Proc. Research Con$ on Shear Strength of Cohesive soils. Boulder, Colorado, 1960, 837-858.
13. M. Garber and R. Baker, Extreme value problems of limiting equilibrioum, J . Geotech Eng. Diu. ASCE, 1979, 105,
(GT10) 1155-1171.
14. R. Baker and M. Garber, Theoretical analysis of the stability of slopes, Geotechnique, 1978, 28, 395441.
15. L. Rendulic, Eingrundgesetz der tonmechanik und sein experimentaller bewwis, Bauingennier, 1937, 18, 459467.
16. D. J. Henkel, The shear strength of saturated remouldedclays, Proc. Research Conf: on Shear Strength of Cohesive
Soils. Boulder, Colorado, 1960, 533-554.
17. D. J. Henkel, Geotechnical considerations of lateral stresses, State of the art paper, Proc ASCE Speciality Conj on
Lateral Stresses and the Design of Earth-Retaining Structures, 1970, 1 4 0 .
18. C. P. Wroth and G. T. Houlsby, Soil mechanics-Property characterization and analysis procedure, Proc. 1 If* lnr.
Conf: on soil Mech. and Found. Eng., San-Fransisco, CA, 1985, 1, 1-55.
19. C. C. Ladd and R. Foot, New design procedure for stability ofsoft clays, J . Geotech. Eng. Diu. ASCE, l974,100(GT7),
736-786.
20. S. Frydman, M. Talesnick, G. Almagor and G. Wiseman, Simple shear testing for the study of the earthquake
response of clay from the Israeli continental slope, Marine Geotechnoloyy 1988, 7, 143-171.
21. R. Baker, Determination of the critical slip surface in slope stability calculations, lnt. J . Numer. Anal. Methods in
Geomech. 1980,4, 333-359.
22. P. S . K. Giam and I. B. Donald, Appropriate optimization technique for failure surface determination in geotechnical
stability analysis, Civil engineering Research Report No. 3/1989, Monash University, Australia.
23. ITASCA, FLAC fast Lagrangian analysis of continua, User Manual, ITASCA Consulting Group Inc., 1991.
24. M. E. Harr, Groundwater and Seepage, McGraw-Hill, New York, 1962.
25. A. Casagrande, Seepage through dams, in Contributions to Soil Mechanics 1925-1940, Published by the Boston Soc.
of Civil Engineers, 1940, pp. 295-336.
26. H. G. Poulos and E. H. Davis, Elastic Solutions,for Soil and Rock Mechanics, Wiley, New York, 1974.
27. E. Spencer, A method of analysis of the stability of embankments assuming parallel inter-slice forces, Geotechnique,
1967, 17, 11-26.
28. R. Baker, Tensile strength, tension cracks and stability of slopes, Soils Foundations, 1981, 21, 1-17.