You are on page 1of 17

BAKER, R. & GARBER, M. (1978). GPofechnique 28, No.

4, 395-41 I

Theoretical analysis of the stability of slopes

R. BAKER* and M. GARBER*

The problem of limiting equilibrium of a slope in a Le probltme plan de I’tquilibre limite d’un talus est
state of plane strain is formulated in terms of the formule dans le cadre du calcul des variations.
variational calculus. Formulated that way, the L’analyse est faite sans aucune hypothtse a priori
analysis is carried out without any a priori assump- concernant la forme de la ligne de glissement ou la
tions with respect to the shape of the slip surface, or repartition des contraintes normales, le long de
cette ligne. On montre que la valeur la plus faible du
the normal stress distribution along it. Thus, on the coefficient de securite est obtenue pour une ligne de
basis of only a formal definition of the concepts of glissement a gtomttrie particulitre. La dtmon-
limiting equilibrium, and factor of safety with respect stration de ce resultat ne font appel qu’aux equations
to strength, it is proven that the minimal factor of d’tquilibre limite et B la definition du coefficient de
safety must occur on slip surfaces with a special securitd unique, Cgal au rapport des caracteristiques
geometrical property. This geometrical property en- reelles et reduites du mattriau. Cette propriete
geometrique particuliere se distingue par le fait que les
sures that the resultant of the infinitesimal normal
forces infinitesimales resultantes des contraintes
and frictional forces either pass through a common normales et tangentielles le long de la ligne de glisse-
point or are parallel to a common direction. It is ment passent par un m&me point ou sont paralleles.
shown that as a result of this geometrical property the II en resulte que la valuer la plus faible du coefficient
minimal factor of safety is independent of the normal de securite est independente de la repartition des
stress distribution along the critical slip surface. In contraintes normales a la ligne de glissement critique.
Dans le cas d’un milieu homogtne et isotrope
the homogeneous and isotropic case theanalysis shows
I’analyse montre que la ligne de glissement critique
that the critical slip surface may be either a log-spiral peut itre soit une spirale logarithmique soit une
or a straight line. In a layered profile the critical slip ligne droite. Dans le cas d’un profil multicouches, la
surface may consist of series of log-spirals that have a ligne de glissement critique peut comprendre soit
common pole or a series of straight lines. In some une serie de spirales logarithmiques a foyer commun
cases the boundary between layers may be part of the soit une serie de lignes droites, ainsi que I’interface
entre les couches dans certains cas. Tous les
critical slip surface. All the results obtained are valid
resultats obtenus sont valables dans le cas d’un sol
for a general non-homogeneous, non-isotropic soil non homogene et non isotrope avec repartition
with arbitrary distribution of pore water pressure and quelconque de la pression intertitielle et des forces
external loads. exterieures.

The study of the stability of slopes in a plane strain state, usually proceeds within the frame-
work of limit equilibrium methods. This essentially simple approach is based on only two
considerations: of the overall equilibrium of a test body, and of a failure condition. Applica-
tions of this approach are at present complicated by a number of assumptions, which, as we
shall show, are not only unwarranted, but actually quite unnecessary.
It is well known that the calculation of the factor of safety requires information regarding
two functions. One of these is the equation of the potential slip surface. The second repre-
sents either the normal stress distribution along this surface, or some property of the forces
acting on vertical planes. The first function will be called the kinematical function, while the
second is termed the stress function. It is our belief that the properties of these functions
should be derived from the principles of the limiting equilibrium approach ratber than arbi-
trarily assumed. The existing applications of the limiting equilibrium approach to slope
stability are briefly reviewed from this point of view.
Discussion on this Paper closes 1 March, 1979. For further details see inside back cover.
* Techinon, Israeli Institute of Technology, Haifa.
396 R. BAKER AND M, GARBER

NOTATION
A integration constant Y’t, Y’b, Y’, required slopes of the extremal
c, 4 cohesion and angle of effective above, below and along a sur-
internal friction respectively face of discontinuity
ci, &, 31 cohesion, pore water pressure, Y’, slope of the surface of discon-
and friction coefficient at the tinuity
centre of the base of the ith Y‘i Y’Cxi)
slice Yi JXxi)
rid/, at arc tan Y’~
the safety functional Qb arc tan y’,
the factor of safety %I arc tan y’,
auxiliary function Y total unit weight
index 7 average total unit weight of the
arc length along y(x) soil above y(x)
number of slices Axi width of the ith slice
unit vectors, normal and tan- Ali dl + (y’i)’ Axi 8
gental to y(x), respectively APxi Px(xiYxi
Pm Pv distributions of horizontal and APvi Py(xWi
vertical external loads along Aw, weight of the ith slice
y(x) respectively 00,4l polar angles of the points x0 and
rr 0 polar co-ordinates X”
R, G auxiliary functionals Lagrange’s undetermined multi-
u pore water pressure distribution pliers
ui UiAli VY77 angles
xcrYc centre of the polar co-ordinate 7. a distribution of tangential and
system normal stress along y(x)
Xi, Yi co-ordinates of the centre of the d mobilized friction angle
base of the ith slice bt, 4, mobilized friction angles above
xorXII end points of y(x) and below a discontinuity
Y(X) potential slip surface trace tan 4
f(x) equation describing the slope the value of 1+5at a discontinuity
surface the 4 value of the layers sur-
YdW equation of a discontinuity rounding a transition zone
Y' dy/dx = tan CL ($1’ $2)

REVIEW OF LIMITING EQUILIBRIUM METHODS


This review follows the attempts to free the limiting equilibrium approach from unnecessary
assumption, so that it can realize its inherent simplicity and generality.
Simplified methods: representative of this group would be the Fellenius method (Fellenius,
1936). This group may not properly be classified as a limiting equilibrium method, since it
does not satisfy a basic element of this approach-namely the equations of equilibrium.
A second group, to which Taylor’s method (Taylor, 1937) belongs, satisfies all equilibrium
requirements, but makes arbitrary assumptions with respect to both functions mentioned
above. Taylor’s method is based on the assumptions that the kinematical function represents
a circular arc, while the stress function is distributed as sin (x).
Generalized method of slices; this group consists mainly of variations of the approach
presented by Morgenstern and Price (1965), and Janbu (1954). In this group the kinematical
THEORETICAL ANALYSIS OF THE STABILITY OF SLOPES 397

function is left unspecified, while the stress function is assumed. No guidance as to how to
locate the critical slip surface is being provided, hence the application of these methods is
extremely tedious.
In order to provide a computation scheme that enables the determination of the critical slip
surface, a variational reasoning has been applied to the generalized methods of slices (Revilla
and Castillo, 1977). A similar type of analysis had been carried out earlier (Dorfman. 1965;
Garber, 1973).
The first attempt to formulate the slope stability problem as a variational problem in terms
of two unspecified functions was made by Kopacsy (1955). A reappraisal of Kopacsy’s
analysis by Baker and Garber (1977a) shows, however, that this analysis contains a number of
serious errors and misconceptions. An improved variational formulation of the slope
stability problem was presented by Baker and Garber (1977b). This formulation applies to
the case of homogeneous and isotropic soil, without pore water pressure or external loads.
For this simplified case Baker and Garber (1977b) showed that a complete analysis of the
slope stability problem may be based on the concept of limiting equilibrium, without any
arbitrary assumptions.

THE PRESENT WORK


Previously, the variational reasoning was used to derive solutions for specific, often over-
simplified, slope stability problems. The present work, which is a generalization and extension
of the analysis of Baker and Garber (1977b), utilizes the variational technique in order to
derive a general theorem, governing the shape of potential slip surfaces. The theorem ob-
tained is valid for the general case of non-homogeneous, non-isotropic soil, with arbitrarily
specified distribution of pore water pressure and external load. Application of this theorem
makes it possible to obtain the minimal factor of safety for almost every practical problem,
using a procedure that is only slightly more laborious than Bishop’s simplified method (Bishop,
1955). The analysis is complete in the sense that all equations of equilibrium are satisfied.

MATHEMATICAL FORMULATION OF THE PROBLEM


The limiting equilibrium state
A mass of soil is considered to be in a state of limiting equilibrium (LE) if Coulomb’s
failure condition is satisfied along a potential slip surface Y(x).
7 = c+(u-24)$ . . . . . . . . . . (1)
where T=T(X) and u=u(x) are the distributions of the tangential and normal stress along
Y(x); r=c(x, Y, Y’), $= tan [4(x, Y, Y’)] and u=u(x, Y) are the distributions of the cohesion, the
effective friction coefficient, and the pore water pressure, respectively, and y’=dy/dx; and if
the equilibrium equations of horizontal and vertical forces and of moments (around the
origin of the (x, y) co-ordinate system, Fig. 1) are satisfied for the mass bounded by the slope
surface Y(x), and the slip surface y(x),
XII
Tcosa-usincr]dl+ pxdx = 0 . . . . . . (2.1)
s ,[ f x0

[T sin cc+a cos IX]dl- X” [p,+p(p-y)] dx = 0 . . . . (2.2)


s1 s x0

, ]Y(T cos a-u sin a)-X(7 sin afu cos a)] dl+ %” [p,p+p,x+y(jj-y)x] dx = 0 (2.3)
f f
where tan a = y’, I is the arc length along y(x); x0, x, a,x,”the end points of y(x); pX =pJx) and
p,=pJx) are the distribution of external loads along the slope surface y(x); 7=7(x, y) is the
398 R.BAKER AND M. GARBER

distribution of the average total unit weight of the soil above y(x). This quantity is related to
the conventional unit weight y, by the relation
Y(X)
c YcGY)+
JYW
7 = 7(X?Y>= jj(x) _y(x)

The dependence of c, (CI,u and y on (x, y) indicates a possible non-homogeneity, while the
dependence of c and I/I on y’ indicates a possible non-isotropy. The functions c, (CI,U, y, pxpv
and j are considered known, while y(x) and a(x) are the unknowns in the analysis.

The factor oj’sajkty


In general, a soil mass of given properties and geometry that is acted upon by a given set of
loads, is not in a state of LE as previously defined. In order to quantify the margin of safety
relative to a state of imminent failure (LE state), one may replace the soil’s real strength
parameters c and # by an artificial one P and (t, for which a state of LE may be realized. There
are many possible ways by which c and $ may be related to C and 4, and still realize a state of
LE. It is customary however, to adjust the real strength parameters by a single factor F in
the following manner
P=clF . . . . . . . . . (3.1)
4 = #/F . . . . . . . . (3.2)
The quantity F, for which a mass of soil attains a state of LE depends in general on the
functions y(x) and u(x) (the kinematical and stress functions) that are being considered.
Therefore F is a functional of two functions. This functional is termed the safety functional,
to be distinguished from the factor of safety F,, which is the minimum value of F.
F, = min F{y(x); u(x)} . . . . . . . (4)
Combining equations (I), (2) and (3), and using the geometrical relations cos (c~)=dx/d/;
sin (a) = y’dx/dl (Fig. l), one arrives at

c X”{[c+(o-u)$]-F[~y’-pX]}dx = 0 . . . . . . (5.1)

X”
s
:;{y’[c+((i-u)$]-F[l)v+jQj-y)-o]}dx = 0 . . . . (5.2)

{(y-y’x)[c+(o-u)$]- F[a(x+y’y)-p,j-p,x-jj(y-y)x]} dx = 0 . (5.3)


s x0
The basic problem of slope stability can now be stated as follows. Find a pair of functions,
y(x) and U(X), that realize the minimum value F, (Factor of safety) of the safety functional F,
subject to the three equations of LE (equations (5)).

VARIATIONAL ANALYSIS
Transformation of the basic problem
It was shown by Baker and Garber (1977b), that the solution of the basic problem formu-
lated above is identical to the solution of the following isoperimetric problem of the calculus
of variation, namely, find a pair of functions y(x) and O(X) that minimizes a functional R

R= “{[c+(“-u)$]-F&y’-pJ}dx . . . . . . (6.1)
s x0
with the condition that
minR=O . . . . . . . (6.2)
THEORETICAL ANALYSIS OF THE STABILITY OF SLOPES 399

and subject to the following two integral constraints:

x”{y’[c+(rr-~)~]-F,[~~+~(~-~)-ul} dx = 0 . . . (7.1)
s x0
%
KY-y’x)k+(a-yMl- F,[a(x+y’y)-p,y-p,x-g(~-y)xlJ dx = 0 . (7.2)
ss a3
Equations (7) represent the conditions of vertical and moment equilibrium. Horizontal
equilibrium is satisfied by virtue of equation (6.2). The factor F, appearing in equations (6.1)
and (7) is a constant to be determined.
Note that the basic system of equations (equations (5)) contains the complete safety func-
tional F, rather than its minimum value F, as in equations (6.1) and (7).
This standard isoperimetric problem is in turn equivalent to the problem of minimizing the
auxiliary functional G defined by

G = IX;gdx = JX; H(~f(u- u)lcI) - F,(v’-P,)J + &{Y’(c+ (u- UN’) - F&J, +?(Y-Y) - u))

+~,((y-y’~)(~+(~-~)lCI)-F,(~(~+~’~)-~,~-~,~-~(~-y)x)}l dx (8)
Subject to satisfying equations (7) and (6.2). The parameters XI and h, appearing in equation
(8) are Lagrange’s undetermined multipliers.
The satisfaction of equation (7) and the condition min R=O imply also min G=O. Thus
the problem is finally reduced to that of finding a pair of functions y(x), a(x) that minimize the
functional G (equation (8)) and adjusting the 3 unknown parameters h,, A2 and F, in such a
way that will ensure the satisfaction of equations (7) and (6.2).
Rearranging the different terms in equation (8) the final system of equations is

G = IX;gdx = IX; {uM(l + J&Y)+Y’(& - X,x)) - Fs((h - ‘\zx) -Y’(] + b))l


+(c- dNl+ ~Y)+Y’(& - Lx)1 + Ml +G9~x-(h - LWv +KF--YJI: dx (9.1)
minG=O . . . . . . . (9.2)

*“{y’(c+(u-u)$)-F,(p,++--y)-u)}dx = 0 . . . . (9.3)
s x0
%I
{(v-y’x)(c+(u-u)~)-F,(u(x+y’y)-p,~-p,x-~(y-y)x)} dx = 0 . (9.4)
s x0

Euler’s equations
The functions y(x) and a(x) that constitute the solution of the system of equations (9) has to
satisfy Euler’s differential equations for the function g, namely

-
d ?g0 ........
1--=
;I;;
au1
%a [
iig
(10.1)

d ag ag o

Since g is independent
1
zv-y=
[
of U’ the first Euler equation
.

reduces to
. . . . . . . (10.2)

aIT= 0
. . . . . . . . . .
Z
Applying equation (11) to the function g (equation (9.1)) results in
t,q( 1 +&y) +y’(& - Azx)] + Fs[(& - &x) -#Cl + k)l = 0 . . . (‘2)
400 R. BAKER AND M. GARBER

Fig. 1. Basic conventions and definitions

This is a first order differential equation for the determination of the family of possible
critical slip surfaces. The study of this equation necessitates the considerations of two
separate cases.

Case (a), (h, # 0)


In this case equation (12) may be written as:
4 (~lF,)ty+(l/h,)l+[(X,/h,)-xl
(13)
Ji = [Y+(l/X,)l-(~I~,)[(~,IX,)-xl . . . . .
This equation may be simplified considerably with the following co-ordinate transformation

Y=
i
-+
2 !+r sin 0 = y,+r sin 0 . . . . . (14.1)

.X= -r cos 0 = x,-r cos 0 . , . . . . (14.2)

Where (r, 8) are polar co-ordinate system with respect to the still unknown point (xc, Y,), (Fig.
1).
and
x, = AJh, . . . . . . . . (15.1)
Y, = - l/A2 . . . . . . . . . (15.2)
In terms of the polar co-ordinates r and 0, equation (13) is simplified to

$ = (#IF& . . . . . . . . . .

Cuse (b), (&!=O)


introducing the condition A,=0 into equation (12) one obtains

The fhmily of potential slip surjbces


The analysis carried out so far indicates that the minimum value (F,) of F may occur along
THEORETICAL ANALYSIS OF THE STABILITY OF SLOPES 401

only lines y(x) that are the solutions of equation (12) or more specifically equations (13) or (17).
Up to this point no assumptions have been made with respect to the dependence of 9 on x, y
and y’. Consequently equations (13) and (17) are valid for a general non-homogeneous, non-
isotropic soil.
The solutions of equations (13) and (17) will however reflect the properties of the function
#=#(x, y, y’) and will be different for different # distributions. It is convenient to write this
solution in the following way:
y=y(A,,h,,A,F,,$,x) . . . . . . . . (18)
where A is the constant of integration.
The solution y(x) was written in this explicit form in order to emphasize its dependence on
the four unknown parameters AI, AZ, A and FS, and the given function #=#(x, y, y’).
Equation (18) represents the family of possible slip surfaces. The nature of this family is
independent of the distributions of c, u, y, 1, pX and pv. This does not imply that the critical
slip surface itself is independent of those distributions, since they will control the values of the
parameters A,, A,, A and F,, and therefore will select from all possible slip surfaces, represented
by equation (18) the one that is critical. The lines given by (18) are called the extremals.

THE BASIC THEOREM


Comparing equations (12) and (9.1) one sees that lines which are the solution of (12),
(the extremals) render G independent of D. It follows that the application of the first
Euler equation alone reduces G from being a functional of two unknown functions, to a
function of four unknown parameters (AI, A,, A, F,) only. It is possible, therefore, to deter-
mine the factor of safety F, by imposing the requirement min G=O, (equation (9.2)) on the
reduced functional (function) G, without specification of the normal stress distribution.
In order to complete the solution of the problem one has to ensure the satisfaction of the
remaining two integral constraints (equations (9.3) and (9.4)). It is easy to see, however, that
every u(x) function that has at least two degrees of freedom (two unspecified parameters) will
satisfy these constraints.
It may be concluded, therefore, that it is possible to obtain a solution of the basic problem
without specification of the normal stress distribution. This result permits the formulation
of the following basic theorem; the minimum value F, (factor of sqfety) qf the safety functional
F, is independent of the normal stress distribution along the critical slip surface y(x).
This theorem is completely general; it is valid for every distribution of 4, c, y, jj, px and py.
The only restriction on its validity is the linearity of the Mohr-Coulomb failure criterion, i.e.
the assumption I/ #f (u).
The basic theorem does not contradict the initial assumption that the safety functional F
depends on both y(x) and a(x). Only if the particular class of functions y(x) that satisfies
equation (12) is considered, the dependence of For o(x) disappears.
It has been observed (Morgenstern and Price, 1965; Taylor, 1937) that if a calculation pro-
cedure satisfies all equations of equilibrium, the resulting factor of safety is often not sensitive
to changes in the assumed form of the stress function. No explanation of this phenomenon
has been suggested previously. It appears that it is, in fact, a consequence of our basic
theorem. If the slip surface for which the comparison of factors of safety has been made
resembles a member of the family of extremals (as it is expected to do if it is a reasonable slip
surface) then, by virtue of the basic theorem, the factor of safety should be nearly independent
of the normal stress distribution along the slip surface (or any other form of the stress function).
Moreover, if the comparison of factors of safety were made on a member of the class of
402 R. BAKER AND M, GARBER

FrictIonal force y
c

Fig. 2. Culmann’s analysis

extremals, the dependence is expected to disappear completely. This could be used as a sort
of experimental check on the validity of the basic theorem.
The expressions which, upon their minimization, yield the factor of safety may be obtained
as follows:
For the case A, # 0, from the combinations of equations (9. l), (9.2), (12), (13) and (15):

-[(x-x,)(P,+Y(~-Y))+P,(~-Y,)I dx = 0 (19)
>
The minimization, in this case, is with respect to x,, y,, and the integration constant, which
may be taken as x0. The lines to which equation (19) is to be applied should satisfy the
differential equation (13).
Similarly, for the case A, = 0, the combination of equations (9. I), (9.2), (12) and (17) yields

min G = min s:,“{(y) [AI:il;21] +h,[g,+y(y-y)]-p,}dx = 0 . (20)

The minimization in this case, is with respect to A,, and the integration constant x,,. The
lines to which equation (20) is to be applied should satisfy the differential equation (17).
The actual factor of safety is the smaller one of the two F, values obtained from equations
(19) and (20). These two equations represent the factor of safety in two alternative modes of
failure.

INTERPRETATION OF THE BASIC THEOREM


The case A2# 0 (rotational mode of failure)
Consider the case $=constant. In this case equation (16) may be integrated
r(0) = A exp [(#/Fs)O] . . . . . . . . (21)
where A is an integration constant. Equation (21) is an equation of a log-spiral, with pole at
the point (x,, y,), thus a log-spiral is one of the two possible slip surfaces for the homogeneous
case.
It is a well known property of the log-spiral function, that the resultant of the elementary
forces ad1 and (#/F&d/ passes through the pole of the spiral. Hence in this case the moment
equation about the pole is independent of u, and may be utilized for the determination of the
factor of safety regardless of the normal stress distribution. The two remaining equilibrium
equations may be satisfied by every (I distribution that has two degrees of freedom. It is
THEORETICAL ANALYSIS OF THE STABILITY OF SLOPES 403

proved, therefore, that the log-spiral analysis as proposed by Rendulic (1935) is not only
convenient, but actually yields the correct result for the case #=constant.
In view of this result, the equation of moment equilibrium is rewritten about the origin of the
polar co-ordinate system (x,, _yc), in the general #=#(x, y, y’) case

x”c(Y-Y~)-Y’(x-x,)(c+(~-u)~)-F,~~((x-~,)+Y’(~~-Y,))
s x0
-(x-.~,)(P~+~(~-Y))-(~-Y,)P,~} dx = 0
This equation may be written in the following form:

j=xy{+(
(Y-i)+y’&x))+F,@) -Y(Y-t))]
+k-W((I’+;)- (x-k))+f,[(+(p,+CyN+ (+&]}dx =0 (22)
Comparing the coefficient of 0 in this equation with the differential equation (12) one sees
that lines that satisfy (12) (i.e. the extremals) render the moment equation (22) independent
of (T. This may occur only if the resultants of the elementary vectors odf and (#/F,)adl all
pass through the centre of the polar co-ordinate system. This result is independent of the 4
distribution, and therefore generally valid.
It is appropriate to call solutions of equation (16) with arbitrary (I/=$(x, y, y’) a generalized
log-spiral family, since they possess the characteristic property of the common log-spiral, i.e.,
the angle between the normal and the radius vector is the local 4 value, where I$ is defined as
the mobilized friction angle, i.e.
4 = arc tan ($/FS) . . . . . . . . . (23)
Introducing y’ from equation (13) into the remaining part of (22), it may be shown that
this equation becomes identical to (19). This provides the physical meaning of (19) as an
expression of moment equilibrium about (x,, yJ. As the factor of safety, in this case, (XfO),
is obtained from the moment equation, it yields the margin of safety against rotational failure.
It is appropriate, therefore, to refer to the case h,#O as the rotational failure mode.

The case h, = 0 (translational mode of failure)


Consider the case #=constant; equation (17) may be integrated to yield
A,+ (#/~A
Y(X) = &Al($,Fs)x+A . . . . . . . .
where A is an integration constant.
This is an equation of a straight line, and thus a straight line is the second possible slip
surface for the homogeneous case. It is evident, that for $=constant, the case /\2 =0 corres-
ponds to Culmann’s method for slope stability calculations (Culmann, 1866).
It is well known that the Culmann approach is independent of the normal stress distribution
along y(x). This is due to the fact that in this case the resultants of the elementary forces ad1
and ($/F&d1 are colinear in a certain 77direction (Fig. 2). Hence the equation of equilibrium
in a direction Y that is normal to 7, must be independent of (J. Since v=or-I$ (Fig. 2) and
tan cz=y’, tan $=4/F,, it can be shown that:
tan (v) = h, . . . . . . . . . (25.1)
tan (7) = - l/X, . . . . . . . . (25.2)
Following these results, the equation of linear equilibrium is written in the Ydirection, in the
general 4=$(x, y, y’) case as follows.
404 R. BAKER AND M. CAREER

xT U(C+(u--)$)--s( by’-p,)] cos v+[y’(c+(u-u)#)-F,(p,+y(j-y)-u)] sin V}dx = 0


s

Introducing into this relation (25.1), and rearranging

X”{4,W +~,y’>-F,(y’-~~~l+(c-cclu~~~
+h,y’)+F,[p,-~,(p,+~(~-y)lJdx = 0 (26)
s x0
Comparing the coefficient of u in this equation with the differential equation (12), specified
for the case h,=O, it is seen that lines satisfying equation (12) (i.e. the extremals) render the
equation of equilibrium in the v direction independent of (I. This is possible only if the
resultant elementary forces are all normal to this direction, or, in other words, are all directed
in the 7 direction (tan q= - I/&). This result is independent of the # distribution, and
therefore generally valid.
The solution of equation (17) possesses the geometrical property that the angle between a
certain direction 7 and the normal to the extremal is always the local 4 value.
Introducing y’ from equation (17) into the remaining part of (26), and comparing the
results with (20), it is found that equations (26) and (20) are identical. This provides a
physical meaning to (20) as an expression for linear equilibrium in the Y direction (tan Y= A,).
As the factor of safety in this case (X, =0) is obtained from an equation of linear equilibrium,
it yields the margin of safety with respect to failure by translation. It is appropriate therefore
to refer to the case X,=0 as the translational mode of failure.

Geometricalfbrmulation of the basic theorem


The results of the last two sections make it possible to express the basic theorem in purely
geometrical terms. To this end, two unit vectors N and T, that are normal and tangential to
y(x), are introduced. The basic theorem can then be stated thus; the minimum value F,
(factor qf sqfety) qf the safety,functional F is realized on lines such that the elementary vectors
N+ ($/FJT, either pass through a common point, or are parallel to a common direction.
This form of the basic theorem actually explains how is it possible for the factor of safety to
be independent of u(x), as required. The two forms of the basic theorem are equivalent, each
implying the other. The geometrical form is however more suggestive and explicit since it
indicates a procedure by which potential extremals may be constructed.

THE SECOND EULER EQUATION


The results obtained so far did not require the application of the second Euler equation
(equation (10.2)). It has been shown by Baker and Garber (1977b) that the second Euler
equation provides a first order differential equation for the determination of the critical u(x)
distribution. In the present work it was found that the factor of safety is independent of u(x),
and that every U(X) function that has at least two degrees of freedom can be considered as a
solution to the basic problem.
Considering the fact that the solution of a first order differential equation contains one
integration constant (degree of freedom). it may be seen that the basic problem admits two
different types of solutions; any function U(X) that has two degrees of freedom and one par-
ticular function (the solution of the second Euler equation) with one degree of freedom only.
Since the family of critical slip surfaces is common to both types of solution, and possess the
geometrical properties discussed before, it is evident that the factor of safety must be the same
in both solutions.
It follows that the two solutions are equivalent and neither one may be considered superior.
THEORETICAL ANALYSIS OF THE STABILITY OF SLOPES 405

Hence the approach taken by Baker and Garber (1977b), and the results obtained here com-
plement rather than contradict each other.

LAYERED PROFILE
The analysis presented so far pertains to the case of continuously varying properties (4 in
particular). In practice, however, such cases are of general interest in which the soil profile
consists of a number of layers such that c, 4 and y are constant inside each layer, with dis-
continuities at the layer boundaries. It is important therefore to consider such cases in detail.
Two relevant theorems apply to this type of variational problem with discontinuities.
(a) ‘Discontinuous solutions must also satisfy Euler’s differential equations’ (Bolza,
1973, p. 36).
(b) ‘The isoperimetric constant X has the same value along the different segments of a
discontinuous solution’ (Bolza, 1973, p. 209).
These two theorems make it possible to conclude immediately that in the present case the
family of extremals y(x) consists of either segments of log-spirals that have a common pole, or
segments of straight lines that have a common (N+ ($/F,)T) direction. The geometrical form
of the basic theorem provides a criterion that the vectors (iV+(#/F,)T) have to satisfy. For
AZ= 0 these vectors have to be parallel to each other, while for X, # 0 they have to pass through
a common point. At a point of discontinuity of #, those two requirements amount to the
same implication, i.e., that at a point of discontinuity the vector (iV+($/F,)T) is common to
both sections of the extremal above and below the discontinuity. This requirement implies
(Fig. 3(a))
a, = ab+(lj,-lj,) . . . . . . . . . (27)
where at, ab, and &, 4, are respectively, the inclination of the extremal, and the mobilized
friction angle, above and below the discontinuity.

THE INTERFACE BETWEEN LAYERS


Transition zone
It is reasonable to assume that at a boundary between layers that can intermix, a narrow
transition zone (TZ) will occur. The distribution of (c, across such a TZ is expected to be
single-valued. The extreme values of this distribution (cI1and #2($1 >&J are assumed to be
equal to that of the surrounding laters, and are located at the boundaries of the TZ (Fig. 3(b)).
With this assumption, the problem of a layered profile reduces to that of a non-homogeneous
but continuous 4 distribution. The determination of potential extremals in such a profile was
discussed previously. Assuming that the (CIdistribution inside the TZ is known, a potential
extremal may be constructed. Two such extremals are shown in Fig. 3(b) as the lines b,-b,
and dl-d4.
In practice however, the TZ is characterized by its centre line yd(x) only (line cl-c2 in Fig.
3(b)) and the assumption that such a zone is thin. As the # distribution is not known it is
impossible to construct the extremal in the TZ. It will be shown that the knowledge ofy,(x)
alone is sufficient for an approximate determination of the effect of the TZ on the stability.
Consider first the extremal bl-b4, which crosses the TZ. As such a zone is assumed to be
thin, the length b,-b, is small, hence the distribution of 1c,along it has only negligible influence
on the factor of safety. It is appropriate therefore to approximate in this case the TZ by a
discontinuity, and construct the extremal using the refraction condition (equation (27)).
The extremal may follow the TZ over a considerable distance (line dl . . . d4 in Fig. 3(b)).
The construction of the extremals in the TZ requires knowledge of the $ distribution in this
406 R. BAKER AND M. GARBER

zone, there are however a number of properties of such an extremal that are independent of
the # distribution namely,

(4 The differential equation for the extremals (equation 13) is satisfied at every point in
the range d2 IX I d3.
(b) The slope of the extremal is continuous in the vicinity of the points d2 and d3. (This
is a consequence of the assumption that (c,is continuous at these points.)
(cl The 4 values mobilized by the extremal (&) at every point in the range d2 ixsd,,
satisfies (cl25 *a 5 &.
Cd)As the TZ is characterized by the function yd(x) only, the best possible assumption for
the geometry of the extremal in the range d2 2 x 2 d3 is to identify it with JJ~(x), i.e.
Y(X) = Yd(X)
Y’(X) 2 Y’d(X)
(4 Conditions (a) and (d) imply that at every point at which the extremal is inside the TZ,
& may be calculated from the differential equation (13)
Y’d(I+ by,) - (A1- b)
*d= Fs(1 f&y,) +y’,(& - h,x)
Combining this equation with the requirement (c) indicates that the extremal will
coincide with yd(x) as long as the following inequality holds
~d(1+X2~d)-(A~-h2x) < $I . .
(28)
*’ ’ Fs(1 + &yd) + y’,(& - &?x) -
At the point where $d equal I,& or &, the extremal will deviate from JJ~(X)into the top
or bottom layer.
(f-1Combination of(b) and (d) shows that the extremal must join yd(x) tangentially.
The conditions (e) and (f) enable the determination of the segment of _Y~(x)that is a part of
an extremal, and the #d value mobilized along this segment. Hence a solution procedure for
this case has been established.
lt can be shown that equation (28) is equivalent to a mathematical model of the TZ as a
discontinuity with a step function as the # distribution across it.

Contact plane
An interface between layers may represent a contact plane between two non-intermixing
layers. In such case the 4 value at the interface is uniquely specified. This $d value may be
outside (usually below) the range #I -I,& (Fig. 3(c)). A rock joint is a typical example of this
type of interface.
Since I,/I~is uniquely assigned, a discontinuity of an arbitrary shape may satisfy the differential
equation of the extremals only at one point, i.e. the extremal may be tangent to such a discon-
tinuity, but not run along it. A straight line discontinuity may however be a segment of an
extremal in a translational mode of failure, and a discontinuity of a logarithmic-spiral shape
may be such a segment in a rotational mode of failure. The conditions in this case are
similar to equation (28) with (cI1and (crzreplaced by the given tid. In this case there exists no
requirement that the extremal approaches the discontinuity tangentially.

Interjbce condition, summary


Refraction of extremals by an interface is governed by equation (27) regardless of the inter-
face type. Based on the arguments presented above, it is possible to formulate the following
THEORETICAL ANALYSIS OF THE STABILITY OF SLOPES
407

\
(4

Bottom layer

04

Top layer

(c)
Fig. 3. The interface between layers: (a) refraction of ertremals; (b) transition zone; (c) contact plane

corollary to the basic theorem : ‘A segment of a discontinuity of the transition zone type may
be a part of an extremal regardless of its geometrical shape. The connection between the
extremal and the discontinuity is smooth with a common tangent.
A segment of a discontinuity of the contact plane type may be part of an extremal if it is a
straight line (translational failure) or a log-spiral (rotational failure). The extremal may
approach the discontinuity without smoothness restrictions.’
Based on this corollary, it is possible to conclude that form a point xA which is common to a
discontinuity and an extremal, the extremal may proceed into the top layer, bottom layer, or
408 R. BAKER AND M. GARBER

along the interface. More than one continuation may be possible from a given point x~,
indicating the existence of more than one extremal with a given set of characterizing parameters
(X,, & and x0). This multiplicity of extremals results from the lack of uniqueness of Ic,at a
discontinuity.
Let ylt, y’, and y’, be the required slopes of the extremal above, below and along the dis-
continuity, the criterion that controls the continuation of the extremal from the point x~, is
conveniently summarized in Fig. 4.

THE COMPUTATION SCHEME


It is not our intention to provide a detailed computation scheme based on the analysis
presented in this work. Here, merely one tentative scheme that may be used for this purpose
is briefly sketched.
To this end equations (19) and (20) may be rewritten in the following way

for the case AZ# 0 and

for the case X,=0.


Using the co-ordinates transformation (equation 14), it is possible to write (29) as
0”
(c- +)r2 d0
s 00
F, = min
xnI(x~-x)[~,+9(~-~)l-~,(~,-~)} dx
s x0
Approximating the integrals in this expression by a finite sum, and taking constant values
of c, +, u and r within each slice, this equation becomes

.5 (C,-~i”i)~/(x,-xi)2+(Yc-Yi)2
F, = min ‘=’ . (31)
I? [dp~i(Yc-~i)-(dp,l+dwi)(X~-X,>l * ' '
i=l

where
C, = c,Al,; U, = uiAli; AP,i = px(~i)A~i;
AP,, = pv(xi)Axi ; Ali = l/l+02 Ax,; Y’i = Y’(Xi)
xi and y, are the co-ordinates of the centre of the base of the ith slice; c1= c(x,, y,); 4, = $(xi, y,);
u, = u(xi, y,); Axi is the breadth of the ith slice; ji =J(xi); Aw, is the weight of the ith slice; n is
the number of slices and i is an index.
Similarly, it is possible to show that equation (30) may be written as

F, = min . . . . . (32)
f [AP,, sin q-(AP,i+Ay) COST]
i=l
THEORETICAL AYAl YSIS OF THF STARILITY OF SLOPES 409

Fig. 4. Construction of an extremal from a point xA on an interface


I the extremal may continue along the interface
B the extremal may continue into the bottom layer
T the extremal may continue into the top layer
T+B the extremal may continue either into the top or the bottom layers
I+ B the extremal may continue either along the interface or into the bottom layers
I+ T the extremal may continue either along the interface or into the top layer
I+ T + B the extremal may continue either into the top layer, the bottom layer or along the interface
l This is an approximate check if the curves y&r) and y(x) are tangential

where ‘1 is related to A1 by equation (25.2). The proposed computation scheme may best be
illustrated in terms of the block diagram presented in Fig. 5.
Inspection of this figure reveals that the present approach lends itself to a computational
scheme that is similar in principle to the one generally used in the simplified Bishop method
(Bishop, 1955). Namely a trial and error search for the minimal factor of safety, over a three
dimensional space consisting two centre co-ordinates and a radius.
The present procedure is however slightly more laborious than the simplified Bishop one,
since the dependence of the potential slip surface on F,, makes it necessary to re-establish the
potential slip surface in each iteration.
More significantly however, the present method should be compared with the Morgenstern-
Price or Janbu’s generalized methods. While the present methods requires minimization in a
three dimensional space, a task that experience show to be possible using a simple trial and
error technique, the generalized methods of slices require a search for the minimal factor of
safety in a space of n+ I dimensions (n being the number of slices). Since an exhaustive
search is clearly impracticable in actual applications of these methods, one is usually content
with a result obtained after a very limited number of trials, The result is on the unsafe side,
since it does not correspond to the critical conditions.
Even so, the foregoing comparison is incomplete, since in the generalized methods of slices
one has to assume the shape of an entire function (the stress function), while the present
approach is free of any arbitrary assumptions.

SUMMARY AND CONCLUSIONS


Starting with the conventional concepts of limiting equilibrium, and factor of safety with
respect to strength, and applying a rigorous analysis without any arbitrary assumptions,
410 R. BAKER AND M. GARBER

Assumevaluesfor xc. Yc,G.Fs


01 0’o Fs

+
Determine Y(x) according
------w to the basvz theorem
and the corollary

use equation (31) or (32) to obtain


a new estimate of the factor of
safety (F,),,,,

Fig. 5. The computation scheme

it was found possible to establish a number of fundamental results of a very general nature.
The factor of safety is independent of the normal stress distribution on the critical slip
surface. This statement constitutes a basic theorem of the limiting equilibrium approach.
Failure may occur in one of two modes (rotational and translational) which corresponds to
two different families of potential slip surfaces. Two differential equations, that govern the
shape of the slip surfaces in the two modes of failure were established.
It was found that the two modes of failure consists of curves with special geometrical
properties. These geometrical properties insure that the elementary forces adl+(tan +/F,jodl
either pass through a common point (rotational failure), or are parallel to a common direction
(translational failure).
These results are general, and are valid for non-homogeneous, non-isotropic soil with
arbitrary distribution of c, (b, y, U, p, pX and py. It was found, moreover, that only the dis-
tribution of 4 can affect the nature of the family of the potential slip surfaces, all other charac-
teristics of the profile merely enable the selection of the most critical member of the family.
In the case of a profile that consists of homogeneous and isotropic layers, it was found that
the critical slip surfaces may consist of either a series of log-spirals that have a common focus,
or a series of straight lines such that the resultant of the normal and frictional forces are all
colinear.
It was found that a discontinuity of the transition zone type may be a part of an extremal
regardless of its geometrical shape. A discontinuity of the contact plane type may be a part
THEORETICAL ANALYSIS OF THE STABILITY OF SLOPES 411

of an extremal, only if it is of a particular shape. In the first case the extremal connects with
the discontinuity smoothly, while in the second case no such restriction exists. The criteria
for a discontinuity to be a slip surface segment were established.
It was proved that for the homogeneous and isotropic case, without pore water pressure or
external loads, the solutions provided by Rendulic (log-spiral), and Culmann (straight-line)
are not only convenient, but correct. The two methods are related to the two possible modes
of failure.
Finally, a simple computational scheme was suggested for the determination of the factor of
safety and the critical slip surface. This computation scheme is onfy slightly more laborious
than the generally used simplified Bishop method.

ACKNOWLEDGEMENT
The Authors thank Professor M. M. Schiffer of Stanford University for spending time dis-
cussing with us various aspects of the variational analysis presented in this Paper.

REFERENCES
Baker, R. & Garber, M. (1977a). Discussion of On slip surface and slope stability analysis by Chen, W. P. and
Snitbhan, N. S. &i/s and Foundafions 17, No. I, 65-68.
Baker, R. & Garber, M. (1977b). Variational approach to slope stability. Proc. Ninfh fnr. Conf. Soil Mech.
Fdn Engng, Tokyo 2,9- 12.
Bishop, A. W. (1955). The use of slip circle in the stability analysis of slopes. GPotechnique 5, 7-17.
Bolza, 0. (1973). Lectures on the calculus of variations. New York: Chelsea Publishing Company.
Culmann, K. (1866). Die Graph&he Static. Zurich.
Dorfman, A. G. (1965). Variational method of investigating the stability of slopes. Vop. Geoferhniki Trans-
port, No. 9, Moscow.
Fellenius, W. (1936). Calculation of the stability of earth dams. Trans. Second Gong. on Large Dums 4,
Washington.
Garber, M. (1973). Variational method of investigating the stability of slopes. Soil Mechanics and Foundation
Engineering 10, No. I, 77-79.
Janbu, N. (1954). Application of composite slip surface for stability analysis. Proc. Eur. Con/I on Stability of
Earth Slopes, Stockholm 3. 43-50.
Kopacsy, J. (1955). Uber die Bruchflachen und Bruchspannungen in den Erdbauten. Gedenbuchfur DrJ.Jaky.
(Ed. by Szechy, K.) Akadimiai Kiado Budapest, 81-99.
Morgenstern, N. R. & Price, V. E. (1965). The analysis of the stability of general slip surfaces. G&o;otechnique
15, No. I, 79-93.
Rendulic, L. (1935). Ein Beitrag zur Bestimmung der Gleitsicherheit. Buuingenier, No. l9/20.
Revilla, J. & Castillo, E. (1977). The calculus of variations applied to stability of slopes. GPotechnique 27,
No. I, I-II.
Taylor, D. W. (1937). Stability of earth slopes. Contributions /o Soil Mechanics. Boston Society of Civil
Engineers, 337-386.

You might also like