Extensions to the generalized method of slices for stability analysis

Zu-Yu CHEN
Water Conservancy and Hydroelectric Power Scientijc Research Institute, P.O. Box 366, Beijing,
People's Republic of China
AND
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by GEORGEWASHUNIVBF on 05/24/13

N. R. MORGENSTERN
Department of Civil Engineering, University of Alberta, Edmonton, Alta., Canada T6G 2G7
Received September 15, 1981
Accepted June 21, 1982

Extensions are suggested to the generalized method of slices that is commonly used in slope stability analysis. It is shown
that restrictions exist on the assumptions used to make the problem statically determinate. In addition, a numerical procedure
has been developed to find the bounds to the factor of safety, subject to additional requirements of physical admissibility. As
a result of these developments it has been possible to produce a revised computer program that appears to overcome the problems
of convergence experienced by other programs in current use. Results obtained with this new analysis confirm the reliability of
several methods of analysis used in practice.

On propose des dCveloppements A la mCthode des tranches gCnCralisCe qui est couramment utilisCe dans l'analyse de stabilitC
de pentes. On montre qu'il existe des restrictions aux hypothkses utilisCes pour rendre le problbme statiquement dCterminC. De
plus, une procCdure numkrique a CtC dCveloppCe pour trouver les limites du facteur de sCcuritC compte tenu d'exigences supplC-
mentaires dladmissibilitC physique. Suite A ces dkveloppements il a CtC possible d'etablir un programme d'ordinateur qui semble
Climiner les problbmes de convergence rencontrks dans d'autres programmes actuellement en usage. Les rksultats obtenus
For personal use only.

avec cette nouvelle analyse confirme la fiabilitC de plusieurs mithodes de stabilitk utilisCes en -pratique.
-
[Traduit par la revue]
Can. Geotech. J., 20, 104-1 19 (1983)

Introduction 1967; Janbu 1973). Since only the shape of the distribu-
Generalized methods of slices are commonly used to tion of one of the unknown functions is assumed, the
investigate the stability of slopes, particularly when the physical reasonableness cannot be checked until the
section is nonhomogeneous. However, the problem is final solution has been found. The selection of the
statically indeterminate and assumptions are necessary assumed function depends to a large extent on intuition
in order to obtain numerical results. Several methods and experience but may be guided by stress analysis.
have been advocated. For example, Janbu (1954, 1973) Since we are basically confronted with an infinite
made assumptions regarding the location of the point of number of possible choices for the assumed function,
action of the interslice force and Morgenstern and Price after performing several calculations we are still uncer-
(1965) and Spencer (1967, 1973) assumed the shape of tain if some relevant solutions are missing or not.
the distribution of the inclination of the interslice force. The purpose of this paper is to overcome the disad-
Sarma (1973) adopted the distribution of the vertical vantages involved in the currently used methods by the
component of the interslice forces. following.
A survey of the commonly used methods indicates the 1. It will be shown that restrictions that have Geen
following. ignored so far exist on the boundary values of the
1. The assumptions made for the different unknown distributive assumption. The search for the solution to
variables involved in the equilibrium equations do not the equilibrium equations is then undertaken with a
result in much difference in the final factor of safety. group of assumed functions that are fixed at both ends.
This is not surprising when considering that the various 2. A method of sensitivity analysis to explore the
methods are based on the same equilibrium equations influence of the assumed functions associated with
and the unknown variables are interrelated. physically reasonable solutions to the equilibrium equa-
2. The assumptions regarding any unknown variable tions will be developed. An efficient computer program
are not unique. A number of functions which lead to a for this method has been coded.
group of solutions satisfying the equilibrium equations With the help of these two extensions, the bounds on
may be assumed. Some of the solutions should be the factor of safety caused by conditions of physical
rejected due to the requirement for physical admissibil- admissibility can be found. The effort required of the
ity (Morgenstern and Price 1965; Whitman and Bailey user is reduced considerably.
0008-3674183 /OlOlO4- l6$Ol .OO/O
01983 National Research Council of CanadaIConseil national de recherches du Canada
 CHEN AND MORGENSTERN

The general equilibrium equations

Limit equilibrium considerations in slope stability
analysis are based on the following.
1. Principle of equilibrium
The sliding mass is divided into a number of slices.
The requirements for force and moment equilibrium
should be fully satisfied for every individual slice.
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by GEORGEWASHUNIVBF on 05/24/13

2 . Mohr-Columb failure criterion

The Mohr-Columb failure criterion holds along the
failure surface,
[I] T = C' + unrtan 4 '
where unl= normal effective stress on the failure plane,
T = shear stress on the failure plane, C' = effective
cohesion, and 4 ' = effective friction angle. FIG. 1. The generalized failure surface.
3. The factor of safety F
This is defined as that value by which the available where
shear strength parameters must be reduced in order to [81 Ac' = ( Y - Y ~ ' ) / ( -
Y z)
bring the soil mass into a state of limiting equilibrium
along a given slip surface. Hence, and y,' = the ordinate of the point of action of the
effective normal force.
[2] T = Cel + (tan +,')unf The general equilibrium equations for a slice can be
For personal use only.

where expressed as
[3] C,' C1/F dG 4
=
cos (4,' - a + p) -
dx
- sin (4,' -a + p) -
dx
G
[4] tan 4,' = (tan +')IF
and F = the factor of safety.
In addition to these considerations, there are condi- = : :1
- -sin (4,' - a ) + q sin (4,' - a )

tions of physical admissibility (Morgenstern and Price

1965) as follows.
1. The shear force on the vertical surface of any slice
should not exceed the shear strength that can be
dW
- ru - sec a sin 4,'
dx
d
+ C,'
d
sec a cos 4,'
1
mobilized along the surface (Fig. I ) , [ l o ] G sin p = - y -(G
dx
cos p) + -
dx
(y,G cos p)
[5] Fv = [ E rtan +,,' +
CaV1(y - z)]/X > F where G = the resultant total force on the vertical side of
the slice, p = the inclination of G to the horizontal,
dW/dx = the weight of the slice per unit width, r, =
[61 Fve = Fv/F pore pressure ratio (Bishop and Morgenstern 1960),
= [E' tan + CaVer(y- z)l/X > 1 and q = vertical surface load.
Equation [9]can be obtained by substituting
where Fv = factor of safety along the vertical surface of
the slice, Fve = relative factor of safety along the [II.] E = Gcos p
vertical surface of the slice, E' = effective normal force
on the vertical surface, X = shear force on the vertical [12] X = G sin P
surface, tan +,,I, CaV1= the average effective strength into the equation of force equilibrium (eq. [ l o ] ) of
parameter on the vertical surface, tan Cave1= tan Morgenstern and Price's paper (1965). However, the
+,,I, Cay' divided by the factor of safety, y = the signs of the various terms of the two equations are not
ordinate of the slip surface, and z = the ordinate of the identical due to the different positive direction of the
slope surface. coordinate systems that have been used. Equation [9]
2. To avoid tensile stresses the line of action of the can also be established by projecting all the forces acting
resultant effective normal force should not lie outside on the slice to the axis A-A which is inclined at an angle
the vertical surface of the slice, i.e. (Fig. 2a), of ( 4 ' - a ) to the horizontal (Fig. 2b). In this case the
[7] 0 <Ac' < 1 resultant of N and S, the force P , which is inclined at an
 106 CAN. GEOTECH. J. VOL. 20, 1983

angle of 0,' to the normal, contributes nothing to the

force equilibrium equation.
Equation [lo] is obtained by considering the moment
equilibrium about the midpoint of the base of the slice.
[I31 (G + AG) cos (P + APMy + Ay)
- ( ~ +r AYJ - UYI
- G cos P(y - y, + 4Ay) + G (sin @)Ax= 0
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by GEORGEWASHUNIVBF on 05/24/13

Neglecting the small magnitudes of order higher than

AG and Ap, [13] reduces to [lo].
The corresponding boundary conditions are:

where a, b are abscissa values of the ends of the sliding

mass.
If the slip surface terminates at a slope surface that is
not vertical, the slices at both ends are triangles rather
than rectangles. It is then required that the value of p and
For personal use only.

A, be fixed at points A and B. A detailed discussion will

be given in a subsequent section.
By virtue of [14]-[17], [lo] can be integrated:

[18] lb a
G(sin p - tan a cos p) dx = 0
For brevity, [9] is rewritten as
dG dP
[I91 -- tan $,' - G = -p(x) sec +,'
dx dx (b)
FIG.2. The force equilibrium of a slice.
where

dW dW
[20] p(x) = -- sin (6,' - a )
dx
+ q sin (0,' - a) - r, - sec a sin 4,'
dx
+ C,' sec a cos +,'

By substituting the boundary condition of [14], the solution to [19] is

[221 G(x) = - exP [Ix

fan (0.' - a + p) -di]
di
IXa
p(6) sec
: ) I exp [- 1 5
a
tan (0: -a
dP
+ p) -
di
di] d6

where 6 and 6 are dummy variables substituting for x.

Substitution of [I51 into [22] yields

[23] 1
a
p(x)s(x) dx = 0

where

[24] s(x) = sec 4,' exp

I
 CHEN AND MORGENSTERN

Substituting [22] into [18] and integrating by parts, we obtain

[25] 1
a
P(X)S(X)t(X) dx = 0

where

[26] t(x) = 1' (sin P - cos P tan a ) exp [I

5
tan (4.' - a dP di]
+ P) - dt
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by GEORGEWASHUNIVBF on 05/24/13

a a d5

Equations [23] and [25] are governing equations for stress on CB , the vertical side of the end element ACB.
the calculation of the factor of safety. If the stress tensor at point A is determined, P will be
The function p(x), which does not contain P(x), fixed as well.
accounts for the geometrical and physical properties of The stresses can be determined by drawing the Mohr
the slope. The function s(x) takes into account the circle as shown in Fig. 3, in which the general case of the
assumption that will be made for P(x) For an assumed end slice ABC with a sloping surface AB and a vertical
function P(x), the value of F that satisfies [23] and [25] load q is considered. Point A in the Mohr circle
may be found. It has been shown (Chen 1981) that [23] represents the stress state along the surface AB of the
and [25] are reducible to special cases, such as the element. Since
wedge slide analysis, Bishop's simplified method
(Bishop 1955), the logarithmic spiral method (Taylor
1948), and Spencer's method (Spencer 1967) if the [32] 7, = q sin y cos y
appropriate simplifying conditions are considered.
For personal use only.

the angle AOC in the Mohr circle (Fig. 3 b) is equal to y.

Boundary values for P(x) and A, It is required that the Mohr circle in Fig. 3b pass
As noted previously, in order to render the problem through point A, be tangent to the Mohr-Coulomb
statically determinate, an assumption about the side failure line GD, and make the angle ABD equal to a -
force can be made. Morgenstern and Price (1965) y. The circle is therefore unique. The stress state on any
assumed that plane of the element is represented by a point on the
Mohr circle that is the intersection point of the Mohr
[27] tan P = hf (x) circle and a straight line passing through point B and
where f (x) is an assumed function and h is a coefficient parallel to the plane concerned, provided the X axis of
to be determined. the Mohr circle is parallel to the direction of the minor
Janbu (1973) assumed that principal stress of the element. For example, the stress
state of the failure surface AC in Fig. 3a is represented
[28] A, = 3 by the point D in the Mohr circle that makes BD inclined
where at an angle of a - y to AB.
Consider point E in Fig. 3b, which makes the angle
[291 A, = (Y - Y,)/(Y - z ) EOC equal to y (E is on the minus side of the Mohr
and y, = the ordinate of the point of action of the total circle). Point E represents the stress state of a surface in
normal force. the element on which
Basically, A, and hf(x) could be any function.
However, the values of A, and hf (x) at the boundary
points A and B should be fully specified if the slip In Fig. 3b, it can be shown that EB is inclined at an
surface terminates at a surface slope that is not vertical. angle of 90" - y to AB. This indicates that the surface
This marks a distinction from the previously published represented by point E in the Mohr circle is the one in the
work. element that is inclined at an angle (90" - y) to the
Point A is a special point where the ratio of X to E is surface AB. This plane is nothing else but the vertical
equal to the ratio of ,T to ox, i.e. (see Fig. 3), surface BC of the element. Hence the stress state on the
vertical surface should satisfy [33]. In other words, the
side force on the vertical surface BC of the end slice
[30] tan pa = x+a
lim - = ,-
E
T
should be parallel to the surface slope AB of the slice if
ax
the width of the slice is sufficiently small.
where pa = the inclination of the total side force G at the The need for determining the boundary values for
boundary point A, and T,u = the shear and normal P(x) is actually based on the requirement for satisfying
 CAN. GEOTECH. J. VOL. 20, 1983
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by GEORGEWASHUNIVBF on 05/24/13

( 0 )

FIG.3. The determination of the inclination of the end slice. Dliagram of: (a) the end slice; ( b )the Mohr circle of the end slice.

the moment equilibrium. For the end slices at A and B, is some vertical surcharge on the slope surface, i.e.,
since the values of G and y - y, are equal to zero, all C(a) f 0 or q(a) # 0 .
terms of first-order magnitude in [13] are zero. The
second-order small magnitudes can no longer be neglec- Case B
For personal use only.

At the end points A and B, ( 1 ) the soil is cohesionless,

ted as was done for normal slices during the derivation of i.e., C(a) = 0 ; (2) there is no surcharge on the slope
the moment equilibrium in [ l o ] .The consideration of surface, i.e., q(a) = 0 ; and (3) the slip surface is not
second-order magnitudes in [13]will lead to the principle
tangent to the slope
of complementary shear stresses that eventually deter-
mines the boundary values of P. In other words, if the yl(a) - zl(a) 0 +
values of p are taken arbitrarily at A and B, the solution
will correspond to one in which the principle of
complementary shear stresses does not hold. The value
of A, at points A can also be determined due to the fact
that CB (Fig. 3a), the area over which the force E is
applied, is infinitesimally small. The restrictions that fix where y1(a),z'(a)are the derivatives of y(x) and z(x)
the boundary values of P and A, are presented here; with respect to x at point A.
formal demonstrations are contained in detailed work
available from the authors. The numerical procedure
Restriction 1 By linearizing the value of a ( x ) and some other
The resultant total force acting on the vertical side of variables for each slice, Morgenstern and Price (1967)
the end slice is parallel to the surface of the slope if the developed a numerical procedure based on the Newton-
width of the end slice is sufficiently small, i.e., Raphson method for solving differential equations. This
[341 Pa = ya method has been refined in the following aspects to
satisfy the required boundary conditions and to make the
where y, and Pa are the inclination of the slope surface iteration more effective.
and side force G, respectively, at points A. As mentioned before, the boundary values of the side
Restriction 2 force function P(x) should be fixed. This can be done by
The point of action of the total normal force on the taking tan P as
vertical side of the end slice is located at the midpoint in
case A and the lower one-third point in case B if the
width of the slice is sufficiently small, i.e., A, = 6 for where f ( x ) is the assumed function, which is equal to
case A and A, = $ for case B, where case A and case B zero at both ends, h is a coefficient to be determined,
are defined as the following. and fo(x) is another function that has the required
values at both ends, i.e.,
Case A
At the end points A or B, the soil is cohesive or there
 CHEN AND MORGENSTERN 109

not close to zero within the tolerable limit, the next

+
iterative values h l Ah and F1 + AF that are sup-
posed to make Gn and M, close to zero were given by
(Morgenstern and Price 1967) as
Basically, f (x) and fo(x) can be arbitrary functions;
however, in some cases, an inappropriate selection of
f(x) and fo(x) will cause difficulties in the iteration
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by GEORGEWASHUNIVBF on 05/24/13

procedure, as will be discussed subsequently.

The iteration procedure starts by substituting assumed
values (Al and F1)into [23] and [25].If hl and F1make
the values of

[do1 Gn = G@i, F I ) = p(x, XI, F I ) ~ ( xhi,

, Fi) dx

Instead of calculating aM,/dF, dGn/dF, aMn/ah,

dG,/ah by linearization of the function a(x), we pro-
ceed here analytically:
For personal use only.

where

1481 k(x) = - [(g+ q) sin

dW
cos P - ru - sec a cos (P - a ) sin +.' + Cel sec a cos +,' cos (p - a )
dx 1
X cos [(g+
/IF
dW
dx sec a sin A' + C,' sec a cos
q) sin (0.' - a ) - ru - +,'I cos (9,' - a + p)}

dP
1491 -= (cos2 P)f (x)
dh

[ ~ O ID~ = tan +, -:xI=-, [tan+eil- '3

where Di, Dti are coefficients accounting for the possi- represent the values at the right and left side at this point
ble discontinuity of 4,' or a at certain point i at which of discontinuity.
the value of +,Ior a changes abruptly from [ai],. The derivation of [44]-[47] may also be found in
to [4ei']r, [ailL; the superscript r and subscript I detailed work available from the authors.
 CAN. GEOTECH. 1. VOL. 20, 1983

/ I
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by GEORGEWASHUNIVBF on 05/24/13

.
,
case 2 case l
For personal use only.

The orocess of iteration

case I case 2
F A F A
2.030 0.010 2.030 0.010
2.068 -0.090 2.038 0.733
2.058 0.115 2.051 0.673
2.065 -0.036 2.051 0.669
1.947 2.481
2 ,051 1.358
2.049 0.4 6 1
2,057 0.163
2.063 0.002
2.079 -0.334
2.069 -0. 165
2.064 - 0.022
2.107 -0.925
1.995 -0.617
2.059 -0.408
2.068 -0.209
2.065 -0.062
2.048 0. 327
2.058 0.108
.. . ..,
FIG. 4. An example showing the selection off (x) and fo(x).
Since the condition of fixing the boundary value by same sign throughout the region (a, b). Figure 4 shows
[38] and [39] is not used in the derivation of [MI-[47], an example in which two different choices off (x) and
the latter are applicable to cases where the boundary fo(x) consistent with the same tan p are used. In case 2,
values of P(x) are not fixed. If it is not desired to fix the the value of tan P will increase or decrease as a whole
boundary value of P(x), for example, when the case of with an increment of X, while in case 1 only part of the
f (x) = 1 of the Morgenstern-Price method is evaluated, value of tan P will be increased and the rest will be
then it is simple to take f (x) = 1 and fo(x) = 0. decrcased. As a result, G, and M , may not be monotonic
Equations [MI-[47] still apply. in A and some difficulties may be experienced in
The problem of convergence is a common concern iteration. The table in Fig. 4 shows that case 2
during the iteration. The refinement for the computation successfully converges but case 1 fails to converge.
of the derivatives described previously, in combination (2) The assumed values of F1and hl are estimated as
with the following techniques, facilitates convergence. closely as possible to the final solution. Some of the
(1)The assumed function f (x)is selected to have the simplified methods such as that of Fellenius could be
 CHEN AND MORGENSTERN 111

used for the estimation of the factor of safety. Chen numerical procedure has converged in every case
(1981) suggested a simplified equation that is applicable investigated.
to generalized failure surfaces and gives reasonable
estimates without the need of iteration. Upper and lower bound solutions for factor of safety
For estimating the value of XI, it is suggested that the By assuming different side force functions, a number
average value of P(x) be taken as equal to the average of solutions for the factor of safety can be found.
inclination of the slip surface, i.e., However, the physically acceptable solutions are lim-
ited within reasonably narrow bounds. In order to find
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by GEORGEWASHUNIVBF on 05/24/13

the bound, it is advantageous to investigate how the

value of F changes with a variation of side force function
By substituting [35] into [52], it is easy to find the P(x) that is required to be physically reasonable.
value of X1 for the first iteration. Suppose there is a set of solutions F* and P*(x) that
Another way of estimating the values of F1and X1 is satisfies the force and moment equilibrium equations
to estimate a number of sets of F1and h l that possibly [23] and [25], i.e.,
cover the accurate values and substitute them into [42]
and [43]. The accurate values of F and X, which make
G, and M, in [42] and [43] equal to zero, will in turn
result in zero values of A F and Ah. The set of F1and X1
that gives the minimum value of AX2 + A F ~is therefore
very likely close to the accurate value of F and A and
makes the problem converge effectively. If a neighbouring solution of F* + AF, P* + Ap
With the refinements described in this section, the also satisfies [23] and [25], we have
For personal use only.

p(F* + AF, P* + AP)s(F* + AF, P* + AP) dx - p(F*, P*)s(F*, P*) dx = 0

p(F* + AF, P* + AP)s(F* + AF, P* + AP)t(F* + AF, P* + AP) dx

- I: p(F*, P*)s(F*, P*)t(F*, P*) dx = 0

Equations [55] and [56] indicate that the variation of G, and M, due to the increment of F * and P*(x) should
+
be zero if F * + A F and P* AP is another set of solutions of [23] and [25].
Let AP(x) take the form

where E is a coefficient that makes Ap sufficiently small in comparison to the corresponding values of P. It can
be shown that

[59] AM, = KmfAF+ K,,E + o(AF, E)

where

t sec2 JI,' - dp df] dx

d'e' -
d F df
 112 CAN. GEOTECH. J. VOL. 20, 1983

+ Ix
a
cos 4.' sec a sec JIet exp [I5
a
tan +e di d(
I
q(<)d t dx

which indicates that if F* + AF and P* + s q ( x ) is a set

S
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by GEORGEWASHUNIVBF on 05/24/13

[641 Bi = - C [tan
i=l
+ei' q(xi)l; + tan +,,' -da)
of solutions of 1231 and 1251,
- - q.(.x.) must be selected in
s such a way that [ti81 holds. Once the value of q ( x )
1
[651 Bti = - i= 1 [ti tan $,it q(xi)Ii satisfying [68] has been found, AF can be obtained by
solving [66] or [67] with a specified value of E that
and o(AF, E ) represents a small magnitude of order makes q ( x ) sufficiently small. The smaller the value of E
higher than and E . The details of the derivation of is, the more accurate the new solution will be. After
[60]-[63] can also be obtained from the authors. performing this procedure several times, it is possible to
substituting [581 and [591 into [551 and [ 5 6 ] , and obtain a new solution that is distinct from the original
neglecting o(AF, E ) , we have solution and satisfies [23] and [25] with the required
accuracy.
For personal use only.

Substitution of [69] into [66] yields and increased on both sides. Thus, the first set of q l ( x )
and q 2 ( x ) will result in an upward or downward
movement of the peak value of P(x).
The second set of q l ( x ) and q 2 ( x ) (curves 3 and 4 in
Fig. 5 ) are selected as cubic polynomial functions with
the peak values on the left and right sides respectively.
The function P(x) is positive on the left side and negative
Basically, q l ( x ) and q 2 ( x )can be arbitrary functions. on the right side if E is positive and m is near 1. This
results in a movement of the peak values of P(x) to the
Attention should be paid to the case where P*(x) is zero
at a certain point. In that case, in order to ensure that A p left. A negative value of E will result in a movement of
is smaller than P*(x), it is advantageous to specify that the peak value to the right.
q l ( x ) and q 2 ( x ) are also equal to zero at that point. Because the behaviour of F,,, the factor of safety
For our purpose here, two sets of q ,(x) and q 2 ( x )may along the vertical surface of the slices, is related to P(x),
be used. the movement of the peak value of P(x) will eventually
The first set of q l ( x )and q 2 ( x )is selected as an elliptic make F,, reach the bound beyond which the value of F,,
and a parabolic function (curves 1 and 2 in Fig. 5 ) . The is less than unity and the requirement of physical
function q ( x ) has positive values in the middle part of admissibility for Fve is not satisfied. From the equation
the region ( a , b) and negative values of both sides if E is
positive and m is less than 2 . This results in an increase / ^ G(sin p - cos P tan a ) dx
of P(x) in the middle part and a decrease on both sides. If
E is negative, P(x) will be decreased in the middle part
[74] y, = y + 'a G cos p
 CHEN AND MORGENSTERN
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by GEORGEWASHUNIVBF on 05/24/13

5911 Unll Wobphl c' cu

Iktr/m3 I @' (kP0) (kPd
2
I 3;: 23 5 27 10 06
19.64
3 17.27 20 7 19 26.35

FIG.6. Example 1.

2 0.5
Curve I q,(x)--( I-(2(x-o)/(b-01-1) 1
For personal use only.

Curve 2 q2(xl=-8(x-b) (x-a)/(b-of

Curve 3 q , ( X I = 6 . 7 5 (x-b12 (x-a~/(b-a)~

Curve 4 g2 ( X I ' 6 . 7 5 ( x - b l ( x - a 1 2 /(b-a?

A fi(xI=f V(x1

FIG.5. The integration function ~ ( x ) .

it can be inferred that an increase of the average value of
P(x) in the region ( a , x ) will result in an increase of the
value of y, at x and in turn a decrease of A,. Therefore, if
q l ( x )and q 2 ( x )are selected as curves 3 and 4, the value
of A,' will probably decrease with a positive value of E
and increase with a negative value of E.
In the examples that follow it will be shown how the
value of A,' or F,, moves to the bound by an appropriate
selection of the set of integration curves q l ( x ) ,q 2 ( x )and
the value of E.

Illustrative examples
A computer program has been coded at the University
of Alberta to undertake the calculation and a plotting
program has also been added to assist in interpreting the
computations.
Example I -20 1 I I I I I
Find the factor of safety of the slope shown in Fig. 6 DO 80 60 40 20 0
for the circular failure surface 1. Also find the factor of Horizontal Distance (m )

safety for circular failure surface 2 under drained FIG.7. The peak value of P(x) moves upward (example 1).
conditions. Curves 1 and 2 were employed; E = 0.025.:
 114 CAN. GEOTECH. J. VOL. 20. 1983
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by GEORGEWASHUNIVBF on 05/24/13
For personal use only.

After 3 ~ntegrotions

--
X

%.

H o r ~ z o n t a l Distance (m 1
FIG.8. The peak value of P(x) moves downward (example
1). Curves 1 and 2 were employed; E = -0.025. Horizontal Distance (m 1
FIG.9. The peak value of P(x) moves to the left (example 1).
Failure s u f a c e I Curves 3 and 4 were employed; E = -0.025.
The analysis using effective strength parameters was
first performed. The failure surface terminates at the with a positive E value of 0.025, the value of F., will
horizontal slope at both ends. A sine function is selected eventually go below unity in the middle part. Figure 7
for f (x), and fo(x) is taken as zero. The values of F and h shows the progress of the integration. After 12 integra-
were found to be 1.465 and 0.360 respectively. The tions, the lowest value of F,, was less than unity; the
curves of F,, and A,' associated with the original case in integration in this direction was then stopped. The
Fig. 7 show that this is a physically reasonable solution. correspondingvalues of the factor of safety change from
However, the value of F,, approaches the F,, = 1 line in 1.465 to 1.471. An integration in the opposite direction
the middle part of the region. If the value of P(x) is can be made by taking E as -0.05 as shown in Fig. 8.
increased inthe middle pa* by employing curves 1 and 2 .-..- of
The valne . F,,.- increases in the middle part and
 CHEN AND MORGENSTERN
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by GEORGEWASHUNIVBF on 05/24/13

.-
2 curve I and 2 12 0 05 0.940
3 curve I and 2 12 -0.05 0.940
4 c u r v e 3 and 4 I2 005 0 940
curve 3 and 4 12 -0.05 0.940

FIG.11. Analysis by the 4 = 0 method (example 2).

In all of the foregoing cases, the values of A,' were

reasonable.
So far, the assumed side force function P(x) covered a
large variation. Each time, it eventually reached the
bound beyond which the solution became unreasonable.
For personal use only.

No factor of safety higher than 1.471 or lower than

1.444 was found. The bounds of the factor of safety
were therefore determined to be 1.444 and 1.471.
As a check on the program, the r$ = 0 method using
the same unit weight and a circular failure surface was
analysed by a similar integration procedure. The value
of the factor of safety in all cases was 0.940 (Fig. 11).
This is consistent with the basic principle that, for the +
= 0 method with a circular slip surface, the resulting
factor of safety is independent of the assumptions made
for the interslice force.
Failure sullface 2
Failure surface 2 exits at the toe of a slope. The value
of P(x) at the left end should be fixed at 3 lo, which is the
inclination of the slope surface at which the slip surface
I I exits (Fig. 6). The original solution gave a value for the
-20 1 I I I
factor of safety of 1.682, but is associated with a
100 80 60 40 20 0
Horizontal Distance (m ) distribution of F,, that is smaller than unity on the left
part due to the relatively high value of P(x) in that area
FIG.10. The peak value of P(x) moves to the right (example (Fig. 12). Curves 3 and 4 were employed to reduce the
1). Curves 3 and 4 were employed; E = 0.04.
value of P(x) on the left part with a value of E of 0.025.
After 12 integrations, the region lower than the F,, = 1
decreases on both sides. The F,, curve eventually line was reduced appreciably; the factor of safety
touches the F,, = 1 line after 12 integrations. The factor became 1.639. Although a further effort to make that
of safety changes from 1.465 to 1.455. unreasonable region even smaller is possible, the factor
Curves 3 and 4 are used to move the peak value of of safety will likely remain around 1.639.
P(x) to the right and left by taking E equal to 0.04 and
-0.025 respectively (Figs. 9 and 10). The values of the Example 2
factor of safety associated with the solutions touching Find the value of the factor of safety for the
the F,, = 1 line were 1.444 and 1.470 after nine and wedge-shaped slide shown in Fig. 13.
three integrations respectively. A similar integration procedure is performed for this
 CAN. GEOTECH. J. VOL. 20. 1983
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by GEORGEWASHUNIVBF on 05/24/13

Original Fz1.682
5-

4 -
Y
/
/ //-
/ //'
,I A f t e r 12 integrations F = 1.639 I / I / / /
For personal use only.

-
X

Q&

'0
- 2 0 70 60 50 40

Horizontal
30

Distance ( m )
20 D 0

FIG. 12. The analysis for failure surface 2 (example 2). Curves 3 and 4 were employed; E = 0.025.

problem. Figure 14 shows the various critical F,, curves

that touch the F,, = 1 line. The factor of safety ranged
between 1.677 and 1.696. It is interesting to note that 6 ,
the value of p at point A at which the inclination of the
failure surface changed abruptly, ranged from 17-23".
Near the right end, the value of A,' is not satisfactory.
As was indicated by Spencer (1973), the location of the
thrust line is always unreasonable near the crown of a
cohesive slope unless a tension crack is considered.
Spencer has suggested a method to determine the tension
crack associated with reasonable values of A,' near the
crown. A further investigation of the problem in
combination with the method developed here would be
FIG. 13. An example of the wedge slide. of interest, but is set aside for further study.
 CHEN AND MORGENSTERN
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by GEORGEWASHUNIVBF on 05/24/13
For personal use only.

"r I 5
curve cum

-20 1 I I
I I
I I I
125 100 75 50 25 0 -2 5
Horizontal Distance (rn )

Curve Integration function Time of integration E F

I original 1.684
2 I ond 2
3 I and 2
4 3 and 4
5 3 and 4

FIG. 14. The analysis for the wedge slide (example 2).
 CAN. GEOTECH. .I.
VOL. 20, 1983

Soil Type
Brown highly weathered
columnar jointed till

w Brown highly weathered

silty clayshale
Stiff grey fissured
sdty clayshale A Piezometric Surface

Alt. layers dark qrev Failure Surface

U Bentonitic c~ayshaie,
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by GEORGEWASHUNIVBF on 05/24/13

& Slope Indicator Cut.off

It. grey sandstone,
carb. clayshale and 300 -.E -5
275
.-8
5;
250 2 n
LU m
0 50 190 150 200 250 300 350 400

Horizontal Distance (rn)

Failure surface 1 2 3 4 5 6 7 8
6' 23' 41' 8O 8O 8O 8O 8O 8O
C 0 0 0 0 0 0 0 0

Moteria l Wet d e n s i t y Saturated density

( k N rn3)
For personal use only.

Til I I 843
Brown
weathered
silty
clay shale 17.55
Grey s i l t y
clay shale 19.8 1
Grey
bentonitic
clay s h a l e 2 1.97
Brown
weathered
clay shale 17.55

FIG. 15. The Edgerton slide (example 3).

Example 3 . Analysis of the Edgerton slide (Fig.15) In addition, an extra assumption must be made to
The Edgerton slide was analysed by Thomson and render the problem statically determinate.
Tweedie (1978) and their case 2 is re-analysed here. A It has been shown formally here that the assumptions
factor of safety of 0.958 was found originally. It is of the inclinations of the interslice forces or the location
interesting to note that in this case the value of A,' rather of the thrust line are subject to certain restrictions at the
than F,, is not satisfactory in the right-hand part of the ends that have been ignored in previous work. Guidance
region (Fig. 16). In order to reduce this unreasonable for satisfying these restrictions are given. A modified
region, curves 3 and 4 were employed with a value of E iterative procedure has been developed that to date has
of 0.015. After 9-12 integrations, the values of A,' over not encountered the convergence problems that occa-
most of the region were reasonable. The corresponding sionally affect other programs in common use.
factor of safety is around 0.960. The factor of safety in any particular problem is not
unique. A procedure has been developed for exploring
formally the bounds of the factor of safety within the
Conclusions limits of physical admissibility. It has been shown that,
The generalized method of slices employed in the consistent with earlier studies, the variation in the factor
analysis of slope stability requires that (1) all conditions of safety when subjected to conditions of physical
of equilibrium be satisfied, (2) the Mohr-Coulomb admissibility is small for all practical purposes. This
failure criterion be satisfied for a specific definition of new analysis confirms the view that variations in the
the factory of safety, and (3) certain conditions of factor of safety between several methods in common use
physical admissibility not be violated. are of little practical significance.
 CHEN AND MORGENSTERN
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by GEORGEWASHUNIVBF on 05/24/13

After 9 integrations

z 3-

-9 1 --- ------ &-F,-

=I
---
Line

07
0 150 300 450 600 760 960 1050 1200 13'50 1500
For personal use only.

0 150 300 450 600 750 900 1050 1200 1350 1500
Horizontal Distonce (rn )

FIG. 16. Search for a reasonable solution (example 3). Curves 3 and 4 were employed; E = 0.015.

A copy of the new computer program is available 1973. Slope stability computations. In Embankment
from the authors. dam engineering. Edited by R. C. Hirschfeld and S. J.
Poulos. Wiley, New York, NY, pp. 47-86.
Acknowledgements MORGENSTERN, N. R., and PRICE,V.E. 1965. The analysis of
The authors wish to thank Prof. Y. N. Chen who has the stability of general slip surfaces. GCotechnique, 15, pp.
carefully investigated and discussed the mathematical 79-93.
deviations. Grateful appreciation is given to Dr. J. 1967. A numerical method for solving the equations of
Simmons for his helpful discussion of the manuscript stability of general slip surfaces. computer ~oumal,9, pp.
and Dr. S. Thomson for his assistance in preparing the 388-393.
material from the Edgerton slide. All the program SARMA, S. K. 1973. Stability analysis of embankments and
slopes. GCotechnique, 23, pp. 423-433.
coding was performed with valuable help from Mr. R. SPENCER,E. 1967. A method of analysis of stability of
Howells. embankments assuming parallel inter-slice forces. GCotech-
BISHOP,A. W. 1955. The use of the slip circle in the stability nique, 17, pp. 11-26.
analysis of slopes. GCotechnique, 5, pp. 7-17. 1973. Thrust line criterion in embankment stability
BISHOP,A. W., and MORGENSTERN, N. R. 1960. Stability analysis. GCotechnigue, 23, pp. 85- 100.
coefficients for earth slopes. GCotechnique, 10, pp. TAYLOR, D. W. 1948. Fundamentals of soil mechanics. John
129-150. Wiley and Sons, New York, NY.
THOMSON, S., ~ ~ ~ T W E ER. D W.
I E ,1978. The Edgerton slide.
CHEN,Z. 1981. On the side force in limit equilibrium analysis
Canadian Geotechnical Journal, 15, pp. 510-521.
of slopes. Internal report, Department of Civil Engineering,
WHITMAN, R. V., and BAILEY,W. A. 1967. Use of computers
University of Alberta, Edmonton, Alta.
for slope stability analysis. Journal of the Soil Mechanics
JANBU,N. 1954. Application of composite slip surfaces for and Foundations Division, ASCE, 93(SM4), pp. 475-498.
stability analysis. Proceedings, European Conference on
Stability of Earth Slopes, Sweden, 3, pp. 43-49.