Cavounidis,S. (1987). Giotechnique37, No.

2, 207-210 TECHNICAL NOTE

On the ratio of factors of safety in slope stability analyses

KEYWORDS: analysis; computation; slopes; stability. Some of the analyses of slope stability in three dimensions, for c, cp soil conditions, produce three-dimensional factors of safety that, in some cases, are smaller than the corresponding twodimensional factors. If the comparison is to be meaningful it should be between minimum factors. In this case it is proven that the threedimensional factor is always greater than or equal to the two-dimensional factor. Furthermore the same holds, a fortiori, for three-dimensional factors corresponding to a surface generated to include the two-dimensional critical line. Arguments about the influence of normal forces on the slip surface in two and three dimensions do not provide valid explanations about how the threedimensional factor of safety can be smaller than the two-dimensional factor. Instead, the reason for this error should rather be attributed to some of the simplifying assumptions that some methods employ.

INTRODUCTION Recently increased interest has been shown in treating the stability analysis of slopes as a threedimensional problem, as it really is. This interest is justified because it may constitute a refinement that would increase the accuracy of the analysis and consequently lead to more economic and safer design. Also, the current computational capacities provide greatly increased possibilities for otherwise tedious and time-consuming solutions. The undrained stability of slopes in three dimensions has been examined by Baligh & Azzouz (1975) for cylindrical slides with conical or ellipsoidal ends. Recently an analytical solution for cylindrical slip surfaces with plane ends has been proposed by Gens, Hutchinson & Cavounidis (1987), some results of which have also been presented at the 11th International Conference on Soil Mechanics and Foundation

Engineering (Cavounidis, 1985). In these analyses the ratio of the factor of safety in three dimensions (FJ to the factor of safety in two dimensions (F,) has been found to be always greater than unity but decreases when the width of the slide increases, tending to unity for infinite width. This is not always the case in the analyses presented for c, cp soil conditions. In one of the early studies of this kind Hovland (1977) extended the ordinary method of slices to three dimensions and suggested that in certain cases the ratio F,/F, can be smaller than unity. This result has been questioned by Azzouz & Baligh (1978). Likewise Chen & Chameau (1983) in what resembles a threedimensional version of Spencer s (1967) method is smaller showed results where the ratio FJF, than unity. In discussing their results they suggested that cohesionless material may, under certain circumstances, lead to such a result. Their conclusion was strongly questioned by Hutchinson & Sarma (1985). In reply to this criticism Chen & Chameau (1985) presented an explanation based on the inclination of the normal force on the base of a column for the three-dimensional case compared with the two-dimensional case. Reference to this argument will be made later. The ratio F,/F, will be discussed here and it will be argued that it must always be greater than or equal to unity.

THE PROBLEM Certain qualifications and clarification need to be made from the outset to avoid confusion about the meaning of certain terms and the context in which they are used. (a) The arguments proposed are solely in terms of limit equilibrium analyses. (b) The twodimensional factor of safety Fzmin is computed in the worst cross-section of a slope, i.e. it is the minimum factor that can be achieved, using a certain method, not on a particular cross-section but on any crosssection of a slope to give the Fzmin value of that slope. Given the geometric characteristics and the material properties there is only one Fzrnin. 207

Discussion on this Technical Note closes on 1 October 1987. For further details see p. ii. * Consultant, Athens.

208

TECHNICAL

NOTES

(c) Any

three-dimensional expressed as
F =

factor

can

be

jzR dz + P(E)
sz D dz

of safety which correspond sections of S, This means that F, = R,JD,

to lines S on other

(1)

< Ri/Di

(4)

where R and D are respectively resisting and driving moments per unit length. (R = R(x, y) and D = D(x, y).) P(E) is the additional resistance due to side forces and end resistance and axis z is perpendicular to the direction of movement (parallel to the axis of rotation). Equation (1) can be written in discrete form as F 3 = ER AZ + P(E) CD AZ (2)

where R and D are respectively the resisting and the driving moments and the index 0 indicates the section giving the line S, while index i indicates any other section. From equation (4) &ID, and, since all R and simple algebra G R,lD, D values are positive, (5) by

R ,<&+RI D, DO
0

(6)
(7) = F, and

three-dimensional factor of safety F3min is not uniquely defined by the slope and the material properties. In addition the side boundaries (width) and the shape of the three-dimensional slip surface (e.g. a cylinder with ellipsoidal ends) have to be given. method for infinite (4 The three-dimensional slide width is virtually the same as the corresponding two-dimensional method. Both Hovland (1977) and Chen & Chameau (1983) conformed to this. Any comparison has to take this into account or to use a twodimensional method that gives even lower factors of safety. Otherwise, a very poor twodimensional method could always be used-or even devised-which gives high factors of safety and these could be compared with the three-dimensional factors computed using a good three-dimensional method giving relatively low factors. (f) The critical three-dimensional surface does not, in general, contain the two-dimensional critical slip line. (g) Any meaningful comparison of factors of safety is between F3min and Fzmin. The proposition to prove is that the ratio of factors of safety is greater than or equal to unity, i.e.

(4 The minimum

<

RO+ CR,
(4) R,/D, G F3min

DO D, + CDi but according to equation R, + ZRi D, + CDi Then FZ G Fjrnin

(9)

but F, corresponding to S, is not the minimum for the slope. Instead, as already defined, Fzmin corresponds to line L, which is not a line on S, Consequently Fzmin G FZ and thus from inequalities (9) and (10) (11) (10)

Fzmin G F3min

PROOF Assume that F3min has been found and that it corresponds to surface S, . Also assume that Fzmin corresponds to line L, where in general L, is not a line on S, . Let S, be a line on a section of S, (a section taken parallel to the movement, i.e. normal to the axis of rotation) such that F, corresponding to S, is the smallest of all the two-dimensional factors

USUAL PRACTICE Very often, the three-dimensional surface considered is that which corresponds and contains the two-dimensional slip line (see for example Chen & Chameau (1983)), i.e. initially line L, corresponding to Flmin is found and then surface L, is generated as a surface with the given shape and boundaries (in accordance with the method and the geometry of the problem) and containing line L, . Then L, corresponds to a factor F, L, is in general not the same as S, which corresponds to F3min. Consequently

and thus F, 2 Fzmin (13)

in accordance with inequality (11). The same result can be obtained directly. If index 0 refers to the critical cross-section for two-

TECHNICAL

NOTES

209

Fl

Fig. 1. Free-body diagram (with assumptions) (after Chen 81 Chameau (1983))

dimensional cross-section

analysis then

and

index

i to any other (14) (15) by simple

normal is

force N at the base of the column

(Fig. 1)

Fzrnin = &ID, RJDi 2 R,JDe Since all R and D values are positive, algebra

(1 + tan a_, + tan u,,) ~~

(20)

for the three-dimensional case while for the twodimensional case, for unit thickness
W

R, -a-

+Ro D, +Do

Ro Do

(16)
(17)

Nm = (1 + tan* a,,)ri2
because aYz = 0. They then reason that driving moments produced by the weight of the soil columns with unit thickness and the same height are the same in both cases. However, N,, is less than N,, and thus the resisting moment is smaller in the threedimensional case , i.e. this is why the threedimensional factor of safety may sometimes be smaller than the two-dimensional factor. This reasoning is based on an incorrect premise. N,, is by definition the normal force at the base of a column (slice) on the critical crosssection for the two-dimensional analysis which corresponds to Fzmin. In any other non-critical cross-section of a slope any soil column will have a force N acting normally to its base, but such forces N are not relevant to a two-dimensional

Ro+CRi,R, D, + ZDi D,
but
F ,

RO+ ERi

3 D, + CDi
and consequently from inequalities and equation (14) F3 2 Fzrnin

(18)
(17) and (18) (19)

NORMAL FORCE In their reply to Hutchinson & Sarma s (1985) discussion, Chen & Chameau (1985) state that the

210

TECHNICAL

NOTES

analysis, i.e. N,,, as defined, does not correspond to any cross-section but only to the worst crosssection: for Chen and Chameau s geometry this is any cross-section of the cylindrical part. In that part aYz = 0 and thus N,, = N,, numerically for unit thickness. In this case W = W,. When, however, ayz # 0 there is no N,, as defined, because N,, corresponds to the worst crosssection and not to any cross-section, i.e. the comparison of factors of safety should not be between the minimum three-dimensional factor and a twodimensional factor corresponding to, for example, a cross-section of the ellipsoidal ends, which is obviously not a critical cross-section in two dimensions. In other cross-sections i, not on the cylindrical part or, more generally, not on the critical crosssection for two-dimensional analysis, N,ni may be but at the same time K must be less than N,,, less than W,. Chen and Chameau s statement, quoted earlier, does not hold because if it did then there would be cross-sections i (i # 0) where RJD, < R,lD, but such a cross-section i then would be the critical two-dimensional cross-section and not crosssection 0, which is contrary to the initial assumption. In the cases where F, < F, the error should be traced to the particular assumptions of the method that produces such a result. Simplifications concerning interslice side forces or lateral forces on the slip surface (on the base of the columns), as seems to be the case discussed in Chen & Chameau (1983), may lead to erroneous decreased values of the resisting moment when the base of a column is laterally inclined.

These conclusions were derived using simple algebra and hold in general for the minimum two- and three-dimensional factors of safety. A comparison between other than the minimum factors may produce incorrect impressions concerning the ratio of factors of safety. ratios which are Methods that give F,/F, smaller than unity either compare inappropriate factors or, more probably, contain simplifying assumptions which neglect important aspects of the problem. Finally simplifying assumptions should be used as a (tested) simplification of a more rigorous method (see for example Bishop s (1955) modified and rigorous method) and not as a way to solve a problem without having any basis for evaluating the effects of the simplifications.

REFERENCES Azzouz, A. S. & Baligh, M. M. (1978). Discussion

on Three-dimensional slope stability analysis method. J. Geotech. Engng Div. Am. Sot. Ciu. Engrs 104,

GT9, 12061208. Baligh, M. M. & Azzouz, A. S. (1975). End effects on stability of cohesive slopes. J. Geotech. Engng Div. Am. Sot. Ciu. Engrs 101,CT1 1, 1105-l 117. Bishop, A. W. (1955). The use of the slip circle in the
stability analysis of slopes. Gtotechnique 5, No. 1, 7-17. Cavounidis, S. (1985). Geologic aspects of slope stability problems. Proc. 11th Int. Conf: Soil Mech. Fdn Engng, San Francisco 5. Chen, R. H. & Chameau, J.-L. (1983). Threedimensional limit equilibrium analysis of slopes. GCotechnique 33, No. 1,314. Chen, R. H. & Chameau, J.-L. (1985). Discussion on Three-dimensional limit equilibrium analysis of slopes. GCotechnique 35, No. 2, 215-216. Gens, A., Hutchinson, J. N. & Cavounidis, S. (1987). Three dimensional analyses of slides in cohesive soils. Submitted to Gtotechnique. Hovland, H. J. (1977). Three-dimensional slope stability analysis method. J. Geotech. Engng Div. Am. Sot. Cio. Engrs 103, GT9,971-986. Hutchinson, J. N. & Sarma, S. K. (1985). Discussion on Three-dimensional limit equilibrium analysis of slopes. GCotechnique 35, No. 2,215. Spencer, E. (1967). A method of analysis of the stability of embankments assuming parallel inter-slice forces. Giotechnique 17, No. 1, 1 l-26.

CONCLUSIONS The three-dimensional factor of safety of a slope is always greater than the two-dimensional factor of the same slope. The three-dimensional factor of safety corresponding to a surface which includes the critical two-dimensional line is even greater than the two-dimensional factor since it, in general, is not even the minimum threedimensional factor.