Professional Documents
Culture Documents
,, Z J - It Was Observed That For A - The Presented Example Problem Dem
,, Z J - It Was Observed That For A - The Presented Example Problem Dem
,, Z J - It Was Observed That For A - The Presented Example Problem Dem
SLOPE-STABILITY ANALYSIS
By Dov Leshchinsky, 1 Member, ASCE, and Ching-Chuan Huang 2
Downloaded from ascelibrary.org by UNIVERSIDADE FEDERAL DO RIO on 04/22/14. Copyright ASCE. For personal use only; all rights reserved.
INTRODUCTION
Natural and manmade slopes are often inhomogeneous and their ge-
ometry varies in space, even along short distances. Therefore, when slides
occur their pattern assumes three-dimensional geometry. It is common,
however, to assess the margin of safety through a 2-D idealization of the
slope; i.e., an "equivalent" plane-strain problem is postulated and analyzed.
The 2-D simplification is to a large extent intuitive, typically corresponding
to the worst-case scenario. Happily, the end result is usually conservative,
though more expensive, since 3-D effects tend to increase stability. How-
ever, ignoring the 3-D effects in research may lead to unsafe conclusions.
For example, in postfailure studies the exclusion of the third dimension in
back-analysis may imply that the strength of the soil is higher than it actually
was; i.e., based on a 2-D back-analysis of the traced idealized
2-D slip surface, the added stability due to end effects is attributed to the soil
shear strength. A second example has to do with laboratory studies of 2-D
earth structure models at failure. These physical models are contained within
a rigid box of limited breadth. Friction and adhesion are likely to develop
between the soil and the sidewalls of the box, thus enhancing the perfor-
mance of the structure as compared to the ideal 2-D case and possibly leading
to overconfidence. This paper presents a generalized 3-D limit-equilibrium-
analysis method. For a general (but symmetrical) slip surface specified by
the user, the corresponding safety factor can be determined with the si-
multaneous satisfaction of all limiting-equilibrium equations. Subsequently,
1
Assoc. Prof, of Civ. Engrg., Univ. of Delaware, Newark, D E 19716.
2
Assoc. Prof., Dept. of Civ. Engrg., Cheng-Kung Univ., Tainan, Taiwan; for-
merly, Postdoctoral Fellow, Dept. of Civ. Engrg., Univ. of Delaware, Newark, D E .
Note. Discussion open until April 1,1993. To extend the closing date one month,
a written request must be filed with the ASCE Manager of Journals. The manuscript
for this paper was submitted for review and possible publication on May 17, 1991.
This paper is part of the Journal of Geotechnical Engineering, Vol. 118, No. 11,
November, 1992. ©ASCE, ISSN 0733/9410792/0011-1748/$1.00 + $.15 per page.
Paper No. 1909.
1748
FORMULATION
FIG. 1. 3-D Sliding Surface: (a) Global Geometry; (b) Differential Area; (c) Sym-
metrical Problem
= 0 (2)
B= (6)
{i + ^Y(z!+wj
and HX, VX, and MY = the equilibrium equations for the horizontal forces
in the x-axis direction, vertical forces in the z-axis direction and moments
about the y-axis, respectively; / = the soil's layer number (/ = 1,2. . . m);
Dt = area of the slip surface, located within layer /, as projected on the x-
y plane [Fig. 1(a)]; z„, zy = partial derivatives of z with respect to x and
y, respectively \zx = dz/dx, zy = dzldy); Z = Z(x, y) = elevation of slope
surface; and 7 = the weighted-average unit weight of the layers of soils in
column (Z-z).
From (l)-(3) it is clear that F = a function of z(x, y), u(x, y), and %{x,
y). Hence, F = a functional. The limiting-equilibrium analysis seeks the
minimum value of F, that is, the factor of safety Fs. To achieve this, the
functions z(x, y), a(x, y), and Q(x, y) that realize Fs = min(.F) must be
known. Finding these critical functions can be done through the variational
calculus by using any one of the three equilibrium equations to minimize
F. However, the other two equilibrium equations must also be satisfied for
F = Fs and its associated z(x, y),a(x,y), and Q(x, y). Therefore, the other
two equations are considered constraints. Using HX to define F, and VZ
and MY as constraints, and expanding to layered soil the approach presented
by Leshchinsky et al. (1985) [who extended the 2-D procedure introduced
by Baker and Garber (1978)], one can show the problem to be equivalent
to the minimization of an auxiliary functional, G, defined as
G = X I g dx dy (7)
1=1 J JDi
where
g = [c, + (ex - u)ty,]B{l + \(zx + ftzr) + |x[z - x(zx + ilzy)]}
+ F > [ z v - \ + (JL(X + zzx)} + ^(Z - z)(\ - iix)} (8)
1751
*L-JL(°L)-±(1L)=O (9)
K
dft dx \dilj dy \my) '
--•f(*L)--f(*L)=o dvo)
do- dx \doJ dy \d(jyJ '
^ d / 3 g \ d / d S \ = Q
v
dz dx \dzj dy \dzyJ '
where ftx = 3ft/to, Q,y = 3ft/5y, ax = da/dx and <ry = d&ldy. Hence, the
application of Euler's equations should enable one to determine fl(x, y)
and cr(;t, y).
Combining (9) with (8), substituting xc = X/u. and zc = -1/u., and
rearranging the terms yield
z
ft = tan 6 = y(Xc - *) - zxzy(z - ze)
zx(xc - x) + (z - zc) + Zy\z - zc)
Combining (10) with (8) will yield the differential equation for the slip
surface z(x, y). Leshchinsky et al. (1985) and Leshchinsky and Baker (1986)
have successfully solved this equation for homogeneous soil. The surface
then is a 3-D generalization of the 2-D log spiral. For this surface, Fs can
be determined without specifying <r(x, y) since the moment-equilibrium
equation written about (xc, zc) is independent of the normal stress (Lesh-
chinsky and Baker 1986). Leshchinsky et al. (1985) showed this surface to
yield a solution equivalent to an upper bound in the context of limit analysis
of plasticity. However, this mathematically derived surface is extremely
restrictive, i.e., it cannot be extended to realistic situations where layered
soil is likely to dictate its trace. Subsequently, similar to the 2-D generali-
zation suggested by Leshchinsky (1990b), z(x, y) in this paper is left to be
specified by the user. This approach is common to all generalized methods
of 2-D slope-stability analysis where one of several available optimization
techniques is used to locate the critical surface. Note that by specifying z(x,
y), 0 in (12) contains only two unknown parameters: xc and zc. Back sub-
stitution of 6(x, y) and z(x, y) into the equilibrium equations [(l)-(3)] gives
three equations with three unknown parameters (xc, zc, and Fs) and one
1752
the critical a(x, y) containing over 100 terms. Although this expression was
developed with the aid of an advanced computer code capable of dealing
with symbolic algebraic derivations, the writers felt uncomfortable in uti-
lizing it. Instead, the terms dg/dz, dg/dzx and dg/dzy were derived manually
in a straightforward manner and integrated into a numerical procedure that
satisfies (11) and thus producing discrete values of cr(x, y), as shown in the
next section. For completeness, these terms are given next:
dg
— = [c, + (CT - u)tyt]B(xc - x) + Fscr(z - zc) + [c, + (CT - u)\\>,]
de
-£• = [c, + (CT - u)ty,]B(xc - x)Q + [c, + (CT - i*H,][(z - zc)
where:
35 = zx{\ + ft2) + (zx + £lzy)[zy(zxn - zv) - 1]
dzx {z2x + z2y + 1)1/2[1 + il2 + (zx + ilzy)2]3/2 {
'
and
dB _ zy{\ + a2) + (zx + Q,zy)[zx(zy - zJX) - O,]
(17)
dzy (z2 + z 2 + i) 1/2 [i + a 2 + (zx + cizy)2Y
Substitution of (13)-(17) into (11) and carrying out the total derivatives
with respect to x and y will result in the differential equation for u(x, y).
One can visualize the complexity of this equation. However, differentiation
with respect to x and y in the numerical scheme shown in the next section
is rather simple.
It should be emphasized thatCT(X,y) is basically a term of the integrands
in the equilibrium equations [(l)-(3)] and only its integrated value must
satisfy these global equations. That is, in the framework of limit-equilibrium
formulation, <j(x, y) does not have to satisfy local stress-equilibrium re-
quirements (e.g., dajdx + dixy/dy + dx^/dz + X = 0, etc., are not satisfied
along the surface). Hence, although the resultant vector associated with the
variationally obtainedCT(X,y) should satisfy global equilibrium and produce
a unique Fs, there are many possible distributions of cr(x, y) capable of
generating the same resultant (Leschinsky and Huang 1992). Consequently,
CT(X, y) enables one to introduce Coulomb's failure criterion in the for-
mulation and thus to attain the fundamental objective: m i n ^ ) while sat-
isfying global limiting equilibrium. As a result, judgment about internal
forces (i.e., location, inclination, and magnitude) that correspond to the
1753
T(JC, y) must be positive, implying that (cr - u) > -(c/tan 4>), everywhere
over the slip surface. Alternatively, if a truncated failure envelope is used,
one can impose (CT - u) > k where k = the tensile strength cutoff discussed
in the next section.
NUMERICAL PROCEDURE
+ {al ~a\
d__
a\
dzx
-w
">i£ -i./
(18)
dx dzx axa2{ax + a2)
Projection of discretized
/ slip surface on x-y plane.
I.I 1,2
2,1 2,2 1,3
Vy
*
-**H SPECIFY a0
1
| ASSUME INITIAL VALUES: F„ xc,~z~^m
1 j
t. , f
j USE EQS. 12 & 6 TO CALCULATE fi(i, j ) & B{i,j)\ |
"IT
COMBINE EQS. 13-23 WITH EQ. 11. SOLVE THE
RESULTED n SIMULTANEOUS LINEAR EQS. FOR a(i,j)
"YES-
-YES
where
1755
bi = y, v - y,.i-i (22)
b2 = y , v + 1 - y.-.j (23)
Combining (13)—(15) and (18)-(19) with (11) gives n linear equations with
n unknown u(i, /'). Solving these simultaneous linear equations produces
cr(/,;). Now, the integration of the three equilibrium equations [(l)-(3)]
can be carried out numerically in a straightforward manner [similar to Lesh-
chinsky and Huang (1992)]. If these equations are satisfied then the initially
assumed values xc, zc, and Fs are correct; if not, however, new values are
assumed (using a routine that solves three simultaneous nonlinear equations)
and the process is repeated until equilibrium is satisfied. Subsequent to the
aforementioned description, the problem is reduced to (3 + n) unknowns
[i.e., xc, zc, Fs, and n times a(i, /)] that are obtained through solving n
simultaneous linear equations and three simultaneous nonlinear equations.
Note in Fig. 3 the term cr0; the location of this normal stress is at the
crest and it is denoted by point 0 on Fig. 2. This stress is introduced into
the numerical scheme through (11) when calculating the terms at point (1,
1). Introduction of rr0 enables the user to somewhat control a(r, j) distri-
bution and thus to produce compliance with Coulomb's failure criterion, if
necessary (i.e., see last section: [CT - u\uj a k). Since tensile stress tends
to develop at the crest, the authors have found that specifying the normal
stress at point 0 is effective in producing this compliance. The user should
specify various values of cr0 until minimum Fs and admissible stresses over
the entire slip surface are obtained [e.g., Leshchinsky and Huang (1992)].
If for all specified <r0 the resulted stress at some point is (<r - u)uj < k, a
tension crack should then be introduced. This situation implies that the
selected slip surface is statically unfeasible.
The computation process should be repeated for all feasible slip surfaces
until the minimal factor of safety is obtained. The results then (including
the slip surface) are considered critical.
RESULTS
To demonstrate the nature of the 3-D solution, a few results are presented.
Additional results are presented in Chowdhury (1992). Rather than intui-
tively specifying an arbitrary 3-D slip surface for this demonstration, it seems
useful to select one that is also a result of the nongeneralized variational
analysis. That is, to choose an extended log-spiral mechanism for which Fs
can be determined through a different procedure as well [see paragraph
following (12)]. Selection of such a surface should provide the reader with
a visual sense of critical variational results as related to homogeneous prob-
lems.
Figs. 4(a)-4(c) show half of the extended log-spiral surfaces obtained for
conical heaps inclined at 30° (Baker and Leshchinsky 1987). Fig. 4(a) illus-
trates the critical slip surface for a c-4> soil while Figs. 4(b)-4(c) represent
the surfaces for cohesive soils. Fig. 4(b) indicates a deep-seated failure for
homogeneous heap and foundation soils. Fig. 4(c) demonstrates the effects
of impenetrable foundation on the critical slip surface and the required
stability number N. Note that 3-D surfaces tend to deepen as the ratio (Nl
tan 4>) increases, similar to the trend resulting from 2-D analysis. Fig. 5(a)
1756
FIG. 4. Extended Log Spiral Critical Slip Surface for 30° Conical Heaps: (a) c-<(>
Soil; (b) Cohesive Soil and Deep-Seated Failure; (c) Cohesive Soil over Impene-
trable Foundation
shows the critical surface predicted for a vertical corner cut comprised of
soil possessing high-friction angle (Leshchinsky and Baker 1986). Fig. 5(b)
shows a plan view of the surface in Fig. 5(a), as well as the direction of the
shear resistance force over the sliding surface. This direction results from
the solution of (12) and, physically, it is in the opposite direction to the
slide movement. Only when t)> = 0° all elemental shear forces over the
extended log spiral surface are parallel (i.e., 0 = 180°).
1757
FIG. 5. Vertical Corner Cut: (a) Critical Slip Surface; (b) Plan View of Direction of
Shear Resistance over Slip Surface
appears "reasonable" along cross sections of the x-z plane [Fig. 8(a)], it is
perhaps unreasonable (i.e., wavey) along sections on the y-z plane [e.g.,
Fig. 8(£>)]. However, the writers obtained several solutions yielding different
combinations of xc, zc, and cr(7,/) with essentially the same Fs. Subsequently,
it is possible that one of these solutions will improve the appearance of cr {x,
v), though without affecting the analysis objective; i.e., m i n ^ ) . Referring
to the discussion in the second paragraph following (17), one realizes that
the actual value of <T(X, y) does not warrant examination except if it violates
the failure criterion.
The factor of safety obtained for the mechanism and problem shown in
Figs. 6 and 7 is 1.31 and it corresponds to v0/yH = -0.13. Since $u = 0,
most 2-D slope-stability-analysis methods will yield identical predictions
when circular slip surfaces are considered. Hence, one can verify that Fs =
1.00 for the equivalent 2-D problem. The traces of the critical 2-D and 3-
D (on the symmetry plane) slip surfaces are shown in Fig. 9. Conducting
2-D stability analysis [e.g., Leshchinsky and Huang (1992)] on the critical
trace of the 3-D problem shown in Fig. 9 yielded a safety factor of 1.01,
nearly the critical 2-D value of 1.00. Consequently, the writers feel this
problem to be instructive, demonstrating the importance of 3-D back-anal-
ysis in post failure investigation. For example, consider a hypothetical case
in which the exact failure pattern shown in Fig. 6 has happened. Selecting
the deepest traced surface, which is also cylindrical to an extent of 0.4H,
for 2-D analysis will yield a factor of safety of about one (i.e., 1.01), thus
explaining the failure and justifying the selection of c„. However, conducting
a 3-D analysis over the entire traced surface will yield a factor of safety of
1.31, implying that failure occurred because N = 0.261/1.31 = 0.2; i.e.,
the actual cu is about 25% lower than initially estimated. Since back-analyses
of failure are often used to assess the adequacy of laboratory and field
1759
*
J. Geotech. Engrg. 1992.118:1748-1764.
J^z/H-
Downloaded from ascelibrary.org by UNIVERSIDADE FEDERAL DO RIO on 04/22/14. Copyright ASCE. For personal use only; all rights reserved.
13
-0.2-•
Lf-20
1 I SSW-WBW
#/ / /
«<&/
/ // // ////
1
0.4'
7 foV
v #/
V/
02-
;;
<—4 j i — —^
x/H -0.2 -0.4 -0.6
C6;
FIG. 7. Discretized Slip Surface for Example Problem: (a) Plan View of Soil "Col-
umns"; (h) Cross Sections
FIG. 8. Resulted Normal Stress Distribution: (a) Along x-z Cross Sections; (b)
Along One y-z Section
CONCLUSIONS
F,(2- D) = 1.00
4> = o "
N = — = 0.2610
•ytl
kz/H
1. ru = 0 1-00 " ^ ^ 5 ? 5 5 5 ^ ^ ^ ^ ^
2. r u = 0.1
iV = — = 0.1 4> = 30°
7±/
0.49
APPENDIX. REFERENCES
Anagnosti,P. (1969). "Three-dimensional stability of fill dams." Proc. 7th Intl. Conf.
on Soil Mech. and Foundation Engrg., International Society of Soil Mechanics and
Foundation Engineering, 2, 275-280.
Azzouz, A. S., and Baligh, M. M. (1976). "Design charts for three-dimensional
stability of cohesive slopes subjected to surcharge loads." Publication No. PB-
275779, National Technical Information Service, U.S. Dept. of Commerce, Spring-
field, Va.
Azzouz, A. S., and Baligh, M. M. (1978). "Three-dimensional stability of slopes."
Publication No. PB-285740, National Technical Information Service, U.S. Dept.
of Commerce.
Azzouz, A. S., and Baligh, M. M. (1983). "Loaded areas on cohesive slopes." /.
Geotech. Engrg. Div., ASCE, 109(5), 726-729.
Azzouz, A. S., Baligh, M. M., and Ladd, C. C. (1981). "Three-dimensional stability
analysis of four embankment failures." Proc. 10th Intl. Conf. on Soil Mech. and
Foundation Engrg., 3, 343-346.
Baker, R., and Garber, M. (1978). "Theoretical analysis of the stability of slopes."
Geotechnique, 28(4), 295-411.
Baker, R., and Leshchinsky, D. (1987). "Stability analysis of conical heaps." Soils
Found., 27(4), 99-110.
Baligh, M. M., and Azzouz, A. S. (1975). "End effects on stability of cohesive
slopes."/. Geotechnical Engrg. Div., ASCE, 101(11), 1105-1117.
Chen, R. H., and Chameau, J. L. (1983). "Three-dimensional limit equilibrium
analysis of slopes." Geotechnique, 32(1), 31-40.
Chowdhurry, S. (1992). "Generalized slope stability analysis," MCE thesis, Uni-
versity of Delaware, Newark, Del.
Elsgolc, L. E. (1962). Calculus of variations. Pergamon Press, Ltd., London, Eng-
land.
Gens, A., Hutchinson, J. N., and Cavounidis, S. (1988). "Three-dimensional analysis
of slides in cohesive soils." Geotechnique, 38(1), 1-23.
Giger, N. W., andKrizek, R. J. (1975). "Stability analysis of vertical cut with variable
corner angle." Soils Found., 15(2), 63-71.
Giger, M. W., and Krizek, R. J. (1976). "Stability of vertical corner cut with con-
centrated surcharge load." J. Geotech. Engrg. Div., ASCE, 102(1), 32-40.
Hovland, H. J.(1977). "Three-dimensional slope stability analysis method." /. Geo-
technical Engrg. Div., ASCE, 103(9), 971-986.
Hungr, O., Salgado, F. M., and Byrne, P. M. (1989). "Evaluation of three-dimen-
sional method of slope stability analysis." Can. Geotech. J., 26(4), 679-686.
Kopacsy, J. (1957). "Three-dimensional stress distribution and slip surface in earth
work at rupture." Proc. 4th Intl. Conf. on Soil. Mech. and Foundation Engrg.,
International Society of Soil Mechanics and Foundation Engineering, 1, 339-342.
Lambe, T. W., and Whitman, R. V. (1969). Soil mechanics. John Wiley Book Co.,
New York, N.Y., 370.
Leshchinsky, D. (1990a). Discussion of: "Three-dimensional stability analysis of
concave slopes in plan view," by Zhang Xing, /. Geotech. Engrg., ASCE, 116(2),
342-345.
Leshchinsky, D. (1990b). "Slope stability analysis: Generalized approach," /. Geo-
tech. Engrg., ASCE, 116(5), 851-867.
Leshchinsky, D., and Baker, R. (1986). "Three-dimensional slope stability: End
effects," Soils Found., 26(4), 98-110.
Leshchinsky, D., Baker, R., and Silver, M. L. (1985). "Three-dimensional analysis
of slope stability," Intl. I. Nwner. Anal. Methods Geomech., 9(2), 199-223.
Leshchinsky, D., and Huang, C. C. (1992). "Generalized slope stability analysis:
interpretation, modification and comparison."/. Geotech. Engrg., ASCE, 118(10).
1763
1764