Вы находитесь на странице: 1из 20

Computers

and Geotechnics

8 (1989)

329-348

A SIMPLEX

ANALYSIS

OF SLOPE STABILITY

J. P. Bardet and M. M. Kapuskar Civil Engineering Department University of Southern California Los Angeles CA 90089-0242

ABSTRACT The downhill simplex algorithm is applied to systematically locate the critical failure mechanism in slopes and to compute the minimum factor of safety. The proposed method is illustrated with three examples of circular and noncircular slope stability and its results are compared to previously published solutions. Although it is not the most efficient optimization procedure, the simplex method is versatile, robust, and simple. It is recommended to enhance slope stability programs by providing for an automatic search of the critical failure mechanisms.

INTRODUCTION Slope stability analyses are routine calculations in geotechnical engineering. They are surveyed and described in several textbooks of soil mechanics

[ 1,2]. Nowadays, slope

stability analyses are performed by computers and no longer by hand calculation. Several slope stability computer programs are available from public or commercial domains. The main problem in slope analysis is to find the critical failure surface which gives the lowest factor of safety. The search for the critical mechanism may be done either by the user, through a tedious trial-and-error procedure, or by the computer. 329
Computersand Geotechnics0266-352X/90/$03.50 @ 1990

Elsevier Science

Publishers

Ltd, England.

Printed

in Great Britain

330 Minimization techniques using dynamic programming and variational methods [3,4]

have been proposed to search for the critical slip surfaces in slope stability computations. This paper presents a simple but general method of optimization which can be applied to most slope stability methods. It is based on the downhill simplex algorithm [5] and on generalized piecewise linear failure mechanisms. The proposed approach may be adapted to simpler types of slope stability analyses, e.g. circular failure. The downhi! the critical failure mechanism is also not novel in computerized simplex method is not claimed to be the most efficient optimization algorithm 161.The automatic search for slope stability analysis. However, the proposed simplex method is simple, robust, versatile and can be quickly adapted to most existing slope stability programs. The first section defines the selected method of slope stability analysis - its generalized piecewise linear failure mechanisms and its factor of safety - and compares it to other slope stability analyses. Then it transforms the factor of safety into a function of several independent variables which can be mathematically optimized. The second section briefly reviews the principle of the downhill simplex method as well as its specific initialization and convergence criterion. The third section applies the present method to three different types of slope analyses and compares the numerical results to previously published solutions.

REVIEW

OF GENERALIZED

SLOPE STABILITY

Definition

of the generalized

failure

mechanism

In general, any failure surface may be approximated by n piecewise linear segments which are defined by n+l points (xi, yi) hereafter referred to as slide points (see Fig. 1). In order to calculate the frictional forces on the failure surface, the failing mass is sectioned into n slices with vertical interfaces. The width of each slice is arbitrarily specified by fixing the horizontal locations x;of the slide points i=2, .... n. However, the coordinates x1 and x,,+, of the starting and ending points of the failure mechanism are left variable within the n+l]. segments [xi. x2] and [x n, xmax The horizontal coordinates xi are assumed to bc ordered by increasing value. The elevation yi of the i* slide point varies in an interval [ymi, ~1. The lower bound y, is defined by selecting an arbitrary bedrock. The upper bound ymaXISspecified by the slope surface geometry.

331

+
x1
In,

Xl

x2

Xi

%+I

X.+1

x::;

FIGURE 1. Generalized piecewise linear failure mechanism for simplex analysis.

The series of coordinates x1. yyi, .... yzi, xrJ:,rdefines the lower boundary of the failure mechanism, while the series xya, yyax, .... yz, xr:t defines its upper boundary.

The i* slide point may be fixed by imposing that y, and yy coincide. The generalized piecewise linear failure mechanisms may be constrained to describe specific types of surfaces such as circles or logarithmic spirals. For instance, in the case of circular failure, the n+l points (xi, yi) may be forced to be on a circle defined by 3 unknowns: the position of its center (xc, y,) and its radius R,. The variation of circular failure may be defined by specifying upper and lower bounds for xc, y, and R, [7]. Specific surfaces introduce obvious advantages by reducing the number of unknowns, but will not be considered hereafter. Definition of factor of safety As commonly assumed in slope stability analysis, the soil is assumed to fail simultaneously, and not progressively, along the failure surface. Its shear strength obeys the Mohr-Coulomb criterion, which is defined by the effective friction angle $I and cohesion

c. The factor of safety F is defined by introducing the friction $* and cohesion cm mobilized on the failure surface:

332 tan tj tan Q = F cm =; The generalized failure mechanism of Fig. 1 is made of n slices of arbitrary width. The forces applied on the i* slice are defined in Fig.2. 0 /I

(1)

* , > .

i+l

2,
t

(Ni -ui Ii ) tan+:+

FIGURE 2.

Forces acting on an individual slice.

ASj = Si+, - Si represents the resultant interface force applied on the i slice and is inclined at an angle Ei with respect to the horizontal. No side forces exist at the left and right extremities of the failing mass. Ri is the resultant of external forces making an angle oi with respect to the vertical. The external forces are the tota loads resulting from the weights of soil and water, concentrated and distributed surface loads, seepage forces, horizontal forces from earthquake effects and other forces such as those applied by ground anchors. N, and T, arc the resultant of the normal and tangential forces acting on the slice base of length 1; and inclination C+with respect to the horizontal. ai and li are computed from the coordinates (xi, yi) of the slide points. Ti is related to the mobilized shear strength as follows:

333
Ti = (Ni - ui li j tan ($7 + cm li

(2)

ui is the average pore water pressure acting on the base of the I siice. In most instances, the pore pressure is calculated by assuming that it varies linearly with the depth tnemmxl from a specified water table. The equilibrium of forces in the direction parallel and perpendicular to AS, &ivcs: 1 N, sin(a,-6 1 N, cos(ni-ii) ) + Ti CO~(CX-6;) + Ri sin(oi-hi) +- T, sin(ai-s,) = -ASi

.di

(:ij

- K, CO:~(W~--~~) 0 =

The equilibrium of moments about the center of the base of the lth slice gives: R, ri = - AS s

Li
liqs. I, 2

(4,

where ri and s, are the arms of RI and AS, about the base center. By combining and 3. AS, becomes:

where (Ri cos(ai-ai) Ai=c0s(a,-6~) + h sin(ai-hi) tan $i - ui li) tan $i + ci !;

Ri sin(aj-wi) Bi=~p~ cos(ai-bi) + k sin(ai-si) tan $

The resultant force and moment about an arbitrary reference point 0 of the internal forces must equal zero:

0 =

2
i=l i=l

ASi cos $

0 = 2

ASi sin hi

(7)

0 = c
61

(ASipi c0s(B,-8~)

+ Riri 1

334 where pi is the distance between the base center of the i* slice and the reference point 0 (see Fig.2). From Eq.7, three factors of safety can be defined for the horizontal and vertical forces and the moment: ( c A; cos 6.

F, =

;I c
i=l n c

Bi cos 6.

Ai sin 6i i;l

<F,=

(8)
Bi sin 6.
I

c
i=l

1
F, =
c

c
i=l

A,p, cos(9,-$)

(B,p, COS(~~-~~) + Ri ri)

i=l

Comparison of slope stability analysis As summarized in Table 1, Eq.8 encompasses,

among others, the methods of

Spencer [8], Janbu [9], Neuber [lo], Gussmann [11], Krey [12], Bishop [13], Terzaghi

[ 141 and Fellenius [15].


Spencer assumes that Si is identical for all slice interfaces (6,= 6, i=I,...,n) and determines the factor of safety F graphically by searching for the angle 6 so that F = F,(6) = Fv(6) = F,(6). Janbu, Gussmann and Neubcr assume 6, = (pmand vertical external loads Krey and Bishop use a circular

(w,=O) and define the factor of safety as F,(6,,6,,...,6,).

failure mechanism with very thin slices, and assume Fi = 0 and oi = 0 for i=l,...,n. They neglect force equilibrium and choose F = Fh?. Terzaghi and Fellenius consider circular failure and select &,=a, and w,=O for i=l,...,n. In cantrast to other methods, their factor of safety F = F&it,62,...,6n) is calculated without an iteration, making their method one of the most commonly used in engineering practice. Additional slope stability methods may be devised from the previous methods by adopting different assumptions for the inclination of the interface forces hi. Seed and Sultan

[ 161 distribute si according to the mobilized shear stresses along the slice base.

335 Alternatively, the hi may be distributed in a sinusoidal manner or calculated by using Spencers thrust line criterion [8]. Selected method of slope stability analysis

The method which is selected in the present analysis assumes (a) that all the interface forces have the same inclination 6, and (b) that the factor of safety F, is larger than F, and Fv. The first hypothesis implies that:

F(6) =F

(6) = F H

(9)

The second assumption is justified when the height-to-width ratio of the slices is not too small. In our analysis, the slice aspect ratio may be controlled by limiting the number of slices. In other words, n cos(a,-wi) -ui li) tan Qi + ci Ii

>

F=

i=l c

(10)
R. & sin(ai-wi) i=l

where I$ = cos(ai-6J + b sin(a&) tan Qi (11)

Eq.10 has to be solved by iteration to yield the factor of safety F. F denotes the factor of safety after k iterations. The initial value flis equal to 1.0, assuming that the initial mechanism is in equilibrium. The following convergence criterion is adopted: FL< (l+rl) F- (12)

where q is a parameter which controls the accuracy. In the numerical applications, q was set to 1~10~. F usually converged after two or three iterations.

336
r,,;+h rl;rec

cos ai (cos ai + ksin

ai tan 4:)

(R.cos

(a.-wi)

- ui 1,) tan o. + c. I

>

1 I

cos (ai-hi)

t k sin (a,-?ii) tan o 1

i=l

Definition of safety function To summarize, the factor of safety F of Eq. 10 is a function of n+l independent variables x,, yZ, ys, . ... y,, and x+~ and is referred to as the safety function hereafter. Finding the minimum factor of safety consists of minimizing the safety function:

337

F(xl, Y,, Y,, .... Y,. xn+]) where

.. .
Y, E [Y;?
X n+le IX?

(13) Y ,I

q;1
x1, y2. y3. . .. . yn, and

For a reason which will be justified later, we scale the coordinates ymin, and i is between 2 and n, zi is:

x,,+l that have an interval of variation different from zero. N variables zi are therefore introduced. When yy#

zi =

2yi - y; - y rn.X - y; yi

i=l, . . .. N

(14)

Zl and %+l are defined similarly from x1 and xn+*. N is the actual degree of freedom of the

mechanism and is less than or equal to n+l. N is equal to n+l when no slide point is fixed. The safety function then becomes: F(z) = F(zl, z2, . .. . zN) where zi E [-l,l] i=l , . . .. N (15)

MULTIDIMENSIONAL Principle of function optimization

SIMPLEX with simplex

MINIMIZATION

The scheme used to minimize F(z) is based on the downhill simplex method proposed by Nedler and Mead [5] and reported by Press et al. [6]. A simplex is the geometrical figure in a N-dimensional space consisting of N+l vertices and all their interconnecting line segments. In two dimensions, a simplex is a triangle, while in three dimensions, it is a tetrahedron. For the N-dimensional problem of Eq. 15, the simplex is defined by N+l vertices 5 in a N-dimensional space (see Fig.3). Each vertex specifies a failure mechanism which has a factor of safety F(z,).

338

n=N=3

FIGURE 3.

Domain of variation of the slide points in the physical space and in the variable coordinate space for a three plane mechanism.

The downhill simplex method takes a series of steps to move the simplex toward a function minimum. The four possible transformations reflection of a vertex zr are:
(16)

zR = (1 + a,) Z - aR zr

reflection plus expansion

z RE = (1 + oRa,,)

Z - oREoR z,

(17)

local contraction

zc =(l+a,)Z-a,zr
N+l

(18)

global contraction

zo = a,

C(zj j&%x

+ zmax)

(19)

where z = bNilZj
i=l J*

(20)

The constants arc aRE and q

control the extent of the simplex modifications

and are

positive. In the slope models tested, a good convergence a RE=0.5 and a,=O.Ol.

was achieved by using a,=0.3,

339 The simplex algorithm proceeds as shown in the flow chart of Fig.4. After initialization of the simplex, the Fist operation is a reflection (Eq. 16) of the vertex zmti which has the lowest factor of safety F(z,~). The second step is either an additional expansion (Eq. 17) or a local contraction (Eq. 18). The third step, global contraction (Eq. 19), is only activated when the local contraction is unsuccessful. This operation moves all vertices towards the vertex zmax which has the highest factor of safety F(zmax). It permits the movement of the simplex in a transverse direction towards the minimum

FIGURE 4.

Flow chart of the multidimensional

downhill simplex algorithm.

340 Initialization of the simplex

The initial simplex is automatically generated within the boundaries -1 and +1 of the N-dimensional space. Its N+l vertices zr are chosen to be on a hypersphere, which has the equation: N r2 = c j=l zb = z z r I (21)

The hypersphere is centered at the origin. Its radius r is chosen conveniently 1. The first N vertices z, to zN are selected to be the intersections the N positive coordinate axes (see Fig.5). =3

between 0 and with

of the hypersphere

FIGURE 5.

Initial scaled simplex in the three dimensional mechanism.

space for a three plane failure

The last vertex z~+~ is the negative intersection of the hyperaxis z1=z2=...=zN with the hypersphere. Therefore, the initial simplex is represented by the following N+l vertices:

(22)

If two or more vertices have identical factors of safety, the initialization is repeated by using a different radius r until the N+l values of F(q) are distinct. The scaling of the original coordinates therefore centers the boundaries about the origin and generates an initial

341

simplex that fills the search domain and can move in all directions. This particular initialization of the simplex was found to accelerate the convergence of the downhill simplex method. Convergence criterion The ultimate goal of the downhill simplex method is to move the simplex to a location where the vertices are centered near or around a minimum of F(z). Unfortunately, the method may converge to a local minimum instead of a global minimum. In contrast to the one-dimensional case, it is not possible to guarantee that the calculated minimum is case. The optimization process is terminated when the global in the multidimensional following criterion is met: F F

max mn 5 a
max+ Fmin

-F

(23)

where E is the convergence control parameter, and

(24)

In the numerical applications, E was set equal to 1~10.~. The iteration process is aborted when the simplex volume becomes very small or when the number of iterations exceeds a given threshold. After successful convergence, vertices:
N+l min = T&i C I r-1

the optimum vertex is computed by averaging all the

(25)

The vertex zti safety F(z,,).

specifies the critical failure mechanism which has the minimum factor of As suggested by Nedler and Mead [?I], the solution z,;, was checked by was observed in the examples which we tested. Failure to

restarting the simplex algorithm from the last simplex produced instead of the one given in Eq.22. Immediate convergence converge after several steps indicates ill behavior of the simplex method or an incorrect definition of the lower and upper bounds for the failure mechanisms.

NUMERICAL

APPLICATIONS

AND

DISCUSSION

The present simplex method is tested by applying it to three different types of failure mechanisms: plane, circular and wedge failures, and by comparing its solutions with

342

analytical and published results. It was implemented in the computer program MULWED ]171. Plane failure in an infinite slope Fig.6 shows a natural slope which is analyzed in [ 161. It is composed of a silt underlain by a fractured limestone.

silt .

d limestons

20 m

FIGURE 6.

Example of plane failure in slopes [18].

The thickness H of the silt layer is smaI1 compared to the sIope Iength so that no deep slip circles develop. For a water table parallel to the slope surface and located at a depth H,, Sanglerat et al derives the factor of safety for a plane failure located at depth H: c + [yl,Hw + y(H-HW)] cos2p tan Q F=2 ]Y,.,H,,,,+

Y~,,W-H,,,)I 33 sin

(26)

where c and Q are the drained shear strength of the silt, p is the slope inclination, rh and r,,i are the unit weights of the silt above and beneath the water table. The following values were selected in the numerical examples: H=lO m, tan p=1/2, yh=18 kN/m3, y,,=20 kN/m3, and 6 = 30. All failure mechanisms are made of n=12 slices with N=l3 degrees of freedom. The upper and lower boundaries of the failure mechanisms are located on the free surface and on the interface between limestone and silt. The basic slice width is 10 m. As summarized in Table 2, five different analyses were performed. For the two first analyses, the soil is cohesionless = 1.155, independent and dry (H,=O, Q=30, c=O). IQ.26 gives F = tan$//tanp of H. In Fig.6, the failure mechanism labelled 1 is not optimized.

343

The corresponding

factor of safety is 1.188, which is slightly larger than the exact value.

The mechanism labelled 2 is optimized after 210 iterations. Compared to the first mechanism, it is almost plane and parallel to the surface. It does not exactly coincide with the free surface due to programming constraints that prevent slices from vanishing completely. The corresponding exact value. Table 2. Comparison of factors of safety for the plane failure Sanglerat et al. [ 181 F exact 1.155 1.571 1.412 1.271 Present method F mm 1.188 1.156 1.579 1.358 1.228 number of Iterations factor of safety is 1.156, which is almost identical to the

CASE 1) c=o 2) c=o 3) HW=Om 4) HW=5m 5) HW=lOm

210 383 255 324

In the case of a partially submerged cohesive silt (c=30 kN/m2 and @=18), three different water depths were considered: H,=O m, 5 m and 10 m. In all cases, the optimized failure mechanisms stretch along the silt- limestone interface. The corresponding factors of safety reported in Table 2 are close to the exact values, which decrease with H, according to Eq.26. Due to the length of the failure mechanism, the end effects are negligible. Circular failure in a bench fill

Fig.7 shows an example of a drained bench which is solved in [l]. The bench of height H=20.4 m is filled with a dry soil of unit weight y,=19.6 kN/m3 in order to restore the original slope with an inclination of p=39.25. The drained shear strength of the fill is Q=30 and c=9.6 kN/m2. The failure mechanisms are made of ten slices, which are spaced as shown in Fig.7. The inclination 6 of interface forces is set equal to 30. As summarized in Table 3, three different analyses are performed. The first one attempts to reproduce the results that Huang obtained by using the Krey-Bishop and Fellenius circular methods. Huang found that the minimum factors of safety were 1.072 and 1.069 for a toe failure, respectively. All slide points are fixed and located along the critical failure surface calculated by Huang. By using the present method without optimization, the factor of safety is found to be 1.108. In the second analysis, the first slide point is fixed at the toe while the others are variable. The optimized failure is slightly different from a circle but has the same factor of safety as Huang. In the third analysis, all slide points are variable. The critical mechanism is found to be circular, but does not pass through the toe. The corresponding factor of safety is 1.082, which indicates that the

344 simplex has converged toward a local rather than a global minimum. Additional analyses show that the inclination 6 of interface forces has little influence on the critical failure mechanism and the minimum factor of safety Fmh; selecting 6 =O, IO, 20 and 40 changes only the third decimal figure of F
mm

(F-1.082)

6.h I I

(F-1.108)

FIGURE 7.

Example of circular failure in slopes [ 11.

Table 3. Comparison of factors of safety for the bench till problem. CRITICAL MECHANISM SURFACE TYPE 10 planes 10 planes planar circle unknown me failure me failure toe failure number of Iterations 191 264

METHOD Present method Present method Krey/Bishop Fellenius [l] [l]

mm

1.082 1.072 1.072 1.069

Wedge failure

in a sloping

core dam

Fig.8 shows the cross section of a dam which was investigated by Seed and Sultan [16] and is reported in [2]. The sand shell has a uniform dry unit weight y,=l7.3 kN/m3 and an average friction angle $=40. The core is inclined at 31.5 and is made of clay with a cohesion c=96 kN/m2. The inclination 6 of interface forces is set equal to 40.

two-wadqe

mechanism

(F- 1.630)

1
FIGURE 8.

100 m

Example of wedge failure in slopes [16].

Based on the observed failure, Seed and Sultan considered the two-wedge mechanism shown in Fig.8. The wedges slide on a plane inclined at 10 and on the core surface. By using the present method without optimization, the Seed and Sultan mechanism gives a factor of safety of 1.630, which is slightly lower than the factor of safety obtained by hand calculation in Lambe and Whitman (F=l.65). A second analysis was performed. As shown in Fig.8, seven instead of two slices were selected and the slide points were variable except at the ends. Their lower bounds were set at the sand-core interface and on the horizontal base line passing through the toe. After 35 iterations, the simplex algorithm produced a critical mechanism which is virtually plane and inclined at 13.2, instead of 10 as assumed by Seed and Sultan. The corresponding factor of safety was 1.645, which nearly duplicates the result that Lambe & Whitman found by hand calculation. Discussion The simplex solutions of generalized plane failures are in good agreement with the results of circular and non-circular methods. The simplex optimization performed well in all the applications tested and never failed to determine a minimum factor of safety. The present method is also versatile for calculating critical mechanisms that are forced to go through a specific location, as in the case of me failure. Given our initialization procedure, it was observed that the number of simplex iterations increases as a power of about 3/2 with the number of slices. For a given tolerance E equal to 1x10, 30 to 45 iterations for a 6-slice mechanism were necessary, compared to 190 to 380 iterations for 12 slices. In the previous examples, the number of slices was selected small to achieve fast calculation time, but large enough to describe potential failure mechanisms realistically. Increasing the number of slices is not expected to influence the results significantly. The

346 terms A,, B,, R, and r1 used in Eq.8 are proportional to the slice width, within a first order approximation. In contrast to the factor of safety F,, F, and Fv are rather insensitive to the provided that the number of slices is appropriate to describe the present method based only on Fn and Fv is expected to number of slices. Therefore, realistic failure mechanisms,

yield results independent of the number of slices. Occasional problems were experienced in generating a vertex for the initial simplex. They were caused by the lack of convergence during the iterative calculation of the safety function F. These problems were remedied by reducing the interface force angle 6, which was arbitrarily set equal to an average friction angle @.

CONCLUSION The present paper unifies various methods of slope stability based on generalized piecewise linear failure mechanisms. It formulates the slope stability analysis into a mathematical problem of optimization that is solved by the simplex algorithm. The simplex method automatically searches for the critical failure mechanism and the minimum factor of safety. The present method may be easily implemented into existing slope stability programs. It was found capable of describing circular, plane and wedge failures. The simplex method may not be the fastest and most efficient technique of optimization, but it is a reliable, robust, versatile and simple tool for searching for the minimum factor of safety of general failure mechanisms in slopes.

REFERENCES Huang, Y.H., Stability of Earth Slopes, Van Nostrand Reinhold Co., (1983) 200, 215-218, 249-254. Lambe, T.W., and R.V. Whitman, Soil Mechanics. %-Version, Series in Soil Engineering, John Wiley & Sons, N. Y., (1979) 333-340, 366-369. Baker, R., Determination of the critical slip surface in slope stability computations, Int. J. for numerical and analvtical methods in geomechanics 4 (1980) 333-359. Martins, J.B., Embankments and slopes by mathematical programing, Numerical methods in geomechanics. Ed. by J.B. Martins, Reidel Co., Holland (1981) 305334. Nedler, J.A., and R. Mead, Simplex Method for Function Minimizations, Journal, 7 (1965) 308. Press, W. H., B.P. Flannery, S.A. Teukolsky, and W.T. Vetterling, Recines. Cambridge University Press, Cambridge, (1986) 289-294. Compw

1. 2. 3. 4.

5. 6.

Numerical

347 7. 8. 9. Bardet, J.P., SLOPE: A Computer Program for Slope Stability Analysis, Denartment of Civil Enrrineerina. Universitv of Southern California. Los Angeles (1989). Spencer, E., 1973, Thrust Line Criterion in Embankment Geotechnique, 23 (1973) 85-100. Stability Analysis,

Janbu, N., Application of Composite Slip Surfaces for Stability Analysis, Proceed es of the Conference on Slope Stability of Earth Slopes. Stockholm, 3 (1954) :3. Neuber, H., Untersuchung der Standsicherheit hoher Boschungen nach der sogenannten Streifenmethode, Fortschritte Geoloeie Rheinland-Westfalen, 15 (1968) 245-262. Gussmann, P., Das allgemeine Lamellenverfahren unter besonderer Beriicksichtigung von Lusseren Kx%ften,Geotechnik, 1 (1978) 68-74. Krey, H.D., Erddruck. Erdwiderstand und Sohn Verlag, Berlin (1926). und Trazfahiekeit des Bauzrundes, W.Ernst

10.

11. 12. 13.

Bishop, A.W., The Use of the Slip Circle in the Stability Analysis of Slopes, Proceedings of the Euronean Conference on Slone Stabilitv of Earth Slones, Stockholm 1(1954) 1-13. Terzaghi, K., Mechanism of Land Slides, Application of Geology to Engineering Practice, Geological Sccietv of America. Berkelev. Harvard Soil Mechanics Series, (1950) 83-123. Fellenius, W., Erdstatische Berechnuneen mit Reibune und Kohasion (Adhlsionl unter Annahme kreiszvlindrischer Gleifflachen, W.Emst und Sohn Verlag, Berlin (1927). Seed, H.B., and H.A. Sultan, Stability Analysis for a Sloping Core Embankment, Journal of the Soil Mechanics and Foundations Division. ASCE, 93 (1967) 69-84. Kapuskar, M.M., and J.P. Bardet, MULWED, A Computer Program for Simplex analysis of slope stability, Deuartment of Civil Eneineerine. Universitv of Southern California. Los Angeles (1989). Sanglerat, G., G. Olivari, and B. Cambou, Practical Problems in Soil Mechanics and Fou dation Engineering (2) Developments in Geotechnical Engineering, Elsevier Scie:ce Publishers Company Inc., N. Y., (1985) 207-212.

14.

15.

16. 17.

18.

NOTATIONS The following symbols are used in this paper:


Cf

mobilized cohesion along the base of slice i; values of factor of safety at initialization and after k steps; factor of safety of the r* vertex 5;

do, Fk F(zr)

348
Ii

base length of slice i; number of slices; number of independent variables for the failure mechanism; total resultant normal force on base of slice i; distance between base center of slice i and point 0; hypersphere radius for initial simplex; perpendicular distance of R, from the base center of slice i; resultant total load force on the base center of slice i; perpendicular distance of ASi from base center of slice i; resultant internal side force on the i* slice interface; increment of side force acting on slice i; resultant mobilized shear force along the base of slice i; average pore pressure acting on the base of slice i; coordinates of slide point i; coordinate of the first and last point of the lower boundary; elevation of the upper boundary at the i* interface;
elevation of the lower boundary at the iintetface;

n N
Ni Pi

r
i i i

i ASi T; i i Yi rnll mm
x1 yi

x*+1 ma1

Y zC>zR, ZRE,ZG

the local contracted, reflected, reflected plus expanded and globally contracted simplex vertex; the r* vertex of a simplex; simplex volume change operators for contraction, reflection, reflection plus expansion; inclination angle of the base of slice i w.r.t. the horizontal; constant and variable inclination angle of ASi w.r.t. to the horizontal; convergence parameter for iteration on factor of safety; convergence parameter for iteration on simplex, effective mobilized angle of friction at the base of slice i; unit weight of water; unit weight of dry, wet, saturated and buoyant soil; inclination angle of Ri w.r.t. the vertical; inclination angle of distance pi w.r.t. the vertical.

Received 1990

15 February

1990;

revised

version

received

18 May

1990;

accepted

18 May

Вам также может понравиться