Академический Документы
Профессиональный Документы
Культура Документы
net/publication/245411578
CITATIONS READS
39 807
2 authors, including:
Sarada K. Sarma
Imperial College London
64 PUBLICATIONS 2,103 CITATIONS
SEE PROFILE
All content following this page was uploaded by Sarada K. Sarma on 02 November 2014.
A new method of finding the critical slip surface in slope Cet article présente le développement d’une nouvelle
stability analysis is developed in this paper, based on the méthode de détermination de la surface de glissement
limit equilibrium technique with added stress acceptabil- critique, dans le cadre de l’analyse de la stabilité des
ity criterion. The stress acceptability criterion is based versants. Elle repose sur la technique d’équilibre-limite
on the limited strength of soil. The slip surface is devel- avec ajout de critère de tolérance de contrainte, lequel
oped in terms of the critical acceleration. The procedure est basé sur la résistance limitée du sol. La surface de
uses stress acceptability as a prerequisite to derive a glissement est développée en terme d’accélération cri-
system of non-linear equations to determine the slip sur- tique. La procédure utilise la tolérance de contrainte
face slice by slice upwards; no prior assumption of the comme pré-requis pour dériver un système d’équations
shape of the surface is needed. This surface simulta- non-linéaires et déterminer ainsi la surface de glissement
neously satisfies the kinematical acceptability criterion. coupe par coupe en remontant. Aucune hypothèse sur la
The whole procedure includes both homogeneous and forme de la surface n’est nécessaire. La surface satisfait
non-homogeneous slopes. In terms of the minimum factor simultanément au critère de tolérance cinématique. L’en-
of safety, the obtained critical slip surfaces compare well semble de la procédure s’applique aussi bien aux versants
with those obtained through methods based on optimisa- non-homogènes qu’aux versants homogènes. En terme de
tion techniques. In addition, the solution provides infor- facteur minimum de sécurité, les surfaces de glissements
mation on the critical acceleration and the kinematically critiques obtenues sont comparables à celles déterminées
acceptable interslice boundaries for the analysis of seis- à l’aide de méthodes basées sur des techniques d’optimi-
mic displacements of slopes using a multi-block sliding sation. La solution fournit également des informations
model. sur l’accélération critique et les limites intercoupes ciné-
matiquement acceptables pour l’analyse de déplacements
KEYWORDS: earthquakes; landslides; limit equilibrium meth- sismiques de versants utilisant un modèle de glissement
ods; slopes multi-blocs.
539
Delivered by ICEVirtualLibrary.com to:
IP: 86.0.218.4
On: Fri, 08 Oct 2010 13:24:01
540 SARMA AND TAN
& Bailey (1967) show that a correct solution from infinite finite element method (Yamagami & Yeta, 1988a; Zou et al.,
possibilities should satisfy the stress acceptability criterion. 1995; Kim & Lee, 1997; Pham & Fredlund, 2003). The
The methods mentioned above look at the internal factor second set of methods was first introduced by Zienkiewicz
of safety in the vertical planes only. No method looks at et al. (1975), and called the ‘strength reduction technique’
other planes within the slice to check their stress states. by Matsui & San (1992): it has been applied to various
Moreover, these conditions are not integral to the solution homogeneous and non-homogeneous slope stability problems
procedure, and rejection of solutions is based on experience. by numerous researchers (e.g. Griffiths & Lane, 1999). How-
ever, the output from such an analysis is not easy for the
engineering profession to interpret.
Methods to find the critical slip surface
The critical slip surface in slope stability analysis is the
surface that produces the minimum factor of safety in static New procedure
analysis or the critical acceleration in pseudo-static analysis. It is known that the limit equilibrium technique of slope
As the process of finding the critical slip surface is linked to stability analysis is statically indeterminate. As shown in the
the process of finding the minimum factor of safety (or Appendix, given a possible slip surface, interslice force
critical acceleration), it is natural to consider using an distributions exist that will produce any desired critical
optimisation method. There are many methods available to acceleration as a solution. However, this distribution (and
determine the critical slip surface; some of these methods therefore the desired critical acceleration) may not be accep-
and their basic features are presented below. table. Therefore the question arises as to whether an accep-
In enumerative schemes, slip surfaces are expressed by table critical acceleration can be found for a given slip
some pattern, and each surface is then analysed and the surface within the limit equilibrium technique. The basic
critical one chosen. This approach is possible for simple feature of the present method is to answer the question: Is it
geometrically definable surfaces such as circular failure possible to find an acceptable slip surface in a slope for a
surfaces (Bishop, 1955), and logarithmic spiral slip surfaces desired critical acceleration? If many acceptable slip sur-
(Lighthall, 1979; Prater, 1979). However, these failure modes faces are found with different desired critical accelerations
are special cases of general failure mode, and are usually then the surface with the minimum critical acceleration is
not the most critical, especially in non-homogeneous slopes. considered as the critical surface.
Variational calculus can be applied to the slope stability A new procedure is developed here to determine the
problem to find the critical factor of safety as an analytically critical slip surface along with the critical acceleration.
expressed function of the slip surface (Revilla & Castillo, Acceleration is applied to the slope to formulate the equa-
1977; Baker & Garber, 1978; De Josselin De Jong, 1980). tions, and the applied acceleration is also the critical accel-
Other calculus-based methods view the location of the eration, because we look for the slip surface where full
critical failure surface as minimising a non-linear function. strength is mobilised. Sarma (2004) formulated this proce-
Within this class of methods, we may include: Belman’s dure in its basic state, and it is now being modified and
dynamic-programming algorithm (Baker, 1980); the steepest enhanced. The procedure is within the framework of the
descent technique (Chen & Shao, 1988); the method of limit equilibrium technique (Sarma, 1979), satisfying accept-
unidirectional search (Celestino & Duncan, 1981); the sim- ability criteria in terms of stresses and kinematics. Moreover,
plex method (Narayan, 1982; Bardet & Kapuskar, 1989); the it provides information on the critical acceleration and the
compound iteration scheme (Li & White, 1986); and the kinematically acceptable internal boundaries for analysis of
gradient search method (Menon et al., 2001). McCombie & seismic displacements using a multi-block sliding model
Wilkinson (2002) mention that multiple minima exist in the (Sarma & Chlimintzas, 2001; Chlimintzas, 2003).
search space, and any calculus-based method is much more
likely to define a false minimum than the true minimum
because it depends on the initial trial, which may be difficult ANALYSIS
to choose. In the proposed method, the slip surface, which comprises
The random search procedure keeps the same objective a series of straight lines, is obtained slice by slice going
function to be minimised as the calculus-based methods but from downhill to uphill. The slip surface and the interslice
uses a random search technique to find the critical slip boundaries are not predefined. The equilibrium of the slice
surface (Boutrup & Lovell, 1980; Chen, 1992; Greco, 1996; and the acceptability criteria (defined later) determine the
Malkawi et al., 2001a, 2001b). However, this still involves slip surface and the interslice boundaries of the slice.
the analysis of a large number of slip surfaces, and a A slip surface in a slope is considered (Fig. 1), which is
solution approaching the optimum is found only by chance. at incipient failure under the action of a desired critical
McCombie & Wilkinson (2002) apply a simple genetic acceleration kc g. Note that the slip surface is unknown at the
algorithm to search for the minimum factor of safety using beginning but the critical acceleration is assumed. The slip
circular slip surfaces. Zolfaghari et al. (2005) apply the surface from a starting point O to point A is assumed solved
same technique to search for non-circular failure surfaces in already, and the corresponding interslice boundary at point
layered slopes. However, it is unable to identify the most A is AD. Therefore all the forces on the slip surface and the
critical factor of safety for some examples given in their interslice boundary up to the point A from O are known. In
paper. Cheng (2003) uses a simulated annealing technique order to define a new point on the slip surface, an arbitrary
for the location of critical circular and non-circular failure slice above the possible slip surface is considered. Within
surfaces under general conditions. Although the results ob- the single slice there are many unknowns, and some assump-
tained from any optimisation technique are mathematically tions have to be made before the problem can be solved
the minimum, some results may not be acceptable from soil mathematically.
mechanics principles.
There are two main sets of methods in static slope
stability analysis based on the finite element procedure to Geometric unknowns
determine the minimum factor of safety and the correspond- The slice i in Fig. 1 is composed of a possible slip surface
ing critical slip surface. The first set of methods searches for AB, a free surface CD, and two internal boundaries AD and
the critical slip surface using stress fields obtained from the BC. The points A and D are known. The aim of the solution
in which (xg i , yg i ) are the coordinates of the centre of þ 12 u iþ1 d iþ1 tanð iþ1 þ łÞ tan 9iþ1 ¼ 0
gravity of the slices. The point of application of Ni is given
by l i measured along the slip surface, and the point of The non-linear equations (19) and (20) along with equa-
application of Ei is given by z i measured along the interslice tions (1) to (4) are solved to obtain the angles Æ i and iþ1 ,
boundary from point A. using the trust-region dogleg method (Powell, 1970) avail-
After the critical slip surface is found, we can start from the able within MATLAB, which improves robustness when
first slice, where z1 ¼ 0. By assuming l i , z iþ1 can be starting far from the solution. The two angles define the
determined or vice versa. Proceeding to the last slice, the last geometry of the new slice. It can be seen from Fig. 3 that
l n is determined from the moment equilibrium of the last slice. the internal factor of safety is greater or equal to 1 on any
The values of z i and l i should lie within the slice, preferably in plane within the slice, and therefore the stress acceptability
the middle third. From analyses performed on many slopes, it is confirmed.
is seen that using the point of application of the normal force Thus, following the same procedure from slice 1, the
based on linear normal stress distribution on the slip surface complete slip surface can be obtained.
does not always produce an acceptable line of thrust, but very An interesting result of this method of solution is that, at
small variation from this assumption is sufficient to produce least for the first slice, the geometry of the slip surface and
an acceptable solution. Similarly, the position of z i does not the interslice boundary obeys the Rankine state, which is an
exactly agree with the linear distribution of stress distribution encouraging sign. For the intermediate slices, the solution
on the interslice boundary (see Example 1 below). This depends on the assumed stress distribution on the interslice
implies that the normal stress distribution is not exactly linear boundaries. The resulting line of thrust is used to check the
but very nearly so. It is possible for us to iterate on the normal robustness of the assumed stress distribution.
stress distribution to get exact results, but the amount of work
needed for this accuracy cannot be justified. Therefore a linear
stress distribution is used. Acceptability criteria for non-homogeneous slopes
In a non-homogeneous slope, a slice may pass through many
soil layers (Fig. 1), which divide the slice into several seg-
Acceptability criterion for homogeneous slope ments. Let us consider a simple example, as shown in Fig. 2.
In a homogeneous slope, the soil parameters are constant. There is a soil layer boundary FF9 passing through the slice at
At this stage, the acceptability criterion can be expressed as point B. Within the interslice boundary side the soil is homo-
follows. It is known that when Ł ¼ 0, the internal plane is geneous, but within the slip surface side it is non-homoge-
either the slip surface or the interslice boundary with mŁ ¼ neous, and so it is necessary to check the variation of mSŁ on the
1. Considering an angle Ł, positive as shown in Fig. 2, internal planes in slip surface side after Æ i and iþ1 are
within the slip surface side or the interslice boundary side, it obtained using equations (19) and (20). The solid line in Fig. 4
is expected that mŁ < 1 for an acceptable solution. It is also shows that mSŁ exceeds unity even if it has a local maximum
expected that if a negative angle Ł is selected, that is, a thin when Ł equal zero. If this happens, the assumption that mSŁ is
section into the body of the slope or into the next slice, the local
maximum
should be modified to
mŁ < 1. Therefore it is expected that mŁ is a maximum max mSŁ ðŁÞ ¼ 1 (21)
when Ł ¼ 0. The condition can be expressed as
dmŁ The new slice needs to be resolved from equations (21) and
¼ 0 and mŁ ¼ 1 when Ł ¼ 0 (18)
dŁ
Applying the above conditions to equation (6) gives
N i þ T i tan 9i 12 ªL2i sec sec 9i sinð9i Æ i Þ 1·00
9i Li
sec 9i sec 9ii
cosð i þ Æ i Þ
0·95
3 cos 9i þ 9ii Æ i i
u i Li
sec 9i cosð9i Æ i i Þ
cosð i þ Æ i Þ
(19) mè 0·90
Li
þ c9ii sec 9i sinð9i Æ i i Þ
cosð i þ Æ i Þ
c9i Li tanðÆ i þ i Þ 0·85
u i Li
Li tan 9i þ 12 ð u i þ u iþ1 Þ
2d i cosð i þ Æ i Þ
0·80
3 Li tanðÆ i þ i Þ tan 9i ¼ 0 0 0·05 0·10 0·15 0·20 0·25 0·30 0·35
è
Similarly, applying the above conditions to equation (9) Fig. 3. mŁ within slip surface side (note that trend within the
gives interslice boundary side is similar)
N n þ T n tan 9n
(20). After the new Æ i and iþ1 are obtained, mSŁ (Ł) is ªL n
þ 2 ªL n þ
1 2
hc sec end sec 9n sinð9n Æ n Þ
checked again, which is shown as a dashed line in Fig. 4, sin Æ n
and it satisfies the criterion.
Similarly, if the interslice boundary side is non-homoge- Ln Ln
þ c9n 1 un ¼0
neous, then tan Æ n 2 tan Æ n
(23)
max mBŁ ðŁÞ ¼ 1 (22)
0·1
EXAMPLES
Example 1 0
10 20 30 40 50
In this problem, a 1 : 5 homogeneous slope 10 m high is Slice number
used, and the soil properties are as shown in Fig. 7. The
pore water pressure coefficient, ru, is assumed to be constant Fig. 8. Variation in º i with number of slice
throughout the slope and equal to 0.2. The critical accelera-
tion is found to be 0.1706g and 0.1719g using the log-spiral
slip surface (Lighthall, 1979) and the circular slip surface
20
12
Critical slip surface
Present slip surface with kc 5 0·1701 16 Line of thrust based on fixed ëi
10
Log-spiral slip surface with kc 5 0·1706 Line of thrust based on variable ëi
Circular slip surface with kc 5 0·1719
8
Vertical distance: m
12
Vertical distance: m
6
8
4
2 4
0 0
c ¢ 5 5 kPa ö¢ 5 20°
ã 5 20 kN/m2 ru 5 0·2
22
24
24 210 0 10 20 30 40 50 60
210 0 10 20 30 40 50 60 Horizontal distance: m
Horizontal distance: m
Fig. 9. Lines of thrust corresponding to the two sets of º i and
Fig. 7. Comparison of critical slip surfaces in a homogeneous inclined interslice boundaries shown for Example 1 (inclined
slope interslice boundaries are shown)
6
After the different slip surfaces are obtained, the stress
c ¢ 5 17 kPa ö¢ 5 21°
acceptability is checked to see which result is more accep-
4
table. Because the details of Yamagami & Ueta’s and
c ¢ 5 5 kPa ö¢ 5 10°
Greco’s results are unknown, Sarma’s (2000) program is used
2
c ¢ 5 35 kPa ö¢ 5 28° to check the interslice factors of safety on the vertical
boundaries for the corresponding critical accelerations:
0
ð Nv U v Þ tan v þ cv Lv
22 Fv ¼ (26)
0 5 10 15 20 Tv
Horizontal distance: m
(a)
The results are shown in Figs 12(a)–(c). These show that
10
only the current procedure gives the interslice factor of
Critical slip surface of current study F 5 1·091 (or kc 5 0·0516)
Critical slip surface of Zolfaghari et al. F 5 1·24
8 (or kc 5 0·0704)
100
Vertical distance: m
60
2 40
0 20
1·8 slip surface and within the slip mass do not conflict with the
limited strength of soil, and the mass can slide along
the slip surface if the applied acceleration is bigger than the
1·6 critical. Factors of safety are found that are comparable to
those obtained from optimisation methods. The solution
provides information on the critical acceleration and the
1·4 kinematically acceptable internal boundaries for analysis of
seismic displacements of slopes using a multi-block sliding
model. In the existing slope stability analysis methods within
1·2
the limit equilibrium technique, the acceptability criteria are
checked after the solution is obtained, and quite often these
1·0
are found to be unacceptable in some parts of the slip
20 40 60 80 100 surface. It is usually left to the experience of the users to
Slice number declare whether the solutions are acceptable. This paper is
(a)
the first time that the acceptability criteria have been rigor-
3·5 ously satisfied.
3·0
ACKNOWLEDGEMENTS
The authors are grateful to Dr John Douglas for his
2·5
comments during the preparation of the paper. Ding Tan
Internal factor of safety
0·5 APPENDIX
The basics of the limit equilibrium technique for critical
acceleration analysis (dry slope) are as follows (Fig. 13).
0
20 40 60 80 The slope is defined by y ¼ g(x).
Slice number The slip surface is defined by y ¼ f (x).
(b) The assumed line of thrust is defined by y ¼ t(x).
The height of the line of thrust from the slip surface z(x) ¼ t f.
2·5
The slope of the slip surface is given by Æ ¼arctan( f 9(x)).
The slip surface is assumed to start at x = 0 and end at x = B.
For an elemental slice of width dx:
2·0
the weight of the slice dw ¼ ª(g f )dx;
the normal force N ¼ n dx secÆ;
Internal factor of safety
g(x)
dx
dx
y f (x) kcdW
E 1 dE
E
dW
t(x)
x X X 1 dX
dz
Fig. 13 Schematic map for a slope and the forces and their acting points in one slice
For simplification, assume no pore water pressure, so that 9n ¼ n . E 0z þ E9 2z9 þ f 9
1 þ af 9
þ E½ z0 þ f 0
Also write (for ease of presentation) a f9
tan9 ¼ a
ð1 þ af 9Þ k c w9 ð a f 9Þ w9 c9 1 þ f 9 2
¼
ð a f 9Þ
Then from equations (27), (28) and (30) we get
k c w0ð g f Þ k c w9ð g9 f 9Þ
n ð1 þ af 9Þdx ¼ dw þ dX c9 f 9 dx (31) (36)
2 2
n ð a f 9Þdx ¼ k c dw þ dE c9dx (32)
1 þ af 9
¼ ªk c ð g f Þ ð g9 f 9Þ ð g f Þ
a f9
Then from equations (31) and (32) we get
1 þ f 9 2 c9
ð a f 9Þðdw þ dX c9 f 9 dxÞ ðg f Þ ª
(33) ð a f 9Þ
ð1 þ af 9Þð k c dw þ dE c9dxÞ ¼ 0
Therefore, for a given slope and a given slip surface, assuming that
Dividing by dx throughout, we get the line of thrust is known and the critical acceleration is also
assumed known, equation (36) is a second-order linear differential
ð a f 9Þð w9 þ X 9 c9 f 9Þ ð1 þ af 9 Þð k c W 9 þ E9 c9Þ ¼ 0 equation with variable coefficients. With the known boundary
conditions that E ¼ 0 at x ¼ 0 and at x ¼ B, the above equation
or can be solved numerically or otherwise. There is a second set of
boundary conditions that X ¼ 0 at x ¼ 0 and at x ¼ B, which are
ð a f 9ÞX 9 ð1 þ af 9ÞE9 automatically satisfied by the solution of E, because z ¼ 0 at x ¼ 0
(34) and at x ¼ B from equation (29b).
¼ ð1 þ af 9Þk c w9 ð a f 9Þw9 c9 1 þ f 92 This solution clearly shows that, without any control on the
interslice forces E or X or both within the slipped mass, we can get
infinite numbers of solutions of kc because, in the above formulation,
We eliminate X from equations (34) and (29c) by differentiating we can choose any kc and any line of thrust. Sarma (1973) also
equation (29c) first and replacing X9 from equation (34): mentions the infinite number of solutions in the limit equilibrium
Eð z0 þ f 0Þ þ E9ð2z9 þ f 9 Þ þ E 0z technique. Some of these solutions may not be acceptable. Therefore
it is essential that some acceptability criterion is provided to curtail
½ð1 þ af 9ÞE9 þ ð1 þ af 9Þ k c w9 ð a f 9Þw9 c9 1 þ f 92 the number of solutions.
The available limit equilibrium solutions apply some sort of
ð a f 9Þ
control over the interslice forces to obtain reasonable results.
k c w 0ð g f Þ k c w9ð g9 f 9Þ However, these still may not be perfectly acceptable. The method
þ þ ¼0 presented in the paper shows a way to apply acceptability criteria
2 2
Rearranging: with rigour.
1 þ af 9
E 0z þ E9 2z9 þ f 9 þ E½ z0 þ f 0
a f9
NOTATION
ð1 þ af 9Þ k c w9 ð a f 9Þ w9 c9 1 þ f 9 2 (35)
¼ c9 cohesion in terms of effective stresses on slip surface
ð a f 9Þ c9 average cohesion in terms of effective stresses on interslice
k c w 0ð g f Þ k c w9ð g9 f 9Þ boundary
c9Ł average cohesion in terms of effective stresses on internal
2 2 plane
E9 normal force on interslice in terms of effective stresses
Note that EŁ implied normal force in terms of total stresses on internal
plane within interslice boundary side
w9 ¼ ªð g f Þ
EL normal force in terms of total stresses on small section on
w 0 ¼ ªð g9 f 9Þ slice boundary
F factor of safety
Fv factor of safety on vertical interslices
Therefore kc critical acceleration