Download as pdf or txt
Download as pdf or txt
You are on page 1of 8

INVITED PAPER

Limit equilibrium, strength summation and strength reduction


methods for assessing slope stability

J. Krahn
GEO-SLOPE International, Calgary, Alberta, Canada

ABSTRACT: Traditional limit equilibrium formulations such as the Bishop, Spencer or Morgenstern-Price
methods have some inherent weaknesses. To overcome these weaknesses, efforts have been directed at using
finite element computed stresses for assessing the stability of slopes. This paper reviews and compares two
methods referred to as the Strength Summation Method and the Strength Reduction Method, and concludes that
the Strength Reduction Method is more suitable for serviceability (deformation) analyses while the Strength
Summation Method is more suitable for stability analyses.

1 INTRODUCTION Limit equilibrium formulations have their short-


comings as was vividly illustrated by Krahn (2003)
Stability is always a key issue in the analysis, assess- in the 2001 R.M. Hardy keynote address in Calgary.
ment or design of most earth structures. After all, the At the heart of the limitations is the fact that limit
prime objective is to ensure that the structure does equilibrium formulations do not in any way consider
not collapse and/or deform to the point that it causes strains and displacements so they certainly do not sat-
unmanageable property damage or in the worst case, isfy strain and or displacement compatibility. This is
the loss of human lives. It is by far the most common the key missing piece of physics. The consequence
type of numerical analysis in geotechnical engineering is that the interslice and slip surface stresses are not
and has become routinely used in practice. necessarily representative of the actual field stresses.
Historically, stability analyses have been completed Instead, the stresses and related forces computed in
using limiting equilibrium (LE) formulations such as a limit equilibrium formulation are the forces which
the well known Bishop, Spencer and Morgenstern- ensure that each slice is in force equilibrium, and that
Price methods.The potential sliding mass is commonly each slice has the same local factor of safety. In real-
discretized into slices as shown in Figure 1. The forces ity, however, the local factor of safety is not constant
on each slice, treated as a free body, are resolved to sat- across the slip surface and is not the same for each
isfy horizontal and vertical force equilibrium together slice.
with the overall moment equilibrium of the entire To overcome the short comings of limit equilibrium
potential sliding mass. The factor of safety is defined formulations, various approaches have been proposed
as the value by which the soil strength must be reduced to use the results from finite element (FE) stress-strain
to achieve the limiting equilibrium condition. computations to assess the stability of earth struc-
tures. Of these, two approaches are gaining favorable
acceptance for use in practice.
One approach, known as the Strength Reduction
method, is based on using the FE results in isolation
without inheriting any legacy from LE procedures. The
factor of safety is obtained by weakening the soil in
steps until the slope “fails.” The factor of safety is
deemed to be the factor by which the soil strength
needs to be reduced to reach failure.
An alternative approach, called the Strength Sum-
mation method, retains some of the concepts inher-
ent in limit equilibrium procedures and combines
Figure 1. Slices and forces in a potential sliding mass. them with finite element computed stresses. Trial slip

311
© 2007 Taylor & Francis Group, London, UK
surfaces are used as in traditional LE analyses. The the finite element computed stresses in a stability anal-
ratio of the available shearing resistance when divided ysis overcomes one of the main weaknesses inherent
by the mobilized shear along the entire trial slip sur- to a limit equilibrium formulation.
face is deemed to be the factor of safety. The trial slip
surface with the lowest factor of safety is considered
to be representative of the critical potential mode of 3 STRENGTH SUMMATION METHOD
failure.
The objective of this paper is to review the details For discussion purposes, using FE stresses in a
of these two approaches and to compare their advan- LE framework will be referred to as the Strength
tages and limitations, as well as to comment on the
applicability of these methods for use in geotechnical
engineering practice. Factor of
Safety
L.E. = 1.145
2 LIMIT EQUILIBRIUM FORCES AND F.E. = 1.318
STRESSES
Downloaded by [Cornell University] at 02:14 05 October 2016

As noted in the introduction, stresses and forces com-


puted using LE methods are not necessarily represen-
tative of the actual insitu stresses. Consider the cases
90
in Figures 2 and 3, which compare the stresses along a 80
slip surface obtained from both a LE and FE calcula-
70
tion. In the toe slip scenario the two stress distributions
Normal Stress

60
vary significantly near the toe where there is a signifi-
50
cant concentration of shear stresses. For the deeper slip
40
surface, LE and FE stress distributions are somewhat
30
similar since there are no significant stress concentra-
20
tions along the slip surface and the normal slip surface
10 L.E. F.E.
stresses are reflective of the overburden.
0
The reason for the differences in the stress dis- 0 5 10 15 20 25 30
tributions and resulting factors of safety lies in the Slice number
fundamental assumptions inherent in limit equilibrium
formulations. LE methods involve iterative procedures Figure 2. Normal stress distributions along a toe slip
since the factor of safety equations are non-liner and surface.
in the more rigorous methods such MorgensternPrice
and Spencer a second level of iterations are required
Factor of
to find the slice forces that result in the same factor Safety
of safety with respect to force and moment equilib-
rium. As pointed out by Krahn (2003), fundamentally L.E. = 1.804
the iterations are required to meet two conditions F.E. = 1.849
which are:
(1) to find the forces acting on each slice so the slice
is in force equilibrium; and
180
(2) to find the forces on each slice that will make the 160
factor of safety the same for each slice. 140
Normal stress

The limit equilibrium method of slices is based 120


100
purely on the principle of statics; that is, the summation
80
of moments, vertical forces, and horizontal forces. The 60
method says nothing about strains and displacements 40
and as a result, it does not satisfy displacement com- 20
L.E. F.E.
patibility. It is this key piece of missing physics that 0
creates many of the difficulties with the limit equilib- 0 5 10 15 20 25 30
rium method. Including some form of a stress-strain Slice number
constitutive relationship such as a stress-strain finite
element analysis, that supplies the missing piece of Figure 3. Normal stress distributions along a deep slip
physics and ensures displacement compatibility. Using surface.

312
© 2007 Taylor & Francis Group, London, UK
Summation Method. The quandary of not getting the then used to obtain the stress distribution along the slip
correct stresses in a limit equilibrium analysis can be surface is as follows:
overcome by using finite element computed stresses
– For slice 1, find the element that encompasses the
along a trial slip surface.
x-y coordinate at the base mid-point of the slice.
Figure 4 shows a simple 45-degree slope discretized
– Compute σx , σy and τxy at the mid-point of the slice
into finite elements. Using a simple gravity turn-on
base.
technique, the stresses in the ground can be com-
– Determine the inclination (α) of the base of the slice
puted. Using a linear-elastic constitutive relationship,
from the limit equilibrium discretization.
the vertical stresses are as shown in Figure 5. The
– Compute the slice base normal and shear stress
basic information obtained from a finite element stress
using ordinary Mohr circle techniques.
analysis is σx , σy and τxy within each element.
– Compute the available shear strength for the com-
Worth noting at this stage is the 50 kPa contour
puted normal stress.
which is not a constant distance from the ground sur-
– Multiply the mobilized shear and available strength
face. The contour is closer to the surface under the toe,
by the length of the slice base to convert stresses
which means that the vertical stress is not just influ-
into forces.
enced by the overburden weight. It is also affected by
– Repeat the process for each slice in succession up
the shear stress within the slope.
to Slice number n.
The finite element computed stresses can be
Downloaded by [Cornell University] at 02:14 05 October 2016

imported into a conventional limit equilibrium anal- Once the mobilized and resisting shear forces are
ysis. The stresses σx , σy and τxy are known within available for each slice, the forces can be integrated
each element and from this information the normal over the length of the slip surface to determine a
and mobilized shear stresses can be computed at the stability factor. The stability factor is defined as:
mid-point of each slice base. The procedure is to take
the known σx , σy and τxy stresses at the Gauss numer-
ical integration point in each element and project and
average the values at the nodes. With σx , σy and τxy
known at the nodes, the same stresses can be computed
at any other point within the element. The procedure where, Sr is the total available shear resistance and
Sm is the total mobilized shear along the entire length
of the slip surface. Similar stability factor expressions
have been presented by others (Kulhawy 1969, Naylor
1982).
1 Figure 6 shows a potential sliding mass discretized
1 10 metres into slices superimposed on the finite element mesh.
Using the Strength Summation method outlined above,
the stability factor for this slip surface is 1.318. This
compares with a Morgenstern-Price factor of safety
of 1.145 (constant interslice function), resulting in a
difference of 15%.
The reason for the difference in the margin of
safety is primarily related to the significantly different
normal stress distributions at the toe of the slip sur-
Figure 4. A finite element mesh for establishing in situ face. The finite element and limit equilibrium normal
stresses.

50

100

200

50

100

200

Figure 5. Vertical stress contours. Figure 6. A toe slip surface on a finite element mesh.

313
© 2007 Taylor & Francis Group, London, UK
stress distributions for this particular slip surface were slip surface while the local safety factors based on the
presented earlier in Figure 2. Localized shear stress FE computed stresses vary along the slip surface. The
concentrations are, of course, not captured in a limit unrealistic LE stresses are reflective of the incorrect
equilibrium formation where the slice base normal is local safety factors.
derived primarily from the slice weight. This is one of In summary, using finite element computed stresses
the limitations of the limit equilibrium method. allows for a variable local safety factor along the
The situation is somewhat different for a deeper slip slip surface which in turn leads to stresses better
as shown in Figure 3. The finite element and limit representative of actual field conditions.
equilibrium normal stress distributions along the slip
surface are much more similar for this case. Conse-
quently the stability factor based on finite element 4 STRENGTH REDUCTION METHOD
stresses is almost the same as the Morgenstern-Price
factor of safety. The stress-based stability factor is As noted in the introduction, an alternative approach
1.849 while the Morgenstern-Price factor of safety is is to artificially weaken the soil in an elastic-plastic
1.804. This shows that, when the normal stress distri- finite element analysis until the slope “fails” (Daw-
bution along the slip surface is fairly representative of son et al 1999, Griffiths & Lane 1999). Numerically,
the actual ground stresses, the limit equilibrium factor this occurs when it is no longer possible to obtain a
of safety is as good as a stress-based factor of safety. converged solution. The finite element equations for
Downloaded by [Cornell University] at 02:14 05 October 2016

Figures 7 and 8 show the actual variation in the local a stress-strain formulation are in essence equations of
safety factors for each slice along the two slip surfaces equilibrium. Not being able to obtain a converged solu-
shown in Figures 2 and 3. As illustrated in these dia- tion therefore infers the system is beyond the point of
grams the LE factor of safety is constant along the limiting equilibrium.
The strength reduction factor (SRF) is defined as,

F.E L.E
1.6

1.5

1.4
where φf and cf are the soil parameters at “failure”.
Factor of safety

Figure 9 shows a 2:1 homogeneous slope which


1.3
has Mohr Coulomb strength parameters of c = 5 kPa
and φ = 28 degrees. Using the Strength Summation
1.2
Method and a gravity turn-on analysis with linear
elastic soil properties, the critical mode of failure is
1.1 represented by the shaded zone in Figure 9. The shaded
zone represents slip surfaces with factors of safety
1.0 between 1.55 and 1.60. The white line in the zone
0 5 10 15 20 25 30 represents the slip with an overall F of S of 1.55.
Slice number The distribution of the local safety factors along
the slip surface is shown in Figure 10. The last slice
Figure 7. Local safety factors for the toe slip. (#30) has a local safety factor of 0.96 indicating that
the FE computed stress distribution represents some
F.E. L.E over-stressing in this area.
4 The next step is to re-distribute the stresses using
an elastic-plastic analysis. The strengths are summed
3
along the slip surface and the local safety factor
Factor of Safety

3
2
2
1
1
0
0 5 10 15 20 25 30
Slice number

Figure 8. Local safety factors for the deep slip. Figure 9. Zone of potential failure.

314
© 2007 Taylor & Francis Group, London, UK
for Slice 30 changes to 1.10. The overall global to obtain a solution as a structure moves towards
factor of safety obtained using the strength summation the point of instability. Smaller and smaller load-re-
procedure changes only slightly from 1.549 to 1.547. distribution steps are required together with an ever
Now, for illustrative purposes, we can apply the increasing number of iterations. Furthermore, there
Strength Reduction Method to the same slope and can be many reasons for lack of convergence besides
reduce the strength in steps, comparing the Strength just the reduction in strength, including incorrect
Summation computed factors of safety (F of S) to the boundary conditions or incorrect initial insitu stresses
computed Strength Reduction Factors (SRF) as shown just to name a few possibilities.
in Table 1. Moreover, the displacement associated with the arti-
The lowest strength for which it was possible to ficial strength reduction as shown in Figure 12 reveals
obtain a converged solution represents a SRF of 1.42 no distinct zone of slip such as is achievable with the
(Case 4). This says that the factor of safety for the Strength Summation result shown in Figure 9.
original “unreduced strength” slope using the Strength The strength reduction method is sometimes touted
Reduction method is 1.42. The more conventional fac- (Griffiths & Lane 1999) as providing a clearer picture
tor of safety based on Strength Summation Method of the potential failure mechanism than using specific
gave a value of 1.55. In terms of criterion for design slip surfaces. The results presented here do not support
this is a significant difference. this claim.
Also, worth noting is that the factor of safety
Downloaded by [Cornell University] at 02:14 05 October 2016

using the Strength Summation Method for the low-


est strength (Case 4) is 1.07. This shows the difficulty 6.00
of obtaining a converged FE solution when the local
safety factor is indeed 1.0 along the entire slip sur- 5.00
Case 1 Case 2 Case 4
face. This is further illustrated by the local safety factor
4.00
Safety Factor

distributions in Figure 11. For the case with the low-


est strength the local safety factor distribution is near 3.00
unity along most of the slip except near the crest.
These results reveal that one of the difficulties 2.00
with the Strength Reduction approach is the imprecise
mechanism of defining the point of non-convergence. 1.00

Numerically, it becomes exponentially more difficult


0.00
0 5 10 15 20 25 30 35
6.00 Slice #
Linear Elastic
5.00
Figure 11. Local safety factors distributions with strength
4.00 reductions.
Safety Factor

3.00

2.00 2
1
1.00 10 m

0.00
0 5 10 15 20 25 30
Slice #

Figure 10. Local safety factors based on linear-elastic


stresses.
2

Table 1. Comparison of F of S and SRF. 1


10 m

 
φ c F of S
Case degrees kPa summation SRF

1 28.00 5.00 1.55 1.00


2 25.00 4.38 1.35 1.14
3 20.80 3.57 1.09 1.40
4 20.53 3.52 1.07 1.42
Figure 12. Displacement field at “failure” magnified 30x.

315
© 2007 Taylor & Francis Group, London, UK
Figure 14. A block slip with an optimized shape.

intended function before reaching the point of col-


lapse. Stated another way, the serviceability of the
structure has failed while the factor of safety is greater
Downloaded by [Cornell University] at 02:14 05 October 2016

than 1.0. A typical example is the permanent deforma-


tion associated with earthquake shaking. The dynamic
Figure 13. A circular slip with an optimized shape. forces may cause the factor of safety to temporarily
fall below 1.0 and/or there may indeed be a strength
Finally, using non-convergence as a criterion to loss due to the generation of excess pore-pressures.
obtain the desired solution is unconventional in engi- In situations where there is indeed a real strength
neering, where generally the practice is to seek con- loss, the strength reduction numerical procedure
verged solutions on which to make design decisions. holds considerable promise. The related stress re-
distribution can potentially provide a reasonable pic-
ture of the permanent deformations arising from the
5 SLIP SURFACE SHAPES actual strength loss provided the strength loss has not
led to a total collapse of the structure.
One of the arguments against the use of trial slip sur- Using the strength reduction stress re-distribution
faces is that the slip surface has a defined geometric procedure seems to have more application for analyz-
shape such as the arc of a circle or piece-wise linear ing permanent deformations than assessing stability.
segments or a combination of the two. This however In summary, the method is more useful for looking at
is no longer a requirement. There are other proce- serviceability than stability.
dures now available whereby the critical slip surface
can have a shape that does not necessarily follow
any defined geometric shape. One of the techniques 7 LINEAR-ELASTIC VERSUS
is called dynamic programming (Yamagami & Ueta ELASTIC-PLASTIC STRESSES
1988). Another method is referred to as optimiza-
tion (Krahn 2004a). In optimization, segments of the When looking at the overall global factor of safety
slip surface are moved within an elliptical search area of a slope, a finite element analysis based on linear-
using a statistical random walk procedure based on the elastic properties is generally more than adequate.
Monte Carlo method. Krahn (2003) presented the case shown in Figure 15.
Figures 13 and 14 show two cases with optimized The local safety factor distributions shown in Fig-
slip surfaces relative to surfaces with pre-defined geo- ure 15 for one particular slip surface are for both a
metric shapes. Not only are the optimized slip surface linear-elastic and an elastic-plastic analysis. The dis-
shapes intuitively more realistic but they also generally tributions vary significantly along the slip surface but
give a lower factor of safety. the global factors of safety are nearly identical. For
The optimization can be used on either limit equi- the linear-elastic case the global factor of safety is
librium formulations or finite element stresses. 1.206 and for the elastic-plastic case the global factor
of safety is 1.212, less than half a percent difference.
A similar conclusion was reached independently by
6 PERMANENT DEFORMATIONS Stianson et al. (2004). After analyzing a wide range
of cases, they concluded, “…that a LE (linear-elastic)
In certain cases earth slope deformations may become stress analysis is sufficient to calculate the overall
sufficiently large so the structure cannot serve its stability of a slope.”

316
© 2007 Taylor & Francis Group, London, UK
8 CONTINUED USE OF LIMIT
EQUILIBRIUM METHODS

Using finite element computed stresses in a slope sta-


Phi = 20
bility analysis overcomes some of the short-comings of
2 traditional limit equilibrium formulations. The down-
C = 5 kPa
1 side is that it is necessary to do a finite element
analysis, which is sometimes not trivial procedure for
some cases. This is particularly true for natural slopes.
Accurately establishing the stress conditions in a nat-
Linear elastic Elastic plastic
ural slope involves modeling the geologic process that
created the slope, which could be a lengthy and com-
2.0 plicated study. A simple gravity turn-on analysis is
usually a rough estimate of the stress conditions in the
1.8 slope. In light of this, doing a limit-equilibrium fac-
1.6
tor of safety analysis is likely as appropriate or even
Factor of Safety

more useful and trustworthy as attempting any finite


1.4 element based approach.
Downloaded by [Cornell University] at 02:14 05 October 2016

Further comment on this is not warranted in the con-


1.2
text of this paper except to note that there is nothing
1.0 wrong with conventional limit equilibrium stability
analyses provided the method is recognized for hav-
0.8 ing inherent limitations and that expectations are not
0.6
placed on the method which it is not able to provide
0 5 10 15 20 25 30 such as realistic insitu stress distributions.

Figure 15. Local safety factor distribution for linear-elastic


and elastic-plastic stress fields.
9 SUMMARY AND CONCLUSIONS

The results from a linear-elastic analysis may mean Limit equilibrium numerical formulations for assess-
that the computed stresses in some zones are higher ing the stability of earth slopes do have some lim-
than the available soil strength. This manifests itself as itations. In an effort to overcome these limitations,
a local factor of safety of less than 1.0 for some slices, methods have been proposed which make use of finite
which is not physically possible. Ideally, non-linear element stress-strain analyses. Of these, two seem to
constitutive models should be used to redistribute the hold some promise for use in engineering practice.
stresses such that the applied stresses do not exceed One is known as the Strength Reduction Method and
the strength. However, using non-linear constitutive the other is referred to in this paper as the Strength
relationships greatly complicates the analysis, primar- Summation Method.
ily because of the associated numerical convergence The Strength Reduction Method is based on artifi-
issues. cially weakening the soil in an elastic-plastic analysis
Ignoring local safety factors that are less than unity until it is no longer possible to achieve convergence,
is not all that serious. Physically, it means that neigh- which is then taken as the point of failure. The factor
boring slices have a local safety factor that is too high. by which the strength is reduced is deemed to be the
Since all the mobilized and resisting shear forces are resulting factor of safety for the original slope.
tallied along the entire slip surface, local irregularities The Strength Summation Method tallies the avail-
are smoothed out and therefore have little effect on the able shear resistance and mobilized shear as deter-
total forces which are used in computing the global mined from a finite element analysis along a trial slip
factor of safety for the entire sliding mass. This is an surface. A ratio of the available shear resistance to the
indirect form of averaging but not nearly to the extent mobilized shear is taken as the factor of safety.
that is inherent in the limit equilibrium formulation The Strength Reduction Method suffers from the
where the factor of safety is the same for all slices. imprecise criteria for defining failure, which is repre-
This is of significance for using finite element com- sented by a state of non-convergence. Convergence is a
puted stresses in practice for stability analyses. The difficult numerical issue as a structure moves towards
fact that a linear-elastic analysis does not result in the point of instability. The problem is serious enough
convergence issues is a great benefit and for this rea- to impede the use of this method in routine practice.
son alone provides a considerable advantage over the The Strength Summation Method together with
Strength Reduction method. a simple linear elastic FE analysis is adequate for

317
© 2007 Taylor & Francis Group, London, UK
assessing the overall global factor of safety of earth out using the GeoStudio 2004 geotechnical software
slopes. The big advantage of this method is that it is packages SLOPE/W and SIGMA/W (Krahn 2004a, b).
not fraught with numerical convergence difficulties. It
also presents a more distinct picture of a failure zone
than the Strength Reduction Method. REFERENCES
The Strength Reduction Method has more applica- Dawson, E.M., Roth, W.H. & Drescher, A. 1999. Slope sta-
tion for assessing serviceability (permanent deforma- bility analysis by strength reduction. Geotechnique 49(6):
tion) than for assessing stability. 835–840.
The perceived limitation of methods based on trial Griffiths, D.V. & Lane, P.A. 1999. Slope stability analysis by
slip surfaces with pre-defined geometric shapes has finite elements. Geotechnique 49(3): 387–403.
now been overcome. Procedures are available for find- Krahn, J. 2003. The 2001 R.M. Hardy Lecture: The Limits
ing critical slip surfaces which have a general shape of Limit Equilibrium Analyses. Canadian Geotechnical
that leads to the lowest factor of safety. Journal 40: 643–620.
Krahn, J. 2004a. Stability Modeling with SLOPE/W; An
Traditional limit equilibrium formulations remain
Engineering Methodology. Calgary: GEOSLOPE Inter-
acceptable for assessing the margin of safety of earth national Ltd., pp. 101–104.
slopes provided the limitations are understood and Krahn, J. 2004b. Stress and Deformation Modeling with
results are not expected that the methods cannot SIGMA/W; An Engineering Methodology. Calgary:
provide. GEO-SLOPE International Ltd.
Downloaded by [Cornell University] at 02:14 05 October 2016

Finally, if a Strength Reduction approach is used to Kulhawy, F.H. 1969. Finite Element Analysis of the Behavior
assess the stability of a slope it would seem prudent to of Embankments. Ph.D. Thesis, University of California
compare the results with a Limit Equilibrium analysis at Berkley, California, USA.
with which we have decades of experience. Until the Naylor, D.J. 1982. Finite elements and slope stability. In
Numerical Methods in Geomechanics. D. Reidel Publish-
Strength Reduction Method has been used routinely in
ing Company.
practice for many years the method should not be used Stianson, J.R., Chan, D. & Fredlund, D.G. 2004. Com-
in isolation. paring slope stability analysis based on linear elastic
or elasto plastic stresses using dynamic programming
techniques. In 57th Canadian Geotechnical Conference,
Québec City, 9 pp.
10 NOTE Yamagami, T. & Ueta, Y. 1988. Search for the critical slip
lines in finite element stress field by dynamic program-
Stability analyses, finite element analyses, Strength ming. In Proceedings of the 6th International Confer-
Summation analyses and Strength Reduction analy- ence on Numerical Methods in Geomechanics, Innsbruck,
ses presented and discussed in the paper were carried pp. 1347–1352.

318
© 2007 Taylor & Francis Group, London, UK

You might also like