Professional Documents
Culture Documents
net/publication/260166270
CITATIONS READS
4 294
4 authors, including:
Some of the authors of this publication are also working on these related projects:
Experimental study on dynamic properties of soil materials in small strain View project
All content following this page was uploaded by Mohammad Khosravi on 14 February 2014.
INTRODUCTION
Seismic instability of soil slopes can be grouped into two main categories (Kramer
1996): 1) Inertial instability, where dynamic stresses that are induced in a soil slope
by earthquake shaking temporarily exceed the strength of the soil, leading to
accumulating slope displacements, and 2) Weakening instability, where earthquake
shaking weakens a soil such that it cannot remain stable under earthquake-induced
stresses (e.g., flow liquefaction and cyclic mobility).
Geo-Congress 2013
Geo-Congress 2013 © ASCE 2013 1311
whether or not more sophisticated analysis techniques are warranted (e.g., Duncan
1992, Duncan and Wright 2005). To conduct these analyses, the effects of an
earthquake on a slope are represented by applying pseudo-static horizontal and/or
vertical accelerations to a potentially unstable mass of soil (Duncan 1992). The factor
of safety against seismic slope failure is then calculated in much the same way as in
static slope stability LE analyses. In a similar fashion as static LE methods, pseudo-
static (earthquake) analysis approaches for assessing seismic slope stability require
prior assumptions for the definition of the failure slip surface and the computation of
the interslice forces. However, a more significant criticism of this analysis approach
is that it only provides an index of stability/instability (i.e. a pseudo-static Factor of
Safety) but no information on the deformations that are associated with the failure of
the slope (Kramer 1996). For transient loading methods such as earthquakes, a
temporarily low factor of safety value does not necessarily indicate a long-term
problem for the slope, and earthquake-induced displacements are a much better
predictive measure of slope performance.
Sliding block based methods, which are typically formulated after the approach
first presented by Newmark (1965), are commonly used to evaluate the expected
displacement that will be obtained when a slope is subjected to a given shaking event.
Unlike in a load-based analysis approach, displacement is used to quantify failure and
damage within a given sliding block model, instead of a factor of safety alone.
Therefore, compared to load-based analysis approaches, displacement-based
techniques offer a more realistic way of estimating likely hazards (e.g., Chlimintzas
2003). Typically, most sliding block models determine permanent deformations by
doubly-integrating those accelerations in a given earthquake time history that are in
excess of the yield acceleration of the slope (KC). Unfortunately, as yield acceleration
values for a given slope are typically determined using LE-based techniques, the
general shortcomings that are associated with load-based techniques are still
applicable for sliding block analyses (e.g., Duncan and Wright 2005).
More recently, as access to cost-effective computing power has increased, there has
been an increasing trend towards the use of more advanced numerical analysis
methods (FE, FD) for both static and seismic analyses of soil slopes (Duncan 1992).
An advantage of numerical analyses over LE procedures is that they can be used to
calculate expected slope deformations if a realistic stress-strain relationship is
assigned to the soil mass. Advanced constitutive models are also capable of
simulating the effect of the actual time history on the response of the slope to the
ground motion, and other time-dependent or deformation-dependent phenomena
during the earthquake (such as excess pore pressure generation). If performed
properly, dynamic numerical approaches can provide more accurate, refined solutions
to the problem than load-based and deformation-based analyses (Kramer 1996).
Geo-Congress 2013
Geo-Congress 2013 © ASCE 2013 1312
Geo-Congress 2013
Geo-Congress 2013 © ASCE 2013 1313
gravitational acceleration to the slope, (b) Apply horizontal body force loading
incrementally until the slope collapses, and (c) Calculate the factor of safety for each
of the incremental horizontal loadings and find the horizontal acceleration
corresponding to FS = 1 (Swan and Seo 1999). This acceleration is the critical (yield)
acceleration.
In the strength reduction method (SRM), the shear strength parameters of the slope
material are successively reduced until slope equilibrium can no longer be maintained
Downloaded from ascelibrary.org by University of California, Davis on 02/13/14. Copyright ASCE. For personal use only; all rights reserved.
(Zeinkiewicz et al. 1975). In this study, the same technique as proposed by Xu et al.
(2009) was utilized to implement the SRM technique into ABAQUS. In this
technique, the Mohr-Coulomb soil strength parameters (c and ϕ ) are defined using a
linearly varying adjustable strength weighting value for the purpose of strength
reduction – in ABAQUS, the “temperature” (θ) strength function was used for this
purpose:
c (θ) = (1 − 0.9 × θ)c′ (1)
ϕ (θ) = tan ((1 − 0.9 × θ)tan(ϕ′)) (2)
It should be noted that, in this approach, the thermal expansion coefficients of the
materials are assumed to be zero; therefore, user induced variations in temperature
will not result in any thermal stresses and strains, but can be used to effectively
reduce the strength of the soil until slope failure occurs. Utilizing these definitions,
the factor of safety as a function of the strength weighting value (“temperature” in
ABAQUS) is defined as (Xu et al. 2009):
1
= (3)
(1 − 0.9 × θ )
Finite element analyses using this technique are performed in five stages: (a)
Establish the geostatic stress field by first applying a vertical gravitational
acceleration to the slope. (b) Apply a test horizontal body force, (c) Increase the
strength weighting factor (“temperature” in ABAQUS) at each node from 0.0 to 1.0.
Using the approach that is described herein, the strength parameters of the soil
material are correspondingly reduced according to Eq. (1) and (2). (d) Calculate the
factor of safety according to the critical value of temperature using Eq. 3. (e) Repeat
steps (b) through (d) for different horizontal body forces to identify the body force
that corresponds to a factor of safety of 1. All the analyses in this study were
performed under plane strain conditions. For the sake of simplicity, no water
pressures were considered in the analyses in this paper. The height of the slope is
assumed to be 20 m. The slopes are inclined at two different slope angles, 2H:1V,
1H:1V (Fig. 1). The finite element mesh was developed using bilinear (4-node)
quadrilateral elements for all cases. The bottom boundary of the model, which
coincides with the firm substratum, was fixed. The lateral boundaries were placed far
enough from the region of the shear zone to avoid any effect on the collapse
mechanism and the FE results. Roller lateral boundaries were considered in the
context of this study (Fig. 1).
Geo-Congress 2013
Geo-Congress 2013 © ASCE 2013 1314
In this study, in order for a compatible comparison with the LE results, an elasto-
plastic constitutive model with the Mohr-Coulomb failure criterion was selected to
characterize the behavior of the soil material. There are five parameters that are used
in conjunction with Mohr-Coulomb model, namely: the elastic modulus (E), the
Poisson’s ratio (ν), the friction angle (φ'), the cohesion (c'), and the dilation angle (ψ).
Trial FE analyses indicated that the elastic modulus and the Poisson’s ratio of the soil
have little effect on the value of the collapse seismic load, provided both elastic
parameters lie in a reasonable range. (e.g., Griffiths and Lane 1999, Loukidis et al
2003). It is also generally accepted that dilatancy in slope stability analyses has a
negligible effect as the problem is relatively unconfined (Zienkiewch et al. 1975,
Griffiths and Lane 1999). Therefore, in this study, the value of the elastic modulus,
Poisson’s ratio, and dilation angle were assumed to be 100 MPa, 0.3, and 0.0 degrees,
respectively (e.g., Griffiths and Lane 1999, Xu et al. 2009, and Loukidis et al. 2003).
The unit weight of the soil was assumed to be 20 kN/m3.
DEFINITION OF FAILURE
Since the introduction of the strength reduction technique by Zienkiewicz et al.
(1975), several criteria have been utilized to define the limit state of a slope.
Currently, there are four criteria which are widely utilized to analyze the stability of a
slope using the FE method, (1) non-convergence of the solution, (2) achievement of a
pre-defined equivalent shear strain or plastic strain magnitude along the potential slip
surface, (3) full extension of the plastic zone from the toe to the crest of the slope, and
(4) rapid increase in nodal displacement on the slope surface. It has been reported that
the “non-convergence of the solution” approach cannot be generalized to all cases, as
in some situations it can overestimate the factor of safety due to the dependency that
this criterion has on a numerical algorithm that is used in many FE programs (e.g., Xu
et al 2011, and Chang and Huang 2005). The second approach described above also
has some limitations, as it requires that a pre-defined strain level be established prior
to the analyses, the magnitude of which can be open to debate and judgment between
various users of this method.
The third approach, full extension of the plastic zone from the toe to the crest of a
slope, is another failure criterion which can be used to visually determine the KC
value. In this study, it was observed that the contours of maximum principal plastic
strains initiate from the toe at lower values of acceleration and fully extend at the
critical seismic acceleration, KC. Moreover, this criterion coincides with a rapid
increase in the maximum plastic shear strain (or maximum displacement). Therefore,
in this study, full-extension of the plastic zone from the toe to the crest of the slope
was adopted to define the occurrence of slope failure. The critical accelerations
Geo-Congress 2013
Geo-Congress 2013 © ASCE 2013 1315
obtained from GIM and LE, as a function of c'/γH and angle of shearing resistance φ '
for two different slope angles, 2H:1V and 1H:1V, are shown in Figure 3. For these
two example slopes, the maximum difference between the yield acceleration obtained
from the FE analyses and LE methods is approximately 18%.
Downloaded from ascelibrary.org by University of California, Davis on 02/13/14. Copyright ASCE. For personal use only; all rights reserved.
2H:1V 1H:1V
Figure 3. Comparison of critical acceleration values determined using the FE
and LE approaches
Geo-Congress 2013
Geo-Congress 2013 © ASCE 2013 1316
In the current study, it was found that the results obtained using three-node
triangular elements could also lead to results which are in a good agreement with the
LE results, a finding which contrasts the conclusions from Xu et al. (2007). More
computational effort was required for meshes comprised of three-node triangular
elements than for meshes having four-node quadrilateral elements. Historically,
square meshes have been shown to have a minimal effect on the geometry of the
failure mechanism for homogeneous simple slopes (Shukha & Baker 2003).
Therefore, in this study, four-node quadrilateral elements were adopted for the FE
slope stability analyses. The results from this study indicate that use of a mesh that is
too fine should be avoided, as fine meshes tend to underestimate the results that are
obtained using FE analysis (as compared to the LE analyses).
Geo-Congress 2013
Geo-Congress 2013 © ASCE 2013 1317
arc assumption that is made in LE analysis, the pseudostatic LE critical surface is not
necessarily identical to the line that passes through the maximum values of plastic
strain, as shown in Figs. 4 and 5. However, it is clear that the FE analysis approach
can successfully predict the critical failure zone within a soil mass that corresponds to
failure circles that have factors of safety that are close to one. Moreover, by using a
critical failure zone rather than a critical failure surface, uncertainties associated with
the true magnitude of the strength parameters and the assumed failure mechanism of
Downloaded from ascelibrary.org by University of California, Davis on 02/13/14. Copyright ASCE. For personal use only; all rights reserved.
o o
KC = 0.200g – 2H:1V, φ' = 20 , c' = 30 kPa KC = 0.310g – 2H:1V, φ' = 20 , c' = 40 kPa
Figure 4. Comparison of safety map and the contours of maximum principal
plastic strains-1V:2H
Of particular concern when using the FE approach for slope stability analysis is the
effect of the associated (where φ = ψ) or non-associated (where ψ = 0) flow rule
formulation on the soil behavior that is predicted by the Mohr-Coulomb elastic-
plastic model, a topic which has been the subject of significant controversy (e.g.,
Griffiths and Lane 1999, Manzari and Nour 2000).
o o
KC = 0.240g – 1H:1V, φ' = 45 , c' = 10 kPa KC = 0.174g – 1H:1V, φ' = 20 , c' = 30 kPa
Figure 5. Comparison of safety map and the contours of
maximum principal plastic strains-1H:1V
Some researchers have observed that the shear zone geometry is affected by
dilatancy, and shear zones in the frictional-cohesive soils with non-associated flow
rule are shallower and less curved than those observed in the associative case
(Loukidis et al. 2003, Wartman and Strenk 2006). In contrast, the failure surfaces
determined using FE analysis are almost identical to those from LE results for the
associative case, where φ = ψ (Wartman and Strenk 2006). Others have argued that,
in slope stability analysis, dilatancy has a negligible effect, as the problem is
relatively unconfined (Ziwnkiewicz et al. 1975, Griffiths and Lane 1999). Moreover,
the critical state of a frictional soil is always accompanied by continuous distortion of
Geo-Congress 2013
Geo-Congress 2013 © ASCE 2013 1318
soil element at no volume change and the assumption of normality is no longer valid.
It has been postulated that this could also lead to less conservative results compared
to those from LE analyses (Manzari and Noor 2000). Although this question is not
resolved here, the pseudo-static slope stability analyses conducted in this study
showed that a safety map failure zone of (1 to 1.03)FS includes the failure surfaces
that are determined using either the non-associated or associative flow rules. The
authors consequently believe that the FE failure zone can be practically considered
Downloaded from ascelibrary.org by University of California, Davis on 02/13/14. Copyright ASCE. For personal use only; all rights reserved.
the same as the one rendered by the critical slip surface, and which is therefore
theoretically within the range of the precision of the most rigorous LE methods.
ACKNOWLEDGEMENTS
This material is based upon work supported by the National Science Foundation
under Grant No. CMMI-0844836. This National Science Foundation grant partially
supported the first and the third authors in conjunction with their work on this project.
REFERENCES
ADAMA Engineering Inc. (2010). ReSSA (3.0): Reinforced and Unreinforced Slope
Stability analysis. Newark, DE, USA.
Geo-Congress 2013
Geo-Congress 2013 © ASCE 2013 1319