Beruflich Dokumente
Kultur Dokumente
net/publication/277238500
CITATIONS READS
20 2,419
1 author:
Ben Leshchinsky
Oregon State University
83 PUBLICATIONS 914 CITATIONS
SEE PROFILE
Some of the authors of this publication are also working on these related projects:
All content following this page was uploaded by Ben Leshchinsky on 26 May 2015.
INTRODUCTION
Limit Equilibrium (LE) analysis has been used to assess slope stability for many
decades. LE techniques consider numerous potential sliding masses - each defined by
assumed slip surfaces. Since the approach is generally statically indeterminate, statical
assumptions are made when assembling the equilibrium equations for the potential
collapsing bodies. By introducing the notion of a safety factor for the entire sliding mass,
these equilibrium equations are solved to render a system that is on the verge of failure.
This optimization process seeks to determine the potential slip surface that produces
Leshchinsky, B. and Ambauen, S. (2015). “Limit Equilibrium and Limit Analysis: Comparison of Benchmark Slope Stability
Problems.” ASCE Journal of Geotechnical and Geoenvironmental Engineering. DOI: 10.1061/(ASCE)GT.1943-5606.0001347.
the lowest safety factor: i.e., the factor of safety and its associated failure surface.
Rigorous LE methods have proven to be a reliable means for assessing slope stability.
However, these methods are based on assumed statics simplifications as well as
prescribed failure geometries, meaning that the approach violates rules of kinematic
admissibility (Michalowski, 1995). While this is not an issue for many slope stability
problems (e.g., simpler geometries, steeper slopes, homogenous slopes where log
spiral or circular arc mechanisms work well), identification of a generalized, critical
collapse mechanism while satisfying equilibrium is challenging for realistically complex
slopes. Subsequently, algorithms for objectively identifying complex geometries of slip
surfaces while using Spencer method have been developed (e.g., Baker 1980,
Celestino and Duncan, 1981).
An attractive aspect of the LE analysis in design is the use of a single global factor, i.e.,
the factor of safety (FS), to characterize the state of stability of a slope. Using the same
notion, Zienkiewicz et al. (1975) introduced the strength reduction method (SRM) into
finite element (FE) analysis. The continuum mechanics-based approach does not
require ‘assumption’ of the geometry of the shear zone and it automatically satisfies
global equilibrium, a requirement for convergence. Since 1975, variations of the SRM
were implemented in finite elements (FE) and finite difference (FD) codes resulting in
extensive comparison of continuum mechanics-based approaches (FE/FD) with LE. In
many cases, the comparison of results demonstrated agreement (Griffiths and Lane,
1999). However, in a few peculiar cases there were disagreements because of
limitations of either the FE/FD or the LE methods used (Griffiths and Lane, 1999). The
evolution of accepting FE/FD as an alternative to LE has resulted from comparative
studies between the methods.
Although FE/FD provides an alternative approach to LE, it has not wholly displaced LE
as the major methodology in design of slopes, likely due to complexity of material
parameters associated with the application of FE/FD analyses. Yet, another alternative
approach to these methods is based on Limit Analysis (LA). It is more robust than LE as
it does not require as many assumptions about formulation of static equilibrium or a
priori slip surfaces. However, implementation of LA can be simpler than FE/FD as it
does not require special constitutive laws, deal with non-intuitive convergence issues or
concern displacements. LA models soil as a material that is perfectly plastic and obeys
an associated flow rule (Yu et al., 1998), employing two different theorems to provide a
solution: upper bound or lower bound plasticity. Lower bound plasticity theorem must
satisfy equilibrium and boundary conditions while not violating the yield criterion at any
point within a given soil mass. In such a case, the soil mass is near the brink of failure
(Chen, 2008). Alternatively, upper bound plasticity (the method used in this study) seeks
to obtain an upper bound for a given collapse load in consideration of a kinematically
admissible collapse surface and an associated ratio of plastic dissipation to that of work
done by applied loads. When the rate of work along a kinematically admissible failure
surface due to external loads is greater or equal to the work done by internal stresses,
the external loading cannot exceed the actual collapse load (Chen, 2008, Yu et al.,
1998). For lower bound, the computed factor of safety is smaller or equal to the actual
theoretical value. Conversely, the computed factor of safety is greater than or equal to
Leshchinsky, B. and Ambauen, S. (2015). “Limit Equilibrium and Limit Analysis: Comparison of Benchmark Slope Stability
Problems.” ASCE Journal of Geotechnical and Geoenvironmental Engineering. DOI: 10.1061/(ASCE)GT.1943-5606.0001347.
the actual theoretical value. The lower and upper bound analyses bracket the exact
solution, a consideration that must be made when applying these approaches to slope
stability.
Slope stability problems with general shaped slip surfaces are highly indeterminate.
That is, there must be some assumptions made to solve the equilibrium equations at a
limit state in order to find the safety factor associated with a test body as defined by a
selected slip surface. All rigorous LE methods (e.g., Janbu, M-P, Spencer) make such
assumptions while satisfying the global equilibrium equations for a given body, some of
which violate kinematical admissibility. Conversely, while computational LA also
satisfies equilibrium, it follows rules in mechanics (plasticity) that are more stringent
than LE, thus requiring an explanation for the implications of a general comparison
between the two approaches as it meets these kinematic requirements. Prior work has
compared LA and LE methods as a means of highlighting scenarios that warrant
appropriate use (Yu et al. 1998), demonstrating good agreement for simplified,
homogeneous slopes, including application of Bishop’s simplified method in a
subsequent revision (Yu et al. 1999). Equivalency between LA and rigorous LE has
been demonstrated by Leshchinsky et al. (1985) in context of static analyses.
Specifically, rigorously obtained LA solutions also satisfy static equilibrium for a sliding
mass, similar to rigorous LE analyses. However, the converse is not true as LE
solutions do not necessarily satisfy all of the rigors of LA (e.g. consideration of kinematic
admissibility). That is, the numerical upper bound approach also produces a
kinematically admissible failure mechanism (in the context of plasticity) which is not
addressed in LE, demonstrating an added level of computational rigor employed by LA.
Hence, numerical solutions obtained from rigorous upper bound analyses satisfy global
equilibrium and therefore, can also be considered as a legitimate rigorous solution in the
context of LE, similar to Spencer, Janbu or Morgenstern-Price methods. With the
proliferation of computers combined with advances in numerical methods, upper bound
LA is becoming a more viable option for complex slope problems (e.g., Kim et al. 2002,
Leshchinsky 2015, Xie and Leshchinsky 2015), generally using a numerical framework
for its application.
Figure 1. Schematic of (a) LE with Dynamic Programming, (b) Spencer Method and (c)
DLO-LA.
Leshchinsky, B. and Ambauen, S. (2015). “Limit Equilibrium and Limit Analysis: Comparison of Benchmark Slope Stability
Problems.” ASCE Journal of Geotechnical and Geoenvironmental Engineering. DOI: 10.1061/(ASCE)GT.1943-5606.0001347.
The objective of this work is to compare the results of a relatively new DLO-LA
approach with the predictions of rigorous LE slope stability analyses, specifically those
using Spencer’s method, a commonly used approach in practice in the US. The LE
predictions are for benchmark cases in which the most critical, general shaped, slip
surface was determined using dynamic programming optimization methodology.
Comparison of these results then can be used to help in assessing the acceptability of
Leshchinsky, B. and Ambauen, S. (2015). “Limit Equilibrium and Limit Analysis: Comparison of Benchmark Slope Stability
Problems.” ASCE Journal of Geotechnical and Geoenvironmental Engineering. DOI: 10.1061/(ASCE)GT.1943-5606.0001347.
computational LA in the realm of slope stability engineering (Yu et al., 1998). Such an
approach has the advantage of relative ease in determining of the critical mechanism
and its associated factor of safety for complex problems, a major design objective.
Acceptance of computational DLO-LA can be facilitated by demonstrating good
agreement with current practice, a realization that it also satisfies the limiting equilibrium
equations for the sliding mass, demonstrating critical failure mechanisms without a priori
kinematics and highlighting scenarios where it may even provide more critical solutions.
Results of DLO-LA are compared to those generated by Baker (1980), Yoo (2006)
and other complex, yet common optimization scenarios. The methodology used by
Baker is detailed in his paper and its critical results are considered as a benchmark for
the complex problems analyzed with LE due to its use of a dynamic programming
algorithm. The results found by Baker range from simple to complex slopes with varying
soil strata, hydraulic conditions and structural constraints. Baker’s results are based on
a pre-determined set of 10 cross-sectional partitions along the width of the problem, and
evenly spaced horizontal partitions along the height of each section, forming
intersections that were potential locations for a given slip surface. Linear segments
between these intersections enable a search which eventually leads to the critical
results based on LE formulations (in this case, Spencer’s approach). Despite the
seemingly coarse discretization by today’s standards, Baker’s results are considered as
a critical, acceptable benchmark in the context of LE and the computational constraints
at the time of analysis. Several comparisons with the results from Baker (1980) were
chosen as a benchmark as it neglects many of the a priori assumptions about slip
surface shape and location employed by common LE slope stability analyses.
Furthermore, complex LE analyses that can account for unusual failure mechanisms
(concave-convex slip surfaces, compound slip surfaces, etc.), partial saturation and
reinforced soil are considered. These examples present an optimal means of
demonstrating the capability of DLO-LA methodology to determine the stability and
failure mode for complex slope stability problems. Although studies comparing of LE
and LA exist, none of these studies have compared or discussed DLO-LA, which
presents a novel means of establishing slope stability in scenarios where highly
complex, non-intuitive solutions are critical.
DLO-LA finds good agreement, both in FS (difference < 3% in all cases) and location of
failure surfaces with Baker’s prior analysis of a slope under various scenarios – as
shown in Fig. 5. Both results tend to find more critical FS values than the results
Leshchinsky, B. and Ambauen, S. (2015). “Limit Equilibrium and Limit Analysis: Comparison of Benchmark Slope Stability
Problems.” ASCE Journal of Geotechnical and Geoenvironmental Engineering. DOI: 10.1061/(ASCE)GT.1943-5606.0001347.
Figure 5. Simple slope analyses a) baseline case; b) underlying weak clay seam with
no pore pressure; c) pore pressure, ru=0.25; d) underlying weak clay seam, ru=0.25; e)
no clay seam, sloped water table; f) clay seam, sloped water table.
A realistic, yet difficult slope stability problem involves slopes dealing with complex
geometry of layered strata. The layout of a soil profile, as well as the interaction
between the various materials can render unconservative results if simplified analyses
using a priori failure kinematics are incorrectly used, especially possible when slip
surface constraints are defined by the designer. Furthermore, the introduction of a
layered soil profile can result in non-intuitive and sophisticated collapse mechanisms,
demanding more complex optimization algorithms to determine collapse geometry. In
order to demonstrate this concept, an earthen dam with varied soil geometry and non-
linear water table was used as an example by Baker in 1980, based on actual
construction specifications (see Fig. 6).
Leshchinsky, B. and Ambauen, S. (2015). “Limit Equilibrium and Limit Analysis: Comparison of Benchmark Slope Stability
Problems.” ASCE Journal of Geotechnical and Geoenvironmental Engineering. DOI: 10.1061/(ASCE)GT.1943-5606.0001347.
An analysis using LA (2000 nodes, approximate nodal density of 0.7 meters) finds
satisfactory agreement in FS (≈2% discrepancy) with Baker’s prior results as shown in
Fig. 7. The failure surfaces are similar, with a minor deviation as the DLO-LA results
demonstrate a failure surface that exits slightly above the toe. The deviations of the slip
surfaces can be attributed to coarse discretization used by Baker, limited by
computational constraints. While the FS values are quite close, the slip surfaces are not
identical. For this scenario, FS is not very sensitive to the location of the slip surface,
agreeing with a well-known observation in slope stability analyses.
Upon comparison of DLO-LA (2000 nodes, approximate nodal density of 0.5 meters)
and Spencer’s analysis with dynamic programming, satisfactory agreement is
demonstrated – Fig. 9. The LA method applied to this model yields a slightly smaller FS
than the results found by Baker (1980): 0.87 and 0.95, respectively. It is noted that
because of computational constraints, Baker’s discretization in using the dynamic
programming was rather coarse; it is entirely possible that by using a finer vertical
sections, Baker’s critical results would have been closer to the LA results. However,
this still demonstrates the benefit of stability approaches that employ fewer assumptions
about failure location and rely less on designer intuition for complex geotechnical
structures.
Leshchinsky, B. and Ambauen, S. (2015). “Limit Equilibrium and Limit Analysis: Comparison of Benchmark Slope Stability
Problems.” ASCE Journal of Geotechnical and Geoenvironmental Engineering. DOI: 10.1061/(ASCE)GT.1943-5606.0001347.
Through their site investigation, sample collection, and subsequent laboratory testing,
Yoo and Jung (2006) obtained the backfill material properties (Table 1) and
Leshchinsky, B. and Ambauen, S. (2015). “Limit Equilibrium and Limit Analysis: Comparison of Benchmark Slope Stability
Problems.” ASCE Journal of Geotechnical and Geoenvironmental Engineering. DOI: 10.1061/(ASCE)GT.1943-5606.0001347.
reinforcement total long-term design strength parameters (Tltds=60 kN/m). These values
were used as inputs for the slope stability analyses conducted in this example. The
reinforced and retained soil types were decomposed granite, while the foundation soil
was a slightly weathered granite rock layer.
Yoo and Jung (2006) acquired pore-water pressure conditions using a transient
seepage model conducted using SEEP (Geostudio, 2004), involving the initial
application of a prescribed unit flux so as to attain realistic negative pore pressures for
the decomposed granite soil. Once the initial conditions were set, the transient seepage
analysis was modelled for a total time span of 108 days, which was the time
documented from the
end of construction
until failure. The pore-
water pressure
conditions for each
time step were used in
calculating the slope
stability with
Morgenstern-Price LE
analysis (applied with
a half-sine interslice
resultant direction
function, 151 surfaces
analyzed, 30 slices).
Three distinctive
scenarios were
selected in this work
to be modelled for
Figure 10. Comparison of failure mechanisms and
comparison using
associated safety factors for MSE wall during varying stages
Limit Analysis (2000
of a transient seepage analysis.
nodes, approximate
nodal density of 0.5 meters): the initial conditions of the transient seepage analysis with
suction present in the unsaturated soil (Fig. 10a), the least stable time step of the
transient analysis where the infiltration saturated the soil (Fig. 10b), and the dry state
with no pore-water pressure included in the analysis (Fig. 10c). The pore-water
pressure conditions for the second and third scenarios were input into DLO-LA analysis
by incorporating suction stresses into the soil’s material strength properties. The suction
stresses were calculated using the seepage model’s pore-water pressures averaged
over the reinforced and retained soil areas of the wall, which provided a reasonable
approximation of its contribution to shear strength in the soil. The results for all three
cases demonstrate close agreement between both FS and collapse mechanisms for
both LE and DLO-LA. The LA slip surfaces tend to avoid penetration of the
reinforcement in these scenarios, demonstrating a slightly more critical failure
mechanism than that of LE, a useful characteristic in the analysis of complex reinforced
earth structures. This example demonstrates that Limit analysis using DLO can provide
Leshchinsky, B. and Ambauen, S. (2015). “Limit Equilibrium and Limit Analysis: Comparison of Benchmark Slope Stability
Problems.” ASCE Journal of Geotechnical and Geoenvironmental Engineering. DOI: 10.1061/(ASCE)GT.1943-5606.0001347.
similar results to that of rigorous LE analyses, while using fewer constraints regarding
slip surface geometry.
Stability analyses that require fewer assumptions about slip surface kinematics present
a benefit in the presence of weak interfacial materials layered between stronger
materials, as can often be the case in slopes made of naturally deposited, residual soils.
The presence of a weak interface, e.g. a clay seam, can result in a collapse mechanism
that is easier to determine with generalized DLO-LA stability methods that do not
adhere to geometrical constraints, such as assumed slip surface geometry.
To demonstrate the
applicability of stability
tools that can
adequately capture
complexities
associated with weak
interfacial materials, a
hypothetical example
was analyzed. That is,
the presence of a
weathered soil
Figure 11. Comparison of failure mechanisms for weathered
(ϕ’=33°, c’=0 kPa,
soil overlying a clay seam and bedrock.
γ=21 kN/m3) overlying
a thin seam of weak clay (ϕ’=0°, c’=12 kPa, γ=19 kN/m3) and a layer of bedrock,
modeled as a rigid material due to the high shear strength of rock in comparison to the
other respective materials (see Fig. 11). Set back from the crest of the slope is a
surcharge of 200 kPa (width of 8 meters), demonstrative of the mat foundation
supporting a structure atop a hill. Using Spencer’s method (151 surfaces analyzed, 30
slices), the critical collapse mechanism is found to be a concave-convex failure surface
that straddles the weak clay interface in between the rock and weathered soil, yielding a
critical FS of 0.85, an unstable slope (see Fig. 11). This critical result was determined
after performing a variety of LE analyses, each with an array of potential entrance and
exit points for analyzed slip surfaces. Use of LA (2000 nodes, approximate nodal
density of 0.5 meters) yields a similar surface and safety factor at 0.84, demonstrating
agreeable results without the challenge of performing a variety of stability analyses to
determine the critical failure mechanism (Spencer, Morgenstern-Price, etc.).
Furthermore, the number of analytical parameters required by these approaches (e.g.
non-intuitive choices of failure restrictions such as entrance and exit points, failure radii,
number of slices, or statical assumptions, such as interslice resultant force inclination
function, etc.) are reduced. Use of DLO-LA, however, requires correct definition of
boundary conditions and nodal density that can adequately discretize a failure surface,
a general reduction in input considerations required for the same problem and similar
results.
Leshchinsky, B. and Ambauen, S. (2015). “Limit Equilibrium and Limit Analysis: Comparison of Benchmark Slope Stability
Problems.” ASCE Journal of Geotechnical and Geoenvironmental Engineering. DOI: 10.1061/(ASCE)GT.1943-5606.0001347.
Another common
situation where
concave-convex
failure mechanisms
may be critical is
when an embankment
is built on top of a
weak clay layer
placed between two
competent layers of
Figure 12. Comparison of failure mechanisms for
soil or rock. Again,
embankment overlying layers of sand and weak clay.
this type of interfacial
problem highlights
discontinuities in strata, emphasized by the presence of weak materials adjacent to
strong materials: a perfect scenario for a generalized analysis unencumbered by
assumptions of slip surface shape or location constraints. In this setting, a choice of LE
approach and assumptions about potential entrance and exit points may not be obvious,
underlining the benefits of an analysis that forgoes assumptions of failure surface
location and kinematics. This situation is shown with an example where an 8-meter tall
sand embankment (ϕ’=30°, c’=0 kPa, γ=20 kN/m3) is rapidly constructed (undrained
conditions) upon a strong sand layer (ϕ’=45°, c’=0 kPa, γ=20 kN/m3), a large weak clay
layer (ϕ’=0°, c’=10 kPa, γ=19 kN/m3) and bedrock (rigid), respectively (see Fig. 12). In
this scenario, the weak clay layer becomes the least resistive path for propagation of
the failure surface until it again penetrates through the top sand layer, where it forms a
passive, failure wedge, or “panhandles”. This scenario was documented by Wright
(2013) as occurring in the presence of varied and complex layers of weak and strong
soils underlying a constructed embankment. In the presented example, the critical
failure mechanism is finally found using a Spencer’s method (151 surfaces analyzed, 30
slices), yielding a nuanced, “panhandling” concave-convex failure mechanism that
partially penetrates the weak clay layer, yielding a factor of safety of 1.00 (see Fig. 12).
A comparison to LA using DLO (2000 nodes, approximate nodal density of 0.5 meters)
demonstrates a slightly more critical failure surface (FS=0.95) with a more defined
collapse mechanism that alternates between concave and convex geometries, showing
the “panhandling” mechanism. The stability results between LA and LE agree well in
terms of FS, but use of DLO forgoes some of the required assumptions of slip surface
location.
CONCLUSIONS
Conventional but rigorous LE stability analyses like Spencer and Morgenstern-Price are
widely accepted as mainstream and satisfactory techniques of determining the stability
of a variety of soil structures, including slopes. This is satisfactory as its ease of
application, appropriate use and history of successful design for homogenous slopes,
steep slopes and simple slopes. However, alternative methods, such as numerical
upper-bound plasticity analysis (i.e. numerical Limit Analysis), can provide equally
Leshchinsky, B. and Ambauen, S. (2015). “Limit Equilibrium and Limit Analysis: Comparison of Benchmark Slope Stability
Problems.” ASCE Journal of Geotechnical and Geoenvironmental Engineering. DOI: 10.1061/(ASCE)GT.1943-5606.0001347.
satisfactory results, especially when employing the DLO technique, which enables
determination of complex and comprehensive (sliding, overturning, global, composite)
collapse mechanisms for non-trivial geotechnical problems without a priori assumptions
about kinematics. That is, the advantage is enhanced with the ability to forgo the non-
intuitive input of failure geometries and slip surface constraints and no need for
assumption of a priori failure mechanisms or interaction properties between slices or
rigid bodies.
LE and DLO-LA both present adequate means of evaluating slope stability for various
scenarios. LE is appealing as it involves simple and well-known computations, but it
requires assumptions from the engineer (kinematics, orientation of interslice forces,
etc.), while DLO-LA is more computationally expensive, but requires fewer assumptions
about the location of collapse. With the progression of powerful, yet affordable computer
processing, engineers are at the point where more computationally intensive analyses
like DLO-LA may be preferable to alternative analyses that require more user input and
potentially erroneous assumptions, especially for complex, yet realistic geotechnical
problems.
Leshchinsky, B. and Ambauen, S. (2015). “Limit Equilibrium and Limit Analysis: Comparison of Benchmark Slope Stability
Problems.” ASCE Journal of Geotechnical and Geoenvironmental Engineering. DOI: 10.1061/(ASCE)GT.1943-5606.0001347.
4. REFERENCES
Wright, S. " 2013 H. Bolton Seed Lecture: Slope Stability Computations." GeoCongress
2013. San Diego, CA. 5 March 2012. Plenary Lecture.
Yoo, C. and Jung, H. (2006). ”Case History of Geosynthetic Reinforced Segmental
Retaining Wall Failure.” J. Geotech. Geoenviron. Eng., 132(12), 1538–1548.
Yu, H.S., R. Salgado, S.W. Sloan and J.M. Kim. (1998) ‘Limit Analysis versus Limit
Equilibrium for Slope Stability.’ ASCE Journal of Geotechnical and
Geoenvironmental Engineering, ASCE, 125(10), 915-918.
Yu, H. S., Salgado, R., Sloan, S. W., & Kim, J. M. (1999). Closure: Limit analysis versus
limit equilibrium for slope stability. Journal of Geotechnical and
Geoenvironmental Engineering, 125(10), 914-918.
Vahedifard, F., Leshchinsky, B. A., Sehat, S., & Leshchinsky, D. (2014). Impact of
Cohesion on Seismic Design of Geosynthetic-Reinforced Earth Structures.
Journal of Geotechnical and Geoenvironmental Engineering, 140(6).
Xie, Y., & Leshchinsky, B. (2015). MSE walls as bridge abutments: Optimal
reinforcement density. Geotextiles and Geomembranes, 43(2), 128-138.
Zienkiewicz, Humpheson, and Lewis, (1975), Associated and non-associated visco-
plasticity and plasticity in soil mechanics. Geotechnique. 25(4), 671-689.