Download as pdf or txt
Download as pdf or txt
You are on page 1of 14

RAPID DRAWDOWN ANALYSIS –

WHAT IS AN ANALYST TO DO?


By John W. France, PE (URS Corporation, Denver),
and Christina J.C. Winckler, (URS Corporation, Denver)

Introduction
In the authors’ experience rapid drawdown analysis of embankment
dams is one of the most misunderstood slope stability analysis cases. As a
result, inappropriate analysis methods have often been used. This paper
begins with a review of the fundamental soil behavior during rapid drawdown.
This is followed by a discussion of analysis methods that have been used,
both appropriate and inappropriate methods. Finally, an example dam is used
to illustrate the differences in results obtained with different methods and some
of the authors’ conclusions are provided.

Fundamental Soil Behavior During Rapid Drawdown


Rapid drawdown of an embankment dam occurs when the impounded
reservoir is lowered rapidly enough that pore water pressures in some or all of
the embankment or foundation soils cannot drain fast enough to
simultaneously reach steady state equilibrium with the new, lower reservoir
water level. Instability of the upstream slope can occur in these
circumstances, and, in fact, documented rapid drawdown failures have
occurred. The susceptibility of an embankment to rapid drawdown instability is
dependent on the rate at which the reservoir elevation is lowered and the
hydraulic conductivity (permeability) of the embankment soils. If the hydraulic
conductivity of any of the soils in the embankment or foundation is low enough
relative to the rate of reservoir drawdown, then the pore water pressures in
those soils will not equilibrate with the new loadings of the lower reservoir. In
these circumstances, the soil will be subjected to undrained loading conditions
and the undrained strength of the soil will control stability. If the undrained
strength of the soil is lower than the drained strength, then instability can result
from rapid drawdown of the reservoir.
Duncan et al, 1992 suggest that the dimensionless time factor can be
used to evaluate whether a soil should be considered undrained during rapid

1
drawdown. The dimensionless time factor, T, is defined by the following
equation:
T = cv t / D2 (Equation 1)
where cv = coefficient of consolidation of the soil, t = time for drawdown, and D
= the length of the drainage path for pore pressure in the soil to dissipate. If T
is greater than 3.0, the dissipation of excess pore pressure during drawdown
will be greater than or equal to 90 percent, and the soil can be assumed to be
fully drained. Otherwise, the soil should be assumed to be subject to
undrained loading. Values of cv can be obtained from laboratory test data, if
available, or approximate values can be estimated from the following Table 1,
which is reproduced from Duncan et al, 1992.
Table 1 – Approximate Values of cv for Various Soil Types
(reproduced from Duncan et al, 1992)
Soil Type Values of cv (ft2/day)
Coarse Sand > 10,000
Fine Sand 100 to 10,000
Silty Sand 10 to 1,000
Silt 0.5 to 100
Compacted Clay 0.05 to 5
Soft Clay < 0.02

In geotechnical practice, stability is typically evaluated using limit


equilibrium analyses methods, most typically methods-of-slices approaches
considering circular or non-circular potential sliding surfaces, as illustrated in
Figure 1. Deformation analyses methods, such as finite element and finite
difference methods are today being more frequently used for stability analysis,
but the use of limit equilibrium methods still dominates practice. The
discussion of rapid drawdown analysis in this paper focuses on limit
equilibrium methods, but the principles apply equally to deformation analysis
methods.
The forces acting on a single slice in the limit equilibrium method are
illustrated in Figure 2. The shear resistance force mobilized at the base of the
slice is defined as:
[ci’ + ( i’ tan ’)]/F (Equation 2)
where ci’ = effective stress cohesion of the soil, ’ = effective stress friction
angle of the soil, i’ = the effective normal stress on the base of the slice,
which equals ( i - ui), where i = the total normal stress on the base of the
slice and ui = the pore water pressure at the base of the slice, and F = the
factor of safety. The numerator in Equation 2 is the Mohr-Coulomb effective

2
stress equation for the shear strength of soil that is familiar to all geotechnical
engineers.

Figure 1 – Limit Equilibrium Stability Analysis: Method of Slices

Figure 2 – Forces on Individual Slices

The application of Equation 2 in stability analysis for steady state


seepage conditions is relatively straight forward. The value of i is a function

3
of the unit weights of the overlying soils and geometry. Values of ci and ’ can
be obtained from field or laboratory tests or estimated from empirical
correlations. The pore water pressure, ui, can be taken from seepage
analyses or estimated from the depth of the base of the slice below the
phreatic surface.
The application of Equation 2 to rapid drawdown analysis is significantly
more difficult. Values of i, ci, and ’, can be estimated for rapid drawdown
analysis as easily as they can be estimated for steady state seepage analysis.
However, it is very difficult to correctly estimate the pore water pressure, ui, for
the rapid drawdown case. The pore pressures during rapid drawdown
conditions are a function of both the changing seepage pattern caused by the
drawdown and the response of the soil to changes in normal and shear
stresses caused by the drawdown. The pore water pressure changes
resulting from normal and shear stress changes are not always properly
recognized, leading to inappropriate analyses. With respect to estimating pore
water pressures during rapid drawdown, Terzaghi and Peck, 1967, provided
the following advice:
“… in order to determine the pore pressure conditions for the
drawdown state, all of the following factors need to be known: the
location of boundaries between materials with significantly different
properties; the permeability and consolidation characteristics of
each of these materials, and the anticipated maximum rate of
drawdown. In addition, the pore pressures induced by the
changes in the shearing stresses themselves … need to be taken
into consideration. In engineering practice, few of these factors
can be reliably determined. The gaps in the available information
must be filled by the most unfavorable assumptions compatible
with the known facts.”
Given the difficulty in estimating pore water pressure responses during
undrained shear caused by rapid drawdown, the authors agree with Duncan et
al, 1992, that the most appropriate analysis methods for this case are those
based on total stress approaches that estimate the undrained shear strengths
of the soils in question. The appropriate undrained shear strengths are a
function of the gradations, densities, void ratios, and stress histories of the
soils as they exist before drawdown. Although it is suggested that undrained,
total stress analyses should be used for rapid drawdown, this does not mean
that the Mohr-Coulomb strength equation (the numerator in Equation 2) does
not apply. Indeed, the strength is still correctly defined by this equation. But in
this instance, it is judged that the strength can be more correctly predicted
than can the pore water pressure. Application of the total stress approach is
simply saying that, at a failure state the pore pressures that would develop in
response to the stress changes caused by rapid drawdown would be those
that would produce the undrained strength consistent with the Mohr-Coulomb
equation.

4
Discussion of Rapid Drawdown Analysis Methods
With the foregoing background discussion of the fundamentals of soil
behavior during rapid drawdown and their relationship to stability analyses,
several methods that the authors have seen employed will be discussed.
Lowered Phreatic Surface and Drained Strengths – The authors
have seen rapid drawdown analyses completed using drained strength
parameters for all materials and an estimated phreatic surface for the drawn
down reservoir condition. For homogeneous embankments of materials that
are not expected to drain during reservoir lowering, the phreatic surface is
often estimated to extend down the upstream face of the embankment from
the original reservoir level to the lowered reservoir level. For zoned
embankments, the phreatic surface between the two reservoir levels in the
upstream section of the dam is often estimated to be along the outer boundary
between undrained and drained materials, for example at the boundary
between and upstream shell and the core. The phreatic surface in the
downstream section of the dam is often assumed to be the same as existed
before drawdown.
This method of analysis is fundamentally incorrect, because it does not
consider the pore water pressures that occur in response to changes in shear
stresses caused by drawdown. Depending on the particular soils involved and
the geometry of the embankment, this method of analysis could either over-
estimate or under-estimate the undrained shear strength available on potential
shear surfaces.
Pore Water Pressures From Transient Seepage Analysis and
Drained Strengths – In this method, drained strengths are used for all
materials and the distribution of pore water pressures in the embankment and
foundation is imported from a transient seepage analysis.
Again, as for the previous method, this method of analysis is
fundamentally incorrect, because it does not consider the pore water
pressures that occur in response to changes in shear stresses caused by
drawdown.
Estimated Values of Undrained Strengths – In this method of
analysis specific values of undrained strengths are estimated for the various
materials that are judged to respond to drawdown in an undrained manner.
Drained strengths are used for materials that are judged to drain during
drawdown. Typically, a single value of undrained strength is assigned to a
single soil material, for example a single value of undrained strength would be
assigned to the embankment core. The selected undrained strengths could be
based on laboratory tests or on empirical relationships.
While this method is more fundamentally correct than either of the
previous two methods, its accuracy is questionable because of the
oversimplification of the strength characterization. The undrained strength
within any particular material would not be expected to be constant, but rather

5
it would be expected to vary with stress history and effective confining stress
before drawdown. This method could produce a conservative estimate of the
factor of safety, if a conservatively low undrained strength is selected. This
could be acceptable if the conservative factor of safety meets minimum factor
of safety requirements, however, the analyst must be confident that the
selected strength is conservatively low. The accuracy of this method could be
improved by subdividing single materials, for example the core, into smaller
zones with individual strengths for each zone selected to reflect the expected
variations in undrained strength, however, this is not a particularly efficient way
to complete stability analysis.
Another significant shortcoming of this method is that at low effective
confining stresses it uses undrained strengths that are higher than drained
strengths. It is typically recommended that the undrained strength not be used
in stability analysis if it is greater than the drained strength. The reason for
this recommendation is that development of the higher undrained strength
requires negative pore water pressures. In reality, these negative pore
pressures may not be able to develop because of cavitation or drainage away
from a relatively thin failure plane. Consequently it prudent to not rely on the
negative pore pressures, and stability analysis methods that include undrained
strengths higher than drained strengths for some slices could be
unconservative.
Undrained Strengths Varying With Effective Confining Stress – In
this approach, the undrained strengths of materials that are not expected to
drain during drawdown are defined as functions of the effective confining
stress of the material. Again, materials that are expected to drain are
characterized by drained strengths.
A number of different approaches have been used to characterize the
undrained strength as a function of effective confining stress. These methods
include ratios of undrained strength to effective confining stress (s/ i’ or c/p’)
and strength envelopes defined in terms of plots of shear strength versus
effective confining stress.
These methods meet the fundamental requirement of defining the
strengths in the materials of concern as undrained strengths, and they
potentially are an improvement over the previous method because they
account for variation of strength with confining stress. However, these
methods also have some fundamental pitfalls.
The effective stress used to define the undrained shear strength vs.
effective confining stress relationship may not be fully compatible with the
effective stress calculated in the stability analysis. For example, if isotropically
consolidated triaxial tests are used to develop an undrained shear strength vs.
effective confining stress envelope, the isotropic consolidation stress is often
used for the effective confining stress. This stress condition is not entirely
compatible with the anisotropic consolidation stresses that exist in the field on
the potential failure plane.

6
A potentially more serious problem occurs if the undrained strength
characterizations are used in single stage stability analysis for the conditions
that exist after drawdown. The undrained strengths of the materials that do
not drain during drawdown are correctly related to the effective confining
stresses that exist before drawdown, because the response of the soil to the
loading is undrained. If a lowered phreatic surface after drawdown is used in
the analysis, the undrained strengths will be calculated using effective
stresses that are higher than existed before drawdown. The resulting
undrained strengths will be overestimated and the resulting calculated factors
of safety will be unconservative. This problem can be addressed using multi-
stage analyses, as described further for the next three methods discussed.
Depending on the specific undrained strength characterization,
methods using relationships between undrained strength and effective stress
can have the same problem for slices at low effective confining stresses
discussed above, because of implicit reliance on negative pore water
pressures.
U.S. Army Corps of Engineers (USACE) 1970 Method – This method
of analysis was the recommended procedure in the 1970 version of the U.S.
Army Corps of Engineers (USACE) slope stability manual EM 1110-2-1902
(USACE, 1970). In this paper the method will be designated the Corps of
Engineers 1970 method. In the 2003 version of EM 1110-2-1902 (USCAE,
2003), the USACE recommends use of the Duncan, Wright, and Wong
method, described below, in place of the Corps of Engineers 1970 method.
The Corps of Engineers 1970 method is a two-stage analysis method.
In the first stage, the consolidation stresses along the potential slip surface are
calculated for the conditions that exist before drawdown. In the second stage,
a stability factor of safety is calculated using the consolidation stresses from
the first stage and a bilinear shear strength envelop representing undrained
strengths versus consolidation stresses. The bilinear strength envelop is
composed of the undrained strength at higher stresses and the drained
strength at lower stresses, as illustrated in Figure 3. As shown in Figure 3, the
strength envelop defined in this manner represents the lower of the undrained
strength and the drained strength at each consolidation stress. The undrained
strength envelop is typically developed from isotropically consolidated,
undrained (IC-U) triaxial tests.
This method of analysis is based on estimated undrained strengths,
and it does not rely on negative pore pressures for the lower confining
stresses, for which the undrained strength is greater than the drained strength.
Although both of these attributes are good, the method has two significant
short comings. First, the consolidation stresses used to calculate the drained
strength are the consolidation stresses before drawdown which are less than
those after drawdown. This produces stability results that are conservative
and sometimes excessively conservative. Second, the undrained strengths
determined from the isotropically consolidated triaxial tests relate to all around

7
consolidation stresses and not to consolidation stresses on the failure plane,
and they do not account for the effects of anisotropic consolidation.

Figure 3 – Shear Strengths in the Corps of Engineer 1970 Method

Lowe and Karafiath Method – This method of analysis is also a two


stage method of analysis (Lowe and Karafiath, 1960a). As in the Corps of
Engineers 1970 method, the purpose of the first stage of the analysis is to
calculate effective stresses before drawdown. However, in the Lowe and
Karafiath method both the effective consolidation stresses on the potential slip
surface and the principal effective stress ratios (Kc = 1’/ 3’) are calculated.
Lowe and Karafiath (Lowe and Karafiath, 1960b) showed that, when the
undrained strength is plotted as shear stress on the failure plane at failure ( ff)
versus effective stress on the failure plane during consolidation ( ’fc), the
relationship between ff and ’fc is a function of Kc, as shown in Figure 4. In
the second stage of the analysis, the consolidation stresses from the first
stage ( ’fc and Kc) and the undrained strength relationships shown in Figure 4
are used to calculate a stability factor of safety.
This method correctly accounts for the relationships between undrained
strength on the failure plane and consolidation stress on the failure plane and
between undrained strength and anisotropic consolidation stresses. However,
the Lowe and Karafiath method uses undrained strengths for the full range of
consolidation stresses, and thus relies on negative pore pressures for those

8
slices where undrained strengths are greater than drained strengths. In some
cases this can produce unconservative results.

Figure 4 – Shear Strengths in the Lowe and Karafiath Method

As originally proposed by Lowe and Karafiath, the undrained shear


strength relationships as a function of anisotropic consolidation stresses
shown in Figure 4 were to be determined from series of isotropically and
anisotropically consolidated triaxial tests. Anisotropically consolidated triaxial
tests can be very difficult, time consuming, and costly. Wong et al, 1983,
found that the anisotropic undrained strength relationships can be reasonably
approximated from the results of isoptropically consolidated udrained triaxial
tests with pore pressure measurements. The Kc = 1.00 envelope shown in
Figure 4 is simply the ff and ’fc relationship taken directly from the undrained
triaxial test results. The Kc = Kf envelope can be developed from the effective
stresses at failure plotted in terms of ff versus ’ff, where ’ff is the effective
normal stress on the failure plane at failure.
Duncan, Wright, and Wong Method – This method is a three stage
method of analysis. The first two stages are the same as those in the Lowe
and Karafiath method, with the Wong et al, 1983, approach used to estimate
the variation of undrained strength relationships with anisotropic consolidation
stresses. The Duncan, Wright, and Wong method includes a third stage of
analysis in which the drained strength is used for all slices where the
undrained strength in the second stage is greater than the drained strength
based on the effective normal stresses on the failure plane after drawdown.

9
This method of analysis includes all of the positive attributes of the
Lowe and Karafiath method and it corrects for the one significant drawback of
that method. In addition, use of the Wong et al, 1983, approach to considering
the effects of anisotropic consolidation stresses avoids the additional difficulty
and cost of anisotropically consolidated triaxial tests.
In the authors’ opinion, the Duncan, Wright and Wong method is the
best available method of analysis for rapid drawdown conditions.
Unfortunately, use of this method can be computationally difficult for many of
the commonly used commercial limit equilibrium stability analysis computer
programs. Some commercially available programs do include this method.
For example the UTEXAS (Wright, 1999) and SLOPE/W (GEO-SLOPE
International, 2008) stability analysis computer programs includes
implementations of the Duncan, Wright, and Wong three stage analysis
method.

Analysis Results For An Example Dam


Several of the analysis methods discussed above were applied to an
example dam to illustrate the potential range of results. The example dam is
illustrated in Figure 5 and consists of a medium size, homogeneous dam of
sandy clay with sandy clay foundations. The strength properties for the
embankment and foundation soils are summarized in the following table.
Drained Strength Parameters c’ = 0, ’ = 29 degrees
Undrained Strength Parameters used
c = 470, = 15 degrees
in Corps of Engineers 1970 Method
Undrained Strength Parameters used
c = 500 psf, = 16 degrees
in Duncan, Wright, and Wong
Average Undrained Strength 1,000 psf

Figure 5 – Example Dam Cross Section

The example dam was analyzed using the following methods:


1. The lowered phreatic surface and drained strength method.

10
2. The estimated undrained strength method.
3. The Corps of Engineers 1970 method.
4. The Duncan, Wright, and Wong method.
The results of the analyses for these methods are illustrated in Figures 6
through 9 and summarized in the following table.
Calculated Minimum
Analysis Method
Factor of Safety
Lowered Phreatic Surface and Drained
1.55
Strength
Estimated Undrained Strength 2.04
Corps of Engineers 1970 1.47
Duncan, Wright, and Wong 1.89

Factor of Safety = 1.55

Figure 6 – Stability Analysis Results: Lowered Phreatic


Surface Drained Strength Method

Factor of Safety = 2.04

Figure 7 – Stability Analysis Results: Estimated


Undrained Strength Method

Beginning with the premise that the Duncan, Wright, and Wong three-
stage analysis produces the most accurate results for rapid drawdown, the
following observations can be made with respect to this case. The Corps of

11
Engineers 1970 method produces a significantly conservative result; a factor
of safety of 1.47 compared with 1.89 from the Duncan, Wright, and Wong
method. In this instance the lowered phreatic surface and drained strength
method produces a result that is similar to that from the Corps of Engineers
1970 method, and the estimated undrained strength method produces an
unconservative factor of safety of 2.04.

Factor of Safety = 1.47

Figure 8 – Stability Analysis Results: Corps


Of Engineers 1970 Method

Factor of Safety = 1.89

Figure 9 – Stability Analysis Results: Duncan,


Wright, and Wong Method

The observation concerning the Corps of Engineers 1970 method is


general, in that this method will always produce results that are conservative
with respect to the Duncan, Wright, and Wong method, although the degree of
conservatism will be different for different specific cases. The observations
concerning the other two methods are not general, but rather could be very
different for other cases. Whether the results for the lowered phreatic surface
and drained strength case will be conservative or unconservative depends on
whether the average drained strength along the failure surface is greater than
or less than the undrained strengths. This in turn will depend on the geometry

12
of the dam and the strength parameters. As discussed previously, this method
of analysis is not fundamentally consistent with the true strength
characteristics during rapid drawdown and is not recommended. Whether the
results for the estimated undrained strength method will be conservative or
unconservative depends on whether the estimated undrained strength is
greater than or less than the weighted average undrained strength along the
failure surface. In this case the estimated undrained strength was apparently
greater than the weighted average undrained strength, producing
unconservative results. This points out the need to be sure that the estimated
undrained strength used in this method is prudently conservative. Back
checking the estimated undrained strength against a weighted average
undrained strength along the critical failure surface calculated by hand is one
possible way to evaluate the conservatism of the estimated undrained
strength.

Conclusions
For soils that will behave undrained during rapid drawdown, the
strengths available to resist instability are a function of the total stresses in the
embankment and foundation soils and the pore water pressure responses in
the soils due to the changes in principal stresses, shear stesses, and seepage
patterns. Although it is relatively simple to calculate the total stresses in the
embankment and foundation soils, it is very difficult to correctly calculate the
pore water pressure responses, especially the pore water pressure responses
to changes in shear stress. It is the authors’ opinion that for this situation, it is
easier and more accurate to attempt to predict the undrained strength that can
be mobilized rather than to attempt to predict pore water pressures. The use
of undrained strength for the analysis does not ignore the all important pore
water pressure responses. Rather, the use of undrained strength implicitly
predicts the pore water pressure response by assuming that the pore water
pressures will be whatever they need to be to produce the predicted undrained
strength for the calculated magnitudes of total stresses. Methods of analysis
that do not consider the undrained strength response of the soils that cannot
drain during the period of drawdown are fundamentally inconsistent with the
conditions being analyzed and should not be used.

In the authors’ opinion, the Duncan, Wong, and Wright three-stage


method of analysis is the most correct method currently available for rapid
drawdown analysis. This method appropriately includes the fundamental
undrained strength behavior of the soil and consideration of stress conditions
on the failure plane. It also prudently uses drained strengths for analysis
slices where the undrained strength is greater than the drained strength, and,
hence, does not rely on the development of negative pore water pressures for
stability. The authors recognize that it is computationally difficult to implement
the Duncan, Wong, and Wright method in many commercially available
stability analysis programs. Some of the other, simpler undrained strength

13
analysis methods can be used, but care must be taken to insure that the
undrained strengths are not unconservatively over-estimated.

References
Duncan, J.M., S.G. Wright, and K.S. Wong, 1992, “Slope Stability During
Rapid Drawdown,” Proceedings of the H. Bolton Seed Memorial Symposium,
Volume 2, No. 4, p. 253-272, B-Tech Publishers, Vancouver, B.C.
GEO-SLOPE International Ltd., 2008, Stability Modeling With SLOPE/W 2007
Version, An Engineering Methodology, Third Edition, March 2008.
Itasca Consulting Group, 1999, FLAC Version 4.0 – Fast Lagrangian Analysis
of Continua, User’s Guide.
Lowe, J. III and Karafiath, L, 1960a, “Stability of Earth Dams Upon
Drawdown,” Proceedings 1st PanAm Conference on Soil Mechanics and
Foundation Engineering, Mexico City, Volume 2.
Lowe, J. III and Karafiath, L, 1960b, “Effect of Anisotropic Consolidation on the
Shear Strength of Compacted Clays,” Research Conference on the Shear
Strength of Cohesive Soils, ASCE, Boulder, Colorado.
U.S. Army Corps of Engineers (USACE), 1970, “Stability of Earth and Rockfill
Dams,” Engineering Manual 1110-2-1902, U.S. Army Engineer Waterways
Experiment Station, Vicksburg, MS
U.S. Army Corps of Engineers (USACE), 2003, “Slope Stability,” Engineering
Manual 1110-2-1902, U.S. Army Corps of Engineers, Washington, DC,
October 3, 2003.
Wong, K.S., Duncan, J.M., and Seed, H.B., 1983, “Comparison of Methods of
Rapid Drawdown Stability Analysis,” Report No. UCB/GT/82-05, Department
of Civil Engineering, University of California, Berkeley; Berkeley, CA,
December 1982, revised July 1983.
Wright, S., 1999, “UTEXAS4 – A Computer Program for Slope Stability
Calculations,” Software Manual, Shinoak Software, Austin, Texas.

14

You might also like