Modernization and Optimization

of Existing Dams and Reservoirs

27th Annual USSD Conference
Philadelphia, Pennsylvania, March 5-9, 2007

Hosted by
U.S. Army Corps of Engineers

On the Cover
The Corps of Engineers Beltzville Lake in East-Central Pennsylvania. In this south-facing photo, from the bottom
to the top, features include the project office, the emergency spillway, the 4,560-foot long embankment, the intake
tower, and a series of ridges of the Appalachian front. Beltzville Lake is on Pohopoco Creek, which drains into the
Lehigh River. The Lehigh Rivers water gap through Blue Mountain can be seen in the background of the photo.
(Photo by Anthony S. Bley.)

U.S. Society on Dams

Vision
To be the nation's leading organization of professionals dedicated to advancing the role of dams
for the benefit of society.
Mission USSD is dedicated to:
Advancing the knowledge of dam engineering, construction, planning, operation,
performance, rehabilitation, decommissioning, maintenance, security and safety;
Fostering dam technology for socially, environmentally and financially sustainable water
resources systems;
Providing public awareness of the role of dams in the management of the nation's water
resources;
Enhancing practices to meet current and future challenges on dams; and
Representing the United States as an active member of the International Commission on
Large Dams (ICOLD).

The information contained in this report regarding commercial projects or firms may not be used for
advertising or promotional purposes and may not be construed as an endorsement of any product or
from by the United States Society on Dams. USSD accepts no responsibility for the statements made
or the opinions expressed in this publication.
Copyright 2007 U.S. Society on Dams
Printed in the United States of America
Library of Congress Control Number: 2007921375
ISBN 978-1-884575-40-2
U.S. Society on Dams
1616 Seventeenth Street, #483
Denver, CO 80202
Telephone: 303-628-5430
Fax: 303-628-5431
E-mail: stephens@ussdams.org
Internet: www.ussdams.org

THREE-DIMENSIONAL FLAC STABILITY AND DEFORMATION ANALYSIS

HOW MUCH CAUTION IS NEEDED?
John W. France, P.E.1

Imran S. Gillani, P.E.2

ABSTRACT

In 2004 Itasca introduced a three-dimensional version of its popular finite difference

analysis program FLAC (Fast Lagrangian Analysis of Continua). The earlier, twodimensional version of FLAC has been widely used for stability and deformation
analyses of embankment dams. The three-dimensional version of the program provides
geotechnical engineers a new tool for the evaluation of embankment dams. Dam
engineers have been particularly interested in applying FLAC3D to the analysis of cases
where liquefiable or weak soils are present in the foundation, but are of limited lateral
extent. In such cases, two-dimensional analysis can underestimate factors of safety and
overestimate deformations, because any liquefiable or weak soils included in the section
are modeled as if they extended infinitely in the direction perpendicular to the twodimensional model cross section. The three-dimensional program allows the effects of
stronger materials at the sides of the liquefiable soils to be considered in the analysis.
In an effort to better understand and calibrate the results obtained from FLAC3D, the
authors completed analyses of a simple case, consisting of a 75-foot-high embankment
founded on a 10-foot thickness of weak soil. The width of the valley was varied and the
variation in factor of safety with valley width was compared with the results of FLAC2D
stability analyses and limit equilibrium stability analyses. As the valley becomes wide,
the FLAC3D results should be comparable to the two-dimensional results. Some
interesting results were found in this comparison. In this paper, the authors will review
their modeling results and provide some observations and conclusions on those results,
including the observation that the FLAC3D results appear to be inherently
unconservative (i.e. higher factors of safety and smaller calculated deformations) when
compared to the comparable FLAC2D and limit equilibrium analyses results.
INTRODUCTION
Over the past decade the computer program Fast Lagrangian Analysis of Continua
(FLAC) has become very popular for numerical analyses of a wide range of geotechnical
problems, including dam embankments under both static and dynamic loadings. The
program that has been widely used for the past decade is a two-dimensional version of
FLAC, referenced in this paper as FLAC2D (Itasca, 1999). In 2002, a three-dimensional
version of FLAC, referenced in this paper as FLAC3D, was introduced (Itasca, 2002).

Vice President and National Dam Technology Leader, URS Corporation, 8181 East Tufts Avenue,
Denver, CO 80112: phone (303) 740-3812: fax (303) 694-3946: john_france@urscorp.com
2
Project Manager, URS Corporation, 8181 East Tufts Avenue, Denver, CO 80112: phone (303) 7403870: fax (303) 694-3946: imran_gillani@urscorp.com

Three-Dimensional FLAC

285

The three-dimensional version of the program is being applied by geotechnical engineers

as a new tool for the evaluation of embankment dams.
The three-dimensional program could be particularly valuable in the analysis of cases
where liquefiable soils, or other low strength soils, are present but are of limited lateral
extent. In such cases, it is generally accepted that two-dimensional analysis can under
estimate factors of safety and overestimate deformations, because any liquefiable soils
included in the section are modeled as if they extend infinitely in the direction
perpendicular to the two-dimensional model cross section. The three-dimensional
program allows the effects of stronger soils at the sides of the liquefiable soils to be
considered in the analysis.
The authors of this paper are interested in using FLAC3D, but first wanted to compare
the results of FLAC3D against more familiar two-dimensional analyses. To make this
comparison, stability analyses and earthquake deformation analyses were completed for a
simplified embankment configuration. This paper begins with a general description of
the FLAC computer program. The simplified embankment dam configuration is then
introduced and the results of stability analyses and earthquake deformation analyses are
presented. Finally, the authors offer some observations and conclusions regarding the
practical implications of the analysis results.
FAST LAGRANGIAN ANALYSIS OF CONTINUA (FLAC)
FLAC, in both two-dimensional and three-dimensional versions, is a two-dimensional,
explicit, finite difference computer program for geotechnical applications. Because
FLAC uses an explicit, time-marching computational routine, it is very well suited for
modeling dynamic deformation analyses, which may involve localized or even general
instability. The explicit, time-marching computational routine allows for modeling of
soils that are at or close to failure much more efficiently than does the implicit finite
element approach. The large strain mode of FLAC allows for inclusion of relatively
large displacements that would be very difficult to compute using the small strain
approach typically used by other programs. Furthermore, the ability to customize the
runs and the working of the actual program through a programming language called FISH
makes FLAC a very effective program for performing complex dynamic deformation
type analyses.
SIMPLIFIED EMBANKMENT CONFIGURATION
The simplified embankment configuration consists of a homogeneous embankment
founded on a single stratum of alluvial soil, all underlain by bedrock, all located within a
symmetrical valley, as illustrated in Figures 1 and 2. Although this case is simplified, it
is not fundamentally different from commonly encountered real cases.
For the simplified example, the embankment height is 75 feet, and the maximum depth of
the alluvial soil stratum is 10 feet. The width of the valley was varied in the three-

286

Modernization and Optimization of Existing Dams and Reservoirs

dimensional analyses to evaluate the effects of the three-dimensional geometry of the

structure and the valley on the results of the FLAC3D analyses.

Figure 1. Simplified Embankment Configuration

Figure 2. Maximum-Height Cross Section for Simplified Embankment Configuration

Two different strength characterizations were used in the analyses, as summarized in
Table 1. Two types of analyses were completed for the simplified embankment: stability
analyses and earthquake deformation analyses.

Three-Dimensional FLAC

287

Table 1. Shear Wave Velocities and Strength Characterizations Used in Analyses

Material
Unsaturated
embankment
Saturated
embankment
Alluvium
Bedrock

Shear Wave
Velocity, fps

Strength Characterization
No. 1
Friction
Cohesion,
c
Angle,

Strength
Characterization No. 2
Friction
Cohesion,
c
Angle,

1200

0 psf

34

34

1200

1600 psf

1400 psf

1200

900 psf

1400 psf

2000

Elastic

0
Elastic

0
Elastic

Elastic

The = 0 strength characterizations for the saturated embankment soils are

representative of those used in analyses of foundation soils that are judged susceptible to
liquefaction. Hence, the stability analyses are representative of post-earthquake
conditions.
STABILITY ANALYSES
For the stability analysis, the FLAC program and the strength reduction method (Dawson
and Roth, 2005) were used. In the strength reduction method, a series of analyses are run
with different values of trial factors of safety, Ftrial. In each case, the strength parameters
(cohesion = c and friction angle = ) used in the deformation model are reduced
according to the following equations:
ctrial = c/Ftrial
trial = arctan(tan/Ftrial)
Ftrial is decreased until calculated deformations suddenly become very large. The value of
Ftrial at which the deformations become large is taken as the factor of safety for the case
being analyzed. The FLAC program actually includes an internal routine that automates
the strength reduction slope stability analysis method.
The FLAC3D model used in the analysis is shown in Figure 3, and a two-dimensional
cross section through the maximum section of the model is shown in Figure 4.
The FLAC3D model was used to calculate factors of safety for a series of analyses with
the valley width (see Figure 1) varied from 150 feet to 1200 feet. The results for Strength
Characterization No. 1 are presented in Figure 5, in the form of a plot of calculated 3D
factor of safety versus valley width. The results for Strength Characterization No. 2 are
presented in Figure 6.
The FLAC3D stability analysis results were compared to two-dimensional stability
factors of safety calculated using three other methods: the limit equilibrium method

288

Modernization and Optimization of Existing Dams and Reservoirs

Figure 3. FLAC3D Model of Simplified Embankment Configuration

Figure 4. FLAC Mesh Through Maximum-Height Cross Section for Simplified

Embankment Configuration
using the computer program UTEXAS3, the finite difference method using FLAC2D,
and the finite difference method using FLAC3D with a slice model. The slice
model in FLAC3D analyzes the behavior of a single row of three-dimensional elements

Three-Dimensional FLAC

289

with roller boundary conditions on the sides. The row of elements is located along the
maximum height section of the dam. The calculated two-dimensional factors of safety
are also shown on Figures 5 and 6, for Strength Characterizations Nos. 1 and 2,
respectively.

Figure 5. Summary of Analysis Results for Strength Characterization No. 1

Figure 6. Summary of Analysis Results for Strength Characterization No. 2

290

Modernization and Optimization of Existing Dams and Reservoirs

The FLAC3D and FLAC2D models were constructed with the same element size, so that
element size would not influence the results.
EARTHQUAKE DEFORMATION ANALYSES
All of the cases evaluated in the stability analyses were also evaluated in the earthquake
deformation analyses. The FLAC models were subjected to an input earthquake motion
at the base of the model. The acceleration time history for this earthquake motion is
shown in Figure 7.

Figure 7. Earthquake Time History Used for Analysis

For Strength Characterization No. 1, the two-dimensional factor of safety is less than 1.0,
indicating that the two-dimensional section is unstable. For the earthquake deformation
analysis for this strength characterization, the foundation strength was initially set at c =
1400 psf, = 0, and then reduced to c = 900 psf, = 0, representing a very simple
model of foundation liquefaction. For Strength Characterization No. 2, the twodimensional factor of safety is greater than 1.0, indicating that the two-dimensional
section is stable. For the earthquake deformation analysis for this strength
characterization, the foundation strength was set at c = 1400 psf, = 0 throughout the
analysis.
All FLAC earthquake deformation analyses were run in large-strain mode.
The permanent horizontal and vertical movements of the crest were selected as a basis of
comparing the results of the earthquake deformation. In all cases, the permanent

Three-Dimensional FLAC

291

horizontal crest movements were in the downstream direction and the vertical crest
movements were settlements.
The calculated permanent horizontal and vertical movements for all analysis cases are
noted on Figures 5 and 6 and are listed in Table 2.
Table 2. Permanent Crest Movements from Earthquake Deformation Analyses
Strength Characterization No. 1

Strength Characterization No. 2

Analysis Case

Factor
of
Safety

Permanent
Horizontal
Movement, ft

Permanent
Vertical
Movement, ft

Factor
of
Safety

Permanent
Horizontal
Movement, ft

Permanent
Vertical
Movement,
ft

FLAC3D 150-ft valley

1.22

3.03

1.18

1.28

2.99

1.23

FLAC3D 300-ft valley

1.11

3.27

1.29

1.21

2.82

1.07

FLAC3D 600-ft valley

1.00

4.08

1.85

1.16

2.76

1.04

FLAC3D 1200-ft
valley

0.96

5.02

2.52

1.13

2.74

1.01

FLAC3D Slice Mode

0.94

4.61

2.33

1.11

2.42

0.85

FLAC2D

0.87

7.99

4.44

1.03

2.91

1.10

UTEXAS3

0.89

--

--

1.04

--

--

OBSERVATIONS
As can be seen in Figures 5 and 6, the calculated FLAC3D stability factors of safety
decrease as the valley width increases; a trend that would be expected. For Strength
Characterization No. 1, the three-dimensional factors of safety decreased from 1.22 for a
150-foot valley width to 0.96 for a 1200-foot valley width. For Strength Characterization
No. 2, the three-dimensional factors of safety decreased from 1.28 to 1.13 for the same
range of valley widths. As the valley becomes wider, the three-dimensional factor of
safety approaches the two-dimensional values, coming very close (within about 0.02) to
the value from the slice model. However, the three-dimensional factors of safety for the
1200-foot wide FLAC models and for the FLAC 3D slice models are higher than the twodimensional values from UTEXAS3 and FLAC2D, which agreed within about 0.02 with
each other. The differences between the FLAC3D factors of safety and the FLAC2D and
UTEXAS3 factors of safety are greater for Strength Characterization No. 1 than for
Strength Characterization No. 2. For Characterization No. 1, the FLAC3D factors of
safety for the wide valley and for the slice method are about 8 percent higher than the
two-dimensional factors of safety. For Characterization No. 2, the FLAC3D factors of
safety for the wide valley and for the slice method are about 3 percent higher than the
two-dimensional factors of safety.
For Strength Characterization No. 1, the calculated crest movements for the FLAC3D
cases increase as the valley widens and the three-dimensional factors of safety decrease,
as would be expected. For the FLAC3D model for the 1200-foot wide valley, the
calculated crest deformations are within 10 percent of those for the FLAC3D slice model,
although curiously the deformations for the full FLAC3D model are slightly greater than
those for the slice model. The calculated crest deformations for the FLAC2D model are
significantly greater than those from either the 1200-foot valley FLAC3D model or the

292

Modernization and Optimization of Existing Dams and Reservoirs

FLAC3D slice model. Comparing the deformations for the FLAC2D model to those for
the FLAC3D model for the 1200-wide valley, the horizontal deformations are 58 percent
higher for FLAC2D and the vertical deformations are 76 percent higher for FLAC2D.
The results for Strength Characterization No. 2 indicated some unexpected results. The
calculated crest movements for the full FLAC3D models actually decreased as the valley
width increased and the factors of safety decreased. All of the horizontal movements
from these models were within about 10 percent of each other and all of the vertical
movements were within about 20 percent of each other. These results would generally be
considered to be essentially the same within the expected accuracy of geotechnical
deformation calculations, however, there is no clear explanation of the unexpected trend
in these calculated deformations. For the FLAC3D model for the 1200-foot wide valley,
the calculated crest deformations are higher than those for the FLAC3D slice model, with
the horizontal deformations about 13 percent higher and the vertical deformations about
19 percent higher. The calculated crest deformations for the FLAC3D model for the
1200-foot wide valley are also less than those from the FLAC2D model, with the
horizontal deformations about 6 percent less and the vertical deformations about 8
percent higher.
CONCLUSIONS
It would be anticipated that the FLAC3D results for the 1200-foot wide valley models
would agree reasonably well with the results of the FLAC2D and UTEXAS3 analyses.
The height to width ratio of these FLAC3D models exceeds 12, for which it would be
anticipated that three-dimensional effects would be minimal. Yet, in this case the
FLAC3D model appears to be producing somewhat higher factors of safety than the twodimensional analyses. Further, the difference appears to be greater for the case where the
two-dimensional factor of safety is less than 1.0 (Strength Characterization No. 1), which
may be the case for which engineers would be most interested in using the program to
analyze three-dimensional effects. The fact that the FLAC3D slice method also produces
factors of safety greater than those from the two-dimensional method suggests that there
may be something inherent in the mathematical formulation of FLAC3D that produces
results that are unconservative relative to two-dimensional analyses.
Similarly, it appears that for this case deformations calculated in the FLAC3D 1200-foot
valley model are less than those calculated in FLAC2D, and the difference is more
pronounced for the case of an initial factor of safety less than 1.0. One unexpected and as
yet unexplained result was the observation of increasing calculated crest deformation
with decreasing valley width for Strength Characterization No. 2. This reverse trend is
perplexing, but may not be of large significance, because the calculated crest
deformations for all of the valley widths for Strength Characterization No. 2 are
essentially equivalent within the limits of accuracy of geotechnical deformation
calculations.
Considering all of the results presented in this paper, the authors conclude that FLAC3D
can help to provide some insights on three-dimensional behavior of cases like that

Three-Dimensional FLAC

293

analyzed for this paper. However, at this time the results should be applied with caution,
because it seems that they are likely unconservative to different degrees when compared
with the results of two-dimensional analyses. This is significant because comparisons
between two-dimensional analyses and observed performance have provided the
profession with the basis used to establish acceptable factors of safety and calculated
deformations. In cases when FLAC3D analyses are used for dam embankment stability
and deformation analyses, it would desirable to complete companion FLAC3D slice
analyses and FLAC2D analyses to help gauge the magnitude of the differences in results
between FLAC3D and FLAC2D.
REFERENCES
Itasca Consulting Group. 1999. FLAC Version 4.0 Fast Lagrangian Analysis of
Continua, Users Guide.
Itasca Consulting Group. 2002. FLAC3D Fast Lagrangian Analysis of Continua in 3
Dimensions, Users Guide.
Dawson, E. and Roth, W. 2005. 3-D Slope Stability Analysis By Strength Reduction.
25th Annual USSD Conference, Salt Lake City, UT, June 6-10, 2005.

294

Modernization and Optimization of Existing Dams and Reservoirs