You are on page 1of 40

Technical Paper by R.J. Bathurst and K.

Hatami

SEISMIC RESPONSE ANALYSIS OF A


GEOSYNTHETIC-REINFORCED SOIL
RETAINING WALL
ABSTRACT: The paper reports results from numerical experiments that were carried out
to investigate the influence of reinforcement stiffness, reinforcement length, and base
boundary condition on the seismic response of an idealized 6 m high geosynthetic-reinforced
soil retaining wall constructed with a very stiff continuous facing panel. The numerical mod-
els were excited at the foundation elevation by a variable-amplitude harmonic motion with
a frequency close to the fundamental frequency of the reference structure. The two-dimen-
sional, explicit dynamic finite difference program Fast Lagrangian Analysis of Continua
(FLAC) was used to carry out the numerical experiments. Numerical results illustrate that
the seismic response of the wall is very different when constructed with a base that allows
the wall and soil to slide freely and when the wall is constrained to rotate only about the toe.
Parametric analyses were also carried out to investigate the quantitative influence of the
damping ratio magnitude used in numerical simulations and the effects of distance and type
of far-end truncated boundary. The response of the same wall excited by a scaled earthquake
record was demonstrated to preserve qualitative features of wall displacement and reinforce-
ment load distribution as that generated using the reference harmonic ground motion applied
at 3 Hz. The lessons learned in this study are of value to researchers using dynamic numerical
modeling techniques to gain insight into the seismic response of reinforced wall structures.

KEYWORDS: Seismic analysis, Numerical modeling, Parametric analysis, Finite


difference, FLAC, Retaining walls, Geosynthetic reinforcement, Metallic reinforcement.

AUTHORS: R.J. Bathurst, Professor, and K. Hatami, Research Associate, Department of


Civil Engineering, Royal Military College of Canada, P.O. Box 17000, STN Forces,
Kingston, Ontario, K7K 7B4, Canada, Telephone: 1/613-541-6000, Ext. 6479, Telefax:
1/613-545-8336, E-mail: bathurst@rmc.ca.

PUBLICATION: Geosynthetics International is published by the Industrial Fabrics


Association International, 1801 County Road B West, Roseville, Minnesota 55113-4061,
USA, Telephone: 1/612-222-2508, Telefax: 1/612-631-9334. Geosynthetics International is
registered under ISSN 1072-6349.

DATES: Original manuscript received 10 February 1998, revised version received 9 March
1998 and accepted 17 March 1998. Discussion open until 1 September 1998.

REFERENCE: Bathurst, R.J. and Hatami, K., 1998, “Seismic Response Analysis of a
Geosynthetic-Reinforced Soil Retaining Wall”, Geosynthetics International, Vol. 5, Nos.
1-2, pp. 127-166.

GEOSYNTHETICS INTERNATIONAL S 1998, VOL. 5, NOS. 1-2 127

Downloaded by [ Purdue Univ Lib TSS] on [19/09/16]. Copyright © ICE Publishing, all rights reserved.
BATHURST AND HATAMI D Seismic Response Analysis of a Geosynthetic-Reinforced Wall

1 INTRODUCTION

In North America, geosynthetic- and metal strip-reinforced soil walls are routinely
designed using limit-equilibrium pseudostatic methods for sites with peak horizontal
ground accelerations ≤ 0.29g (AASHTO 1996; FHWA 1996). A limitation of pseudo-
static methods is that they cannot consider the effects of duration of seismic loading,
frequency content, acceleration amplification through the backfill soil, and foundation
condition on the development of reinforcement loads and structure deformations (Ba-
thurst and Alfaro 1997; Bathurst and Cai 1995). Displacement methods developed
from classical Newmark sliding-block models have been proposed to predict seismic
load-induced deformation of reinforced structures (Cai and Bathurst 1996; Ling et al.
1997a,b). However, pseudostatic limit-equilibrium methods for design against col-
lapse and pseudostatic displacement methods for design against excessive deforma-
tions are not satisfactory if the objective is to investigate the coupled effects of
reinforcement properties, structure geometry, and foundation condition on reinforced
soil wall performance under a prescribed seismic event. Unfortunately, only limited
physical data from reduced-scale shaking table tests are available to guide the develop-
ment of rational models.
Carefully conceived and executed numerical experiments offer the possibility to im-
prove the understanding of the effects of dynamic loading on reinforced soil structures
and to demonstrate the influence of the primary component properties (e.g. reinforce-
ment stiffness, number of reinforcement layers, base condition, wall geometry, and fac-
ing type) on the system response to an earthquake.
Numerical experiments were carried out to investigate the influence of reinforcement
stiffness, reinforcement length, and toe restraint condition on the seismic response of
an idealized 6 m high geogrid-reinforced soil retaining wall constructed with a very stiff
continuous facing panel. The wall height, number of reinforcement layers, and rein-
forced soil volumes are typical of actual structures in the field. The wall was subjected
to base excitation using a variable-amplitude harmonic motion with a frequency close
to the fundamental frequency of the reference structure. The frequency of the applied
input base acceleration is representative of a typical predominant frequency of me-
dium- to high-frequency content earthquakes. The excitation frequency was chosen to
generate relatively large displacements and reinforcement loads during base shaking
and thus help identify performance differences between models with different proper-
ties, but at the same time ensure that the models were numerically stable.
The two-dimensional, explicit, dynamic finite difference program Fast Lagrangian
Analysis of Continua (FLAC 1995) was used to perform the numerical experiments.
Analyses were carried out after first confirming that the FLAC program gives similar
predictions to those reported in the literature using a finite element method (FEM) tech-
nique applied to the same idealized reinforced wall structure under static loading (Rowe
and Ho 1997).
A second set of parametric analyses was carried out with a range of material damping
ratios, variable widths of numerical grid, and different far-end truncated boundary
conditions to investigate the effect of these model parameters on the predicted wall re-
sponse. The seismic response of the reinforced wall model subjected to a range of har-
monic ground motion frequencies was also examined. Finally, the response of the wall

128 GEOSYNTHETICS INTERNATIONAL S 1998, VOL. 5, NOS. 1-2

Downloaded by [ Purdue Univ Lib TSS] on [19/09/16]. Copyright © ICE Publishing, all rights reserved.
BATHURST AND HATAMI D Seismic Response Analysis of a Geosynthetic-Reinforced Wall

to a scaled earthquake record with the same peak acceleration as the reference harmonic
ground motion was investigated.

2 PREVIOUS RELATED WORK

2.1 Shaking Table Tests

Chida et al. (1985) carried out a series of experiments on a half-scale shaking table test
model of a metal strip-reinforced soil wall. The physical model was constructed with
four equal height incremental concrete panels for a total wall height of 3 m. A 1.4 m high
unreinforced sloped fill was placed over the reinforced section that incorporated eight,
4 m long steel strip reinforcement layers. The physical model was 5.2 m from the toe to
the back of the shaking table container. Hence, the width of model to height of facing
ratio was approximately 1.7 with only a 1.2 m column of unreinforced soil between the
back of the reinforced soil zone and the back of the shaking table container. The back
vertical boundary of the experiment comprised a rigid wall attached directly to the shak-
ing table. The toe of the wall was constrained horizontally. The model base was excited
in a series of experiments using a sinusoidal input acceleration with different frequen-
cies ranging from f = 2 to 7 Hz and peak base accelerations ranging from approximately
amax = 0.1g to 0.4g. Resulting maximum reinforcement loads were approximately the
same magnitude in all layers and were observed to increase linearly with increasing peak
base acceleration for frequencies less than 7 Hz and peak base accelerations less than
approximately 0.4g. A maximum frequency of 7 Hz for a half-scale model corresponds
to approximately 5 Hz at prototype scale. For tests carried out with an estimated base
acceleration of amax = 0.4g at f = 2 Hz, and amax = 0.17g at f = 7 Hz, there was a nonuniform
distribution of the dynamic load increment in the reinforcement layers with large values
developed in the layers at the top of the structure. Here, the dynamic load increment in
a reinforcement layer is the difference between the maximum tensile load under seismic
loading and the maximum tensile load under static loading. Peak acceleration amplifi-
cation between the base of the wall and the top of the wall was observed to increase by
a factor of two at a frequency f = 5 Hz (f = 3.5 Hz at prototype scale).
Murata et al. (1994) reported the results of shaking table tests carried out on a rein-
forced embankment model with a crest width of 3.45 m and contained by two 2.5 m high
walls constructed with gabion baskets and an outer continuous concrete panel. The
model was subjected to harmonic and actual earthquake records. The harmonic record
was observed to generate larger deformations than the earthquake record.
Sakaguchi (1996) carried out shaking table tests on a 1.5 m high model test of a rein-
forced wall. The tests were constructed with lightweight blocks and five layers of geo-
grid reinforcement. The tests showed that wall displacements and permanent strains in
the reinforcement accumulated with time during a 4 Hz sinusoidal base acceleration
record applied for 7 seconds (peak accelerations up to approximately 0.5g were applied
to the model). The reinforced zone was observed to act as a monolithic body with no
evidence of a yield surface propagating across the reinforcement layers even after large
wall displacements developed.

GEOSYNTHETICS INTERNATIONAL S 1998, VOL. 5, NOS. 1-2 129

Downloaded by [ Purdue Univ Lib TSS] on [19/09/16]. Copyright © ICE Publishing, all rights reserved.
BATHURST AND HATAMI D Seismic Response Analysis of a Geosynthetic-Reinforced Wall

2.2 Dynamic Numerical Modeling

Segrestin and Bastick (1988) used the FEM program SUPERFLUSH to generate nu-
merical results that showed excellent agreement with the results of half-scale shaking
table tests reported by Chida et al. (1985). However, the material properties used to
model these physical tests were assumed values since actual material properties were
not reported by Chida et al.
Segrestin and Bastick (1988) used the same program to predict the seismic response
of two hypothetical full-scale reinforced walls (6 and 10.5 m high) that used large artic-
ulated concrete facing panels and steel strips as the soil reinforcement. The walls were
constrained at the toe but seated on regions that simulated three different foundation
materials (hard rock, stiff soil, and loose soil). The width of the numerical model is not
reported by Segrestin and Bastick (1988). The location of maximum loads in the rein-
forcement layers during simulated seismic shaking was generally not at the connec-
tions. The results showed that dynamic load increments carried by reinforcement layers
increased with depth below the crest of the wall.
Yogendrakumar et al. (1991) used a modified version of the dynamic FEM program
TARA-3 to carry out a similar study of the influence of seismic loading on a 6 m high
steel strip-reinforced soil wall. The program modifications included the introduction
of a hysteretic load-strain model for the reinforcement elements. The width of the finite
element mesh was 3.3 times the height of the wall. The numerical results showed that
maximum reinforcement loads during shaking occurred at the connections with the
wall, and significant dynamic load increments were generated when the first 10 seconds
of the 1940 El Centro earthquake record (scaled to a peak acceleration of 0.2g) was ap-
plied to the base of the model. The magnitude of the dynamic load increment in each
reinforcement layer close to the back of the wall facing was observed to increase, almost
linearly, with depth below the crest of the wall. The maximum amplification factor (i.e.
the ratio of the maximum acceleration in the structure to the peak input base accelera-
tion) was approximately 1.4 and occurred at the top of the wall.
Cai and Bathurst (1995) used the same modified version of TARA-3 to investigate
the response of a 3.2 m high segmental retaining wall reinforced with a polymeric geo-
grid subjected to a scaled 1940 El Centro earthquake record. The width to height ratio
of the FEM of the structure was approximately 4.2, and the far-end truncated boundary
was treated as a free-field energy transmitting boundary (i.e. a boundary condition that
simulates an infinitely wide domain with respect to elastic wave transmission). The nu-
merical results showed that wall displacements and reinforcement loads accumulated
with time during seismic shaking. Displacement and reinforcement amplitudes were in-
significant in magnitude compared to the permanent values predicted at the end of base
shaking. In contrast to the earlier work by Yogendrakumar et al. (1991) that investigated
walls with a stiffer facing and stiffer reinforcement, the numerical results for the discrete,
deformable segmental retaining wall models showed that reinforcement loads under
both static and dynamic loading conditions were attenuated at the connections. Also,
the dynamic load increments in the reinforcement layers did not increase linearly with
depth below the crest of the wall. Base acceleration amplifications were very small (≤
1.2) which is likely due to the low height of the walls that were investigated.

130 GEOSYNTHETICS INTERNATIONAL S 1998, VOL. 5, NOS. 1-2

Downloaded by [ Purdue Univ Lib TSS] on [19/09/16]. Copyright © ICE Publishing, all rights reserved.
BATHURST AND HATAMI D Seismic Response Analysis of a Geosynthetic-Reinforced Wall

2.3 Summary

The very limited number of experimental and numerical studies identified above, the
limited scope of each study, the different constitutive models and numerical codes used,
and the wide range of results illustrate that the current understanding of the seismic re-
sponse of reinforced soil walls is incomplete.
The shaking table tests carried out by Chida et al. (1985) represent an important set
of experimental results. Nevertheless, the extrapolation of their results to the field case
must be done with caution. In particular, the very narrow volume of material in the re-
tained soil zone and the rigid far-end boundary will influence model response as demon-
strated in the current paper.
Current practice in North America (AASHTO 1996; FHWA 1996) with regard to em-
pirical rules to calculate dynamic load increments in reinforcement layers is based on
the interpretation of numerical results reported by Segrestin and Bastick (1988). The
simulations performed by Segrestin and Bastick (1988) were restricted to one type of
reinforcement (metal strips), one reinforcement length, and a common footing condi-
tion. Segrestin and Bastick (1988) clearly state that the numerical simulation results do
not apply to other reinforcement materials (e.g. geosynthetics). Based on a review of
the limited data available in the literature, Bathurst and Alfaro (1997) have also noted
that the applicability of empirical rules developed using numerical simulation results
of the seismic response of metal strip-reinforced soil walls may not be applicable to
nominal identical walls constructed with a less stiff (polymeric) reinforcement.
The current study is a first step toward identifying a numerical method and systematic
approach that can ultimately be used to validate or improve current seismic design prac-
tice for continuous panel walls constructed with a range of soil reinforcement products.

3 FLAC PROGRAM

FLAC is an explicit, dynamic, finite difference code based on the Lagrangian calcula-
tion scheme that is well suited for modeling large distortions, material collapse, and the
dynamic response of earth structures. Complete descriptions of the numerical formula-
tion are reported by Cundall and Board (1988). Several built-in constitutive models are
available in the FLAC package. Users can also implement their own models. Other ad-
vantages of using FLAC for seismic analysis is the simplicity of applying seismic load-
ing anywhere within the problem domain and the excellent post-processing
capabilities. Prior to the time of this study, FLAC had not been used to investigate the
seismic response of reinforced soil walls even though the program is widely used by
geotechnical and mining engineers for a range of problems. The results of some initial
FLAC modeling for seismic response analysis of reinforced slopes and geofoam seis-
mic buffers are summarized in the paper by Bathurst and Alfaro (1997) and represent
the only related work to date.

4 COMPARISON OF FEM RESULTS WITH FLAC RESULTS

Prior to carrying out the parametric studies that are the focus of the current paper, se-
lected FEM results for a geosynthetic-reinforced continuous panel wall at the end of

GEOSYNTHETICS INTERNATIONAL S 1998, VOL. 5, NOS. 1-2 131

Downloaded by [ Purdue Univ Lib TSS] on [19/09/16]. Copyright © ICE Publishing, all rights reserved.
BATHURST AND HATAMI D Seismic Response Analysis of a Geosynthetic-Reinforced Wall

construction were compared to results using the FLAC program. This comparison was
undertaken to develop confidence that the FLAC program can give reasonably similar
results to carefully conceived and executed FEM models simulating a static load condi-
tion. The example, plane strain reinforced wall selected for this purpose was the base
case example reported by Rowe and Ho (1993, 1997) and Ho (1993). While it would
be desirable to compare FLAC results to the physical test results reported by Chida et
al. (1985), this is not possible because no material properties were given in the report
by Chida et al.

4.1 Continuous Panel Wall Models

The reference continuous panel wall is 6.0 m high with six uniformly spaced rein-
forcement layers (Figure 1). The wall facing was modeled as a continuous concrete pan-
el with a thickness of 0.14 m. The bulk and shear modulus values of the wall were Kw
= 11,430 MPa and Gw = 10,430 MPa, respectively. Poisson’s ratio for the panel material
was taken as νw = 0.15. The panel was hinged at its base, as illustrated in Figure 1. The
soil was modeled as a purely frictional, elastic-plastic material with a Mohr-Coulomb
failure criterion and nonassociated flow rule. The friction angle of the soil was Ô = 35_,
dilatancy angle ψ = 6_, and unit weight γ = 20 kN/m3. These properties and dimensions
were the same as those reported by Rowe and Ho (1993) and Ho (1993). Similarly, the
following model was used in the current study and by Rowe and Ho (1993) to calculate
the elastic modulus of the soil, Es :
m
Es
Pa
= K Pσ 
3

a
(1)

where: K = 460 and m = 0.5 are constant coefficients; Pa = atmospheric pressure; and
σ3 = minor principal effective stress in the soil. A constant Poisson’s ratio value for the

Interface column
L = 4.25 m Fixed boundary
Reinforcement

Facing
panel H=6 m

Hinge Fixed boundary


B=15 m
Interface
layer
Figure 1. Numerical grid for the reference static load case.

132 GEOSYNTHETICS INTERNATIONAL S 1998, VOL. 5, NOS. 1-2

Downloaded by [ Purdue Univ Lib TSS] on [19/09/16]. Copyright © ICE Publishing, all rights reserved.
BATHURST AND HATAMI D Seismic Response Analysis of a Geosynthetic-Reinforced Wall

soil material was assumed (νs = 0.3). However, for simplicity, in the current study, the
elastic modulus was held constant for the duration of the numerical experiment, includ-
ing construction, using values calculated immediately prior to prop removal and taking
σ3 as the horizontal earth pressure in each soil layer. Hence, soil modulus and Poisson’s
ratio values were not updated as was done by Rowe and Ho (1993, 1997).
The reinforcement layers were modeled using linear, elastic-plastic cable elements
with negligible compressive strength and an equivalent cross-sectional area of 0.002
m2. The stiffness of the reinforcement was taken as J = 2,000 kN/m. The tensile yield
strength of the reinforcement was set to Ty = 200 kN/m to ensure that reinforcement rup-
ture was not a failure mechanism and to be consistent with the Rowe and Ho (1993,
1997) model. The interface between the reinforcement (cable elements) and the soil
was modeled by a grout material of negligible thickness with an interface friction angle
δg = 35_. The bond stiffness and bond strength of the grout were taken as kb = 2×106
MN/m/m and sb = 103 kN/m, respectively. The interface and grout properties were se-
lected to simulate a perfect bond between the soil and reinforcement layers.
The results of the FEM simulation of reinforced continuous panel walls have been
demonstrated to be sensitive to mesh construction details and material properties at the
reinforcement-wall connections (Rowe and Ho 1997; Andrawes and Yogarajah 1994).
In the current study, a simple connection model was adopted that involved attaching the
end of the cable elements (reinforcement) to a single grid point at the back surface of
the continuous panel region.
The wall-soil interface was modeled using a thin soil column, 0.05 m thick, directly
behind the facing panel. A no-slip boundary was used between the thin soil column and
the facing panel. The soil-wall interface column material was assigned a friction angle
Ôi = 20_ and a dilatancy angle ψi = 0. A similar thin soil layer was introduced at the base
of the soil region but was assigned the same properties as the reinforced and retained
soil materials. These interface zones in the numerical grid were introduced to match,
as closely as possible, the Rowe and Ho (1993, 1997) model that used interface ele-
ments to model the same boundaries.
The soil and reinforcement elements were constructed in layers, while the continuous
panel was braced horizontally using rigid external supports. The panel supports were
then released in sequence from the top of the structure. Similar to the work by Rowe
and Ho (1993, 1997), no attempt was made to model compaction-induced stresses in
this static load simulation (or in the dynamic simulations presented in Section 5).
The differences between the current simulations and the reference FEM model are:
(i) the facing panel-soil interface and soil-foundation interface were modeled using thin
layers of soil rather than zero thickness interface elements; and (ii) the elastic modulus
of the soil was kept at a constant value (corresponding to the end of soil placement) at
all stages during the numerical experiment and was not updated.
The choice of thin soil layers to model soil interfaces was made because zero-thick-
ness interfaces are not permissible in FLAC (Version 3.30) in combination with inter-
secting free-field boundaries used in dynamic modeling. Hence, the use of thin layers
of soil to model selected interfaces for static loading was preferred. Finally, the use of
a constant elastic modulus for the soil reduces computation time for both static and dy-
namic loading simulations using FLAC.

GEOSYNTHETICS INTERNATIONAL S 1998, VOL. 5, NOS. 1-2 133

Downloaded by [ Purdue Univ Lib TSS] on [19/09/16]. Copyright © ICE Publishing, all rights reserved.
BATHURST AND HATAMI D Seismic Response Analysis of a Geosynthetic-Reinforced Wall

4.2 Comparison of Results

The values of lateral displacement of the wall face and normalized (axial) connection
loads in reinforcement layers after release of the external props are plotted in Figure
2. The corresponding results reported by Ho (1993) are also presented in the Figure 2.
The displacement profiles from both studies are in close agreement as illustrated in Fig-
ure 2a. The computed values of connection load, Tc , in Figure 2b have been normalized
with respect to the theoretical value of the Rankine active soil pressure at the bottom
of the wall. Connection loads are in close agreement over the top half of the wall with
lower loads calculated using the FLAC model for the bottom half of the wall height
when compared to the corresponding FEM results. The differences may be due to the
calculation of soil elastic modulus and the treatment of the wall-soil interface. While,
not attempted here, agreement between results could be improved by adopting an up-
dated stress dependent modulus of elasticity in FLAC models for the soil similar to that

(a) 6

4 FLAC simulation
Elevation (m)

1 Ho (1993)

0
0 5 10 15 20 25 30 35
Displacement (mm)
(b) 6
Ho (1993)
5

4
Elevation (m)

2
FLAC simulation
1

0
0.0 0.1 0.2 0.3 0.4 0.5
Normalized connection load, Tc / Ka γ H

Figure 2. Comparison of FLAC results with finite element model results reported by Ho
(1993) for the end-of-construction condition: (a) wall displacements; (b) normalized axial
loads in the reinforcement at the wall connections.

134 GEOSYNTHETICS INTERNATIONAL S 1998, VOL. 5, NOS. 1-2

Downloaded by [ Purdue Univ Lib TSS] on [19/09/16]. Copyright © ICE Publishing, all rights reserved.
BATHURST AND HATAMI D Seismic Response Analysis of a Geosynthetic-Reinforced Wall

adopted by Rowe and Ho (1993, 1997). For the purposes of the comparative parametric
analyses in Section 6, this was deemed unnecessary when compared to the benefit of
keeping the details of dynamic modeling as simple as possible and to minimize com-
putational time.

5 DYNAMIC MODELING USING FLAC

Dynamic analyses of a reinforced soil wall subjected to simulated horizontal founda-


tion shaking due to an earthquake were carried out using numerical models with the
same height and number of reinforcement layers as those described for the static load
FLAC model in Section 4. The numerical grid for the reference geometry in the current
study is illustrated in Figure 3.
The test series in the current study can be divided into two sets. One set of data is fo-
cused on the influence of material properties and dimensions on seismic response of the
numerical models. The corresponding variables include: reinforcement stiffness, rein-
forcement length, and toe restraint condition (Table 1). A second set of data is focused
on the influence of the type and location of the far-end truncated boundary and magni-
tude of material damping ratio on numerical results (Table 2).
Several additional numerical analyses were carried out to investigate the influence
of frequency of the reference harmonic base input acceleration function on wall re-
sponse. Finally, a numerical analysis using the initial six seconds of an actual scaled
earthquake accelerogram was carried out to investigate quantitative and qualitative dif-
ferences in the seismic response of the idealized reinforced wall due to harmonic excita-
tion and a typical earthquake record.

Right edge of numerical grid and


Thin soil interface column free-field transmitting boundary
Very stiff
B = 40 m
facing
panel 10 m
Reinforcement
Layer 6
Non-yielding
5
region
4
H=6m
3
2
Hinge 1
Thin horizontal
1m soil layer (sliding
case only)
L = 4.2 m Fixed boundary
Very stiff foundation (fixed case only)
Free-field transmitting boundary Base acceleration
Figure 3. Numerical grid for the reference reinforced soil wall with a fixed-base condition.

GEOSYNTHETICS INTERNATIONAL S 1998, VOL. 5, NOS. 1-2 135

Downloaded by [ Purdue Univ Lib TSS] on [19/09/16]. Copyright © ICE Publishing, all rights reserved.
BATHURST AND HATAMI D Seismic Response Analysis of a Geosynthetic-Reinforced Wall

Table 1. Parametric values for the influence of the reinforcement length, reinforcement
stiffness, and the base condition.

Run Base condition L/H J (kN/m) Run Base condition L/H J (kN/m)
1 Fixed 0.7 500 11 Fixed 1 500
2 Fixed 0.7 1,000 12 Fixed 1 1,000
3 Fixed 0.7 2,000 13 Fixed 1 2,000
4 Fixed 0.7 9,000 14 Fixed 1 9,000
5 Fixed 0.7 69,000 15 Fixed 1 69,000
6 Sliding 0.7 500 16 Sliding 1 500
7 Sliding 0.7 1,000 17 Sliding 1 1,000
8 Sliding 0.7 2,000 18 Sliding 1 2,000
9 Sliding 0.7 9,000 19 Sliding 1 9,000
10 Sliding 0.7 69,000 20 Sliding 1 69,000

Notes: B = 40 m; c = 5%; free-field, far-end boundary.

Table 2. Parametric values for the influence of the base condition, far-end boundary
condition, boundary distance, and soil damping ratio.

Run Base condition Far-end boundary condition B (m) c (%)


21 Fixed Rigid stationary 7.5 5
22 Sliding Rigid stationary 7.5 5
23 Fixed Rigid stationary 15 5
24 Sliding Rigid stationary 15 5
25 Fixed Rigid stationary 25 5
26 Sliding Rigid stationary 25 5
27 Fixed Rigid stationary 40 5
28 Sliding Rigid stationary 40 5
29 Fixed Free-field 40 10
30 Fixed Free-field 40 20
31 Fixed Rigid forced 40 5

Notes: L/H = 0.7; J = 2,000 kN/m.

5.1 Numerical Grid and Problem Boundaries

The numerical grid for the reference geometry was selected to represent an infinitely
wide region. The width of the backfill, B, for the reference geometry was extended to 40
m beyond the back of the facing panel and a free-field boundary condition was applied
at the vertical truncated boundaries at the left and right edges of the grid to allow for the
radiation of elastic waves to the far field. The right edge of the grid contains a 10 m wide
non-yielding zone. The width of the reinforced zone, L, for the reference geometry was
selected to give L/H = 0.7 where H is the height of the wall. This is a typical minimum
reinforcement ratio for static design of reinforced soil walls (e.g. FHWA 1996).

136 GEOSYNTHETICS INTERNATIONAL S 1998, VOL. 5, NOS. 1-2

Downloaded by [ Purdue Univ Lib TSS] on [19/09/16]. Copyright © ICE Publishing, all rights reserved.
BATHURST AND HATAMI D Seismic Response Analysis of a Geosynthetic-Reinforced Wall

The base condition of the wall was either fixed (i.e. the toe of the wall was slaved to
the foundation but was free to rotate) or free to slide horizontally and rotate about the
toe. The results of experimentally measured toe loads from a full-scale, continuous pan-
el wall has demonstrated that these loads can be very large for walls with a hinged toe
and hence toe restraint can significantly add to the capacity of continuous panel walls
to carry earth loads (Bathurst 1993). For sliding cases, the wall model was seated on
a thin, 0.05 m thick, region of soil that was extended across the full width of the numeri-
cal grid. The layer performed a similar function to sliding interface elements in FEM
work and was required to ensure that models representing walls without horizontal toe
restraint (i.e. sliding-wall cases) were not artificially restrained during shaking. For the
fixed-base condition, the wall and soil regions were connected directly to a foundation
base comprising a 1 m thick layer of very stiff material.
The length of the reinforcement, L, was varied to give L/H = 0.7 or 1.0 and, thus, the
influence of the reinforced soil volume on the system response could be examined
(Table 1). Ho and Rowe (1996) have shown that for uniformly spaced reinforcement
there is little effect of reinforcement length on the maximum tensile loads in reinforce-
ment layers for L/H ≥ 0.7 and static loading conditions.
In order to examine the influence of grid width on numerical results, the location of
the far-end boundary behind the wall was varied using values of B = 7.5, 15, and 25 m
applied to the reference geometry (Table 2). Finally, several additional runs were car-
ried out to examine the influence of the type of far-end truncated boundary on the sys-
tem response to simulated seismic loading.

5.2 Material Properties

The backfill soil properties used in the dynamic analyses were identical to those re-
ported for the static load FLAC model described in Section 4.1 with the exception that
this material was assigned constant values of bulk modulus (Ks = 27.5 MPa or, equiva-
lently, Es = 33 MPa) and shear modulus (Gs = 12.7 MPa). Constant values were selected
in order to minimize the number of problem parameters. The non-yielding material was
assigned the same properties as the backfill soil with the exception that a very large co-
hesion was used. A non-yielding region was necessary since FLAC does not allow a
free-field boundary to be in contact with yielded material. The location of the left hand
boundary of the non-yielding region for most analyses was selected to ensure that it did
not intersect the active soil zone that develops immediately behind the reinforced soil
zone during seismic shaking. The foundation zone in the fixed-base models was as-
signed the same material properties as the concrete facing panel.
The friction angle of the interface soil column between the reinforced soil zone and
the panel wall was set to Ôi = 20_ with the remaining soil properties matching the
properties of the backfill soil. For sliding cases, the continuous horizontal thin layer at
the toe elevation of the wall was assigned the same soil properties as the backfill soil.
The reinforcement was modeled using cable elements and grout material as described
in Section 4.1.
In the parametric analyses, the linear elastic stiffness value for the cables varied over
a range of values from J = 500 to 69,000 kN/m (Table 1). The lowest values in Table
1 are typical of the index stiffness of some low strength polymeric geogrids and the
highest value was selected to correspond to steel strip reinforcement. The range of

GEOSYNTHETICS INTERNATIONAL S 1998, VOL. 5, NOS. 1-2 137

Downloaded by [ Purdue Univ Lib TSS] on [19/09/16]. Copyright © ICE Publishing, all rights reserved.
BATHURST AND HATAMI D Seismic Response Analysis of a Geosynthetic-Reinforced Wall

stiffness values investigated also captures the magnitude of the initial stiffness values
of a typical, woven polyester geogrid and a typical high density polyethylene (HDPE)
geogrid under cyclic loading at a frequency of 3 Hz (Bathurst and Cai 1994). However,
the yield strength of the reinforcement in all cases was kept constant at Ty = 200 kN/m,
which is well above the magnitude of the maximum reinforcement load recorded in the
simulations.

5.3 Seismic Loading

Soil construction and prop release were identical to those described for the static load-
ing case in Section 4.1. After static equilibrium was achieved, the full width of the
foundation was subjected to the variable-amplitude harmonic ground motion record il-
lustrated in Figure 4. This acceleration record was applied horizontally to all nodes at
the bottom of the soil zone at equal time intervals of ∆t = 0.05 s. The accelerogram has
both increasing and decaying peak acceleration portions and is expressed as:

u(t) = βe −αt t ζ sin (2πft)


..
(2)

where: α = 5.5, β = 55, and ζ = 12 are constant coefficients; f = frequency; and t = time.
The peak amplitude of the input acceleration is 0.2g, and the frequency, f = 3 Hz, was
selected to represent a typical predominant frequency of medium- to high-frequency
content earthquakes (Figure 4). A frequency of 3 Hz was also chosen because it gave
stable numerical results in all simulations while generating large displacements and
large reinforcement loads within a relatively short simulation time (6 seconds). Large
displacements and reinforcement loads were judged to be useful by the authors of the
current paper in order to identify the relative influence of primary variables on seismic
response of a reinforced soil wall of typical height and reinforced soil volume. Howev-
er, it is important to note that the ground acceleration function selected in the current
study is not representative of actual earthquake records that have a range of frequency
content and typically longer durations. However, the selection of a single earthquake
record, scaled to a target peak acceleration, was judged to add more complexity to the
proposed work and would require greater run times without the assurance that potential-
ly large seismic-induced effects would result. The authors of the current paper are inves-

2
Acceleration (m/s2)

---1

---2
0 1 2 3 4 5 6
Time (s)
Figure 4. Base acceleration history.

138 GEOSYNTHETICS INTERNATIONAL S 1998, VOL. 5, NOS. 1-2

Downloaded by [ Purdue Univ Lib TSS] on [19/09/16]. Copyright © ICE Publishing, all rights reserved.
BATHURST AND HATAMI D Seismic Response Analysis of a Geosynthetic-Reinforced Wall

tigating the seismic response of geosynthetic-reinforced soil retaining walls to actual


earthquake records using a faster computer platform.
The fundamental frequency of vibration for a two-dimensional, linear elastic medium
of width B and height H contained by two, rigid vertical boundaries and a rigid base and
subject to horizontal base excitation is given by Wu (1994) as:

f 11 = 1
4H
Gρ 1 + 1 −2 vHB 2
(3)

where: f11 = frequency in Hz corresponding to the first mode shape of the medium in
both the horizontal and vertical directions; G = shear modulus; ρ = density; and v = Pois-
son’s ratio of the elastic medium. In the limit of an infinitely wide medium (B→∞),
Equation 3 becomes the well known expression for the fundamental frequency of a one-
dimensional elastic medium with height H. For the one-dimensional case with soil layer
height H = 6 m and the elastic soil properties described in Section 5.2, Equation 3 gives
f11 = 3.32 Hz. This value is close to the frequency of the harmonic input acceleration
record (f = 3 Hz) used in most numerical simulations in the current investigation. Hence,
the large deformations reported in Section 6 may be expected because the aspect ratio
(H/B) of the numerical grids used in the simulations is small (i.e. Equation 3 with H/B
= 0.15 gives f11 = 3.43 Hz.).
Richardson and Lee (1975) proposed that the fundamental period, T1 , of reinforced
soil walls constructed with steel strip reinforcement can be estimated empirically using
the following nondimensional equation:

T 1 = 0.020H to 0.033H (4)

where H is the height of the wall in metres and T1 is in seconds. According to Equation
4, the expected fundamental frequency of most wall models in the current study ranges
from 5.1 to 8.3 Hz, which is significantly greater than the applied harmonic base fre-
quency of f = 3 Hz. A possible explanation for the difference in predicted fundamental
frequencies using Equations 3 and 4 is that the empirical relationship by Richardson and
Lee (1975) is applicable to walls retaining a relatively narrow soil volume beyond the
reinforced zone. The influence of the choice of base excitation frequency on the seismic
response of the 6 m high wall in the current study and implications of the frequency con-
tent on numerical modeling results in general are discussed further in Section 6.3.
The total duration of the input excitation was limited to 6 seconds (Figure 4) in order
to minimize computation time. Only a horizontal acceleration record was applied
whereas, in an actual seismic event, vertical acceleration components may be expected.
Vertical accelerations are typically ignored in pseudostatic design of wall structures in
North America and Japan (Bathurst and Alfaro 1997). It is important to note that the
harmonic ground motion used in the current study is much more aggressive on the system
response than a true earthquake record with the same peak acceleration and comparable
peak velocity, duration, and predominant frequency. Hence, the numerical analyses re-
sults reported in the current study are interpreted largely in relative and qualitative terms.
Elgamal et al. (1996) have proposed that for conventional reinforced concrete cantile-
ver wall-backfill systems less than 10 m in height, and subject to typical situations of
seismic excitation, a viscous damping ratio of 5% is conservative. Hence, a damping ra-

GEOSYNTHETICS INTERNATIONAL S 1998, VOL. 5, NOS. 1-2 139

Downloaded by [ Purdue Univ Lib TSS] on [19/09/16]. Copyright © ICE Publishing, all rights reserved.
BATHURST AND HATAMI D Seismic Response Analysis of a Geosynthetic-Reinforced Wall

tio of c = 5% was chosen for the reference case in the parametric analyses reported in the
current study. However, a number of analyses were carried out to examine the influence
of damping ratio on the system response by using values of c = 10 and 20% (Table 2). In
all of the simulations, damping was applied to both the soil and facing panel regions;
however, the influence of the relatively small panel mass was considered negligible.
The program was executed on a 200 MHz personal computer. Computer runs required
approximately 3 to 6 hours for construction and prop release, and approximately 12
hours for the 6 second base acceleration loading record.

6 RESULTS OF SEISMIC ANALYSES

6.1 Influence of Reinforcement Stiffness, Reinforcement Length, and Toe


Restraint Condition

6.1.1 Wall Displacements

Example horizontal displacement histories of the wall facing for typical simulations
are presented in Figure 5. The datum for horizontal wall displacements, ∆x, was taken
at the end of construction following prop release. Results are shown for the two cases
of fixed- and sliding-base conditions with L/H = 1 and reinforcement stiffness J = 1,000
kN/m. The displacement histories show that the permanent outward displacement of
the wall increases monotonically with time during application of the input acceleration.
The amplitudes of motion are small compared to the magnitude of the permanent out-
ward displacement at the end of seismic shaking. The qualitative displacement-time
features described for Figure 5 are typical results for all of the simulation runs.
Wall displacement profiles predicted at the end of the excitation period are shown in
Figure 6 for the range of conditions summarized in Table 1. The maximum displace-
ment at the top of the wall is greater for the fixed-base condition than for the sliding-
base condition due to the larger magnitude of tilting that occurs for the fixed-toe
condition. For a given base condition (fixed or sliding), the total wall displacements
diminish with increasing reinforcement stiffness. Similarly, for a given base condition
and reinforcement stiffness, there is less total wall displacement for L/H = 1 compared
to configurations with L/H = 0.7.
The variation of maximum lateral displacement at the top of the facing panel with
reinforcement stiffness, reinforcement length, and base condition is summarized in Fig-
ure 7. At the end of construction, there is essentially no influence of reinforcement
length on the maximum wall displacements for the fixed-base condition. For the com-
parable cases with a sliding base, there is a small effect of reinforcement length on the
displacements with the maximum wall displacement being slightly greater for L/H =
0.7 compared to the case with L/H = 1. However, the major influence on wall displace-
ments at the end of construction is the reinforcement stiffness, particularly for stiffness
values of 2,000 kN/m or less.
Figure 7 shows that permanent maximum wall displacements are significantly larger
after base shaking. For the fixed-base condition, the displacements at the top of the wall
are influenced by both the reinforcement stiffness and the volume of the reinforced soil
zone for the range of parameters investigated. The amount of displacement was less

140 GEOSYNTHETICS INTERNATIONAL S 1998, VOL. 5, NOS. 1-2

Downloaded by [ Purdue Univ Lib TSS] on [19/09/16]. Copyright © ICE Publishing, all rights reserved.
BATHURST AND HATAMI D Seismic Response Analysis of a Geosynthetic-Reinforced Wall

(a)

Normalized horizontal displacement, Dx / H


0.10
J = 1,000 kN/m Soil surface Top
0.09 L/H = 1 6
0.08 Fixed base
0.07 Reinforcement layer 5
0.06 ∆x
4
0.05
0.04 3
H
0.03
2
0.02
0.01 1 Bottom
0.00
Foundation surface
---0.01
(b) 0.10
Normalized horizontal displacement, Dx / H

0.09 J = 1,000 kN/m


L/H = 1
0.08 Sliding base Reinforcement layer
0.07
0.06 ∆x Top
Soil surface 6
0.05 5
4
0.04 H 3
2
0.03
1 Bottom
0.02
0.01
0.00
Foundation surface
---0.01
0 1 2 3 4 5 6
Time (s)
Figure 5. History of normalized horizontal wall displacements, ∆x/H, at selected
elevations: (a) fixed-base condition; (b) sliding-base condition.
Note: Datum taken at the end of construction following external prop release.

with L/H = 1 compared to shorter reinforcement length models (i.e. L/H = 0.7). For slid-
ing-wall cases, the influence of reinforcement length and reinforcement stiffness on
wall displacements was significantly less than for the fixed-base condition.
In summary, the plots in Figures 6 and 7 show that the toe restraint condition has a
greater influence on the magnitude of maximum wall displacements for the given input
acceleration record than reinforcement length and reinforcement stiffness. Wall dis-
placements are greater for the fixed-base condition compared to the sliding-base condi-

GEOSYNTHETICS INTERNATIONAL S 1998, VOL. 5, NOS. 1-2 141

Downloaded by [ Purdue Univ Lib TSS] on [19/09/16]. Copyright © ICE Publishing, all rights reserved.
BATHURST AND HATAMI D Seismic Response Analysis of a Geosynthetic-Reinforced Wall

Relative Total
(a) 6
J (kN/m)
5 9,000
69,000

Elevation (m)
4 J (kN/m)
J (kN/m) 500
3 500 1,000
1,000 2,000
2 2,000 9,000
69,000
1
0
(b) 6
J (kN/m)
5 500
1,000
Elevation (m)

4 2,000
9,000
3 69,000 J (kN/m)
500
2 1,000
2,000
1 9,000
69,000
0
(c) 6
5
Elevation (m)

4
J (kN/m)
3 500
1,000
2 2,000
9,000
1
69,000
0
(d) 6
5
Elevation (m)

4
J (kN/m)
3 500
1,000
2 2,000
9,000
1 69,000

0
0.00 0.02 0.04 0.06 0.08 0.10
Normalized horizontal displacement, ∆x / H

Figure 6. Total and relative normalized wall displacement at the end of seismic shaking
for walls with different reinforcement stiffness: (a) L/H = 0.7, sliding-base condition;
(b) L/H = 1, sliding-base condition; (c) L/H = 0.7, fixed-base condition; (d) L/H = 1,
fixed-base condition.
Note: Datum taken at the end of construction following external prop release.

142 GEOSYNTHETICS INTERNATIONAL S 1998, VOL. 5, NOS. 1-2

Downloaded by [ Purdue Univ Lib TSS] on [19/09/16]. Copyright © ICE Publishing, all rights reserved.
BATHURST AND HATAMI D Seismic Response Analysis of a Geosynthetic-Reinforced Wall

0.7
End of
0.6 seismic Fixed base
Maximum displacement (m) shaking Sliding base
0.5

0.4

0.3
L/H = 0.7, sliding base
0.2 End of L/H = 1.0, sliding base
construction
L/H = 0.7, fixed base
0.1 L/H = 1.0, fixed base

0.0
100 1000 10000 100000
Reinforcement stiffness, J (kN/m)

Figure 7. Influence of the reinforcement stiffness, J, reinforcement length, L, and base


condition on the maximum wall displacements at the end of construction and after
seismic shaking.

tion when all other parameter values are equal. However, the influence of the toe restraint
condition on wall displacements reduces as the stiffness of the reinforcement increases.

6.1.2 Reinforcement Loads

A cross section showing the true-scale deformation of the reinforced zone in an exam-
ple simulated wall after base shaking is shown in Figure 8. Superimposed on Figure 8
are bar graphs of reinforcement loads at the same point in time. In all of the tests, rein-
forcement loads were greatest at the connections after base shaking. This trend can be
attributed to the progressive downward movement of the reinforced soil zone relative
to the continuous wall panel during base excitation and the pinned reinforcement-wall
connection detail adopted in the model.
An example of normalized axial load histories in the reinforcement layers at the con-
nections is shown in Figure 9. The results are shown for the fixed- and sliding-base condi-
tions for walls with J = 1,000 kN/m and L/H = 1. Connection loads, Tc , can be seen to
accumulate with time during shaking of the base and this qualitative feature was ob-
served in all simulation runs. However, the distribution and magnitude of reinforcement
loads over the entire time record was very different between fixed- and sliding-base
cases in the current study.
The distributions and magnitudes of the maximum recorded load, Tmax , in each rein-
forcement layer at the end of construction (static loading) and during base shaking (dy-
namic loading) are plotted in Figure 10. Each maximum load value for a reinforcement
layer corresponds to the maximum tensile load recorded along the entire length of that

GEOSYNTHETICS INTERNATIONAL S 1998, VOL. 5, NOS. 1-2 143

Downloaded by [ Purdue Univ Lib TSS] on [19/09/16]. Copyright © ICE Publishing, all rights reserved.
BATHURST AND HATAMI D Seismic Response Analysis of a Geosynthetic-Reinforced Wall

Grid boundary
B = 40 m

H=6m

L = 4.2 m

Figure 8. True-scale deformed numerical grid and the reinforcement loads at the end of
seismic shaking (t = 6 s) for the fixed-base condition with L/H = 0.7 and J = 2,000 kN/m.
Note: Maximum reinforcement load of 65 kN/m in bar graphs occurs in the second layer from the
bottom.

layer. The maximum dynamic load values generally occurred at or close to the connec-
tions at a time corresponding to the end of the excitation record (Figure 10).
The static load distributions for all of the reinforcement stiffness cases show a trend
toward increasing reinforcement loads with increasing reinforcement stiffness for the
fixed-base condition (Figures 10a and 10b). Also shown in Figure 10 are the theoretical
values for the normalized static load in each reinforcement layer using limit-equilibri-
um methods based on Rankine (Ka = f(Ô)) and Coulomb earth pressure theory (Ka = f(Ô,
δ)), and a contributory area approach to distribute reinforcement loads. For the fixed-
base condition, the linear distribution values from theory contain the range of predicted
reinforcement loads but do not capture the general trend which is reasonably uniform.
The limit-equilibrium solutions under-predict the reinforcement loads close to the top
of the wall and over-predict the magnitude of loads toward the base of the wall. These
qualitative observations are consistent with the results reported by Rowe and Ho (1997)
who investigated the influence of reinforcement stiffness on the distribution and magni-
tude of reinforcement loads under static conditions.
The static load distributions from numerical analyses with a sliding-base condition
(Figures 10c and 10d) are more dispersed than for the fixed case. The trends of the cal-
culated static load values with elevation from numerical analyses for the sliding-base
cases are better captured using the linear distributions from Rankine and Coulomb earth
pressure theories than for the fixed-base cases. However, the magnitudes of the rein-
forcement load using earth pressure theories are consistently lower than predicted val-
ues from numerical simulations.
There is no significant influence of reinforcement length on the magnitude and dis-
tribution of reinforcement loads under static loading apparent in the data in Figure 10.

144 GEOSYNTHETICS INTERNATIONAL S 1998, VOL. 5, NOS. 1-2

Downloaded by [ Purdue Univ Lib TSS] on [19/09/16]. Copyright © ICE Publishing, all rights reserved.
BATHURST AND HATAMI D Seismic Response Analysis of a Geosynthetic-Reinforced Wall

(a) 0.5
J = 1,000 kN/m Reinforcement layer
L/H = 1 3
0.4
4
2
0.3 Bottom 1
Tc / Ty
0.2 5

0.1 6
Top
0.0
(b) 0.5
J = 1,000 kN/m Reinforcement layer
L/H = 1
0.4 1
Bottom
0.3
2
Tc / Ty

3
0.2 4
5
0.1
6
Top
0.0
0 1 2 3 4 5 6
Time (s)
Figure 9. History of the normalized connection loads in the reinforcement layers:
(a) fixed-base condition; (b) sliding-base condition.
Note: Tc is the connection load and Ty = 200 kN/m is the yield strength of the reinforcement.

The maximum reinforcement load distributions recorded during dynamic loading are
highlighted by the hatched regions in Figure 10. In all cases, the reinforcement loads
were larger under dynamic loading than the calculated loads for the end-of-construction
(i.e. static) condition. The increase in reinforcement load in any layer increases with
reinforcement stiffness.
For the fixed-base condition (Figures 10a and 10b), the trend of the reinforcement
load can be seen to increase with depth below the wall crest for J = 9,000 and 69,000
kN/m while the trend of the reinforcement loads for walls with lower stiffness reinforce-
ment is more uniform with depth. Moreover, there is essentially no influence of rein-
forcement length on the magnitude of the maximum loads recorded for the
fixed-foundation cases except for the stiffest reinforcement case with L/H = 1 which
gives consistently higher reinforcement loads than the otherwise identical configura-
tion with L/H = 0.7.
The trends of the reinforcement load with elevation under dynamic loading for slid-
ing-base cases (Figures 10c and 10d) are qualitatively different from the corresponding

GEOSYNTHETICS INTERNATIONAL S 1998, VOL. 5, NOS. 1-2 145

Downloaded by [ Purdue Univ Lib TSS] on [19/09/16]. Copyright © ICE Publishing, all rights reserved.
BATHURST AND HATAMI D Seismic Response Analysis of a Geosynthetic-Reinforced Wall

(a) (1) Ka = f(f), (2) Ka = f(f, d) (b) (1) Ka = f(f), (2) Ka = f(f, d)
6 6
6, top reinforcement layer 6, top reinforcement layer
5 5
5 J (kN/m) 5 J (kN/m)
Elevation (m)

4 500 4 500
4 1,000 4 1,000
(2) 3 (2)
3 (1) 2,000 (1) 2,000
3 9,000 3 9,000
2 69,000 2 69,000
2
2
1 Dynamic 1 Dynamic 1
1
End of construction (static) End of construction (static)
0 0
0.0 0.2 0.4 0.6 0.8 0.0 0.2 0.4 0.6 0.8

(c) (d)
(1) Ka = f(f), (2) Ka = f(f, d) (1) Ka = f(f), (2) Ka = f(f, d)
6 6
6, top reinforcement layer 6, top reinforcement layer
5 5
5 J (kN/m) 5 J (kN/m)
Elevation (m)

4 500 4 (2)
500
(2)
4 1,000 4 1,000
3 2,000 3 (1) 2,000
(1) 3 9,000 3 9,000
2 69,000 2 69,000
2 2
1 Dynamic 1 Dynamic
1 1
End of construction (static) End of construction (static)
0 0
0.0 0.2 0.4 0.6 0.8 0.0 0.2 0.4 0.6 0.8
Normalized maximum reinforcement load, Tmax / Ty

Figure 10. Influence of the reinforcement stiffness, J, reinforcement length, L, and base
condition on the maximum load in each reinforcement layer: (a) L/H = 0.7, fixed-base
condition; (b) L/H = 1, fixed-base condition; (c) L/H = 0.7, sliding-base condition; (d) L/H
= 1, sliding-base condition.
Note: Tmax = maximum tensile load recorded along the entire length of the reinforcement layer and Ty =
200 kN/m is the yield strength of the reinforcement.

results for the fixed cases because dynamic loads are observed to increase in magnitude
with depth below the crest of the wall for all reinforcement stiffness values.
Figure 11 shows the normalized dynamic load increment , ∆T, recorded in all simula-
tions. Dynamic load increment values were calculated by subtracting from the maxi-
mum load, Tmax , used to generate the maximum dynamic load curves in Figure 10, the
corresponding maximum initial static load values and then normalizing the dynamic
load increment value with the yield strength of the reinforcement, Ty . The predicted
dynamic load increments using the current AASHTO (1996) method with a peak hori-
zontal ground acceleration of 0.2g are also shown in Figure 11.
For the fixed-base condition (Figures 11a and 11b), the empirical AASHTO (1996)
method underestimates the magnitude of dynamic load increments, ∆T. The magnitude

146 GEOSYNTHETICS INTERNATIONAL S 1998, VOL. 5, NOS. 1-2

Downloaded by [ Purdue Univ Lib TSS] on [19/09/16]. Copyright © ICE Publishing, all rights reserved.
BATHURST AND HATAMI D Seismic Response Analysis of a Geosynthetic-Reinforced Wall

(a) (b)
6 6
6, top reinforcement layer 6, top reinforcement layer
5 5
5 J (kN/m) 5 J (kN/m)
Elevation (m)

4 500 4 500
4 1,000 4 1,000
3 2,000 3 2,000
3 9,000 3 9,000
2 69,000 2 69,000
2 2
1 1
(1) 1 (1) 1
0 0
0.0 0.2 0.4 0.6 0.8 0.0 0.2 0.4 0.6 0.8

(c) (d)
6 6
6, top reinforcement layer 6, top reinforcement layer
5 5
5 J (kN/m) 5 J (kN/m)
Elevation (m)

4 500 4 500
4 1,000 4 1,000
3 2,000 3 2,000
3 9,000 3 9,000
2 69,000 2 69,000
2 2
1 1
(1) 1 (1) 1
0 0
0.0 0.2 0.4 0.6 0.8 0.0 0.2 0.4 0.6 0.8
Normalized dynamic load increment, ∆T / Ty
Figure 11. Influence of the reinforcement stiffness, J, reinforcement length, L, and base
condition on the reinforcement dynamic load increment, ∆T : (a) L/H = 0.7, fixed-base
condition; (b) L/H=1, fixed-base condition; (c) L/H = 0.7, sliding-base condition; (d) L/H
= 1, sliding-base condition.
Note: (1) = AASHTO (1996) method.

of the underestimation increases with stiffness of the reinforcement. In addition, the


trend of the linear increase in ∆T with increasing depth below the crest of the wall using
the empirical AASHTO method does not reflect the trend of the data except for the stif-
fest reinforcement cases (i.e. J = 9,000 and 69,000 kN/m).
For the sliding-base condition (Figures 11c and 11d), the empirical AASHTO (1996)
method predicts values of ∆T that are in the range of values for the lowest reinforcement
stiffness cases but increasingly underestimates the magnitude of ∆T as reinforcement
stiffness becomes larger. However, the trend of the data for the magnitude of the dynam-
ic load increment, ∆T, with depth below the crest of the wall can be argued to be qualita-
tively similar to the empirical AASHTO results.
The influence of reinforcement elevation, base condition, and reinforcement stiffness
is summarized in Figure 12. The following observations can be made on the data corre-

GEOSYNTHETICS INTERNATIONAL S 1998, VOL. 5, NOS. 1-2 147

Downloaded by [ Purdue Univ Lib TSS] on [19/09/16]. Copyright © ICE Publishing, all rights reserved.
BATHURST AND HATAMI D Seismic Response Analysis of a Geosynthetic-Reinforced Wall

(a) 1.0
0.8
Tmax / Ty Bottom
1
0.6 2
0.4 3
4
5
0.2 6
Top
0.0
(b) 1.0
Bottom
0.8 1
Tmax / Ty

0.6 2
3
0.4 4
5
0.2 6
Top
0.0
(c) 1.0
0.8 Bottom
Tmax / Ty

1
0.6 2
3
0.4 4
5
0.2 6
Top
0.0
(d) 1.0
0.8 Bottom
1
Tmax / Ty

0.6 2
3
0.4 4
5
0.2 6
Top
0.0
100 1000 10000 100000
Reinforcement stiffness, J (kN/m)
Figure 12. Variation of the normalized maximum tensile load, Tmax / Ty , in the
reinforcement layers during seismic shaking versus reinforcement stiffness, J,
reinforcement length, L, and base condition: (a) L/H = 0.7, fixed-base condition; (b) L/H
= 1, fixed-base condition; (c) L/H = 0.7, sliding-base condition; (d) L/H = 1, sliding-base
condition.

sponding to the end of base shaking using the applied harmonic input acceleration re-
cord: (i) reinforcement loads increase in magnitude from top to bottom of the wall; (ii)
reinforcement loads are more sensitive to the magnitude of the reinforcement stiffness
at lower reinforcement elevations than at the top of the wall; and (iii) reinforcement
loads at the bottom of the wall are more sensitive to reinforcement stiffness for the
fixed-base condition than for the sliding-base condition.
However, as demonstrated in Sections 6.2 and 6.3, quantitative values from numeri-
cal simulations (such as reinforcement loads) will vary widely as a result of characteris-

148 GEOSYNTHETICS INTERNATIONAL S 1998, VOL. 5, NOS. 1-2

Downloaded by [ Purdue Univ Lib TSS] on [19/09/16]. Copyright © ICE Publishing, all rights reserved.
BATHURST AND HATAMI D Seismic Response Analysis of a Geosynthetic-Reinforced Wall

tics of the base input acceleration record, fundamental frequency of the model, and
height to width ratio of the excited structure. In addition, the frequency of the harmonic
ground motion, which is just below the fundamental frequency of the reference struc-
tures, can be expected to generate large reinforcement loads.
The observation that the empirical AASHTO (1996) method underestimates the mag-
nitude of the dynamic load increments, ∆T, suggests that this empirical method may be
unsafe for the design of walls that are excited close to the fundamental frequency of the
structure during an earthquake.

6.1.3 Distribution of Backfill Accelerations

The distribution and magnitude of peak accelerations in the backfill soil is of interest
in pseudostatic seismic design methods because a coherent distribution of the ground
acceleration is considered to be responsible for the additional destabilizing force that
must be resisted by reinforced structures during a seismic event.
The acceleration responses in the soil at different elevations were filtered using an
elliptic, low-pass filter with a cut-off frequency, flim = 10 Hz (Bellanger 1989). This filter
function was applied to all acceleration response records to exclude the spurious, high-
frequency acceleration peaks associated with the reflection of waves in the simulation
runs. The significant portion of seismic energy during actual earthquakes is also im-
parted at frequencies below 10 Hz (Elgamal et al. 1996). Hence, the upper cut-off fre-
quency of flim = 10 Hz was used to capture a significant portion of the response frequency
content in the numerical simulations.
The distribution and magnitude of filtered accelerations recorded in the soil immedi-
ately behind the wall-soil interface column (i.e. close to the back of the facing panel)
are summarized in Figure 13. Acceleration amplification over the depth of the backfill
is noticeably larger at the soil surface for the fixed-base condition as compared to the
sliding-base condition. The large amount of acceleration amplification at the surface
of the backfill soil may be partly attributed to the assumption of a cohesionless backfill
and the relatively low damping value used in most analyses. The generally larger accel-
eration values for the fixed-base condition are also apparent in Figure 14 where the
magnitude of the average amplification value is plotted against reinforcement stiffness.
The average acceleration amplification value is in the range of 2.0 to 2.8 and is relative-
ly insensitive to the range of reinforcement stiffness values used in the current study.
Analyses of the data confirmed that the back-calculated values of the amplification fac-
tor were insensitive to the cut-off frequency above 10 Hz.

6.1.4 Failure Zones

Figure 15 shows typical plots of shear zones within the reinforced soil zone and in
the retained soil for typical fixed- and sliding-base conditions. In this numerical study,
there was no evidence of a well-defined failure surface intersecting all reinforcement
layers as may be expected from conventional tied-back wedge and nonlinear slip surface
methods of analysis (Bathurst and Alfaro 1997). This was true even for models with
the lowest reinforcement stiffness (i.e. J = 500 kN/m). Rather, the reinforced soil zone
acted as a parallel-sided monolithic mass. Further investigation is required to determine
if the pattern of internal failure will change with greater reinforcement spacings.

GEOSYNTHETICS INTERNATIONAL S 1998, VOL. 5, NOS. 1-2 149

Downloaded by [ Purdue Univ Lib TSS] on [19/09/16]. Copyright © ICE Publishing, all rights reserved.
BATHURST AND HATAMI D Seismic Response Analysis of a Geosynthetic-Reinforced Wall

(a) 6.0
5.0

Elevation (m)
4.0
J (kN/m)
3.0 Peak 69,000
foundation 9,000
2.0
acceleration 2,000
1.0 1,000
500
0.0
(b)
6.0
5.0
Elevation (m)

4.0 J (kN/m)
3.0 Peak 69,000
foundation 9,000
2.0 acceleration 2,000
1.0 1,000
500
0.0
0.0 0.2 0.4 0.6 0.8 1.0 1.2
Acceleration (g)
Figure 13. Distribution of the peak horizontal accelerations recorded in the reinforced
soil zone for different reinforcement stiffness values, J, and base conditions (L/H = 0.7):
(a) fixed-base condition; (b) sliding-base condition.
Note: Acceleration response filter cut-off frequency, flim = 10 Hz.
Mean base acceleration amplification

3.0
Fixed base
2.5

2.0
Sliding base
1.5

1.0

0.5

0.0
100 1000 10000 100000
Reinforcement stiffness, J (kN/m)
Figure 14. Variation of the mean base acceleration amplification in the reinforced soil
zone versus the reinforcement stiffness, J, and the base condition.
Notes: Acceleration response filter cut-off frequency, flim = 10 Hz; damping ratio c = 5%.

150 GEOSYNTHETICS INTERNATIONAL S 1998, VOL. 5, NOS. 1-2

Downloaded by [ Purdue Univ Lib TSS] on [19/09/16]. Copyright © ICE Publishing, all rights reserved.
BATHURST AND HATAMI D Seismic Response Analysis of a Geosynthetic-Reinforced Wall

Grid boundary
(a) 40 m
12 m

6.0 m 31_
23_

20_

4.2 m
Grid boundary
(b) 40 m
13 m

6.0 m

31_
23_
7_
4.2 m
Figure 15. Shear zones at t = 6 s, with L/H = 0.7 and J = 2,000 kN/m: (a) fixed-base
condition; (b) sliding-base condition.
Note: Dark shading indicates relatively large shear strains.

Large shear strains were recorded at the wall-soil interface and at the reinforced re-
tained soil interface. The failure volume in each simulation can be approximated by a
bilinear wedge with a break point at the back of the reinforced soil zone. The break point
was observed to be at a higher elevation for the fixed-base condition when compared
to the sliding-base condition.
Also shown in Figure 15 are the linear failure surfaces in the retained soil that are pre-
dicted from solutions for slip surface orientation using Mononobe-Okabe earth pressure
theory (Okabe 1924; Zarrabi 1979; Bathurst and Alfaro 1997). Orientations of 23 and
31_ from the horizontal correspond to computed mean wedge accelerations of approxi-
mately 0.5g and 0.4g, respectively, and are in reasonably good agreement with shear
zone boundaries. Hence, pseudostatic equilibrium methods may be useful to estimate
minimum widths for numerical grids if the influence of yielded soil zones on the wall
response is to be captured in numerical simulations.

GEOSYNTHETICS INTERNATIONAL S 1998, VOL. 5, NOS. 1-2 151

Downloaded by [ Purdue Univ Lib TSS] on [19/09/16]. Copyright © ICE Publishing, all rights reserved.
BATHURST AND HATAMI D Seismic Response Analysis of a Geosynthetic-Reinforced Wall

6.2 Influence of Boundaries and Damping on Numerical Results

6.2.1 Far-End Truncated Boundary Condition

The reference, far-end truncated boundary condition in the current study is a free-field
boundary condition applied at the vertical left and right edges of the grid (Figure 3).
This boundary condition simulates an infinitely wide domain with respect to elastic
wave transmission. Two less complicated boundary conditions were also investigated
and applied at the right hand boundary: (i) rigid, forced boundary condition - a rigid
vertical boundary with the base acceleration function applied to all grid points from the
foundation elevation to the soil surface; (ii) rigid, stationary boundary - a rigid vertical
boundary that was fixed horizontally during the entire simulation.
The influence of boundary condition is illustrated in Figure 16 for walls with a fixed-
base condition. The data show that for the three conditions examined, the free-field

(a) 6

5 Rigid, stationary
Elevation (m)

4 Rigid, forced
3 Free-field
2 J = 2000 kN/m
L/H = 0.7
1 B = 40 m
Fixed base
0
0.00 0.02 0.04 0.06 0.08 0.10
Normalized horizontal displacement, ∆x / H
(b) 6

5 J = 2000 kN/m Free-field


L/H = 0.7
Elevation (m)

4 B = 40 m
Fixed base
3

2 Rigid, forced

1 Rigid, stationary

0
0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35
Normalized connection load, Tc / Ty
Figure 16. Influence of the model far-end condition on the wall response with the
fixed-base condition: (a) normalized wall displacements; (b) normalized maximum
reinforcement connection loads.
Note: Displacement datum taken at the end of construction following external prop release.

152 GEOSYNTHETICS INTERNATIONAL S 1998, VOL. 5, NOS. 1-2

Downloaded by [ Purdue Univ Lib TSS] on [19/09/16]. Copyright © ICE Publishing, all rights reserved.
BATHURST AND HATAMI D Seismic Response Analysis of a Geosynthetic-Reinforced Wall

condition results in the largest wall displacements and the largest connection loads at
the end of seismic shaking. The rigid stationary condition results in the least wall dis-
placement and reinforcement loads. The latter result is considered to be due to the hor-
izontal constraint imposed on the backfill soil volume at the right hand boundary. The
reduced maximum reinforcement load at the base of the wall with a fixed-forced
boundary compared to the free-field case is consistent with the trend of the results from
dynamic finite element modeling work reported by Richardson and Lee (1975) who
investigated boundary effects for steel strip-reinforced walls with thin-wall metallic
facings and a fixed base.

6.2.2 Model Width

The influence of width B of numerical grids on dynamic response was examined us-
ing models with J = 2000 kN/m, L/H = 0.7 and a stationary rigid truncated far-end
boundary condition. Figure 17 shows that the magnitude of lateral wall displacements

B/H = 1.3 2.5 4.2 6.7


B = 7.5 m 15 m 25 m 40 m
(a) 6

5
Elevation (m)

1 J = 2000 kN/m
L/H = 0.7
0
(b) 6

5
Elevation (m)

1 J = 2000 kN/m
L/H = 0.7
0
0.00 0.02 0.04 0.06 0.08
Normalized horizontal displacement, ∆x / H
Figure 17. Influence of the model width, B, on the normalized wall displacements:
(a) fixed-base condition; (b) sliding-base condition.
Notes: Stationary rigid far-end boundary condition; displacement datum taken at the end of construction
following external prop release.

GEOSYNTHETICS INTERNATIONAL S 1998, VOL. 5, NOS. 1-2 153

Downloaded by [ Purdue Univ Lib TSS] on [19/09/16]. Copyright © ICE Publishing, all rights reserved.
BATHURST AND HATAMI D Seismic Response Analysis of a Geosynthetic-Reinforced Wall

is greatly influenced by the volume of soil behind the reinforced soil zone for the refer-
ence base input acceleration record used in the current study. The larger the mass of
backfill soil the greater the wall displacements at the end of seismic shaking. The corre-
sponding maximum axial loads in the reinforcement at different elevations are plotted
in Figure 18. Not surprisingly, the maximum axial loads calculated for all reinforce-
ment layers (except the top layer) also increase in value with increasing width B. The
loads in the bottom reinforcement layer for each base condition vary by approximately
a factor of two over the range of B values examined. Nevertheless, the effect of parame-
ter B on calculated displacements and reinforcement loads can be seen to diminish as
the width of the soil model increases (e.g. compare cases with B = 25 and 40 m in Fig-
ures 17 and 18).
The magnitude of reinforced wall displacements and reinforcement loads increases
with the width of the backfill model (for the same input motion) as a direct result of the
increase in the mass of the model and the width of the excitation boundary at the founda-
tion level. The significant dependence of dynamic response of the wall on the width of
the backfill model would not be observed if the excitation boundary of the retaining
wall system under horizontal ground motion was limited to the facing panel and the
truncated boundary of the model was placed at sufficiently large distance from the fac-
ing panel. It follows that the size of the backfill model may be a dominating factor in
the response analysis of the reinforced retaining walls subjected to a prescribed ground
motion. For the range of parameters investigated in this study, the model width, B, was
more important than the type of far-end truncated boundary condition employed in nu-
merical simulations.
The shear plots in Figure 15 suggest that a width B = 15 m (or ratio B/H = 2.5 ) is
sufficient to capture the influence of the yielded zone in the retained soil during seismic
shaking. If the mass of soil beyond the yielded zone is ignored, then the quantitative
results for reinforcement loads and wall displacements are very much less than those
reported earlier for the reference case B = 40 m (B/H = 6.7). However, the width of the

6
Fixed base
5 Sliding base
Elevation (m)

4 B B/H H/B
40 m 6.7 0.15
3 25 m 4.2 0.24
15 m 2.5 0.40
2 7.5 m 1.3 0.80

0
0.0 0.1 0.2 0.3 0.4 0.5
Tc / Ty
Figure 18. Influence of model width, B, and base condition on normalized maximum
reinforcement connection loads (J = 2,000 kN/m, L/H = 0.7).
Note: Stationary rigid far-end boundary condition.

154 GEOSYNTHETICS INTERNATIONAL S 1998, VOL. 5, NOS. 1-2

Downloaded by [ Purdue Univ Lib TSS] on [19/09/16]. Copyright © ICE Publishing, all rights reserved.
BATHURST AND HATAMI D Seismic Response Analysis of a Geosynthetic-Reinforced Wall

yielded zone can be expected to increase with base acceleration according to Mono-
nobe-Okabe theory. Hence, more work remains to be done to develop rules to select a
representative volume of retained soil in combination with an appropriate far-end trun-
cated boundary condition.

6.2.3 Damping

A dynamic shear damping ratio of c = 5% was selected as the reference value for the
soil in the current study (i.e. Table 1). To simplify numerical modeling this value was
assigned to all material zones except the cable elements. In reality, dynamic damping
of a soil mass significantly increases with the amplitude of vibration and a reduction
in effective confining pressure. For example, Saxena et al. (1988) proposed the fol-
lowing equation for small strain dynamic shear damping of uncemented sands ex-
pressed in percent:

Pσ 
−0.38

ξ = 9.22 0
η 0.33 (5)
a

where: σo = effective confining pressure; Pa = atmospheric pressure; and η = dynamic


shear strain in percent. Equation 5 illustrates that dynamic shear damping of the soil can
theoretically reach very high values toward the surface of the backfill where soil con-
finement stress approaches zero. Results of other studies (Kramer 1996; Ishihara 1996)
also indicate a strong dependence of the damping ratio of sands on both cyclic shear
strain amplitude and magnitude of effective confining pressure. Hence, it can be argued
that a constant value of c = 5% applied to the entire soil zone may lead to an overestima-
tion of the magnitude of model deformation and acceleration response.
The influence of damping ratio was investigated by repeating selected simulation
runs with c = 10 and 20%. The effects of damping ratio on wall displacements and rein-
forcement loads at the end of seismic shaking are shown in Figures 19a and 19b. The
lateral displacement of the wall at the end of base shaking is up to 40% less for the case
with c = 20% as compared to the reference case of c = 5%. The corresponding reduction
in maximum reinforcement loads (Figure 19b) is approximately 20%. Nevertheless, the
data show that qualitative features of model deformation and reinforcement load dis-
tributions are similar in all three cases.
The effect of damping ratio on base acceleration amplification is shown in Figure 20.
The data show that increasing the damping ratio from c = 5 to 20% can reduce the peak
acceleration at selected locations in the soil zone by up to 40%. The corresponding re-
duction in mean base acceleration amplification factor is approximately 25%. Never-
theless, qualitative trends are preserved in each case (i.e. generally increasing peak
acceleration with height above the foundation base).
It is worth noting that the influence of the magnitude of damping ratio on the dynamic
response of the models in this study is not unexpected since the frequency of the applied
base input acceleration record is close to the fundamental frequency of the reference
models. Theoretically, the magnitude of viscous damping ratio is frequency dependent.
However, an additional simulation run was carried out that showed that introducing the

GEOSYNTHETICS INTERNATIONAL S 1998, VOL. 5, NOS. 1-2 155

Downloaded by [ Purdue Univ Lib TSS] on [19/09/16]. Copyright © ICE Publishing, all rights reserved.
BATHURST AND HATAMI D Seismic Response Analysis of a Geosynthetic-Reinforced Wall

(a) 6
c = 10%
5
c = 20%

Elevation (m)
4
c = 5%
3

2 J = 2000 kN/m
L/H = 1.0
1 B = 40 m
Fixed base
0
0.00 0.02 0.04 0.06 0.08 0.10
Normalized horizontal displacement, ∆x / H
(b) 6

5
c = 5%
J = 2000 kN/m
Elevation (m)

4 L/H = 1.0
B = 40 m
3 Fixed base
c = 20%
2

1 c = 10%

0
0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35
Tc / Ty
Figure 19. Influence of the soil damping value, c , on the wall displacements and the
magnitude and distribution of the normalized peak reinforcement connection loads: (a) wall
displacements; (b) normalized maximum connection loads.
Notes: Acceleration response filter cut-off frequency, flim = 10 Hz; free-field, far-end boundary condition;
displacement datum taken at the end of construction following external prop release.

6.0 c = 20%
5.0
Peak
Elevation (m)

4.0 foundation c = 10%


acceleration
3.0
c = 5%
2.0 J = 2000 kN/m
1.0 L/H = 1
Fixed base
0.0
0.0 0.2 0.4 0.6 0.8 1.0 1.2
Acceleration (g)
Figure 20. Influence of the soil damping value, c, on the magnitude and distribution of
peak horizontal accelerations recorded in the reinforced soil zone.
Note: Acceleration response filter cut-off frequency, flim = 10 Hz.

156 GEOSYNTHETICS INTERNATIONAL S 1998, VOL. 5, NOS. 1-2

Downloaded by [ Purdue Univ Lib TSS] on [19/09/16]. Copyright © ICE Publishing, all rights reserved.
BATHURST AND HATAMI D Seismic Response Analysis of a Geosynthetic-Reinforced Wall

damping ratio at a slightly different frequency had little effect on the calculated model
displacements and reinforcement loads.
Finally, it can be noted that the greatest accelerations in Figure 20 correspond to the
near-surface soil locations. This observation is consistent with Equation 5 which can
be rearranged to show that for a given damping value, the strain amplitude will increase
with lower confining pressures.

6.3 Influence of Base Acceleration Record on Numerical Results

6.3.1 Frequency of Harmonic Input Acceleration Record

Wave propagation theory for one and two-dimensional linear elastic media discussed
in Section 5.3 suggests that the frequency of base excitation adopted in the present study
(f = 3 Hz) may be close to the fundamental frequency of the 6 m high wall. The influence
of frequency on numerical simulation results was investigated by carrying out a series
of runs with an input frequency f = 2.5, 3, 3.2, 3.4, 3.5, 3.7, 4, and 5 Hz. Figure 21a shows
the variation of maximum horizontal displacement at the crest of a wall with input fre-
quency. Numerical simulations were stable for all frequencies except 3.4 and 3.5 Hz.
Hence, the selected harmonic input acceleration record adopted as the base case in the
present study (f = 3 Hz) is close to but below the fundamental (critical) frequency of
the reference structures with B = 40 m (H/B = 0.15). The vertical lines in the hatched
zone in the figures correspond to fundamental frequencies predicted by Equation 3 for
model dimensions with height to width ratio of H/B = 0 (one dimensional), 0.15 (refer-
ence case geometry) and 0.8 (minimum width model). Equation 3 for geometries ap-
proaching the one dimensional case (H/B → 0) proved to be a good predictor of
resonance in numerical simulations (i.e. the difference between predicted fundamental
frequency for the reference case geometry with H/B = 0.15 and the one dimensional
case is very small). According to Equation 3, the fundamental frequency of two dimen-
sional elastic media increases with increasingly narrower regions (larger H/B ratios).
This effect may explain why progressively lower magnitudes of reinforcement load
were recorded with decreasing model width B in Figure 18.
Figure 21b illustrates the effect of input frequency on the magnitude of maximum re-
inforcement loads during harmonic shaking. The results of numerical analyses show
that the magnitude of reinforcement loads is influenced by input ground motion fre-
quency with a general reduction in reinforcement loads as the input frequency diverges
from the fundamental frequency of the structure.

6.3.2 Example Earthquake Input Acceleration Record

A numerical simulation run was carried out on the reference wall model with a
fixed-base condition using the first 6 seconds of the horizontal component of the 1940
El Centro earthquake acceleration record scaled to a peak acceleration of 0.2g. The
first 6 seconds of the El Centro accelerogram contains the peak ground acceleration
and the significant portion of the record. While both the harmonic record and scaled El
Centro (truncated) record have the same peak acceleration (0.2g) and the same applied
duration (6 seconds) in this study they are different with respect to ground motion char-
acteristics (i.e. frequency content, duration of strong ground motion, and peak ground

GEOSYNTHETICS INTERNATIONAL S 1998, VOL. 5, NOS. 1-2 157

Downloaded by [ Purdue Univ Lib TSS] on [19/09/16]. Copyright © ICE Publishing, all rights reserved.
BATHURST AND HATAMI D Seismic Response Analysis of a Geosynthetic-Reinforced Wall

Predicted fundamental frequency from Equation 3


(a) 3.32 3.43 4.5
0.10
H/B = 0

Normalized horizontal displacement


0.09 H/B = 0.15
0.08 Reference case
in current
0.07 investigation
(f = 3 Hz) H/B = 0.8
0.06
Dx / H

Maximum ∆x
0.05
0.04
H
0.03
0.02
B
0.01
H = 6 m, B = 40 m
0.00
(b) 0.8
Normalized maximum reinforcement load

H/B = 0
0.7 H/B = 0.15 H/B = 0.8

0.6 Bottom 1

0.5 2
Reinforcement
Tmax / Ty

layer 3
0.4 4
0.3
5
0.2 J = 69,000 kN/m
L/H = 0.7
0.1 Fixed base 6
B = 40 m Top
0.0
0 1 2 3 4 5 6
Base input frequency (Hz)
Figure 21. Influence of the frequency of the harmonic base input record on the wall
response: (a) maximum horizontal displacement recorded at the top of the wall, ∆x / H;
(b) normalized maximum reinforcement load, Tmax / Ty .
Note: Displacement datum taken at the end of construction following external prop release.

velocity). The intent here is to illustrate that qualitative features of wall displacement
and reinforcement loads under an actual input earthquake are similar to those reported
in this paper using a variable-amplitude harmonic ground motion with a single fre-
quency. Figure 22a illustrates that wall displacements are cumulative with time during
shaking and that the amplitudes of motion are small with respect to the permanent dis-
placements at the end of the simulation (compare Figure 22a to Figure 5a). Figure 22b

158 GEOSYNTHETICS INTERNATIONAL S 1998, VOL. 5, NOS. 1-2

Downloaded by [ Purdue Univ Lib TSS] on [19/09/16]. Copyright © ICE Publishing, all rights reserved.
BATHURST AND HATAMI D Seismic Response Analysis of a Geosynthetic-Reinforced Wall

(a) 0.04

Normalized horizontal displacement


J = 2000 kN/m
L/H = 0.7 Top
0.03 Fixed base

∆x
Dx / H

0.02
Mid-height
H
0.01

0.00
0 1 2 3 4 5 6
Time (s)
(b) 6
Harmonic record
5
El Centro record
Elevation (m)

4 (scaled to 0.2g)

2
J = 2000 kN/m
1 L/H = 0.7
Fixed base
0
0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35
Tc / Ty
Figure 22. Response of the wall with a fixed-base condition to the initial 6 s of the El
Centro base acceleration record scaled to 0.2g: (a) history of wall displacements;
(b) reinforcement connection loads at the end of the applied shaking record.
Notes: Scaled El Centro record from the S00E component of the Imperial Valley accelerogram
(epicentral distance = 8 km); displacement datum taken at the end of construction following external
prop release.

shows that the trend of the connection loads is the same for the El Centro and harmonic
input base acceleration cases.
In this study, no attempt was made to compare the effects of synthetic earthquake re-
cords and actual earthquake records on the seismic response of idealized reinforced soil
walls. However, the magnitude of displacements and reinforcement loads predicted us-
ing the reference harmonic record in this study are not unexpected since the frequency
of applied ground motion was close to the fundamental frequency of the models. Actual
earthquake records contain a range of significant frequencies that will typically be less
aggressive on wall response than a harmonic record with a single frequency selected
to be close to the fundamental frequency of the model structure. Hence, while harmonic

GEOSYNTHETICS INTERNATIONAL S 1998, VOL. 5, NOS. 1-2 159

Downloaded by [ Purdue Univ Lib TSS] on [19/09/16]. Copyright © ICE Publishing, all rights reserved.
BATHURST AND HATAMI D Seismic Response Analysis of a Geosynthetic-Reinforced Wall

earthquake records offer simplicity in parametric analyses of the type reported in this
paper, quantitative results may not be typical of actual earthquake response.
Finally, the difference in magnitude of reinforcement loads shown in Figure 22b
points to a limitation in the current AASHTO (1996) pseudostatic design method used
to estimate dynamic reinforcement loads under earthquake. Both excitation records
have the same peak ground acceleration and based on the AASHTO method would re-
sult in the same magnitude of dynamic load at a reinforcement layer which is clearly
not the case in Figure 22b.

7 CONCLUSIONS

The results of parametric analyses of a reinforced soil wall using program FLAC have
been reported. The program was first demonstrated to give similar results to a well-doc-
umented FEM model that was used to simulate the static load response of a 6 m high
reinforced soil wall constructed with a continuous facing panel. Parametric seismic
analyses were carried out on a similar 6 m high reinforced soil wall constructed with
two different foundation conditions, a range of geosynthetic reinforcement stiffness
values, and two different reinforcement lengths. Qualitative features of the dynamic re-
sponse of this idealized wall to a variable-amplitude harmonic base input acceleration
record are summarized below:
1. Wall displacements and reinforcement loads accumulated during base shaking. The
amplitudes of wall deformation and reinforcement load during base shaking were
small compared to permanent values calculated at the end of the input record. Simi-
lar qualitative responses were calculated when the first 6 seconds of the El Centro
earthquake accelerogram scaled to 0.2g were applied to the structure.
2. The magnitude of total wall displacement at the wall crest and relative wall dis-
placement with respect to the wall toe at the end of base excitation were less for a
reinforced wall that was free to slide at the base than for a wall that could only rotate
about the toe.
3. The magnitude of permanent wall displacement diminished with increasing rein-
forcement stiffness and increasing reinforcement length. However, for models sub-
jected to the reference harmonic base input motion at 3 Hz, the greatest influence on
the magnitude of wall displacement was the foundation condition (i.e. whether the
reinforced soil system was free to slide or was constrained to only rotate at the toe).
4. The introduction of a soil column at the back of the facing panel simulating a wall-
soil interface with reduced frictional resistance resulted in the maximum reinforce-
ment loads being generated at the connections. This observation was true at the end
of construction of the wall (static loading) and during base shaking.
5. The reference harmonic base input record resulted in additional tensile loads being
generated in reinforcement layers that were significantly larger than values result-
ing from static loading alone. The magnitude of additional dynamic-induced loads
in reinforcement layers was observed to increase with increasing stiffness of the
reinforcement.

160 GEOSYNTHETICS INTERNATIONAL S 1998, VOL. 5, NOS. 1-2

Downloaded by [ Purdue Univ Lib TSS] on [19/09/16]. Copyright © ICE Publishing, all rights reserved.
BATHURST AND HATAMI D Seismic Response Analysis of a Geosynthetic-Reinforced Wall

6. The magnitude and distribution of dynamic-induced reinforcement loads was in-


fluenced by the base condition. For walls with a sliding base, dynamic load incre-
ments increased in a generally linear fashion with depth below the crest of the wall
regardless of reinforcement stiffness. For walls with a fixed base, the dynamic load
increments did not increase with depth and were attenuated in the lower layers for
walls with a reinforcement stiffness J ≤ 2000 kN/m. For stiffer reinforcements, a
linear trend of the dynamic reinforcement loads similar to that noted for sliding-base
cases was observed. Further study is required to investigate if these trends are appli-
cable to the same models subjected to other input ground motions.
7. Horizontal ground acceleration was amplified with height above the base founda-
tion. For the reference geometry, boundary conditions and base input motion the
mean amplification factor ranged from 2 to 2.8 for models with a damping ratio of
5%. However, the magnitude of amplification was shown to be influenced by the
magnitude of damping ratio used in the numerical models.
8. The soil in the retained soil zone was observed to yield during shaking and the in-
clination of the failure surface in this region was reasonably well predicted by
Mononobe-Okabe theory.
9. There was no evidence of an interior shear surface propagating from the heel of the
facing panel and intersecting all reinforcement layers as is assumed in conventional
pseudostatic methods of analyses.
The quantitative results reported here illustrate that the predicted seismic response
of reinforced soil walls using program FLAC and the reference variable-amplitude har-
monic base input acceleration record are strongly influenced by the magnitude of
damping ratio for the soil and the type of far-end boundary condition adopted. However,
the greatest influence on wall response is the choice of base ground motion record ap-
plied to the structure. The difference between the frequency of the base excitation re-
cord and the fundamental frequency of the model is the most important factor
determining wall response to seismic excitation. The large displacements and rein-
forcement loads reported in this paper are due to the observation that the applied fre-
quency of base excitation is close to the fundamental frequency of the model with the
reference geometry. Wave propagation theory for two-dimensional elastic media shows
that a reduction in model width leads to an increase in the magnitude of the fundamental
frequency of the model. This theoretical result explains the large reduction in wall dis-
placements and reinforcement loads calculated for walls with a reduced soil mass width
but excited at the reference 3 Hz used in this study. The combination of width of model
and frequency of the applied base acceleration record has important implications to dy-
namic numerical modelling of reinforced soil walls. The use of excessively narrow re-
tained soil volumes to reduce numerical grid size and computation time may result in
an under-estimation of wall displacements and reinforcement loads. Conversely, very
wide numerical grids may generate excessively large deformations and reinforcement
loads. As a preliminary recommendation, it may be prudent to select a numerical grid
width that will capture the volume of the yielded soil in the retained soil zone predicted
by Mononobe-Okabe theory.
The results of the current investigation identify deficiencies in the current empirical
AASHTO (1996) method that is commonly used to estimate the magnitude of dynamic
load increment, ∆T, in reinforcement layers. The linear trend of the dynamic load incre-

GEOSYNTHETICS INTERNATIONAL S 1998, VOL. 5, NOS. 1-2 161

Downloaded by [ Purdue Univ Lib TSS] on [19/09/16]. Copyright © ICE Publishing, all rights reserved.
BATHURST AND HATAMI D Seismic Response Analysis of a Geosynthetic-Reinforced Wall

ment with depth below the crest predicted by the AASHTO method was not observed
in all simulations for walls constructed with stiffness values comparable to geosynthetic
reinforcement materials. In addition, the AASHTO method predicts the same magnitude
of dynamic load in a reinforcement layer for the same peak horizontal ground accelera-
tion regardless of other characteristic values of the accelerogram. In contrast, the magni-
tude of dynamic load increment, ∆T, was demonstrated to be sensitive to frequency of
base excitation and model width. It is expected that other characteristics of ground mo-
tion (e.g. frequency content, duration of strong ground motion, and peak ground veloc-
ity) will also influence the magnitude of seismic-induced loads in reinforcement layers.
In fact, the empirical AASHTO (1996) method may be unsafe for design of walls that
are excited close to the fundamental frequency of the structure during an earthquake.
Ultimately, numerical simulation results of the type demonstrated here may be used
to verify or modify current pseudostatic methods of analysis and design. However, nu-
merical results must be checked against physical measurements from large-scale shak-
ing table tests and further numerical work must be carried out to investigate response
features of idealized reinforced soil walls subjected to ground motions representing a
range of actual earthquake records.

ACKNOWLEDGMENTS

The writers wish to acknowledge the efforts of the reviewers of this paper whose com-
ments greatly improved the original submission. The funding for the work reported in
the paper was provided by the Department of National Defence (Canada).

REFERENCES

AASHTO, 1996, “Standard Specifications for Highway Bridges”, 16th Edition, Amer-
ican Association of State Highway and Transportation Officials, Washington, DC,
USA, 677 p.
Andrawes, K.Z. and Yogarajah, I., 1994, “Effects of Reinforcement Connections on the
Behaviour of Reinforced Soil Retaining Walls”, Computer Methods and Advances
in Geomechanics, Siriwardane, H.J. and Zaman, M.M., Editors, Balkema, Vol. 2,
Proceedings of the Eighth International Conference on Numerical Methods in Geo-
mechanics, Morgantown, West Virginia, USA, May 1994, pp. 1313-1318.
Bellanger, M, 1988, “Digital Processing of Signals: Theory and Practice”, Second Edi-
tion, John Wiley and Sons, New York, New York, USA, 384 p.
Bathurst, R.J., 1993, “Investigation of Footing Restraint on Stability of Large-Scale Re-
inforced Soil Wall Tests”, Proceedings of the 46th Canadian Geotechnical Confer-
ence, Regina, Manitoba, Canada, September 1993, pp. 389-398.
Bathurst, R.J. and Alfaro, M.C., 1997, “Review of Seismic Design, Analysis and Per-
formance of Geosynthetic Reinforced Walls, Slopes and Embankments”, Earth Rein-
forcement, Ochiai, H., Yasufuku, N. and Omine, K., Editors, Balkema, Vol. 2,
Proceedings of International Symposium on Earth Reinforcement, Fukuoka, Kyushu,
Japan, November 1996, Keynote Lecture, pp. 887-918.

162 GEOSYNTHETICS INTERNATIONAL S 1998, VOL. 5, NOS. 1-2

Downloaded by [ Purdue Univ Lib TSS] on [19/09/16]. Copyright © ICE Publishing, all rights reserved.
BATHURST AND HATAMI D Seismic Response Analysis of a Geosynthetic-Reinforced Wall

Bathurst, R.J. and Cai, Z., 1995, “Pseudo-static Seismic Analysis of Geosynthetic Rein-
forced Segmental Retaining Walls”, Geosynthetics International, Vol. 2, No. 5, pp.
789-832.
Bathurst, R.J., and Cai, Z., 1994, “In-isolation Cyclic Load-Extension Behavior of Two
Geogrids”, Geosynthetics International, Vol. 1, No. 1, pp. 3-17.
Cai. Z. and Bathurst, R.J., 1996, “Seismic-induced Permanent Displacement of Geo-
synthetic Reinforced Segmental Retaining Walls”, Canadian Geotechnical Journal,
Vol. 31, pp. 937-955.
Cai, Z. and Bathurst, R.J., 1995, “Seismic Response Analysis of Geosynthetic Rein-
forced Soil Segmental Retaining Walls by Finite Element Method”, Computers and
Geotechnics, Vol. 17 No. 4, pp. 523-546.
Chida, S., Minami, K. and Adachi, K., 1985, “Tests with Regard to the Stability of the
Fill Constructed by the Reinforced Earth Technique”, 13 p. (unpublished report trans-
lated from Japanese)
Cundall, P. and Board, M., 1988, “A Microcomputer Program for Modelling Large-
Strain Plasticity Problems”, Computer Methods and Advances in Geomechanics, Si-
riwardane, H.J. and Zaman, M.M., Editors, Balkema, Vol. 3, Proceedings of the Sixth
International Conference on Numerical Methods in Geomechanics, pp. 2101-2108.
Elgamal, A-W., Alampalli, S. and Van Laak, P., 1996, “Forced Vibration of Full-Scale
Wall-Backfill System”, Journal of Geotechnical Engineering, Vol. 122, No. 10, pp.
849-858.
FHWA, 1996, “Mechanically Stabilized Earth Walls and Reinforced Soil Slopes Design
and Construction Guidelines”, Federal Highway Administration (FHWA) Demon-
stration Project 82, (Elias, V. and Christopher, B.R.), Washington, DC., USA, 364 p.
FLAC, 1995, “Fast Lagrangian Analysis of Continua”, Version 3.30, Itasca Consulting
Group, Inc., 708 South Third Street, Minneapolis, Minneapolis, USA.
Ho, S.K., 1993, “A Numerical Investigation into the Behaviour of Reinforced Soil
Walls”, Ph.D. Thesis, University of Western Ontario, London, Canada, 408 p.
Ho, S.K. and Rowe, R.K., 1996, “Effect of Wall Geometry on the Behaviour of Rein-
forced Soil Walls”, Geotextiles and Geomembranes, Vol. 14, No. 10, pp. 521-541.
Ishihara, K., 1996, “Soil Behaviour in Earthquake Engineering”, Clarendon Press, Ox-
ford University, New York, USA, 350 p.
Kramer, S.L., 1996, “Geotechnical Earthquake Engineering”, Prentice-Hall, Upper
Saddle River, New Jersey, USA, 653 p.
Ling, H.I., Leshchinsky, D. and Perry, E.B., 1997a, “A New Concept on Seismic Design
of Geosynthetic-Reinforced Soil Structures: Permanent-Displacement Limit”, Earth
Reinforcement, Ochiai, H., Yasufuku, N. and Omine, K., Editors, Balkema, Vol. 1,
Proceedings of International Symposium on Earth Reinforcement, Fukuoka, Kyushu,
Japan, November 1996, pp. 797-802.
Ling, H.I., Leshchinsky, D. and Perry, E.B., 1997b, “Seismic Design and Performance
of Geosynthetic-Reinforced Soil Structures”, Géotechnique, Vol. 47, No. 5, pp.
933-952.

GEOSYNTHETICS INTERNATIONAL S 1998, VOL. 5, NOS. 1-2 163

Downloaded by [ Purdue Univ Lib TSS] on [19/09/16]. Copyright © ICE Publishing, all rights reserved.
BATHURST AND HATAMI D Seismic Response Analysis of a Geosynthetic-Reinforced Wall

Murata, O., Tateyama, M. and Tatsuoka, F., 1994, “Shaking Table Tests on a Large Geo-
synthetic-Reinforced Soil Retaining Wall Model”, Recent Case Histories of Perma-
nent Geosynthetic-Reinforced Soil Retaining Walls , Tatsuoka, F. and Leshchinsky,
D., Editors, Balkema, Proceedings of Seiken Symposium No. 11, Tokyo, Japan, No-
vember 1992, pp. 259-264.
Okabe, S., 1924, “General Theory on Earth Pressure and Seismic Stability of Retaining
Wall and Dam”, Doboku Gakkaishi - Journal of the Japan Society of Civil Engineers,
Vol. 10, No. 6, pp. 1277-1323. (translated from Japanese)
Richardson, G.N. and Lee, K.L., 1975, “Seismic Design of Reinforced Earth Walls”,
Journal of the Geotechnical Engineering Division, Vol. 101, GT2, pp. 167-188.
Rowe, R.K. and Ho, S.K., 1993, “A Review of the Behavior of Reinforced Soil Walls”,
Earth Reinforcement Practice, Ochiai, H., Hayashi, S. and Otani, J., Editors, Balke-
ma, Vol. 2, Proceedings of International Symposium on Earth Reinforcement Prac-
tice, Fukuoka, Kyushu, Japan, November 1992, pp. 801-830.
Rowe, R.K. and Ho, S.K., 1997, “Continuous Panel Reinforced Soil Walls on Rigid
Foundations”, Journal of Geotechnical and Geoenvironmental Engineering, Vol.
123, No. 10, pp. 912-920.
Sakaguchi, M., 1996, “A Study of the Seismic Behavior of Geosynthetic-Reinforced
Walls in Japan”, Geosynthetics International, Vol. 3, No. 1, pp. 13-30.
Saxena, K.S., Avramidis, A.S. and Reddy, K.R., 1988, “Dynamic Moduli and Damping
Ratios for Cemented Sands at Low Strains”, Canadian Geotechnical Journal, Vol.
25, pp. 353-368.
Segrestin, P. and Bastick, M., 1988, “Seismic Design of Reinforced Earth Retaining
Walls - the Contribution of Finite Element Analysis,” Theory and Practice of Earth
Reinforcement, Yamanouchi, T., Miura, N. and Ochiai, H., Editors, Proceedings of
the International Geotechnical Symposium on Theory and Practice of Earth Rein-
forcement, Fukuoka, Kyushu, Japan, October 1988, pp. 577-582.
Wu, G., 1994, “Dynamic Soil-Structure Interaction: Pile Foundations and Retaining
Structures”, Ph.D. Thesis, Department of Civil Engineering, University of British
Columbia, Vancouver, British Columbia, Canada, 213 p.
Yogendrakumar, M., Bathurst, R.J. and Finn, W.D. L., 1991, “Response of Reinforced
Soil Walls to Earthquake Loads”, Proceedings of the IX PanAmerican Conference on
Soil Mechanics and Foundation Engineering, Chile, August 1991, 10 p.
Zarrabi, K., 1979, “Sliding of Gravity Retaining Wall During Earthquakes Considering
Vertical Acceleration and Changing Inclination of Failure Surface”, M.Sc. Thesis,
Department of Civil Engineering, Massachusetts Institute of Technology, Cam-
bridge, Massachusetts, USA, 140 p.

NOTATIONS

Basic SI units are given in parentheses.

amax = peak base acceleration (m/s2)

164 GEOSYNTHETICS INTERNATIONAL S 1998, VOL. 5, NOS. 1-2

Downloaded by [ Purdue Univ Lib TSS] on [19/09/16]. Copyright © ICE Publishing, all rights reserved.
BATHURST AND HATAMI D Seismic Response Analysis of a Geosynthetic-Reinforced Wall

B = width of numerical grid (m)


Es = elastic modulus for soil (Pa)
f = frequency (Hz)
flim = upper cut-off frequency (Hz)
f11 = frequency of first mode shape of a two-dimensional elastic medium (Hz)
Gs = shear modulus for soil (Pa)
Gw = shear modulus for concrete panel material and foundation base (Pa)
H = height of wall (m)
J = reinforcement stiffness (N/m)
K = constant (dimensionless)
Ka = coefficient of active earth pressure (dimensionless)
Ks = bulk modulus for soil (Pa)
Kw = bulk modulus for concrete panel material and foundation base (Pa)
kb = grout-soil interface stiffness (N/m/m)
L = base width of reinforced soil zone (m)
m = constant (dimensionless)
Pa = atmospheric pressure (Pa)
sb = grout-soil bond strength (N/m)
T = reinforcement load (N/m)
Tc = reinforcement-wall connection load (N/m)
Tmax = maximum tensile load recorded along the length of a reinforcement layer
(N/m)
Ty = yield strength of reinforcement (N/m)
T1 = fundamental period (s)
t = time (s)
ü(t) = horizontal acceleration (m/s2)
α, β, ζ = constants (dimensionless)
c = damping (dimensionless)
∆T = reinforcement dynamic load increment
(∆T= Tmax (dynamic) - Tmax (static)) (N/m)
∆t = time step (s)
∆x = horizontal wall displacement following prop release at end of
construction (m)
δ = interface friction angle (Coulomb earth pressure theory) (_)
δg = grout-soil interface friction angle (_)
Ô = friction angle for soil (_)
Ôi = wall-soil interface column friction angle (used in numerical modeling)
(_)

GEOSYNTHETICS INTERNATIONAL S 1998, VOL. 5, NOS. 1-2 165

Downloaded by [ Purdue Univ Lib TSS] on [19/09/16]. Copyright © ICE Publishing, all rights reserved.
BATHURST AND HATAMI D Seismic Response Analysis of a Geosynthetic-Reinforced Wall

γ = bulk unit weight of soil (N/m3)


η = dynamic shear strain (%)
ν = Poisson’s ratio (dimensionless)
νs = Poisson’s ratio for soil (dimensionless)
νw = Poisson’s ratio for panel wall (dimensionless)
ρ = density (kg/m3)
σ3 = minor principal effective stress (Pa)
σo = effective confining pressure (Pa)
ψ = dilatancy angle for soil (_)
ψi = wall-soil interface column dilatancy angle (_)

ABBREVIATIONS

AASHTO: American Association of State Highway and Transportation Officials


FEM: Finite Element Method
FHWA: Federal Highway Administration
FLAC: Fast Lagrangian Analysis of Continua

166 GEOSYNTHETICS INTERNATIONAL S 1998, VOL. 5, NOS. 1-2

Downloaded by [ Purdue Univ Lib TSS] on [19/09/16]. Copyright © ICE Publishing, all rights reserved.

You might also like