See discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/238790017
Seismic Response Analysis of a GeosyntheticReinforced Soil Retaining Wall
Article in Geosynthetics International · August 1998
DOI: 10.1680/gein.5.0117
CITATIONS
102
READS
169
2 authors:
Some of the authors of this publication are also working on these related projects:
Peformance based design of reinforced soil structure View project
Sensor-Enabled Geosynthetics (SEG) View project
Richard J. Bathurst
Royal Military College of Canada
232 PUBLICATIONS 5,281 CITATIONS
SEE PROFILE
Kianoosh Hatami
University of Oklahoma
78 PUBLICATIONS 980 CITATIONS
SEE PROFILE
All content following this page was uploaded by Richard J. Bathurst on 06 June 2017.
The user has requested enhancement of the downloaded file. All in-text references underlined in blue are added to the original document
and are linked to publications on ResearchGate, letting you access and read them immediately.GEOSYNTHETICS INTERNATIONAL S 1998, VOL. 5, NOS. 1-2 127
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
boundaryconditiononthe seismic response ofanidealized6mhighgeosynthetic-reinforced
soil retaining wallconstructed witha verystiff continuous facing panel.The numericalmodels 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-dimensional, 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
recordwas demonstratedtopreserve qualitative features of walldisplacement andreinforcement loaddistribution as that generatedusing the reference harmonic ground motionapplied
at3Hz.The lessons learnedinthis studyare ofvalue toresearchers usingdynamic 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: [email protected].
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 Internationalis
registered under ISSN 1072-6349.
DATES: Originalmanuscriptreceived10February1998,revisedversionreceived 9March
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.BATHURST AND HATAMI D Seismic Response Analysis of a Geosynthetic-Reinforced Wall
128 GEOSYNTHETICS INTERNATIONAL S 1998, VOL. 5, NOS. 1-2
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 pseudostatic 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 (Bathurst 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 collapse and pseudostatic displacement methods for design against excessive deformations 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 development of rational models.
Carefully conceived and executed numerical experiments offer the possibility to improve the understanding of the effects of dynamic loading on reinforced soil structures
and to demonstrate the influence of the primary component properties (e.g. reinforcement stiffness, number of reinforcement layers, base condition, wall geometry, and facing type) on the system response to an earthquake.
Numerical experimentswere carriedout toinvestigate theinfluence ofreinforcement
stiffness, reinforcement length, and toe restraint condition on the seismic response of
an idealized 6m highgeogrid-reinforced soilretaining wallconstructed witha verystiff
continuous facing panel. The wall height, number of reinforcement layers, and reinforced 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 medium- 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 properties, 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
predictionsto those reported in the literature using a finite element method (FEM)techniqueappliedtothesame idealizedreinforced wallstructure understatic loading(Rowe
and Ho 1997).
Asecond set of parametric analyseswas 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 response. The seismic response of the reinforced wall model subjected to a range of harmonic ground motion frequencies was also examined. Finally, the response of the wallBATHURST AND HATAMI D Seismic Response Analysis of a Geosynthetic-Reinforced Wall
GEOSYNTHETICS INTERNATIONAL S 1998, VOL. 5, NOS. 1-2 129
to a scaled earthquake record with the same peak acceleration asthe reference harmonic
ground motion was investigated.
2 PREVIOUS RELATED WORK
2.1 Shaking Table Tests
Chida et al. (1985) carried out a seriesof experimentson 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 panelsfor a total wall height of 3 m. A1.4 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 shaking 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 frequencies 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 inall layersand wereobserved toincrease linearlywith increasingpeak
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 layerswith large values
developed in the layers at the top of the structure. Here, the dynamic load increment in
a reinforcement layer isthe difference between the maximum tensile load under seismic
loading and the maximum tensile load under static loading. Peak acceleration amplification 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 reinforced 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 reinforced wall. The tests were constructed with lightweight blocks and five layers of geogrid 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.BATHURST AND HATAMI D Seismic Response Analysis of a Geosynthetic-Reinforced Wall
130 GEOSYNTHETICS INTERNATIONAL S 1998, VOL. 5, NOS. 1-2
2.2 Dynamic Numerical Modeling
Segrestin and Bastick (1988) used the FEM program SUPERFLUSH to generate numerical 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 articulated 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 reinforcement layers during simulated seismic shaking was generally not at the connections. 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 incrementswere generated when the first 10 seconds
of the 1940 El Centro earthquake record (scaled to a peak acceleration of 0.2g) was applied 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 acceleration) 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 geogrid 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 numerical results showed that wall displacements and reinforcement loads accumulated
with time during seismic shaking. Displacement and reinforcement amplitudes were insignificant 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
wallswithastifferfacingandstifferreinforcement,thenumericalresultsfor thediscrete,
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.BATHURST AND HATAMI D Seismic Response Analysis of a Geosynthetic-Reinforced Wall
GEOSYNTHETICS INTERNATIONAL S 1998, VOL. 5, NOS. 1-2 131
2.3 Summary
The very limited number of experimental and numerical studies identified above, the
limited scope of each study, thedifferent constitutivemodels andnumerical codesused,
and the wide range of results illustrate that the current understanding of the seismic response 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 retainedsoilzoneandtherigid far-endboundary willinfluence modelresponse asdemonstrated in the current paper.
Current practice in North America (AASHTO 1996; FHWA 1996) with regard to empirical 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 condition. 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 steptoward identifyinga numericalmethod andsystematic
approach that canultimately beused tovalidate orimprove currentseismic designpractice for continuous panel walls constructed with a range of soil reinforcement products.
3 FLAC PROGRAM
FLACisanexplicit, dynamic, finitedifference codebased onthe Lagrangiancalculation scheme that is well suited for modeling large distortions, material collapse, and the
dynamic response of earth structures. Complete descriptions of the numerical formulation 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 advantages of using FLAC for seismic analysis is the simplicity of applying seismic loading 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 seismic 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, selected FEM results for a geosynthetic-reinforced continuous panel wall at the end ofBATHURST AND HATAMI D Seismic Response Analysis of a Geosynthetic-Reinforced Wall
132 GEOSYNTHETICS INTERNATIONAL S 1998, VOL. 5, NOS. 1-2
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 condition. 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 reinforcement layers (Figure 1). The wall facing was modeled as a continuous concrete panel 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 :
Es (1)
P
a
= K P σ3 am
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
Hinge
Figure 1. Numerical grid for the reference static load case.
L = 4.25 m
H=6 m
Reinforcement
Interface column
Facing
panel
Fixed boundary
Fixed boundary
Interface
layer
B=15 mBATHURST AND HATAMI D Seismic Response Analysis of a Geosynthetic-Reinforced Wall
GEOSYNTHETICS INTERNATIONAL S 1998, VOL. 5, NOS. 1-2 133
soil material was assumed (νs = 0.3). However, for simplicity, in the current study, the
elasticmoduluswasheldconstantforthe durationofthenumerical experiment,includingconstruction, usingvaluescalculatedimmediatelypriortopropremoval andtaking
σ3 asthe horizontal earthpressure ineachsoil layer. Hence, soil modulus andPoisson’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
strengthofthereinforcementwassettoTy=200kN/mtoensurethat reinforcementrupture 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
wasmodeledbya groutmaterial ofnegligible thicknesswithaninterface frictionangle
δ
g = 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 selected to simulate a perfect bond between the soil and reinforcement layers.
The results of the FEM simulation of reinforced continuous panel walls have been
demonstratedtobe sensitive tomeshconstructiondetailsandmaterial properties at the
reinforcement-wall connections (Rowe and Ho 1997; Andrawes and Yogarajah 1994).
Inthecurrentstudy,asimpleconnectionmodelwasadoptedthatinvolvedattachingthe
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
behindthe facingpanel. Ano-slipboundarywasusedbetweenthe thinsoil columnand
the facing panel. The soil-wall interface column material was assigneda frictionangle
Ôi =20_andadilatancyangle ψi =0. Asimilarthinsoil layerwasintroducedat thebase
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 elements to model the same boundaries.
Thesoilandreinforcementelementswereconstructedinlayers,whilethecontinuous
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)thefacingpanel-soilinterfaceandsoil-foundationinterfaceweremodeledusingthin
layers of soil rather thanzero thicknessinterface 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-thickness interfaces are not permissible in FLAC (Version 3.30) in combination with intersecting 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 modulusfor the soil reducescomputation time for bothstatic anddynamic loading simulations using FLAC.BATHURST AND HATAMI D Seismic Response Analysis of a Geosynthetic-Reinforced Wall
134 GEOSYNTHETICS INTERNATIONAL S 1998, VOL. 5, NOS. 1-2
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 Figure 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 updated stress dependent modulus of elasticity in FLAC models for the soil similar to that
Elevation (m)
6 5 4 3 2 1 0
0 5 10 15 20 25 30 35
FLAC simulation
(a)
6 5 4 3 2 1 0
0.0 0.1 0.2 0.3 0.4 0.5
FLAC simulation
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.
(b)
Normalized connection load, Tc / Ka γ H
Ho (1993)
Ho (1993)
Displacement (mm)
Elevation (m)BATHURST AND HATAMI D Seismic Response Analysis of a Geosynthetic-Reinforced Wall
GEOSYNTHETICS INTERNATIONAL S 1998, VOL. 5, NOS. 1-2 135
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 computational time.
5 DYNAMIC MODELING USING FLAC
Dynamic analyses of a reinforced soil wall subjected to simulated horizontal foundation 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 focused on the influence of material properties and dimensions on seismic response of the
numerical models. The corresponding variables include: reinforcement stiffness, reinforcement 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 magnitude 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 response. Finally, a numerical analysis using the initial six seconds of an actual scaled
earthquake accelerogram was carried out to investigate quantitative and qualitative differences in the seismic response of the idealized reinforced wall due to harmonic excitation and a typical earthquake record.
L = 4.2 m
Thin horizontal
soil layer (sliding
case only)
Hinge
Thin soil interface column
Very stiff
facing
panel
Figure 3. Numerical grid for the reference reinforced soil wall with a fixed-base condition.
H = 6 m
B = 40 m
Right edge of numerical grid and
free-field transmitting boundary
Free-field transmitting boundary
Very stiff foundation (fixed case only)
Base acceleration
Reinforcement
5 4 3 2 1
Layer
10 m
Non-yielding
region
1 m
Fixed boundary
6BATHURST AND HATAMI D Seismic Response Analysis of a Geosynthetic-Reinforced Wall
136 GEOSYNTHETICS INTERNATIONAL S 1998, VOL. 5, NOS. 1-2
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).BATHURST AND HATAMI D Seismic Response Analysis of a Geosynthetic-Reinforced Wall
GEOSYNTHETICS INTERNATIONAL S 1998, VOL. 5, NOS. 1-2 137
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 ofexperimentally measuredtoe loadsfrom afull-scale, continuouspanel 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 numerical 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 reinforcement 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 carried out to examine the influence of the type of far-end truncated boundary on the system response to simulated seismic loading.
5.2 Material Properties
The backfill soil properties used in the dynamic analyses were identical to those reported 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, equivalently, 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 cohesion 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 assigned 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 ofBATHURST AND HATAMI D Seismic Response Analysis of a Geosynthetic-Reinforced Wall
138 GEOSYNTHETICS INTERNATIONAL S 1998, VOL. 5, NOS. 1-2
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)
geogridundercyclic loadingat a frequencyof3Hz (Bathurst andCai 1994). However,
the yield strength of the reinforcement in all cases was kept constant at Ty = 200 kN/m,
whichiswell above the magnitude ofthe maximum reinforcement loadrecordedinthe
simulations.
5.3 Seismic Loading
Soilconstructionandpropreleasewereidenticaltothosedescribedforthestaticloading case in Section 4.1. After static equilibrium was achieved, the full width of the
foundationwassubjectedtothe variable-amplitude harmonic groundmotion recordillustrated in Figure 4. This acceleration record was applied horizontally to all nodes at
the bottom ofthe 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 tζ sin (2πft) (2)
where: α= 5.5, β = 55, and ζ= 12 are constant coefficients; f = frequency; and t = time.
The peakamplitude ofthe input accelerationis0.2g, andthe frequency, f =3Hz,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 inorder toidentify the relative influence of primaryvariables onseismic
response ofa reinforcedsoil wall oftypical height and reinforcedsoil volume. However, 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
proposedworkandwouldrequiregreaterruntimeswithouttheassurancethatpotentiallylargeseismic-inducedeffectswouldresult.Theauthorsofthecurrentpaperareinves-
---2
---1
2 1 0
0 1 2 3 4 5 6
Time (s)
Figure 4. Base acceleration history.
Acceleration (m/s2)BATHURST AND HATAMI D Seismic Response Analysis of a Geosynthetic-Reinforced Wall
GEOSYNTHETICS INTERNATIONAL S 1998, VOL. 5, NOS. 1-2 139
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:
f11 = 1 (3)
4H
Gρ
1 +1 − 2 vH B2
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; andv = Poisson’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 onedimensional 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 simulationsin 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:
T1 = 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 frequency of f = 3 Hz. A possible explanation for the difference in predicted fundamental
frequenciesusing Equations3 and 4 isthat 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 content 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 componentsmay 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
harmonicgroundmotionusedinthecurrentstudyismuchmoreaggressiveon thesystem
response than a true earthquake record with the same peak acceleration and comparable
peak velocity, duration, and predominant frequency. Hence, the numerical analyses resultsreportedinthecurrentstudyareinterpretedlargelyin relativeand qualitativeterms.
Elgamal et al. (1996) have proposed that for conventional reinforced concrete cantilever wall-backfill systems less than 10 m in height, and subject to typical situations of
seismic excitation, a viscous damping ratio of 5% isconservative. Hence, a damping ra-BATHURST AND HATAMI D Seismic Response Analysis of a Geosynthetic-Reinforced Wall
140 GEOSYNTHETICS INTERNATIONAL S 1998, VOL. 5, NOS. 1-2
tio of c = 5% was chosen for the reference case in the parametric analysesreported 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 runsrequired
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 outward 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 displacement at the top of the wall is greater for the fixed-base condition than for the slidingbase 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 issummarized in Figure 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 comparable 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 displacements 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 lessBATHURST AND HATAMI D Seismic Response Analysis of a Geosynthetic-Reinforced Wall
GEOSYNTHETICS INTERNATIONAL S 1998, VOL. 5, NOS. 1-2 141
---0.01
0.00
0.01
0.02
0.03
0.04
0.05
0.06
0.07
0.08
0.09
0.10
---0.01
0.00
0.01
0.02
0.03
0.04
0.05
0.06
0.07
0.08
0.09
0.10
0 1 2 3 4 5 6
Reinforcement layer
Figure 5. History of normalized horizontal wall displacements, ∆x/H, at selected
elevations: (a) fixed-base condition; (b) sliding-base condition.
Time (s)
6 5 4 3 2 1
Soil surface
Reinforcement layer
(a)
(b)
21
5 3
6 4
Soil surface
Foundation surface
J = 1,000 kN/m
L/H = 1
Sliding base
∆x
H H
∆x
Bottom
Bottom
Top
Top
Note: Datum taken at the end of construction following external prop release.
J = 1,000 kN/m
L/H = 1
Fixed base
Foundation surface
Normalized horizontal displacement, Dx / H Normalized horizontal displacement, Dx / H
with L/H = 1 compared to shorter reinforcement length models (i.e. L/H = 0.7). For sliding-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 displacements are greater for the fixed-base condition compared to the sliding-base condi-BATHURST AND HATAMI D Seismic Response Analysis of a Geosynthetic-Reinforced Wall
142 GEOSYNTHETICS INTERNATIONAL S 1998, VOL. 5, NOS. 1-2
6 5 4 3 2 1 06 5 4 3 2 1 0
0.00 0.02 0.04 0.06 0.08 0.10
6 5 4 3 2 1 0
500
1,000
2,000
9,000
69,000
Normalized horizontal displacement, ∆x / H
Elevation (m) Elevation (m)
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.
(a)
(b)
6 5 4 3 2 1 0
Relative Total
J (kN/m)
500
1,000
2,000
9,000
69,000
J (kN/m)
500
1,000
2,000
J (kN/m)
500
1,000
2,000
9,000
69,000
9,000
69,000
Note: Datum taken at the end of construction following external prop release.
J (kN/m)
J (kN/m)
Elevation (m) Elevation (m)
(c)
(d)
J (kN/m)
500
1,000
2,000
9,000
69,000
J (kN/m)
500
1,000
2,000
9,000
69,000BATHURST AND HATAMI D Seismic Response Analysis of a Geosynthetic-Reinforced Wall
GEOSYNTHETICS INTERNATIONAL S 1998, VOL. 5, NOS. 1-2 143
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.
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
100 1000 10000 100000
Reinforcement stiffness, J (kN/m)
End of
seismic
shaking
Fixed base
L/H = 0.7, sliding base
L/H = 1.0, sliding base
L/H = 1.0, fixed base
End of
construction
L/H = 0.7, fixed base
Maximum displacement (m)
Sliding base
tionwhenallotherparametervaluesareequal.However,theinfluenceofthetoerestraint
condition on wall displacements reduces as the stiffness of the reinforcement increases.
6.1.2 Reinforcement Loads
Across section showing the true-scale deformation of the reinforced zone in an example 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, reinforcement 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 connectionsisshowninFigure9.Theresultsareshownforthefixed-and sliding-baseconditions 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 observed 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 reinforcement layer at the end of construction (static loading) and during base shaking (dynamic 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 thatBATHURST AND HATAMI D Seismic Response Analysis of a Geosynthetic-Reinforced Wall
144 GEOSYNTHETICS INTERNATIONAL S 1998, VOL. 5, NOS. 1-2
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.
Grid boundary
H = 6 m
B = 40 m
L = 4.2 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 connections 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-equilibrium 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 fixedbase 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 ofreinforcement stiffnesson the distribution andmagnitude 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 calculated static load values with elevation from numerical analyses for the sliding-base
cases are better captured using the linear distributionsfrom Rankine and Coulombearth
pressure theories than for the fixed-base cases. However, the magnitudes of the reinforcement load using earth pressure theories are consistently lower than predicted values from numerical simulations.
There is no significant influence of reinforcement length on the magnitude and distribution of reinforcement loads under static loading apparent in the data in Figure 10.BATHURST AND HATAMI D Seismic Response Analysis of a Geosynthetic-Reinforced Wall
GEOSYNTHETICS INTERNATIONAL S 1998, VOL. 5, NOS. 1-2 145
0.0
0.1
0.2
0.3
0.4
0.5
0 1 2 3 4 5 6
0.0
0.1
0.2
0.3
0.4
0.5
Figure 9. History of the normalized connection loads in the reinforcement layers:
(a) fixed-base condition; (b) sliding-base condition.
342
5 6
(a)
(b)
J = 1,000 kN/m Reinforcement layer
L/H = 1
J = 1,000 kN/m
L/H = 1
Time (s)
1 234 5 6
Reinforcement layer
Top
Bottom
Note: Tc is the connection load and Ty = 200 kN/m is the yield strength of the reinforcement.
Tc
/ Ty Tc / Ty
Top
Bottom 1
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 underdynamic loadingthan the calculated loadsfor 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 ofthe reinforcement loads forwalls withlower stiffnessreinforcement is more uniform with depth. Moreover, there is essentially no influence of reinforcement 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 configuration with L/H = 0.7.
The trends of the reinforcement load with elevation under dynamic loading for sliding-base cases (Figures 10c and 10d) are qualitatively different from the correspondingBATHURST AND HATAMI D Seismic Response Analysis of a Geosynthetic-Reinforced Wall
146 GEOSYNTHETICS INTERNATIONAL S 1998, VOL. 5, NOS. 1-2
6 5 4 3 2 1 0
0.0 0.2 0.4 0.6 0.8
6 5 4 3 2 1 0
0.0 0.2 0.4 0.6 0.8
6 5 4 3 2 1 0
0.0 0.2 0.4 0.6 0.8
6 5 4 3 2 1 0
0.0 0.2 0.4 0.6 0.8
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.
1
Elevation (m)
2
1
2
3
4
3
2,000
1,000
500
9,000
69,000
(b)
(2) (1)
J (kN/m)
Dynamic
4
5
6, top reinforcement layer 6, top reinforcement layer
5
2
3
4
1
5
4
3
2
J (kN/m)
(c) (d)
Normalized maximum reinforcement load, Tmax / Ty
6, top reinforcement layer
5
6, top reinforcement layer
Elevation (m)
(a)
2,000
1,000
500
9,000
69,000
End of construction (static)
J (kN/m)
(2) (1)
Dynamic
End of construction (static)
(1)
(2)
Dynamic
End of construction (static)
2,000
1,000
500
9,000
69,000
1
(1)
(2)
Dynamic
End of construction (static)
2,000
1,000
500
9,000
69,000
J (kN/m)
(1) Ka = f(f), (2) Ka = f(f, d) (1) Ka = f(f), (2) Ka = f(f, d)
(1) Ka = f(f), (2) Ka = f(f, d) (1) Ka = f(f), (2) Ka = f(f, d)
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 simulations. Dynamic load increment values were calculated by subtracting from the maximum 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 horizontal 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 magnitudeBATHURST AND HATAMI D Seismic Response Analysis of a Geosynthetic-Reinforced Wall
GEOSYNTHETICS INTERNATIONAL S 1998, VOL. 5, NOS. 1-2 147
6 5 4 3 2 1 0
0.0 0.2 0.4 0.6 0.8
6 5 4 3 2 1 0
0.0 0.2 0.4 0.6 0.8
6 5 4 3 2 1 0
0.0 0.2 0.4 0.6 0.8
1
Elevation (m)
Normalized dynamic load increment, ∆T / Ty
(c)
2
3
4
5
6, top reinforcement layer
6 5 4 3 2 1 0
0.0 0.2 0.4 0.6 0.8
1 1
Elevation (m)
J (kN/m)
2,000
1,000
500
9,000
69,000
(a) (b)
J (kN/m)
(1)
2
3
4
5
6, top reinforcement layer
2
3
4
5
6, top reinforcement layer
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.
2,000
1,000
500
9,000
69,000
(1)
(d)
(1) (1)
J (kN/m)
2,000
1,000
500
9,000
69,000
1
2
3
4
5
6, top reinforcement layer
J (kN/m)
2,000
1,000
500
9,000
69,000
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 stiffest 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 valuesof ∆Tthat are in the range ofvalues forthe lowest reinforcement
stiffness cases but increasingly underestimates the magnitude of ∆T as reinforcement
stiffness becomeslarger. However, the trendof the data forthe magnitude of the dynamic load increment, ∆T, with depth below the crest of the wall canbe arguedto be qualitatively similar to the empirical AASHTO results.
The influence ofreinforcement elevation, base condition, and reinforcement stiffness
is summarized in Figure 12. The following observations can be made on the data corre-BATHURST AND HATAMI D Seismic Response Analysis of a Geosynthetic-Reinforced Wall
148 GEOSYNTHETICS INTERNATIONAL S 1998, VOL. 5, NOS. 1-2
0.0
0.2
0.4
0.6
0.8
1.0
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.
0.0
0.2
0.4
0.6
0.8
1.0
0.0
0.2
0.4
0.6
0.8
1.0
0.0
0.2
0.4
0.6
0.8
1.0
100 1000 10000 100000
Reinforcement stiffness, J (kN/m)
Bottom
Top
123456
(a)
(b)
(c)
(d)
Tmax / Ty
Bottom
Top
1 23456
Tmax / Ty
Bottom
Top
123456
Tmax / Ty
Bottom
Top
12 3456
Tmax / Ty
sponding to the end of base shaking using the applied harmonic input acceleration record: (i) reinforcement loads increase in magnitude from top tobottom ofthe 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 numericalsimulations(suchasreinforcementloads)willvarywidelyasaresult ofcharacteris-BATHURST AND HATAMI D Seismic Response Analysis of a Geosynthetic-Reinforced Wall
GEOSYNTHETICS INTERNATIONAL S 1998, VOL. 5, NOS. 1-2 149
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 structures, can be expected to generate large reinforcement loads.
The observation that the empirical AASHTO(1996) method underestimates the magnitude 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-passfilter with a cut-off frequency, flim = 10 Hz (Bellanger 1989). Thisfilter
function was applied to all acceleration response records to exclude the spurious, highfrequency acceleration peaks associated with the reflection of waves in the simulation
runs. The significant portion of seismic energy during actual earthquakes is also imparted at frequencies below 10 Hz (Elgamal et al. 1996). Hence, the upper cut-off frequency 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 immediately 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 acceleration 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 relatively 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 factor 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
layersasmay be expected fromconventional tied-backwedge andnonlinear slipsurface
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 isrequired to determine
if the pattern of internal failure will change with greater reinforcement spacings.BATHURST AND HATAMI D Seismic Response Analysis of a Geosynthetic-Reinforced Wall
150 GEOSYNTHETICS INTERNATIONAL S 1998, VOL. 5, NOS. 1-2
0.0
1.0
2.0
3.0
4.0
5.0
6.0
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.
0.0
1.0
2.0
3.0
4.0
5.0
6.0
0.0 0.2 0.4 0.6 0.8 1.0 1.2
Acceleration (g)
Elevation (m) Elevation (m)
J (kN/m)
69,000
9,000
2,000
1,000
500
Peak
foundation
acceleration
Peak
foundation
acceleration
0.0
0.5
1.0
1.5
2.0
2.5
3.0
100 1000 10000 100000
Figure 14. Variation of the mean base acceleration amplification in the reinforced soil
zone versus the reinforcement stiffness, J, and the base condition.
Reinforcement stiffness, J (kN/m)
Fixed base
Sliding base
Mean base acceleration amplification
J (kN/m)
69,000
9,000
2,000
1,000
500
Note: Acceleration response filter cut-off frequency, flim = 10 Hz.
Notes: Acceleration response filter cut-off frequency, flim = 10 Hz; damping ratio c = 5%.
(a)
(b)BATHURST AND HATAMI D Seismic Response Analysis of a Geosynthetic-Reinforced Wall
GEOSYNTHETICS INTERNATIONAL S 1998, VOL. 5, NOS. 1-2 151
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.
12 m
(a)
4.2 m
6.0 m
20_
40 m
Grid boundary
23_
6.0 m
40 m
Grid boundary
4.2 m
(b)
13 m
31_
7_
Note: Dark shading indicates relatively large shear strains.
23_
31_
Large shear strains were recorded at the wall-soil interface and at the reinforced retained soil interface. The failure volume in each simulation can be approximated by a
bilinearwedge with a break point at the back ofthe reinforcedsoil zone. The breakpoint
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 predicted from solutions forslip surface orientation usingMononobe-Okabe earthpressure
theory (Okabe 1924; Zarrabi 1979; Bathurst and Alfaro 1997). Orientations of 23 and
31_ from the horizontal correspond to computed mean wedge accelerations of approximately 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.BATHURST AND HATAMI D Seismic Response Analysis of a Geosynthetic-Reinforced Wall
152 GEOSYNTHETICS INTERNATIONAL S 1998, VOL. 5, NOS. 1-2
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 afree-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 fixedbase condition. The data show that for the three conditions examined, the free-field
Normalized horizontal displacement, ∆x / H
6 5 4 3 2 1 0
0.00 0.02 0.04 0.06 0.08 0.10
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.
Free-field
Rigid, forced
Rigid, stationary
6 5 4 3 2 1 0
0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35
Rigid, forced
Free-field
Rigid, stationary
Elevation (m) Elevation (m)
Normalized connection load, Tc / Ty
J = 2000 kN/m
L/H = 0.7
B = 40 m
Fixed base
J = 2000 kN/m
L/H = 0.7
B = 40 m
Fixed base
(a)
(b)
Note: Displacement datum taken at the end of construction following external prop release.BATHURST AND HATAMI D Seismic Response Analysis of a Geosynthetic-Reinforced Wall
GEOSYNTHETICS INTERNATIONAL S 1998, VOL. 5, NOS. 1-2 153
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 displacement and reinforcement loads. The latter result is considered to be due to the horizontal 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 using 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
Figure 17. Influence of the model width, B, on the normalized wall displacements:
(a) fixed-base condition; (b) sliding-base condition.
(a)
(b)
6 5 4 3 2 1 06 5 4 3 2 1 0
0.00 0.02 0.04 0.06 0.08
Normalized horizontal displacement, ∆x / H
J = 2000 kN/m
L/H = 0.7
J = 2000 kN/m
L/H = 0.7
Elevation (m) Elevation (m)
B = 7.5 m 15 m 25 m 40 m
B/H = 1.3 2.5 4.2 6.7
Notes: Stationary rigid far-end boundary condition; displacement datum taken at the end of construction
following external prop release.BATHURST AND HATAMI D Seismic Response Analysis of a Geosynthetic-Reinforced Wall
154 GEOSYNTHETICS INTERNATIONAL S 1998, VOL. 5, NOS. 1-2
is greatly influenced by the volume of soil behind the reinforced soil zone for the reference 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 corresponding maximum axial loads in the reinforcement at different elevations are plotted
in Figure 18. Not surprisingly, the maximum axial loads calculated for all reinforcement 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 parameter 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 Figures 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 massof the model and the width of the excitation boundary at the foundation 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 facing 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 numerical 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 5 4 3 2 1 0
0.0 0.1 0.2 0.3 0.4 0.5
Sliding base
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).
B
15 m
25 m
40 m
Elevation (m)
7.5 m
B/H
6.7
4.2
2.5
1.3
Tc / Ty
Fixed base
H/B
0.15
0.24
0.40
0.80
Note: Stationary rigid far-end boundary condition.BATHURST AND HATAMI D Seismic Response Analysis of a Geosynthetic-Reinforced Wall
GEOSYNTHETICS INTERNATIONAL S 1998, VOL. 5, NOS. 1-2 155
yielded zone can be expected to increase with base acceleration according to Mononobe-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 truncated 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 following equation for small strain dynamic shear damping of uncemented sands expressed in percent:
ξ = 9.22 P σ0 a−0.38η0.33 (5)
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 confinement 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 overestimation 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 reinforcement 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 distributions 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 reduction in mean base acceleration amplification factor is approximately 25%. Nevertheless, 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 theBATHURST AND HATAMI D Seismic Response Analysis of a Geosynthetic-Reinforced Wall
156 GEOSYNTHETICS INTERNATIONAL S 1998, VOL. 5, NOS. 1-2
Figure 19. Influence of the soil damping value, c , on the wall displacements and the
magnitude and distribution of thenormalized peak reinforcement connection loads: (a) wall
displacements; (b) normalized maximum connection loads.
6 5 4 3 2 1 0
0.00 0.02 0.04 0.06 0.08 0.10
6 5 4 3 2 1 0
0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35
Tc / Ty
Elevation (m) Elevation (m)
(a)
(b)
Normalized horizontal displacement, ∆x / H
c = 20%
J = 2000 kN/m
L/H = 1.0
B = 40 m
Fixed base
J = 2000 kN/m
L/H = 1.0
B = 40 m
Fixed base
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.
0.0
1.0
2.0
3.0
4.0
5.0
6.0
0.0 0.2 0.4 0.6 0.8 1.0 1.2
Acceleration (g)
J = 2000 kN/m
L/H = 1
Fixed base
Elevation (m)
Peak
foundation
acceleration
Figure 20. Influence of the soil damping value, c, on the magnitude and distribution of
peak horizontal accelerations recorded in the reinforced soil zone.
c = 20%
c = 5%
c = 10%
c = 20%
c = 5%
c = 10%
Note: Acceleration response filter cut-off frequency, flim = 10 Hz.
c = 5%
c = 10%BATHURST AND HATAMI D Seismic Response Analysis of a Geosynthetic-Reinforced Wall
GEOSYNTHETICS INTERNATIONAL S 1998, VOL. 5, NOS. 1-2 157
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 frequencyf = 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 frequency. 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 (reference case geometry) and 0.8 (minimum width model). Equation 3 for geometries approaching 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 dimensional 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 reinforcement loads during harmonic shaking. The results of numerical analyses show
that the magnitude of reinforcement loads is influenced by input ground motion frequency 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 characteristics (i.e. frequency content, duration of strong ground motion, and peak groundBATHURST AND HATAMI D Seismic Response Analysis of a Geosynthetic-Reinforced Wall
158 GEOSYNTHETICS INTERNATIONAL S 1998, VOL. 5, NOS. 1-2
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0 1 2 3 4 5 6
0.00
0.01
0.02
0.03
0.04
0.05
0.06
0.07
0.08
0.09
0.10
H = 6 m, B = 40 m
H
Reference case
in current
investigation
(f = 3 Hz)
H/B = 0
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 .
(a)
J = 69,000 kN/m
L/H = 0.7
Fixed base
B = 40 m
Maximum ∆x
H/B = 0.15
H/B = 0.8
B
Predicted fundamental frequency from Equation 3
Reinforcement
layer
1 2 34 5
6
Bottom
Top
(b)
H/B = 0
H/B = 0.15 H/B = 0.8
3.32 3.43 4.5
Note: Displacement datum taken at the end of construction following external prop release.
Normalized horizontal displacement
Dx / H
Normalized maximum reinforcement load
T
max / Ty
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 frequency. 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 displacements at the end of the simulation (compare Figure 22a to Figure 5a). Figure 22bBATHURST AND HATAMI D Seismic Response Analysis of a Geosynthetic-Reinforced Wall
GEOSYNTHETICS INTERNATIONAL S 1998, VOL. 5, NOS. 1-2 159
6 5 4 3 2 1 0
0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35
El Centro record
(scaled to 0.2g)
(a)
0.00
0.01
0.02
0.03
0.04
0 1 2 3 4 5 6
J = 2000 kN/m
L/H = 0.7
Fixed base
Tc / Ty
Elevation (m)
(b)
J = 2000 kN/m
L/H = 0.7
Fixed base
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.
Harmonic record
Time (s)
Top
Mid-height
H
∆x
Normalized horizontal displacement
Dx / H
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 records and actual earthquake records on the seismic response of idealized reinforced soil
walls. However, the magnitude of displacements and reinforcement loads predicted using 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 harmonicBATHURST AND HATAMI D Seismic Response Analysis of a Geosynthetic-Reinforced Wall
160 GEOSYNTHETICS INTERNATIONAL S 1998, VOL. 5, NOS. 1-2
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 result 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-documented 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 response 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. Similar 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 displacement 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 reinforcement stiffness and increasing reinforcement length. However, for models subjected 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 wallsoil interface with reduced frictional resistance resulted in the maximum reinforcement 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 resulting from static loading alone. The magnitude of additional dynamic-induced loads
in reinforcement layers was observed to increase with increasing stiffness of the
reinforcement.BATHURST AND HATAMI D Seismic Response Analysis of a Geosynthetic-Reinforced Wall
GEOSYNTHETICS INTERNATIONAL S 1998, VOL. 5, NOS. 1-2 161
6. The magnitude and distribution of dynamic-induced reinforcement loads was influenced by the base condition. For walls with a sliding base, dynamic load increments 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 applicable to the same models subjected to other input ground motions.
7. Horizontal ground acceleration was amplified with height above the base foundation. 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 inclination 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 harmonic 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 applied to the structure. The difference between the frequency of the base excitation record and the fundamental frequency of the model is the most important factor
determining wall response to seismic excitation. The large displacements and reinforcement loads reported in this paper are due to the observation that the applied frequency 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 displacements 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 dynamic numerical modelling of reinforced soil walls. The use of excessively narrow retained 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-BATHURST AND HATAMI D Seismic Response Analysis of a Geosynthetic-Reinforced Wall
162 GEOSYNTHETICS INTERNATIONAL S 1998, VOL. 5, NOS. 1-2
ment with depth below the crest predicted by the AASHTO method was not observed
inallsimulationsfor wallsconstructed withstiffness valuescomparable togeosynthetic
reinforcementmaterials.Inaddition,theAASHTOmethodpredictsthesamemagnitude
of dynamic load in a reinforcement layer for the same peak horizontal ground accelerationregardlessofothercharacteristicvaluesoftheaccelerogram.Incontrast,themagnitude 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 motion (e.g. frequency content, duration of strong ground motion, and peak ground velocity) 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, numerical results must be checked against physical measurements from large-scale shaking 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 toacknowledge theefforts ofthe reviewersof thispaper whosecomments 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, American 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 Geomechanics, Morgantown, West Virginia, USA, May 1994, pp. 1313-1318.
Bellanger, M, 1988, “Digital Processing of Signals: Theory and Practice”, SecondEdition, John Wiley and Sons, New York, New York, USA, 384 p.
Bathurst,R.J.,1993,“Investigation ofFooting Restrainton Stabilityof Large-ScaleReinforced Soil Wall Tests”, Proceedings of the 46th Canadian Geotechnical Conference, Regina, Manitoba, Canada, September 1993, pp. 389-398.
Bathurst, R.J. and Alfaro, M.C., 1997, “Review of Seismic Design, Analysis and Performance of Geosynthetic Reinforced Walls,Slopes andEmbankments”, Earth Reinforcement, Ochiai, H., Yasufuku, N. and Omine, K., Editors, Balkema, Vol. 2,
ProceedingsofInternational Symposiumon EarthReinforcement, Fukuoka,Kyushu,
Japan, November 1996, Keynote Lecture, pp. 887-918.BATHURST AND HATAMI D Seismic Response Analysis of a Geosynthetic-Reinforced Wall
GEOSYNTHETICS INTERNATIONAL S 1998, VOL. 5, NOS. 1-2 163
Bathurst, R.J. and Cai, Z., 1995, “Pseudo-static Seismic Analysis of Geosynthetic Reinforced 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 Geosynthetic Reinforced Segmental Retaining Walls”, Canadian Geotechnical Journal,
Vol. 31, pp. 937-955.
Cai, Z. and Bathurst, R.J., 1995, “Seismic Response Analysis of Geosynthetic Reinforced 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 translated from Japanese)
Cundall, P. and Board, M., 1988, “A Microcomputer Program for Modelling LargeStrain Plasticity Problems”, Computer Methods and Advances in Geomechanics, Siriwardane, 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 StabilizedEarth Wallsand ReinforcedSoil SlopesDesign
and Construction Guidelines”, Federal Highway Administration (FHWA) Demonstration 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 Reinforced Soil Walls”, Geotextiles and Geomembranes, Vol. 14, No. 10, pp. 521-541.
Ishihara, K., 1996, “Soil Behaviour in Earthquake Engineering”, Clarendon Press, Oxford 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.BATHURST AND HATAMI D Seismic Response Analysis of a Geosynthetic-Reinforced Wall
164 GEOSYNTHETICS INTERNATIONAL S 1998, VOL. 5, NOS. 1-2
Murata, O., Tateyama, M. and Tatsuoka, F., 1994, “Shaking Table Tests ona Large Geosynthetic-Reinforced Soil Retaining Wall Model”, Recent Case Histories of Permanent Geosynthetic-Reinforced Soil Retaining Walls , Tatsuoka, F. and Leshchinsky,
D., Editors, Balkema, Proceedings of Seiken Symposium No. 11, Tokyo, Japan, November 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, Balkema, Vol. 2, Proceedings of International Symposium on Earth Reinforcement Practice, 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 Reinforcement, 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, Cambridge, Massachusetts, USA, 140 p.
NOTATIONS
Basic SI units are given in parentheses.
amax = peak base acceleration (m/s2)BATHURST AND HATAMI D Seismic Response Analysis of a Geosynthetic-Reinforced Wall
GEOSYNTHETICS INTERNATIONAL S 1998, VOL. 5, NOS. 1-2 165
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) (_)
δ
g = grout-soil interface friction angle (_)
Ô = friction angle for soil (_)
Ôi = wall-soil interface column friction angle (used in numerical modeling)
(_)BATHURST AND HATAMI D Seismic Response Analysis of a Geosynthetic-Reinforced Wall
166 GEOSYNTHETICS INTERNATIONAL S 1998, VOL. 5, NOS. 1-2
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
View publication stats