Professional Documents
Culture Documents
14.1. Introduction The rst analytical treatment of the inuence of seismic-induced forces on
the stability of earth retaining structures can be traced to the work of
Sabro Okabe in his landmark paper (Okabe, 1924). Since this seminal
work there has been a large body of research on the development of
analytical methods that consider the potentially large forces that exert
additional destabilizing forces on earth retaining walls, slopes, dams
and embankments during earthquakes. The vast majority of this work
has been focused on conventional earth structures. The analysis methods
that have been proposed include:
. pseudo-static rigid body analyses that are variants of the original
MononobeOkabe approach
. displacement methods that originate from Newmark sliding block
models
. dynamic nite element/nite dierence methods.
However, with the growing use of geosynthetics in reinforced soil walls,
slopes and embankments, the need to extend current methods of analysis
for conventional structures under seismic loading to geosynthetic-
reinforced systems in similar environments has developed. A concurrent
need has been the requirement to select properties of the component
materials that represent rapid and/or cyclic loading conditions.
This chapter is an extended and updated version of a state-of-the-art
review paper by Bathurst and Alfaro (1996) that appears as a keynote
paper in the Proceedings of the International Symposium on Earth
Reinforcement, IS-Kyushu '96, Fukuoka, Kyushu, Japan in November
1996. This chapter presents selected published works related to the
properties of cohesionless soil, geosynthetic reinforcement and facing
components under cyclic loading, and summarizes the important features
of current analytical and numerical methods for the seismic analysis and
design of geosynthetic-reinforced soil walls and slopes. The scope of the
chapter is restricted to structures seated on rm foundations for which
settlement and collapse of the foundation materials are not a concern.
Non-surcharged structures with simple geometry are considered and
the reinforced and retained soils are assumed to be homogeneous,
unsaturated and cohesionless.
An important component of recent work in the eld of seismic
performance of reinforced walls and slopes has been the use of carefully
conducted numerical studies to gain insight into the performance of
reinforced walls and slopes under simulated seismic loading. This chapter
highlights numerical modelling investigations by the authors and others,
and identies the implications of the results to current design practice.
Many of the examples that highlight important issues related to seismic
performance of geosynthetic-reinforced soil-retaining structures are
328 Geosynthetics and their applications
14.2.1. Soil
14.2.1.1. Strength properties (Coulomb friction angle)
Pseudo-static, pseudo-dynamic and displacement (Newmark) methods
introduced later in the chapter describe cohesionless soil strength according
to the Coulomb failure criterion. The selection of an appropriate value of
soil friction angle, , becomes an issue in these methods, particularly with
respect to the choice of peak, p , or residual (constant volume), cv ,
strength values. A review of the literature suggests that for dry cohesionless
soils the rate of loading used in direct shear or triaxial tests has negligible
eect on shear strength (Bachus et al., 1993). For example, Schimming and
Saxe (1964) used a direct shear device to test Ottawa sand under both static
and dynamic conditions. No signicant dierence in strength envelopes
was recorded (Fig. 14.1). Conventional practice using Newmark methods
is to assume that the cohesionless soil friction angle does not change
during an earthquake.
where Gmax is the maximum shear modulus and smax is the maximum
shear strength. The equation for the unloading curve from the point
(cr ; sr ) at which the loading reverses direction is given by:
s sr c" cr
f 14:2
2 2
or
c" cr
Gmax
s sr 2
14:3
2 1 Gmax =2smax jc" cr j
The shape of the unloadingreloading curve is shown in Fig. 14.2. The
tangent shear modulus, Gt , for a point on the backbone curve is given by:
Gmax
Gt 14:4
1 Gmax =smax jc" j2
and at a stress point on an unloading or reloading curve:
Gmax
Gt 14:5
1 Gmax =2smax jc" cr j2
The response of the soil to uniform conning pressure is assumed to be
non-linear elastic and dependent on the mean normal stress. Hysteretic
behaviour, if any, is neglected in this mode. The tangent bulk modulus,
Bt , is expressed in the form:
rm n
Bt Kb Pa 14:6
Pa
where Kb is the bulk modulus constant, Pa is the atmospheric pressure in
units consistent with mean normal eective stress rm , and n is the bulk
modulus exponent.
References to variations on the above model and other advanced con-
stitutive models can be found in the textbook by Kramer (1996a).
Soilgeosynthetic composites
Chen et al. (1996) carried out dynamic triaxial tests on reinforced sand
samples. They found that the shear modulus of reinforced specimens
Geosynthetic-reinforced soil walls and slopes 331
where T a is the axial load per unit width (e.g. kN/m), Ji is the initial load
modulus, ea is the axial strain, and eaf is the axial strain at failure. The
details of the model parameters are shown in Fig. 14.6(a). The tangent
load modulus, J, on the initial loading curve is calculated as:
dT a ea
J Ji 1 14:8
dea eaf
rule. The equation for the unloading curve from point A (er ; T r ), or for
the reloading curve from point B at which the load reverses direction,
is given by:
Ta Tr Jur ea er =2
14:11
2 1 Jur =2T max jea er j
where Jur is the unload stiness dened in terms of the initial load stiness
according to Jur kJi and k is a constant.
Bonaparte et al. (1986) cautioned that the strain at rupture for HDPE
geogrids will decrease with increasing rate of loading and, hence, inuence
the choice of rupture load in limit state design. Only one of the tests shown
in Fig. 14.3 was taken to rupture due to equipment limitations; so, possible
rate eects cannot be quantied here. The reduction of rupture load
capacity for HDPE geogrids under high rates of loading also has
implications to Newmark sliding block methods of analysis where large
cumulative displacements may be computed (see Section 14.3.3).
Moraci and Montanelli (1997) also carried out cyclic load tests on
HDPE specimens at frequencies in the range 0:11 Hz and at dierent
load amplitudes. They found that for the HDPE material the unload
reload stiness value decreased with increasing load amplitude and
increased with greater loading frequency. They observed that the
unloadreload stiness of the HDPE reinforcement materials for load
amplitudes less than 60% of the reference tensile strength was approxi-
mately 1:5 to 2 times the secant stiness from monotonic tensile strength
tests. This observation can be used to estimate the unloadreload stiness
of the specic HDPE geogrid investigated. They also found that the
stiness value of HDPE specimens was greater for specimens that were
cycle loaded from a minimum load equal to 20% or 40% of the maximum
applied load (i.e. prestressed specimens) than for specimens that were
fully unloaded during each load cycle.
Ling et al. (1998) carried out strain-controlled cyclic loading tests on
virgin and prestressed specimens of three commonly used geogrids
manufactured from HDPE, polypropylene (PP) and woven PET. The
cyclic strain rate was kept at the 10%/min rate in conformance with
the ASTM D4595 method of test. They found that the reload stiness
of all the polymeric materials examined at any given load level increased
with the number of loading cycles. The magnitude of the stiness increase
was greater at higher load levels. They also concluded that the index
strength loadstrain curve from static loading tests was in reasonable
agreement with the backbone curve for each material under low
frequency cyclic loading. The results of cyclic loading tests on the three
dierent reinforcement materials showed that the index strength of PP
geogrid specimens was not changed signicantly as a result of cyclic
loading. In contrast, the post-cyclic tensile strength values of HDPE
and woven PET type geogrids increased with the number of cyclic
loads and load amplitude. Finally, Ling et al. (1998) proposed the follow-
ing hyperbolic formula to estimate the accumulated reinforcement strain,
e eo , from the number of load cycles, Nl , for a given load intensity level:
Nl
e eo 14:12
Nl
where eo is the strain developed during primary loading, and and are
constants.
An implication to seismic design of the results of standard monotonic
loading wide-width tensile tests and the cyclic load data reviewed here, is
that initial and secant stiness values for uniaxial HDPE and woven PET
336 Geosynthetics and their applications
Pullout tests
The simplest pullout model in limit-equilibrium based methods of
analysis takes the form (e.g. Public Works Research Institute (PWRI,
1992)):
Tpull 2La Ci rv tan 14:13
where Tpull is the pullout capacity, La is the anchorage length, rv is the
vertical stress acting over the anchorage length, is the friction angle
of the soil, and Ci is the interaction coecient that is interpreted from
the results of pullout tests. In the United States, a combination of
terms are used to calculate default values of Ci based on the type of
geosynthetic, aperture size, and d50 of the conning soil. The reader is
referred to FHWA (1996) guidelines for details.
A large amount of data can be found on the pullout behaviour of
geotextiles and geogrids in combination with cohesionless soils (e.g.
Farrag, 1990). Bachus et al. (1993) reported the results of constant rate
of displacement (static) pullout test results on four dierent geogrids in
sand. Most tests gave interaction coecient values equal to, or slightly
in excess of, 1:0. Increasing the rate of loading from 1 to 150 mm/min
did not result in signicant changes in interaction coecient values.
338 Geosynthetics and their applications
Repeated load interface shear tests can also be carried out using the
NCMA (Simac et al., 1993) methodology for blockblock shear
response. Static testing shows that interface shear behaviour can be
inuenced by the presence of a geosynthetic layer. Cai and Bathurst
(1996a) assumed that static interface shear values were reasonable
in sliding block analyses for systems that provide positive interlock in
the form of shear keys, pins and other forms of connectors (Section
14.3.4.1).
14.3. Seismic Analytical and numerical approaches for the seismic analysis of
analysis and design reinforced walls, slopes and embankments can be divided into the follow-
ing categories:
of walls and slopes
(a) pseudo-static methods
(b) displacement methods
(c) dynamic nite element/nite dierence methods.
In this chapter, global stability modes of failure for walls are not
addressed.
where is the peak soil friction angle (peak ), is the wallslope face
inclination (positive in a clockwise direction from the vertical), is the
mobilized interface friction angle at the back of the wall (or back of the
reinforced soil zone), is the backslope angle (from horizontal), and
is the seismic inertia angle given by:
1 kh
tan 14:16
1 kv
Quantities kh and kv are horizontal and vertical seismic coecients,
respectively, expressed as fractions of the gravitational constant, e.g.
Seed and Whitman (1970) decomposed the total (active) earth force,
P AE , calculated according to equations (14.14) and (14.15) into two
components representing the static earth force component, P A , and the
incremental dynamic earth force due to seismic eects, Pdyn . Hence:
P AE P A P dyn 14:17
or
1 kv K AE K A K dyn 14:18
where KA is the static active earth pressure coecient, and Kdyn is the
incremental dynamic active earth pressure coecient. Closed-form
approximate solutions for the orientation of the critical planar surface
from the horizontal, AE , have been reported by Okabe (1924) and
Zarrabi (1979). These solutions can be expressed as follows:
1 A D
AE tan 14:19
E
where:
A tan
p
D A A B B C 1
E 1 C A B 14:20
B 1=tan
C tan
Equation (14.19) can be used to calculate the orientation of the
assumed active failure plane within the reinforced soil mass and in the
retained soil. However, the result of pseudo-static analyses of the type
described here have been shown to lead to excessively long reinforcement
lengths if reinforcement layers are required to extend beyond the internal
failure plane. Current practice in North America is to assume that the
orientation of the internal failure plane for reinforcement design is
described by static load conditions (i.e. AE (kh kv 0)) (AASHTO,
1998; FHWA, 1996; NCMA Bathurst, 1998). Koseki et al. (1998a)
and Tatsuoka et al. (1998) have proposed a pseudo-static design
method that results in internal failure planes that are steeper than those
calculated using a rigorous interpretation of the extended Coulomb
wedge approach.
Geosynthetic-reinforced soil walls and slopes 343
Bathurst and Cai (1995) have proposed the total active earth pressure
distribution illustrated in Fig. 14.12 for external, internal and facing
stability analyses of reinforced segmental retaining walls. The normalized
elevation of the resultant total earth force varies over the range
1=3 < md < 0:6 depending on the magnitude of Kdyn . The assumed
pressure distribution is based on a review of the literature for conven-
tional gravity retaining wall structures in North America, where the
dynamic increment is typically taken as acting at 0:6H above the base
of the wall. The total pressure distribution is identical to that recom-
mended for the design of exible anchored sheet pile walls under seismic
loads (Ebeling and Morrison, 1993), and is used in AASHTO (1998) and
FHWA (1996) design guidelines for reinforced soil wall structures. In the
absence of ground acceleration, the distribution reduces to the triangular
active earth pressure distribution due to soil self-weight. The inuence of
reinforcement stiness and ground motion on the distribution and line of
action of active earth forces under static and dynamic loading has been
investigated through numerical modelling by Bathurst and Hatami
(1999a) and is discussed in Section 14.3.4.2.
mass and the retained soil zone will not reach peak values at the same time
during a seismic event. Christopher et al. (1989) proposed the following
expression for horizontal backlls:
PIR 0:5kh
H 2 14:22
:
where 0 6 based on recommendations for reinforced walls that use
steel reinforcement strips (Segrestin and Bastick, 1988). Cai and Bathurst
(1995) proposed an expression that gives similar results for typical L=H
ratios for segmental walls:
PIR kh W R 14:23
where 0:6. AASHTO (1998) interims propose that P IR be calculated
using equation (14.22) with 1 and that the external dynamic active
earth force component, P dyn , be reduced by 50%. North American
practice is to reduce dynamic factors of safety against sliding and over-
turning to 75% of the static factor of safety values in recognition of the
transient nature of seismic loading. The calculation method for P IR
and reduction of static factors of safety described above for AASHTO
has been adopted for pseudo-static seismic design of reinforced segmental
retaining walls by the NCMA (Bathurst, 1998).
Dynamic factors of safety are also reduced in Japan (PWRI, 1992;
GRB, 1990; Koga and Washida, 1992). However, factor of safety calcu-
lations for wall base sliding in Japan do not consider any reduction in
inertial force, P IR (i.e. equation (14.23) is used with 1). In order to
further reduce conservatism in the Japanese approach for base sliding,
Fukuda et al. (1994) have proposed ignoring the dynamic force incre-
ment, P dyn , and restricting seismic loading contributions to the gravity
mass term, P IR , only. Overturning criteria for walls are restricted to
ensuring that the resultant force acting at the base of the reinforced
mass, WR , falls within L=3 of the base midpoint for walls subject to earth-
quake. FHWA (1996) guidelines for geosynthetic-reinforced walls also
omit overturning as a potential failure mode for geosynthetic-reinforced
soil walls. However, to be consistent with current static design of
reinforced segmental retaining walls (Simac et al., 1993), overturning is
considered for seismic design of this class of structure (Bathurst, 1998).
Bathurst et al. (1997) used the NCMA pseudo-static method to produce
design charts for the preliminary evaluation of seismic resistance of
segmental reinforced soil-retaining walls on rm foundations. The
charts are presented as the ratio of dynamic to static safety factor
values for peak horizontal ground accelerations up to 0.5g and soil
friction angle values in the range 258 < peak < 458.
Rankine failure plane ( =4 =2) for vertical walls, and Coulomb
theory with a Coulomb angle according to equation (14.19) (using kh
kv 0 in equation (14.16)) for walls with a facing batter greater than
108. The dynamic earth force is calculated as P dyn kh WA , where
WA is the weight of the static internal failure wedge. The distribution
of the dynamic tensile reinforcement load increment, T dyn , is weighted
based on total anchorage length in the resistance zone according to:
N
X
T dyn i Pdyn Lai = Laj 14:24
j 1
where:
1
A1 14:27
sin 1 tan f cos 1
B1 tan f sin 1 cos 1 14:28
The quantity is the inter-wedge shear mobilization ratio and varies over
the range 0 1. Parameter f is the factored soil friction angle
expressed as:
f tan1 tan =FS 14:29
The horizontal out-of-balance force, PAE , is calculated as:
P AE P 1 kh W 2 B2 A2 1 kv W 2 V 1 14:30
where:
1
A2 14:31
tan f sin 2 cos 2
B2 tan f cos 2 sin 2 14:32
By setting FS 1 (i.e. f ), an equivalent total active earth pressure
coecient for the most critical trial geometry (i.e. trial search that
yields a maximum value for P AE in the slope) can be calculated as:
K AE 2PAE =
H 2 14:33
This approach has been used by Bonaparte et al. (1986) to produce
seismic design charts for geosynthetic-reinforced soil slopes. The total
required P design strength of the horizontal layers of reinforcement is
taken as T i P AE . The two-part wedge approach with 0 is used
by the Geogrid Research Board (GRB, 1990) to calculate KAE according
to equations (14.30) and (14.33) for internal stability calculations.
The two-part wedge analysis degenerates to a single wedge analysis by
restricting trial searches to 1 2 and setting 0. All three solutions
(MO, single and two-part wedge) give the same solution for the
horizontal component of total earth force when 0. In addition,
direct sliding mechanisms, including those generated at the base of the
reinforced soil mass or along reinforcement layers, can be analysed
using the two-part wedge approach.
An alternative strategy that extends the general approach used by
Woods and Jewell (1990) for statically loaded slopes to the seismic case
(Bathurst, 1994) is to rewrite equation (30) as:
P
B1 A1 T i1 X
P AE P 1 kh W 2 T i2
tan f B1 A1
B2 A2 1 kv W 2 V 1 14:34
The factor of safety for a given two-part wedge geometry corresponds to
the value of FS that yields PAE 0. The factor of safety for a slope
corresponds to the minimum value of FS from a search of all potential
failure geometries. It should also be noted that in this approach, the
same global FS is applied to the reference design tensile strength of the
reinforcement and pullout capacity dened by equation (14.13). Equation
(14.34) illustrates that the value of FS against collapse is independent of
the location of the reinforcement layers for 0.
Ling et al. (1996) presented design charts for calculating geosynthetic-
reinforcement strength and length against direct sliding using a two-part
wedge mechanism. Ling et al. (1997) maintained that the direct sliding
Geosynthetic-reinforced soil walls and slopes 351
failure mode governs reinforcement length design for lower layers as seis-
mic acceleration increases.
Tatsuoka et al. (1998) concluded that the two-part wedge geometry is a
valid failure geometry for geosynthetic-reinforced soil walls with a full
height rigid facing and short reinforcement lengths based on shaking
table tests. The pattern and location of the failure shape is controlled
by reinforcement length. Tatsuoka et al. (1998) proposed a modied
two-part wedge method. They concluded that the size of failure wedge
from the modied two-part wedge method was typically smaller than
what would be predicted from conventional two-part wedge analysis
and more realistic according to experimental observations.
Ismeik and Guler (1998) considered the contribution of vertical, full-
height concrete panel facing rigidity on wall stability using a two-part
wedge analysis. Their method allows the contribution of the facing
rigidity to be included explicitly to reduce the reinforcement loads and
reinforcement lengths that would otherwise be larger without the contri-
bution of the structural facing.
is given by:
R A etan f 14:35
For an assumed surface (i.e. for any three independent parameters
dening a log spiral, xp , yp and A), the moment equilibrium equation
about the pole, P, can be explicitly written as:
X
M p 1 kv Wxc xp kh Wyp yc P AE yp yAE
0 14:36
Note that the moment about the log spiral pole is independent of the
distribution of normal and shear stresses over the log spiral because
their resultant must pass through the pole. The point of application of
the components of seismic inertial forces is taken at the centre of the
failure mass. The critical mechanism corresponds to the trace that
yields the maximum value of PAE required to satisfy equation (14.36).
Clearly, the elevation, yAE , of the equivalent out-of-balance horizontal
force PAE inuences the magnitude of PAE . Here, it is assumed a priori
that yAE H=3. The equivalent dynamic active earth pressure coecient,
KAE , can be calculated using equation (14.33) with FS 1 (i.e. f ).
In practice, the factor of safety against collapse of aPreinforced slope
can be determined by replacing P AE (yp yAE ) with T i (yp yi ) in
equation (14.36) and nding the minimum valuePfor FS from a search
of all potential failure geometries that yields MP 0. This value
corresponds to the minimum factor of safety for the reinforced soil
slope (Leshchinsky, 1995). The formulation of equation (14.36) illustrates
that the FS against collapse is a function of the location of the reinforce-
ment layers.
Ling et al. (1997) used a log spiral failure pattern in tie-back internal
stability calculations. Ling and Leshchinsky (1998) extended the method
to calculate the stability and permanent displacement of geosynthetic-
reinforced soil walls under the combined eect of horizontal and vertical
ground acceleration. They considered three dierent modes of failure in
their analysis:
(a) tie-back/compound failure
(b) direct sliding
(c) pullout.
They assumed a log spiral failure shape in their pseudo-static analyses of
the tie-back/compound failure mechanism. As in all pseudo-static
methods of analyses, the method can be expected to result in conservative
design because a momentary acceleration-induced force is assumed to act
permanently on the wall. However, they argue that the inherent conserva-
tism in the method is required since possible acceleration amplication is
disregarded.
where:
H
Qh t mzaz; t dz 14:46
0
(c) interface shear between facing units with or without the presence
of a geosynthetic inclusion (Fig. 14.25(b)).
A summary of calculation results for the geosynthetic-reinforced soil wall
structure shown in Fig. 14.26 is given in Table 14.1 assuming 358.
The material properties for the facing units have been taken from
large-scale laboratory tests carried out at the Royal Military College of
Canada (RMCC). The blockgeosynthetic interface shear properties
(au ; u ) were selected to represent a system with relatively low interface
shear capacity in order to generate a worst case set of displacement
predictions. The EW (908) horizontal ground acceleration component
recorded at Newhall Station (California Strong Motion Instrumentation
Program) during the 17 January 1994 Northridge earthquake (M 6:7)
was used as the input earthquake data. The record shows a peak
horizontal ground acceleration of km 0:60. The total permanent
displacement at the wall face at each elevation from the initial static
position was estimated by adding the layer displacement to the cumula-
tive displacement below that layer. The layer displacement was taken as
Layer Displacement: mm
Newmark Empirical
8 154 206
7 47 70
6 29 49
5 25 41
4 25 41
3 25 41
2 24 36
1 21 29
Base sliding 11 15
Controlling mechanism is facing shear, otherwise internal
sliding controls.
Geosynthetic-reinforced soil walls and slopes 361
Reinforced slopes
The dynamic response of reinforced and unreinforced soil slope models
with c properties resting on a rm foundation was determined by
the senior author and co-workers using a modied version of the
TARA-3 program (Finn et al., 1986). The slopes were 12 m high with a
side slope of 1:1 (Yogendrakumar et al., 1991). One slope was lightly
reinforced with 12 m long polymeric reinforcement layers with a vertical
spacing of 2 m. The nite element representation of the reinforced soil
slope is shown in Fig. 14.27(a). The reinforcement was modelled using
the non-linear quadratic equation with linear (non-hysteretic) unload
reloading behaviour described in Section 14.2.2.2. The slope was
The geometry of the failed mass was a combined two-part wedge (Figs
14.34(a)(b)) which is similar to the observed failure geometry for walls
with a similar reinforcement to height ratio in shaking table studies
reported by Tatsuoka et al. (1998) (Fig. 14.34(c)). The top wedge in the
numerical modelling cited here extended beyond the reinforced zone at
an angle that was consistent with the predicted value from Mononabe
Okabe theory considering acceleration amplication over the depth of
backll.
Bathurst and Hatami (1998b) carried out two groups of parametric
analyses on physical and numerical model parameters. In the parametric
analyses on physical parameters, the reinforcement stiness values
ranged from very sti geogrids to metallic reinforcement. Bathurst and
Hatami compared the reinforcement load distribution behind the facing
with the distribution predicted from Coulomb and Rankine earth
pressure theories. They found that the load distributions for geosyn-
thetic-reinforcement materials in the lower stiness range (e.g.
J < 2000 kN/m) were essentially uniform over the height of the wall
with xed toe condition and deviated from the linear distribution
predicted from the two earth pressure theories (Fig. 14.35). The eect
of reinforcement stiness on the load distribution behind the wall has
an important implication to the pseudo-static seismic design of geo-
synthetic-reinforced soil walls. The dynamic load distribution behind
metallic-reinforced soil walls is triangular, whereas it is essentially uni-
form for the case of less sti, polymeric reinforcement. The load distribu-
tion determines the local failure mode of the facing in segmental retaining
walls. Bathurst and Hatami found that the inuence of reinforcement
stiness and length on wall response was larger for the xed-toe case
compared to a toe that was pinned but free to slide laterally. Bathurst
and Hatami found that acceleration amplication in the backll was
slightly greater for the xed toe condition compared to the case where
the toe was free to slide.
The numerical model parameters investigated included the backll far-
end boundary condition, backll width and viscous damping ratio.
Geosynthetic-reinforced soil walls and slopes 369
Bathurst and Hatami found that the inuence of backll width on calcu-
lated response of the wall was signicant for both toe restraint conditions
when B=H < 5, where B is the width of the numerical grid and H is the
height of the wall facing.
Bathurst and Hatami (1999a) examined the change in elevation of the
reinforcement load resultant with reinforcement stiness under static
(end of construction) and seismic (end of input ground motion) loading
conditions using numerical simulation. Figure 14.36 shows that the
resultant reinforcement load elevation under both static and dynamic
loading conditions is generally less for higher reinforcement stiness
values. Furthermore, for static load conditions and a given reinforcement
stiness, the normalized elevation of the load resultant, ms , is lower in
taller wall models. An important implication of the trend in the data in
Fig. 14.36(a) to limit-equilibrium based design of walls under static
loading is that the assumption of a triangular load distribution may be
most applicable for very sti reinforcement systems (i.e. steel strip
reinforced walls) and may not be applicable for extensible reinforcement
systems (i.e. geosynthetic-reinforced soil walls). The curves in Fig.
14.36(b) show that the normalized elevation of the load resultant
during base shaking is always lower than the corresponding static load
370 Geosynthetics and their applications
case. For dynamic load conditions and a given reinforcement stiness, the
normalized elevation of the load resultant, md , increases with increasing
wall height for extensible reinforcement systems (opposite trend to
static loading case). The trend described here is consistent with the results
of the pseudo-dynamic method described in Section 14.3.2.
It should be noted that the value of md for the dynamic case in Fig.
14.36(b) is based on total reinforcement loads recorded along each
reinforcement layer during base shaking. The peak loads in dierent
reinforcement layers are not necessarily time coincident. The numerical
results discussed here have potential implications to conventional,
limit-equilibrium seismic design of reinforced soil walls with propped
panel wall facings. AASHTO (1998) and FHWA (1996) guidelines recom-
mend that a trapezoidal distribution be assumed for the total dynamic
Geosynthetic-reinforced soil walls and slopes 371
geometry parameters included the wall height and the backll width. The
intensity of the ground motion, characterized by peak ground acceleration,
was also varied.
The results of the analyses showed that equation (14.47) provided a
reasonable estimate for the fundamental frequency of reinforced-soil
retaining wall systems with wide uniform backll subjected to moderately
strong ground motion (e.g. g 0:2g in their study). Among the two-
dimensional approaches examined, the frequency formula proposed by
Wu and Finn (1996; 1999) gave the closest agreement to the fundamental
frequency value inferred from numerical results (Fig. 14.37(a)). The
fundamental frequency values from two-dimensional continuum models
were shown to approach values based on one-dimensional theory for
signicantly wide backll (e.g. B=H > 5 see Fig. 14.37(b)).
Earlier numerical simulation work by Bathurst and Hatami (1998b)
had demonstrated that reinforcement stiness, reinforcement length
and toe restraint condition could have a signicant inuence on the
magnitude of reinforcement forces and wall displacements of
reinforced-soil wall models during a simulated seismic event. However,
the results of the study by Hatami and Bathurst (2000) using the same
numerical models demonstrated that these variables did not signicantly
aect the fundamental frequency of reinforced-soil wall models with a
wide range of structural component values. Hatami and Bathurst
found the numerical results of model walls' fundamental frequency to
be less sensitive to the backll width compared to theoretical closed-
form predictions. They attributed the reason for the reduced eect of
the backll width partly to the soil plasticity in the near-eld behind
the facing panel which would reduce the geometrical eect of a purely
elastic backll on wall response.
14.4. Physical Model tests for seismic studies fall into two categories:
testing of model (a) reduced-scale shaking table tests
walls and slopes (b) centrifuge tests subjected to base shaking.
Both shaking table and centrifuge model tests share certain drawbacks,
among the most recognized of which are similitude and boundary eects.
Reference Model details Observed behaviour and implications to design and analysis
Koga et al., 1988; 1:01:8 m high models with vertical Deformations decreased with increasing reinforcement stiness
Koga and Washida, and inclined slopes at 1/7 scale. and density, and decreasing face slope angle. Failure volumes
1992 Sandbags with wrapped-face facing. were shallower for reinforced structures. Relative reduction in
Non-woven geotextile, plastic nets deformation of reinforced structures compared to unreinforced
and steel bars with sandy silt backll structures increased with steepness of the face. Circular slip
method agrees well with experimental results except for
steep-faced models
Murata et al., 1994 2:5 m high 1/2 scale model walls Increase in reinforcement forces due to shaking was very small.
with gabion/rigid concrete panel Reinforcement loads increased towards the front of the wall.
walls. Geogrid with dry sand Acceleration amplication was negligible up to mid-height of
backll. Horizontal shaking using wall but increased to about 1:5 at the top. Amplication
sinusoidal and scaled earthquake behaviour was similar for reinforced and unreinforced zones.
record. Base accelerations up to The reinforced zone behaved as a monolithic body. Sinusoidal
0.5g at 3:4 Hz base input resulted in greater deformations than scaled
earthquake record. Rigid facing adds to wall seismic resistance
Sugimoto et al., 1:5 m high model embankment with Reinforced models more stable than unreinforced. Proposed
1994; Telekes et al., sand bags and wrapped-face slope similitude rules for small and large strain deformation modelling.
1994 surface. Geogrid reinforcement Largest amplication recorded at crest of models. Failure of
with sand backll. Model scales 1/6 structures was progressive from top of structure downward.
and 1/9. Sinusoidal and scaled Reinforcement forces increased linearly with acceleration up to
earthquake record. Base start of failure. Failure mechanism dicult to predict using
acceleration up to 0.5g at 40 Hz proposed scaling rules. Under seismic loading conditions, there
was a tendency for shallow slopes to fail compared to steeper
ones. Scale eects due to vertical stress and apparent cohesion
of backll soil inuenced the relative performance of steep-faced
and shallow-faced models
Budhu and 0:72 m high model wall with Sliding progressed with increasing acceleration from the top
Halloum, 1994 wrapped-face facing. Geotextile geotextile/sand interface to the bottom layer. No consistent
with dry sand backll. Base decreasing trend of critical acceleration was observed with
acceleration in increments of 0.05g increasing spacing to length ratio. Critical acceleration
at 3 Hz proportional to the soil/geotextile interface friction value
Sakaguchi et al., 1:5 m high model walls. One Wrapped-face wall behaved as a rigid body and failed at a higher
1992; Sakaguchi, wrapped-face and four unreinforced acceleration than unreinforced structures. However, at smaller
1996 rigid concrete panel walls. Geogrid accelerations (due to sti facing panels) the displacements of the
with dry sand backll. Sinusoidal unreinforced structures were less. A base input acceleration of
loading with base acceleration up to 0:32 g delineated stable wall performance from yielding wall
0.72g at 4 Hz performance for the reinforced structure. Residual strains were
greatest closest to the face. Concluded that more rigid light-
weight modular block facings may be eective in reducing
reinforcement loads
Koseki et al., 1998b 0:50:53 m high propped-panel Overturning was observed to be the main failure mode. Simple
models, phosphor-bronze shear deformation of reinforced zone was observed. The ratio of
reinforcement strips (with observed and predicted critical seismic coecients
L=H 0:4) connected together in a (corresponding to 5% lateral displacement) was about 1:05 for
grid form. One uniform length uniform reinforcement model and 1:15 for the model with
model and one model with extended extended reinforcement layer length at the top. These ratio
reinforcement length at the top. 5 Hz values were larger than the values for conventional retaining
sinusoidal base acceleration with wall models (values less than one) tested in the same study.
stepwise increase in amplitude Walls on shaking tables were more stable than on equivalent
tilting tables. Observed failure plane angle was steeper than the
predicted value
Matsuo et al., 1998 11:4 m high models with hard Walls showed larger margin of safety when subjected to
facing panel. Reinforcement length, recorded ground motion compared to sinusoidal base
L=H 0:4 and 0:7. One model with acceleration. Did not observe failure of the model walls in spite
inclined facing. 5 Hz sinusoidal of predicted factors of safety that were less than 1
base acceleration with stepwise
increase in amplitude. In addition,
recorded ground motion was
applied
Geosynthetic-reinforced soil walls and slopes 375
1995) and lead to the requirement, in some cases, to increase the number
and length of reinforcement layers close to the top of reinforced wall
structures based on limit-equilibrium design.
Bathurst et al. (1996) and Pelletier (1996) have reported the results of a
series of shaking table tests that examined seismic resistance of model
reinforced segmental retaining walls. The tests were focused on the
inuence of interface shear properties on facing column stability, which
was identied as an important design consideration based on pseudo-
static methods of analysis (Bathurst and Cai, 1995). A set of 1/6 scale
model walls were constructed inside a plexiglas box and were 2400 mm
long by 1400 mm wide by 1020 mm high. Similitude rules proposed by
Iai (1989) were used to scale the model components and geometry. A
typical test conguration is illustrated in Fig. 14.38. The models were
constructed with concrete blocks 100 mm wide (toe to heel) by 160 mm
wide by 34 mm high. Five layers of a weak geogrid (HDPE bird fencing)
were used to model the reinforcement. The backll was a standard
laboratory silica #40 sand prepared at a relative density of 67%. The
four test congurations used are summarized in Table 14.3. The dier-
ences between the tests are related to interface shear capacity and wall
batter. Interfaces identied as frictional in Table 14.3 derive shear
capacity solely from sliding resistance at the interface. These interfaces
represent a very poor facing column detail with respect to shear capacity.
In two of the tests, the interfaces were xed at some locations in order to
simulate systems with high shear capacity at all or selected facing column
interfaces (i.e. positive interlock due to eective shear keys, pins or other
types of connectors). Each test was subjected to a staged increase in base
input motion resulting in the accelerationtime record shown in Fig.
14.39. The base input frequency was kept constant at 5 Hz. At the proto-
type scale, this frequency corresponds to 2 Hz.
The inuence of interface shear capacity and facing batter can be seen
in Fig. 14.40. The vertical wall with xed interface construction (high
shear capacity at each interface) required the greatest input acceleration
to generate large wall displacements during staged shaking (Test 4).
The vertical wall with poor interface shear at all facing unit elevations
performed worst (Test 1). However, the resistance to wall displacement
was improved greatly for the weakest interface condition by simply
increasing the wall batter (Test 3). The vertical wall with poor interface
properties only at the geosynthetic layer elevations (Test 2) gave a
displacement response that fell between the results of walls constructed
with uniformly poor interface shear properties (Test 1) and the nominally
identical structure with uniformly good interface shear properties (Test
4). The resistance of the facing column to horizontal base shaking
improved with increasing shear capacity between dry-stacked modular
blocks or by increasing the wall batter.
The results of this study conrmed that measured accelerations were
not uniform throughout the soil-wall system. Large acceleration ampli-
cations as high as 2:2 were recorded, particularly at the top of the
unreinforced portion of the facing column. Observed critical accelera-
tions to cause failure of the wall models were compared to predictions
based on the analysis method proposed by Bathurst and Cai (1995).
The measured peak acceleration at the middle wall height or at the top
of the backll surface was shown to give more accurate estimates of
critical acceleration to be used in pseudo-static analysis. The total load
in the reinforcement layers was estimated to be only a very small
percentage of the tensile capacity of the reinforcement layers. The test
results showed that, while critical accelerations to cause incipient collapse
of the wall models could be predicted reasonably well, the actual failure
Fig. 14.40.
Displacement close to top
of wall versus peak base
input acceleration
Geosynthetic-reinforced soil walls and slopes 377
Matsuo et al. (1998) carried out shaking table tests on six reduced scale,
reinforced soil walls models. They used a polypropylene material for the
geogrid reinforcement. The variables in the test walls were reinforcement
length (L=H 0:4; 0:7), wall height (H 1 m and 1:4 m), wall facing type
(incremental and propped) and facing slope (vertical and battered). The
input base acceleration was sinusoidal with a frequency of 5 Hz and
with a stepwise increase in amplitude. In addition, the NS component
of recorded ground motion at Kobe Maritime Observatory was applied
to one test model. The model walls showed larger margins of safety
when subjected to recorded ground motion compared to sinusoidal
base acceleration. Matsuo et al. (1998) suggested that predominant
frequency of the base accelerations also contributed to the dierence in
wall response magnitude to sinusoidal versus recorded base accelerations.
Matsuo et al. predicted the magnitude of wall horizontal displacement
subjected to recorded input acceleration using sliding block and cumula-
tive damage concepts. The predicted permanent displacement magnitude
from the sliding block approach was found to be only about a fourth of
the measured displacement. The predicted displacement magnitude from
the cumulative damage approach was about a fth of the measured value.
Matsuo et al. pointed to the eect of ground motion predominant
frequency (not included in the above approaches) and shear deformation
in the reinforced zone among the possible reasons for the dierence
between the predicted and measured values for the wall displacement in
their tests. They also observed that the model walls subjected to base
acceleration remained stable in spite of predicted factors of safety that
were less than unity. They attributed the stability of the test walls in
spite of low factors of safety to ductile behaviour of the walls. Matsuo
et al. found that increasing the reinforcement length ratio L=H from
0:4 to 0:7 was the most eective method to reduce the wall deformation.
In addition, the horizontal displacement at the top of the walls with a
continuous facing panel was greater than the corresponding displacement
in discrete facing walls. They found this result unexpected. However,
large lateral displacement at the top of propped-panel walls with xed
toe condition subjected to base acceleration has also been observed in
numerical simulation studies (e.g. Bathurst and Hatami, 1998b).
and recorded ground motions were used as the input base acceleration.
Nova-Roessig and Sitar found that the amplication of acceleration in
the backll depended on the amplitude of input base acceleration. They
measured acceleration amplication as great as 2:5 for 0:1g base accelera-
tion. The model slopes showed deamplication when they were subjected
to stronger (e.g. PGA > 0:35g) input accelerations. These results were
consistent with observations of Matsuo et al. (1998) on 1g shaking
table tests on walls with hard facing. Nova-Roessig and Sitar found
that the model slopes under base acceleration deformed in a ductile
manner with considerable amount of shear deformation near the crest
and with no distinct failure surface. This observation is also consistent
with the observations by Matsuo et al. (1998) for reinforced soil walls
on 1g shaking table tests (see Section 14.4.1). Nova-Roessig and Sitar
suggested that the lack of a well-dened shear failure surface in reinforced
soil slopes subjected to base acceleration contradicts the routine
assumption of a distinct failed mass behind the reinforced zone in limit-
equilibrium-based design methods. They proposed that deformation-
based approaches should be adopted for the seismic design of
reinforced-soil walls and slopes.
14.5. Seismic The generic term `geofoam' has recently entered geosynthetic terminology
buers to describe expanded foams used in geotechnical applications (Horvath,
1995). Horvath proposed that geofoam panels could be used against
rigid wall structures (e.g. basement walls) to reduce seismic-induced stresses
that would otherwise overstress rigid wall structures.
To the best of the authors' knowledge, the rst application of this
technology in North America was reported by Inglis et al. (1996).
Panels of low density expanded polystyrene (EPS) from 450 to 610 mm
thick were placed against rigid basement walls up to 9 m in height at a
site in Vancouver, British Columbia. Analyses using the FLAC program
showed that a 50% reduction in lateral loads could be expected (Fig.
14.41) during a seismic event compared to a rigid wall solution. The
design challenge using this technique is to optimize the thickness of the
buer panels for a candidate geofoam material so that the horizontal
compliance under peak loading is just sucient to minimize lateral
earth pressures without excessive lateral deformations. In addition, the
ideal properties of the geofoam are adequate compressive stiness
under static loading conditions but with a compressive yield plateau
that will just be exceeded under the design seismic lateral stresses.
Horvath has recognized that the technique described here may be an
economical solution to the problem of retrotting existing rigid wall
structures that do not satisfy modern seismic design codes.
14.6. Observed 14.6.1. North American experience (Northridge 1994 and Loma
performance of Prieta 1989)
reinforced soil Sandri (1994) conducted a survey of reinforced soil segmental retaining
walls greater than 4:5 m in height in the Los Angeles area immediately
walls and slopes after the Northridge Earthquake of 17 January 1994 (moment
during magnitude 6:7). The results of the survey showed no evidence of
earthquakes visual damage to nine of eleven structures located within 23113 km of
the earthquake epicentre. Two structures (Valencia and Gould Walls)
showed tension cracks within and behind the reinforced soil mass that
were clearly attributable to the results of seismic loading. Bathurst and
Cai (1995) analysed both structures and noted that minor cracking at
380 Geosynthetics and their applications
the back of the reinforced soil zone could be attributed to the attening
of the internal failure plane predicted using MO theory. The facing
columns for all walls were intact even though peak horizontal ground
accelerations as great as 0.5g were estimated at one site.
A similar survey of three geosynthetic-reinforced walls and four
geosynthetic-reinforced slopes by White and Holtz (1996) after the
same earthquake revealed no visual indications of distress. Stewart et al.
(1994) report that slope indicator measurements at the toe of a 24 m high
geogrid-reinforced slope, which was estimated to have sustained peak
horizontal ground accelerations of 0.2g, showed no movement. Some
unreinforced crib walls and unreinforced segmental walls were observed
to have developed cracks in the backll during the same survey by
Stewart et al. They concluded that concrete crib walls may not perform
as well as more exible retaining wall systems under seismic loading.
Similar good performance of several geosynthetic reinforced soil walls
and slopes during the 1989 Loma Prieta earthquake (Richter
magnitude 7:1) was reported by Eliahu and Watt (1991) and Collin
et al. (1992).
(a) The depth, strength and stiness of the foundation soil may have
a greater inuence on the internal and external stability of
reinforced soil slopes and walls than the design of the reinforced
mass in isolation. Parametric analyses are required to investigate
the inuence of the foundation condition on seismic performance.
(b) The design methodologies that are currently used in the United
States for geosynthetic-reinforced soil walls have been based
largely on the results of numerical modelling of reinforced
structures constructed with inextensible reinforcement (steel
strips). Recent numerical studies by the authors conrm that
the general approach is not valid for reinforced soil wall structures
constructed with relatively less sti geosynthetic products.
Further numerical and experimental work is required to investi-
gate the validity of pseudo-static analysis methods that predict
increased reinforcement lengths at the top of reinforced walls
and slopes.
(c) Ground motion amplication (or attenuation) through retained
soils plays a major role in generating additional dynamic loads
on geosynthetic reinforcement and wall-facing components.
More work is required to oer guidance on the appropriate distri-
bution of incremental seismic forces to be applied to extensible
reinforcing elements and to establish the inuence of system sti-
ness (i.e. the combined eect of reinforcement stiness, number of
reinforcement layers, facing stiness and height of structure) on
this distribution. Numerical models calibrated against the results
of carefully conducted large shaking table tests or small-scale
centrifuge tests are possible research strategies to meet this goal.
(d) The single most important characteristic determining the seismic
response of reinforced soil walls is the fundamental frequency
of the structure, namely the predominant frequency of the
design seismic event. The calculation of the fundamental
frequency of a reinforced wall structure in a seismic area should
be part of the analysis and design process. Simple expressions
are available to carry out this evaluation.
(e) A number of design methodologies have been proposed in the
United States and Japan for the seismic design of walls and
slopes that can lead to important dierences in the required
number/strength, location and length of reinforcement layers.
Comparative analyses should be carried out to examine the
relative conservatism (or non-conservatism) of the proposed
methodologies.
( f ) Geosynthetic-reinforced segmental retaining walls in seismic
areas oer unique challenges to the designer because of their
modular facing column construction. These structures involve
analyses not required for other retaining wall systems. The
experience of the authors is that the economic potential of these
systems in seismic areas will not be fully realized until condence
is developed through proven design methodologies for these
structures.
(g) The design engineer will continue to be attracted to relatively
simple seismic design tools based on pseudo-static and displace-
ment methods for the design and analysis of routine walls and
slopes under modest seismic loads. Nevertheless, the results of
sophisticated numerical models carried out by experienced
modellers oer the possibility of rening simple models to mini-
mize unwarranted conservatism.
Geosynthetic-reinforced soil walls and slopes 383
14.7.1. Acknowledgements
The funding for the work reported in this chapter was provided by the
Department of National Defense (Canada) through an Academic
Research Program (ARP), Directorate Infrastructure Support (DIS/
DND) and Natural Sciences and Engineering Research Council of
Canada. The authors thank Professors H. Ochiai, R. D. Holtz, T.
Akagi, F. Tatsuoka, J. DiMaggio and J. Nishimura for the provision
of many useful references, and Professor S. L. Kramer for permission
to publish results of FLAC analyses carried out at the University of
Washington, USA. The contribution of former post-doctoral research
associates Dr Z. Cai and Dr M. Yogendrakumar to the research program
at RMCC is also gratefully acknowledged as are the eorts of former
graduate students M. McLay and M. Pelletier. The authors would also
like to thank M. Simac and T. Allen for many fruitful discussions on
the general topic of segmental walls.
References
Allen, T. M. (1993). Issues regarding design and specication of segmental block-
faced geosynthetic walls. Transportation Research Record, 1414, 611.
AASHTO (1998). Interims: Standard specications for highway bridges. American
Association of State Highway and Transportation Ocials, Washington, DC,
USA.
ASTM (1996). Designation D4595: Standard test method for tensile properties of
geotextiles by the wide-width strip method. 1996 Annual Book of ASTM
Standards, Section 4, Construction, (4.09), American Society for Testing and
Materials, West Conshohocken, Pennsylvania, USA.
Bachus, R. C., Fragaszy, R. J., Jaber, M., Olen, K. L., Yuan, Z. and Jewell, R.
(1993). Dynamic response of reinforced soil systems. Engineering Research
Division, US Department of the Air Force Civil Engineering Support Agency,
March 1993, 1 & 2, Report ESL-TR-92-47.
Bathurst, R. J. (1994). Reinforced soil slopes and embankments. Technical Notes
for Computer Programs GEOSLOPE and GEOPLOT.
Bathurst, R. J. (1998). NCMA segmental retaining wall seismic design procedure
supplement to design manual for segmental retaining walls. National Concrete
Masonry Association, Herdon, Virginia, USA.
Bathurst, R. J. and Alfaro, M. C. (1996). Review of seismic design, analysis
and performance of geosynthetic reinforced walls, slopes and embankments.
Proceedings of the Earth Reinforcement International Symposium on Earth
Reinforcement. Fukuoka, Kyushu, Japan, pp. 887918.
Bathurst, R. J. and Cai, Z. (1994). In-isolation cyclic load-extension behavior of
two geogrids. Geosynthetics International, 1, No. 1, 317.
Bathurst, R. J. and Cai, Z. (1995). Pseudo-static seismic analysis of geosynthetic-
reinforced segmental retaining walls. Geosynthetics International, 2, No. 5, 787
830.
Bathurst, R. J., Cai, Z. and Pelletier, M. J. (1996). Seismic design and perfor-
mance of geosynthetic reinforced segmental retaining walls. Proceedings of the
10th Annual Symposium of the Vancouver Geotechnical Society. Vancouver,
British Columbia, Canada.
Bathurst, R. J., Cai, Z. and Simac, M. R. (1997). Seismic performance charts for
geosynthetic reinforced segmental retaining walls. Proceedings of the Geosyn-
thetic '97. Long Beach, California, USA, pp. 10011014.
Bathurst, R. J. and Hatami, K. (1998a). Inuence of reinforcement stiness,
length and base condition on seismic response of geosynthetic reinforced
384 Geosynthetics and their applications