Interpretation of dynamic retaining wall model tests in light of elastic and

plastic solutions
Christos Giarlelis
n
, George Mylonakis
Department of Civil Engineering, University of Patras, 26500 Rio, Greece
a r t i c l e i n f o
Article history:
Received 14 January 2010
Received in revised form
27 June 2010
Accepted 5 July 2010
a b s t r a c t
The dynamic response of rigid and exible walls retaining dry cohesionless soil is examined in light of
experimental results and analytical elastodynamic and limit analysis solutions. Following a brief review
of the problem, experimental ndings from three different testing programs on retaining walls are
presented, and compared with theoretical predictions based on the above-mentioned approaches.
Reasonable agreement is found depending on the assumptions. It is shown that wall exibility which
is not taken into account in classical design approaches should be considered to establish the point of
application of seismic thrust on the wall. Detailed calculations and set of graphs and charts are
presented, which highlight salient aspects of the problem.
& 2010 Elsevier Ltd. All rights reserved.
1. Background
Despite extensive research carried out over the years, the
dynamic behavior of retaining structures is far from being well
understood. Signicant unresolved issues, both of practical and
theoretical nature persist, which appear to have impeded
advances in design methods and seismic regulations. These
include: (1) dependence of soil thrust on magnitude of earthquake
acceleration; (2) importance of wall kinematics on the distribu-
tion of earth pressures; (3) importance of the dynamics of the
backll; (4) description of boundary condition pertaining to wave
radiation away from the wall; (5) role of pore water pressure and
associated loss of strength behind and under the wall; (6)
importance of pre-existing and stress-induced inhomogeneities
in the soil; (7) inuence of construction processes on the above.
Evidence of a lack of understanding comes from post-earthquake
investigations (notably in Kobe [1] and Chi-Chi [2]) which have
reported extensive damage on a number of retaining structures,
which were thought to have been properly designed against
seismic action.
Past analyses of dynamic response of earth-retaining systems
can be roughly classied into two main groups: (a) limit-state
analyses, in which the wall is considered to displace and/or rotate
sufciently at the base to fully mobilize the shearing strength of
the backll and (b) elastic analyses, in which the wall is
considered to be xed at the base, while the backll is presumed
to respond in a linearly elastic or viscoelastic manner. Represen-
tatives of the rst group is the classical MononobeOkabe (MO)
approach [3,4] and its variants (Seed and Whitman [5], Richards
and Elms [6], Nadim and Whitman [7], Dubrova [8]), which have
found widespread acceptance in practice (e.g., ATC [9], EC-8 [10]).
Representatives of the second group are the contributions of
Matsuo and Ohara [11], Wood [12,13], Arias et al. [14] and
Veletsos and Younan [1519].
In a large number of reports (a list of references is available in
Giarlelis [20]), results from experimental studies are presented in
an effort to evaluate the predictions of theoretical analyses
mainly the ones in the rst group. However, little effort has been
put in interpreting experimental results in light of elastic
solutions. This is potentially important as elastic solutions
incorporate the effect of wall kinematics, which is missing from
the limit analysis solutions. This need provided the initial
motivation for the herein-reported work.
Following a review of limit-state analyses, a brief presentation
is made of the available elastic solutions, especially the ones
developed by Arias et al. [14] and later extended by Veletsos
and Younan [1519]. The article then focuses in comparing
experimental results against the predictions of these solutions.
The comparisons are carried out by means of two key parameters:
(1) the magnitude of the force exerted on the wall by the soil
(base shear); (2) the point of application of this force
(overturning moment). It is shown that, while the estimation
of these parameters under limit-state analyses is problematic
particularly regarding the point of application of the force this is
not the case with the elastic solutions. More importantly,
theoretical predictions based on elastic solutions appear to
compare better to experimental results than those based on
limit-state analyses, as demonstrated in the ensuing.
Contents lists available at ScienceDirect
journal homepage: www.elsevier.com/locate/soildyn
Soil Dynamics and Earthquake Engineering
0267-7261/$ - see front matter & 2010 Elsevier Ltd. All rights reserved.
doi:10.1016/j.soildyn.2010.07.002
n
Correspondence address: 9 Kariotaki, Athens 11141, Greece.
E-mail address: giarlelis@alumni.rice.edu (C. Giarlelis).
Soil Dynamics and Earthquake Engineering 31 (2011) 1624
2. Limit-state analyses
For over 80 years, the familiar MO formula has dominated the
design of retaining structures for earthquake loading. An exten-
sion of Coulombs theory for gravitational loading, the method
makes use of pseudostatic inertial forces in the backll to account
for the effects of the horizontal and vertical components of
ground shaking. The resulting accelerations are assumed to be
constant within the soil mass, while the wall is presumed to
be capable of yielding sufciently to allow active conditions in the
backll. A soil wedge bounded by the wall and a failure plane in
the backll, will, therefore, start moving towards the wall,
inducing an active thrust on the latter.
For a vertical wall retaining horizontal backll and in the
absence of cohesion or surcharge, the MO solution is given by
the well-known formula [38]:
K
AE

2P
AE
gH
2
1k
v

cos
2
fc
e
_ _
cosc
e
cos dc
e
_ _
17

sin df sin fc
e
cos dc
e
_ _ _
2
1
where P
AE
denotes the combined seismic-gravitational active limit
thrust on the wall (unitsF/L) and K
AE
is the corresponding earth
pressure coefcient. In the above representation and hereafter, d
denotes the roughness of the wallsoil interface, f denotes the
soil friction angle, g the unit weight of the soil material and H the
height of the wall. The angle
c
e
tan
k
h
1k
v
_ _
1
2
is the inclination of the resultant pseudostatic body force in the
soil, k
h
and k
v
being the seismic body forces in the horizontal and
vertical directions, respectively.
Following the case of gravitational loading under Rankine
conditions, the distribution of dynamic pressures is assumed to
vary proportionally with depth, resulting in a point of application
of the dynamic thrust, known as the effective height, at 1/3 of wall
height.
Early experiments by Mononobe and Matsuo [3] conrmed the
magnitude of the forces predicted by the MO analysis. Later
investigations by Jacobsen [21], Matsuo [22], Ishii et al. [23] as
well as the more recent experiments by Sherif et al. [2426], have
led to similar conclusions.
Unlike its magnitude, however, the point of application of
dynamic thrust assumed in the MO method has not shared
comparable acceptance by researchers. Experiments by Jacobsen
[21] and later by Matsuo [22], showed that the dynamic
component of pressures acts at about 2H/3, while Ishii et al.
[23] found that the point of application of the total force may
well be at 0.4H, a result that was also advocated by Sherif et al.
[2426]. Prakash and Basavanna [27] demonstrated analytically
that the dynamic pressure distribution is denitely not hydro-
static, and that it may well be parabolic. Current practice for the
effective height is based on the widely accepted proposal by Seed
and Whitman [5] where the point of application of the dynamic
component of soil thrust is considered to be at 0.6H. This result
was based partly on the ndings of Prakash and Basavanna [27]
and the experiments referred above, and partly on the elastic
solutions derived by Wood [12,13] for non-yielding rigid walls.
The above practice is also adopted by ATC [9] while in EC-8 [10],
the point of application of the total force is suggested to be at
0.5H.
In all aforementioned studies, no account has been taken of the
inertia of the wall itself, and no attempt is made to estimate the
magnitude of the resulting dynamic wall displacements. Both
were rst considered by Richards and Elms [6], who, by making
use of Newmarks sliding block concept [28], provided the means
for relating the wall displacement to the level of the horizontal
ground shaking. Their approach permits design of retaining walls
for controlled displacements and, along with the original MO
formula, provides the basis of all modern guidelines (ATC [9], EC-8
[10]). A more accurate MO type formula, based on stress elds,
has been recently proposed by Mylonakis et al. [29]. For the
aforementioned conditions, this solution is expressed by
K
AE
1k
v

1
cosd
1sinf cos D
2
d
1sinf cos D
1
c
e
_ _
_ _
exp 2y
E
tanf 3
where
2y
E
D
2
7 D
1
*
d
_ _
c
e
, D
1
*
sin
1
sinc
e
sinf
_ _
, D
2
sin
1
sind
sinf
_ _
426
In the above equations, y
E
, D
1
, D
2
are problem angles
(measured in radians). The rest of the symbols have the same
meaning as before. A more complete version of Eq. (3) encom-
passing active and passive conditions as well as inclined walls,
inclined backll and surcharge is provided in [29]. An interesting
discussion (and an independent derivation of Eq. (3) for passive
conditions) can be found in Lancellotta [30].
It should be noted that the true dynamic behavior of the soil
wall system is not adequately captured in limit-state solutions,
since these are based on the assumption of constant acceleration
in the backll and neglect the wave propagation in the soil
medium. Furthermore, the wall in these solutions is presumed to
be rigid and capable of yielding sufciently to induce an active
state of stress in the backll. These approaches are thus not
applicable to walls that do not satisfy the indicated assumptions.
It is also worth mentioning in this context that the limit of
sufcient yielding for dynamically excited systems, especially for
tilting walls, is not particularly clear. Accordingly, the basic
assumption of an active condition may not be fully realized and
the resulting wall forces may be higher than those predicted by
the MO method.
In design practice, it is generally accepted that the limitations
and weaknesses of the limit-state methods of analysis can be
compensated by a wise choice of the parameters involved and the
selection of an appropriate factor of safety. However, without
the use of rational formulations that may better dene the
response of such systems such as the elastic solutions, it is
difcult to determine the factor of safety needed to compensate
for the inaccuracies that may result from applying limit-state
methods.
3. Elastic analyses
In an attempt to gain insight into the response of soilwall
systems for which the assumptions underlying the limit-state
methods are not satised, many researchers resorted to elastic
solutions. Early elastic analyses dealt with non-yielding rigid
walls. Matsuo and Ohara [11] arrived at a formula that determines
the dynamic earth pressure on the wall by solving an elastic wave
equation. However, the accuracy of their analysis could not be
conrmed. Wood [12,13] provided analytical solutions and
comprehensive numerical data for the response of a stratum of
nite length excited uniformly along its base and its two vertical
boundaries. His analysis was implemented by modal super-
position and led to rather complicated expressions involving
double trigonometric series which are not suitable for routine use
by engineers.
C. Giarlelis, G. Mylonakis / Soil Dynamics and Earthquake Engineering 31 (2011) 1624 17
Scott [31] proposed a simple model in which the wall is
considered to be connected by a series of massless horizontal
elastic springs to a vertical cantilever shear-beam that represents
the action of the soil at the far-eld. However, this model ignores
the radiation damping capacity of the medium as well as the
ability of the backll to transfer forces by horizontal shearing
action to the base and thus does not adequately describe the
action of the system. Arias et al. [14] used a simplied
representation of the elastic medium and provided relatively
simple analytical expressions for the wall pressures induced by
both harmonic and seismic excitations. However, their analyses
were restricted to rigid walls.
More recently, Veletsos and Younan [15,16], presented simple
approximate expressions for the response of soilwall systems
with non-yielding rigid walls retaining an elastic soil medium. In
a next step [1719], the same authors took into account the effect
of wall exibility on dynamic pressures and associated forces and
moments. The accuracy of these solutions has been recently
veried by nite element analysis by Wu and Finn [32] and
Psarropoulos et al. [33]. Theodorakopoulos et al. [34,35] extended
these solutions for unyielding rigid walls retaining a saturated
poroelastic soil backll.
By denition, elastic solutions do not account for the true
nonlinear hysteretic behavior of the soil. Similarly, they are not
applicable to walls that can slide. However, apart from their direct
applicability to the cases where elastic conditions do prevail,
these solutions may provide a safe upper limit for the wall
pressures and highlight the role of the various factors that
inuence the problem. Even for walls that can slide, the elastic
solutions are valuable in identifying wall pressures and forces
associated with initiation of sliding. So when one chooses to
use a limit-state analysis for a certain problem, it would be
desirable to compare the results against those obtained by an
elastic analysis.
4. Elastic analyses by Veletsos and Younan
The brief description that follows presents only the elements
of these analyses necessary for the comparisons with the
experimental studies. More detailed information regarding the
methodology and the theoretical background can be found in
the original publications of Veletsos and Younan [1519] and also
in Wu and Finn [32] and Li [36].
4.1. Soilwall model
The system examined (Fig. 1) consists of a semi-innite,
uniform layer of viscoelastic material of height H, that is free at its
upper surface, bonded to a rigid base and retained along its
vertical boundary by a uniform, exible cantilever wall that is
elastically constrained against rotation at its base by means of a
rotational constraint of stiffness R
y
. The bases of both the wall and
the soil stratum are presumed to experience a space-invariant
horizontal motion, the acceleration of which at any time is x
g
t,
and its maximum value is

X
g
. Material damping for the medium is
considered to be of the constant hysteretic type.
The properties of the soil stratum are dened by its mass
density r
s
, shear modulus of elasticity G, Poissons ratio n and
material damping factor d, which is considered to be frequency-
independent and the same for both shearing and axial deforma-
tions. The properties of the wall are dened by its thickness t
w
,
mass per unit of surface area m
w
, Youngs modulus of elasticity E
w
,
Poissons ratio n
w
and material damping factor d
w
, which is
considered to be the same for both the wall in exure and the
rotational base constraint.
The primary parameters governing the response of the system
are the relative exibility of the wall and retained medium
dened by
d
w

GH
3
D
w
7
and the relative rotational exibility
d
y

GH
2
R
y
8
where R
y
denotes the modulus of the rotational spring at the base
of the wall. The parameter D
w
, in Eq. 7, represents the exural
rigidity per unit of length of the wall in plane strain conditions,
dened as
D
w

E
w
I
w
1n
w
2

E
w
t
w
3
121n
w
2

9
Also affecting the response are the characteristics of the base
motion. For harmonic excitation, the response is controlled by the
frequency ratio o/o
1
, obeing the cyclic frequency of base motion
(and steady-state response) and o
1
the fundamental cyclic
natural frequency of the stratum considered to act as an
unconstrained vertical shear-beam. For an arbitrary transient
excitation, the relevant stratum property is its fundamental
natural frequency f
1
o
1
/2p, or the corresponding period
T
1
2p/o
1
.
In the particular case of a rigid wall, the VY solution can be
cast in the following form [15]:
DK
AE

2 P
AE
P
A

gH
2
k
h
32c
s
p
3

1id
_
1
n 1,3,5:::
1
n
3
1

1f
2
n
id
_
10
and
h
H

2
p

1
n 1,3,5:::
1
n1=2
n
3
1

1f
2
n
id
_
_
_
_
_
_
_

1
n 1,3,5:::
1
n
3
1

1f
2
n
id
_
_
_
_
_
_
_
1
11
where
c
s

1n2n
_ and f
n

1
n
o
o
1
12; 13
are dimensionless functions. In the above equations, h stands for
the point of application of seismic thrust measured from the base
of the wall. A more general, yet more complex, solution
encompassing wall exibility is provided in Refs. [18] and [19].
It should be noted that the above solution expresses solely the
dynamic component of seismic thrust, i.e., the difference between
the overall seismic active thrust and the active thrust due to
gravitational action. Fig. 1. Soilwall system considered in the elastic solutions of VeletsosYounan.
C. Giarlelis, G. Mylonakis / Soil Dynamics and Earthquake Engineering 31 (2011) 1624 18
4.2. Static response
For the comparisons with the test results made in the
following sections, only the responses obtained for excitations
for which the dominant frequencies are extremely small
compared to the fundamental frequency of the soilwall system
(i.e., for values of o/o
1
-0 or f
1
-N) are of importance. Such
low-frequency excitations and the resulting effects are referred to
as static and are identied with the subscript st. This term
should not be confused with that normally used to represent the
effects of gravity forces. With the static response determined, the
maximum value of a dynamic effect for an arbitrary transient
excitation is expressed as the product of the corresponding static
effect and an appropriate dynamic response (i.e., amplication or
deamplication) factor.
Tables 1A and B present normalized values of the static base
shear (Q
b
)
st
, and of the height, h, from the base to the centroid of
the wall force for different values of the relative wall exibility
parameter, d
w
, and the rotational base exibility factor, d
y
,
respectively. For the calculation of these values, the wall is
considered to be massless (i.e., m
w
0), Poissons ratio for the soil
is taken as n1/3 and the damping factors for the soil and wall are
taken as D0.1 and D
w
0.04, respectively (i.e., as 5% and 2% of
critical damping). In Table 1A, d
y
0 and in Table 1B, d
w
0.
Considering that

X
g
k
h
g and that grg then
r

X
g
H
2
k
h
gH
2
14
Hence, static base shear can be dened in terms used in the
equivalent expression of MO.
As it can be noticed from both tables, the point of application
of the dynamic thrust, the effective height, lies in a wide range of
values starting from 0.6H for an unyielding wall and going even
below H/3 for a very exible one.
In the particular case of a rigid wall, the static solution
simplies to
DK
AE
k
h
15
which suggests that the normalized seismic thrust is practi-
cally equal to the inertial force acting in a square region of soil of
unit thickness and height equal to that of the wall.
The corresponding point of application of the force is 0.625H
(i.e., h/H0.625).
5. Comparisons with experimental results
In a large number of reports [20], the results of experimental
studies have been presented in an effort to verify the predictions
of theoretical analyses. The majority of these studies were
orientated to the conrmation of limit-state analyses, so in most
cases valuable information for the calculation of some of the soil
wall system parameters (the exural rigidity of the wall, E
w
I
w
, and
the shear modulus of the soil, G) used in the elastic analyses of
VY is missing. In other cases the appropriate conditions for the
application of elastic solutions are not satised (the wall base is
not xed against translation). From most of the available reports,
the above two requirements are fullled in the works of Sherif
et al. [2426], Bolton and Steedman [37] and Andersen et al. [38]
(also in Al-Homoud et al. [39]). In the following sections the
results of the elastic and plastic solutions will be compared to
those of the above three studies.
5.1. Experiments by Sherif et al.
Sherif et al. [2426] carried out shaking table tests on an
unyielding wall, considered to be rigid, retaining dry dense sand.
The measured quantities were the total dynamic force on the wall
and its point of application and both were compared to the ones
calculated by the elastic solutions of Matsuo and Ohara [11] and
Wood [12,13]. The main conclusion was that elastic solutions
overestimate the dynamic thrust. However, it can be derived from
their report (e.g., Fig. 16) that this applies mainly to the solutions
of Matsuo and Ohara and not to Wood s solution. In addition, as it
will be shown in this work, the wall used is not perfectly rigidso
the comparison with both these elastic solutions that do not take
into account the effect of relative wall exibility is not appro-
priate.
The ratio of soil length to wall height in the experiment,
L/H1.8, is not sufciently large to consider the soil as semi-
innite, as required for the application of the elastic solutions of
VY. To overcome the problem, the effect of wall exibility from
these solutions will be combined with Woods solution for
unyielding walls, which takes into account the ratio of L/H. This
can be performed by applying a pertinent correction factor, which
is determined from the analysis presented below.
The wall used was made of structural aluminum, had a
thickness of 127 mm (5 in.) and retained a soil stratum of
1020 mm (3 ft, 4 in.) height. The soil density was r1.66 Mg/m
3
leading to unit weight g16.3 kN/m
3
(104 lb/ft
3
). The friction
angle of the soil was measured at 411 and that of the soilwall
interface at 23.51. The input motion was sinusoidal, with a
frequency of 3.5 Hz and an acceleration coefcient varying from
k
h
0.10.5. The structural properties of the wall and the shear
modulus of the soil are not provided, but can be estimated from
the available information.
The modulus of elasticity for aluminum can be considered
7110
6
kN/m
2
(1.4910
9
psf) and its Poissons ratio equal to
1/3. The shear modulus of the soil can be computed from the
Table 1
Normalized values of static base shear (Q
b
)
st
(inertial component, for o/o
1
-0, in
addition to gravitational) and of effective wall height, h, for massless walls with
different exibilities, d
w
and d
y
, respectively, [1519].
Part A Part B
d
w Q
b

st
=r

XgH
2 h/H d
y Q
b

st
=r

XgH
2 h/H
0 0.940 0.599 0.0 0.940 0.599
1 0.838 0.553 0.5 0.693 0.552
2 0.770 0.517 1.0 0.566 0.512
3 0.720 0.488 1.5 0.489 0.477
4 0.683 0.464 2.0 0.437 0.447
5 0.653 0.444 2.5 0.399 0.420
6 0.628 0.426 3.0 0.371 0.396
7 0.607 0.411 3.5 0.349 0.375
8 0.590 0.398 4.0 0.331 0.356
9 0.574 0.386 4.5 0.317 0.339
10 0.561 0.376 5.0 0.305 0.323
12 0.537 0.358 5.5 0.294 0.309
14 0.518 0.343 6.0 0.286 0.296
16 0.502 0.331 6.5 0.278 0.284
18 0.488 0.320 7.0 0.271 0.273
20 0.475 0.311 7.5 0.265 0.263
25 0.450 0.292 8.0 0.260 0.253
30 0.429 0.278 9.0 0.251 0.236
35 0.412 0.267 10.0 0.244 0.221
40 0.397 0.257 11.0 0.238 0.208
45 0.384 0.249 12.0 0.233 0.196
50 0.373 0.242 13.0 0.228 0.186
55 0.363 0.236 14.0 0.224 0.176
60 0.353 0.231 15.0 0.221 0.168
80 0.324 0.214 20.0 0.209 0.135
C. Giarlelis, G. Mylonakis / Soil Dynamics and Earthquake Engineering 31 (2011) 1624 19
empirical expression of Seed and Idriss [40]
G 1000K
2

s
o
_
psf 16
where s
o
is the mean effective stress in the medium, i.e., 7 kN/m
2
(150 psf). Regarding K
2
, a value of 65 can be adopted based on the
high density of the sand, as inferred from the unit weight and the
friction angle. This value need not be reduced given the small
levels of strain in the medium during dynamic excitation (average
shear strain at 0.5g maximum acceleration of the order of 10
5
).
The shear modulus is estimated from Eq. (16) at 38 MPa
(810
5
psf).
The natural frequency of the stratum can be evaluated from
the formula
f
1

1
4H

G
r

17
which results in f
1
37 Hz. As the imposed frequency of 3.5 Hz is
very small compared to the fundamental frequency of the
stratum, the response can be safely considered to be static.
Considering a reasonable variation of about 720% in the value
of G that is a range from 30 to 46 MPa, the relative exibility of
the wall and the retained medium, d
w
, as calculated from Eq. (7),
varies from 2.3 to 3.5 so the system is denitely not rigid.
From Table 1A, the normalized values of static base shear are
found to vary from 0.75 to 0.7. For d
w
0, the corresponding
normalized value is 0.94, so the correction factor due to the
system exibility ranges from 0.75/0.940.8 to 0.7/0.940.75,
respectively. This suggests that the experiment results
should have been compared to values 2025% smaller than
the ones found from Woods elastic solution for an unyielding
wall.
In Fig. 2, the original and corrected values the shaded area
bounded by two lines dening the limits of correction of Woods
solution based on the VY results is presented. While the
corrected results are still higher than the ones found in
the experiment, the agreement between records and predictions
has been improved, due to the effect of wall exibility. The
predictions of the Mylonakis et al. [29] stress solution (Eq. (3)) are
also plotted for comparison.
Regarding the point of application of the dynamic thrust, Sherif
et al. report that it was found to be slightly rising with increase in
level of ground shaking, from 0.51H for k
h
0.1 to 0.53H for
k
h
0.3 and to 0.54H for k
h
0.5. A possible explanation, in the
realm of the present analysis, for this rise is that, as the intensity
of the shaking increases, higher strains develop in the backll so
effective soil shear modulus decreases. In light of Eq. (7), the value
of d
w
drops, leading to an increase in effective height according to
Table 1A.
In the absence of more detailed information on dynamic
strains, the reduction in shear modulus and the associated
elastic parameters cannot be safely determined. So using the
elastic solutions, a constant value in the range of 0.480.51H
corresponding to d
w
varying from 3.5 to 2.3, respectively, can be
estimated from Table 1A. Obviously, values in that range are
comparable to the observed ones.
Looking at the variation in the dynamic thrust and moment
with k
h
, the elastic solution appears to provide a linear correlation
while the measured results present a hyperbolic-like correlation.
The linear dependence should be viewed as a range of values
bounded between the upper and lower band estimates for G of
46 and 30 MPa. As mentioned previously, as the intensity of the
shaking increases, higher strains develop in the backll so
effective soil shear modulus decreases and the system becomes
stiffer. So the actual variation is likely to be nonlinear as shown
qualitatively in Figs. 2 and 3 by a dashed line. It is fair to mention
that this variation is weaker than the observed one.
In Fig. 3, the dynamic moments calculated from the elastic,
limit state and stress solutions are compared to the ndings
of Sherif et al. This comparison is more favorable to the
elastic solutions than the previous one, due to the good
agreement between theory and experiments regarding the point
of application of the dynamic thrust.
5.2. Experiments by Bolton and Steedman
Bolton and Steedman [37], performed a series (RSS30) of
centrifuge tests on xed-based aluminum cantilever walls
retaining dry sand backll in a 80g eld measuring dynamic and
total base moments. The input acceleration consisted of a
sequence of 10 similar, nearly sinusoidal cycles of approximately
the same dominant frequency of 80 Hz, as concluded from their
report. They compared the measured moments with those
Fig. 2. Comparison between the ndings of Sherif et al. and the results of the elastic, limit state and stress solutions regarding the dynamic thrust.
C. Giarlelis, G. Mylonakis / Soil Dynamics and Earthquake Engineering 31 (2011) 1624 20
obtained from the MO equation considering the dynamic
component of the wall force to act at H/3 and not at 0.6H as
suggested by Seed and Whitman [5] and making due provision
for the dynamic moment induced by the self-weight of the wall.
They conclude that the comparison shows good agreement. In this
section it will be shown that there is also good agreement
between the experimental results and the analytical solutions for
exible cantilever walls of VY, and that the drop in the point of
application of dynamic thrust, as suggested by Bolton et al., stems
from the exibility of the wall.
The xed cantilever aluminum wall had a height of 175 mm.
The backll was very dense sand with a relative density of
D
r
96%, r1.77 Mg/m
3
and a friction angle of approximately
501. The structural properties of the wall (E
w
,I
w
) and the shear
modulus of the soil were not provided, so they had to be
estimated by the authors.
In their report, Bolton and Steedman present the variation in
the top deection of the wall as the centrifuge acceleration
increases from 1g to 80g. They compare the experiment results to
the ones predicted by the MO method and calculate the value of
K
AE
cos d. Using this value, the quantity E
w
I
w
can be computed as
follows:
For hydrostatic distribution of wall pressures, as assumed by
Bolton and Steedman, the deection of the wall is given by the
easy-to-derive formula
u
o

q
o
H
4
30E
w
I
w
18
where q
o
is the pressure at the bottom.
The base shear is given by
Q
b

1
2
gH
2
K
AE
cosd
1
2
q
o
H 19
where K
AE
is the well-known parameter of the MO formula and d
is the soilwall friction angle. The value of K
AE
cos d as reported by
Bolton et al. is 0.12, while the top deection of the wall, u
o
,
measured in their report is 0.22 mm. From the soil density of
r1.77 Mg/m
3
, the unit weight of the soil in the eld of 80g is
gr(80g)1390 kN/m
3
. Then from Eq. (19), it follows that
q
o
30.10 kN/m
2
and from Eq. (18) that E
w
I
w
4.23 kN m
2
/m.
Poissons ratio for aluminum is practically equal to 1/3. So from
Eq. (9), D
w
4.75.
Shear modulus can be evaluated from Eq. (16) for a mean
effective stress of s
o
110 kPa. The parameter K
2
can be
determined from the SeedIdriss diagram [40], as function of
the average shear strain in the backll, g, for different values
of relative density. The value of g can be related to the deection
at the top of the wall through the simple expression [20]
g %u
o
=H 20
Since the maximum top wall deection u
o
0.22 mm, it
follows that gE10
3
; using the diagram K
2
17 and then the
shear modulus G can be estimated at about 38 MPa.
The fundamental frequency of the stratum can be evaluated
from Eq. (17) as f
1
210 Hz at 80g. The ratio of the imposed
frequency to the fundamental one is, thereby, f
e
/f
1
80/2100.38,
so for all practical purposes the response is static.
The relative wall exibility is determined from Eq. (7) to be
d
w
43.1 (a very exible system). From Table 1A, the correspond-
ing value of normalized base shear for a massless wall is about
0.39. Then the maximum dynamic base shear increment is
DQ
b
0:389gk
h
H
2
21
applied at a distance 0.25H from the base. The results of the above
equation and the moments induced at 0.05H from the base for
each of the three experiments are tabulated in Table 2.
Bolton and Steedman compare their results to those obtained
by the MO method with the point of application of the dynamic
thrust at H/3. In addition, they take into account the dynamic base
moment induced by the wall inertia (i.e., M
b
0.5g
a
k
h
d
a
H
2
) where
g
a
is the unit weight of the wall material and d
a
the wall thickness
(corresponding to t
w
of VY). This results in additional dynamic
moments of 53, 69 and 67 N mm/mm for the three experiments,
so the total dynamic moments are 136, 184 and 178 N mm/mm,
respectively. It is noted, however, this procedure does not account
for the effect of soilstructure interaction and overestimates the
effect of wall inertia.
In Table 2 the results of the elastic solutions and Mylonakis
et al. [29] stress solution are compared to the values reported by
Bolton and Steedman. Also, the dynamic moments calculated by
the MO method with point of application taken at 0.6H, as
suggested by Seed and Whitman [5] and also at H/3 as originally
proposed by MO and also advocated by Bolton and Steedman,
are tabulated. The comparison is also presented in Fig. 4.
Fig. 3. Comparison between the ndings of Sherif et al. and the results of the elastic, limit state and stress solutions regarding the dynamic moment.
C. Giarlelis, G. Mylonakis / Soil Dynamics and Earthquake Engineering 31 (2011) 1624 21
Two noteworthy inconsistencies can be observed in the
experimental ndings: rst, it would be expected that tests 2
and 6 would result in closer values of dynamic base moments, as
the intensity of shaking is almost identical in the two tests.
Second, these moments should be higher than those of test 1
where the intensity of shaking is lower. Explaining this counter-
intuitive trend is not straightforward.
Overall, Bolton and Steedman conclude that their results
conrm the validity of the MO methodboth for the wall force
and for the point of application at H/3. Obviously, the same can be
concluded regarding the comparison with the elastic solutions.
5.3. Experiments by Andersen et al.
Andersen et al. [38], performed a series of centrifuge tests
GA3EQ1, GA6EQ1 and GA5EQ1 (3, 6 and 5) on tilting wall models
retaining dry sand backll in a 80g eld.
The wall made of aluminum was essentially rigid and
elastically constrained against rotation at the base. It was
152 mm (6 in.) high, 152 mm (6 in.) wide and retained a stratum
of dry sand of relative density D
r
84%, height H145 mm
(5.69 in.) and unit weight g16.8 kN/m
3
(107 lb/ft
3
). The input
acceleration consisted of a sequence of 10 nearly sinusoidal major
cycles of approximately the same dominant frequency of 125 Hz,
preceded and followed by several less severe pulses.
Measured quantities included wall accelerations and wall
forces, but not the point of application of dynamic thrust because,
as they mention, it could not be determined accurately.
Andersen et al., compared the experiment results with those
obtained by the MO method and concluded that MO under-
predicts the measured values for the forces by nearly 30%. Had
this comparison been done with the values calculated by the
elastic solutions of VY, a very good agreement would have been
found as shown below.
The shear modulus of the soil can be calculated from the
fundamental frequency from the amplitude spectrum for the
angular wall acceleration presented in the report of Andersen
et al. and their comments. For test 3, besides the input frequency
of 125 Hz, the frequencies of 240 and 360 Hz were observed and
they consider the former to be the fundamental. For test 6, besides
the input frequency of 125 Hz, the frequencies of 240, 360 and
600 Hz were present so they note that further study is needed to
determine the fundamental frequency. For test 5, the amplitude
spectrum is not provided. However, it is reasonable to conclude
that the fundamental frequency for all tests is 240 Hz by taking
into account the maximum values of the displacement of wall tip
reported: 3.6 mm (0.143 in.), 1.93 mm (0.076 in.) and 0.533 mm
(0.021 in.) for tests 3, 6 and 5, respectively. The corresponding
associated soil strains computed from Eq. (20) are approximately
2410
3
, 1310
3
and 3.510
3
and from the SeedIdriss
diagram [40] it can be derived that to these high values
correspond close values of shear modulus and hence, not
signicantly different values of fundamental frequency.
For the fundamental frequency of 240 Hz, the shear wave
velocity of the soil is given by
V
s
4Hf
1
22
resulting in V
s
139 m/s (455 ft/s). Then from the relation
G rV
s
2
23
the corresponding shear modulus is determined at 33 MPa
(69010
3
psf).
For test 6, the rotational stiffness at the base of the wall was
14,200 N m/rad (10,500 lb ft/rad), which divided by the wall
width leads to a stiffness per unit of length of R
y
14,200/
0.15293,400 N/rad (21,000 lb/rad). So the relative wall ex-
ibility dened by Eq. (2), was d
y
7.38, a rather high value. The
ratio of the imposed frequency to the fundamental one is
f
e
/f
1
125/2400.52, so the response cannot be considered as
static (low frequency), but it is dened as the product of the static
value multiplied by an amplication factor.
Table 2
Computation of dynamic moments, DM
b
, at 0.05H from the base by elastic, limit state and stress solutions for the different tests performed by Bolton and Steedman. In
parentheses are given the percentile difference between the measured and the various analytical solution magnitudes.
Test k
h Dynamic thrust DQ
b
(N) Dynamic moment DM
b
(N mm/mm)
MO [3,4] VY [18] MO, thrust at
H/3 [3,4]
MO, thrust at
0.6H [5]
VY, thrust at
0.252H [1519]
Stress solution [29] Measured [37]
1 0.184 1660 3130 82 (39%) 160 (19%) 107 (20%) 111 (17%) 135
2 0.241 2320 4100 115 (4%) 224 (86%) 141 (17%) 162 (35%) 120
6 0.234 2240 3980 111 (23%) 215 (139%) 137 (51%) 157 (74%) 90
Fig. 4. Dynamic moments at 0.05H from the base computed by elastic, limit state and stress solutions compared to the values measured by Bolton and Steedman.
C. Giarlelis, G. Mylonakis / Soil Dynamics and Earthquake Engineering 31 (2011) 1624 22
The peak value of the static base shear per unit length of wall
for a Poissons ratio of n
w
1/3 can be calculated using Table 1B. It
is
Q
b

st

0:266r

X
g
H
2
1020N=m 69:5lb=ft
_ _
24
For the ratio of frequencies calculated above and since the
excitation can be considered as harmonic, the amplication factor
found from the elastic solutions of VY is 1.26. So the maximum
base shear per unit wall length is
Q
b

1:26 Q
b

st

1280N=m 87:5lb=ft
_ _
25
This value should be compared to the test results appearing in
Table 6.2 of Andersen et al. [38]. The maximum value of the total
force (including the static, residual stress effect and dynamic
components) was 854 N (192 lb). The corresponding static force,
including the residual stress effect, was 649 N (146 lb) leading to
a maximum dynamic force of DP
AE
854649205 N (46 lb).
Then the corresponding base shear per unit of wall length is
Q
b

205
0:152
cosd 1280N=m 87:5lb=ft
_ _
26
where d is the friction angle, 181.
The same procedure is followed for tests 3 and 5 and the
results are presented in Table 3 and in Fig. 5. In both cases the
base shear has been calculated by multiplying the dynamic thrust
by cos d, for d181 and 101, respectively.
Overall the comparison shows that the agreement is excellent
for tests 6 and 5. It is not that favorable for test 3, probably due to
the high nonlinear action taking place as a result of the high
rotational exibility.
6. Conclusions
Following a brief review of limit-state methods, elements of
the elastodynamic solutions advanced by Veletsos and Younan
were discussed. The study focused on comparisons between
available experimental results on retaining walls and predictions
of the aforementioned elastic solutions.
Comparisons were made with the works of Sherif et al.
[2426], Bolton et al. [37] and Andersen et al. [38]. These
comparisons focused on the magnitude and point of application
of dynamic soil thrust. In all cases, reasonable agreement between
analysis and experimental measurement was shown. It should be
noted that Sherif et al. and Bolton et al. arrive at different
conclusions regarding the point of application of the dynamic
thrustthe former reporting it to vary from 0.51H to 0.54H,
whereas the latter to be approximately H/3. This discrepancy was
explained convincingly by means of the wallsoil exibility
parameter employed in the elastic solutions. Applying this
parameter in the aformentionted experimental works, it was
found that the soilwall systems examined were not rigid.
It is noted that the measured results from the shaking table
experiments compare far better with the limit-state solutions,
whereas for the centrifuge tests the results seem to match better
with the elastic solution. A possible explanation is that centrifuge
experiments may be more accurate in terms of the actual scale.
However, no safe conclusion can be drawn since the number of
experiments is limited.
In light of the above results, elastic solutions appear to deserve
more credit than normally granted in these problems. Apart from
direct applicability when elastic response prevails (e.g., in base-
ment walls), these solutions seem to guarantee a conservative
upper limit for wall pressures. Contrary to traditional limit-state
analyses, the elastic solutions elucidate the role of wall kinematics
quantied through the relative soilwall exibility parameter d
w
and the rotational base exibility parameter d
y
.
Acknowledgements
The rst author is indebted to Professor Anestis Veletsos, for
his guidance during the early phases of this study. He would also
like to thank Dr. Adel Younan and Dr. Panos Dakoulas for many
Table 3
Computation of Dynamic Base Shear by elastic, limit state and stress solutions for the different tests performed by Andersen et al. In parentheses the percentile difference
between the measured and the various analytical solution magnitudes is given.
Test R
y
(N/rad) d
y
, VY [1519] Dynamic base shear (N)
VY [1519] MO [34] Stress solution [29] Measured [38]
GA3EQ1 46200 14.90 823 (24%) 993 (8%) 1095 (1%) 1083
GA6EQ1 93400 7.38 1275 (0%) 993 (22%) 1095 (14%) 1277
GA5EQ1 279000 2.47 1065 (3%) 995 (4%) 1150 (11%) 1034
Fig. 5. Dynamic base shear computed by elastic, limit state and stress solutions compared to the values measured by Andersen et al.
C. Giarlelis, G. Mylonakis / Soil Dynamics and Earthquake Engineering 31 (2011) 1624 23
helpful comments. The study reported herein was partially
supported by a graduate fellowship from Rice University and a
research grant from Brookhaven National Laboratory. Sincere
appreciation is expressed for this support. Last, but not least, the
authors would like to thank the anonymous reviewers for many
valuable comments that helped in improving the manuscript.
References
[1] Japanese Geotechnical Society (JGS). Geotechnical aspects of the January 17
1995 Hyogoken-Nambu earthquake. Soils and Foundations, Special issue no.
2, 1998.
[2] EERI, Chi-Chi, Taiwan, Earthquake of September 21, 1999, Reconnaisance
Report. Earthquake Spectra 17, Supplement A, April 2001.
[3] Mononobe H, Matsuo M. On the determination of earth pressures during
earthquakes. Proceedings of World Engineering Congress, vol. 9; 1929.
[4] Okabe S. General theory of earth pressure and seismic stability of retaining
wall and dam. Journal of Japanese Society of Civil Engineering 1926;12:1.
[5] Seed HB, Whitman RV. Design of earth retaining structures for dynamic loads.
Proceedings of ASCE specialty conference on lateral stresses in the ground
and the design of earth retaining structures. Ithaca, NY: Cornell University,
New York; 1970. p. 10347.
[6] Richards R, Elms DG. Seismic behavior of gravity retaining walls. Journal of
Geotechnical Engineering 1979;105(GT4):449464.
[7] Nadim F, Whitman RV. Seismically induced movement of retaining walls.
Journal of Geotechnical Engineering 1983;109(7):91531.
[8] Dubrova GA. Interaction of soil and structures. Rehnoy Transport, Moscow,
USSR, 1963.
[9] ATC. Seismic design guidelines for highway bridges. ATC-6, Palo Alto, CA, 1981.
[10] EC-8. Design provisions for earthquake resistance of structures. Part 5, 1994.
[11] Matsuo H, Ohara S. Lateral earth pressure and stability of quay walls during
earthquakes. Proceedings of 2nd world conference on earthquake engineer-
ing, vol. I; 1960. p. 16581.
[12] Wood JH. Earthquake induced soil pressures on structures. Rep. EERL 73-05,
Earthquake Engineering Research Laboratory, California Institute of Technol-
ogy, 1973.
[13] Wood JH. Earthquake induced pressures on rigid wall structure. Bulletin of
New Zealand Society of Earthquake Engineering 1975;8:17586.
[14] Arias A, Sanchez-Sesma FJ, Ovando-Shelley E. A simplied elastic model for
seismic analysis of earth-retaining structures with limited displacements. In:
Prakash S, editor. International conference on recent advances in geotechni-
cal earthquake engineering and soil dynamics, vol. 1. Rolla: University of
Missouri; 1981. p. 23540.
[15] Veletsos AS, Younan AH. Dynamic soil pressures on rigid retaining walls.
Earthquake Engineering & Structural Dynamics 1994;23(3):275301.
[16] Veletsos AS, Younan AH. Dynamic modeling and response of soil-wall
systems. Journal of Geotechnical Engineering 1994;120(12):2155179.
[17] Veletsos AS, Younan AH. Dynamic soil pressures on vertical walls. In: Prakash
S, editor. Proceedings 3rd international conference on recent advances in
geotechnical earthquake engineering and soil dynamics, vol. III. Rolla, MO:
University of Missouri; 1995. p. 1589604.
[18] Veletsos AS, Younan AH. Dynamic response of cantilever retaining walls.
Journal of Geotechnical Engineering 1997;123(2):16172.
[19] Younan AH, Veletsos AS. Dynamic response of exible retaining walls.
Earthquake Engineering & Structural Dynamics 2000;29(12):181544.
[20] Giarlelis CM. Dynamic response of soil wall systems. MSc Thesis, Rice
University, Houston, TX, 1997.
[21] Jacobsen LS. Appendix D, The Kentucky Project, Technical report no. 13,
Tennessee Valley Authority, 1951.
[22] Matsuo M. Experimental study on the distribution of earth pressures acting
on a vertical wall during earthquakes. Journal of Japanese Society of Civil
Engineering 1941;no 2:27.
[23] Ishii Y, Arai H, Tsuchida H. Lateral earth pressure in an earthquake.
Proceedings of 2nd world conference on earthquake engineering, vol. I;
1960. p. 21130.
[24] Sherif MA, Ishibashi I, Lee CD. Earth pressure against rigid retaining walls.
Journal of Geotechnical Engineering, 108; 1982 p. 67995.
[25] Sherif MA, Fang YS. Dynamic earth pressures against rotating and non-
yielding retaining walls. University of Washington, Soil Engineering Research
Report 23, 1983.
[26] Sherif MA, Fang YS. Dynamic earth pressures on rigid walls rotating about the
base. Proceedings of 8th world conference on earthquake engineering, vol. 3.
San Francisco, CA, USA. Englewood Cliffs, New Jersey: Prentice Hall; 1984.
p. 9931000.
[27] Prakash S, Basavanna BM. Earth pressure distribution behind retaining wall
during earthquake. Proceedings of 4th world conference on earthquake
engineering. 1969.
[28] Newmark NM. Effects of earthquakes on dams and embankments. 5th
Rankine Lecture, Geotechnique, vol. XV, No. 2. The Institution of Civil
Engineers, London, England, 1965. p. 13960.
[29] Mylonakis G, Kloukinas P, Papantonopoulos C. An alternative to the
MononobeOkabe equation for seismic earth pressures. Soil Dynamics and
Earthquake Engineering 2007;27(10):95769.
[30] Lancellotta R. Geotechnical Engineering. 2nd edition. Taylor & Francis; 2009.
[31] Scott RF. Earthquake-induced earth pressures on retaining walls. Proceedings
of 5th world conference on earthquake engineering. Rome: International
Association for Earthquake Engineering; 1973. p. 161120.
[32] Wu G, Finn WDL. Seismic lateral pressures for design of rigid walls. Canadian
Geotechnical Journal 1999;36(3):50922.
[33] Psarropoulos P, Klonaris G, Gazetas G. Seismic earth pressures on rigid and
exible retaining walls. Soil Dynamics and Earthquake Engineering 2005;25:
795809.
[34] Theodorakopoulos DD, Chassiakos AP, Beskos DE. Dynamic pressures
on rigid cantilever walls retaining poroelastic soil media. Part I: rst method
of solution. Soil Dynamics and Earthquake Engineering 2001;21(4):31538.
[35] Theodorakopoulos DD, Chassiakos AP, Beskos DE. Dynamic pressures on
rigid cantilever walls retaining poroelastic soil media. Part II: second method
of solution. Soil Dynamics and Earthquake Engineering 2001;21(4):339364.
[36] Li X. Dynamic analysis of rigid walls considering exible foundation, 125.
ASCE; 1999. p. 8036.
[37] Bolton MD, Steedman RS. The behavior of xed cantilever walls subjected to
lateral shaking. Symposium on the application of centrifuge modeling to
geotechnical design, Manchester, 1618 April 1984, Rotterdam: AA Balkema,
Boston.
[38] Andersen G, Whitman R, Germaine J. Tilting response of centrifuge modeled
gravity retaining wall to seismic shaking. Report R87-14, Department of Civil
Engineering, MIT, Cambridge, MA, 1987.
[39] Al-Homoud AS, Whitman RV. Comparison between FE prediction and results
from dynamic centrifuge tests on tilting gravity walls. Soil Dynamic and
Earthquake Engineering 1995;14:25968.
[40] Seed HB, Idriss IM. Soil moduli and damping factors for dynamic response
analyses. UCB/EERC-70/10, Earthquake Engineering Research Center.
Berkeley: University of California; 1970.
C. Giarlelis, G. Mylonakis / Soil Dynamics and Earthquake Engineering 31 (2011) 1624 24