geosciences

Article
Numerical Modelling of Structures Adjacent to
Retaining Walls Subjected to Earthquake Loading
Xiaoyu Guan and Gopal S. P. Madabhushi *
Schofield Centre, Department of Engineering, University of Cambridge, Trumpington Street,
Cambridge CB2 1PZ, UK; xg257@cam.ac.uk
* Correspondence: mspg1@cam.ac.uk; Tel.: +44-12-2376-8053




Received: 28 September 2020; Accepted: 26 November 2020; Published: 3 December 2020


Abstract: In an urban environment, it is often necessary to locate structures close to existing retaining
walls due to congestion in space. When such structures are in seismically active zones, the dynamic
loading attracted by the retaining wall can increase. In a novel approach taken in this paper, finite
element-based numerical analyses are presented for the case of a flexible, cantilever sheet pile wall
with and without a structure on the backfill side. This enables a direct comparison of the influence
exerted by the structure on the dynamic behaviour of the retaining wall. In this paper, the initial static
bending moments and horizontal stresses prior to application of any earthquake loading are compared
to Coulomb’s theory. The dynamic behaviour of the retaining wall is compared in terms of wall-top
accelerations and bending moments for different earthquake loadings. The dynamic structural
rotation induced by the differential settlements of the foundations is presented. The accelerations
generated in the soil body are considered in three zones, i.e., the free field, the active and the passive
zones. The differences caused by the presence of the structure are highlighted. Finally, the distribution
of horizontal soil pressures generated by the earthquake loading behind the wall, and in front of the
wall is compared to the traditional Mononobe-Okabe type analytical solutions.

Keywords: earthquake; finite element modelling; retaining walls; structures

1. Introduction
The increasing population in large cities and towns around the world place a high demand on
available land. In such situations, excavations are required next to existing buildings to construct new
metros, railways, highways or new buildings with basement floors. Retaining walls are often used to
create the necessary difference in ground levels with backfills placed behind the retaining walls or
excavations carried out in front of them. In the former case, new buildings could be constructed close
to the retaining wall, while, in the latter case, where existing buildings are present behind the wall,
the retaining walls have to protect from any damage due to ground movement. The location of these
structures can often be quite close to the retaining wall, often only a few meters away. An example of a
retaining wall close to an existing structure is shown in Figure 1. In this figure the retaining wall is
anchored using soil nails, but many cases exist where cantilever retaining walls are used. The designer
has to evaluate the additional soil-structure interaction effects due to the presence of the structure on
the backfill side.

Geosciences 2020, 10, 486 ; doi:10.3390/geosciences10120486 www.mdpi.com/journal/geosciences

Geosciences 2020, 10, 486 2 of 18
Geosciences 2020, 10, x FOR PEER REVIEW 2 of 18

Figure
Figure 1.
1. A
A structure
structure in
in close
close proximity
proximity to
to aa retaining
retaining wall
wall [1]
[1] (courtesy
(courtesy Aarsleff
Aarsleff Ground
Ground Engineering
Engineering
Ltd. (Newark, UK)).
Ltd. (Newark, UK)).

In seismic regions of the world, additional additional loading

loading is expected
expected on on the retaining
retaining wallwall due to
earthquake loading. In In these
these extreme
extreme loading
loading cases,
cases, itit can
can bebe anticipated
anticipated that additional lateral loads
act on
onthetheretaining
retaining wall duedue
wall to increased
to increasedhorizontal stresses,
horizontal additional
stresses, inertialinertial
additional loads from thefrom
loads building,
the
etc. Thereetc.
building, is aThere
significant dynamic soil-structure
is a significant interactioninteraction
dynamic soil-structure due to the due
actionto of
theearthquake loads. It is
action of earthquake
imperative
loads. It is that designers
imperative provide
that adequate
designers provide walladequate
sections for wallthesections
retaining forwall
thetoretaining
withstand these
wall to
additional these
withstand loads.additional
Additionally, theAdditionally,
loads. wall movements the can
wallhave adverse effects
movements can have on adverse
the foundations
effects onof the
structure causing
foundations of theitstructure
to settle causing
and rotate.it toThe
settleinteraction
and rotate. between the retaining
The interaction betweenwalls and
the structures
retaining walls is
also structures
and important is from
alsothe bridge engineering
important from the bridge perspective.
engineering For example, the spandrel
perspective. For example,wallstheof spandrel
masonry
bridges,
walls acting as retaining
of masonry walls, as
bridges, acting interact strongly
retaining walls, with the backfill
interact strongly soils.
with Pelà
theetbackfill
al. [2] and D’Amato
soils. et
Pelà et al.
al. [3] have investigated the seismic behaviour of masonry arch bridges. Similarly,
[2] and D’Amato et al. [3] have investigated the seismic behaviour of masonry arch bridges. Similarly, the use of static and
dynamic
the use oftests onand
static bridges
dynamicwas recommended
tests on bridges based
was on the modal analyses
recommended based on by the
Marcheggiani
modal analyseset al. [4].
by
In all these cases,
Marcheggiani the[4].
et al. consideration
In all these of the interaction
cases, the considerationbetweenofthe theabutment
interaction walls and the
between thesoil backfill
abutment
is important,
walls and the particularly
soil backfill during strongparticularly
is important, earthquake duringloading.strong earthquake loading.
There have been examplesexamples of damagedamage to to retaining
retaining wallswalls inin previous
previous earthquakes—for
earthquakes—for example,example,
a sea front retaining wall collapsed during the Bhuj earthquake in 2001 [5]. Zeng and Steedman Steedman [6]
dynamic centrifuge
used dynamic centrifuge modelling
modelling to investigate
investigate flexible, cantilever
cantilever retaining
retaining walls subjected
subjected to
earthquake loading
earthquake loading with
with dry and saturated backfills. Madabhushi Madabhushi and and Zeng
Zeng [7,8] investigated
investigated the
same retaining walls using the finite element method in the time domain using a fully coupled code
Swandyne [9]. They
called Swandyne They show
show thatthat the
the finite
finite element
element (FE) (FE) code
code SWANDYNE
SWANDYNE was able to capture
response observed
the dynamic response observed in in the centrifuge
centrifuge tests
tests both
both in terms of the wall displacements and
dynamic bending moments. Cilingir Cilingir et et al.
al. [10]
[10] extended
extended this this work to investigate anchored retaining
walls in dry soils. TheirTheir research
research shows
shows that
that SWANDYNE was able to capture the magnitude of the
anchor forces recorded in the dynamic centrifuge test well, in addition to the bending moments and
wall displacements.
displacements. More More recent
recent improvements
improvements in in the
the numerical
numerical modelling
modelling of of retaining walls were
carried out by Callisto and
carried and Soccodato
Soccodato [11] [11] and
and Conti
Conti and and Viggiani [12] with the aim of developing
improved design methods. Conti and Viggiani Viggiani [12] used the the “threshold
“threshold or yield
yield acceleration”
acceleration” concept
to estimate
estimatethe thelateral
lateralwall
walldisplacements
displacements and andthe accumulation
the accumulation of the of
wallthedisplacements over different
wall displacements over
different cycles of an earthquake. Yeganeh et al. [13] carried out numerical analyses of a high-rise
building behind an anchored wall, observing more damage in the model where the real building was
Geosciences 2020, 10, 486 3 of 18

cycles of an earthquake. Yeganeh et al. [13] carried out numerical analyses of a high-rise building
Geosciences 2020, 10, x FOR PEER REVIEW 3 of 18
behind an anchored wall, observing more damage in the model where the real building was simulated
instead
simulatedof being
instead simplified
of beingassimplified
surcharge as loading. Although
surcharge the dynamic
loading. Althoughbehaviour
the dynamic of retaining
behaviour wallsof
was investigated by many researchers previously [6–8], very little research exists
retaining walls was investigated by many researchers previously [6–8], very little research exists in in the literature that
investigated
the literaturethe thatdynamic performance
investigated the dynamic of the retaining walls
performance of theinretaining
the presencewallsofina the
structure
presence on ofthea
backfill
structureside, in close
on the proximity
backfill side, intoclose
the retaining
proximitywall.to the retaining wall.
In
In many cases, the soil on the backfill side,
many cases, the soil on the backfill side, as
as well
well as
as in
in front
front of
of the
the retaining
retaining wall,
wall, may
may be be
saturated,
saturated, such
such asas in
in the
the case
case ofof water-front
water-front structures.
structures. Earthquake
Earthquake loading
loading cancan lead
lead toto full
full or
or partial
partial
liquefaction
liquefaction of of such
such soils.
soils. In
In this
this paper,
paper, the
the focus
focus is
is on
on dry
dry sandy
sandy soils
soils and
and the
the performance
performance of of the
the
retaining
retaining wall is studied for this relatively simple soil case, so that the effect of the structure on the
wall is studied for this relatively simple soil case, so that the effect of the structure on the
backfill
backfill can
can bebe easily
easily identified.
identified. TheThe presence
presence of of pore
pore fluid
fluid in
in the
the saturated
saturated soil
soil case
case would
would leadlead toto
additional loading due to hydrodynamic action on the retaining wall, as well
additional loading due to hydrodynamic action on the retaining wall, as well as degradation in soil as degradation in soil
stiffness
stiffnessdue
dueto toexcess
excessporeporepressure
pressuregeneration.
generation.
Overall,
Overall, this paper adopts a novel approach
this paper adopts a novel approach where
where an an identical
identical flexible
flexible retaining
retaining wallwall system
system
will
will be analysed with and without a structure placed on the backfill side. This allows for
be analysed with and without a structure placed on the backfill side. This allows for aa direct
direct
comparison
comparison between
between the the retaining
retaining wall
wall on
on its
its own
own andand the
the changes
changes induced
induced by by the
the presence
presence of of the
the
structure and its foundations.
structure and its foundations.

2.
2. Numerical
Numerical Modelling
Modelling
In
In this
this paper,
paper, finite
finite element
element modelling
modelling is is carried
carried out
out using
using the
the Swandyne
Swandyne code code [9],
[9], which
which is is aa
general-purpose FE code for problems in Geomechanics. As explained
general-purpose FE code for problems in Geomechanics. As explained earlier, two models are earlier, two models are analysed
with and without
analysed with and a structure
without abehindstructurethebehind
retaining thewall. The dimensions
retaining of the model
wall. The dimensions ofwith a structure
the model with
are shown in
a structure areFigure
shown 2. in
The retaining
Figure 2. Thewall had a retained
retaining wall had soil height soil
a retained (h) of 3.6 m,(h)and
height the m,
of 3.6 wallanditself
the
penetrated to a depth (d)
wall itself penetrated to of 5.76 m,
a depth (d)giving
of 5.76anm, h/dgiving
ratio of
an0.625. Theof
h/d ratio soil was The
0.625. modelled
soil was as amodelled
uniform as drya
sand bed dry
uniform on either
sand side
bed of onthe retaining
either side ofwall.
theThe structure
retaining was
wall. Themodelled
structureas awas
sway frame with
modelled as aa sway
high
centre of gravity. In addition, loading on the roof was chosen to be asymmetric,
frame with a high centre of gravity. In addition, loading on the roof was chosen to be asymmetric, with larger loading
applied to the
with larger columns
loading closertotothe
applied thecolumns
retaining wall.toThe
closer the columns
retainingofwall.
the structure
The columns wereofsupported
the structure on
individual strip footings.
were supported The right-hand
on individual side The
strip footings. structural footing
right-hand was
side placed about
structural footing2.1was
m from
placedtheabout
back
edge of the retaining wall. In Swandyne FE code, two distinct steps are required
2.1 m from the back edge of the retaining wall. In Swandyne FE code, two distinct steps are required for any dynamic
analyses. Initially,analyses.
for any dynamic a static run must be
Initially, executed
a static run to establish
must the geostatic
be executed stresses
to establish theingeostatic
the soil and for the
stresses in
model
the soiltoand
establish
for theequilibrium of forces.equilibrium
model to establish In the second step, the
of forces. In final stress state
the second step,from the static
the final stressrun is
state
utilised as the initial stress state to start the dynamic run.
from the static run is utilised as the initial stress state to start the dynamic run.

Figure 2. Dimensions
Figure2. Dimensions of
of aa retaining
retaining wall
wall system
system (unit:
(unit: m).
m).

2.1. Finite Element Discretisation

2.1. Finite Element Discretisation
The finite element discretisation used in the analyses described in this paper was carried out using
The finite element discretisation used in the analyses described in this paper was carried out
eight noded isoparametric, quadrilateral elements with nine gauss points for integration. This enabled
using eight noded isoparametric, quadrilateral elements with nine gauss points for integration. This
accurate modelling of the soil elements, as well as the structural elements forming the retaining wall and the
enabled accurate modelling of the soil elements, as well as the structural elements forming the
retaining wall and the sway frame structure. Slip elements were utilised next to the retaining wall on
both the active and passive sides and below the structure. The material properties used in these
analyses are presented in Section 2.2.
Geosciences 2020, 10, 486 4 of 18

Geosciences 2020, 10, x FOR PEER REVIEW 4 of 18

sway frame structure. Slip elements were utilised next to the retaining wall on both the active and passive
sidesAn
andoverview
below theofstructure.
the FE discretisation is shown in
The material properties Figure
used 3a for
in these the case
analyses are of a retaining
presented wall only
in Section 2.2.
and Figure 3b for the case of a retaining wall with a sway frame structure behind it.
An overview of the FE discretisation is shown in Figure 3a for the case of a retaining wall only In each of these
FE
andmeshes,
Figure 3bthere were
for the case1690
of aand 1716 wall
retaining solidwith
nodes, respectively,
a sway with two
frame structure behinddegrees of freedom
it. In each of theseper
FE
node.
meshes,The number
there of elements
were 1690 and 1716 insolid
eachnodes,
of these meshes were
respectively, with 781
twoand 789, respectively.
degrees of freedom per The element
node. The
sizes
numberandoftime stepsin
elements ineach
the dynamic analyses
of these meshes were
were 781chosen to respectively.
and 789, allow for theThetransmission
element sizes of Sand
h waves
time
from
steps the bedrock
in the dynamic towards
analysesthewere
ground surface
chosen following
to allow for thethe recommendations
transmission of Sh wavesof Semblat
from theetbedrock
al. [14]
and Haigh
towards theetground
al. [15].surface following the recommendations of Semblat et al. [14] and Haigh et al. [15].

(a) (b)
Figure 3. Finite element (FE) discretization: (a) a retaining wall only and (b) a retaining wall with an
Figure 3. Finite element (FE) discretization: (a) a retaining wall only and (b) a retaining wall with an
adjacent structure on the backfill.
adjacent structure on the backfill.
In these FE analyses, the base nodes of the meshes shown in Figure 3a,b were fixed in both the x
and yIndirections.
these FE analyses,
This was the base nodes
necessary, sinceofthe
the meshes shown
earthquake in Figure
loading 3a,b were
was applied fixedaccelerations
as nodal in both the
x and y directions. This was necessary, since the earthquake loading was applied as nodal
at all the fixed nodes in the meshes. In these analyses, a simple sinusoidal motions and sine sweep
accelerations
motions wereatusedall the
as fixed nodes
the input in the meshes.
accelerations In these
to the analyses,
base nodes. Thea simple
reasonsinusoidal motions
for this choice and
of input
sine sweep motions were used as the input accelerations to the base nodes. The reason for this choice
motion is to determine the evolution of the peak bending moments at a time instant and the evolution
of input motion is to determine the evolution of the peak bending moments at a time instant and the
of horizontal earth pressures next to the retaining wall, with and without the structure on the backfill.
evolution of horizontal earth pressures next to the retaining wall, with and without the structure on
The edge boundaries were modelled as fixed in the x direction and free in the y direction. While these
the backfill. The edge boundaries were modelled as fixed in the x direction and free in the y direction.
boundary conditions are satisfactory for the static analyses, it is recognised that unwarranted wave
While these boundary conditions are satisfactory for the static analyses, it is recognised that
reflections may occur when earthquake loading is applied. In previous research, e.g., Madabhushi
unwarranted wave reflections may occur when earthquake loading is applied. In previous research,
and Zeng [7,8] and Cilingir et al. [10], absorbing boundaries were used in the numerical analyses to
e.g., Madabhushi and Zeng [7,8] and Cilingir et al. [10], absorbing boundaries were used in the
simulate the boundaries in the dynamic centrifuge tests. In this paper, the lateral extent of the boundary
numerical analyses to simulate the boundaries in the dynamic centrifuge tests. In this paper, the
value problem was sufficiently far from the active and passive zones of the boundary. The material
lateral extent of the boundary value problem was sufficiently far from the active and passive zones
and Rayleigh damping in the numerical analyses were considered to be large enough to absorb the
of the boundary. The material and Rayleigh damping in the numerical analyses were considered to
wave reflections. Nonetheless, it must be pointed that some error will occur in the dynamic analyses
be large enough to absorb the wave reflections. Nonetheless, it must be pointed that some error will
due to the wave reflections from the vertical boundaries.
occur in the dynamic analyses due to the wave reflections from the vertical boundaries.
The flexural properties and densities of different components used in these analyses are presented
The flexural properties and densities of different components used in these analyses are
in Table 1. The flexural stiffness is expressed for the retaining wall and the structural columns and is
presented in Table 1. The flexural stiffness is expressed for the retaining wall and the structural
expressed as “per m”, as the analyses were carried out with a plane strain formulation. The average
columns and is expressed as “per m”, as the analyses were carried out with a plane strain formulation.
density of the slab was calculated to allow for the additional mass added to the right-hand edge of the
The average density of the slab was calculated to allow for the additional mass added to the right-
structure (Figure 3b) to provide additional structural loads for the right-hand side columns. This allows
hand edge of the structure (Figure 3b) to provide additional structural loads for the right-hand side
for the inclusion of a degree of asymmetry in the structure. The resulting bearing pressures on the left
columns. This allows for the inclusion of a degree of asymmetry in the structure. The resulting
and right footings are also included in Table 1.
bearing pressures on the left and right footings are also included in Table 1.
Geosciences 2020, 10, 486 5 of 18

Table 1. Structural properties.

Structural Element Flexural Stiffness (EI) (MNm 2 /m) Density (kg/m 3 ) Bearing Pressure (kPa)
structural columns 66.68 2800 -
slab Rigid 5906.7 -
strip footing (left) Rigid 2800 83.0
strip footing (right) Rigid 2800 117.0
retaining wall 34.02 2800 -

2.2. Constitutive Models

In the analyses described in this paper, three types of constitutive models were used. The simple
“elastic” model was used for the sway frame structure including the roof slab and the footings. Similarly,
the retaining wall was also modelled as an “elastic” material. The constitutive parameters ascribed
to these materials are presented in Table 2. Interface elements were used between the retaining wall
and the soil both on the active and passive sides. These were modelled as “slip elements”, with the
properties shown in Table 2.

Table 2. Details of constitutive parameters.

Value
Parameter Definition
Interface Structure/Retaining
Dry Sand
Elements Wall
Constitutive model Mohr-Coulomb V Slip Elastic Type of constitutive model used
Young’s modulus 50 MPa 50 MPa 70 GPa Soil stiffness for static equilibrium
Young’s modulus Soil stiffness for damping in
50 MPa 50 MPa 70 GPa
(dynamic) dynamic analyses
Links strains in horizontal and
Poisson’s ratio 0.3 0.3 0.15
vertical directions
Uniaxial yield
100 Pa - - Cohesion
stress
Friction angle
30º 16.4º - To obtain critical state failure line
(critical state)
Dilatancy angle 2º - - To obtain the peak friction angle
Work-hardening The slope of the stress vs. yield
100 - -
modulus strain
For the calculation of material
Void ratio 0.8 - -
density

For modelling soil behaviour in FE analyses of geotechnical problems, it is important to account

for the plastic behaviour of soil. Accordingly, many elasto-plastic models are used in the FE analysis.
One of the simplest soil models is the Mohr-Coulomb type model. Such a model has been adapted
to handle the elastic-perfectly plastic behaviour of soil, which has been implemented in the FE code
SWANDYNE. Various versions of such models are available in this code, with increasing levels of
complexity. The analyses presented in this paper were carried out using Mohr-Coulomb V model.
The main features of the model are briefly outlined below, and the actual parameters that are used are
presented in Table 2.
Since the primary concern in these analyses is with dry, sandy soil, the nonassociated flow rule
must be used. Many researchers have established that sandy soils follow the nonassociative flow
rule [16–18]. The Mohr-Coulomb V model in SWANDYNE captures the nonassociative Mohr-Coulomb
elastic-perfectly plastic response. The Mohr-Coulomb V constitutive model was used for cyclic loading
problems investigated by previous researchers [7,10,19]. The main advantages of this constitutive
model are discussed below:
• The variation of bulk and/or shear modulus is considered with mean confining effective stress.
Geosciences 2020, 10, 486 6 of 18

• Cohesion can be included. The stress state will be cut off if the mean effective confining stress
is more negative than the allowable cohesion. In the analyses presented in this, only a nominal
cohesion of 100 Pa was used.
• The plastic potential can have a different slope with the yield surface.
• A smooth fit to the triaxial compression and triaxial extension state, so there is no corner or
singularity in the π-plane [20].
• Strain hardening of the soil is incorporated in this model, so that the hardening of soil that occurs
cycle by cycle during an earthquake load can be captured.

Some of the limitations of the Mohr-Coulomb model are given below:

• As the Mohr-Coulomb model is a simple model, the shapes of the yield surface and plastic
potential surface are similar.
• This model cannot be used for saturated soil conditions, as it does not capture the volumetric
strains under cyclic loading, and hence, no excess pore pressure build-up occurs.

In the present analyses, a nominal cohesion of 100 Pa was used primarily to achieve numerical
stability. Additionally, the interface friction between the retaining wall and the dry sand was chosen to
be about 0.5 the peak friction angle of the sand.
In this paper, only uniform dry sand was considered. However, in case of waterfront structures,
the shallow portions of the backfill above the water table can be under suction. Ng et al. [21] discussed
the variation of soil stiffness and its influence on the retaining walls under such conditions. Although
it is beyond the scope of this paper, it would be interesting to consider the influence of such suction
dependent shear modulus and its variation under cyclic loading.

3. Static Equilibrium of the Retaining Wall

As explained earlier, before Swandyne code can be used to run any dynamic analyses, it requires a
static run to establish the geostatic stresses. Such runs were carried out for the present analyses, and in this
section, the static behaviour of the retaining wall on its own will be compared to the case where a structure
is present on the backfill side. It must be pointed out that, in running the static analyses, the retaining wall
was wished-in-place, i.e., the construction process of the retaining wall was not taken into account.

3.1. Wall Deflections and Bending Moments

The two important parameters in the retaining wall design are the static wall deflections on
completion of the wall construction and the bending moments that are generated in the wall section.
In Figure 4, the normalised static wall deflections are plotted against the normalised depth. Both
these parameters are normalised with the total height of the wall Hwall used in these analyses. In
Figure 4, the normalised wall deflections obtained for the case of the retaining wall only (i.e., “no
structure”) are compared with the case when the sway frame structure was present on the backfill
side (see Figure 3b). Clearly, the wall deflections increased substantially when the structure is in place
behind the retaining wall. The wall tip deflection increased nearly by a factor of three.
In Figure 4, the increased curvature of the wall can also be seen. This implies the bending
moments in the wall, which are proportional to the curvature, should also increase. In Figure 5, the
normalised bending moments are plotted against the normalised depth. The bending moments M are
normalised by the unit weight of the soil and wall height (H3 ) to obtain a nondimensioned variable.
The same normalisation will be used for the results from dynamic analyses presented later. In Figure 5,
it can be seen that the normalised bending moments increased substantially in the presence of the
structure. Further, it can also be seen that the location of the peak bending moment was just below the
excavation depth for the retaining wall-only case. This location of peak bending moment seems to
shift downwards in the presence of the structure in the backfill. In general, the shapes of the bending
moment curves are as expected in a static analysis.
 In Figure 4, the normalised static wall deflections are plotted against the normalised depth. Both these
parameters are normalised with the total height of the wall Hwall used in these analyses. In Figure 4,
the normalised wall deflections obtained for the case of the retaining wall only (i.e., “no structure”)
are compared with the case when the sway frame structure was present on the backfill side (see
Figure
Geosciences 2020,3b).
10, Clearly,
486 the wall deflections increased substantially when the structure is in place behind
7 of 18
the retaining wall. The wall tip deflection increased nearly by a factor of three.

Geosciences 2020, 10, x FOR PEER REVIEW 7 of 18

In Figure 4, the increased curvature of the wall can also be seen. This implies the bending
moments in the wall, which are proportional to the curvature, should also increase. In Figure 5, the
normalised bending moments are plotted against the normalised depth. The bending moments M are
normalised by the unit weight of the soil and wall height (H3) to obtain a nondimensioned variable.
The same normalisation will be used for the results from dynamic analyses presented later. In Figure
5, it can be seen that the normalised bending moments increased substantially in the presence of the
structure. Further, it can also be seen that the location of the peak bending moment was just below
the excavation depth for the retaining wall-only case. This location of peak bending moment seems
to shift downwards in the presence
Figure of thedeflection
4. Static structure in
thethe
of the backfill.
retaining In general, the shapes of the
wall.
Figure 4. Static deflection of retaining wall.
bending moment curves are as expected in a static analysis.

Figure
Figure 5. Normalised
5. Normalised staticbending
static bendingmoments
moments in
inthe
theretaining
retainingwall.
wall.

3.2. Horizontal Stresses

3.2. Horizontal and and
Stresses Earth Pressures
Earth Pressures

It would It would be interesting

be interesting to compare
to compare thehorizontal
the horizontal stresses
stresses mobilised
mobilisedinin thethe
soilsoil
for for
the the
two two
cases. cases.
In Figure 6a, the horizontal stress contours in the soil body are plotted for
In Figure 6a, the horizontal stress contours in the soil body are plotted for the “retaining wall-only” the “retaining wall-only”
case incasetheinunit
the unit of kPa. In this figure, it can be seen that, generally, the horizontal stresses increase
of kPa. In this figure, it can be seen that, generally, the horizontal stresses increase
with depth. However, there is a bulge of increased horizontal stress in front of the retaining wall on
with depth. However, there is a bulge of increased horizontal stress in front of the retaining wall
the passive side, which is required to establish the equilibrium of the retaining wall. There are also
on thesmall
passive side, which is required to establish the equilibrium of the retaining wall. There are
drops in the horizontal stresses on the active side. In Figure 6b, an equivalent contour plot for
also small
the horizontalinstresses
drops the horizontal
in the casestresses
with theon the active
structure on theside. In Figure
backfill 6b, an equivalent
side is presented. contour
In this figure,
plot forthethe horizontal
horizontal stresses
contours in the
are not verycase withonthe
uniform thestructure
active sideon
due the
to backfill sideofisthe
the presence presented.
two footings In this
figure,ofthe thehorizontal
structure. Thecontours are not
horizontal verybelow
stresses uniform on the active
the footings side due
are expected to the presence
to increase due to the of the
increased
two footings ofbearing pressureThe
the structure. of the footings, causing
horizontal stresseshorizontal
below the stress bulbsare
footings to form in thisto
expected region.
increaseIt is due
to the also interesting
increased to see
bearing that, onofthe
pressure thepassive
footings,side,causing
the horizontal stresses
horizontal stressincrease
bulbs just in front
to form of the
in this region.
It is also interesting to see that, on the passive side, the horizontal stresses increase just incase.
wall. A much larger pressure bulb is formed in this region than in the “retaining wall-only” front of
Again, this larger pressure bulb is required to establish equilibrium in the “structure on backfill” case.
the wall. A much larger pressure bulb is formed in this region than in the “retaining wall-only” case.
Again, thisInlargeraddition to thebulb
pressure general distribution
is required of the horizontal
to establish equilibriumstresses
in theshown in Figure
“structure 6a,b, it iscase.
on backfill”
Inpossible
addition to obtain the horizontal
to the general earth pressures
distribution acting at each
of the horizontal gauss
stresses point in
shown of Figure
the soil 6a,b,
elements
it is next
possible
to obtain the horizontal earth pressures acting at each gauss point of the soil elements nextcan
to the retaining walls from the FE analyses. In addition, the Coulomb’s earth pressure coefficients to the
be calculated using the following well-known Equations (1) and (2) for active and passive cases:
retaining walls from the FE analyses. In addition, the Coulomb’s earth pressure coefficients can be
calculated using the following well-known Equations sin(α(1)
− φ)and (2) for active and passive cases:
K =⎛ sinα ⎞ , (1)
2
 sin(α + δ) +sin(α−ϕ) + δ)sin(φ − β)
sin(φ
⎝  sin α sin(α − β) ⎠
Ka =  √ r  , (1)
 sin(ϕ+δ) sin(ϕ−β) 
sin(α+δ)+ 
sin(α−β)
Geosciences 2020, 10, 486 8 of 18

Geosciences 2020, 10, x FOR PEER REVIEW 8 of 18

 2
Geosciences 2020, 10, x FOR PEER REVIEW 

 sin(α+ϕ)  8 of 18
sin α 
Kp =  √ sin(α + φ)
r  . (2)
sin(ϕ+δ) sin(ϕ+β) 
sin(α−δ)− sinα sin(α−β)

K =⎛ ⎞ .

(2)
sin(α + φ)
sin(φ + δ)sin(φ + β)
In Figure 7, the static distributions sin(α − δ) − sinα
of the earth pressures
K =⎝⎛ sin(α − β)acting⎠⎞on. the retaining wall (2) for the
“wall-only” case and the “structure on the backfill” sin(φ
case + δ)sin(φ
are plotted,+ β)both on the active side and the
In Figure 7, the static distributions sin(α
of−the
δ) earth
− pressures
⎝ sin(α −acting
β) on⎠the retaining wall for the
passive side. Thecase
“wall-only” theoretical earth pressure
and the “structure on thedistributions
backfill” casegiven by Equations
are plotted, both on (1)
the and (2)side
active are and
alsothe plotted
in this In Figure
figure. OnThe 7, the
the static
active distributions
side, earthofpressures
the pressure the earth forpressures acting on the retaining wallclose
for the
passive side. theoretical earth distributionsthe “wall-only”
given by Equations case areand
(1) pretty
(2) are alsoto the
“wall-only” case and the “structure on the backfill” case are plotted, both on the active side and the
Coulomb’s
plotted active earth pressure
in this figure. On the activeline.side,
However,
the earth for the “structure
pressures on the backfill”
for the “wall-only” case, the
case are pretty closeactive
passive side. The theoretical earth pressure distributions given by Equations (1) and (2) are also
to the Coulomb’s active earth pressure line. However, for the “structure
earth pressure shows a pressure bulge between 2–4-m depth. This is due to the increased horizontal on the backfill” case, the
plotted in this figure. On the active side, the earth pressures for the “wall-only” case are pretty close
active
pressures earth pressure shows a pressure bulge between 2–4-m depth. This is due to the increased
to thecoming
Coulomb’sfromactive
the foundations
earth pressure of line.
the structure.
However, On the “structure
for the passive side, for backfill”
on the the “wall-only
case, the case”,
it canhorizontal
be seen pressures
that the comingpressures
passive from the foundations
do not fully ofmobilise
the structure. On theto
compared
active earth pressure shows a pressure bulge between 2–4-m depth. This is due to the increased
passive
the side, for thedistribution.
Coulomb’s “wall-
only case”, it can be seen that the passive pressures do not fully mobilise compared to the Coulomb’s
This ishorizontal
well-known,pressuresandcoming
only partial
from the passive pressures
foundations of theare mobilised
structure. On thetopassive
attain side,
equilibrium.
for the “wall-The FE
distribution. This is well-known, and only partial passive pressures are mobilised to attain
analyses
onlyalso
case”,show
it canthat the
be seen passive
that earth
the passive pressure
pressures mobilisation
do depends
not fully mobilise on
compared
equilibrium. The FE analyses also show that the passive earth pressure mobilisation depends on the
the tovertical stress
the Coulomb’s and
distribution.
the lateral This is well-known, and only partial passive pressures are mobilised to attain
verticalstrain.
stress Passive
and the earthlateral pressures are larger
strain. Passive when larger
earth pressures are strains are mobilised
larger when larger strainsdueare to wall
equilibrium.
deflection and due The aFEpeak
reach analyses also show
at about a depth that the passive earth this
pressure mobilisation depends on the
mobilised to wall deflection and reach aof 5.5
peak m.about
at Below a depthdepth, they
of 5.5 m. Belowstartthis
to depth,
decrease theyas the
vertical stressreduce.
mobilised and the lateral strain. Passive earth pressures are larger when larger strains are
start tostrains
decrease as the mobilised strains reduce.
mobilised due to wall deflection and reach a peak at about a depth of 5.5 m. Below this depth, they
start to decrease as the mobilised strains reduce.

(a) (b)
Figure 6. Horizontal
Figure stresses
6. Horizontal (a) in the
stresses soilsoil
in the for: (a)(a)the
for: the“‘retaining
“‘retaining wall-only” case
wall-only” case and(b)
(b)and (b)the
the“structure
“structure on
on the backfill”
the backfill” case. case.
Figure 6. Horizontal stresses in the soil for: (a) the “‘retaining wall-only” case and (b) the “structure
on the backfill” case.

Figure 7. Horizontal earth pressures acting on the retaining wall.

Figure 7. Horizontal
Figure 7. Horizontalearth
earthpressures
pressures acting onthe
acting on theretaining
retaining wall.
wall.
 Geosciences 2020, 10, x FOR PEER REVIEW 9 of 18
Geosciences 2020, 10, 486 9 of 18

In the case of “structure on the backfill”, much larger passive pressures are mobilised, as
required
In the to establish
case the horizontal
of “structure static equilibrium.
on the backfill”, much largerThis is consistent
passive pressureswith the bending
are mobilised, moments
as required
toobserved
establishearlier in Figurestatic
the horizontal 5. Asequilibrium.
seen in Figure
This7,iseven in thiswith
consistent case,the
thebending
passive moments
pressuresobserved
are much
smaller than the Coulomb’s passive pressure line given by Equation (2).
earlier in Figure 5. As seen in Figure 7, even in this case, the passive pressures are much smaller than
the Coulomb’s passive pressure line given by Equation (2).
4. Dynamic Response of the Retaining Wall
4. Dynamic Response of the Retaining Wall
Following the static runs, dynamic analyses were carried out using the same constitutive models
andFollowing
parameters theshown in Table
static runs, 2. The
dynamic final stress
analyses werestate from
carried outthe static
using therun
samewas used as the
constitutive initial
models
stress
and state in the
parameters dynamic
shown runs.
in Table Several
2. The different
final stress stateearthquake loadings
from the static run waswere considered,
used such
as the initial as a
stress
sine sweep motion and sinusoidal motions of different amplitudes, to replicate moderate
state in the dynamic runs. Several different earthquake loadings were considered, such as a sine sweep and strong
earthquakes.
motion and sinusoidal motions of different amplitudes, to replicate moderate and strong earthquakes.

4.1.
4.1.Acceleration
AccelerationTime
TimeHistories
Histories
InInFigure
Figure8,8,the
theaccelerations
accelerationsrecorded
recordedatatthe thetop
topofofthe
theretaining
retainingwall
wallare
arepresented,
presented,along
alongwith
with
the
theinput
inputmotion
motionapplied
appliedatatthe
the base
base nodes. This input
nodes. This input motion
motionwas wasaa1-Hz
1-Hzsinusoidal
sinusoidalmotion
motionwith
with10
10cycles
cyclesofofloading.
loading.ItIthas
hasaapeak
peak acceleration
acceleration of of about 0.23 g. g. The input
input motion
motionwaswasobtained
obtainedfrom
froma a
dynamic
dynamiccentrifuge
centrifugetest
testconducted
conductedearlier
earlieratat60
60ggusing
usingthetheservo-hydraulic
servo-hydraulicearthquake
earthquakeactuator
actuator[16].
[16].
InInthis figure, the wall top accelerations are presented for the cases of “wall-only” and “structure
this figure, the wall top accelerations are presented for the cases of “wall-only” and “structure on on the
backfill”. As expected,
the backfill”. the peak
As expected, theaccelerations are larger
peak accelerations are for the case
larger with
for the thewith
case structure on the backfill.
the structure on the
Additionally, in both cases,
backfill. Additionally, the input
in both cases, accelerations are amplified
the input accelerations by the timebythey
are amplified the reach the wall
time they reachtop.
the
The amplification is larger when the structure is present on the backfill.
wall top. The amplification is larger when the structure is present on the backfill.

Figure8.8.Wall
Figure Walltop
topaccelerations.
accelerations.

4.2. Dynamic Bending Moments

4.2. Dynamic Bending Moments
The additional loading applied by the earthquake causes dynamic bending moments in the
The additional loading applied by the earthquake causes dynamic bending moments in the
retaining wall. In Figure 9, the envelopes of the maximum normalised bending moments are plotted
retaining wall. In Figure 9, the envelopes of the maximum normalised bending moments are plotted
against normalised depth for the cases of “wall-only” and “structure on the backfill” subjected to the
against normalised depth for the cases of “wall-only” and “structure on the backfill” subjected to the
same input motion shown in Figure 8. In Figure 9, it can be seen that the dynamic bending moments
same input motion shown in Figure 8. In Figure 9, it can be seen that the dynamic bending moments
are much larger for the case of the retaining wall with a structure on the backfill. It is also possible to
are much larger for the case of the retaining wall with a structure on the backfill. It is also possible to
compare this figure with the static bending moments shown in Figure 5 and conclude that the dynamic
compare this figure with the static bending moments shown in Figure 5 and conclude that the
bending moments are much larger. In fact, they increased by a factor of almost 2.5 during this 0.23-g
dynamic bending moments are much larger. In fact, they increased by a factor of almost 2.5 during
earthquake for the case with a structure on the backfill. In addition, the location of the peak bending
this 0.23-g earthquake for the case with a structure on the backfill. In addition, the location of the peak
moment appears to have shifted deeper relative to the static case.
bending moment appears to have shifted deeper relative to the static case.
Geosciences
Geosciences 2020, 10, x FOR PEER REVIEW 1010of 18
2020,10,
Geosciences2020, 10,486
x FOR PEER REVIEW 10 of
of18
18

Figure 9. Envelopes of a maximum, normalised dynamic bending moment.

Figure9.9.Envelopes
Figure Envelopesof
ofaamaximum,
maximum,normalised
normaliseddynamic
dynamicbending
bendingmoment.
moment.

However,
However,the
However, thepeak
the peakbending
peak bendingmoments
bending momentsmay
moments maynot
may notoccur
not occurat
occur atatthe
thesame
the sametime
same timeat
time atatall
allelevations
all elevationsof
elevations ofthe
of the
the
retaining
retaining wall.
retainingwall. Therefore,
wall.Therefore,
Therefore, in Figure
in Figure 10,
10, the
in Figure the normalised
10,normalised bending
bending
the normalised moment
bending moment profiles
profiles profiles
moment are plottedare plotted
areatplotted atat
different
different
time time
instants
different instants
timemarked marked
instantsonmarked on
the input the input
motion
on the motion
plotted
input plotted
motionatplottedat
the bottomthe bottom
at theofbottom of the
the figure.
of the figure.
Time Time
instants
figure. instants
Time 1instants
and 3
1were
and 3 were
1 andchosen chosen
3 wereatchosen at 0 acceleration
0 acceleration (maximum
at 0 acceleration (maximum +ve
+ve and −ve
(maximum and −ve velocity),
velocity),
+ve and and
and timeand
−ve velocity), time
instants instants 2 and
2 and 4 were
time instants 4 4were
2 andchosen
were
chosen
atchosen atatmaximum
maximum +ve and+ve
maximum −ve
+ve and −ve
−veaccelerations
accelerations
and (zero
(zerovelocity).
(zero velocity).
accelerations velocity).

0 0 0 0
No structure No structure No structure No structure
0.2 With a structure 0.2 With a structure With a structure With a structure
0.2 0.2
z/Hwall

z/Hwall

0.4 0.4

z/Hwall
0.4 0.4
wall
z/H

0.6 0.6 0.6 0.6

0.8 0.8 0.8 0.8

At time point 1 At time point 2 At time point 5 At time point 6
1 1 1 1
0 0.02 0.04 0 0.02 0.04 0 0.02 0.04 0 0.02 0.04
M/ H 3 M/ H 3 M/ H 3 M/ H 3

0 0 0 0
No structure No structure No structure No structure
0.2 With a structure 0.2
With a structure With a structure With a structure
0.2 0.2
z/Hwall

z/Hwall

0.4 0.4 0.4 0.4

z/Hwall

z/Hwall

0.6 0.6 0.6 0.6

0.8 0.8 0.8 0.8

At time point 3 At time point 4 At time point 7 At time point 8
1 1 1 1
0 0.02 0.04 0 0.02 0.04 0 0.02 0.04 0 0.02 0.04
M/ H 3 M/ H 3 M/ H 3 M/ H 3

Input Input
0.2 0.2
4 8
0.1 0.1
Acc. (g)

Acc. (g)

0 0
1 3 57
-0.1 -0.1

-0.2 -0.2 6
2
-0.3 -0.3
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
Time (s) Time (s)

(a)
(a) (b)
(b)
Figure 10. Dynamic bending moment profiles at different time instants: (a) during an earlier cycle and
Figure
(b) 10. Dynamic
during bending moment profiles at different time instants: (a) during an earlier cycle and
Figure 10.aDynamic
later cycle.
bending moment profiles at different time instants: (a) during an earlier cycle and
(b) during a later cycle.
(b) during a later cycle.
In Figure 10a, the bending moment variations through time instants 1 to 4 during an earlier cycle
for both cases of “wall-only”
InInFigure and “structure on the backfill” are presented.1There is no great reversal of
Figure10a,
10a,the
thebending
bendingmoment
momentvariations
variationsthrough
throughtime
timeinstants
instants 1toto44during
duringananearlier
earliercycle
cycle
bending
for moments, as the retaining wall is locked into a particular curvature due to the soil. However,
forboth
bothcases
casesofof“wall-only”
“wall-only”andand“structure
“structureononthe
thebackfill”
backfill”are
arepresented.
presented.There
Thereisisno
nogreat
greatreversal
reversal
thebending
of peak bending moment does increase from time point 1 to 4, with small drops in value at time
of bendingmoments,
moments,asasthe theretaining
retainingwall
wallisislocked
lockedinto
intoaaparticular
particularcurvature
curvaturedue duetotothe
thesoil.
soil.
However, the peak bending moment does increase from time point 1 to 4, with small
However, the peak bending moment does increase from time point 1 to 4, with small drops in value drops in value
Geosciences 2020,10,
Geosciences 2020, 10,486
x FOR PEER REVIEW 11 of
11 of 18
18

at time points 2 and 3. This is due to the direction of soil acceleration and ensuing changes in dynamic
points 2 and 3. This is due to the direction of soil acceleration and ensuing changes in dynamic earth
earth pressure. The accumulation of bending moments can continue with the increasing number of
pressure. The accumulation of bending moments can continue with the increasing number of loading
loading cycles throughout the earthquake loading. This is illustrated in Figure 10b, in which the
cycles throughout the earthquake loading. This is illustrated in Figure 10b, in which the bending
bending moment variations are plotted for a later loading cycle at time points 5–8. In this figure, it
moment variations are plotted for a later loading cycle at time points 5–8. In this figure, it can be seen
can be seen that the peak bending moments for both cases (i.e., with and without the structure on the
that the peak bending moments for both cases (i.e., with and without the structure on the backfill)
backfill) increased relative to the earlier cycle presented in Figure 10a. The retaining wall can accrue
increased relative to the earlier cycle presented in Figure 10a. The retaining wall can accrue more
more deflections away from the backfill during the dynamic loading but cannot move back into the
deflections away from the backfill during the dynamic loading but cannot move back into the soil. This
soil. This is why the bending moments can either increase when deflection of the wall increases (and,
is why the bending moments can either increase when deflection of the wall increases (and, hence, the
hence, the curvature of the wall increases) or can remain the same but can never decrease by a
curvature of the wall increases) or can remain the same but can never decrease by a significant amount.
significant amount.
5. Soil Response
5. Soil Response
Accelerations can be obtained at any of the solid nodes in the FE discretisation shown in Figure 3a,b.
Accelerations can be obtained at any of the solid nodes in the FE discretisation shown in Figure
In Figure 11, the acceleration-time histories obtained in a soil column in the free field, far away from
3a,b. In Figure 11, the acceleration-time histories obtained in a soil column in the free field, far away
the retaining wall, are presented. The depth of the node below the ground surface is shown in each
from the retaining wall, are presented. The depth of the node below the ground surface is shown in
plot. It can be seen that the input accelerations amplify as they propagate upwards to the ground
each plot. It can be seen that the input accelerations amplify as they propagate upwards to the ground
surface. Additionally, there is very little difference between the “wall-only” case and the “structure on
surface. Additionally, there is very little difference between the “wall-only” case and the “structure
the backfill” case, supporting that this column represents the far-field assumption reasonably well.
on the backfill” case, supporting that this column represents the far-field assumption reasonably well.
Similar acceleration-time history plots can be obtained for the nodes on the active side and passive
Similar acceleration-time history plots can be obtained for the nodes on the active side and passive
side. It may be more instructive to plot the amplification factor defined as the peak acceleration at a
side. It may be more instructive to plot the amplification factor defined as the peak acceleration at a
node normalised by the peak input acceleration. Such a plot is presented in Figure 12 for the free-field,
node normalised by the peak input acceleration. Such a plot is presented in Figure 12 for the free-
active and passive sides.
field, active and passive sides.

Figure 11. Acceleration

Figure 11. Acceleration time
time histories
histories in
in the
the free-field
free-field soil.
soil.

Referring
Referring to Figure 11,
to Figure 11, the
theamplification
amplificationfactor
factorincreases
increasesfrom
from 1 at
1 at thethe bottom
bottom to about
to about 1.5 1.5 at
at the
the
groundground surface
surface in the
in the free
free field.
field. Such
Such amplificationinindry,
amplification dry,relatively
relativelyloose
loose sand
sand layers
layers has
has been
been
previously observed both in earthquake case histories, as well as other numerical
previously observed both in earthquake case histories, as well as other numerical analyses [7,8]. analyses [7,8]. In the
In
active zonezone
the active behindbehindthe the
retaining
retaining wall, a difference
wall, a difference in in
amplification
amplificationfactors
factorsisisnoticed
noticedforforthe
the case
case
of
of “structure
“structure on on the
the backfill”.
backfill”. The
The amplification
amplification that
that occurs
occurs towards
towards thethe ground
ground surface
surface isis almost
almost
completely
completely stopped by the presence of the structure. This may be due to the additional stresses
stopped by the presence of the structure. This may be due to the additional vertical vertical
applied
stresses by the footings
applied by the increasing the confining
footings increasing pressurespressures
the confining in this zone, forcing
in this zone,theforcing
soil to the
act more
soil tolike
act
amore
rigidlike
body. Similarly, on the passive side, a small attenuation of the amplification
a rigid body. Similarly, on the passive side, a small attenuation of the amplification factors factors is noticed.
It
is isnoticed.
not clearIt why
is nottheclear
presence
why of thestructure
presence should influence
of structure the soil
should responsethe
influence on soil
the passive
response side,
onandthe
this aspect may require further investigation.
passive side, and this aspect may require further investigation.
Geosciences 2020, 10, x FOR PEER REVIEW 12 of 18
Geosciences 2020, 10, 486 12 of 18
Geosciences 2020, 10, x FOR PEER REVIEW 12 of 18

Figure 12. Amplification factor in the free-field, active and passive sides.
Figure12.
Figure 12.Amplification
Amplificationfactor
factorin
inthe
thefree-field,
free-field,active
activeand
andpassive
passivesides.
sides.
6. Dynamic Response of the Structure
6.6.Dynamic
Dynamic Response
Response of
ofthe
the Structure
Structure
One of the objectives of this paper is to investigate the response of the structure placed adjacent
to a retaining
One
Oneofofthe wall on
theobjectives the
objectivesofof backfill.
this
this Theisresponse
paper
paper to
is to of thethe
investigate
investigate structure
response
the response was investigated
of the
of structure
the structure firstplaced
placed byadjacent
applying to a
adjacent
sine
to a sweep
retaining wallinput
retaining on themotion
wall the in
backfill.
on theresponse
The
backfill. range ofof0.1–2
The response Hz.
the structure
of theThe peak
was acceleration
investigated
structure firstmagnitude
was investigated by applying
first byof this sweep
a sine inputa
applying
motion
input is
motion onlyin 0.05
the g.
range As ofin the
0.1–2 case
Hz. of
The the
peak sinusoidal
acceleration motion,
magnitude
sine sweep input motion in the range of 0.1–2 Hz. The peak acceleration magnitude of this input this input
of this motion
input was
motion taken
is only from
0.05 a
g.
previous
As in theis
motion dynamic
case
only of the centrifuge
0.05sinusoidal
g. As in the test of
motion, Madabhushi
case this
of theinput et al.
motion was
sinusoidal [22]. In Figure
takenthis
motion, from13, the input
a previous
input motion
motiondynamic
was taken and its Fast
centrifuge
from a
Fourier
test of transform
Madabhushi (FFT)
et al. are
[22]. plotted
In Figure at the
13, the bottom.
input The
motion acceleration
and its
previous dynamic centrifuge test of Madabhushi et al. [22]. In Figure 13, the input motion and its Fast response
Fourier of the
transform roof
(FFT) of the
are sway
plotted
Fast
frame
at the and
bottom.its FFT
The are plotted
acceleration at the
responsetop in
of Figure
the roof 13.
of Due
the to
sway theframedifferential
Fourier transform (FFT) are plotted at the bottom. The acceleration response of the roof of the sway and its settlement
FFT are plottedof the
at footings
the top in
as a consequence
Figure
frame 13.
andDue ofare
to the
its FFT the lateralat
differential
plotted displacement
settlement ofof
the top in Figure the
the retaining
footings
13. Due as wall
to athe describedofearlier,
consequence
differential the of
the lateral
settlement structure
displacement also
the footings
experiences
of the retaining vertical
wall movements,
described which
earlier, theare also
structure shown
also in Figure
experiences 13.
as a consequence of the lateral displacement of the retaining wall described earlier, the structure also It can
vertical be seen
movements, that the
which magnitude
are also
of
showntheseinvertical
experiences Figure accelerations
13. Itmovements,
vertical can be seen is quite
thatsmall
which the and shown
aremagnitude
also moreof towards
these
in Figure the13.higher-frequency
vertical Itaccelerations
can be seen that is end.
the Moreover,
quite small and
magnitude
itofcan
more be seen
towards that
the the frequencies
higher-frequency above
end. 0.25
Moreover, Hz are
it cansignificantly
be seen that
these vertical accelerations is quite small and more towards the higher-frequency end. Moreover, amplified
the by
frequencies the time
above they
0.25 Hzreach
are
the roof
significantly slab. There
amplified is also
by the a sharp
time theyspike
reach in the
the amplification
roof slab. There at
is 1
also
it can be seen that the frequencies above 0.25 Hz are significantly amplified by the time they reach Hz,a which
sharp is
spike the
in natural
the frequency
amplification at
of
1the the
Hz,roof sway
which slab.frame.
is the natural
There is alsofrequency
a sharpofspikethe sway
in theframe.
amplification at 1 Hz, which is the natural frequency
of the sway frame.

Figure 13. Structural response to a sine sweep motion.

Figure 13. Structural response to a sine sweep motion.
Geosciences 2020, 10, x FOR PEER REVIEW 13 of 18
Geosciences 2020, 10, 486 13 of 18

The dynamic response of the structure is also investigated during a stronger 1-Hz sinusoidal
earthquake motionresponse
The dynamic introduced earlier.
of the In Figure
structure is also14,investigated
the input acceleration and the1-Hz
during a stronger horizontal and
sinusoidal
vertical accelerations
earthquake recorded on
motion introduced the roof
earlier. slab are14,
In Figure presented.
the inputThe input acceleration
acceleration in this caseand
and the horizontal has
a maximum
vertical value of 0.23
accelerations g. However,
recorded as this
on the roof slabisare
applied at 1 Hz,
presented. Thewhich
inputis acceleration
the natural frequency ofhas
in this case the
asway frame, value
maximum large of
accelerations were registered
0.23 g. However, by the roof
as this is applied at 1 slab. In the ishorizontal
Hz, which the naturaldirection,
frequencythese
of
accelerations
the reach
sway frame, a magnitude
large of were
accelerations nearlyregistered
0.4 g, while,
by thein the
roofvertical
slab. Indirection, a peakdirection,
the horizontal acceleration of
these
0.1 g was registered.
accelerations The FFT plots
reach a magnitude in Figure
of nearly 0.4 g,14while,
also confirm the amplification
in the vertical of accelerations
direction, a peak accelerationatof1
Hz.g was registered. The FFT plots in Figure 14 also confirm the amplification of accelerations at 1 Hz.
0.1

Figure 14. Structural

Figure 14. Structural response
response to
to aa sinusoidal
sinusoidal motion.
motion.

Another
Anotherinteresting
interestingobservation
observationthat thatwas
wasobtained
obtainedfromfromthese
these dynamic
dynamic analyses
analyses was
was the
the rotation
rotation
suffered by the structure. The structure suffered rotation due to the outward
suffered by the structure. The structure suffered rotation due to the outward movement movement of the retaining
of the
wall creating
retaining walla settlement trough in the
creating a settlement backfill
trough in and a consequent
the backfill and a differential
consequentsettlement
differential between the
settlement
left and right footings. In Figure 15, the rotation suffered by the structure obtained from
between the left and right footings. In Figure 15, the rotation suffered by the structure obtained from the differential
settlement of the
the differential footings is
settlement of plotted, alongis with
the footings the along
plotted, sinusoidal
with input motion. input
the sinusoidal In thismotion.
figure, In
it can
this
be seen itthat
figure, canthe
bestructure
seen thataccumulates
the structurerotation in eachrotation
accumulates cycle ofinthe input
each motion
cycle of theand accrues
input motiona total,
and
residual
accrues asettlement of about
total, residual 4◦ . It isof
settlement also interesting
about 4°. It is to seeinteresting
also that the rotation ceases
to see that thetorotation
accrue ceases
after the
to
end of the
accrue strong
after cycles
the end in the
of the input
strong motion.
cycles in the input motion.
Geosciences 2020, 10, 486 14 of 18
Geosciences 2020, 10, x FOR PEER REVIEW 14 of 18

Figure 15. Rotation

Figure 15. Rotation of
of the
the structure
structure due
due to
to aa sinusoidal
sinusoidal input
input motion.
motion.
7. Dynamic Earth Pressures
7. Dynamic Earth Pressures
The earthquake loading will induce additional dynamic earth pressures on the retaining wall.
The earthquake loading will induce additional dynamic earth pressures on the retaining wall.
Traditionally, limit equilibrium methods such as the Mononobe-Okabe (M-O) method developed
Traditionally, limit equilibrium methods such as the Mononobe-Okabe (M-O) method developed by
by Mononobe [23] and Okabe [24] are used to calculate the dynamic lateral earth pressures. In the
Mononobe [23] and Okabe [24] are used to calculate the dynamic lateral earth pressures. In the M-O
M-O method, the dynamic inertial forces are considered as additional static forces applied to the soil
method, the dynamic inertial forces are considered as additional static forces applied to the soil
wedge. The dynamic active and passive earth pressure coefficients can be calculated by modifying the
wedge. The dynamic active and passive earth pressure coefficients can be calculated by modifying
Coulomb’s equations given earlier in Equations (1) and (2), as follows:
the Coulomb’s equations given earlier in Equations (1) and (2), as follows:
2
cos(2ϕ(φ−−
cos −−
αα θ)θ)
K
K ae =
ae =  q 2 , 2 , (3)
2 sin(ϕ+δ) sin(ϕ−β−θ)
cos(θ) cos (α) cos(δ + α + θ) 1 + sin(φ( + δ)sin(φ ) ( − β )− θ) (3)
cos(θ)cos2 (α)cos(δ + α + θ) [1 + √ cos δ+α+θ cos β−α ]
cos(δ + α + θ)cos(β − α)
cos2 (ϕ + α − θ)
Kpe = 2 , (4)
2 (φ
 q
cos ( θ ) cos2 (α) cos(δ − cos
α + θ ) +
1 α − θ)
+
sin(ϕ+δ) sin(ϕ+β−θ)
K pe = cos(δ−α+θ) cos(β−α) 2,
sin(φ + δ)sin(φ + β − θ) (4)
and noting that cos(θ)cos2 (α)cos(δ − α + θ) [1 + √ ]
kh cos(δ − α + θ)cos(β − α)
tan θ = . (5)
(1 − kv )
and noting that
Using the above equations, it is possible to have the limiting dynamic earth pressure lines as
shown in Figure 16, along with the maximum k hearth pressures obtained from the dynamic
tan dynamic
θ= . (5)
analyses for the sinusoidal input motion with a peak (1acceleration
− kv) of 0.23 g. It should be pointed out
that the M-O
Using themethod
above will result in
equations, it an increasedtoactive
is possible have pressure linedynamic
the limiting and a reduced passive pressure
earth pressure lines as
line relative to the static earth pressures predicted by Coulomb’s equations
shown in Figure 16, along with the maximum dynamic earth pressures obtained from the dynamicshown in Figure 7. It can
be
analyses for the sinusoidal input motion with a peak acceleration of 0.23 g. It should be pointed the
seen that the dynamic passive earth pressures are much larger in the case of the structure on out
backfill
that thecompared
M-O method to the “wall-only”
will result in ancase. These are
increased alsopressure
active larger thanlinethe
andstatic earth pressures
a reduced for both
passive pressure
cases, with the
line relative to peak horizontal
the static earth pressure
earth pressures reaching
predicted nearly 200 equations
by Coulomb’s kPa for theshown
case with a structure
in Figure on
7. It can
the backfill.
be seen that This is an increase
the dynamic passiveofearth
nearly 50 kPa compared
pressures to the static
are much larger in the earth
case ofpressure for thisoncase.
the structure the
For both cases, the dynamic passive earth pressures are lower than those predicted by
backfill compared to the “wall-only” case. These are also larger than the static earth pressures for the M-O method,
although
both cases, this linethe
with on the
peakpassive side is
horizontal closer
earth to this limiting
pressure reachingearthnearly pressure
200 kPa line,
forcompared to thea
the case with
structure on the backfill. This is an increase of nearly 50 kPa compared to the static earth pressure for
 Geosciences 2020, 10, x FOR PEER REVIEW 15 of 18

this case. For both cases, the dynamic passive earth pressures are lower than those predicted15byofthe
Geosciences 2020, 10, 486 18
M-O method, although this line on the passive side is closer to this limiting earth pressure line,
compared to the static earth pressures to the Coulomb passive pressure line shown in Figure 7.
static earth the
However, pressures to the
dynamic Coulomb
active passive pressure
earth pressures line shown
on the backfill side in Figure 7.
obtained byHowever, the dynamic
the numerical analyses
active earth pressures on the backfill side obtained by the numerical analyses are
are larger than those predicted by the M-O line for both the “wall-only” and “structure on larger than those
the
predicted
backfill” by the The
cases. M-Obulge
line for
in both theearth
active “wall-only”
pressureand “structure
at about 2.75-mon depth
the backfill” cases.
attributed The bulge
earlier to the
inpresence
active earth
of thepressure atstill
structure about 2.75-m
persists indepth attributed
the dynamic earlier to the presence of the structure still
analyses.
persists in the dynamic analyses.

Figure
Figure 16. Distributionofofmaximum
16. Distribution maximum horizontal
horizontal earth
earth pressures
pressures acting
acting on retaining
on the the retaining
wall. wall.
M-O:
M-O: Mononobe-Okabe.
Mononobe-Okabe.

In addition to the maximum horizontal earth pressures, it is possible to trace the evolution of these
In addition to the maximum horizontal earth pressures, it is possible to trace the evolution of
earth pressures through a cycle of earthquake loading. In Figure 17a, the horizontal earth pressures
these earth pressures through a cycle of earthquake loading. In Figure 17a, the horizontal earth
are plotted on the active and passive sides at four different time instants, 1–4, for an early cycle.
pressures are plotted on the active and passive sides at four different time instants, 1–4, for an early
These time instants are noted on the input acceleration at the bottom and were explained earlier with
cycle. These time instants are noted on the input acceleration at the bottom and were explained earlier
respect to Figure 10. Unlike the bending moments, the horizontal earth pressures can increase and
with respect to Figure 10. Unlike the bending moments, the horizontal earth pressures can increase
decrease depending on the earthquake loading direction. For example, following the passive earth
and decrease depending on the earthquake loading direction. For example, following the passive
pressure for the “structure on the backfill” case, it can be seen that, between time instants 1 and 2,
earth pressure for the “structure on the backfill” case, it can be seen that, between time instants 1 and
the peak horizontal earth pressure increased from 100 to 140 kPa. However, between time instants
2, the peak horizontal earth pressure increased from 100 to 140 kPa. However, between time instants
3 and 4, this earth pressure decreases from 130 to 105 kPa. Similar observations can be made for the
3 and 4, this earth pressure decreases from 130 to 105 kPa. Similar observations can be made for the
“wall-only” case. Additionally, similar increases in active earth pressure can be seen on the backfill
“wall-only” case. Additionally, similar increases in active earth pressure can be seen on the backfill
side in Figure 17a. For comparison, the horizontal earth pressures are plotted for a later cycle at time
side in Figure 17a. For comparison, the horizontal earth pressures are plotted for a later cycle at time
instants 5–8, as shown in Figure 17b. It is interesting to note that the magnitude of earth pressures is
instants 5–8, as shown in Figure 17b. It is interesting to note that the magnitude of earth pressures is
much larger for this later cycle of earthquake loading, particularly on the passive side. For example,
much larger for this later cycle of earthquake loading, particularly on the passive side. For example,
for the structure in the backfill case, the magnitude of peak horizontal earth pressure increased from
for the structure in the backfill case, the magnitude of peak horizontal earth pressure increased from
125 kPa to 180 kPa for time instants 2 and 6, respectively (i.e., peak negative acceleration points). This
125 kPa to 180 kPa for time instants 2 and 6, respectively (i.e., peak negative acceleration points). This
increase in horizontal earth pressure may be due to the accumulation of locked-in stresses cycle by
increase in horizontal earth pressure may be due to the accumulation of locked-in stresses cycle by
cycle and partly due to the increase in soil density resulting from earthquake-induced settlements. The
cycle and partly due to the increase in soil density resulting from earthquake-induced settlements.
location of the peak horizontal earth pressure shifted to a deeper elevation in this later cycle. Similar
The location of the peak horizontal earth pressure shifted to a deeper elevation in this later cycle.
observations can be made for other time instants as well. On the active side, the increases are much less
Similar observations can be made for other time instants as well. On the active side, the increases are
pronounced. However, the active earth pressures for the “no structure” case seem to have increased
much less pronounced. However, the active earth pressures for the “no structure” case seem to have
more than the “with structure” case in this later cycle, especially at deeper elevations. This aspect
increased more than the “with structure” case in this later cycle, especially at deeper elevations. This
requires further investigation. It should be noted that, in all cases, the M-O predictions underestimate
aspect requires further investigation. It should be noted that, in all cases, the M-O predictions
the horizontal earth pressure on the active side.
underestimate the horizontal earth pressure on the active side.
Geosciences2020,
Geosciences 10,x486
2020,10, FOR PEER REVIEW 1616ofof18
18

(a) (b)

Figure 17. Evolution of the horizontal earth pressures through: (a) an earlier earthquake loading cycle
Figure 17. Evolution of the horizontal earth pressures through: (a) an earlier earthquake loading cycle
and (b) a later earthquake loading cycle.
and (b) a later earthquake loading cycle.
8. Conclusions
8. Conclusions
In this paper, static and dynamic analyses were carried out on a retaining wall in dry sand with
an h/dIn ratio
this paper,
of 0.625.static andwere
These dynamic analyses
compared to awere carriedaout
case where swayonframe
a retaining wallwas
structure in dry sandon
placed with
the
an h/d ratio
backfill of 0.625.
adjacent to theThese were compared
retaining to a case
wall. The main whereofathis
objective sway frame was
research structure was placed
to investigate theon the
strong
backfill adjacent
soil-structure to the retaining
interaction that maywall. Thedue
occur main
to objective
the presenceof this research
of the was Using
structure. to investigate
the staticthe strong
analyses,
soil-structure interaction that may occur due to the presence of the
it was shown that the deflections and bending moments in the retaining wall increased due to structure. Using the static
the
analyses, it was shown that the deflections and bending moments in the retaining
presence of the structure on the backfill. In addition, the horizontal pressure bulb that developed in wall increased due
to theofpresence
front the retaining of the wallstructure on the
also increased backfill.
in size whenIntheaddition,
structurethewashorizontal
present on pressure
the backfill. bulb that
Further,
developed in front of the retaining wall also increased in size when the structure
the horizontal earth pressures obtained from these analyses were compared to Coulomb’s predictions. was present on the
backfill.
UsingFurther, the horizontal
the dynamic analyses, earth pressures
the response obtained
of the from
top of the these analyses
retaining wall was were
foundcompared
to increaseto in
Coulomb’s predictions.
the presence of the structure on the backfill. The maximum bending moments due to the earthquake
Using
loading the dynamic
increased analyses, the
in magnitude. Byresponse of the
considering thetop of the retaining
bending momentswall was found
at different timeto increase
instants,
in the presence of the structure on the backfill. The maximum
it was shown that these bending moments locked in and did not reduce in value. The dynamic bending moments due to the
earthquake loading increased in magnitude. By considering the bending
response of the soil was considered next, and it was shown that the amplification ratio below moments at different time
the
instants, it was shown that these bending moments locked in and did not
structure on the backfill was smaller than for the “wall-only” case, suggesting more “rigid body”-like reduce in value. The
dynamic response
accelerations below ofthe
the footings
soil was considered
owing to thenext, and it was
increased shownstress.
confining that the
Theamplification ratio below
structural response was
the structure on the backfill was smaller than for the “wall-only” case, suggesting
considered next, and the large amplification of accelerations of the roof slab was explained. Finally, more “rigid body”-the
like accelerations below the footings owing to the increased confining stress.
dynamic earth pressures acting on the retaining wall were considered both in terms of their maxima, The structural response
was considered
as well next, and with
as their evolution the large
time.amplification
Unlike the bendingof accelerations
moments,ofit thewasroof
shownslabthat
wasthe explained.
dynamic
Finally, the dynamic
earth pressures can both earth pressures
increase andacting
decreaseon depending
the retaining onwall were considered
the direction bothaccelerations.
of the input in terms of
their maxima, asdynamic
The maximum well as their
earthevolution
pressureswith
weretime. Unliketothe
compared thebending moments,
traditional it was shown
M-O method. that
It was seen
the dynamic earth pressures can both increase and decrease depending on the direction of the input
accelerations. The maximum dynamic earth pressures were compared to the traditional M-O method.
Geosciences 2020, 10, 486 17 of 18

that the M-O method overpredicts the dynamic passive earth pressures, while it underpredicts the
active earth pressures.

Author Contributions: X.G. carried out the numerical analyses reported in this paper, with help from the second
author. G.S.P.M. helped prepare the first draft of this manuscript, while the first author created all the figures
shown in this paper. All authors have read and agreed to the published version of the manuscript.
Funding: This research received no external funding.
Acknowledgments: The first author would like to thank the financial support from China Scholarship Council
and Cambridge Trust for her PhD studies.
Conflicts of Interest: The authors declare no conflict of interest.

References
1. Aarsleff. Available online: www.aarsleff.co.uk (accessed on 20 September 2020).
2. Pelà, L.; Aprile, A.; Benedetti, A. Seismic assessment of masonry arch bridges. Eng. Struct. 2009, 31,
1777–1788. [CrossRef]
3. D’Amato, M.; Laterza, M.; Casamassima, V.M. Seismic performance evaluation of multi-span existing
masonry arch bridge. Open Civ. Eng. J. 2017, 5, 1191–1207. [CrossRef]
4. Marcheggiani, L.; Clementi, F.; Formisano, A. Static and dynamic testing of highway bridges: A best practice
example. J. Civ. Struct. Health Monit. 2020, 10, 43–56.
5. Madabhushi, S.P.G.; Patel, D.; Haigh, S.K. Geotechnical Aspects of the Bhuj Earthquake. In EEFIT Report;
Madabhushi, S.P.G., Haigh, S.K., Eds.; Institution of Structural Engineers: London, UK, 2005; pp. 1–22.
6. Zeng, X.; Steedman, R.S. On the behaviour of quay walls in earthquakes. Géotechnique 1993, 43, 417–431.
[CrossRef]
7. Madabhushi, S.P.G.; Zeng, X. Seismic Response of Flexible Cantilever Retaining Walls with Dry Backfill.
Geomech. Geoeng. 2006, 1, 275–290. [CrossRef]
8. Madabhushi, S.P.G.; Zeng, X. Simulating Seismic Response of Cantilever Retaining Walls with Saturated
Backfill. J. Geotech. Geoenviron. Eng. 2007, 133, 539–549. [CrossRef]
9. Chan, A.H.C. A Generalised Fully Coupled Effective Stress Based Computer Procedure for Problems in Geomechanics;
Swandyne User Manual: Swansea, UK, 1988.
10. Cilingir, U.; Haigh, S.K.; Madabhushi, S.P.G.; Zeng, X. Seismic behaviour of anchored quay walls with dry
backfill. J. Geotech. Geoenviron. Eng. 2011, 6, 227–235. [CrossRef]
11. Callisto, L.; Soccodato, F.M. Seismic design of flexible cantilevered retaining walls. J. Geotech. Geoenviron. Eng.
2010, 136, 344–354. [CrossRef]
12. Conti, R.; Viggiani, G.M. A new limit equilibrium method for the pseudo-static design of embedded
cantilevered retaining walls. Soil Dyn. Earthq. Eng. 2013, 50, 143–150. [CrossRef]
13. Yeganeh, N.; Bazaz, J.B.; Akhtarpour, A. Seismic analysis of the soil–structure interaction for a high rise
building adjacent to deep excavation. Soil Dyn. Earthq. Eng. 2015, 79, 149–170. [CrossRef]
14. Semblat, J.F.; Brioist, J.J.; Luong, M.P. Mean Stress Dependent Damping of Seismic Waves in Sand; Proceedings of
the Earthquake Geotechnical Engineering: Rotterdam, The Netherlands, 1999; pp. 95–100.
15. Haigh, S.K.; Ghosh, B.; Madabhushi, S.P.G. A numerical investigation into effects of single and multiple
frequency earthquake input motions. J. Soil Dyn. Earthq. Eng. 2003, 23, 691–704.
16. Desai, C.S.; Siriwardane, H.J. Constitutive Laws for Engineering Materials with Emphasis on Geologic Materials;
Prentice Hall: Englewood Cliffs, NJ, USA, 1984.
17. Wood, D.M. Geotechnical Modelling; Taylor and Francis: London, UK, 2001.
18. Zhang, J.; Stewart, D.P.; Randolph, M.F. Modeling of Shallowly Embedded Offshore Pipelines in Calcareous
Sand. J. Geotech. Geoenviron. Eng. 2002, 128, 363–371. [CrossRef]
19. Madabhushi, S.P.G.; Zeng, X. Behaviour of gravity quay walls subjected to earthquake loading. Part II:
Numerical Modelling. JNL Geotech. Eng. 1998, 124, 418–428. [CrossRef]
20. Pande, G.N.; Zienkiewicz, O.C. Soil Mechanics, Transient & Cyclic Loads Constitutive Relations & Numerical
Treatment; Wiley: Chichester, NH, USA, 1982.
21. Ng, C.W.W.; Zheng, G.; Ni, J.; Zhou, C. Use of unsaturated small-strain soil stiffness to the design of wall
deflection and ground movement adjacent to deep excavation. Comput. Geotech. 2020, 119, 103375. [CrossRef]
Geosciences 2020, 10, 486 18 of 18

22. Madabhushi, S.P.G.; Haigh, S.K.; Houghton, N.E.; Gould, E. Development of a servo-hydraulic earthquake
actuator for the Cambridge Turner Beam Centrifuge. Int. J. Phys. Model. Geotech. 2012, 12, 77–88. [CrossRef]
23. Mononobe, N. Earthquake-proof construction of masonry dams. In Proceedings of the World Engineering
Conference, Tokyo, Japan, 19–20 October 1929; pp. 275–278.
24. Okabe, S. General theory of earth pressures. J. Jpn. Soc. Civ. Eng. 1926, 12, 1083–1088.

Publisher’s Note: MDPI stays neutral with regard to jurisdictional claims in published maps and institutional
affiliations.

© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC BY) license (http://creativecommons.org/licenses/by/4.0/).