Geotextiles and Geomembranes 44 (2016) 489e503

Contents lists available at ScienceDirect

Geotextiles and Geomembranes

journal homepage: www.elsevier.com/locate/geotexmem

Modelling of geocell-reinforced subballast subjected to cyclic loading

M. Mahdi Biabani a, b, Buddhima Indraratna c, b, *, 1, Ngoc Trung Ngo d, b, 2
a
Centre for Geomechanics and Railway Engineering, Faculty of Engineering, University of Wollongong, Wollongong City, NSW 2522, Australia
b
ARC Centre for Excellence for Geotechnical Science and Engineering, Australia
c
Distinguished Professor in Civil Engineering, Centre for Geomechanics and Railway Engineering, University of Wollongong, Wollongong, NSW 2522,
Australia
d
School of Civil, Mining and Environmental Engineering, University of Wollongong, Wollongong City, NSW 2522, Australia

a r t i c l e i n f o a b s t r a c t

Article history: This paper presents the experimental and numerical studies of geocell-reinforced subballast subjected to
Received 18 September 2015 cyclic loading. A series of laboratory experiments were conducted using a large-scale prismoidal triaxial
Received in revised form apparatus that was subjected to relatively low confining pressures of s0 3 ¼ 10e30 kPa and a frequency of
18 January 2016
f ¼ 10 Hz. Numerical simulations were performed using the commercial finite element package ABAQUS
Accepted 14 February 2016
Available online 10 March 2016
in three dimensions to realistically model cellular confinement, and to study the effectiveness of geocell
reinforcement on subballast. A cyclic loading with a periodic and positive full-sine waveform was
adopted to model the geocell-reinforced subballast, which is similar to the load carried out in the lab-
Keywords:
Geosynthetics
oratory. The results of numerical modelling agreed well with the experimental data, and showed that
Geocell reinforcement geocell could effectively decrease the lateral and axial deformations of the reinforced subballast. The
Subballast numerical model was also validated by the field data, and the results were found to be in good agree-
Cyclic loading ment, indicating that the proposed model was able to capture the load-deformation behaviour of geocell-
Plane strain reinforced subballast under cyclic loading. A parametric study was also carried out to evaluate the effect
Numerical modelling of the subballast strength and geocell stiffness on the mobilized tensile strength in the geocell mattress.
It was found that the maximum mobilized tensile stress occurs on the subballast with the lowest degree
of stiffness. Also the results revealed that lateral displacement decreased further by increasing geocell
stiffness, and geocell with a relatively low stiffness performs very well compared to the geocell with a
higher stiffness.
© 2016 Elsevier Ltd. All rights reserved.

1. Introduction maintenance and rehabilitation of track substructure (Indraratna

et al., 2013). However, to compete with other transportation
Railway networks are one of the major transport systems used modes and meet the ever growing demand for public and freight
for carrying passengers, and transporting freight and bulk com- transport, the railway industry will face challenges to improve the
modities between major mines and ports in many countries track operational efficiency and decrease maintenance and infra-
worldwide. Considering an acceptable ride quality, relatively low structure costs. The foundation of a conventional ballasted track
cost, and growing demand from industry and commuters, railways consists of granular material layers that help to transmit and
have become more popular than other modes of transportation. distribute the induced cyclic load to the underlying subgrade at an
Nevertheless, the sustainable development of rail infrastructure acceptable or controlled stress level (Suiker et al., 2005; Selig and
requires a significant amount of cost associated with track Waters, 1994). To date, reinforcing track substructure using a
planar reinforcement is commonly deployed as it has been proven
to reduce the axial and lateral deformation of ballast and subballast
* Corresponding author. Faculty of Engineering and Information Sciences, Uni-
layers, and to improve the stability of track substructure under
versity of Wollongong, Wollongong, NSW 2522, Australia. Tel.: þ61 2 4221 3046; cyclic train loading (Ngo et al., 2014; Indraratna et al., 2011a,b;
fax: þ61 2 4221 3238. Kwon and Tutumluer, 2009; Atalar et al., 2001). Past studies have
E-mail addresses: mmb958@uowmail.edu.au (M.M. Biabani), indra@uow.edu.au shown that cellular reinforcement can provide much better lateral
(B. Indraratna), trung@uow.edu.au (N.T. Ngo).
1 confinement to infill granular soils than planar reinforcements
Tel.: þ61 2 4221 3046; fax: þ61 2 4221 3238.
2
Tel.: þ61 2 4221 4892; fax: þ61 2 4221 3238. (Indraratna et al., 2015; Hegde and Sitharam, 2015; Huang et al.,

http://dx.doi.org/10.1016/j.geotexmem.2016.02.001
0266-1144/© 2016 Elsevier Ltd. All rights reserved.
490 M.M. Biabani et al. / Geotextiles and Geomembranes 44 (2016) 489e503

2011; Han et al., 2011). The performance of geocell mattress in the performance of geocell reinforcement under cyclic loading is
stabilizing different types of infill soils subjected to monotonic the key requirement, which is needed for its design and application
loading has been investigated in several studies (Biabani and in ballasted rail tracks.
Indraratna, 2015; Wang et al., 2013; Tafreshi and Dawson, 2012; The development of a numerical model is inevitable in order to
Yang et al., 2010; Pokharel et al., 2010; Saride et al., 2009). A establish proper design guidelines based on safety and economic
summary of research outcomes of selected past studies is given in considerations. Considering the computational effects involved, a
Table 1. In addition, there are a few studies available, which have two-dimensional (2D) model often become more popular than a
investigated the performance of granular material in plane-strain three-dimensional (3D) model for plane strain conditions (Hegde
condition (Radampola et al., 2008; Wanatowski et al., 2008; and Sitharam, 2013; Mehdipour et al., 2013; Yu and Sloan, 1997).
Radampola, 2006; Peters et al., 1988). However, understanding An equivalent composite approach has often been used to model

Table 1
Summary of research outcomes of previous studies.

Material Reinforcement Test specimen scale Research methodology Salient research outcomes Limitations Reference
type type

Sand and Geocell 900  900  600 Numerical simulation 1) The geometry of the geocell has a significant Monotonic loading only, Saride et al.
clay L  W  H (mm) (FLAC3D). impact on the load carrying capacity and hence cannot interpret (2009)
reducing the settlement of the soil bed. cyclic loading behaviour.
2) The results revealed that having three layers
of planar geogrids can be led to provide
optimum performance improvement.
Clayey Geocell 50,000  25,000 Experimental and 1) Locally available material can be used as (1) Monotonic loading Latha and
sand W  H (mm) numerical investigation of infill material in the absence of granular (2) Equivalent composite Rajagopal
and soft 1000 and 2000 mm geocell supported material. model (i.e. soil and geocell (2007)
clay Geocell height embankment (GEOFEM) 2) Performance of the reinforced embankment are integrated as one
was significantly improved by increasing the material)
aspect ratios (optimum aspect ratio of 1.0)
Sand Geocell and 900  900  600 Experimental results and 1) Geocell reinforcement was found to be more 1) Monotonic loading Latha and
planar geogrid L  W  H (mm) numerical investigation on effective than other types of reinforcement. 2) The mobilised stress Somwanshi
the bearing capacity of 2) Numerical results confirmed that by over the geocell was not (2009)
square footings. transferring the footing load to deeper depth, investigated.
stress and strain underneath of the footing will
be markedly reduced.
Gravel Geocell 1524  610  546 Numerical modelling of 1) Providing geocell reinforcement 1) Confining pressure Leshchinsky
L  W  H (mm) behaviour of railway significantly reduced vertical deformation, assumed constant during and Ling
ballasted structure with particularly for material with lower quality. the entire simulation. (2013a,b)
geocell 2) Geocell reinforcement successfully arrested 2) Diamond shaped
confinement lateral spreading along the slope of the railroad geocell pockets, which are
substructure. different to actual geocell
configuration.
Aggregate Geocell 1000  840  1000 Numerical modelling for 1) A three-dimensional mechanisticeempirical 1) Confining pressure Yang et al.
and L  W  H (mm) geocell-reinforced unpaved (MeE) model for geocell-reinforced unpaved remains constant during (2013)
sand Geocellthickness ¼ 100, roads (FLAC3D). roads was developed. the entire simulation.
150 mm 2) A compaction-induced residual stress in the 2) Diamond shaped
base layer was determined using the hysteretic geocell pockets for
k0-loading model. simplicity
Sand and Geocell and 450  450  600 Numerical modelling of 1) Tensile strength had a significant impact on Monotonic loading only. Hegde and
clay geogrid L  W  H (mm) geocell-reinforced sand footing strength, compared to other Sitharam
(FLAC3D). reinforcement properties. (2014);
2) Performance of the foundation was (2015)
improved further by proving additional planar
geogrid.
Gravel Geogrid 300  200  400 Numerical modelling of 1) It was found that pullout force to be greater Geocells not used. Ferellec and
D  W  H (mm) ballast and geogrid for the clumps than for the spheres. McDowell
interaction in pullout 2) Much more localised deformation of the (2012)
testing (DEM). geogrid
observed as result of stronger grid-particle
interlock.
Gravel Geogrid 700  300  450 Numerical modelling of 1) settlement of ballast decreased significantly 1) Geocells were not used. Chen et al.
L  W  H (mm) geogrid-reinforced ballast due to geogrid. 2) Limited number of (2012)
under cyclic loading (DEM). 2) The optimum location for the geogrid was cycles.
found to be at 100 mm above the base
(confined test) and 50 mm from the subballast
(unconfined test).
Sand Geocell 480  380  100 Numerical modelling of 1) Bearing capacity of the foundation increased 1) Study is limited to a Han et al.
L  W  H (mm) geocell-reinforced sand significantly due to geocell reinforcement. single geocell pocket. (2008)
3D
(FLAC ). 2) Maximum displacement and tension were 2) Monotonic loading.
found to be close to the bottom of the geocell
pocket.
Sand Geocell 2000  2000  700 Experimental results of 1) The optimum embedded depth of first layer 1) Limited number of Moghaddas
L  W  H (mm) rubberesoil mixture and of geocell and vertical spacing of geocell layers cycles. Tafreshi
geocell under repeated were about 0.2 times of loading plate diameter. et al. (2014)
loading. 2) The maximum and plastic deformation
increased by increasing number of load cycles.
 M.M. Biabani et al. / Geotextiles and Geomembranes 44 (2016) 489e503 491

geocell-reinforced soil in a 2D environment by modelling the and it was used to investigate the stressestrain behaviour of the
reinforced soil as a new layer with improved strength and stiffness unreinforced and geocell-reinforced subballast subjected to cyclic
(Hegde and Sitharam, 2013; Latha and Somwanshi, 2009; Bathurst loading (Indraratna et al., 2015). The area of the test specimen in the
and Knight, 1998). However, a 2D model cannot accurately capture prismoidal triaxial chamber was selected based on Australian stan-
the additional confinement developed through circumferential dard gauge for heavy haul track. Therefore, in the direction perpen-
strains due to the complex shape of the geocell mattress (with its dicular to sleepers (tie), 400 mm symmetrically on each side of one
honeycomb like structure). Very limited number of studies have rail (i.e. 800 mm) was taken, and this is equal to 1/3 of the total
been investigated the performance of geocell-reinforced soil in a 3- sleeper length (l) of 2400 mm (also termed as effective sleeper length
dimensional framework under monotonic loading (Hegde and by Jeffs and Tew, 1991); and (ii) in the direction parallel to sleeper
Sitharam, 2014). Also, to the authors knowledge, there has only (tie), a distance equalling the sleeper spacing of 600 mm was
been limited research carried out on the effect of geocell mattress considered (Fig. 2(a)). This explains the plan area of
on railway substructure, where the benefits of geocell subjected to 800 mm  600 mm of the test specimen. The subballast material had
cyclic loading has not been studied in details either in laboratory or a total depth of 450 mm, of which the upper 150 mm was stabilised
numerical modelling (Leshchinsky and Ling, 2013a; Wang et al., by geocell. The material for subballast used in this study was a locally
2013; Mehdipour et al., 2013; Tavakoli Mehrjardi et al., 2012; available crushed basalt, collected from a quarry near Wollongong
Choudhury, 2009). The development of a numerical model that (NSW, Australia). The particle size distribution adopted for the sub-
has been calibrated accurately by laboratory and field measure- ballast was within the rail industry specified range (D50 ¼ 3.3 mm,
ments is thereby inevitable to understand the performance of Dmax ¼ 19 mm, Dmin ¼ 0.075 mm, Cu ¼ 16.3, Cc ¼ 1.3, gd ¼ 19 kN/m3).
geocell-reinforced subballast and to propose a proper design A predetermined mass of subballast was placed inside the cubical
guideline for ballasted rail track, while considering the confine- box in several layers and compacted in dry conditions using a
ment effect of the geocell. An attempt was made in this study to vibratory hammer to achieve a relative density (DR) of about 77%,
carry out large-scale cubical tests of geocell-reinforced subballast which is representative of the density of subballast in the field
and to develop a 3D numerical model to simulate the composite (gbulk ¼ 20.5 kN/m3). A geocell mattress was placed onto the surface
system, and capture the actual geometry of geocell pockets and its of the subballast. All the specimens were prepared until the layer of
additional confinement to the subballast. subballast reached a final height of 450 mm. A geocell mattress made
from polyethylene materials, that was connected at the joints to
2. Experimental study create a three-dimensional cellular form (i.e. depth ¼ 150 mm, ul-
timate tensile strength ¼ 9.5 kN/m (ASTM D4885),
In order to obtain a more realistic understanding of subballast thickness ¼ 1.3 mm, density ¼ 950 kg/m3) was used.
under cyclic loading, experimental work was conducted to mimic the The experiments were conducted under plane strain condition,
true field conditions, where the intermediate stress differs from the where any lateral movement in the longitudinal direction (parallel
minor principal stress (s0 2 s s0 3 ). As a result, the large-scale pris- to the track) was restricted (ε2 ¼ 0). The walls were allowed to
moidal triaxial apparatus (800 mm long, 600 mm wide and 600 mm move laterally in the direction parallel to the sleeper (or tie)
high) was designed and built at the University of Wollongong (Fig. 1), (ε3 s 0), to simulate a long straight section of track. The

Fig. 1. Cubical triaxial used in this study.

492 M.M. Biabani et al. / Geotextiles and Geomembranes 44 (2016) 489e503

Unit cell 3. Finite element analysis

Model tests were simulated in the finite element method (FEM)

sleeper (tie)
where the material properties obtained from laboratory tests and
model geometry followed the cubical apparatus carried out in the
laboratory (800 mm  600 mm  450 mm), as illustrated in
Geocell reinforced subballast Fig. 2(a) and (b). The size of the specimen represents the effective
sleeper length (le ¼ 800 mm) and distance between two adjacent
Ballast 300 mm sleepers (tie in North America) (b ¼ 600 mm) with a depth of
150 mm h ¼ 450 mm (Indraratna et al., 2015; Jeffs and Tew, 1991). A cyclic
300 mm loading stress, caused by a train and exerted beneath the ballast,
Subgrade
was applied directly onto the subballast layer, where the loading
800 mm characteristics were similar to those used in the laboratory. As
(a)
result, no ballast or sleeper (tie) was considered in this simulation.
Geocell mattress In the current analysis, a subballast layer was modelled with a
depth of 450 mm to represent variations in the subballast height as
measured in the field. One of the limitations of the prismoidal test
chamber is that it may not be deep enough to represent the actual
subgrade depth in the field, and therefore, the actual measured
strains in the prismoidal test chamber with 450 mm depth may not
450 mm be truly representative of the field conditions except where the
Pressure exerted
by train subgrade is rock or very stiff and dense gravel.

3.1. Material properties

(b) An elasto-plastic material with non-associative behaviour was

used to model the subballast. To capture the elasto-plastic behav-
Fig. 2. Finite element idealisation of typical rail environment. iour of subballast, the DruckerePrager yield criterion was adopted
in this analysis. The shear strength parameters (i.e., f, j) could be
obtained using appropriate triaxial equipment (Indraratna et al.,
laboratory tests were conducted in a stress-controlled manner, 2015; Bolton, 1986). A linear elastic-perfectly plastic material was
where the magnitudes of the cyclic stresses were computed based used to model the geocell mattress. The elastic properties of the
on 30 tons/axle load (Jeffs and Tew, 1991; Atalar et al., 2001). To geocell strips were determined in the laboratory and then incor-
examine the effects confining pressure on the behaviour of sub- porated into a finite element model in ABAQUS. A hexagonal shape
ballast, cyclic drained triaxial tests were conducted at confining was used to model the geometry of the geocell pockets, as it is
pressures (s0 3 ) of 5, 10, 15, 20, 30 kPa and frequencies (f) of 10, 20, similar to the actual curvature of the geocell mattress used in the
and 30 Hz. During the laboratory experiment, the lateral move- laboratory. Since, subballast is never clean or perfectly granular, as
ment of the vertical walls in the direction of the intermediate it is always contaminated with small amount of cohesive clay. The
principle stress (s0 2 ) was prevented by locking the castors (i.e. actual measured cohesion values from large-scale triaxial tests
ε2 ¼ 0), while the vertical walls in the orthogonal (i.e. transverse) (depending upon the extent of clay fouling) can sometimes exceed
direction of minor principle stress (s0 3 ) were allowed to move 5 kPa as shown by Tennakoon (2012). Accordingly, a small value of
laterally. Initially, a monotonic strain-controlled load was applied cohesion (2 kPa) was used in the simulations to improve numerical
to the specimen at a rate of 1 mm/min until a mean level of cyclic stability and to ensure that the model converged in a reasonable
deviator stress was attained. Subsequently, a stress controlled computation time while not critically affecting the results. A sum-
cyclic loading using a positive full-sine waveform was applied to mary of the material properties used in the analysis is provided in
the specimens where a maximum and minimum stress of Table 2. Considering the extensive computation time required to
qmax ¼ 166 kPa and qmin ¼ 41 kPa was used to simulate subballast run a model under cyclic loading, the simulations were carried out
under a heavy haul freight network operating in NSW (Indraratna up to 10,000 cycles, where most of the subballast deformation had
et al., 2015). already occurred, as seen in the laboratory.
The laboratory outcomes revealed that confining pressure and
frequency have a significant impact on the behaviour of the 3.2. Boundary conditions
granular material under cyclic loading. The laboratory results
confirmed that under cyclic loading, geocell mattress can offer To simulate field track conditions, lateral displacement in the
additional confinement (Ds0 3 ) to the infill material (i.e. other than direction of the intermediate principal stress s0 2 (i.e. parallel to the
the confining pressure available from sleepers and shoulder tracks) was constrained (ε2 ¼ 0). Meanwhile, the model was
ballast), and help to decrease the axial and lateral deformations allowed to move in the direction of the major principal stress s0 1
(Indraratna et al., 2015). Also it is noted that due to cyclic loading, (i.e. vertical settlement) and the minor principal stresses s0 3 (i.e.
the magnitude of Ds0 3 would be increase as the number of load parallel to the sleepers) (ε1 , ε3 s 0). The base of the model was
cycle increases and when the densely compacted infill material restricted to any displacement, as shown in Fig. 3(a). To obtain an
dilates and thereby increases the magnitude of hoop stress. A optimum size mesh for the model, a sensitivity analysis of mesh
summary of the key aspects of experimental outcomes can be density was carried out for a model with a varying number of el-
briefly summarised as: (i) the mobilised hoop stress of the geocell ements. Vertical settlements of the geocell-reinforced subballast
pockets is generated as a result of the dilation of the infilled soil obtained from the models with different numbers of elements were
during shearing; and (ii) the magnitude of hoop stress varies with compared at a load cycle of N ¼ 1000 (Fig. 3(b)). The results showed
the geocell modulus. that by increasing the number of elements, the vertical settlement
 M.M. Biabani et al. / Geotextiles and Geomembranes 44 (2016) 489e503 493

Table 2
Finite element properties of subballast and geocell used in current study.

Subballast& subgrade

Density (kg/m3) Poisson's ratio, n Elastic modulus (MPa) Cohesion yield stress (kPa) Angle of dilation, j (degree) Friction angle, f (degree)

1955 0.3 4e40 2 9 39

Geocell

Density (kg/m3) Poisson's ratio, n Secant modulus (3% strain), Egeocell (GPa)

950 0.3 0.3e5

Plate

Density (kg/m3) Poisson's ratio, n Elastic modulus, Eplate (GPa)

2000 0.3 500

Subballastegeocell interface parameters

Eslip tolerance Friction coefficient, d

0.005 2/34

Y including the hard normal contact as used by Leshchinsky and Ling

(2013a). Given the highly random nature of particle orientations
X within the subballast assembly, it was assumed that the angle of
shearing resistance between the aggregates and geocell mattress
Z was isotropic, considering that the membrane texture is usually
uniform. In the numerical simulation, the contact between the
infilled granular soil and geocell wall was modelled as an interface
element with a fixed angle of shearing resistance. The interface
3
friction angle (f ¼ 39 ) between the infill and cell wall was
3
determined in the laboratory using large-scale direct shear tests.
The direct shear tests were conducted at relatively low normal
FEM model
stress that varied from 1 to 45 kPa. As a result, the interface contact
for both horizontal and vertical direction was modelled by
assuming 2/3 of the interface friction angle (f ¼ 39 ).
Displacement in the direction
3.3. Loading conditions
to Z axis is restricted ( 2=0)
(a)
In order to have uniform deformation, initially a constant
confining pressure (s0 3 ) was applied to the unit cell prior to cyclic
loading to simulate mean pressure due to the geostatic stresses
inherent in a railway track. Two loading stages were used to
simulate the cyclic loading as conducted in the laboratory. A static
load with a magnitude of smean ¼ 104 kPa was applied to the top of
the specimen in the strain-controlled model. The maximum stress
in the cyclic loading pattern was chosen to provide the most
possible critical stress that could be applied to the subballast as
3
measured by Indraratna et al. (2015), which the details are given in
the Appendix. After completing the static loading, a cyclic load was
superimposed onto the monotonic loading at different confining
pressures, as shown in Fig. 4. During the application of cyclic load,
each load cycle was returned to a fully unloaded stage to represent a
passing train wheel, where a periodic and positive full sine wave-
form was considered. The cyclic load was performed in a stress
controlled mode with a frequency of f ¼ 10 Hz, where the maximum
and minimum load amplitudes were 166 kPa and 41 kPa, respec-
(b) tively (Fig. 4) (Indraratna et al., 2015; Jeffs and Tew, 1991). Numerical
simulation were carried out at relatively low confining pressures
Fig. 3. (a) Typical boundary condition for unit cell and (b) vertical settlement predicted (5 kPa  s0 3  30 kPa) to simulate the typical small confinement
by FE model with different number of elements.
provided by the shoulder ballast and sleeper (tie) in actual tracks.
(SV) increases, and when the number of elements is beyond 9,000,
the increment in vertical settlement can be negligible. Based on this 4. Results and discussions
preliminary finding, a FE model with 9380 elements with 12,624
nodes (ABAQUS) was selected to simulate the geocell-reinforced 4.1. Vertical settlement
subballast. The type of elements chosen were C3D8R (i. e. eight-
node reduced integration element). The interaction between sub- Fig. 5(a) and (b) show the vertical settlement contours for
ballast and geocell strips was modelled with contact elements, unreinforced and geocell-reinforced subballast subjected to
494 M.M. Biabani et al. / Geotextiles and Geomembranes 44 (2016) 489e503

However, placing a geocell mattress beneath the footing

significantly decreased the magnitude of vertical settlement in the
reinforced subballast, where settlement was only SV ¼ 2.57 mm for
the reinforced specimen, compared to SV ¼ 8.45 mm for the unre-
inforced specimen [Fig. 5(b)]. Maximum settlement occurred close
to the edge of the specimen (i.e. the subballast material was outside
the cellular mattress); but the vertical deformation of subballast at
the centre of the geocell mattress was much less than unreinforced
subballast, indicating the effectiveness of geocell in reducing the
stress to the lower soil layer.
Fig. 6(a) and (b) show a comparison between the vertical set-
tlement (SV) of unreinforced and geocell-reinforced subballast ob-
tained from the FEM model, and that measured in the laboratory at
different confining pressures (s0 3 ¼ 10e20 kPa) and frequency of
f ¼ 10 Hz. The vertical deformation obtained from the FEM simu-
lation agrees with the experimental data. Most of deformation in
unreinforced subballast occurs in the early stage of loading
(N  5000 cycles). This is because immediate settlement takes place

Fig. 4. Schematic illustration of cyclic loading.

3
cyclic loading (frequency of f ¼ 10 Hz; confining pressure of
3
s0 3 ¼ 10 kPa) at a load cycle of N ¼ 10,000. The numerical results
3
indicate that the subballast experiences the highest vertical
deformation of about SV ¼ 8.45 mm under the footing surface
(Fig. 5(a)), and this gradually decreases with the depth, where
settlement at a depth of h ¼ 200e250 mm is about
SV ¼ 5.28 mm.

Controlled steady state

(a)

3
3
3

(b)
Fig. 5. FEM predicted vertical settlements: (a) unreinforced; and (b) geocell-reinforced Fig. 6. Vertical deformation of (a) unreinforced and (b) geocell-reinforced subballast
subballast. against number of cycles (N): laboratory measurements vs. model predictions.
 M.M. Biabani et al. / Geotextiles and Geomembranes 44 (2016) 489e503 495

during the initial number of cycles. The experimental and numer-

ical results both confirm that the rate of settlement diminishes to a
controlled steady state after N  5000 cycles. The numerical and
experimental results also confirm that the magnitude of SV de-
creases by increasing s0 3 . Fig. 6(a) shows that the axial settlement
decreases from SV ¼ 17.5 mm to SV ¼ 6 mm by increasing the
confining pressure from s0 3 ¼ 10 kPa to s0 3 ¼ 20 kPa.
At a given confining pressure, the magnitude of SV in the rein-
forced assembly is much less than that of the unreinforced spec-
imen [Fig. 6(b)]. In fact the experimental and numerical results
confirm that the geocell has the highest impact on the specimens at
lower confining pressures (s0 3  15 kPa). They also verify that at a
given load cycle the magnitude of SV decreases by increasing s0 3 (i.e.
at the load cycle N ¼ 10,000, vertical settlement in the reinforced
specimen was about SV ¼ 12.5 mm and it was 6.0 mm when
s0 3 ¼ 10 kPa and 20 kPa, respectively). A marginal reduction in axial
settlement is observed in the reinforced specimen by increasing
confining pressure of about s0 3  20 kPa, where similar perfor-
mance is observed in the unreinforced and reinforced specimens.

4.2. Lateral spreading

Fig. 7 shows the variations of lateral spreading (SL) with the

depth of unreinforced subballast under a confining pressure of
s0 3 ¼ 10 kPa and a frequency of f ¼ 10 Hz at different load cycles,
that was obtained from the FE model and measured experimen-
tally. At a given number of load cycles the lateral spreading of
subballast increases with depth until it reaches a maximum
displacement at a depth of about h ¼ 250e300 mm, and then SL
decreases at a lower depth (h < 250 mm). A similar trend also
occurred at other load cycles, where the subballast spread slightly
at the base of the assembly (h ¼ 450 mm).
The beneficial effect of geocell can be highlighted in terms of
minimizing lateral spreading SL, as shown in Fig. 8(a) and (b).
Almost all lateral displacement within the geocell has been arrested Fig. 8. Typical lateral deformation profile of reinforced geocell-reinforced subballast
(a) predicted by FEM and (b) at different depth and number of cycles.
[Fig. 8(a)]. This figure also shows that the degree of SL is markedly
minimized for the subballast beneath the reinforced layer, indi-
cating that the geocell can effectively improve the performance of layer also increases. The value of SL reaches a maximum at a
the reinforced soil and the soil beneath this layer. depth of approximately subballast height of h ¼ 200e250 mm,
Fig. 8(b) presents the variations of lateral spreading (SL) with the where it is much lower than the unreinforced specimen. This figure
depth of reinforced subballast under a confining pressure of shows that at a given number of load cycles there is almost no
s0 3 ¼ 10 kPa at different numbers of load cycles. As expected, the lateral spreading of subballast within the geocell due to the addi-
magnitude of lateral spreading increases as the number of cycles tional confinement provided by the geocell.
(N) increases and lateral spreading beneath the geocell-reinforced The beneficial effects of the geocell mattress can best be inter-
preted by comparing the lateral displacement (SL) of unreinforced
and reinforced subballast obtained from experimental and nu-
ε merical modelling under different confining pressures
(s0 3 ¼ 10e20 kPa) at f ¼ 10 Hz [Fig. 9(a) and (b)]. A good agreement
is observed between the laboratory and numerical results. Under a
low confining pressure (i.e. s0 3  15 kPa), unreinforced subballast
Subballast experiences considerable lateral deformation, while the magnitude
of SL decreases with an increase in s0 3 [Fig. 9(a)]. In general, rein-
forced assembly exhibits less lateral displacement than the unre-
inforced assembly at a given confining pressure and load cycle,
while the specimen reinforced with the geocell performs better
σ′3 under a low confining pressure (i.e. s0 3  15 kPa). Marginal
improvement is observed in the reinforced specimen at a higher
confining pressure of s0 3  20 kPa [Fig. 9(b)].
Based on the results obtained from the tests and numerical
modelling, settlement (Rs) and lateral (RL) deformation reduction
factors for different confining pressures were introduced and pro-
vided in Table 3. The beneficial effect of the geocell to reduce the
subballast deformation is clearly reflected by Rs and RL. Indeed the
maximum beneficial effect of geocell reinforcement can be seen at a
Fig. 7. Variation of lateral displacements with depth of unreinforced subballast. low confining pressure which is often observed in the field. At a low
496 M.M. Biabani et al. / Geotextiles and Geomembranes 44 (2016) 489e503

4.3. Distribution of stress in the subballast

σ′ The deviator stress contours (q) of subballast reinforced with

σ′3 3
σ′3 geocell under a confining pressure of s0 3 ¼ 10 kPa and a frequency
of f ¼ 10 Hz during loading and unloading process (smax ¼ 166 kPa
and smin ¼ 41 kPa) at a load cycle of N ¼ 10,000 are presented in
Fig. 10(a) and (b). During the loading stage stress concentration
occurs inside the geocell [i.e. point C in Fig. 10(a)], because the
subballast is confined inside these pockets and prevented from
displacement. The deviator stress is much less in the middle of the
geocell mattress (point D-in Fig. 10(a)) compared to the edges.
Also Fig. 10(a) highlights the impact of the testing condition (i.e.
plane strain) during the simulation. Under loading, the granular
material can spread laterally in the transverse direction. However,
in the direction parallel to the track (longitudinal direction),
(points A & B), the lateral displacement is prevented (ε2 ¼ 0), with
a corresponding intermediate stress of s0 2 ¼ 120e140 kPa, which
is markedly greater than s0 3 (25e45 kPa). This can be attributed to
the impact of intermediate stress (s0 2 ) and boundary condition in
plane strain condition (ε2 ¼ 0). However, for the remainder of the
specimen, the geocell mattress successfully captures the intensity
(a)

σ′3
σ′3
σ′3

(b)
Fig. 9. Lateral displacement of (a) unreinforced and (b) reinforced subballast again
number of cycles (N): laboratory measurements vs. model predictions.

confining pressure (s0 3  10 kPa), geocell can reduce the defor-

mation of subballast (approximately 25% and 35% reduction for
settlement and lateral displacement, respectively), and this value
significantly decreases as the confining pressure increases
(approximately only 10% and 16% reduction for settlement and Fig. 10. Stress distribution during (a) loading and (b) unloading of unit cell in a geocell-
lateral displacement, respectively, for s0 3 ¼ 30 kPa). reinforced subballast.

Table 3
A summary of RL and RS obtained from the numerical results.

Factor f (Hz) Confining pressure, s0 3 (kPa)

5 10 15 20 30

Settlement reduction factor, Rs (%) 10 31.25 25 14.55 12.67 10.5

SV ðunreinforcedÞ SVðreinf orcedÞ
RS ¼ SV ðunreinf orcedÞ  100
Lateral displacement reduction factor, RL (%) 10 34.75 30.32 31.91 22.5 16.25
SLðunreinf orcedÞ SLðreinforcedÞ
RL ¼ SLðunreinf orcedÞ  100
 M.M. Biabani et al. / Geotextiles and Geomembranes 44 (2016) 489e503 497

of applied cyclic loading and less stress was transferred to the and lateral spreading (SL) in a reinforced specimen with subballast
lower layer of soil, which again confirms the effectiveness of a of different compressive strengths (4 MPa  Esubballast  40 MPa)
geocell mattress, when it is placed beneath the footing. during the loading stage. It is observed that the maximum mobi-
One of the benefits of numerical modelling is investigating the lized tensile stress occurs on the subballast with the lowest degree
effect of intermediate stress (s0 2 ). The numerical results show that of stiffness (Esubballast ¼ 4 MPa). This is because the infill soil be-
the magnitude of stress that developed in the direction of s0 2 [side comes stiffer by transferring the cyclic load as hoop stress (mobi-
EFGH in Fig. 10(b)] is much higher than s0 3 (about lizes tensile stress) to the geocell, where lowering the stiffness of
s0 2 ¼ 60e80 kPa). This can be justified because of the boundary the infill materials enables a greater hoop stress in the geocell. As
condition which has led to the accumulation of stress in this di- expected, lateral spreading decreases significantly as the stiffness of
rection. With conventional design criteria, the degree of inter- the subballast is increased.
mediate stress is usually assumed to be equal to the minor The effect of the subballast strength and stiffness of geocell can
principal stress (s0 2 ¼ s0 3 ), but this simulation shows there is a also be evaluated by comparing the reduction factor (RL) of lateral
remarkable dissimilarity between s0 3 and s0 2 when the experi- spreading in reinforced subballast. Fig. 13(b) shows the magnitude
ment is carried out in plane strain condition, which is similar to of RL for geocell with different stiffness and subballast strength
field conditions. under s0 3 ¼ 30 kPa and at N ¼ 10,000 cycles. As shown by Fig. 13(b),
for different subballast stiffness (Esubballast), the value of RL was
4.4. Distribution of stress in the geocell mattress varied between 32  RL  20%. The maximum value of RL >30%
occurred at Esubballast  10 MPa. In addition, for the soil with rela-
The tensile strength of geocell is an important parameter, tively higher strength, RL ¼ 20% could still be achieved.
affecting the performance of geocell-reinforced subballast, where The reinforced specimen exhibited a wider range of RL that
it is usually assumed to be constant in conventional design corresponded to variations in the stiffness of geocell (Egeocell), where
practices (Indraratna et al., 2015; Leshchinsky and Ling, 2013b). the value of RL varied from 30% up to about 75%, depending on the
However, the results obtained from numerical modelling indicate stiffness of geocell used. These results indicate the effectiveness of
that during cyclic loading, the mobilized tensile stress of geocell cellular confinement, where a larger RL occurs in the geocell with a
changes considerably at the loading and unloading stages, where higher stiffness. One practical implication of this study is that the
there are maximum and minimum tensile stresses at the loading rail industry can continue to use subballast of relatively low
and unloading stages respectively [Fig. 11(a) and (b)]. Fig. 11(a) compressive strength that can be improved by geocell, while still
shows that during the loading stage, maximum tensile stress is ensuring an acceptable performance at a lower manufacturing cost.
mobilized in the geocell due to preventing the infill subballast Moreover in the absence of high-strength granular materials, the
from excessive lateral spreading. The middle of the geocell strip numerical simulations presented in this study can be used in the
(e.g. point A) in the direction parallel to the minor principal stress preliminary design of track substructure where a wide range of
(s0 3 ) has experienced the highest degree of mobilized tensile granular materials and geocell mattresses with different strengths
stress. Fig. 11(a) also shows that minimum tensile stress occurs in and stiffness can be considered.
the direction parallel to the intermediate principal stress s0 2 (e.g.
point C), where the geocell mattress is not allowed to move in this 6. Parametric study
direction (i.e. parallel to the sleepers). This study verifies that
stress over the geocell strip is distributed non-uniformly across 6.1. Strength of the subballast
the geocell where the mobilized tensile stresses in the middle
pocket (Point B) are considerably less than those in the sur- Since the supply of aggregates with high strength is limited,
rounding pockets. Compared to the loading stage, the mobilized the use of granular materials with low strength in combination
tensile stress in the geocell during the unloading stage is also non- with the geosynthetic reinforcement is inevitable. Perhaps the
uniform, although the observed magnitude is much lower, as greatest advantage of numerical analysing is that it provides an
shown in Fig. 11(b). This study shows that the improved perfor- insight into the behaviour of specimens with different properties.
mance of geocell-reinforced subballast is controlled mostly by the By taking advantage of this, a model of geocell-reinforced sub-
mobilized tensile stress of the reinforcement where the ballast with a wide range of stiffnesses
maximum mobilized tensile stress is much less than the ultimate (4 MPa  Esubballast  40 MPa) was simulated to evaluate the
tensile strength of the geocell. performance of geocell on the subballast having varying stiff-
The influence of confining pressure on the mobilized tensile nesses. A range of stiffnesses (4 MPa  Esubballast  40 MPa) was
stress in the geocell was also investigated by comparing the selected to represent soft soils to the very stiff aggregates
mobilized tensile stress at a corresponding confining pressure of commonly used as subballast materials. The results showed that
s0 3 ¼ 5e30 kPa. Fig. 12 shows that the mobilized tensile stress an improvement in the behaviour of reinforced subballast
reaches its highest magnitude at the lowest confining pressure decreased as the stiffness of subballast increased, which is in
(s0 3 ¼ 5 kPa) and decreases significantly as s0 3 increases (i.e. geocell agreement with the previous study (Biswas et al., 2013). Fig. 14(a)
mobilized a tensile stress of around 2.5 MPa and 0.5 MPa under a shows that the lateral displacement of unreinforced subballast
confining pressure of s0 3 ¼ 5 kPa and 30 kPa, respectively). This decreases as the subballast stiffness increases, while a specimen
observation confirms the fact that the beneficial effect of geocell is with very low strength experiences significant lateral displace-
mobilized under the low confining pressure often seen in actual ment (i.e. SL ¼ 7.50 mm for Esubballast ¼ 4 MPa compared to
track conditions. SL ¼ 1.50 mm for Esubballast ¼ 40 MPa). This figure also shows that
regardless of the strength of the subballast, maximum lateral
5. Practical implications spreading occurs at the depth of h ¼ 250e300 mm. Nevertheless,
in the reinforced specimen, the inclusion of geocell decreased
The influence of subballast stiffness on the behaviour of geocell- lateral spreading markedly when the subballast had relatively low
reinforced subballast was investigated by comparing the lateral stiffness (i.e. Esubballast  10 MPa) [Fig. 14(b)]. The presence of the
spreading and mobilized tensile strength of geocell. Fig. 13(a) geocell mattress makes compaction to high levels of Relative
shows the tensile stress mobilized in the geocell (Egeocell ¼ 0.3 GPa) Density (>95%) quite difficult, and therefore the current material
498 M.M. Biabani et al. / Geotextiles and Geomembranes 44 (2016) 489e503

Fig. 11. Tensile stress mobilized in geocell mattress (a) loading and (b) unloading stage subjected to cyclic loading.

stress and the associated confining pressure. Marginal improve-

ment was observed in the reinforced subballast compared to the
unreinforced ballast at a higher modulus (30 MPa  Esubballast).

6.2. Stiffness of geocell

The variety of polymeric material available, such as high-density

polyethylene (HDPE) or novel polymeric alloy (NPA), made it
necessary to investigate the influence of geocell stiffness on the
performance of a reinforced subballast assembly. This was inves-
tigated by simulating geocell with a range of stiffness that varied
from 0.3 GPa to 5 GPa to represent a variety of materials including
HDPE, NPA, and structural steel. The lateral spreading of subballast
under geocell with varying stiffness is presented in Fig. 15. As
geocell stiffness increases the lateral displacement of subballast
decreases in every case, while unreinforced subballast assembly
exhibits the highest lateral displacement. The degree of lateral
spreading of soil beneath the reinforced layer is reduced further by
increasing the stiffness of geocell, indicating the effectiveness of the
σ′ geocell on the lower soil layer. This figure also shows that geocell
with a relatively low stiffness Egeocell ¼ 0.3 GPa performs very well
Fig. 12. Mobilized tensile stress at the geocell in reinforced subballast at different compared to the geocell with a higher stiffness.
confining pressures s0 3 .

7. Model validation
with a bulk unit weight of 20.5 kN/m3 (approx. dry unit weight of
18.5 kN/m3, optimum moisture content ¼ 11%) cannot be By using the proposed analytical model (Indraratna et al.,
considered as a highly dense subballast, but its dilation angle of 9 2015), the model results compared well with different sets of
(Table 2) still provides sufficient dilation to induce geocell hoop experimental data reported elsewhere (Leshchinsky and Ling,
 M.M. Biabani et al. / Geotextiles and Geomembranes 44 (2016) 489e503 499

σ′
Subballast

(a)
(a)

Geocell+
Subballast

(b)

(b) Fig. 14. Lateral spreading of (a) unreinforced and (b) geocell-reinforced subballast at
different strengths.

Fig. 13. (a) Mobilized tensile stress at the geocell in reinforced subballast with
different stiffness and (b) reduction factor of lateral spreading for geocell-reinforced
subballast. long  356 mm width  25 mm thick) under N ¼ 50,000 cycles. A
FE model with 12,661 elements (C3D8R) and 21,038 nodes was
used for the analysis. An elasto-plastic material with non-
2013a). The model was validated numerically (Fig. 16) to associative behaviour was used to model the subballast, where
demonstrate that the current FE model is a reasonable repre- the internal friction angle and dilatancy angle of granular material
sentation of an actual specimen of subballast reinforced with were chosen as 45 and 15 , respectively. The base of the model
geocell, where the settlements and lateral displacements ob- was restricted from any displacement in order to model a concrete
tained from the model were compared with those measured foundation as conducted in the laboratory; but each side was free
experimentally. A model test with a truncated square pyramid to move to simulate field conditions. The model was then vali-
with dimensions of 1524 mm  1524 mm  546 mm was simu- dated by the experimental and numerical data presented by
lated. For the purpose of validation purpose, a hexagonal shaped Leshchinsky and Ling (2013a).
geocell mattress with a single layer 200 mm thick was developed. Fig. 17(a)e(c) present comparisons of the predicted and
Cyclic loading amplitudes of 35 kPa  q  175 kPa and measured vertical and lateral displacements of unreinforced and
70 kPa  q  350 kPa were used for the unreinforced and rein- geocell-reinforced granular material, where a reasonably good
forced models, respectively. The cyclic load was applied on the top agreement is found between the FEM predictions and the experi-
surface of the model via a rigid plate (plate size: 356 mm mental results. The unreinforced model shows a vertical
500 M.M. Biabani et al. / Geotextiles and Geomembranes 44 (2016) 489e503

ε 8. Conclusion

The behaviour of subballast reinforced with geocell subjected to

cyclic loading was investigated using a series of large-scale pris-
Geocell+ moidal triaxial apparatus and numerical modelling. The tests and
Subballast numerical simulations were carried out to mimic the actual track
conditions. The model was compared with the experimental data
and a reasonably good agreement was achieved. The numerical
results were also validated by the published data. The proposed
FEM model can be effectively used to predict the performance of
subballast reinforced with a geocell mattress. A parametric study
was also conducted to investigate the effects of subballast and
geocell stiffness on the tensile strength mobilized in geocell
mattress. The following conclusions can be drawn from this study:

1) The numerical results agree well with the experimental results,

highlighting the effectiveness of ‘mobilised additional confine-
ment’ under cyclic loading, attributed to the inclusion of geocell,
Fig. 15. Comparison of lateral spreading of unreinforced and geocell-reinforced sub- whose hoop stress related confining pressure increases with the
ballast at different geocell stiffness. number of cycles. The test results indicate that the current
three-dimensional FE modelling can successfully simulate ver-
tical and lateral deformation of geocell-reinforced subballast
displacement of about SV ¼ 120 mm, while only around SV ¼ 60 mm
under cyclic loading at varying confining pressure over time. At
was predicted for the reinforced specimen. Compared to the model
a given confining pressure, deformation in the reinforced as-
prediction presented by Leshchinsky and Ling (2013a), the current
sembly was considerably less than that in the unreinforced
model provides better predictions, where most of vertical settle-
subballast. The numerical and experimental results also
ments occur within a range of N ¼ 10,000 load cycles [Fig. 17(a)].
confirmed that at a given load cycle the deformation decreased
This figure also shows that the current FE analysis accurately cap-
as the confining pressure increased. The observed benefit of
tures the decreasing rate of settlement as the number of load cycles
geocell was maximum at the lowest confining pressure; after
is increased. By using a hexagonal shaped geocell mattress that is
which it decreased as the confining pressure increased and
similar to the actual shape used in the laboratory, the model is able
became marginal when the confining pressure was 30 kPa.
to capture more realistic mobilized tensile stress in the geocell
2) Unlike the conventional design practices which consider a uni-
strips.
form distribution of stress over the geocell, this study epito-
Lateral displacements at the top and bottom of the model test
mized that the mobilized tensile stress was distributed non-
are captured in the simulation and compared with the published
uniformly across the geocell, where the maximum mobilized
data, as shown in Fig. 17(b) and (c). In general, the current FE
tensile stress occurred in the direction parallel to the interme-
simulations match well with the experimental data, apart from the
diate principal stress during the loading stage. Also it was
initial modulus predicted for the reinforced models that is slightly
confirmed that under cyclic loading, the tensile stress varied
greater than that for the models tested by Leshchinsky and Ling
during loading and unloading cycles, and the ultimate tensile
(2013a). This can be attributed to differences in boundary condi-
stress of a geocell mattress was never reached during the
tions between the FEM simulation and laboratory observations as
loading period of this study, i.e. after 10,000 cycles.
well as the particle degradation not considered in the current
3) The cellular confinement induced by the geocell mattress was
analysis. Although there are some disparities between the FEM
effective in reducing the lateral spreading of the infilled material
simulations and laboratory data, the proposed FEM model was able
even when the granular mass confined within the geocells was
to model the advantages attained from using geocell reinforce-
not of good quality rockfill with very apparently high friction
ment, including lower deformation and greater strength and
angles (e.g. highly angular quarried aggregates with internal
stiffness.

Fig. 16. Typical FEM mesh of geometry of model and geocell mattress used for validation.
 M.M. Biabani et al. / Geotextiles and Geomembranes 44 (2016) 489e503 501

(a)

(b) (c)
Fig. 17. Model prediction compared with experiment and numerical results of (a) vertical and lateral deformation at (b) the top and (c) bottom of the embankment.

friction angles often exceeding 50 ). This implies that marginal well with a geocell inclusion. Furthermore, an increase in the
materials can be used as subballast when improved by the use of geocell stiffness reduced the deformation in the subballast,
geocells, and this is clearly beneficial in railway practice where while a stiffer geocell exhibited less lateral displacement. This
high quality aggregates may not be available locally. study would encourage practising engineers to use a subballast
4) A parametric study was carried out to evaluate the effect of of relatively low compressive strength improved with geocell,
subballast and geocell stiffness on performance of a composite while still ensuring an acceptable performance at a lower
system. It was found that in reinforced specimen lateral manufacturing cost.
displacement was reduced further when the stiffness of sub- 5) Maximum vertical displacement (SV) occurred directly under
ballast had lower stiffness (i.e. Esubballast  10 MPa), indicating the footing where the cyclic loading was at its maximum. The
that subballast with low compressive strength could perform intensity of SV decreased at a lower depth. The numerical results
502 M.M. Biabani et al. / Geotextiles and Geomembranes 44 (2016) 489e503

also showed that maximum lateral displacements in the unre- ѱ dilation angle (degrees)
inforced specimen occurred at a depth of about
h ¼ 250e300 mm, and then SL decreases at a lower depth Appendix. Calculation of applied stress on the subballast
(h < 250 mm). As a result, utilizing geocell in the subballast layer
will reduce excessive axial and lateral deformation. For calculating the maximum contact pressure on the top of
subballast surface, a nominal axle load of 294 kN (30 tons) was
Acknowledgements assumed, and this is equivalent to a vertical pressure of 147 kN. The
design wheel load can then be calculated using (Li and Selig, 1998):
The financial support received from the Cooperative Research
Centre (CRC) for Rail Innovation to conduct this research is grate- Pd ¼ f,Ps (1)
fully appreciated. The authors are grateful for Alan Grant, Labora-
tory Manager and Ritchie McLean, Technical Officer at GRE for their where, Ps is the static wheel load (kN), Pd is the design wheel load
assistance during laboratory testing. (kN), and 4 is the impact factor (dimensionless) given as
(Indraratna et al., 2011a,b):
List of symbols
f ¼ 0:0052V=D þ1 (2)
W
b distance between two adjacent sleepers(mm)
Cc coefficient of curvature By substituting train speed (v) as 73 km/h (corresponding to
Cr apparent cohesion (kPa) f ¼ 10 Hz), and wheel diameter Dw as 0.97 m, the design wheel load
Cu uniformity coefficient Pd was determined to be about 200e208 kN. Considering 50% of
D10 diameter for 10% fine by weight (mm) pressure transmitted to the adjacent sleepers (varies from 50 to 60%
D20 diameter for 20% fine by weight (mm) as shown by Atalar et al., 2001), the rail seat load (qr) can be ob-
D30 diameter for 30% fine by weight (mm) tained about 100e104 kN. Assuming a uniform distribution of
D50 average particle size (mm) stress, the contact pressure at the ballastesleeper interface (Pa) can
D60 diameter for 60% fine by weight (mm) then be computed as (Jeffs and Tew, 1991):
Dmax maximum particle size (mm) qr
Dmin minimum particle size (mm) Pa ¼ F (3)
BL 2
DR relative density (%)
Dr depth of reinforcement (mm) where, F2 is a factor depending on track maintenance and sleeper
E elastic modulus (MPa) type (F2 ¼ 1), B is the width of sleeper (B ¼ 260 mm), l is the total
Esubballast subballast strength (MPa) length of sleeper (l ¼ 2400 mm) and L is the effective length of
Egeocell geocell stiffness (GPa) sleeper. By assuming the effective length of sleeper as one third of
f frequency (Hz) the total sleeper length (Jeffs and Tew, 1991), Eq. (3) becomes:
h specimen height (mm)  
d interface coefficient 3qr
Pa ¼ F (4)
ks normalized confinement ratio Bl 2
RL lateral spreading reduction factor (%)
According to Japanese Track Standards, considering L as 2d,
Rs settlement reduction factor (%)
where, d is distance between the rail head center and edge of the
L total length of sleeper (mm)
sleeper (d ¼ 500 mm), following equation can be used to obtain Pa
le effective length of sleeper supporting the load (mm)
(Atalar et al., 2001):
M geocell modulus (kN/m)
Mm mobilized geocell modulus (kN/m) qr
Pa ¼ F (5)
N number of cycles 2dB 2
qmax maximum amplitude (kPa)
Considering Eqns. (3)e(5), a maximum of Pa (i.e. Pa ¼ 492 kPa)
qmean mean load (kN)
was considered. Considering the sleeper area (L ¼ 800 mm,
qmin minimum amplitude (kPa)
B ¼ 260 mm) and a ballast depth of 300 mm, the stress on top of
SL lateral spreading (mm)
subballast can then be calculated of about 160e170 kPa using
SV vertical deformation (mm)
Boussinesq elastic theory. A minimum amplitude ðsmin Þ of 41 kPa
gd dry unit weight of the soil (kN/m3)
was selected to represent in situ unloaded track state.
d interface friction angle of subballast-geosynthetic
(degree)
ε2 lateral strain parallel to intermediate principal stress (%) References
ε3 lateral strains parallel to minor principal stress (%)
Atalar, C., Das, B.M., Shin, E.C., Kim, D.H., 2001. Settlement of geogrid-reinforced
εv volumetric strain (%) railroad bed due to cyclic load. In: Proc., 15th Int. Conf. on Soil Mechanics
εpv plastic volumetric strain (%) and Geotechnical Engineering, vol. 3. CRC Press, Boca Raton, FL, pp. 2045e2048.
yg Poisson's ratio of geocell Bathurst, R.J., Knight, M.A., 1998. Analysis of geocell reinforced-soil covers over
large span conduits. Comput. Geotech. 22 (3e4), 205e219.
s1 major principal stress (kPa) Biabani, M.M., Indraratna, B., 2015. An evaluation of the interface behaviour of rail
s0 2 intermediate principal stress (kPa) subballast stabilised with geogrids and geomembranes. Geotext. Geomembr. 43
s0 3 minor principal stress (kPa) (3), 240e249.
Ds0 3 additional confining pressure (kPa) Biswas, A., Krishna, A.M., Dash, S.K., 2013. Influence of subgrade strength on the
performance of geocell-reinforced foundation systems. Geosynth. Int. 20 (6),
scyc cyclic deviator stress (kPa) 376e388.
smax maximum stress (kPa) Bolton, M.D., 1986. The strength and dilatancy of sands. Ge otechnique 36 (1),
smean mean stress (kPa) 65e78.
Chen, C., McDowell, G.R., Thom, N.H., 2012. Discrete element modelling of cyclic
smin minimum stress (kPa) loads of geogrid-reinforced ballast under confined and unconfined conditions.
f internal friction angle of unreinforced soil (degree) Geotext. Geomembr. 35, 76e86.
 M.M. Biabani et al. / Geotextiles and Geomembranes 44 (2016) 489e503 503

Choudhury, J., 2009. Track Reconditioning Guidelines. TMC 403 engineering Mehdipour, I., Ghazavi, M., Moayed, R.Z., 2013. Numerical study on stability analysis
manual. RailCorp, Chippendale, NSW, Australia. of geocell reinforced slopes by considering the bending effect. Geotext. Geo-
Ferellec, J.-F., McDowell, G.R., 2012. Modelling of ballastegeogrid interaction using membr. 37 (0), 23e34.
the discrete-element method. Geosynth. Int. 19 (6), 470e479. Mehrjardi, G.T., Tafreshi, S.M., Dawson, A.R., 2012. Combined use of geocell rein-
Han, J., Pokharel, S.K., Yang, X., Manandhar, C., Leshchinsky, D., Halahmi, I., forcement and rubber soil mixtures to improve performance of buried pipes.
Parsons, R.L., 2011. Performance of geocell-reinforced RAP bases over weak Geotext. Geomembr. 34, 116e130.
subgrade under full-scale moving wheel loads. J. Mater. Civ. Eng. 23 (11), Moghaddas Tafreshi, S.N., Khalaj, O., Dawson, A.R., 2014. Repeated loading of soil
1525e1534. containing granulated rubber and multiple geocell layers. Geotext. Geomembr.
Han, J., Yang, X.M., Leshchinsky, D., Parsons, R.L., 2008. Behavior of geocell- 42, 25e38.
reinforced sand under a vertical load. Transp. Res. Rec. (2045), 95e101. Ngo, N.T., Indraratna, B., Rujikiatkamjorn, C., 2014. DEM simulation of the behaviour
Hegde, A., Sitharam, T.G., 2013. Experimental and numerical studies on footings of geogrid stabilised ballast fouled with coal. Comput. Geotech. 55, 224e231.
supported on geocell reinforced sand and clay beds. Int. J. Geotech. Eng. 7 (4), Peters, J.F., Lade, P.V., Bro, A., 1988. Shear band formation in triaxial and plane strain
346e354. tests. In: Donaghe, R.T., Chaney, R.C., Silver, M.L. (Eds.), Advanced Triaxial
Hegde, A., Sitharam, T.G., 2014. 3-Dimensional numerical modelling of geocell Testing of Soil and Rock. ASTM, West Conshohocken, PA, pp. 604e627.
reinforced sand beds. Geotext. Geomembr. 1e11. Pokharel, S.K., Han, J., Leshchinsky, D., Parsons, R.L., Halahmi, I., 2010. Investigation
Hegde, A.M., Sitharam, T.G., 2015. Three-dimensional numerical analysis of geocell- of factors influencing behavior of single geocell reinforced bases under static
reinforced soft clay beds by considering the actual geometry of geocell pockets. loading. Geotext. Geomembr. 28 (6), 570e578.
Can. Geotech. J. 52 (9), 1396e1407. Radampola, S.S., 2006. Evaluation and Modelling Performance of Capping Layer in
Huang, J., Bhandari, A., Yang, X., 2011. Numerical modelling of geosynthetic- Rail Track Substructure (Ph.D. thesis). Central Queensland Univ., Rockhampton,
reinforced earth structures and geosynthetic-soil interactions. Geotech. Eng. QLD, Australia.
42 (1), 42e55. Radampola, S.S., Gurung, N., McSweeney, T., Dhanasekar, M., 2008. Evaluation of the
Indraratna, B., Salim, W., Rujikiatkamjorn, C., 2011a. Advanced Rail Geotechnology properties of railway capping layer soil. Comput. Geotech. 35 (5), 719e728.
e Ballasted Track. CRC Press/Balkema. Saride, S., Puppala, A.J., Sitharam, T.G., Gowrisetti, S., 2009. Numerical simulation of
Indraratna, B., Ngo, N.T., Rujikiatkamjorn, C., 2011b. Behavior of geogrid-reinforced geocell-reinforced sand and clay. Proc. Inst. Civ. Eng. e Ground Improv. 162 (4),
ballast under various levels of fouling. Geotext. Geomembr. 29 (3), 313e322. 185e198.
Indraratna, B., Ngo, N.T., Rujikiatkamjorn, C., 2013. Studying the deformation of coal Selig, E.T., Waters, J.M., 1994. Track Geotechnology and Substructure Management.
fouled ballast stabilised with geogrid under cyclic load. J. Geotech. Geo- Thomas Telford, London.
environmental Eng. e ASCE 139 (8), 1275e1289. Suiker, A.S.J., Selig, E.T., Frenkel, R., 2005. Static and cyclic triaxial testing of ballast
Indraratna, B., Biabani, M., Nimbalkar, S., 2015. Behavior of geocell-reinforced and subballast. J. Geotech. Geoenvironmental Eng. 771e782. http://dx.doi.org/
subballast subjected to cyclic loading in plane-strain condition. J. Geotech. 10.1061/(ASCE)1090-0241(2005)131:6(771).
Geoenvironmental Eng. 141 (1), 04014081e16. Moghaddas Tafreshi, S.N., Dawson, A.R., 2012. A comparison of static and cyclic
Jeffs, T., Tew, G.P., 1991. A Review of Track Design Procedures: Sleepers and Ballast, loading responses of foundations on geocell reinforced sand. Geotext. Geo-
vol. 2. Railways of Australia BHP Research, Melbourne Laboratories, Melbourne, membr. 32, 55e68.
Australia. Tennakoon, N.C., 2012. Geotechnical Study of Engineering Behaviour of Fouled
Kwon, J., Tutumluer, E., 2009. Geogrid Base Reinforcement with Aggregate Interlock Ballast (Ph.D. thesis). University of Wollongong, Australia.
and Modelling of Associated Stiffness Enhancement in Mechanistic Pavement Wanatowski, D., Chu, J., Lo, R.S.C., 2008. Strain-softening behaviour of sand in
Analysis. Transportation Research Record 2116. Transportation Research Board, strain path testing under plane-strain conditions. Acta Geotech. 3 (2),
Washington, DC, pp. 85e95. 99e114.
Latha, G.M., Rajagopal, K., 2007. Parametric finite element analyses of geocell- Wang, G.Y., Zhang, J.P., Zhao, J.W., 2013. Numerical analysis of geocell protective
supported embankments. Can. Geotech. J. 44 (8), 917. slope stability. Applied Mechanics and Materials 353, 635e639.
Latha, G.M., Somwanshi, A., 2009. Effect of reinforcement form on the bearing ca- Yang, X., Han, J., Parsons, R.L., Leshchinsky, D., 2010. Three-dimensional numerical
pacity of square footings on sand. Geotext. Geomembr. 27 (6), 409e422. modelling of single geocell-reinforced sand. Front. Archit. Civ. Eng. China 4 (2),
Leshchinsky, B., Ling, H., 2013a. Effects of geocell confinement on strength and 233e240.
deformation behavior of gravel. J. Geotech. Geoenvironmental Eng. 139 (2), Yang, X., Han, J., Leshchinsky, D., Parsons, R., 2013. A three-dimensional
340e352. mechanistic-empirical model for geocell-reinforced unpaved roads. Acta Geo-
Leshchinsky, B., Ling, H.I., 2013b. Numerical modelling of behavior of railway bal- tech. 8 (2), 201e213.
lasted structure with geocell confinement. Geotext. Geomembr. 36 (0), 33e43. Yu, H.S., Sloan, S.W., 1997. Finite element limit analysis of reinforced soils. Comput.
Li, D., Selig, E.T., 1998. Method for railroad track foundation design, I: development. Struct. 63 (3), 567e577.
J. Geotech. Geoenvironmental Eng. e ASCE 124 (4), 316e322.