Geotextiles and Geomembranes 28 (2010) 33–43

Contents lists available at ScienceDirect

Geotextiles and Geomembranes

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

Investigation of geotextile–soil interaction under a cyclic vertical load

using the discrete element method
Anil Bhandari, Jie Han*
Civil, Environmental, and Architectural Engineering (CEAE) Department, The University of Kansas, Lawrence, KS 66045, USA

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

Article history: Geotextiles are often used in roadway construction as separation, filtration, and reinforcement. Their
Received 21 August 2008 performance as reinforcement in geotextile-reinforced bases depends on geotextile–soil interaction. This
Received in revised form paper investigates the geotextile–soil interaction under a cyclic wheel load using the Discrete Element
11 August 2009
Method (DEM). In this study, soil was modeled as unbonded particles using the linear contact stiffness
Accepted 21 September 2009
model, and the geotextile was modeled as bonded particles. The micro-parameters of the soil and the
Available online 30 October 2009
geotextile were determined using biaxial tests and a tensile test, respectively. The influence of the
placement depth and the stiffness of the geotextile on the performance of the reinforced base was
Keywords:
Geotextile investigated. The DEM results show that the depth of the geotextile significantly affected the degree of
Interaction interaction between the geotextile and the soil. Under the applied cyclic vertical load, the geotextile
Cyclic wheel load developed a low tensile strain. The effect of the stiffness of the geotextile on the deformation was more
Discrete Element Method (DEM) significant when the geotextile was placed at a shallower location than when placed at a deeper location.
Stiffness Ó 2009 Elsevier Ltd. All rights reserved.

1. Introduction The effect of the location and number of geosynthetic layers on

Geosynthetics have become more commonly used for subgrade
improvement by placing the reinforcement at the interface
between base and subgrade or base reinforcement by placing the
reinforcement within the base course for roadway applications,
such as unpaved and paved roads and railroads. The benefit of the
geosynthetic in the improvement of roadways depends on the
interaction of the geosynthetic and the soil. The performance of
a geogrid-reinforced base is governed by thickness, stiffness, and
shape of rib; size, shape, and rigidity of aperture; and junction
strength, and secant modulus, and stability of a geogrid (Webster,
1992). However, the geosynthetic with superior tensile strength
does not necessarily ensure the improved performance. For
example, plate load tests conducted on lightweight shale aggre-
gates by DeMerchant et al. (2002) showed that the performance of
the reinforced aggregate depended on the stiffness of the geogrid
and the load level. Similar findings were reported by Han et al.
(2008a,b) using a moving wheel load test. These researchers
highlighted the importance of the interaction between the geo-
synthetic and the soil on the performance of a geosynthetic-
reinforced soil.
* Corresponding author. Tel.: þ1 7858643714; fax: þ1 7858645631.
E-mail address: jiehan@ku.edu (J. Han).
the performance of the reinforced section has also been investi-
gated. By summarizing the findings of previous experimental
research, Perkins and Ismeik (1997) concluded that the proper
location of the geosynthetic within the base course depended on
the magnitude of the applied load and the strength of the subgrade
layer. A paved road subjected to a heavier traffic load would
perform better if the geosynthetic reinforcement was placed at
a greater depth. On the other hand, the paved road subjected to
a lighter traffic load would perform better if the geosynthetic
reinforcement was placed at a shallower depth. While numerical
simulations of all the studies reported by Perkins and Ismeik (1997)
would be a daunting task, simple models with changing the depth
of the geosynthetic are helpful in understanding the influence of
the geosynthetic placement on the performance of the reinforced
section.
The stiffness of the geosynthetic is an important mechanical
parameter that governs the strain in the geosynthetic under a given
loading condition. It is expected that a stiff geosynthetic would
develop a lesser strain than a flexible geosynthetic. The full-scale
field tests performed by Hufenus et al. (2006) showed that the
geosynthetic underwent a low strain of 2% and developed a tensile
force up to 8–15 kN/m. In another study, Miura et al. (1990) found
that the maximum tensile strain of the geogrid was 0.3% under
a plate load. These investigations suggest that the strain in the
geosynthetic in roadways is typically small (mostly less than 2%),

0266-1144/$ – see front matter Ó 2009 Elsevier Ltd. All rights reserved.
doi:10.1016/j.geotexmem.2009.09.005
34 A. Bhandari, J. Han / Geotextiles and Geomembranes 28 (2010) 33–43
however, it is not well understood how the stiffness of the geo-
synthetic affects the interaction between the geosynthetic and the
soil.
Numerical methods, both continuum and discontinuum
approaches, have been used to evaluate the geosynthetic–soil
interaction. In the continuum approach, material properties are
averaged over a finite integral volume and the domain is discretized
into finite zones to evaluate the nodal strains and stresses (Munjiza,
2004). Continuum-based finite element or finite difference
methods are beneficial for studying the overall behavior of the
system; however, they do not provide insight into micro-scale
responses. Several constitutive models have been used to replicate
the response of the system. For example, Saad et al. (2006) used
three constitutive models for four materials in geosynthetic-rein-
forced pavements, namely, a linear elastic model for the asphalt
concrete and the geosynthetic reinforcement, the Drucker–Prager
model for the granular base, and the Cam Clay model for the
subgrade to investigate the effects of the geosynthetic reinforce-
ment on the fatigue and rutting of the pavements. In some models,
input parameters may not have any physical meaning. In addition,
the continuum approach does not account for the effects of particle
angularity and particle rolling and sliding on the performance of
the geosynthetic-reinforced structures. The discontinuum
approach handles the problem at micro-scale; a force-displace-
ment law is applied at individual grains, and very few parameters
are required to replicate the response. All parameters have physical
meaning. Hence, the Discrete Element Method (DEM), based on the
discontinuum approach, is suitable for solving geosynthetic–soil
interaction problems since it addresses major drawbacks of the
finite element or finite difference approach discussed above.
This paper presents a DEM analysis to investigate the interaction
of the geotextile and the soil and its effect on the performance of
the geotextile-reinforced base subjected to a cyclic vertical load.
When a geotextile is placed within a base course for base rein-
forcement, installation damage to the geotextile in the field should
be carefully examined and considered. This study was conducted to
demonstrate how the DEM modeling can be used to analyze the
geotextile–soil interaction under a cyclic vertical load. The same
approach can be used to simulate geogrid-reinforced bases, which
is currently under development.
2. Discrete Element Method (DEM) for geosynthetic–soil
interaction
The DEM is an ideal tool for studying the interaction between
geogrid and aggregate (Konietzky et al., 2004). Most DEM studies
have so far focused on the geogrid–soil interaction in large triaxial
tests, pullout tests, and plate load tests. The following review on the
geogrid–soil interaction analyzed by DEM is equally relevant to the
geotextile–soil interaction.
2.1. Large triaxial test
Konietzky et al. (2004) modeled the triaxial tests on geogrid-
reinforced samples in three dimensions using spheres to study the
confinement zone of geogrids. The back analysis of triaxial tests
showed that the increased strength of the reinforced sample cannot
be attributed only to the tensile strength of geogrids. The additional
increased strength must have been derived from the confinement
effect of the geogrids. For aggregates with particle sizes between
0.6 and 20 mm, a punched-drawn biaxial geogrid had a confine-
ment influence zone of 100 mm on either side of the geogrid. They
noted that beyond this influence zone, the contact forces in
aggregates did not change significantly. Even though this study
captured the force-displacement behavior in the triaxial tests and
demonstrated interlocking effects of the geogrid with aggregates,
this study did not consider the effect of the aggregate shape on the
performance of the reinforced sample.
The subsequent study by McDowell et al. (2006) showed the
importance of simulating aggregate shapes on the peak strength
and dilation behavior of railway ballasts. McDowell et al. (2006)
modeled a 300 mm (diameter)  600 mm (height) triaxial sample
using the clumps of overlapping spheres to represent the actual
shape of aggregates. This study showed that the ability of the
clumps to capture the peak strength and dilation behavior of the
ballast is superior to the spheres. In addition, McDowell et al.
(2006) investigated the effects of multiple geogrids (i.e., three
layers) on the displacement of the sample as compared with
a single layer of geogrid reinforcement. The numerical results
showed that the sample with a single geogrid had 50% more axial
and radial displacements than those with three geogrid layers.
2.2. Pullout test
Pullout test is a common experimental method for evaluating
the geosynthetic–soil interaction. This test was simulated by
Chareyre and Villard (2002). Before this simulation, the micro-
mechanical properties of the materials were calibrated based on
the simulation of two-dimensional biaxial compression tests of soil
and a geogrid pullout test in a horizontal plane for interface char-
acterization. Using the calibrated parameters, Chareyre and Villard
(2002) simulated the force-displacement curves for the geogrid
anchored in sand and silt. Large fluctuations in the pullout force
were observed when the geogrid was anchored in sand. However,
the force-displacement curve was smooth with a constant period-
icity in force fluctuations when the geogrid was anchored in silty
sand. The anchorage failure mechanism was qualitatively captured
in both cases. Villard and Chareyre (2004) conducted similar
pullout tests in cohesive and cohesionless soils but using L- or
V-shaped anchorage of the geotextile in their laboratory models.
The interlocking effect of a geogrid helps mobilize the intrinsic
capacity of aggregates, which depends upon the aperture size of the
geogrid and the particle size of aggregates. Using a 3D DEM model
to simulate laboratory pullout tests, McDowell et al. (2006)
demonstrated that the ratio of aperture size to particle diameter
played an important role in the peak strength and the minimum
displacement required to mobilize this peak strength. For aggre-
gates of 40 mm in diameter, the optimum interlock effect was
observed at the ratio of 1.4. Recent cyclic plate loading tests con-
ducted by Brown et al. (2007) on geogrid-reinforced ballast
confirmed this ratio. They observed the least settlement of the
geogrid-reinforced ballast at the aggregate diameter of 50 mm and
the grid opening between 60 and 80 mm. In the numerical study of
McDowell et al. (2006), the distribution of the average shear force
recorded inside the model showed that the interlocking effect of
the geogrid was confined to a relatively narrow width (i.e.,
approximately 20 cm). The ratio of the peak shear and normal
forces at the end of the test to their initial values reached 10. Fig. 1
clearly demonstrates the confinement effect of the geogrid on
aggregates after the pullout test. This conclusion is consistent with
that obtained by Konietzky et al. (2004).
2.3. Plate load test
A plate load test is commonly used to evaluate bearing capacity
and stiffness of foundations. This test can also be used to evaluate
the benefits of placing geosynthetics in the foundations (Douglas
and Valsangkar, 1992; Kelly et al., 1995; Raymond, 2002; Tingle and
Jersey, 2005; Han et al., 2008a). The DEM analysis was conducted to
investigate particle movement and contact forces between the
 A. Bhandari, J. Han / Geotextiles and Geomembranes 28 (2010) 33–43 35

0. 35 Rutting tests were conducted on the bases with and without

Before pull-out a geotextile sheet at depths of 25.0 and 12.5 mm below the surface.
0. 30
During pull-out A load of 88 N was applied on a wheel moving on a pressured hose
0. 25 with a diameter of 25.0 mm for 8000 cycles (one cycle equals to two
Zerosurcharge
passes). A pressured hose is a rubber pipe with air pressure inside,
0. 20
13.2 kPa surcharge which is used to apply pressure on the base to simulate the effect of
0. 15 35.4 kPa surcharge tire pressure. Rut depths were measured at different cycles. Fig. 2
0. 10 shows the plot of rut depth versus number of cycles. It is clearly
shown that the rut depth increased significantly at the initial stage
Depth (m)

0. 05 of tests while it slowly increased beyond 2000 cycles. Fig. 2 shows

0. 00 the benefit of the geotextile in the reduction of the rut depth.
Details of these tests can be found in the paper by Han et al.
-0.0 5 (2008b).
-0.1 0 The above experiment evaluated the rut depth of geosynthetic-
reinforced and unreinforced bases, in which the interaction
-0 .1 5
Geogrid Confinement between geosynthetic and sand is three-dimensional in nature. The
-0 .2 0 numerical modeling discussed in the succeeding sections, however,
location zone
-0 .2 5 is two-dimensional. Therefore, the purpose of this study was not to
match the experimental results, instead, it was to qualitatively
-0 .3 0 capture the mechanisms involved in this kind of test and explain
-0 .3 5 the phenomena from the DEM analysis. The importance of the
0 0.4 0 .8 1.2 1 .6 depth and the stiffness of the geotextile on the performance of the
Average shear f orce (N) reinforced section under a cyclic wheel load were also studied.

Fig. 1. Average shear force before and at end of test (modified from McDowell et al.,
2006).
4. Numerical modeling

interacting particles under plate loading (Lim and McDowell, 2005; Particle Flow Code (PFC2D) used in this study was developed by
Lu and McDowell, 2007). Itasca based on the discrete element model (Cundall and Strack,
Lu and McDowell (2007) compared contact forces and particle 1979). PFC2D uses rigid body and soft contact approaches. The soft
movement using two particle shapes: spheres and angular clumps contact approach allows deformation calculations at the contacts.
of spheres. The current version of Particle Flow Code (PFC3D) PFC2D utilizes two successive cycles to compute the forces and the
allows the simulation of particle shape using clumps of bonded displacements of the particles. The motion of each particle is
spheres which is different from using polyhedral or elliptical sha- calculated from resultant contact and body forces acting on the
ped particles. Shear strength of the granular plane assemblies could particle using Newton’s 2nd law of motion. The contact and body
be increased by approximately 60% compared to the granular plane forces are then updated for the resulted motion by applying the
assemblies of circular particles with the introduction of eccentricity force-displacement law (Itasca, 2004). The model preparation and
on particles (Rothenburg and Bathurst, 1993). The increased shear the determination of the micro-mechanical parameters for the
strength of the eccentric particles was mainly due to their ability to model used in this study are discussed below.
form higher coordination number (Rothenburg and Bathurst, 1992,
1993). However, Lu and McDowell (2007) used angular clumps and
observed an increased contact force between particles and an
improved rolling resistance that provided more interlocking among
particles. Furthermore, the use of properly shaped aggregates
closely modeled the force-deformation response of ballast particles
in a box plate load test. When smooth spheres were used, the
particles continued rolling even during unloading. On the other
hand, when the aggregates were simulated with angular clumps,
the particles rolled in the opposite direction during unloading to
that during loading.

3. Experimental background

The DEM analysis in this study was based on the experimental

work done by Han et al. (2008b). In their tests, the Asphalt Pave-
ment Analyzer (APA) machine, originally used to test rutting and
fatigue behavior of hot mix asphalt samples in cylinders or beams
in test molds, was modified to evaluate the geosynthetic–soil
interaction. The original test mold of the APA machine was
modified into an aluminum box at dimensions of
0.38 m  0.45 m  0.10 m to enable the placement of a geosynthetic
sheet in a base layer. Poorly graded, subrounded river sand with
a mean grain size of 2.6 mm was used in the laboratory tests. The
bases were prepared in four layers at 70.0% relative density. Fig. 2. Rut depth versus number of cycles (Han et al., 2008b).
36 A. Bhandari, J. Han / Geotextiles and Geomembranes 28 (2010) 33–43

4.1. Determination of micro-mechanical parameters

The microscopic properties of materials are usually calibrated or

determined using known macroscopic responses. Numerical biaxial
tests and shear tests are commonly used to evaluate the micro-
scopic properties of geo-materials (for example, Chareyre and Vil-
lard, 2002; Zhang and Thornton, 2007). In this study, biaxial tests
were used to determine the micro-parameters of sand to have
reasonable mechanical properties. A tensile test was used to cali-
brate the micro-parameters of the geotextile. The stiffness and
other parameters of the contact bond strength of the cylindrical
particles were calculated using the following relationships (Itasca,
2004; Potyondy and Cundall, 2004):

kn ¼ 2tEc
fn ¼ 2t sc R (1)
fs ¼ 2t sc R
where kn is the normal stiffness of the particles, t is the thickness of
particles along the plane of paper, Ec is Young’s modulus of the
particle–particle contact, fn is the parameter of the contact bond
normal strength (sc) between particles, R is the radius of the Fig. 3. Deviatoric stress versus axial strain from PFC2D biaxial test simulations.
particle, fs is the parameter of the contact bond shear strength (sc)
between particles. fn and fs are expressed in the unit of force.
in the DEM model. The mechanical strength of the geotextile was
4.1.1. Sand characterized by its tensile strength. The contact bond force
The biaxial sample, used to determine the micro-mechanical between the geotextile particles corresponds to its tensile strength.
parameters of sand, had a width of 0.15 m and height of 0.3 m. The Similarly, the normal stiffness between the geotextile particles
uniform-sized cylindrical particles of 4.0 mm in diameter were corresponds to its tensile stiffness at a strain. The microscopic
used for this determination using the biaxial test. The medium parameters for the geotextile were calibrated using a tensile test.
dense particle assembly was created by setting porosity (n) of 0.16. Only a portion of the particles generated for the tensile test is
The input parameters presented in Table 1 were used to perform shown in Fig. 6 due to space limitation. Seventy-five particles were
the biaxial test as outlined in PFC2D manual (Itasca, 2004). generated to cover a span (L) of 75 mm without any overlap
The deviatoric stress (s1  s2) versus axial strain and volumetric between the particles. The input parameters are shown in Table 1.
strain versus axial strain are shown in Figs. 3 and 4. These plots The starting and ending particles in the row were given a small
were obtained at three confining pressures: 5, 10, and 15 kPa. Low velocity (V) of 1E  8 m/step and the resulted forces were
confining stresses were selected to mimic typical stress levels in measured. The DEM results were compared with the strength of the
a base course under its overburden stress. The q–p0 plot of the test is geotextile as mentioned on the product specification (HP370-Ten-
shown in Fig. 5, and the friction angle of the assembly was calcu- cate Mirafi) by the manufacturer. Fig. 7 shows a reasonable agree-
lated using Equation (2). ment between the DEM and the experimental results. It is worth
From the q–p0 plot, the friction angle of the sand is calculated as: mentioning that the stiffness of a viscoelastic material such as
geotextile depends on the strain rate (Walters et al., 2002).
q
¼ sin f ¼ 0:534
p0 (2)

f ¼ 32:2 :
The obtained friction angle of the assembly (f ¼ 32.2 ) was much
smaller than the friction angle of the particle/particle contact
(m ¼ tan fc; fc ¼ 40.4 ). The selected m value resulted in a friction angle
of the assembly typical to sand used in practice; therefore, it is satis-
factory. The differences between the friction angle of the assembly
and the contact friction angle of particles have been observed in
experimental and numerical simulations of biaxial compression tests
(Bathurst and Rothenburg, 1990; Masson and Martinez, 2000).

4.1.2. Geotextile
A woven geotextile was used in the laboratory test. The geo-
textile was modeled using bonded particles of 1.0 mm in diameter
Table 1
Micro-mechanical properties for DEM analysis.

Material Ec (N/m2) kn/ks sc*2R (kN/m) sc*2R (kN/m) m

Soil 1.5E þ 8 2 – – 0.85
Geotextile 3.5E þ 8 1 47.3 47.3 0.68

Ec ¼ Young’s modulus of the particle, kn ¼ normal stiffness, ks ¼ shear stiffness, sc is

the contact bond normal strength, sc is the contact bond shear strength between
particles, and m is the friction angle of particle/particle contact. Fig. 4. Volumetric strain versus axial strain from PFC2D biaxial test simulations.
 A. Bhandari, J. Han / Geotextiles and Geomembranes 28 (2010) 33–43
Fig. 5. q versus p0 plot.
However, the current model used an elastic spring at the contact of
two particles (contact-bond model) and the viscoelastic behavior of
the geotextile was not captured.
4.2. Numerical modeling of unreinforced and reinforced bases
For the numerical simulation, a box was created with the same
dimensions in length and height (0.38 m  0.10 m) as those used in
the laboratory tests (Fig. 8). The box was enclosed by four walls.
First, the box was divided into two compartments to ease the
placement of the geotextile. The dimensions of the compartments
were chosen so as to mimic the location of the geotextile in the
laboratory tests. Cylindrical particles (0.38 m long along the plane
of paper) were generated simultaneously in both compartments.
The radius expansion technique, as described in Itasca (2004), was
chosen to obtain the required 2-D porosity (n ¼ 0.16). The selected
porosity was based on the previous research conducted by Bhan-
dari et al. (2008), in which the authors noticed a qualitative
agreement between the laboratory and numerical results of the
moving wheel load tests at a porosity of 0.16.
The geotextile layer was placed by generating the particles
inside the guided walls (Fig. 8). After the generation of the geo-
textile layer, the walls were deleted and a hose was created at the
top of the assembly by particles with the same size and properties
as the geotextile. The particles were then subjected to the gravity
loading, and the inter-particle friction was introduced at this stage.
The assembly was cycled to meet the equilibrium of forces. It
should be noted that the radius expansion technique generates
large forces on the wall, and hence the forces should be brought to
an initial at-rest condition. Lateral walls were moved in opposite
directions using the numerical servo mechanism until the desired
lateral wall force (equal to the lateral earth thrust at the earth
Fig. 6. Particle assembly for a tensile test simulation.
37
Fig. 7. Calibration of the geotextiles using a tensile strength test.
pressure coefficient of 1.0) was attained. The vertical and horizontal
stresses were monitored at five locations across the horizontal
plane at the mid-depth of the base (50 mm). The average coefficient
of lateral earth pressure across the horizontal plane for different
models was close to 1.0 and the difference was less than 0.3;
therefore, there was some variation of the normal contact level in
the horizontal direction. This result was the consequence of
controlling the porosity of the assembly rather than the stress level
within the sample during the numerical sample preparation.
However, the wall movement required to achieve the target wall
forces was very small (approximately 0.2 mm) and had a marginal
effect on the porosity of the assembly. The displacements of all
particles were set to zero at this stage so that the vertical defor-
mation and the particle movement measurement would not be
affected by the displacements during the sample preparation. The
particles attained a low velocity (max. velocity ¼ 1E  5 m/s) and
the velocity vectors were randomly oriented ensuring the equilib-
rium of the assembly. The time step for the analysis was 8.3E  7,
which means the system requiring 8.3E7 computational cycles for
an additional displacement of 0.01 mm (Itasca, 2004). Obviously,
such a small velocity of the particles would have a negligible effect
on the numerical results.
On top of the hose, a wheel (diameter ¼ 40 mm) was created at
the mid-length to apply a cyclic load of 353 N in the DEM model.
The cyclic load was applied vertically and no horizontal movement
was allowed. The contact normal stiffness value and the shear
stiffness value of this wheel were the same as those of the particles.
The dead load of the wheel was 36.5 N (note that the cylindrical
particles were 0.38 m long along the plane of paper, and the iron
wheel (r ¼ 7800 kg/m3) was considered). Hence in the cyclic model
test, the following loadings were applied:
Condition Load (N)
Loading 389.5
Unloading 36.5
A 353 N load was applied to the wheel on the assembly. The
assembly was solved until the ratio of the maximum unbalanced
force to the maximum contact force reached the value of 0.01
(default in the PFC2D). The load was removed from the wheel and
the assembly was again solved in the similar way. This loading
and unloading process was repeated for 25 cycles (i.e., 25 loading
38 A. Bhandari, J. Han / Geotextiles and Geomembranes 28 (2010) 33–43
Fig. 8. DEM model of APA test simulation.
and 25 unloading). The number of cycles was chosen after some
initial trials that would achieve a constant deformation.
On each loading and unloading of the wheel, the vertical
deformation was calculated. To evaluate the vertical deformation,
the new position of the wheel was monitored. Four cases were
considered for the numerical simulation. In the first reference case,
the geotextile was not placed and the model was run without any
reinforcement. In the second case, the geotextile was placed at
a depth of 12.5 mm below the top surface. In the third case, the
geotextile was placed at a depth of 25.0 mm below the top surface.
The particles of small diameter used to model the geotextile creates
a slippage plane (Bhandari et al., 2008), therefore, their effect was
studied using a sheet of tiny particles without any bonding strength
at 25 mm below the top surface as the fourth case. The above four
cases are named as Case-I, Case-II, Case-III, and Case-IV, respec-
tively and will be referred accordingly in the paper. The first three
cases qualitatively represent the laboratory tests. The stiffness of
the geotextile was changed in a parametric study to investigate its
effect on the geotextile–soil interaction. Whenever a geotextile was
used, the principal stresses of the particles that represented the
geotextile were calculated. For all numerical models, the same
procedure was followed to prepare the samples so that the results
can be compared without any adverse effect of model preparation.
As random generation of particles was used to create the numerical
models, the uniqueness of the results can be questionable. The
uniqueness of the results was verified by creating another random
assembly with same material properties, boundary and loading
conditions to achieve the same numerical results.
Fig. 9. Deformation versus number of cycles at particle porosity of n ¼ 0.16 (medium dense sand).
5. Analysis of results
5.1. Vertical permanent deformation
The vertical permanent deformation versus the number of
cycles is shown in Fig. 9. The model with a sheet of tiny particles
(Case-IV) had the maximum deformation followed by those with
the geotextile at a depth of 25.0 mm (Case-III), without a geotextile
(Case-I), and with the geotextile at a depth of 12.5 mm (Case-II). In
the experimental study (Han et al., 2008b), the geotextile placed at
a depth of 25.0 mm (Case-III) resulted in less deformation than the
unreinforced case (Case-I) while in the numerical study, the geo-
textile placed at a depth of 25.0 mm had higher deformation than
the unreinforced case (Case-I). The reason for this difference may
be attributed to different particle shapes and size distributions.
Angular particles with different sizes in the experiment tend to
interact with the geotextile better to minimize lateral movement of
particles.
Fig. 9 shows non-smooth curves with steps in the two cases
with the geotextile placed at a depth of 25.0 mm (Case-III) and
without a geotextile (Case-I). These results could be due to a limited
number of particles and a formation of a quasi-stable configuration
of the particles during simulation, which collapsed with additional
loading and unloading steps. The use of a larger number of particles
is expected to improve the smoothness of the curves but require
significant computation time. Towards the end of simulation, an
increase of cycles did not increase the deformation. This phenom-
enon shows that 25 loading–unloading cycles adopted in this study
 A. Bhandari, J. Han / Geotextiles and Geomembranes 28 (2010) 33–43
Fig. 10. Loading–unloading cycles for a sheet of tiny particles at a depth of 25 mm.
are adequate to achieve a stable deformation state. In the discrete
element analysis of railway degradation, Lobo-Guerrero and Vallejo
(2006) noticed a stable permanent deformation after 60 cycles,
which are higher than that obtained in this study. Different model
heights, loading conditions, and porosities may result in this
difference. Lobo-Guerrero and Vallejo (2006) used a 0.6-m high
model generated at a porosity (n) of 0.2, which requires more cycles
for a stable permanent deformation than this study.
The loading–unloading cycles for two cases with the maximum
and minimum vertical deformations are shown in Figs. 10 and 11. In
both cases, the initial cycles consist of elastic and plastic defor-
mations. With an increase of cycles, the elastic deformations
became dominant and the hysteresis behavior diminished. The
load-deformation behavior as depicted in Figs. 10 and 11 shows that
the geotextile at a depth of 12.5 mm doubled the stiffness of the
model as compared with a sheet of tiny particles. A similar
comparison, when the geotextiles were placed at depths of 12.5 and
25.0 mm, shows the stiffness ratio (k12.5/k25.0) of 1.42, where k12.5
Fig. 11. Loading–unloading cycles for a geotextile at a depth of 12.5 mm.
39
Fig. 12. Maximum tension on the geotextiles.
and k25.0 are the stiffness of the models when reinforced with
geotextile at the depths of 12.5 mm and 25.0 mm respectively.
5.2. Stresses in geotextile
The stresses (sx, sy, and sxy) in the particles that represented the
geotextile are recorded at the end of 25 cycles (a loading stage). The
principal stresses in the geotextile can be calculated from these
stresses using the following equation.
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 
sx þ sy sx  sy 2 2
s1;2 ¼  þsxy (3)
2 2
where, s1,2 are the major and minor principal stresses.
The major principal stress corresponds to the tension in the
geogrid, which can be calculated as:
T ¼ s1  tg (4)
where T is the tension and tg is the thickness of the geotextile (i.e.,
diameter of the particles that represented the geotextile ¼ 1 mm).
The variation of the tension in the geotextile along the width of
the model is shown in Fig. 12. The maximum tension in the geo-
textile increased by 2.3 times when it was placed at a depth of
12.5 mm as compared with that at a depth of 25.0 mm.
The tensile strain (31) of the geotextile is related to its tensile
stress (s1) by the following relation:
s1 s1
31 ¼ ¼ (5)
E ðJ=tÞ
where, 31 is the tensile strain, s1 is the tensile stress, J is the tensile
stiffness, and t is the thickness of the geotextile.
The maximum tensile strains developed in the geotextiles were
0.038 and 0.0174% when the geotextiles were placed at depths of
12.5 and 25.0 mm, respectively. This result proves that only a small
fraction of the geotextile strength was mobilized in the reinforced
base under the simulated cyclic loading. The computed strains in
the geotextile in this study were lower than those of Miura et al.
(1990) and Hufenus et al. (2006) as previously discussed in the
Introduction. The reasons for the lower strains in this study are (1)
a lower vertical load of 298.5 N was applied and (2) the subgrade
was firm. Under such a low strain, the benefit of the geotextile as
a tensioned membrane is minor. Therefore, the contribution of the
40 A. Bhandari, J. Han / Geotextiles and Geomembranes 28 (2010) 33–43

Fig. 13. Displacement vectors of particles without a geotextile (the maximum displacement ¼ 13.26 mm and the deformation under the wheel ¼ 11.7 mm).

Fig. 14. Displacement vectors of particles for a geotextile at a depth of 25 mm (the maximum displacement ¼ 15.7 mm and the deformation under the wheel ¼ 14.6 mm).

geotextile in the base is to provide vertical confinement to particles
above the geotextile, i.e., separation, which will be discussed in the
following section.
5.3. Role of the geotextile in particle movement
Figs. 13–16 present displacement vectors of soil particles due to
the cyclic wheel load for four cases. The magnitudes of the
displacement vectors were based on the same scale as that in the
worst case (Case-IV), which had the maximum vertical deforma-
tion. Fig. 13 shows that the soil particles in the unreinforced section
under the load moved downward and laterally, in which a large
portion of particles moved downward. However, Figs. 14 and 15
both show that the geotextile helped minimize vertical movement
of particles as compared with those in the unreinforced section in
Fig. 13. The enlarged view of the particle displacement vectors
clearly illustrates the role of geotextile in preventing the vertical
movement of particles (Fig. 14). When the geotextile was placed at
a depth of 12.5 mm, the geotextile attained a curved shape and the
vertical displacement of the particles continued beneath the geo-
textile layer. However, the contact bond between particles that
represented the geotextile remained intact (Fig. 15). The enlarged

Fig. 15. Displacement vectors of particles for a geotextile at a depth of 12.5 mm (the maximum displacement ¼ 11.9 mm and the deformation under the wheel ¼ 11.7 mm).
 A. Bhandari, J. Han / Geotextiles and Geomembranes 28 (2010) 33–43
Fig. 16. Displacement vectors of particles for a sheet of tiny particles at a depth of 25.0 mm (the maximum displacement ¼ 20.4 mm and the deformation under the
wheel ¼ 18.3 mm).
view of the displacement vectors in Fig. 15 also shows a slippage of
particles towards the left side of the model. As shown in Fig. 16, the
largest particle movement occurred when a sheet of tiny particles
with a smaller contact friction (0.8 times the contact friction of
sand) and zero bonding strength existed. This result is anticipated
because the sheet of tiny particles with weaker friction contact
facilitates horizontal as well as vertical movements of particles. The
case with the sheet of tiny particles does not exist in the field,
however, it sets up a baseline for understanding the mechanisms
involved in the geotextile. In the geotextile-reinforced base, the
geotextile had a relatively strong bonding strength but weak fric-
tion resistance. As a result, two opposing mechanisms existed:
restricting the vertical movement of particles and facilitating the
horizontal movement of particles. If the vertical movement of
particles restricted by the geotextile is larger than that due to the
facilitated horizontal movement of particles, less deformation
would occur (for example, the geotextile placed at a depth of
12.5 mm (Case-II)) and vice versa (for example, the geotextile
placed at a depth of 25 mm (Case-III)). When the geotextile was
placed at a depth of 25.0 mm (Case-III), the vertical deformation on
the surface was mostly cumulative from the soil above the
Fig. 17. Deformation versus ratio of stiffness of the geotextile.
41
geotextile. Therefore, the effect of the geotextile in reducing the
vertical deformation is limited. However, when the geotextile was
placed at a depth of 12.5 mm (Case-II), the vertical deformation on
the surface was cumulative from the soil above and below the
geotextile. Therefore, the existence of the geotextile at this depth
prevents the particles from moving upward outside the loading
area and downward within the loading area, thus reducing the
vertical deformation.
5.4. Influence of the stiffness of a geotextile on interaction
During the DEM analysis, it was observed that the geotextile
particles always held together and the tensile failure of the geo-
textile was not an issue. Therefore, the stiffness of the geotextile is
more relevant than the bond strength to the performance of the
geotextile-reinforced sand.
Fig. 17 shows the effect of the stiffness of a geotextile on the
vertical deformation under the cyclic load. To investigate the effect
of the stiffness of a geotextile, the stiffness in the baseline case
(J ¼ 350 kN/m) was increased by 10 and 100 times. The plotted
Fig. 18. Loading–unloading cycles for the geotextile with 100J at a depth of 12.5 mm
(J is the stiffness of geotextile in the baseline case).
42 A. Bhandari, J. Han / Geotextiles and Geomembranes 28 (2010) 33–43
vertical deformations were obtained at the end of 25 cycles (the
loading stage) at different stiffness ratios as compared with those in
the baseline case. The tenfold increase in the stiffness of the geo-
textile had a marginal benefit in minimizing the vertical deforma-
tion for the geotextile either at a depth of 12.5 or 25.0 mm. Further
increase of the stiffness of the geotextile by 100 times showed an
improved benefit in minimizing the vertical deformation. However,
the benefit was less significant for the geotextile at the depth of
25.0 mm as compared with that at the depth of 12.5 mm.
Fig. 18 shows loading–unloading cycles of the reinforced base
with hundred-fold stiffness of the geotextile placed at a depth of
12.5 mm. As compared with the geotextile placed at the same depth
(12.5 mm) in the baseline case (Fig. 11), Fig. 18 shows both elastic
and plastic deformations developing within the entire loading–
unloading cycles. Even though the geotextile with a high stiffness
could control the plastic deformation at the initial loading–
unloading cycles, the hysteresis behavior continued up to the end of
the analysis due to the continuous re-arrangement of the particles.
Therefore, the stiffer geotextile promoted continuous re-arrange-
ment of the particles.
6. Limitations
This paper discussed the DEM analysis of geotextile and soil
interaction using an elastic stiffness model of cylindrical particles.
The main purpose of this study was to capture the qualitative
behavior of the granular base subjected to a cyclic vertical load. Few
simplifications on the numerical models were needed to obtain the
simulation results in a reasonable time. The real three-dimensional
problem was simplified to a two-dimensional plane assembly. As
mentioned by Rothenburg and Bathurst (1992), the two-dimen-
sional plane assembly undergoes dilation at a higher rate than the
three-dimensional assembly since the two-dimensional assembly
can dilate in one direction only. A parameter as simple as Poisson’s
ratio for the two-dimensional assembly is different from the three-
dimensional assembly (Bathurst and Rothenburg, 1988). Skermer
and Hillis (1970) and Lekarp et al. (2000) found that an appropriate
gradation of the granular particles showed the maximum shear
strength. This study used uniform particles of cylindrical shape,
which resulted in a lower strength of the granular base. The geo-
textile reinforcement of a weaker base would show more benefits
than that of the otherwise strong base. The irregular shape of the
particles can be introduced by bonding cylindrical shaped particles
in a certain fashion to form clusters. The irregular shape can also be
simulated using clumps (Lu and McDowell, 2007) or ellipses
(Rothenburg and Bathurst, 1992, 1993). The model considering the
influence of particle shape is under development by the authors
and will be published in a future publication. In addition, breakage
of particles has not been considered in this study. Inclusion of all
these factors would make the DEM study closer to the reality but
require significant computation time. On the other hand, results
from two-dimensional models while not as quantitatively reliable
as those from three-dimensional models, are nonetheless useful
and help understand the general behavior and performance of
geotextile-reinforced granular bases.
7. Conclusions
This DEM study investigated the geotextile–soil interaction
under a cyclic vertical load. It is shown that the development of
tensile stresses in the geotextile helped improve the performance
of the reinforced section. The geotextile placed at a depth of
25.0 mm had small tensile stresses and was less effective in mini-
mizing the surface deformation caused by the cyclic vertical load as
compared with the geotextile placed at a depth of 12.5 mm.
The DEM results show that the geotextile restrained the parti-
cles from vertical movement. At the same time, the geotextile
facilitated horizontal movement of particles since it had lower
frictional resistance to the soil particles when rolling and sliding on
the geotextile. Hence, the benefit of the geotextile in minimizing
the vertical deformation depends upon the above combined effects,
which are related to the location of the geotextile.
The DEM analysis also demonstrated that a low tensile strain
developed in the geotextile under cyclic vertical loading. An
increase in the stiffness of the geotextile showed a marginal
improvement of the performance, particularly when the geotextile
was placed at a depth of 25.0 mm.
To generate actual quantitative results, 3-D DEM models
considering particle shape, gradation, and crushing are necessary,
however, they require significant computation time.
Acknowledgements
This research was jointly supported by the University of Kansas,
Transportation Research Institute from Grant # DT0S59-06-
G-00047, provided by the US Department of Transportation –
Research and Innovative Technology Administration and the
Geosynthetic Institute under its GSI Fellowship Program. This
support is greatly appreciated. The authors are thankful for
Prof. A. Misra for his valuable suggestions during the preparation of
this paper. The authors also appreciate the reviewers’ excellent
comments and suggestions, which have helped improve the quality
of this paper.
References
Bathurst, R.J., Rothenburg, L., 1988. Micromechanical aspects of isotropic granular
assemblies with linear contact interactions. Journal of Applied Mechanics,
Transactions ASME 55 (1), 17–23.
Bathurst, R.J., Rothenburg, L., 1990. Observations on stress-force-fabric relationships
in idealized granular materials. Mechanics of Materials 9 (1), 65–80.
Bhandari, A., Han, J., Parsons, R.L., 2008. DEM analysis of geotextile–soil interaction
under wheel loading. In: Paper Accepted for Presentation and Publication at the
Research Symposium on the Characterization and Behavior of Interfaces (CBI),
Atlanta.
Brown, S.F., Kwan, J., Thom, N.H., 2007. Identifying the key parameters that influ-
ence geogrid reinforcement of railway ballast. Geotextiles and Geomembranes
25 (6), 326–335.
Chareyre, B., Villard, P., 2002. Discrete element modeling of curved geosynthetic
anchorages with known macro-properties. In: Paper Presented at the Numer-
ical Modeling in Micromechanics Via Particle Methods, Gelsenkirchen,
Germany.
Cundall, P.A., Strack, O.D.L., 1979. A discrete numerical model for granular assem-
blies. Geotechnique 29 (1), 47–65.
DeMerchant, M.R., Valsangkar, A.J., Schriver, A.B., 2002. Plate load tests on geogrid-
reinforced expanded shale lightweight aggregate. Geotextiles and Geo-
membranes 20 (3), 173–190.
Douglas, R.A., Valsangkar, A.J., 1992. Unpaved geosynthetic-built resource access
roads. Stiffness rather than rut depth as the key design criterion. Geotextiles
and Geomembranes 11 (1), 45–59.
Han, J., Yang, X.M., Leshchinsky, D., Parsons, R.L., 2008a. Behavior of geocell-rein-
forced sand under a vertical load. Journal of Transportation Research Board
2045, 95–101.
Han, J., Zhang, Y.-Z., Parsons, R.L., 2008b. Development of a performance-based
laboratory test method for evaluating geosynthetic–soil confinement. In: Paper
Presented at the CD-Rom Publication, the 87th TRB Annual Meeting, January 13
to 17, Washington, DC.
Hufenus, R., Rueegger, R., Banjac, R., Mayor, P., Springman, S.M., Bronnimann, R.,
2006. Full-scale field tests on geosynthetic reinforced unpaved roads on soft
subgrade. Geotextiles and Geomembranes 24 (1), 21–37.
Itasca, 2004. Particle Flow Code in Two Dimensions. Itasca Consulting Group, Inc.,
Minnesota.
Kelly, D., Fairfield, C., Sibbald, A., 1995. Geosynthetics for the improvement of
unpaved roads. Highways and Transportation 42 (7–8), 13–15.
Konietzky, H., te Kamo, L., Groeger, T., Jenner, C., 2004. Use of DEM to model the
interlocking effect of geogrids under static and cyclic loading. In: Paper Pre-
sented at the Numerical Modeling in Micromechanics via Particle Methods,
Kyoto, Japan.
Lekarp, F., Isacsson, U., Dawson, A., 2000. State of the art. I: resilient response of
unbound aggregates. Journal of Transportation Engineering 126 (1), 66–75.
 A. Bhandari, J. Han / Geotextiles and Geomembranes 28 (2010) 33–43
Lim, W.L., McDowell, G.R., 2005. Discrete element modelling of railway ballast.
Granular Matter 7, 19–29.
Lobo-Guerrero, S., Vallejo, L.E., 2006. Discrete element method analysis of railtrack
ballast degradation during cyclic loading. Granular Matter 8 (3–4), 195–204.
Lu, M., McDowell, G.R., 2007. The importance of modelling ballast particle shape in
the discrete element method. Granular Matter 9 (1–2), 69–80.
Masson, S., Martinez, J., 2000. Multiscale simulations of the mechanical
behaviour of an ensiled granular material. Mechanics of Cohesive-Frictional
Materials 5 (6), 425–442.
McDowell, G.R., Harireche, O., Konietzky, H., Brown, S.F., Thom, N.H., 2006. Discrete
element modelling of geogrid-reinforced aggregates. Geotechnical Engineering
159 (GEI), 35–48.
Miura, N., Sakai, A., Taesiri, Y., Yamanouchi, T., Yasuhara, K., 1990. Polymer grid
reinforced pavement on soft clay grounds. Geotextiles and Geomembranes
9 (1), 99–123.
Munjiza, A., 2004. The Combined Finite-discrete Element Method. John Wiley and
Sons, England.
Perkins, S.W., Ismeik, M., 1997. Synthesis and evaluation of geosynthetic-reinforced
base layers in flexible pavements: part I. Geosynthetics International 4 (6),
549–604.
Potyondy, D.O., Cundall, P.A., 2004. A bonded-particle model for rock. International
Journal of Rock Mechanics and Mining Sciences 41 (8 SPECISS), 1329–1364.
Raymond, G.P., 2002. Reinforced ballast behaviour subjected to repeated load.
Geotextiles and Geomembranes 20 (1), 39–61.
43
Rothenburg, L., Bathurst, R.J., 1992. Micromechanical features of granular assem-
blies with planar elliptical particles. Geotechnique 42 (1), 79–95.
Rothenburg, L., Bathurst, R.J., 1993. Influence of particle eccentricity on micro-
mechanical behavior of granular materials. Mechanics of Materials 16 (1–2),
141–152.
Saad, B., Mitri, H., Poorooshasb, H., 2006. 3D FE analysis of flexible pavement with
geosynthetic reinforcement. Journal of Transportation Engineering 132 (5),
402–415.
Skermer, N.A., Hillis, S.F., 1970. Gradation and shear characteristics of four cohe-
sionless soils. Canadian Geotechnical Journal 7 (1), 62–68.
Tingle, J.S., Jersey, S.R., 2005. Cyclic plate load testing of geosynthetic-reinforced
unbound aggregate roads. Transportation Research Record (1936), 60–69.
Villard, P., Chareyre, B., 2004. Design methods for geosynthetic anchor trenches on
the basis of true scale experiments and discrete element modeling. Canadian
Geotechnical Journal 41, 1193–1205.
Walters, D.L., Allen, T.M., Bathurst, R.J., 2002. Conversion of geosynthetic strain to
load using reinforcement stiffness. Geosynthetics International 9 (5–6),
483–523.
Webster, S.L., 1992. Geogrid Reinforced Base Courses for Flexible Pavements for
Light Aircraft: Test Section Construction, Behavior Under Traffic, Laboratory
Tests, and Design Criteria. U.S. Department of Transportation and Federal
Aviation Administration.
Zhang, L., Thornton, C., 2007. A numerical examination of the direct shear test.
Geotechnique 57 (4), 343–354.