You are on page 1of 16

Influence of Selection of Soil and Interface Properties

on Numerical Results of Two Soil–Geosynthetic


Interaction Problems
Yan Yu1 and Richard J. Bathurst2

Abstract: Numerical modelers are often faced with the challenge to numerically reproduce the physical behavior of soil–geosynthetic
Downloaded from ascelibrary.org by University of Toronto on 11/10/16. Copyright ASCE. For personal use only; all rights reserved.

interaction problems using assumed values for missing soil and interface model parameters. This paper examines two such examples
using the finite-difference method, (1) horizontal pullout of a geosynthetic (geogrid) reinforcement layer in a pullout box and (2) a geosyn-
thetic (geotextile)-reinforced soil layer over a void. It also presents the results of parametric sensitivity analyses for the missing soil and
interface model parameter values and identifies the values that give the best agreement with measured data. The paper further demon-
strates that correct modeling of geometrical nonlinearity is key to accurately predicting the performance of geosynthetic-reinforced soil
systems controlled by the tensioned membrane effect. The lessons learned in this study will be of interest to numerical modelers during
numerical model design of geosynthetic-reinforced soil systems. DOI: 10.1061/(ASCE)GM.1943-5622.0000847. © 2016 American
Society of Civil Engineers.
Author keywords: Geosynthetics; Numerical modeling; Pullout test; Voids; Geosynthetic-soil interaction; Interfaces; FLAC.

Introduction the reinforcement layer are sensibly horizontal. Examples of


physical and numerical analysis of geosynthetic-reinforced and
Numerical modeling of geosynthetic-reinforced soil systems is steel-reinforced pullout tests are the work of Alam et al. (2014),
now widely used for the design of field structures, prediction of Bathurst and Ezzein (2016), Ezzein and Bathurst (2014), Fannin
measured responses, and in research to generate synthetic data for and Raju (1993), Moraci and Recalcati (2006), Ochiai et al.
the purpose of filling in knowledge gaps on the behavior of these (1996), Perkins and Edens (2003), Rouse et al. (2014), Wilson-
systems. However, in almost all cases, assumptions must be made Fahmy and Koerner (1993), and Yogarajah and Yeo (1994),
regarding the choice of some or all of the parameter values for the among others. The practical application of pullout testing and
soil and soil-reinforcement interaction constitutive models that modeling is the performance of the reinforcement in the anchor-
are used in advanced numerical codes (computer programs). The age zone of mechanically stabilized earth walls (Allen and
accuracy of numerical outcomes when compared with physical Bathurst 2014a, b; Damians et al. 2015; Karpurapu and Bathurst
measurements may be sensitive to the choice of magnitude of the 1995; Rowe and Skinner 2001; Yu et al. 2015a, b, 2016a, b).
missing parameters in some cases and negligible in others. This The second example of geosynthetic-reinforced soil struc-
paper considers two examples in which parameter values for the tures examined in this paper is a geosynthetic-reinforced soil
component constitutive models in the finite-difference method layer over a void. For this case, large vertical displacements of
(FDM) program FLAC are not available. The numerical results the reinforced soil layer above the void can occur and the rein-
using a range of missing parameter values from physical tests forcement layers will displace both horizontally and vertically
generating tensile loads in the horizontal anchorage zone and in the
reported in the literature are compared with the matching meas-
vertically deformed reinforcement layer over the void. Examples of
ured physical test results.
the behavior of these systems can be found in the papers by
The first example is a geogrid reinforcement layer in a pullout
Briançon et al. (2005), Briançon and Villard (2008), Bridle and
box subjected to horizontal pullout loads. The geosynthetic rein-
Jenner (1997), Feng and Lu (2015), Galve et al. (2012), Giroud et al.
forcement layer is placed horizontally within the soil mass to
(1990), Wang et al. (1996), Jones and Cooper (2005), Ponomaryov
carry axial tensile loads. Vertical differential displacements
and Zolotozubov (2014), Villard and Briançon (2008), Villard et al.
between the soil and reinforcement are negligible in the rein-
(2009), Villard et al. (2000), and Wang et al. (2009), among others.
forced soil mass. Hence, axial strains and loads generated within
The second example in this paper with the unsupported geosynthetic
located just above a void can be considered as the limiting case for
1
Postdoctoral Fellow, GeoEngineering Centre at Queen’s-RMC, Dept. geosynthetic-reinforced embankments over soft zones (Blanc et al.
of Civil Engineering, Royal Military College of Canada, Kingston, ON, 2014; Benmebarek et al. 2015; Cao et al. 2016; Xu et al. 2016; Yu et
Canada K7K 7B4. E-mail: yan.yu@queensu.ca al. 2016c).
2
Professor, GeoEngineering Centre at Queen’s-RMC, Dept. of Civil Unless noted otherwise, the numerical simulations in this
Engineering, Royal Military College of Canada, Kingston, ON, Canada
study were performed in the large-strain mode. Readers are
K7K 7B4 (corresponding author). E-mail: bathurst-r@rmc.ca
Note. This manuscript was submitted on May 6, 2016; approved on referred to the FLAC manual (Itasca 2011) for more details
September 21, 2016; published online on November 8, 2016. Discussion regarding FLAC small-strain and large-strain mode options. The
period open until April 8, 2017; separate discussions must be submitted lessons learned in this paper are valuable to numerical modelers
for individual papers. This paper is part of the International Journal of investigating soil–geosynthetic interaction problems of the type
Geomechanics, © ASCE, ISSN 1532-3641. described in the paper using the program FLAC.

© ASCE 04016136-1 Int. J. Geomech.

Int. J. Geomech., 04016136


Geogrid Pullout Test Example Numerical Modeling Details
Reinforced soil walls with planar soil-reinforcement elements are,
Physical Tests for practical purposes, plane strain problems and are most often ana-
lyzed and designed as such using analytical methods (AASHTO
The numerical geogrid pullout models are based on physical tests
2014; BSI 2016) or numerical methods (Damians et al. 2015;
reported by Moraci and Recalcati (2006). The tests were performed
Karpurapu and Bathurst 1995; Rowe and Skinner 2001; Yu et al.
in a pullout box 1.70 m long, 0.60 m wide, and 0.68 m high
2015a, 2016a, among many others). The only practical experimen-
[Fig. 1(a)]. The geogrid specimens were placed at middepth in 0.6
tal method for deducing soil-geogrid interface properties for design
m of granular soil. Four direct shear box tests were performed with
and analysis is to perform pullout or direct interface shear tests in
the same soil prepared to 95% of Standard Proctor maximum dry
mechanical devices of limited width. Moraci and Recalcati (2006)
density matching the as-compacted soil in the pullout tests. The se-
reviewed the literature for lessons about how to construct and per-
cant friction angle for the soil from direct shear tests varied from
form pullout tests that approach plane strain conditions as far as
f ds = 48° at s v = 10 kPa to f ds = 42° at s v = 100 kPa.
Moraci and Recalcati (2006) tested three different extruded inte- practical and adopted these recommendations in their experimen-
Downloaded from ascelibrary.org by University of Toronto on 11/10/16. Copyright ASCE. For personal use only; all rights reserved.

gral uniaxial high-density polyethylene (HDPE) geogrids. In the tal program. Rouse et al. (2014) compared the measured and
current study, the tests with the highest tensile strength product back-calculated results of the pullout resistance of two planar
(GG3) are used because load-displacement curves up to pullout fail- inclusions and concluded that plane strain conditions prevail in a
ure are available for all applied vertical stresses. Tests with GG1 typical pullout test box at large displacements. Based on the pre-
and GG2 materials ruptured before pullout was achieved when the vious comments, the numerical geogrid pullout test models were
applied stress was 100 kPa. Based on the reported data for the in- performed using the two-dimensional (2D) version of the pro-
isolation tensile tests performed at a 1 mm/min displacement rate, gram FLAC and plane strain boundary conditions (Itasca 2011).
the tensile stiffness of Geogrid GG3 was J = 1,903 kN/m (i.e., The plane strain FLAC numerical grid and boundary conditions
secant stiffness calculated using tensile load at 2% axial strain), for the soil within the pullout box are shown in Fig. 1(b). A total of
and the ultimate tensile strength was Tult = 118 kN/m. The GG3 3,376 zones for the soil were used. The soil domain was fixed in x-
geogrid specimens in the pullout tests had a length of 1.15 m and and y-directions at the bottom boundary and in the x-direction at
a width of 0.58 m. The pullout tests were performed under four both the left and right vertical boundaries. The pullout clamp
different surcharge pressures using an air bag. The corresponding extended 0.35 m into the soil from the left pullout box boundary.
vertical stresses at the geogrid elevation were s v = 10, 25, 50, and The purpose of using 0.25-m long upper and lower sleeves in the
100 kPa. The tensile load in each geogrid specimen was measured original test was to reduce the effect of the front pullout box bound-
using a load cell placed between an electric jack and the clamping ary on the measured pullout loads. However, to simplify the model
system. the soil boundaries above and below the clamp in Fig. 1(b) were
fixed in the y-direction, and the soil boundary at the sleeve ends was
fixed in the x-direction [Fig. 1(c)]. Using finer numerical grids than
those shown in Fig. 1 (and also in the void case study that follows
Surcharge pressure later) had negligible influence on the numerical results for the
examples and boundary conditions examined.
0.30 m 0.30 m

Upper sleeve The reported geogrid pullout loads from Moraci and Recalcati
Clamp Geogrid
(2006) were corrected loads that account for soil-clamp friction.
Soil

Pullout load
Thus, no attempt was made to model the interaction between the
Lower sleeve clamp and soil. A single beam element was used to model the clamp.
The two nodes of the beam element were initially fixed in the x- and
0.25 m y-directions and then advanced to the left in the x-direction with the
0.10 m 1.15 m same pullout velocity.
(a)
A total of 115 cable elements were used to model each geo-
1.70 m
grid specimen. The first geogrid cable node was slaved to the
Surcharge pressure right beam node at the end of the clamp to generate the same x-
and y-displacements during pullout.
0.30 m 0.30 m

The soil was modeled as a linear elastic-plastic material (i.e.,


Geogrid Mohr-Coulomb model). The plane strain secant friction angle used
Soil

Pullout load in each numerical simulation was computed using the transforma-
y tion f = tan−1(1.2  tan f ds) (Kulhawy and Mayne 1990) applied
to the direct shear friction angle at the corresponding stress level.
x
(b) Clamp The conversion of direct shear test (and triaxial test) peak friction
angles to plane strain values that better reflect field conditions for
typical reinforced soil walls has been recommended by Allen and
Bathurst (2014a, b) and Yu et al. (2016a). The soil Young’s modu-
lus and cohesion were not reported by Moraci and Recalcati (2006).
Thus, a range of soil Young’s modulus E = 10 to 100 MPa (Budhu
(c)
2010) and cohesion c = 1–5 kPa were examined. The influence of
Fig. 1. Example showing horizontal geogrid specimen subjected to
stress-dependent stiffness of the granular soil on the pullout loads
horizontal pullout load: (a) schematic of physical pullout test; was investigated by comparing numerical outcomes under the same
(b) matching FLAC numerical grid and boundary conditions; (c) local applied vertical stress but with different soil Young’s moduli. The
soil boundary conditions around the clamp granular soil properties are summarized in Table 1. The choice of
clamp properties does not influence numerical outcomes because

© ASCE 04016136-2 Int. J. Geomech.

Int. J. Geomech., 04016136


the clamp is subjected to a prescribed horizontal velocity only. clamp by applying the same pullout velocity to the two beam
The interface properties between the granular backfill and geogrid nodes.
were also unknown. The interface was modeled using the grouted
cable element available in FLAC. The interface frictional and cohe- Influence of a Granular Soil Young’s Modulus
sive shear strength components were computed by applying the
same strength reduction factor (Ri) as that applied to the granular Fig. 2 shows the influence of the value of the granular soil Young’s
soil friction angle and cohesion, where Ri = 0.5–0.8. Grout stiffness modulus on predicted geogrid pullout load-displacement response
values of Ks,gg = 1–100 MN/m/m were selected in the parametric under different applied vertical pressures. The numerical results
analyses to follow. The cable elements were assigned tensile yield show that the Young’s modulus ranging from 10 to 100 MPa had
strength equal to the geogrid ultimate tensile strength (i.e., negligible influence on numerical results. Yu et al. (2016a) com-
118 kN/m). More details about the calculation of cable properties pared results using a linear elastic-plastic soil model (with con-
are provided by Yu et al. (2015b) and in the footnotes of Table 2. stant soil stiffness) and a nonlinear elastic-plastic model (with
Each pullout test simulation started by activating 0.1-m-thick soil stress-dependent soil stiffness) to predict the behavior of two well-
layers in sequence matching the actual test construction method, instrumented geogrid-reinforced modular block walls (Allen and
Downloaded from ascelibrary.org by University of Toronto on 11/10/16. Copyright ASCE. For personal use only; all rights reserved.

until the final thickness of 0.6 m was achieved. Nevertheless, acti- Bathurst 2014a, b). They found that a properly selected single value
vating the total 0.6-m-thick soil layer in one step had negligible of the Young’s modulus for the backfill soil implemented in a linear
influence on the numerical outcomes. The clamp beam element elastic-plastic model gave similar wall-facing displacements and
and geogrid cable elements were added to the model when the reinforcement strains using the nonlinear elastic-plastic soil model.
soil thickness reached 0.3 m. After the surcharge pressure was Comparisons of measured and numerical results in Fig. 2 show that
applied, the model was solved to force equilibrium. Next, the a simple linear-elastic model with Mohr-Coulomb failure criterion is
specimen was subjected to a constant rate of displacement at the able to reproduce the physical load-displacement performance of the
geogrid pullout tests to practical satisfactory accuracy.

Table 1. Granular Soil Property Values for Geogrid Pullout Test Example
Influence of the Granular Soil Cohesive
Property Value Strength Component
Unit weighta [ g (kN/m3)] 15.4 It is common practice to add a very small cohesive soil strength
Young’s modulus [E (MPa)] 10–100 component to frictional soil backfills in soil-structure interaction
Poisson’s ratio () 0.3 problems, such as reinforced soil walls and slopes, to ensure numer-
Friction angleb [ f (°), s v (kPa)] 47, 100 ical stability at unconfined domain boundaries (1–2 kPa) (Alam et
49, 50 al. 2014; Rowe and Skinner 2001; Yu et al. 2015a, 2016a–c). The
51, 25 influence of the granular soil cohesive component on the geogrid
53, 10 pullout load-displacement response is shown in Fig. 3. Increasing
Dilation anglec [ c (°), s v (kPa)] 17, 100 the soil cohesion from 1 to 5 kPa generated a detectable increase in
19, 50 the geogrid pullout loads as expected, but the effect was minor. Fig.
21, 25 3 shows that the calculated pullout load-displacement responses
23, 10 using cohesion of 1 kPa agree well with the measured data.
Cohesion [c (kPa)] 1–5
a
g = 95%  g dmax. Influence of the Strength Reduction Factor
b
f = tan−1(1.2  tan f ds) for plane strain boundary condition. The strength reduction factor (Ri) is used to calculate the shear
c
c = f − 30°. strength parameter values of the interface between the soil and geo-
synthetic based on the soil friction angle and cohesion (Yu et al.
2015b). Thus, the grout frictional resistance is calculated as f gg =
Table 2. Cable Element Property Values for Geogrid Pullout Test
Example tan−1(Ri  tan f ), where f = sand backfill friction angle. The grout
cohesive strength component (Cgg) is computed as Cgg =
Cable element parameter Value Ri  c  Pgg, where c = sand backfill cohesion and Pgg = cross-
a
Tensile stiffness [J (kN/m)] 1,903 section perimeter of the geogrid with an out-of-plane width
Young’s modulusb [Egg (MPa)] 951.5 of 1 m (Pgg = 2 m for the plane strain condition). Interface
Cross-sectional area [Agg (m2/m)] 2  10−3 properties are summarized in Table 2.
Exposed perimeter [Pgg (m)] 2 Fig. 4 shows the influence of the shear strength reduction factor
Tensile yield strengtha [Tult (kN/m)] 118 on the geogrid pullout load-displacement responses. The numerical
Grout stiffness [Ks,gg (MN/m/m)] 1–100 results using Ri = 0.5–0.8 show that increasing the shear strength
Grout frictional resistancec [ f gg(°), s v (kPa)] 28.2–40.6, 100 reduction factor increased the pullout loads. The best agreement
29.9–42.6, 50 between the calculated and measured pullout load-displacement
31.7–44.6, 25 responses was obtained for a strength reduction factor Ri = 0.67.
33.5–46.7, 10
Grout cohesive strengthd [Cgg (kN/m)] 1.34–6.7 Influence of the Soil-Geogrid Interface Stiffness
a
With an out-of-plane width of 1 m [for tensile yield strength a unit of Fig. 5 shows the influence of cable element grout stiffness on the
force was used in the FLAC manual (Itasca 2011)]. geogrid pullout load-displacement responses. The numerical
b
Egg = J/Agg. results show that the use of grout stiffness Ks,gg = 1 MN/m/m
c
f gg = tan−1(Ri  tan f ) with Ri = 0.5 to 0.8. underestimated the pullout loads. However, the influence of grout
d
Cgg = Ri  c  Pgg. stiffness greater than 20 MN/m/m on the pullout loads is

© ASCE 04016136-3 Int. J. Geomech.

Int. J. Geomech., 04016136


Measured (Moraci and Recalcati 2006)
Calculated (E = 10 MPa)
Calculated (E = 30 MPa)
Calculated (E = 100 MPa)
150 150
σv= 10 kPa σv= 25 kPa

Pullout load (kN/m)

Pullout load (kN/m)


Large-strain mode
100 c = 1 kPa 100
Ri = 0.67
Ks,gg = 20 MN/m/m
50 50

0 0
0 25 50 75 100 0 25 50 75 100
Downloaded from ascelibrary.org by University of Toronto on 11/10/16. Copyright ASCE. For personal use only; all rights reserved.

Geogrid front end displacement (mm) Geogrid front end displacement (mm)
(a) (b)

150 150
σv= 50 kPa σv= 100 kPa

Pullout load (kN/m)


Pullout load (kN/m)

100 100

50 50

0 0
0 25 50 75 100 0 25 50 75 100

Geogrid front end displacement (mm) Geogrid front end displacement (mm)
(c) (d)

Fig. 2. Influence of the soil Young’s modulus on calculated pullout load-displacement responses

Measured (Moraci and Recalcati 2006)


Calculated (c = 1 kPa)
Calculated (c = 3 kPa)
Calculated (c = 5 kPa)
150 150
σv= 10 kPa σv= 25 kPa
Pullout load (kN/m)

Pullout load (kN/m)

Large-strain mode
100 100
E = 30 MPa
Ri = 0.67
Ks,gg = 20 MN/m/m
50 50

0 0
0 25 50 75 100 0 25 50 75 100

Geogrid front end displacement (mm) Geogrid front end displacement (mm)
(a) (b)

150 150
σv= 50 kPa σv= 100 kPa
Pullout load (kN/m)
Pullout load (kN/m)

100 100

50 50

0 0
0 25 50 75 100 0 25 50 75 100

Geogrid front end displacement (mm) Geogrid front end displacement (mm)
(c) (d)

Fig. 3. Influence of soil cohesive strength component on calculated pullout load-displacement responses

© ASCE 04016136-4 Int. J. Geomech.

Int. J. Geomech., 04016136


Measured (Moraci and Recalcati 2006)
Calculated (Ri = 0.5)
Calculated (Ri = 0.67)
Calculated (Ri = 0.8)
150 150
σv= 10 kPa σv= 25 kPa

Pullout load (kN/m)

Pullout load (kN/m)


Large-strain mode
100 E = 30 MPa 100
c = 1 kPa
Ks,gg = 20 MN/m/m

50 50

0 0
Downloaded from ascelibrary.org by University of Toronto on 11/10/16. Copyright ASCE. For personal use only; all rights reserved.

0 25 50 75 100 0 25 50 75 100

Geogrid front end displacement (mm) Geogrid front end displacement (mm)
(a) (b)

150 150
σv= 50 kPa σv= 100 kPa

Pullout load (kN/m)


Pullout load (kN/m)

100 100

50 50

0 0
0 25 50 75 100 0 25 50 75 100

Geogrid front end displacement (mm) Geogrid front end displacement (mm)
(c) (d)

Fig. 4. Influence of strength reduction factor on calculated pullout load-displacement responses

negligible. The calculated pullout load-displacement responses pullout capacity of the most heavily surcharged test is controlled by
using Ks,gg = 20 MN/m/m agreed well with the measured data, the tensile strength of the reinforcement, which can be expected for
and this value is used from now on. pullout tests that are performed under large surcharge load to ulti-
mate tensile strength ratios (Huang and Bathurst 2009).
Influence of the FLAC Strain Mode Fig. 7 shows the measured and calculated geogrid pullout loads
at four different applied vertical stresses. Increasing the applied ver-
The parametric analyses described in previous sections for the tical stress on the geogrid increased the ultimate geogrid pullout
geogrid pullout test demonstrated that the measured pullout load- capacity. The results show that the calculated geogrid pullout loads
displacement responses were generally in good agreement with agree well with the measured peak and large-displacement pullout
numerical predictions using large-strain mode and backfill Young’s loads reported by Moraci and Recalcati (2006) when sand and inter-
modulus E = 30 MPa, backfill cohesion c = 1 kPa, strength reduction face parameter values that are not available in their study were taken
factor Ri = 0.67, and cable grout stiffness Ks,gg = 20 MN/m/m. Fig. 6 as E = 30 MPa (the backfill Young’s modulus), c = 1 kPa (the back-
shows the influence of the FLAC large-strain and small-strain modes fill cohesion), Ri = 0.67 (the strength reduction factor), and Ks,gg =
on the geogrid pullout loads. From a practical point of view, the 20 MN/m/m (the cable element grout stiffness) and using the peak
influence of the choice of FLAC small-strain and large-strain modes plane strain soil friction angle.
on the calculated pullout load-displacement responses is judged to
be negligible for the example geometry, physical test displacement
rate, boundary conditions, and the simple constitutive models for the Summary of Geogrid Pullout Test Model Results
component materials in this example.
Generally, good agreement is judged to have been achieved
between physical and numerical results using the set of soil and
Measured and Calculated Peak and interface parameter values identified earlier. This good agreement
Large-Displacement Geogrid Pullout Loads using this set of parameters was achieved regardless of whether the
simulations were performed in small-strain or large-strain mode.
Previous data plots have shown that the peak pullout capacity of Better predictions could be possible using other soil-geogrid inter-
each test can be reasonably well predicted even though linear face models and/or more advanced tensile load models for the geo-
elastic-plastic (Mohr-Coulomb) models adopted for the soil and the grid, such as the time-dependent and strain-dependent stiffness
soil-geogrid interface are very simple. The predicted ultimate model by Walters et al. (2002) and Allen and Bathurst (2014b) and/

© ASCE 04016136-5 Int. J. Geomech.

Int. J. Geomech., 04016136


Measured (Moraci and Recalcati 2006)
Calculated (Ks,gg = 1 MN/m/m)
Calculated (Ks,gg = 20 MN/m/m)
Calculated (Ks,gg = 100 MN/m/m)
150 150
σv= 10 kPa σv= 25 kPa

Pullout load (kN/m)

Pullout load (kN/m)


100 Large-strain mode 100
E = 30 MPa
c = 1 kPa
Ri = 0.67
50 50

0 0
Downloaded from ascelibrary.org by University of Toronto on 11/10/16. Copyright ASCE. For personal use only; all rights reserved.

0 25 50 75 100 0 25 50 75 100

Geogrid front end displacement (mm) Geogrid front end displacement (mm)
(a) (b)

150 150
σv= 50 kPa σv= 100 kPa

Pullout load (kN/m)


Pullout load (kN/m)

100 100

50 50

0 0
0 25 50 75 100 0 25 50 75 100

Geogrid front end displacement (mm) Geogrid front end displacement (mm)
(c) (d)

Fig. 5. Influence of cable element grout stiffness on calculated pullout load-displacement responses

or the rate-dependent models by Ezzein et al. (2015) and Kongkitkul surcharge increments of q2 = 5.3 and 12.4 kPa were applied to the
et al. (2014). However, these models require additional in-isolation gravel layer surface directly over the void. The strains in the geo-
tensile test data that were not available. synthetic reinforcement at different locations were measured using
optical fiber strain sensors.
The linear-elastic Mohr-Coulomb constitutive model was used
Geotextile-Reinforced Soil Layer over a Void Example for the gravel layer and supporting soil. Using the reported interface
friction angle between the gravel and the geotextile, the interface
Numerical Modeling Details reduction factor was computed as Ri = tan( f upper)/tan( f g) = tan
(30°)/tan(44°) = 0.6. Thus, the friction angle of the supporting
The second example is a geotextile-reinforced soil (gravel) layer soil was back-calculated to be f s = tan−1[tan( f lower)/Ri] = tan−1
over a void [Fig. 8(a)] based on a physical test reported by Villard [tan(25°)/0.6] = 38° using the same value of Ri = 0.6 for the inter-
and Briançon (2008). The reinforced soil layer at the location of the face between the supporting soil and the geotextile. The cohesive
void in the figure was initially supported by two inflated air bags. strength component of the supporting soil was not reported, and in
The gravel layer had a thickness of hg = 0.5 m and was compacted the current study values of cs = 2–10 kPa were examined. The
to a unit weight of g g = 17.0 kN/m3. Villard et al. (2009) reported a gravel layer with a Young’s modulus from Eg = 10–100 MPa and
friction angle of f g = 44° for the gravel. The geotextile had a tensile cohesion cg = 1–5 kPa was investigated. The gravel and supporting
stiffness of Jgt = 1,100 kN/m in the longitudinal direction and a ten- soil properties are summarized in Table 3.
sile rupture strength of Tf,gt = 125 kN/m measured at 12% of strain. The simulations were run using FLAC with an out-of-plane
The supporting soil (original ground) was assigned a Young’s mod- width of 1 m matching the plane strain conditions of the original
ulus of Es = 30 MPa by Villard and Briançon (2008). No other infor- physical test. The in-plane horizontal length of 10 m starting from
mation for the supporting soil is reported. The interface friction the left fixed end of the geotextile was modeled [Fig. 8(a)]. Two dif-
angle was reported as f upper = 30° between the gravel and geotex- ferent model cases were performed as described in the following:
tile, and f lower = 25° between the supporting soil and geotextile. 1. Case 1 modeled the geotextile using 200 beam elements. The
The physical test began by deflating the air bags, resulting in a gravel layer and supporting soil were discretized into 6,000
void under the geotextile and continued by surcharging the gravel zones [Fig. 8(b)]. The geotextile was fixed at the left end and
layer in the vicinity of the void. During the surcharging stage, the free at the right end. All beam nodes were defined using x- and
gravel layer surface over a 2-m distance adjacent to the left side of y-coordinates except the beam node at the center of void (x =
the void was surcharged first (q1 = 16 kPa). Thereafter, two 3 m), which was defined using a grid number. It should be

© ASCE 04016136-6 Int. J. Geomech.

Int. J. Geomech., 04016136


Measured (Moraci and Recalcati 2006)
Calculated (Large-strain mode)
Calculated (Small-strain mode)
150 150
σv= 10 kPa σv= 25 kPa

Pullout load (kN/m)

Pullout load (kN/m)


100 E = 30 MPa 100
c = 1 kPa
Ri = 0.67
Ks,gg = 20 MN/m/m
50 50

0 0
0 25 50 75 100 0 25 50 75 100
Downloaded from ascelibrary.org by University of Toronto on 11/10/16. Copyright ASCE. For personal use only; all rights reserved.

Geogrid front end displacement (mm) Geogrid front end displacement (mm)
(a) (b)

150 150
σv= 50 kPa σv= 100 kPa

Pullout load (kN/m)


Pullout load (kN/m)

100 100

50 50

0 0
0 25 50 75 100 0 25 50 75 100

Geogrid front end displacement (mm) Geogrid front end displacement (mm)
(c) (d)

Fig. 6. Influence of FLAC small-strain and large-strain mode on calculated pullout load-displacement responses

150
Reinforcement ultimate tensile strength:
Tult = 118 kN/m

Large-strain mode
Pullout load (kN/m)

100 E = 30 MPa
c = 1 kPa
Ri = 0.67
Ks,gg = 20 MN/m/m

50
Measured data from Moraci and Recalcati (2006)

Measured (Peak)
Measured (Large-displacement)
Calculated
0
0 20 40 60 80 100

Applied vertical stress (kPa)

Fig. 7. Measured and calculated peak and large-displacement pullout loads at different applied vertical pressures

© ASCE 04016136-7 Int. J. Geomech.

Int. J. Geomech., 04016136


noted that the model did not converge if the beam node at x = 3 interaction between the geotextile and surrounding soil was simu-
m was not defined using the grid number even when the model lated using grouted cable elements. For Case 2, the use of cable
was executed in large-strain mode. The supporting soil bottom elements for the geotextile required an interface between the
boundary was fixed in x- and y-directions. The left and right gravel layer and the supporting soil. This interface was assigned
sides of the model were fixed in the x-direction. The soil– the same properties as the lower interface between the geotextile
geosynthetic interactions were modeled using interfaces beam elements and the supporting soil in Case 1.
between the gravel layer and geotextile and between the sup- The interface normal stiffness kn = 50–1,000 MPa/m and shear
porting soil and geotextile. For Case 1, the beam elements stiffness ks = 1–50 MPa/m were investigated for Case 1 using beam
were assigned a zero moment of inertia. elements to model the geotextile. For Case 2, the cable grout stiff-
2. Case 2 is similar to Case 1 with the exceptions that the geotextile ness was calculated using Ks,gt = ks  Pgt, where Pgt = cross-section
was modeled using 200 FLAC cable elements (all nodes were perimeter of the geotextile with an out-of-plane width of 1 m (Pgt =
defined using x- and y-coordinates) and only one interface was 2 m for the plane strain condition). For example, if the interface
generated between the gravel layer and the supporting soil. The shear stiffness between the geotextile and surrounding soil from
Case 1 is ks = 10 MPa/m, the cable grout stiffness for Case 2 is cal-
Downloaded from ascelibrary.org by University of Toronto on 11/10/16. Copyright ASCE. For personal use only; all rights reserved.

Surcharge pressure culated as Ks,gt = ks  Pgt = 10  2 = 20 MN/m/m. Table 4 summa-


q1 rizes the interface property values used for the two cases. The beam
q2 Gravel layer
and cable element property values for the geotextile are summarized
Fixed end 0.5 m in Table 5. Details regarding the use of beam and cable elements to
0.5 m
(Geotexle) Geotexle model soil reinforcement materials in the program FLAC can be
Void
(with air bags inially) found in the paper by Yu et al. (2015b).
4.0 m Each FLAC model was solved to force equilibrium before acti-
Supporng soil vating the void. After creating the void below the geotextile, each

Table 4. Interface Property Values for Geotextile over a Void Example


2.0 m 2.0 m
(a) Interface parameter Value
10.0 m
Normal stiffness [kn (MPa/m)] 50–1,000
Surcharge pressure
q1 Shear stiffness [ks (MPa/m)] 1–50
q2 Gravel layer Gravel layer-geotextile
0.5 m Friction angle [ f upper (°)] 30
Fixed end 0.5 m Dilation angle [ c upper (°)] 0
(Geotexle) Geotexle
Void Cohesiona [cupper (kPa)] 0.6–3
Supporting soil-geotextile
4.0 m
Supporng soil Friction angle [ f lower (°)] 25
Dilation angle [ c lower (°)] 0
(b) Cohesionb [clower (kPa)] 1.2–6
a
cupper = Ri  cg with Ri = 0.6.
Fig. 8. Example showing a horizontal geotextile subjected to unbal- b
clower = Ri  cs with Ri = 0.6.
anced vertical stresses on upper and lower sides of the geotextile:
(a) geotextile-reinforced soil layer over void; (b) FLAC numerical grid
and boundary conditions
Table 5. Beam and Cable Element Property Values for Geotextile over a
Void Example
Table 3. Gravel Layer and Supporting Soil Property Values for
Property Value
Geotextile over a Void Example
a
Tensile stiffness [Jgt (kN/m)] 1,100
Property Value
Young’s modulusb [Egt (MPa)] 550
Gravel layer Cross-sectional area [Agt (m2/m)] 2  10−3
Unit weight [ g g (kN/m3)] 17.0 Exposed perimeter [Pgt (m)] 2
Young’s modulus [Eg (MPa)] 10–100 Tensile yield strengtha [Tf,gt (kN/m)] 125
Poisson’s ratio ( g) 0.3 Beam elements (Case 1)
Friction angle [ f g (°)] 44 Moment of inertia [Igt (m4)] 0
Dilation anglea [ c g (°)] 14 Cable elements (Case 2)
Cohesion [cg (kPa)] 1–5 Grout stiffnessc [Ks,gt (MN/m/m)] 2–100
Supporting soil Grout frictional resistanced [ f gt (°)] 30
Unit weight [ g s (kN/m3)] 16.0 Grout cohesive strengthe [Cgt (kN/m)] 1.2–6
Young’s modulus [Es (MPa)] 30 a
Out-of-plane width of 1 m [for tensile yield strength a unit of force was
Poisson’s ratio ( s) 0.3
used in the FLAC manual (Itasca 2011)].
Friction angle [ f s (°)] 38 b
Egt = Jgt/Agt.
Dilation angleb [ c s (°)] 8 c
Ks,gt = ks  Pgt with ks defined in Table 4 and Pgt = 2 m for the plane strain
Cohesion [cs (kPa)] 2–10
condition.
a
c g = f g −30°. d
f gt = f upper with f upper defined in Table 4.
b
c s = f s −30°. e
Cgt = cupper  Pgt with cupper defined in Table 4.

© ASCE 04016136-8 Int. J. Geomech.

Int. J. Geomech., 04016136


model was solved to reach force equilibrium again. Then the distrib- increased with decreasing supporting soil cohesion. For example,
uted surcharge q1 = 16 kPa was applied to the left-side anchorage the calculated geotextile vertical displacements at x = 3 m were
zone and solved to force equilibrium, and the procedure was about 0.27, 0.23, and 0.21 m for the case of supporting soil cohesion
repeated for surcharge increments q2 = 5.3 and 12.4 kPa located cs = 2, 5 and 10 kPa, respectively. The best agreement between the
directly over the void. measured and calculated data was obtained for cs = 5 kPa.
As in the previous pullout test example cases, no attempt was
made to model the strain and time-dependent behavior of the geotex- Influence of the Soil-Geosynthetic Interface
tile in this study because the necessary test data were not available. Normal Stiffness
The influence of the soil-geosynthetic interface normal stiffness on
Influence of the Gravel Layer Young’s modulus
geotextile vertical displacements is shown in Fig. 12. The results
Fig. 9 shows the influence of the gravel layer Young’s modulus on the indicate that the interface normal stiffness kn = 50–1,000 MPa/m
geotextile vertical displacements after removing the air bags and cre- had a minor influence on geotextile vertical displacements when the
ating a void below the geotextile. The numerical results demonstrate interface shear stiffness was fixed at ks = 10 MPa/m. The use of
Downloaded from ascelibrary.org by University of Toronto on 11/10/16. Copyright ASCE. For personal use only; all rights reserved.

that the value of the gravel layer Young’s modulus has negligible interface normal stiffness kn = 50 or 100 MPa/m resulted in slightly
influence on the geotextile vertical displacements. The calculated geo- better agreement between the calculated and measured data.
textile vertical displacements for the gravel layer Young’s modulus,
Eg, ranging from 10 to 100 MPa were in good agreement with the Influence of the Soil-Geosynthetic Interface
measured data from Villard and Briançon (2008). Shear Stiffness
Fig. 13 shows the influence of the soil-geosynthetic interface
Influence of the Gravel Layer Cohesive
shear stiffness on geotextile vertical displacements. Values of
Strength Component
soil-geosynthetic interface shear stiffness ks = 1–50 MPa/m had
The influence of the gravel layer cohesive strength component on the a minor influence on the geotextile vertical displacements when
geotextile vertical displacements is shown in Fig. 10. Increasing the the interface normal stiffness was fixed at kn = 100 MPa/m. A
gravel layer cohesion decreased the geotextile vertical displacements. slightly better agreement between the calculated and measured
For example, the calculated geotextile vertical displacement at x = results was obtained using interface shear stiffness in the range
3 m was about 0.18 m for cg = 5 kPa and increased to about 0.21 and ks = 1–10 MPa/m.
0.23 m when cg = 3 and 1 kPa, respectively. The calculated geotextile
vertical displacements using the gravel cohesion value cg = 1 kPa Influence of Soil-Geosynthetic Interface Normal and
were in good agreement with measured data. Shear Stiffness
For soil-geosynthetic interaction problems, the interface stiffness
Influence of Supporting the Soil Cohesive
can be related to the stiffness of the surrounding soils (Itasca 2011;
Strength Component
Yu et al. 2015b). An increase in the interface normal stiffness gener-
Fig. 11 shows the influence of the supporting soil cohesive strength ally results in an increase in the interface shear stiffness. Fig. 14
component on geotextile vertical displacements. The numerical shows the influence of soil-geosynthetic interface normal and shear
results show that the calculated geotextile vertical displacements stiffness with fixed ratio kn = 10 ks on geotextile vertical

Horizontal distance from geotextile left end, x (m)


1 2 3 4 5
0.0
Void edge
Geotextile vertical displacement (m)

Case 1
0.1
Large-strain mode
cg = 1 kPa
cs = 5 kPa
ks = 10 MPa/m
0.2 kn = 100 MPa/m

Measured data from


Villard and Briançon (2008)
0.3
Measured (after removing air bags)
Calculated (Eg = 10 MPa)
Calculated (Eg = 40 MPa)
Calculated (Eg = 100 MPa)
0.4

Fig. 9. Influence of the gravel layer Young’s modulus on geotextile vertical displacements after creating a void (2 m ≤ x ≤ 4 m)

© ASCE 04016136-9 Int. J. Geomech.

Int. J. Geomech., 04016136


Horizontal distance from geotextile left end, x (m)
1 2 3 4 5
0.0
Void edge

Geotextile vertical displacement (m)


0.1 Case 1
Large-strain mode
Eg = 40 MPa
cs = 5 kPa
ks = 10 MPa/m
0.2 kn = 100 MPa/m
Downloaded from ascelibrary.org by University of Toronto on 11/10/16. Copyright ASCE. For personal use only; all rights reserved.

Measured data from


Villard and Briançon (2008)
0.3
Measured (after removing air bags)
Calculated (cg = 1 kPa)
Calculated (cg = 3 kPa)
Calculated (cg = 5 kPa)
0.4

Fig. 10. Influence of the gravel layer cohesive strength component on the geotextile vertical displacements after creating a void (2 m ≤ x ≤ 4 m)

Horizontal distance from geotextile left end, x (m)


1 2 3 4 5
0.0
Void edge
Geotextile vertical displacement (m)

0.1 Case 1
Large-strain mode
Eg = 40 MPa
cg = 1 kPa
ks = 10 MPa/m
0.2 kn = 100 MPa/m

Measured data from


Villard and Briançon (2008)
0.3
Measured (after removing air bags)
Calculated (cs = 2 kPa)
Calculated (cs = 5 kPa)
Calculated (cs = 10 kPa)
0.4

Fig. 11. Influence of the supporting soil cohesive strength component on geotextile vertical displacements after creating a void (2 m ≤ x ≤ 4 m)

displacements. The results show that the influence of interface nor- the FLAC small-strain mode resulted in failure of the model to
mal stiffness kn = 50–500 MPa/m with fixed ratio kn = 10 ks on the converge. After 10,000 calculation steps, the maximum calcu-
geotextile vertical displacements is negligible. lated vertical displacement of the geotextile in Fig. 15(a) was
about 1.51 m. When the model was run in the FLAC large-
strain mode, force equilibrium was reached with a maximum
Influence of Small-Strain and Large-Strain
calculated vertical displacement of 0.23 m for the geotextile
Mode Options
[Fig. 15(b)].
Fig. 15 shows the calculated displacement vectors for Case 1 In some applications there may be the temptation to use the
with beam elements used to model the geotextile. The use of small-strain mode rather than large-strain mode in FLAC (or

© ASCE 04016136-10 Int. J. Geomech.

Int. J. Geomech., 04016136


Horizontal distance from geotextile left end, x (m)
1 2 3 4 5
0.0
Void edge

Geotextile vertical displacement (m)


Case 1
0.1 Large-strain mode
Eg = 40 MPa
cg = 1 kPa
cs = 5 kPa
ks = 10 MPa/m
0.2
Downloaded from ascelibrary.org by University of Toronto on 11/10/16. Copyright ASCE. For personal use only; all rights reserved.

Measured data from


Villard and Briançon (2008)

Measured (after removing air bags)


0.3 Calculated (kn = 50 MPa/m)
Calculated (kn = 100 MPa/m)
Calculated (kn = 500 MPa/m)
Calculated (kn = 1000 MPa/m)
0.4

Fig. 12. Influence of soil-geotextile interface normal stiffness on geotextile vertical displacements after creating a void (2 m ≤ x ≤ 4 m)

Horizontal distance from geotextile left end, x (m)


1 2 3 4 5
0.0
Void edge
Geotextile vertical displacement (m)

Case 1
0.1 Large-strain mode
Eg = 40 MPa
cg = 1 kPa
cs = 5 kPa
kn = 100 MPa/m
0.2

Measured data from


Villard and Briançon (2008)

Measured (after removing air bags)


0.3 Calculated (ks = 1 MPa/m)
Calculated (ks = 5 MPa/m)
Calculated (ks = 10 MPa/m)
Calculated (ks = 50 MPa/m)
0.4

Fig. 13. Influence of soil-geotextile interface shear stiffness on geotextile vertical displacements after creating a void (2 m ≤ x ≤ 4 m)

equivalently without mesh updating in FEM models) because of Measured and Calculated Geotextile
the reduced computation time to reach convergence. However, Vertical Displacements
for situations in which there is large geometrical nonlinearity
the two approaches can give very different numerical outcomes. Based on the results of parametric analyses presented in the previ-
An example is the case of a geosynthetic-reinforced soil ous sections, the good agreement between the measured and calcu-
embankment over soft or weak foundations (Benmebarek et al. lated geotextile vertical displacements can be obtained using the
2015; Cao et al. 2016; Xu et al. 2016; Yu et al. 2016c). The gravel Young’s modulus Eg = 40 MPa, gravel cohesive strength
example of a geotextile over a void in the current study can be component cg = 1 kPa, supporting soil cohesive strength component
understood to be an extreme case for geosynthetic-reinforced cs = 5 kPa, interface normal stiffness kn = 100 MPa/m, and interface
embankments. shear stiffness ks = 10 MPa/m. Fig. 16 shows the deformed

© ASCE 04016136-11 Int. J. Geomech.

Int. J. Geomech., 04016136


Horizontal distance from geotextile left end, x (m)
1 2 3 4 5
0.0
Void edge

Geotextile vertical displacement (m)


Case 1
0.1 Large-strain mode
Eg = 40 MPa
cg = 1 kPa
cs = 5 kPa
0.2
Downloaded from ascelibrary.org by University of Toronto on 11/10/16. Copyright ASCE. For personal use only; all rights reserved.

Measured data from


Villard and Briançon (2008)
0.3
Measured (after removing air bags)
Calculated (kn = 10ks = 50 MPa/m)
Calculated (kn = 10ks = 100 MPa/m)
Calculated (kn = 10ks = 500 MPa/m)
0.4

Fig. 14. Influence of soil–geotextile interface normal and shear stiffness on geotextile vertical displacements after creating a void (2 m ≤ x ≤ 4 m)

geotextile shape after forming a void below the geotextile. The


measured data and calculated results using a FEM code capable of
modeling large displacement are taken from Villard and Briançon
(2008). The FLAC Case 1 model with beam elements for the geo-
textile provided better prediction than the FLAC Case 2 with cable
elements and may be judged to give an improved prediction when
compared with the FEM result.

Measured and Calculated Geotextile Strains

Fig. 17 shows the calculated geotextile strains after removing


the air bags to create the underlying void. Also shown in Fig. 17
(a) are the available measured data from the physical test. The nu-
merical results indicate that the calculated geotextile strains
from Villard and Briançon (2008) gave maximum geotextile
strains at the left and right edges of the void (i.e., at x = 2 and
4 m in Fig. 17), whereas FLAC Cases 1 and 2 predicted the maxi-
mum geotextile strains above the middle of the void. The calcu-
lated strains using FLAC Case 1 (beam elements for the geotex-
tile) agree well with the physical data measured shortly after
removing the supporting air bags. The numerical results from
FLAC Case 2 (cable elements for the geotextile) are in good
agreement with the measured geotextile strains measured
5 months after forming the void. A coupled FEM and discrete-
element method (FEM-DEM) code used to simulate the same
physical test gave the same numerical outcome with the maxi-
(b) mum geotextile strain located above the middle of the void
(Villard et al. 2009).
Fig. 15. Calculated displacement vectors with beam elements for geo-
Measured and predicted geotextile strains are plotted in Fig.
textile reinforcement (Case 1): (a) FLAC small-strain mode (final force
18(a) for the case of the geotextile anchorage zone between x = 0
equilibrium is not reached, and the maximum vertical displacement of
and 2 m surcharged to q1 = 16 kPa and q2 = 5.3 kPa applied to the
the geotextile is about 1.51 m after 10,000 steps); (b) FLAC large-strain
gravel surface over the void (between x = 2 and 4 m). As expected,
mode (final force equilibrium is reached, and the maximum vertical dis-
the strains are larger than those plotted in Fig. 17. Increasing the sur-
placement of the geotextile is about 0.23 m) (Note: Vectors are drawn
to the same scale)
charge from q2 = 5.3–12.4 kPa increased the geotextile strains even
further, as shown in Fig. 18(b).

© ASCE 04016136-12 Int. J. Geomech.

Int. J. Geomech., 04016136


Horizontal distance from geotextile left end, x (m)
1 2 3 4 5
0.0
Void edge

Geotextile vertical displacement (m) 0.1


Large-strain mode

0.2
Downloaded from ascelibrary.org by University of Toronto on 11/10/16. Copyright ASCE. For personal use only; all rights reserved.

Measured data and calculated FEM results


from Villard and Briançon (2008)
0.3
Measured (after removing air bags)
Calculated (FEM)
Calculated (FLAC, Case 1)
Calculated (FLAC, Case 2)
0.4

Fig. 16. Measured and calculated geotextile vertical displacements after creating a void (2 m ≤ x ≤ 4 m)

Horizontal distance from geotextile left end, x (m)


0 1 2 3 4 5 6 7 8 9 10
0.0

Void edge
0.5

1.0
Geotextile strain (%)

Large-strain mode
1.5

2.0
Measured data and calculated FEM results
2.5 from Villard and Briançon (2008)

3.0 Measured (shortly after removing air bags)


Measured (5 months after removing air bags)
Calculated (FEM)
3.5
Calculated (FLAC, Case 1)
Calculated (FLAC, Case 2)
4.0

Fig. 17. Measured and calculated geotextile strains after removing the air bags to create an underlying void (2 m ≤ x ≤ 4 m)

From a practical point of view, the three numerical models for FEM-DEM code), whereas the FEM results from Villard and
the surcharge cases in Figs. 17 and 18 performed well in predicting Briançon (2008) predicted the largest strains at the edge of the void.
the maximum geotextile strains after forming the void below the These differences in geotextile strain distributions above the void
geotextile. However, there are differences in the shape of the geo- can be attributed to the limitation of the FEM model used by Villard
textile strain distributions. For example, FLAC Cases 1 and 2 pre- and Briançon (2008), which cannot capture the load transfer mecha-
dicted the largest strains at the middle of the void (which is also the nism between the soil and geotextile at locations in which the bot-
same result reported by Villard et al. 2009 using their coupled tom of the geotextile is unsupported.

© ASCE 04016136-13 Int. J. Geomech.

Int. J. Geomech., 04016136


Horizontal distance from geotextile left end, x (m)
0 1 2 3 4 5 6 7 8 9 10
0.0
Void edge
0.5

Geotextile strain (%) 1.0 Large-strain mode

1.5 q1 = 16 kPa
q2 = 5.3 kPa
2.0
Downloaded from ascelibrary.org by University of Toronto on 11/10/16. Copyright ASCE. For personal use only; all rights reserved.

2.5 Measured data and calculated FEM results


from Villard and Briançon (2008)

3.0
Measured
Calculated (FEM)
3.5
Calculated (FLAC, Case 1)
Calculated (FLAC, Case 2)
4.0
(a)

Horizontal distance from geotextile left end, x (m)


0 1 2 3 4 5 6 7 8 9 10
0.0
Void edge
0.5

1.0
Large-strain mode
Geotextile strain (%)

1.5 q1 = 16 kPa
q2 = 12.4 kPa
2.0

2.5 Measured data and calculated FEM results


from Villard and Briançon (2008)

3.0
Measured
Calculated (FEM)
3.5
Calculated (FLAC, Case 1)
Calculated (FLAC, Case 2)
4.0
(b)

Fig. 18. Measured and calculated geotextile strains for surcharge pressure over geotextile anchor zone (0 ≤ x ≤ 2 m) q1 = 16 kPa, and surcharge pres-
sure directly above the void (2 m ≤ x ≤ 4 m) is (a) q2 = 5.3 kPa and (b) q2 = 12.4 kPa

Conclusions program FLAC will depend on the choice of soil and geosynthetic
constitutive models selected, which may be problematic for param-
The FEM and FDM are attractive numerical techniques to model eter values that are not available from original physical test
the performance of soil–geosynthetic interaction problems, such as documentation.
geosynthetic/steel pullout tests, geosynthetic-reinforced soil walls This paper reports the modeling details and the sensitivity
and embankments, and geosynthetic-reinforced layers over voids. and accuracy of numerical results using a range of assumed soil
However, the accuracy of numerical models using the FDM and interface property values for two different geosynthetic-

© ASCE 04016136-14 Int. J. Geomech.

Int. J. Geomech., 04016136


reinforced soil interaction examples: (1) horizontal pullout of a Notation
reinforcing element (geogrid) in the anchorage zone of a reinforced
wall simulated by a pullout test and (2) a geotextile-reinforced soil The following symbols are used in this paper:
layer over a void. Based on the results reported in this paper, the hg ¼ thickness of gravel layer over a void (m);
following conclusions can be summarized: J ¼ secant stiffness of geogrid (N/m);
• For granular materials, such as sand and gravel, the use of the
q, q1, q2 ¼ surcharge pressure (Pa);
soil Young’s modulus between 10 and 100 MPa had negligible
influence on the geosynthetic responses for the examples Ri ¼ strength reduction factor (dimensionless);
examined and all other conditions unchanged. Tult ¼ ultimate tensile strength (N/m);
• A small cohesion (e.g., 1 kPa) is recommended when modeling x, y ¼ Cartesian coordinates (m);
the granular materials in soil–geosynthetic interaction problems s v ¼ vertical stress at elevation of the geogrid speci-
to ensure numerical stability at the unconfined soil boundaries. men (Pa);
• The interface between geosynthetic and surrounding soils can be
f ¼ plane strain peak friction angle for pullout test
modeled by using normal stiffness kn = 100 MPa/m and shear
Downloaded from ascelibrary.org by University of Toronto on 11/10/16. Copyright ASCE. For personal use only; all rights reserved.

model (°); and


stiffness ks = 10 MPa/m. The latter value corresponds to the grout
stiffness Ks = 20 MN/m/m when using FLAC cable elements f ds ¼ peak friction angle from direct shear box test (°).
with the geosynthetic continuous in the out-of-plane direction.
• The shear strength of soil-geosynthetic interfaces can be
References
obtained by applying a strength reduction factor Ri = 0.67 to
the surrounding soil if the measured shear strength of soil- AASHTO. (2014). LRFD bridge design specifications, 7th Ed., Washington,
geosynthetic interfaces is not reported from the physical DC.
tests. This is also the default value recommended by Alam, M. J. I., Lo, S. R., and Karim, M. R. (2014). “Pull-out behaviour of
AASHTO (2014). steel grid soil reinforcement embedded in silty sand.” Comput. Geotech.,
• The numerical results from the geogrid pullout test simulations 56, 216–226.
in which the geogrid specimen deformed only in the horizontal Allen, T., and Bathurst, R. (2014a). “Design and performance of 6.3-m high
direction showed that the choice of FLAC small-strain or block-faced geogrid wall designed using the K-stiffness method.” J.
large-strain mode had a minor quantitative influence on pullout Geotech. Geoenviron. Eng., 10.1061/(ASCE)GT.1943-5606.0001013,
load-displacement response. Numerical results were in reason- 04013016.
able practical agreement with measured values, particularly Allen, T. M., and Bathurst, R. J. (2014b). “Performance of an 11 m high
block-faced geogrid wall designed using the K-stiffness method.” Can.
for tests with small confining pressures, at small initial dis-
Geotech. J., 51(1), 16–29.
placements (say less than 20 mm), and at peak pullout load. Bathurst, R. J., and Ezzein, F. M. (2016). “Geogrid pullout load–strain
• The numerical analyses of an unsupported geotextile-reinforced
behaviour and modelling using a transparent granular soil.” Geosynth.
soil layer over a void demonstrated that force equilibrium (nu- Int., 23(4), 271–286.
merical convergence) was not achieved in numerical models Benmebarek, S., Berrabah, F., and Benmebarek, N. (2015). “Effect of geo-
using the FLAC small-strain mode. This is because the horizon- synthetic reinforced embankment on locally weak zones by numerical
tal structure (beam) elements using the FLAC small-strain mode approach.” Comput. Geotech., 65, 115–125.
can only generate horizontal axial loads; hence, force balance Blanc, M., Thorel, L., Girout, R., and Almeida, M. (2014). “Geosynthetic
cannot be satisfied between horizontal loads and the vertical reinforcement of a granular load transfer platform above rigid inclu-
stresses acting on the horizontal structure elements. However, sions: Comparison between centrifuge testing and analytical model-
force equilibrium was achieved using the large-strain mode in ling.” Geosynth. Int., 21(1), 37–52.
FLAC. The calculated geotextile deformations and strains using Briançon, L., Nancey, A., and Villard, P. (2005). “Development of
Geodetect: A new warning system for the survey of reinforced earth
the FLAC large-strain mode were in good agreement with the
constructions.” Stud. Geotech. Mech., 27, 21–32.
measured data from a physical test. Briançon, L., and Villard, P. (2008). “Design of geosynthetic-reinforced
More advanced models for the soil, geosynthetic, and soil- platforms spanning localized sinkholes.” Geotext. Geomembr., 26(5),
geosynthetic interaction could be used in numerical modeling of 416–428.
the type investigated in this study. Examples include nonlinear Bridle, R. J., and Jenner, C. G. (1997). “Polymer geogrids for bridging min-
elastic-plastic models with strain softening for the soil and time- ing voids.” Geosynth. Int., 4(1), 33–50.
dependent and strain-dependent geosynthetic stiffness models. BSI (British Standards Institution). (2016). “Code of practice for strength-
However, the relatively simple models used in this study are easy ened/reinforced soil and other fills.” BS8006, London.
to understand, need a minimum number of input parameters, and Budhu, M. (2010). Soil mechanics and foundations, 3rd Ed., Wiley,
at the same time result in satisfactory agreement with physical Hoboken, NJ.
test results when the latter are available. Finally, this study used Cao, W. Z., Zheng, J. J., Zhang, J., and Zhang, R. J. (2016). “Field test of a
geogrid reinforced and floating pile-supported embankment.” Geosynth.
the program FLAC to perform numerical simulations. The inter-
Int., 23(5), 348–361.
ested reader is directed to the paper by Yu et al. (2015b) who
Damians, I.. Bathurst, R., Josa, A., and Lloret, A. (2015). “Numerical analy-
investigated the influence of choice of the numerical approach sis of an instrumented steel reinforced soil wall.” Int. J. Geomech., 10
[i.e., the FDM (program FLAC) and the FEM] on numerical soil- .1061/(ASCE)GM.1943-5622.0000394, 04014037.
structure interaction results. Ezzein, F., and Bathurst, R. J. (2014). “A new approach to evaluate soil-
geosynthetic interaction using a novel pullout test apparatus and
transparent granular soil.” Geotext. Geomembr., 42(3), 246–255.
Acknowledgments Ezzein, F., Bathurst, R. J., and Kongkitkul, W. (2015). “Non-linear load-
strain modeling of polypropylene geogrids during constant rate-of-strain
The work reported in this paper was supported by a grant from the loading.” Polym. Eng. Sci., 55(7), 1617–1627.
Natural Sciences and Engineering Research Council of Canada Fannin, R. J., and Raju, D. M. (1993). “On the pullout resistance of geosyn-
(NSERC). thetics.” Can. Geotech. J., 30(3), 409–417.

© ASCE 04016136-15 Int. J. Geomech.

Int. J. Geomech., 04016136


Feng, S. J., and Lu, S. F. (2015). “Deformation analysis of a geosynthetic Villard, P., and Briançon, L. (2008). “Design of geosynthetic reinforce-
material subjected to two adjacent voids.” Geotext. Geomembr., 43(4), ments for platforms subjected to localized sinkholes.” Can. Geotech. J.,
317–331. 45(2), 196–209.
FLAC [Computer software]. Itasca Consulting Group, Inc., Minneapolis. Villard, P., Chevalier, B., Le Hello, B., and Combe, G. (2009). “Coupling
Galve, J. P., Gutierrez, F., Guerrero, J., Alonso, J., and Diego, I. (2012). between finite and discrete element methods for the modelling of earth
“Optimizing the application of geosynthetics to roads in sinkhole-prone structures reinforced by geosynthetic.” Comput. Geotech., 36(5), 709–717.
areas on the basis of hazard models and cost-benefit analyses.” Geotext. Villard, P., Gourc, J. P., and Giraud, H. (2000). “A geosynthetic reinforce-
Geomembr., 34, 80–92. ment solution to prevent the formation of localized sinkholes.” Can.
Giroud, J. P., Bonaparte, R., Beech, J. F., and Gross, B. A. (1990). “Design Geotech. J., 37(5), 987–999.
of soil layer-geosynthetic systems overlying voids.” Geotext. Walters, D. L., Allen, T. M., and Bathurst, R. J. (2002). “Conversion of geo-
Geomembr., 9(1), 11–50. synthetic strain to load using reinforcement stiffness.” Geosynth. Int.,
Huang, B., and Bathurst, R. J. (2009). “Evaluation of soil-geogrid pullout 9(5–6), 483–523.
models using a statistical approach.” Geotech. Test. J., 32(6), 489–504. Wang, F., Han, J., Miao, L. C., and Bhandari, A. (2009). “Numerical analy-
Itasca. (2011). FLAC: Fast Lagrangian analysis of continua 7.0 user’s sis of geosynthetic bridged and drilled shafts-supported embankments
manual, Itasca Consulting Group, Inc., Minneapolis. over large sinkholes.” Geosynth. Int., 16(6), 408–419.
Downloaded from ascelibrary.org by University of Toronto on 11/10/16. Copyright ASCE. For personal use only; all rights reserved.

Jones, C. J. F. P., and Cooper, A. H. (2005). “Road construction over voids Wang, M. C., Feng, Y. X., and Jao, M. (1996). “Stability of geosynthetic-
caused by active gypsum dissolution, with an example from Ripon, reinforced soil above a cavity.” Geotext. Geomembr., 14(2), 95–109.
North Yorkshire, England.” Environ. Geol., 48(3), 384–394. Wilson-Fahmy, R. F., and Koerner, R. M. (1993). “Finite element model-
Karpurapu, R. G., and Bathurst, R. J. (1995). “Behaviour of geosynthetic re- ling of soil-geogrid interaction with application to the behavior of geo-
inforced soil retaining walls using the finite element method.” Comput. grids in a pullout loading condition.” Geotext. Geomembr., 12(5),
Geotech., 17(3), 279–299. 479–501.
Kongkitkul, W., Chantachot, T., and Tatsuoka, F. (2014). “Simulation of geo- Xu, C., Song, S., and Han, J. (2016). “Scaled model tests on influence fac-
synthetic load–strain–time behaviour by the non-linear three-component tors of full geosynthetic-reinforced pile-supported embankments.”
model.” Geosynth. Int., 21(4), 244–255. Geosynth. Int., 23(2), 140–153.
Kulhawy, F. H., and Mayne, P. W. (1990). “Manual on estimating soil prop- Yogarajah, I., and Yeo, K. C. (1994). “Finite element modelling of pullout
erties for foundation design.” Rep. EL-6800, Electric Power Research tests with load and strain measurements.” Geotext. Geomembr., 13(1),
Institute, Palo Alto, CA. 43–54.
Moraci, N., and Recalcati, P. (2006). “Factors affecting the pullout behavior Yu, Y., Bathurst, R., and Allen, T. (2016a). “Numerical modelling of the
of extruded geogrids embedded in a compacted granular soil.” Geotext. SR-18 geogrid reinforced modular block retaining walls.” J.
Geomembr., 24(4), 220–242. Geotech. Geoenviron. Eng., 10.1061/(ASCE)GT.1943-5606.0001438,
Ochiai, H., Otani, J., Hayashic, S., and Hirai, T. (1996). “The pull-out resist- 04016003.
ance of geogrids in reinforced soil.” Geotext. Geomembr., 14(1), 19–42. Yu, Y., Bathurst, R. J., Allen, T. M., and Nelson, R. (2016b). “Physical
Perkins, S. W., and Edens, M. Q. (2003). “Finite element modeling of a geo- and numerical modelling of a geogrid reinforced incremental concrete
synthetic pullout test.” Geotech. Geol. Eng., 21, 357–375. panel retaining wall.” Can. Geotech. J., in press.
Ponomaryov, A., and Zolotozubov, D. (2014). “Several approaches for the Yu, Y., Bathurst, R. J., and Damians, I. P. (2016c). “Modified unit cell
design of reinforced bases on karst areas.” Geotext. Geomembr., 42(1), approach for modelling geosynthetic-reinforced column-supported
48–51. embankments.” Geotext. Geomembr., 44(3), 332–343.
Rouse, P. C., Fannin, R. J., and Taiebat, M. (2014). “Sand strength for back- Yu, Y., Bathurst, R. J., and Miyata, Y. (2015a). “Numerical analysis of a
analysis of pull-out tests at large displacement.” Geotechnique, 64(4), mechanically stabilized earth wall reinforced with steel strips.” Soils
320–324. Found., 55(3), 536–547.
Rowe, R. K., and Skinner, G. D. (2001). “Numerical analysis of geosyn- Yu, Y., Damians, I. P., and Bathurst, R. J. (2015b). “Influence of choice of
thetic reinforced retaining wall constructed on a layered soil founda- FLAC and PLAXIS interface models on reinforced soil-structure interac-
tion.” Geotext. Geomembr., 19(7), 387–412. tions.” Comput. Geotech., 65, 164–174.

© ASCE 04016136-16 Int. J. Geomech.

Int. J. Geomech., 04016136

You might also like