Professional Documents
Culture Documents
Research Paper
a r t i c l e i n f o a b s t r a c t
Article history: Three-dimensional simulations of interface direct shear test between geomembrane and sand are con-
Received 28 December 2016 ducted using DEM in this study. An overlapping technique is established to reasonably simulate the
Received in revised form 11 July 2017 roughness of geomembrane surface. The simulation results agree reasonably well with the published
Accepted 24 August 2017
experimental results, such as stress-strain relationship, interface shear strength, thickness of shear band.
Available online xxxx
The method is a practical tool to investigate the micro-mechanical interactions between geomembrane
and sand, including particle displacement, void ratio and stress state, number of interparticle contacts,
Keywords:
and distribution, magnitude and orientation of contact force during the shearing process.
Surface roughness
DEM
Ó 2017 Published by Elsevier Ltd.
Interface direct shear test
Numerical simulation
Triaxial test
http://dx.doi.org/10.1016/j.compgeo.2017.08.019
0266-352X/Ó 2017 Published by Elsevier Ltd.
Please cite this article in press as: Feng S-J et al. Micro-mechanical analysis of geomembrane-sand interactions using DEM. Comput Geotech (2017), http://
dx.doi.org/10.1016/j.compgeo.2017.08.019
2 S.-J. Feng et al. / Computers and Geotechnics xxx (2017) xxx–xxx
Please cite this article in press as: Feng S-J et al. Micro-mechanical analysis of geomembrane-sand interactions using DEM. Comput Geotech (2017), http://
dx.doi.org/10.1016/j.compgeo.2017.08.019
S.-J. Feng et al. / Computers and Geotechnics xxx (2017) xxx–xxx 3
i j Table 1
ks ks
Ks ¼ i j
ð4Þ Calibrated micro-mechanical parameters of sand and geomembrane.
ks þ ks
Numerical triaxial shear test
where kn and ks are the normal stiffness and the shear stiffness of an Density of sand particles, qs (kg/m3) 2640 106
entity, respectively; the superscripts i and j denote the two entities D50 (mm) 0.74
Cu 1.69
in contact. Cc 0.99
Geomembrane is modeled with bonded particles. Similarly, the kn/r (Pa) 1 108
interaction between geomembrane and platen is also modeled kn/ks 1.5
with bonded contact model. Two entities in contact are initially Friction coefficient between sand particles, fs 0.7
Numerical tensile test
assumed to be bonded together. The bond between two entities
Density of geomembrane, qg (kg/m3) 940 106
may undergo normal and shear forces, which can be calculated Particle diameter, dg (mm) 1.53
by Eqs. (1) and (2), as well as bending and twisting moments kn/r (Pa) 1.09 105
[20]. The bonds are assumed to be unbreakable in this study. kn/ks 2.5
Therefore, the motion of cohesionless sand particles and Numerical interface direct shear test
Friction coefficient between geomembrane and sand particles, fg 0.37
geomembrane is governed by the contact stiffnesses of particles
(kn and ks) and the interparticle friction coefficient, f. In Section 3,
the micro-mechanical parameters of sand and geomembrane are
calibrated by triaxial and tensile tests, respectively, and the friction
coefficient between geomembrane and sand particles is calibrated
by interface direct shear test. The contact stiffness of platens is 600
assumed to be the same as that of the sand particles. Confining pressure
100 kPa
500 150 kPa
2.2. Discrete element modeling of geomembrane 200 kPa
Deviatoric stress (kPa) 400
Geomembranes are continuum material with smooth or rough
surface, and the surface roughness has been identified as a primary
300
factor influencing interface shear strength [22]. Uesugi and Kishida
[23] proposed a roughness parameter, normalized roughness (Rn),
to quantify the surface roughness: 200
Rmax
Rn ¼ ð5Þ 100
D50
where D50 is the mean diameter of sand particles; Rmax is the verti- 0
0 2 4 6 8 10
cal peak-to-valley distance in Fig. 1 with reference to a sand particle
with D50. The surface can be defined as ‘smooth’ for Rn < 0.02, ‘inter- Axial strain (%)
mediate’ for 0.02 Rn 0.5, and ‘rough’ for Rn > 0.5 [24]. Fig. 3. Stress-strain relationships in simulation of triaxial test.
To reproduce the surface of geomembrane, an effective method
is developed as follows. As shown in Fig. 1, every four adjacent
bonded geomembrane particles, with their centers placed at the
four corners of a square with side length of l, form a ‘‘valley”,
and the geomembrane particle vertex is the ‘‘peak”. Since sand par-
ticles cannot directly contact the ‘‘valley bottom” when placed on 30
the geomembrane, Rmax is then determined as
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi (13 %, 23.1 kN/m)
2 25
l
Rmax ¼ ðr þ RÞ ðr þ RÞ2 ð6Þ
2
Tensile stress (kN/m)
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi
R 2R
l ¼ 2r 2 1þ Rn 2R2n ¼ D50 2 1 þ Rn 2R2n ð8Þ 0
r D50 0 5 10 15 20
Strain (%)
In this way, geomembrane with various roughness can be simulated
by adjusting the value of l. Fig. 4. Stress-strain relationship in numerical tensile test of geomembrane.
Please cite this article in press as: Feng S-J et al. Micro-mechanical analysis of geomembrane-sand interactions using DEM. Comput Geotech (2017), http://
dx.doi.org/10.1016/j.compgeo.2017.08.019
4 S.-J. Feng et al. / Computers and Geotechnics xxx (2017) xxx–xxx
(a) 1000
Experimental σy=10 kPa σy=100 kPa σy=500 kPa
DEM σy=10 kPa σy=100 kPa σy=500 kPa
10
1
0 1 2 3 4 5
Shear displacement (mm)
(b) 300
o
Peak stress (Experimental) Fitting line (19.8 )
o
Peak stress (DEM) Fitting line (17.8 )
250 o
Residual stress (Experimental) Fitting line (14.4 )
o
Residual stress (DEM) Fitting line (14.0 )
200
Shear stress (kPa)
150
100
50
0
0 100 200 300 400 500 600
Normal stress (kPa)
Fig. 5. Comparison between the numerical and the experimental interface direct shear tests: (a) shear stress versus shear displacement; (b) shear stress versus normal stress.
2.3. 3D modeling of interface direct shear test The ASTM standard D5321/D5321M-14 [19] indicates that the
box length, L, or width, W, should be at least 300 mm or 15 times
The model configuration of the interface direct shear test is D85 of coarse sand, whichever is greater; and the box thickness, H,
shown in Fig. 2. The geomembrane is bonded to the bottom platen should be at least 60 mm or 6 times the maximum particle size of
to prevent relative sliding between them. The bottom platen can the coarse sand, whichever is greater. Jacobson et al. [25] con-
freely move in x direction but is confined in y and z directions. Sand ducted a series of DEM simulations of direct shear test with L/D50
particles are positioned against the geomembrane and packed ranging from 6.5 to 177, the results indicated that L/D50 33 gave
within a box consisting of five smooth and rigid platens. The four well-defined shear band. Frost et al. [12] studied the extent of
vertical platens are kept fixed during the whole shearing process, shear zone at the interface between geomembrane and sand
while the top platen can freely move in vertical direction so that through laboratory tests, and found that the shear zone extended
a prescribed normal stress, ry, can be achieved. The interface direct 2 particle diameters from the interface for smooth geomembrane
shear test is performed by moving the bottom platen with a and 4 particle diameters for slightly textured geomembrane, and
displacement rate, vx. 6 particle diameters for moderately/heavily textured geomem-
Both the sand particle size and the geomembrane thickness brane. Therefore, in this study, the width and the length of the sand
adopted in the numerical simulations are identical to those of sample are set as 35 times D50, and the height is 10 times D50. In
the real materials. On the one hand, it is unrealistic to perform sim- order to achieve quasi-static condition within a reasonable compu-
ulations using the same sample size as that in the laboratory test tational time, the particle density is scaled up by a factor of 106 as
due to the limited computation capacity. On the other hand, the suggested by Thornton [26] and Zhao et al. [27]. In this way, the
sample size should be large enough since the numerical results particle velocity and acceleration could be significantly reduced,
would be inaccurate due to the boundary effects [25]. Therefore, while the strain and stress within the sample are not affected [27].
it is critical to set a reasonable sample dimensions to balance accu- The numerical interface direct shear test consists of two stages:
racy and computation cost. sample preparation and direct shear. During the sample preparation
Please cite this article in press as: Feng S-J et al. Micro-mechanical analysis of geomembrane-sand interactions using DEM. Comput Geotech (2017), http://
dx.doi.org/10.1016/j.compgeo.2017.08.019
S.-J. Feng et al. / Computers and Geotechnics xxx (2017) xxx–xxx 5
Table 2
Simulation scheme of interface direct shear tests between geomembrane and sand. Rn: normalized roughness of geomembrane surface; ry: normal stress (kPa); e0: initial void
ratio; sp: peak shear stress (kPa); sr: residual shear stress (kPa); t: shear band thickness (D50); h: vertical deformation of the sand sample during shearing (mm).
Fig. 6. Sample deformation during the direct shear test: (a) oblique view (the color of a soil particle reflects the magnitude of particle displacement); (b) side view (the color is
used to locate soil particles). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
Please cite this article in press as: Feng S-J et al. Micro-mechanical analysis of geomembrane-sand interactions using DEM. Comput Geotech (2017), http://
dx.doi.org/10.1016/j.compgeo.2017.08.019
6 S.-J. Feng et al. / Computers and Geotechnics xxx (2017) xxx–xxx
(a)
S =0.0 mm S =5.0 mm
(b)
0.020 0.020
S=0.0 mm S=5.0 mm
0.015 0.015
0.010 0.010
0.005 0.005
0.000 0.000
0 5 10 15 20 25 0 5 10 15 20 25
Distance from left wall (mm) Distance from left wall (mm)
(c)
S =0.0 mm S =5.0 mm
Fig. 8. Force chains in the sample: (a) top view of the force chains; (b) side view of the force chains; (c) orientations of the force chains in the x-y plane.
geomembrane were 23.1 kN/m and 13%, respectively. The interface which is very important for controlling the behavior of granular
direct shear tests were conducted at normal stresses of 10, 100, materials [29]. But the sand particles are assumed to be spherical
and 500 kPa. The sand samples were sheared at a displacement and particle rotation is assumed to be free in ESyS-Particle.
rate of 1.3 mm/min to a total displacement of 12.5 mm. Nonetheless, since the shearing mostly occurs in the
geomembrane-sand interface in the interface shear tests (see Sec-
tion 4, within the shear box, particles placed at the geomembrane-
3.1. Calibration of micro-mechanical parameters of sand sand interface have the largest shear displacement during simula-
tions), the influence of this discrepancy on interface shear behavior
Simulation of triaxial test is performed using DEM to calibrate is not significant. The micro-mechanical parameters of the sand are
the micro-mechanical parameters of sand. The simulation then determined and summarized in Table 1. The stress-strain
procedures are similar to those of the sample preparation stage relationships in the simulation are shown in Fig. 3.
in Section 2.3. After the equilibrium state is reached, deviatoric
stress is applied. In the simulation, particle number highly influ-
ences the results. However, if the particle number exceeds 8000, 3.2. Calibration of micro-mechanical parameters of geomembrane
the results do not change much [26,28]. Therefore, in this study,
8324 particles are randomly generated within a cubic space Numerical tensile test is carried out to obtain the contact stiff-
(width = length = height = 14.5 mm) to perform the simulation. nesses of particles forming the smooth HDPE geomembrane (kn
D50, Cu and Cc (see Table 1) of the sand are the same as those in and ks). 1600 particles are bonded together to simulate the smooth
DeJong and Westgate [11], thus the grain-size distribution of sand geomembrane with dimensions of 10.36 10.36 mm. In order to
in the simulation is almost the same as that in the laboratory test. generate a smooth surface (Rn < 0.02), the particle distance, l, is
The simulation is conducted under three confining pressures: 100, set as 0.258 mm. A central part of the sample marked with red1
150 and 200 kPa. The friction angle of sand is adopted as the index color (see Fig. 4) with a width of 5.18 mm is used to measure the ten-
to evaluate the calibration. By trial and error, the obtained friction sile strain during the simulation, which ensures that the deformation
angle at critical state is 30.6°, while the measured value in labora- is representative.
tory test is 33.4°. The reason for this discrepancy is as follows. Real
sand particles have irregular shapes, and there exists rolling resis- 1
For interpretation of color in Fig. 4, the reader is referred to the web version of
tance when relative movement happens between two particles, this article.
Please cite this article in press as: Feng S-J et al. Micro-mechanical analysis of geomembrane-sand interactions using DEM. Comput Geotech (2017), http://
dx.doi.org/10.1016/j.compgeo.2017.08.019
S.-J. Feng et al. / Computers and Geotechnics xxx (2017) xxx–xxx 7
During the dragging process, a small displacement rate of at a very small displacement (e.g., S = 0.1 mm for the case of
107 mm/s is applied at one side of the geomembrane, while the 100 kPa), then slightly decreases, and finally sustains around a
other side is fixed. The stress-strain relationship is adopted to eval- stable (i.e., residual) value. The peak and residual shear stresses
uate the calibration. As shown in Fig. 4, the tensile stress reaches are then plotted against the respective normal stress to obtain fail-
23.1 kN/m when the tensile strain is 13%, which is consistent with ure envelopes, as shown in Fig. 5b. Fig. 5 shows that the simulation
the results in the laboratory test. The corresponding micro- results match the experimental results reasonably well, which
mechanical parameters of geomembrane are then determined indicates that the interface direct shear test can be well simulated
and summarized in Table 1. It should be noted that the relationship by the present method and further confirms that the micro-
between force and strain of the HDPE geomembrane in the labora- mechanical parameters determined from the calibration work in
tory test was nonlinear [30], which was caused by the complicated Sections 3.1 and 3.2 are reasonable. In the following part, the
properties of the geomembrane. However, since geomembrane is micro-mechanical interactions between geomembrane and sand
bonded to the bottom platen in the numerical interface direct will be comprehensively investigated using the present method.
shear test, the effect of the difference is negligible.
Simulation of interface direct shear test is conducted under The sand and geomembrane with calibrated parameters are
three normal stresses: 10, 100 and 500 kPa, the same as those in adopted to conduct the parametric study. The input parameters
the laboratory tests. The sand samples are sheared at a displace- are summarized in Table 1. The simulation scheme of the interface
ment rate of 1.3 mm/min to a total displacement of 5.0 mm. By direct shear tests is summarized in Table 2. Three groups are
trial and error, the friction coefficient between geomembrane adopted to investigate the influences of roughness of geomem-
and sand particles, fg, is determined as 0.37 (see Table 1). The com- brane surface, normal stress, and initial void ratio on the interac-
parison between the numerical and the experimental results is tions between geomembrane and sand. In each group, particle
illustrated in Fig. 5. As shown in Fig. 5a, the shear stress rapidly displacement, number of interparticle contacts, and force chain
increases with shear displacement (S) and reaches a peak value are adopted to evaluate the micro-mechanical interactions.
Fig. 9. Vertical distributions of soil particle displacement at the end of the test with different normalized roughnesses of geomembrane surface: (a) horizontal displacement;
(b) vertical displacement.
Please cite this article in press as: Feng S-J et al. Micro-mechanical analysis of geomembrane-sand interactions using DEM. Comput Geotech (2017), http://
dx.doi.org/10.1016/j.compgeo.2017.08.019
8 S.-J. Feng et al. / Computers and Geotechnics xxx (2017) xxx–xxx
4.1. Shear behavior at the geomembrane-sand interface the shear band in this case is determined as 3.0 times D50. The
method is also applied in the other cases to locate the shear band.
The case with Rn = 0.02 in Group I in Table 2 is adopted to con- The force chains in the sample are illustrated in Fig. 8. In order
cretely illustrate the shear behavior at the geomembrane-sand to measure the mean contact force, the sample is uniformly
interface. Fig. 6 presents deformation of the sample during the divided into 17 parts along the horizontal direction, the mean con-
direct shear test. Shear distortion concentrates within a narrow tact force in part i, Fi , is calculated as
zone, which is termed as shear band in this study, and no shear dis- Pmi Pnk c
jf j
tortion is observed outside the shear band. Fig. 7 shows displace- Fi ¼ Pmi c¼1 k
k¼1
ð9Þ
ment vectors of the sand particles at the end of the test. The n
k¼1 k
displacement largely concentrates within the shear band. As
where mi is the number of particles with their center located in part
shown in Fig. 7a, particles within the shear band generally move c
i; nk is the contact number of particle k; jf k j is the absolute value of
along the shear direction, most of the particles outside the shear
the contact force acting on particle k through contact c. At the
band are driven upward due to dilation. Comparing Figs. 7a and
beginning of the test (S = 0.0 mm), the distribution of force chains
Fig. 7b, it is clear that the components of the displacement vectors
exhibits a homogeneous pattern, the mean contact force uniformly
in z direction are significantly smaller than those in x and y direc-
distributes along the horizontal direction (Fig. 8b), the orientation
tions due to the confining effect in z direction (Fig. 2).
of the contact forces is nearly vertical (Fig. 8c). During the shearing
In order to quantitatively locate the shear band, the sample is
process, the shear load is transferred from the geomembrane to the
uniformly divided into 100 parts with depth, the average horizon-
sand by frictional resistance, the force chains accordingly change. At
tal displacement of the sand particles in part i, Si, is measured. If Si/
the end of the test (S = 5.0 mm), strong force chains concentrate in
St 0.01 (St = 5.0 mm is the total displacement), part i is then
the left part of the sample, and the orientation of the contact forces
viewed as part of the shear band. In this way, the thickness of
changes from nearly vertical to diagonal or even nearly horizontal.
(a)
(a) 8
5Q
Rn=0.02 (S=0.1 mm)
7
Rn=0.02 (S=5.0 mm)
Rn=0.25 (S=0.1 mm)
Distance from bottom (mm)
6
Rn=0.25 (S=5.0 mm)
5 Rn=0.50 (S=0.1 mm) 5Q
4
Rn=0.50 (S=5.0 mm)
2
5Q
0
0 50 100 150 200 250 300
(b) 8
Number of sliding contacts
Rn=0.02 (S=0.1 mm)
(b) 7
Rn=0.02 (S=5.0 mm)
1800
Total contact Sliding contact Locked contact Rn=0.25 (S=0.1 mm)
Distance from bottom (mm)
6
Number of the interface contacts
0 0
0 1 2 3 4 5 6 0.04 0.05 0.06 0.07 0.08 0.09 0.10
Shear displacement (mm) Mean contact force (N)
Fig. 10. Influence of the normalized roughness of geomembrane surface on the Fig. 11. Influence of the normalized roughness of geomembrane surface on: (a)
number of: (a) sliding contacts in sand sample with depth; (b) contacts at the force chains at the end of the interface direct shear test; (b) vertical profiles of mean
geomembrane-soil interface. contact force.
Please cite this article in press as: Feng S-J et al. Micro-mechanical analysis of geomembrane-sand interactions using DEM. Comput Geotech (2017), http://
dx.doi.org/10.1016/j.compgeo.2017.08.019
S.-J. Feng et al. / Computers and Geotechnics xxx (2017) xxx–xxx 9
4.2. Effect of roughness of geomembrane surface on interface shear between geomembrane surface and sand, which inevitably leads
behavior to larger displacement of the sand particles.
In order to investigate the evolution of microstructure induced
Group I in Table 2 is adopted to investigate the effect of by shearing, the evolution of sliding and locked contacts is studied.
roughness of geomembrane surface on the interface shear Sliding contact is defined as a contact that overcomes the friction
behavior. Since D50 is constant in this study, different normalized force, otherwise it is a locked contact. The variations of number
roughnesses of geomembrane surface, Rn, are achieved by setting of sliding contacts in the sand sample with depth under different
different particle distances, l (Eq. (8)). The vertical profiles of sand roughnesses are shown in Fig. 10a. The number of sliding contacts
particle displacement at the end of test are shown in Fig. 9. Positive is counted at shear displacements of 0.1 and 5.0 mm. Before shear-
horizontal displacement means particle moves in the same direc- ing, there is no sliding contact. Once the shearing process begins,
tion as the shear direction, and positive vertical displacement the number of sliding contacts sharply increases. At a specific dis-
means particle moves upward. The regulation is valid for other fig- placement, it increases with depth, especially in the shear band,
ures. Most particles move toward the shear direction and upward. which is consistent with the displacement results in Fig. 9. Besides,
Both the horizontal displacement and the vertical displacement the number of sliding contacts in the sand sample generally
overall increase with depth. Moreover, obvious gradient of vertical increases with increasing Rn (Fig. 10a), the reason is that the more
displacement is observed within the shear band (Fig. 9b), hence the intense deformation in the sand sample caused by larger Rn (Fig. 9)
dilation largely occurs within the shear band. With the increase of inevitably creates more sliding contacts in the sand sample. How-
Rn, both the vertical displacement and the horizontal displacement ever, the number of sliding contacts at the interface shows a con-
increase. The thickness of shear band also increases with increas- trary trend (Fig. 10b) due to the interlocking between
ing Rn, forming approximately 3.0 times D50 in the case of geomembrane surface and sand caused by larger Rn. As shown in
Rn = 0.02, and around 4.4 times D50 in the case of Rn = 0.5 (Table 2), Fig. 10b, the number of total contacts at the interface also increases
which is consistent with the results in the experimental observa- with increasing Rn, the reason is that larger Rn means larger contact
tions using similar materials [12]. The measured sp/ry and sr/ry area between geomembrane surface and sand at the interface.
and degree of dilation (h) also increase with increasing Rn (see To further investigate the effect of roughness of geomembrane
Table 2). The reason is that larger Rn means stronger interaction surface on the shear behavior, force chains at the end of the test
Fig. 12. Vertical distributions of soil particle displacement at the end of the test under different normal stresses: (a) horizontal displacement; (b) vertical displacement.
Please cite this article in press as: Feng S-J et al. Micro-mechanical analysis of geomembrane-sand interactions using DEM. Comput Geotech (2017), http://
dx.doi.org/10.1016/j.compgeo.2017.08.019
10 S.-J. Feng et al. / Computers and Geotechnics xxx (2017) xxx–xxx
(a) 8
(a)
σy=10 kPa (S =0.1 mm)
7
0 σy=10 kPa (S =5.0 mm)
0.90 0
(b) e0=0.66, σy =10 kPa 0 100 200 300 400 500 600
e0=0.66, σy =100 kPa
0.85 Number of sliding contacts
e0=0.66, σy =500 kPa
e0=0.62, σy =100 kPa (b)
e0=0.74, σy =100 kPa
0.80 1800
SSL
Void ratio
0.65 1000
800
0.60
1 10 100 1000 600
Mean principal stress (kPa)
400
Fig. 13. (a) Definition of observation zone M; (b) state diagram of sand during the
shearing process. 200
0
0 1 2 3 4 5 6
Shear displacement (mm)
are presented in Fig. 11a. The distribution of force chains for rela-
tively smooth surface (Rn = 0.02) is quite uniform, while there are Fig. 14. Influence of the normal stress on the number of: (a) sliding contacts in sand
more strong force chains in the left part of sand sample for sample with depth; (b) contacts at the geomembrane-soil interface.
Please cite this article in press as: Feng S-J et al. Micro-mechanical analysis of geomembrane-sand interactions using DEM. Comput Geotech (2017), http://
dx.doi.org/10.1016/j.compgeo.2017.08.019
S.-J. Feng et al. / Computers and Geotechnics xxx (2017) xxx–xxx 11
σy=100 kPa (S =5.0 mm) ing that the influence of e0 on the peak and residual shear stresses
5 is not significant (Table 2), which is consistent with the experimen-
σy=500 kPa (S =0.1 mm) tal observations [11]. The reason is that the concerned sand parti-
4
σy=500 kPa (S =5.0 mm) cles are spherical and the plastic deformation of geomembrane is
not considered.
As shown in Fig. 13b, the initial w is negative in the dense
3 (e0 = 0.62) and medium dense (e0 = 0.66) samples, but positive in
the loose sample (e0 = 0.74). Therefore, dilation occurs in the dense
2 (e0 = 0.62) and medium dense (e0 = 0.66) samples, while compres-
sion occurs in the loose sample (e0 = 0.74). This is also confirmed
1 by the fact that more sand particles move upward with decreasing
e0 (see Fig. 16b). In Fig. 16, both the displacement and the thickness
of shear band increase with increasing e0. The reason is that larger
0
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 e0 means sand particles can easily rearrange, leading to larger hor-
izontal displacement. It is also observed that sp and sr overall
Mean contact force (N)
increase with increasing thickness of shear band (Table 2), which
Fig. 15. Influence of the normal stress on: (a) force chains at the end of the interface further confirms that the interface shear strength is strongly linked
direct shear test; (b) vertical profiles of mean contact force. to the thickness of shear band.
The influence of e0 on the contacts is shown in Fig. 17. In
Fig. 17a, the number of sliding contacts overall increases with
increasing e0 for S = 0.1 mm, but the difference is not significant
is the porosity within the observation zone; V(p) is the volume of for S = 5 mm. As shown in Fig. 17b, the numbers of total contacts
particle (p); xi(p) and xi(c) are the locations of a particle centroid at the interface of medium dense (e0 = 0.66) and dense (e0 = 0.62)
and its contact, respectively; ni(c,p) is the unit normal vector directed samples are larger than that of loose sample (e0 = 0.74) at the early
from a particle centroid to its contact location; and Fj(c) is the force shearing stage but they are almost the same for all the three sam-
acting at contact (c). Fig. 13b is the state diagram of sand in the ples when the shear displacement is large enough. Hence, initial
observation zone during shearing. Each dot in Fig. 13b represents void ratio greatly influences the initial number of particle contacts
an e-P0 [P0 = 1/3 (r11 + r22 + r33)] state and is recorded for every but has relatively small influence on it after the sand structure is
1.0 mm of shear displacement except for the triaxial test results. disturbed by the shearing process.
As shown in Fig. 13b, the initial w is negative in all the cases with The influence of initial void ratio on force chains at the end of
e0 = 0.66, and the absolute value of w decreases with increasing nor- the test is shown in Fig. 18a. The force chains in the dense sample
mal stress. Hence dilation is observed in all the cases with e0 = 0.66 (e0 = 0.62) are relatively uniform, while there are more strong force
and the most intense dilation occurs in the case of ry = 10 kPa. chains in the left part of loose sample (e0 = 0.74), since larger sand
The variations of number of sliding contacts in the sand sample displacement in loose sample (see Fig. 16) makes more sand parti-
with depth under various normal stresses are shown in Fig. 14a. cles concentrate in the left part. The vertical profiles of mean con-
The number of sliding contacts also decreases with increasing nor- tact force are shown in Fig. 18b. An interesting phenomenon is
mal stress. Fig. 14b illustrates the evolution of different types of observed. At the displacement of 0.1 mm, the initial void ratio
contacts at the interface with shear displacement. Similarly, the makes little difference of the profiles of mean contact force for
number of sliding contacts decreases with increasing normal dense (e0 = 0.62) and medium dense (e0 = 0.66) samples, while
stress. On the contrary, the numbers of total contacts and locked loose sample (e0 = 0.74) tends to have larger mean contact force
contacts increase with increasing normal stress. In the case of in the sand. However, at the displacement of 5 mm, the profiles
Please cite this article in press as: Feng S-J et al. Micro-mechanical analysis of geomembrane-sand interactions using DEM. Comput Geotech (2017), http://
dx.doi.org/10.1016/j.compgeo.2017.08.019
12 S.-J. Feng et al. / Computers and Geotechnics xxx (2017) xxx–xxx
Fig. 16. Vertical distributions of soil particle displacement at the end of the test with different initial void ratios: (a) horizontal displacement; (b) vertical displacement.
of medium dense (e0 = 0.66) and loose (e0 = 0.74) samples differ tude and orientation of contact force during the shearing
slightly, especially in the lower part, while dense sample process. During the shearing process, the displacement lar-
(e0 = 0.62) has larger mean contact force near the geomembrane- gely concentrates within the shear band. The thickness of
sand interface but has smaller mean contact force in the sand. shear band ranges from 2.7 to 4.4 times D50 of sand and is
strongly linked to the interface shear strength.
5. Summary and conclusions (2) Surface roughness is an important factor influencing inter-
face shear behavior. With the increase of normalized rough-
Three-dimensional simulations of interface direct shear test ness (Rn), stronger interaction between geomembrane and
between geomembrane and sand are conducted using DEM in this sand is developed, leading to larger displacement of sand
study. An overlapping technique is established to reasonably sim- particles, increasing the thickness of shear band (from 3.0
ulate the roughness of geomembrane surface. The experimental to 4.4 times D50), the ratio of interface shear strength to nor-
work conducted by DeJong and Westgate [11] is adopted to cali- mal stress (from 0.365 to 0.451 and 0.286 to 0.334 at the
brate the micro-mechanical parameters of sand and geomembrane. peak and residual states, respectively) and the number of
The calibration results of the interface direct shear test indicate sliding contacts.
that interface direct shear test can be well simulated by the present (3) Normal stress is another important factor influencing the
method and the obtained micro-mechanical parameters are rea- interface shear behavior. Since higher normal stress means
sonable. Three groups of numerical simulation are carried out to stronger confining effect, the increasing normal stress leads
investigate the influences of roughness of geomembrane surface, to the decrease of horizontal and vertical displacements and
normal stress, and initial void ratio on the interface shear behavior. the number of sliding contacts, the ratio of interface shear
Some major conclusions can be drawn as follows: strength to normal stress decreases from 0.41 to 0.318 and
0.3 to 0.256 at peak and residual states, respectively, the
(1) The method is a practical tool to investigate the micro- thickness of shear band also decreases from 3.4 to 2.7 times
mechanical interactions between geomembrane and sand, D50. Under relatively high normal stress, the force chains are
including particle displacement, void ratio and stress state, strengthened with displacement although the number of
number of interparticle contacts, and distribution, magni- sliding contacts does not change much.
Please cite this article in press as: Feng S-J et al. Micro-mechanical analysis of geomembrane-sand interactions using DEM. Comput Geotech (2017), http://
dx.doi.org/10.1016/j.compgeo.2017.08.019
S.-J. Feng et al. / Computers and Geotechnics xxx (2017) xxx–xxx 13
(a) 8 (a)
e0=0.62 (S =0.1 mm) H
7
e0=0.62 (S =5.0 mm)
e0=0.66 (S =0.1 mm)
Distance from bottom (mm)
6
e0=0.66 (S =5.0 mm)
5 e0=0.74 (S =0.1 mm) H
e0=0.74 (S =5.0 mm)
4
3
H
2
0
0 50 100 150 200 250 300 (b) 8
Number of sliding contacts
e0=0.61
=0.62 (S=0.1 mm)
800 2
200
1
0
0.04 0.05 0.06 0.07 0.08 0.09 0.10
0 Mean contact force (N)
0 1 2 3 4 5 6
Shear displacement (mm) Fig. 18. Influence of the initial void ratio on: (a) force chains at the end of the
interface direct shear test; (b) vertical profiles of mean contact force.
Fig. 17. Influence of the initial void ratio on the number of: (a) sliding contacts in
sand sample with depth; (b) contacts at the geomembrane-soil interface. NCET-13-0421. The writers would like to greatly acknowledge all
these financial supports and express the most sincere gratitude.
(4) With the increase of initial void ratio (e0) from 0.62 to 0.74,
horizontal and vertical displacements increase, the thickness References
of shear band slightly increases from 2.8 to 3.2 times D50.
[1] Abdelaal FB, Rowe RK, Islam MZ. Effect of leachate composition on the long-
The influence of e0 on the peak and residual shear stresses term performance of a HDPE geomembrane. Geotext Geomembr 2014;42
is not significant. The change in state parameter reveals that (4):348–62.
dilation occurs in the dense (e0 = 0.62) and medium dense [2] Mccartney JS, Zornberg JG. Analysis of a large database of gcl-geomembrane
interface shear strength results. J Geotech Geoenviron Eng 2009;135
(e0 = 0.66) samples, while compression occurs in the loose
(2):209–23.
sample (e0 = 0.74). Moreover, the initial void ratio greatly [3] Chai JC, Miura N, Hayashi S. Large-scale tests for leachate flow through
influences the initial number of particle contacts but has rel- composite liner due to geomembrane defects. Geosynth Int 2005;12
(3):134–44.
atively small influence on it after the sand structure is dis-
[4] Mitchell JK, Seed RB, Seed HB. Kettleman hills waste landfill slope failure. I:
turbed by the shearing process. liner-system properties. J Geotech Eng 1990;116(4):647–68.
[5] Koerner RM, Soong TY. Stability assessment of ten large landfill failures. Geo-
Denver 2000:1–38.
[6] Qian XD, Shi JY, Hui L, Zhu YB. Failure interface behavior of multilayer landfill
liner system. Chin J Geotech Eng 2011;33(6):840–5 [in Chinese].
Acknowledgments [7] Feng SJ, Chen YM, Gao GY. Analysis on translational failure of landfill along the
underlying liner system. Chin J Geotech Eng 2007;29(1):20–5 [in Chinese].
[8] Fishman KL, Pal S. Further study of geomembrane/cohesive soil interface shear
Much of the work described in this paper was supported by the behavior. Geotext Geomembr 1994;13(9):571–90.
National Natural Science Foundation of China under Grant Nos. [9] Dove JE, Frost JD, Han J, Bachus RC. The influence of geomembrane surface
41222021 and 41572265, the National Basic Research Program of roughness on interface strength. Proc Geosynth 1997;2:863–76.
[10] Ling HI, Pamuk A, Dechasakulsom M, Mohri Y, Burke C. Interactions between
China (973 Program) under Grant No. 2012CB719803, the Program PVC geomembranes and compacted clays. J Geotech Geoenviron Eng 2001;127
for New Century Excellent Talents in University under Grant No. (11):950–4.
Please cite this article in press as: Feng S-J et al. Micro-mechanical analysis of geomembrane-sand interactions using DEM. Comput Geotech (2017), http://
dx.doi.org/10.1016/j.compgeo.2017.08.019
14 S.-J. Feng et al. / Computers and Geotechnics xxx (2017) xxx–xxx
[11] DeJong JT, Westgate ZJ. Role of overconsolidation on sand-geomembrane [21] Guo N, Zhao J. The signature of shear-induced anisotropy in granular media.
interface response and material damage evolution. Geotext Geomembr Comput Geotech 2013;47:1–15.
2005;23(6):486–512. [22] DeJong JT, Frost JD, Sacs M. Relating quantitative measures of surface
[12] Frost JD, Kim D, Lee SW. Microscale geomembrane-granular material roughness and hardness to geomaterial interface strength. In: Proceedings of
interactions. KSCE J Civil Eng 2012;16(1):79–92. geo-eng 2000 conference; 2000.
[13] Cundall PA, Strack OD. A discrete numerical model for granular assemblies. [23] Uesugi M, Kishida H. Frictional resistance at yield between dry sand and mild
Géotechnique 1979;29(1):47–65. steel. Soils Found 1986;26(4):139–49.
[14] Lai HJ, Zheng JJ, Zhang J, Zhang RJ, Cui L. Dem analysis of ‘‘soil” – arching within [24] Paikowsky SG, Player CM, Connors PJ. A dual interface apparatus for testing
geogrid-reinforced and unreinforced pile-supported embankments. Comput unrestricted friction of soil along solid surfaces. Geotech Test J 1995;18
Geotech 2014;61:13–23. (2):168–93.
[15] Wang Z, Jacobs F, Ziegler M. Visualization of load transfer behaviour between [25] Jacobson DE, Valdes JR, Evans TM. A numerical view into direct shear specimen
geogrid and sand using PFC2D. Geotext Geomembr 2014;42:83–90. size effects. Geotech Test J 2007;30(6):512–6.
[16] Ngo NT, Indraratna B, Rujikiatkamjorn C. DEM simulation of the behaviour of [26] Thornton C. Numerical simulations of deviatoric shear deformation of granular
geogrid stabilised ballast fouled with coal. Comput Geotech 2014;55 media. Géotechnique 2000;50(1):43–53.
(1):224–31. [27] Zhao T, Dai F, Xu NW, Liu Y, Xu Y. A composite particle model for non-spherical
[17] Cheng H, Yamamoto H, Thoeni K. Numerical study on stress states and fabric particles in DEM simulations. Granular Matter 2015;17:763–74.
anisotropies in soilbags using the DEM. Comput Geotech 2016;76:170–83. [28] Salot C, Gotteland P, Villard P. Influence of relative density on granular
[18] Kazempoor S, Noorzad A, Mahboubi A, Mirghasemi A. Numerical modeling of materials behavior: DEM simulations of triaxial tests. Granular Matter
smooth geomembrane-soil interaction shear behaviour by distinct element 2009;11(4):221–36.
method. Berlin Heidelberg: Springer; 2008. [29] Jiang MJ, Yu HS, Harris D. A novel discrete model for granular material
[19] ASTM. D5321/D5321M: standard test method for determining the shear incorporating rolling resistance. Comput Geotech 2005;32(5):340–57.
strength of soil-geosynthetic and geosynthetic-geosynthetic interfaces by [30] Koerner RM. Designing with geosynthetics. Xlibris Corporation; 2012.
direct shear. West Conshohocken, USA: ASTM International; 2014. [31] Itasca CG. PFC 3D-user manual. Minneapolis: Itasca Consulting Group; 2005.
[20] Wang Y, Alonso-Marroquin F. A finite deformation method for discrete [32] Been K, Jefferies MG. A state parameter for sands. Géotechnique 1985;35
modeling: particle rotation and parameter calibration. Granular Matter (2):99–112.
2009;11(5):331–43.
Please cite this article in press as: Feng S-J et al. Micro-mechanical analysis of geomembrane-sand interactions using DEM. Comput Geotech (2017), http://
dx.doi.org/10.1016/j.compgeo.2017.08.019