Download as pdf or txt
Download as pdf or txt
You are on page 1of 13

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/264856761

Discrete Element Modeling for Granular Materials

Article in Electronic Journal of Geotechnical Engineering · January 2012

CITATIONS READS
11 4,035

2 authors, including:

Xinghua Wang
Central South University
50 PUBLICATIONS 665 CITATIONS

SEE PROFILE

All content following this page was uploaded by Xinghua Wang on 30 August 2015.

The user has requested enhancement of the downloaded file.


Discrete Element Modeling for Granular
Materials

Guangcheng Yang
Ph.D. candidate, School of Civil Engineering, Central South University,
Changsha, 410075, China
e-mail: yanggc23@gmail.com

Xinghua Wang
Professor, School of Civil Engineering, Central South University, Changsha,
410075, China

ABSTRACT
Conventional numerical methods, such as finite element method (FEM) and finite difference
method (FDM), are grid-based and macro-scale modeling approaches which are incapable of
capturing microstructure phenomenon in soils. Discrete element method (DEM) models the
system at the micromechanical level, which is a desirable technique to investigate microscale
behavior in soils. The application of DEM to the simulation for mechanical behavior of
granular materials is presented in this paper. The nonlinear Hertz contact model is
implemented in the DEM code. A direct shear test is modeled and compared to experimental
data, showing a good agreement between numerical simulation and lab observation. The
validated model is used to simulate triaxial shear test and dynamic torsional shear test. The
effects of shear strain along with loading frequency and particle friction on damping ratio are
investigated..
KEYWORDS: Discrete element method, granular material, torsional shear test.

INTRODUCTION
Granular materials are the second most important material in nature (the first one being
water). The study of how granular materials deform or flow from micromechanical perspective is
of interest in applications as diverse as geotechnical engineering, pharmaceutical industry, energy
production and other industrial processes. Computational modeling can provide insights into
mechanisms of granular material behavior, playing an important role in understanding natural
processes involved and predicting the outcomes of various events in specific conditions.
Conventional computational methods, such as FEM and FDM (Zienkiewicz and Taylor. 2005;
Clausen et al. 2006), are grid-based methods and based on macroscale continuum framework,
which fail to capture the micromechanical behavior in granular materials (Cleary et al. 2004).
Liquefaction, caused by the decrease of effective stress in saturated soil under dynamic loading,
can result in significant loss of lives and properties. However, the process of the development of

- 2463 -
Vol. 17 [2012], Bund. Q 2464

dynamic loading induced pore pressure is difficult to be predicted using macroscale approaches
due to the limitation of constitutive models (Robertson and Wride, 1998). Another example of the
importance that micro-mechanism needs to be investigated is the initiation and development of
landslides (Cleary et al., 2004). Analytical methods and numerical techniques based on
continuum framework have been widely applied to the evaluation of landslides (Ham et al. 2009).
Nevertheless, the conventional methods play a limited role in helping understand under what
circumstances the landslide may occur and how far the landslide will run from the initiation
position, which are crucial considerations in risk assessment and protective-strategy formation.
Another drawback associated with grid based methods is that they have difficulties in the
simulation of problems involving extremely large deformation (Liu et al. 2007), which are
commonly encountered with in geotechnical engineering. In FEM, the interpolation functions can
produce highly accurate solutions if the meshes are of regular shape. The extremely large
deformation may significantly degrade and distort the computational grids, leading to inaccurate
solutions and in most cases the failure of computation. Various methods have been developed to
overcome this drawback in FEM, among which, the adaptive remeshing method is a relatively
mature one (Khoei and Lewis, 1999). However, this method requires the identification of updated
deformed shape and mapping field variables to the deformed shape (Bui et al, 2008), which is
obviously problematic when a complicated material model is used.
As a purely Lagrangian approach, DEM tracks materials by using a set of discrete particles
instead of grids, avoiding the numerical problems associated with mesh distortion. In addition,
DEM models a system on a micro-mechanical level, capable of capturing micro-mechanisms in
geotechnical material behavior. Originally developed by Cundell (1977) to solve rock mechanics
problems in mining engineering, the application of DEM has been found in many other research
areas and industries as diverse as pharmaceutical industry, agriculture, geotechnical engineering
and energy production (Cleary 2004). DEM represents materials as discrete particles which match
real material particle sizes and preferably their shapes (Hopkins et al. 1991). The computation
algorithm involves the detection of particle collisions and calculation of contact forces. Once the
net force on each particle is determined, the equations of motion are solved for the motion of
particle system (Barker 1994).
The developed 3D model in this paper is validated against a direct shear test. A good
agreement between numerical simulation and experimental data is observed. The validated model
is then used to simulate triaxial shear test and dynamic torsional shear test. The effects of shear
strain along with loading frequency and particle friction on damping ratio are numerically
investigated.

METHODOLOGY

Calculation algorithm
DEM simulates the mechanical behavior of a system comprised of a collection of arbitrarily
shaped and arranged particles. The particles displace independent of one another, and only
interact at the contacts and interfaces between particles. The computation involves the application
of Newton’s second law to particles and a force-displacement law at contacts. Newton’s second
law is used to determine the motion of each particle arising from the contact and body forces
acting on it, while the force-displacement law is used to update the contact forces arising from the
relative motion at each contact. The calculation cycle requires the repeated application of the law
of motion to each particle and a force-displacement law to each contact. The calculation cycle is
Vol.
V 17 [201
12], Bund. Q 246
65

illustrrated in Figure 1. At the start of each h time step, tthe set of conntacts is deteected from thhe
knowwn particle possitions. The foorce-displacem ment law is thhen applied too each contacct to update thhe
contacct forces baseed on the relattive motion or
o overlappingg distance bettween the twoo entities at thhe
contacct and the contact constituutive model. Next,
N the law
w of motion iss applied to eeach particle tto
updatte its velocity and positionn based on thee resultant forrce and momeent arising froom the contact
forcess and body foorces acting onn the particle..

Fig
gure 1: Calcculation cyclle for DEM

Force-dis
F splaceme
ent law
The
T force-disp placement law w relates thee relative dissplacement bbetween two particles at a
contacct to the conttact force actiing on the parrticles. As shhown in Figurre 2, the forcee-displacemennt
law applies at a coontact and can n be describeed in terms o f a contact pooint, , lyinng on a contact
plane that is defineed by a unit normal
n vectorr, . The conntact point iss within the innterpenetratioon
volumme of the two o particles. The
T normal veector on a coontact betweeen two particcles is directeed
along the line betw ween ball cen nters. The contact force iss decomposedd into a norm mal componennt
g in the direcction of the normal vector,, and a shearr component acting in the contact planne.
acting
The force-displacement law relates r thesee two compoonents of foorce to the correspondinng
components of thee relative disp placement viaa the normal and shear stifffnesses at thhe contact. Thhe
momeent arising froom the normaal and shear fo orce can be caalculated by:

= × (1)

wheree F is the force vector, R is the contaact position vvector with rrespect to thee center of thhe
particcle, and indiccates the indeex of contact points.
p
Vol.
V 17 [201
12], Bund. Q 246
66

Figu
ure 2: Contaact law betweeen two partticles (after IItasca, 2003))

After
A the calcuulation of forrce and moment on each ccontact pointt on a particlee, the resultannt
force and rotation moment
m on thhe particle aree determined..

Law of motio
on
The
T motion off a single rigid d particle is determined
d moment vectorrs
byy the resultannt force and m
acting
g on it, descrribed in term ms of the tran nslational mootion of a pooint in the paarticle and thhe
rotatio
onal motion of o the particlee. The translaational motionn of the centeer of mass is described witth
respecct to its positiion, , velociity, , and accceleration, . The rotatioonal motion off the particle is
descriibed with resp pect to its an
ngular velocityy, , and anngular accelerration, . Thhe equations oof
motioon can be expressed as two vector equations: oone relates tthe resultantt force to thhe
translational motio on; and the other relates th he resultant m
moment to thhe rotational motion, whicch
can bee expressed as:
a

= (22)

= (33)

wheree and are


a the resultaant force and
d rotational m
moment on paarticle , respectively; is
the mass,
m and is the moment of
o inertia with
h respect to thhe mass centeer of the particcle.
The
T leap-frog time integraation method
d is employeed to updatee the particlee velocity annd
position.

MODEL
L VALID
DATION
The
T developed d numerical model
m is valid
dated by a dirrect shear testt conducted bby Yan (20100).
The particle
p assemmblage is prepared by rando omly generatiing particles iin a containerr with 10 cm iin
lengthh, 10 cm in width
w and 3.71
1 cm in heigh ht, as shown inn Figure 3. A vertical forcce is applied oon
the to
op of shear bo ox, which remmains unchang ged during exxperiment. Thhe soil specim men undergoees
static shearing with h bottom shear box movin ng horizontallly in a very smmall velocityy of 3 mm/minn.
The particle
p frictio
on is assumedd to be 0.7 in this
t simulatioon. Figure 4 shhows the com mparison of thhe
experriment and sim mulation resuults for shear stresses undeer various noormal forces. The numerical
resultts indicate thaat the shear strength
s of sooil increases as the increaase of normaal force, whicch
Vol.
V 17 [201
12], Bund. Q 246
67

follow
ws the trendd observed in n experimennts. The com mputational sshear stress vs. horizonttal
displaacement curvees show a goo
od agreementt with that obttained from eexperiments.

Figuree 3: Numerical model off the direct sshear test (affter Yan, 20110)

Figuree 4: Shear stress and horrizontal displlacement (affter Yan, 20110)

APPLICA
A ATION OF
O NUM MERICAL
L MODE
EL TO CO
OMPLEX
X
SIMU
ULATION
NS

Triaxia
al shear ttest
In
n DEM, con ntact models play an im mportant role in the accuuracy of sim mulations. Thhe
parammeters in these contact models,m such as normal aand shear stiiffness, frictiion coefficiennt
betweeen particles, are based on n micro-mech hanical perspeective, whichh are difficultt to be directlly
obtain ned from geo otechnical exp periments. Th herefore, the parameter caalibration is nnecessary prioor
to sim
mulation of larrger problemss containing the t same mateerials as the ssample. Triaxial shear test is
a commmonly used method
m to dettermine strength parameteers in a macroo-fashion. A DDEM modelinng
of triaaxial shear tesst is carried out
o to match th he micro-scalle parameterss with the maccro-scale onees.
Vol.
V 17 [201
12], Bund. Q 246
68

As sh
hown in Figu ure 5, the num
merical particcle assemblagge containingg 2886 DEM particles witth
diameeter ranging from
fr 1.5 to 2 mm
m confined in a 40 mm bby 20 mm cyllinder containner.

Figuree 5: Particle assemblage for triaxial ttest

The
T confining g pressure iss set to 1 Mp pa and remaains unchangged during nuumerical tesst.
As a Lagrangian approach,
a it iss difficult to apply a presssure boundaryy on discrete particles. Thhis
type ofo boundary is i accounted for by applying a movingg wall on the vertical sidee of the modeel.
The velocity
v of thee moving walll is determineed by the conntact force bettween particlees and the waall
in ord der to produce desirable confining prressure. A hoorizontal walll boundary, moving alonng
verticcal direction, is applied onn the top to produce
p the ddeviatoric strress in the triiaxial test. Thhe
volummetric strain and axial strrain under ellastic load iss shown in F Figure 6, shoowing a lineaar
relatio
onship. As th he increase off axial deviatooric stress, thhe soil samplee undergoes a shear failurre,
indicaating a nonelaastic deformaation. Figure 7 shows that the axial devviatoric stresss changes nonn-
linearrly with the axial
a strain unnder the elasstic-plastic loaad. The horizzontal red linne is confininng
pressuure, which iss constant du uring simulattion. The tri--axial test siimulation inddicates a goood
repressentation of granular
g materrial behavior under shear.

Figure 6: Volumetric
V strain vs. axiaal strain for elastic load (After Itascaa, 2003)
Vol.
V 17 [201
12], Bund. Q 246
69

Figuree 7: Axial deeviatoric streess vs. axial strain (Afteer Itasca, 20003)

Dyn
namic torsional sshear tes
st
Soil materials exhibit intern
nal friction beetween particcles as the maaterials underrgo non-elasttic
deform
mation. The energy dissip pated in non-elastic deform mation can bbe determinedd by hysteressis
loop under cyclic loads. The cyclic
c torsion
nal shear testt is a most ccommonly ussed method tto
mine damping
determ g in geotechnnical materialss. Figure 8 shhows a typicall hysteresis looop. and
are th
he area of hyysteresis loopp and triangle, respectivelly. The damp mping ratio is calculated bby
equatiion (4) shown
n below.

Figure 8: Typical
T hysteeresis loop (aafter Kramerr, 1996)
Vol.
V 17 [201
12], Bund. Q 247
70

= (44)
4

A DEM model is developed to simulatee dynamic torrsional shear test. As show wn in Figure 9,
the particle
p assemmblage, contaained in a cylinder
c conttainer, consissts of 8015 particles witth
diameeter ranging from
f 0.75 mm m to 4.0 mm m. The dynam mic torsional shear test is carried out bby
applying a harmon nic rotation att the top of asssemblage, ass shown in Fiigure 10. Thee shear strain is
calcullated using th
he following equation:
e

= ∙ / (55)

wheree , and are a the rotatioonal angle, cy


ylinder radiuss and specimeen height, resspectively. Thhe
confin
ning pressuree is set at 7ee5 Pa, remain ning unchangged during thhe test. Tablee 1 shows thhe
materrial propertiess and loading parameters in
n the simulatiions.

Fig
gure 9: Particle assemblaage for dynaamic torsionaal shear test

= ∙ ( ∙ )

= ∙ /

Fixed at the bottom

Figure 10: Soil co


olumn subjeccted to a torssional excitaation at the ttop
Vol. 17 [2012], Bund. Q 2471

Table 1: Model parameters


Porosity 0.4
Sample size × =5 × 17.5
Density 2700 /
Harmonic Excitation function = ∙ ( )
=2 ∙
= ∙ 2 ∙ ∙ cos( )
where is the amplitude in torsion

The effects of shear strain along with loading frequency and particle friction coefficient on
damping ratio are numerically investigated.

Effect of shear strain


As the soil is plastically deformed, the energy dissipation is caused by internal particle
friction as well as particle rearrangement. Therefore, the damping ratio increases as shear strain
increases (Kramer, 1996). This behavior is simulated by applying harmonic torsion with different
amplitudes at the top of the soil column, leaving all other parameters constant. As shown in
Figure 11, higher torsional shear strain results in larger hysteresis loop area indicating higher
damping ratio. The computed damping ratio for torsional strain of 0.06% is 0.063, whereas the
damping is 0.29 for torsional strain of 0.35%.

Effect of particle friction coefficient


The simulations are conducted for various particle friction coefficients. Figure 12 presents the
variation of damping ratio with particle coefficient. It is shown that the damping ratio decreases
as the particle friction increases, which is opposite to what our intuition may indicate. The energy
dissipation in coarse sand is caused by particle rearrangement, whereas higher particle friction
limits the movement of particles. As a consequence, the damping ratio for sand with higher
friction is lower.

Effect of torsional frequency


The influence of loading frequency on damping ratio is shown in Table 2. It is indicated that
the damping ratio decreases as the frequency increases at low loading frequency, which is
consistent with the experimental observation from Shibuya et al (1995). The impacts of loading
frequency on damping ratio at medium and high frequency are not investigated in our simulation
since higher frequency requires extremely small time step which is computationally demanding.
Vol.
V 17 [201
12], Bund. Q 247
72

(a) Hysterresis loop fo


or torsional sshear strain 00.35%

(b) Hysterresis loop fo


or torsional sshear strain 00.06%

Figure 11:
1 Damping
g under diffeerent shear sstrain
Vol. 17 [2012], Bund. Q 2473

0.08

Damping ratio 0.06

0.04

0.02

0.00
0.2 0.3 0.4 0.5 0.6 0.7 0.8
Particle friction coefficient

Figure 12: Influence of particle friction coefficient on damping ratio

Table 2: Effect of torsional frequency on damping ratio

Frequency ( ) Strain ( ) Damping ratio


200 0.35% 0.29
50 0.35% 0.31

CONCLUSIONS
The application of DEM to the simulation for mechanical behavior of granular materials is
presented in this paper. A direct shear test is modeled and compared to experimental data,
showing a good agreement between numerical simulation and lab observation. The validated
model is used to simulate triaxial shear test and dynamic torsional shear test. The effects of shear
strain along with loading frequency and particle friction on damping ratio are investigated. The
numerical results show that larger shear strain results in higher damping ratio, which is consistent
with previous experimental observation. Furthermore, both loading frequency and particle friction
have impacts on damping ratio of granular materials. However, the effects of loading frequency
are not as significant as our intuition may indicate. The most unexpected results are the damping
ratio decreases as the particle friction increases for coarse sand. As a result, the damping ratio is
not an intrinsic material property. On the contrary, it depends on strain magnitude as well as
loading frequency, particle shape and confining pressure.
The numerical results suggest that DEM could be a powerful method to investigate micro-
scale behavior in granular materials.
Vol. 17 [2012], Bund. Q 2474

REFERENCES
1. Barker, G. C. (1994) “Computer simulations of granular materials.” Granular matter: an
interdisciplinary approach (ed. A. Mehta). Springer.
2. Bui, H.H., Fukgawa, R., Sako, K. and Ohno, S. (2008) “Lagrangian mesh-free particle
method (SPH) for large deformation and post-failure flows of geomaterial using elasticplastic
soil constitutive model.” International Journal for Numerical and Analytical Methods in
Geomechanics, 32 (12), 1537–1570.
3. Clausen, J., Damkidle, L. and Andersen, L. (2006) “Efficient return algorithms for associated
plasticity with multiple yield planes.” International Journal for Numerical Methods in
Engineering, 66, 1036–1059.
4. Cleary, P.W. and Parkash, M. (2004) “Discrete-Element Modelling: Methods and
Applications in the Environmental Sciences.” Philosophical Transactions: Mathematical,
Physical and Engineering Sciences, 362(1822), 2003-2030.
5. Cleary, P. W. (2004) “Large scale industrial DEM modeling.” Engineering Computation. 21,
169–204.
6. Cundall, P. and Strack, O. (1979) “A discrete numerical model for granular assemblies.”
Geotechnique, 29, 47–65.
7. Ham, G., Rohn, J., Meier, T. and Czurda, K. (2009) “Finite element simulation of a slow
moving natural slope in the upper-austrain alps using a visco-hypoplastic constitutive model.”
Geomorphology, 103, 136–142.
8. Hopkins, M., Hibler, W. and Flato, G. (1991), “On the simulation of the sea ice ridging
process.” Journal of Geophysics Research, 96, 4809–4820.
9. Itasca (2003). Particle Flow Code, PFC3D, release 3.0. Itasca Consulting Group, Inc.,
Minneapolis, Minnesota.
10. Khoei, A.R. and Lewis, R.W. (1999) “Adaptive finite element remeshing in a large
deformation analysis of metal powder forming.” International Journal for Numerical Methods
in Engineering, 45, 801-820.
11. Kramer, S.L. (1996), Geotechnical Earthquake Engineering, Prentice Hall.
12. Liu, J., Chen, H., Ewing, R. and Qin, G. (2007) “An efficient algorithm for characteristic
tracking on two dimensional triangular meshes.” Computing, 80, 121–136.
13. Robertson, P.K. and Wride, C.E. (1998) “Evaluating Cyclic Liquefaction Potential using the
cone penetration test.” Canadian Geotechnical Journal, Ottawa, 35(5), 442-459.
14. Yan, Y. and Ji, S. (2010) “Discrete element modeling of direct shear tests for a granular
material”, International Journal for Numerical and Analytical Methods in Geomechanics, 34,
978-990.
15. Zienkiewicz and Taylor (2005) “The Finite Element Method for Solids and Structural
Mechanics (5 ed.)” Elsevier

© 2012 ejge

View publication stats

You might also like