Powder Technology 158 (2005) 58 – 68

www.elsevier.com/locate/powtec

Three-dimensional discrete element simulations in hoppers and silos

Tibor J. Goda, Fritz Ebert*
Chair of Particle Technology and Fluid Mechanics, University of Kaiserslautern, Kaiserslautern D-67653, Gottlieb-Daimler-Straße Geb. 44, Germany
Available online 26 May 2005
Dedicated to Professor Otto Molerus on the occasion of his 70th birthday

Abstract

The initial fill and the subsequent discharge of containers containing granular material have been modeled using the discrete element
method (DEM). In particular, the behavior of granular material in three dimensions is accounted for. The aim is to study the distributions of
normal wall forces and pressures developed at the end of filling process (static state) and those developed during outflow from the containers
(dynamic state). A hopper, a silo with a hopper bottom, and a flat-bottomed silo have been chosen for DE analysis. The granular material is
modeled as a particle assembly consisting of mono-sized, spherical, cohesionless particles. Movement of individual particles during the
outflow caused by gravity was also studied. The simulation results obtained in the present work are realistic and show clearly the advantages
of DEM to understand the complex flow behavior of granular materials.
D 2005 Elsevier B.V. All rights reserved.

Keywords: Discrete element analysis; Granular flow; Wall pressure

1. Introduction Two-dimensional DE analyses provide satisfactory results

Containers with a variety of shapes are widely used in
industry for storage of granular materials. At the end of the
initial fill and during the subsequent discharge of a
container, pressures acting on the walls and wall stresses
can be predicted on the basis of both analytical and
numerical calculations. Recently, two numerical methods
have been widely used to study the mechanical behavior of
granular materials in containers: the finite element method
(FEM) and the discrete element method (DEM). The latter is
a powerful numerical method, in which the motion of each
individual particle is determined on the basis of all the
forces acting upon it [1]. Contrary to continuum techniques,
it simulates effects at particle level.
There are many reports about the investigations of
container flows, but most of them deal with two-dimen-
sional simulation and model particles as circular rods [2– 4]
for the purpose of simplifying mathematical treatment and
reducing computer memory requirements. In most cases,
analyses are performed by commercial simulation programs.
* Corresponding author. Tel.: +49 631 205 2415; fax: +49 631 205 3055.
E-mail address: ebert@mv.uni-kl.de (F. Ebert).
0032-5910/$ - see front matter D 2005 Elsevier B.V. All rights reserved.
doi:10.1016/j.powtec.2005.04.019
for several problems, but there are many phenomena, which
require three-dimensional simulation. This has been empha-
sized, for example, in Dürr et al. [5]. Two- and three-
dimensional discrete element simulations have been carried
out for filling and discharge under gravity from model
hoppers in Langston et al. [6]. The granular material is
modeled as an assembly of non-cohesive discs or spheres.
The results obtained are, in general, in very good agreement
with established literature empirical predictions.
The main aim of the present study is to investigate the
behavior of granular material after the initial fill (static state)
and during the discharge (dynamic state) of different
containers. For this purpose, a generally applicable three-
dimensional DE computer program, developed by the
authors, was used. The source code was written in C++
and compiled using the Visual C++ Version 6.0 compiler.
The main advantages of the program are as follows: both the
translational and the rotational motion of the particles are
considered, the numerical integration of the Newton
equations is performed by means of the efficient Feyn-
man – Newton method, and, finally, it allows users to
construct arbitrary geometry from planar walls with three
or four vertexes.
 T.J. Goda, F. Ebert / Powder Technology 158 (2005) 58 – 68 59

In a recent study, a system of 40,000 mono-sized cellulose

acetate spheres of diameter d = 6 mm was considered,
although the program developed allows us to investigate
also the effects of polydisperse particles on the flow
behavior. For numerical investigations, the following con-
tainer types have been chosen: a pyramidal hopper (Fig. 1a),
a silo with a hopper bottom (Fig. 1b), and a vertical-sided silo
with a flat bottom (Fig. 1c). The hopper has an opening angle
h = 20- and a square cross section. The silo with a hopper
bottom is 72d in height, has a square cross section with a size
of 30d, and a converging section with a hopper half angle of
h = 35-. The flat-bottomed silo is 85d in height and has a
square cross section with a size of 27d. In all the cases
studied, the size of the outlet is 8d and it has a square cross
section. With these sizes, the height of the densely packed
configuration of particles at rest is approximately the same in
each case. The only external force acting on the system is the
(a) gravity ( g = 9.81 m/s2). At the beginning of the simulations,
the containers were filled by regularly located, non-over-
lapping particles. While particles had no initial velocity in the
case of the hopper, a small random initial velocity was
assigned to each particle in the other two cases. Subsequent-
ly, the systems were allowed to settle in order to obtain
densely packed configurations before the container opening.
At the end of settling, the static packing height of the
particles is about 0.3 m in all the cases studied. The outlets
were kept closed during the settling process. Based on
preliminary calculations, a settling time of 1.5 s was specified
for each simulation. At t = 1.5 s, the outlets were opened and,
at the same time, the discharge of containers was started.

2. Simulation technique

(b) Discrete element modeling of granular materials involves

following the movement of all particles and the detection
their collisions with other particles and with their environ-
ment. In order to calculate the motion of each particle, all
the forces acting on a particle are summed and the resulting
equations of motion are numerically integrated with a fixed
time step. If the time step chosen is sufficiently small, it can
be assumed that, during a single time step, the instantaneous
motion of each particle is determined only by its immediate
neighbors with which it is in contact [1]. This assumption
greatly reduces computer memory requirement.

2.1. Equations of motion

Each particle has two types of motion: translational and

rotational. These motions can be described by Newton’s
second law given by:

(c) mv̇v ¼ ~ ðFn þ Ft Þ þ mg ð1Þ

i
Fig. 1. Geometry details of the containers studied: (a) pyramidal hopper, (b)
silo with a hopper bottom, and (c) silo with a flat bottom (d is the particle x ¼ ~M
hẋ ð2Þ
diameter). i
60 T.J. Goda, F. Ebert / Powder Technology 158 (2005) 58 – 68

where m, h, v, and x are, respectively, the mass, moment of

inertia, velocity, and angular velocity of the particle. The
forces considered are the gravitational force mg, and the
contact forces in normal ( F n) and in tangential directions
( F t). The torque M is generated by tangential contact forces
and causes the particle to rotate because the contact forces
act at the contact point and not at the particle center.
Generally, each particle is in contact with several other
particles or walls; therefore all the contact forces acting on a
particle must be summed. This operation is indicated by
index i. The dot in Eqs. (1) and (2) refers to time derivation.
The suffixes ‘‘n’’ and ‘‘t’’ mean the components corres-
ponding to the normal and tangential directions, respective-
ly. In the present study, the Feynman –Newton method with
a fixed time step of Dt = 5107 s was used to integrate the
equations of motion belonging to the translational as well as
the rotational motions. The most important advantages of
this method are fastness and great accuracy. The calculation
procedure can be summarized as follows: the position of a
particle is calculated at the end of a time step by using
velocity in the middle of a Dt interval:


1
xðt þ Dt Þ ¼ xðt Þ þ v t þ Dt Dt ð3Þ
2
where x(t) is the particle position at time t. In order to find
the velocity halfway through a Dt interval, the following
equation is used:



1 1
v t þ Dt ¼ v t  Dt þ aðt ÞDt ð4Þ
2 2
where a is the acceleration of the particle. To start the
calculation, the following equation can be used:


1 Dt
v Dt ¼ vð0Þ þ að0Þ ð5Þ
2 2
Fig. 2. Position of particles during settling in the case of hopper: (a) at t = 0 s
2.2. Modeling of contact forces (initial configuration), and (b) at t = 1.327 s.

In order to calculate contact forces between contacting Time [s]

entities, simple contact models were applied in the simula- 0 0,5 1 1,5
tions. When two particles, or a particle and a wall, are in 10
contact, the normal component of the contact force, which is
perpendicular to the contact plane, is given by the Hertz – 1
Total kinetic energy [J]

Kuwabara – Kono equation [7]:

Fn ¼  kn n3=2 1=2
n  c n nn v n ð6Þ 0,1
where k n is a stiffness parameter connecting to the elastic
properties, n n is the overlap between the contacting entities, 0,01

Table 1 0,001
Material properties and contact parameters
Acetate particle Parameters of the contact models
0,0001
Young’s modulus [N/m2] 3.2  109 k n [N/m3/2] 9  107
Poisson’s ratio 0.28 c n [kg/m1/2 s] 190
Density [kg/m3] 1319 k t [N/m] 9.45  104 0,00001
c t [kg/s] 0 Fig. 3. Total kinetic energy (translational and rotational) of the system as a
l 0.3 function of time in the case of hopper.
 T.J. Goda, F. Ebert / Powder Technology 158 (2005) 58 – 68 61

c n is a damping parameter characterizing the inelastic

behavior, and v n is the normal component of the relative
velocity defined at the contact point. Since the cohesion is
neglected in the present study, the normal contact force
calculated is always repulsive. The tangential contact force
or the friction force is given by:

Ft 4 ¼  kt nt  ct vt ð7Þ

where k t and n t are the stiffness and elongation of a

‘‘virtual’’ spring, c t is a damping parameter, and v t is the

Fig. 5. Absolute
 velocity distribution
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi at t = 2.45 s in the case of hopper
(in m/s); Vabs ¼ v2x þ v2y þ v2z .

tangential component of the relative velocity defined at

the contact point. In order to obtain a non-zero frictional
force at rest (i.e., in the case of v t = 0), a ‘‘virtual’’
tangential spring (‘‘Cundall spring’’ [1]) is located
between the contacting entities when two particles, or a
particle and a wall, start to touch one another. The total
elongation of this spring is set to zero at the initiation of
the contact and then it is integrated over the entire
collision time:
Z ti
nt ðti Þ ¼ nt ðti1 Þ þ vt dt: ð8Þ
ti1

When contacting entities are no longer in contact with

each other, the ‘‘virtual’’ spring is removed. According to
the Coulomb criterion, the maximum possible value of the

Fig. 4. Flow patterns at different simulation times in the case of hopper: (a) Fig. 6. Absolute angular velocity distribution at t = 2.45 s in the case of
at t = 1.5 s (static state at the end of the filling process), (b) at t = 2.55 s, and hopper (in s1) (the depicted quantity is derived in the same manner as in
(c) at t = 4.5 s. Fig. 5).
62 T.J. Goda, F. Ebert / Powder Technology 158 (2005) 58 – 68

Fig. 7. Distributions of normal wall forces and mean wall pressures acting on the walls of the hopper at the end of the filling process (at t = 1.5 s). The
expression normal wall force is a synonym for the normal component of particle/wall contact force.

frictional force is proportional to the normal force multiplied We assumed that the maximum impact velocity during the
by the coefficient of friction l: simulations is about v max = 1 m/s. The collision time t n
belonging to this velocity can be determined by the
Ft ¼ minðFt 4  sgnðvt ÞljFn jÞ: ð9Þ following equation:


In the case of sliding, the elongation of the tangential meff 2=5 1=5
tn ¼ 3:21 vmax ð11Þ
spring is frozen at a value of lF n/k t that is in accordance kn
with the Coulomb criterion. The material properties of the
where m eff = (m 1m 2)/(m 1+m 2) is the effective mass. The
acetate sphere used in the present study and the parameters
value of the tangential stiffness parameter is provided by the
of the contact models are listed in Table 1. In the case of
following equation:
particle/particle collision, the equations used for calculation


of the contact parameters were adopted from [7] and are kt
1=2
summarized here as follows: t¼p 1 þ mR2 =I ð12Þ
meff
4 pffiffiffiffiffiffiffiffi
kn ¼ Reff Eeff ð10Þ where I = 2/5mR 2 is the moment of inertia and t = t n. On the
3 basis of test calculations, a value of 190 kg/m1/2s was
where R eff = (R 1R 2)/(R 1+R 2) is the effective radius and E eff selected for the normal damping parameter in order to
is related to Young’s modulus E and Poisson’s ratio m of achieve e n å 0.87 [7]. e n is the coefficient of normal
contacting spheres through 1/E eff = (1  v 12)/E 1+(1  v 22)/E 2. restitution. Since parameter c t has no clear physical

Fig. 8. Mean wall pressure distribution along the walls in the case of hopper outflow (at t = 1.8 s).
 T.J. Goda, F. Ebert / Powder Technology 158 (2005) 58 – 68 63

Fig. 9. Position of particles during settling in the case of silo: (a) at t = 0 s (initial configuration), (b) at t = 0.1 s, and (c) at t = 1 s.

interpretation [7], its value was chosen to be equal to zero in
the present analysis. As a first assumption, the same
stiffness and damping parameters were used for both the
particle/particle and the particle/wall collisions in the
present study.
3. Results
3.1. Hopper
Fig. 2 shows snapshots of the settling process. From this
point, in figures showing the actual position of particles,
only half of the particle assembly is visualized in order to
obtain an insight into the interior of the granular material.
The same way of presentation is followed for all the cases
studied. At the beginning of simulation, the particles with
zero initial velocity are located regularly within the hopper.
Due to the friction and damping forces, the motion of
particles, caused by gravity during the settling, decays as the
time increases. The total kinetic energy of the system as a
function of time (Fig. 3) represents clearly this process. As it
can be seen in Fig. 3, the system has a small kinetic energy
at the so-called static state. This is a consequence of the
relatively short settling time.
The time scale to reach an equilibrium state (see Fig. 4a)
is about 1.5 s. At this time, the outlet was opened and, at the
same time, the hopper discharge was started. To indicate the
characteristic motion of the granular material, one portion of
the particles was shaded by dark gray. Fig. 4 shows
snapshots of the granular material flowing out of the hopper.
The motion of the particles can be followed easily by the
movement of the interfaces between the strata. The
interfaces deform across the whole cross section of the

Fig. 10. Flow patterns at different simulation times in the case of silo: (a) at t = 1.5 s (static state at the end of the filling process), (b) at t = 2 s, and (c) at
t = 3.25 s.
64 T.J. Goda, F. Ebert / Powder Technology 158 (2005) 58 – 68
Fig. 11. The v y velocity component at t = 2.48 s in the case of silo (in m/s).
hopper and have a characteristic ‘‘V’’ shape within a section
located parallel to the y – z plane. Distributions of the
absolute velocity and the absolute angular velocity at
t = 2.45 s are shown in Figs. 5 and 6. The gray code of
the individual particles indicates the magnitude of the
depicted quantity. Since all the particles have the same
diameter and mass, the tendencies of the translational and
rotational kinetic energy distributions were similar to Figs. 5
and 6. As anticipated, the most intensive flow appears near
and above the outlet. Near the walls, the particles move
slower than in the center region of the granular material.
In the case of static state (t = 1.5 s), the distributions of
normal wall forces and mean wall pressures along the walls
as a function of height from the bottom of the hopper are
shown in Fig. 7. The distributions of the normal wall forces
are represented on the basis of the normal component of
particle/wall contact forces. In this study, the expression
normal wall force is used as a synonym for the normal
component of particle/wall contact force. The sum of the
normal forces acting on the horizontal wall of Fig. 7 is
~F bottom = 5.17 N. For the right and the left walls of Fig. 7,
these values are as follows: ~F right = 34.17 N, ~F left = 33.65
N. As it can be concluded, mostly the inclined walls carry
the weight of the granular material. The wall pressures were
averaged on a wall segment of height dz (see Fig. 1a). In the
present study, dz is equal to 2d. For each evaluation, the
segment boundary was moved by half particle diameter in
vertical direction (z-direction). In the case of a given wall,
the height dz corresponds to a trapezoid-shaped wall
segment. In order to calculate the mean wall pressure, as a
first step, the normal forces acting on the actual wall
segment were summed. The mean wall pressure is defined
as the total force acting perpendicular to a wall segment
divided by the area of the wall segment. Since the mean wall
pressure distribution is calculated from the normal compo-
nent of the particle/wall contact forces, it shows a
fluctuation along the wall. It must be noted that the
fluctuation is strongly affected by the averaging method
applied to the pressure calculation. It can be seen clearly that
there is no significant deviation from a linear pressure
dependence in Fig. 7. As it is well known, the wall pressures
acting on the walls of a container filled with any fluid
decrease linearly with increase of height because no shear
forces are exerted between the fluid and the solid boundary
if the fluid is at rest. Contrary to fluids, in the case of real
granular material, the static wall pressures reach a constant,
height-independent value at a sufficiently large depth below
the free surface because the shear forces transfer one portion
of the weight of the granular material to the walls. This
tendency cannot be identified in Fig. 7. The reason for this
is that the system modeled is not sufficiently high to show
this tendency unambiguously; so only the linear approxi-
mation of the wall pressure distribution valid in the
neighborhood of the free surface is described.
In order to determine the wall pressure distribution
during the outflow, the pressures were averaged both on a
wall segment of height dz and over 1000 time steps. In the
case of hopper, the mean wall pressure distribution during
discharge is shown in Fig. 8. The results reported in this
study are in accord with both the two-dimensional ones of
Ristow and Herrmann [2] and the analytical calculation of
Walters [8].
3.2. Silo with a hopper bottom
In this case, a small, randomly distributed velocity is
assigned to the particles in order to obtain an irregular
densely packed configuration before the silo opening.
During settling, the total kinetic energy– time function has
a similar tendency as shown in Fig. 3. The snapshots of the
settling process, shown in Fig. 9, illustrate clearly the
motion of the particles. Fig. 9b and c show an enlarged
segment of the silo. The static, densely packed configuration
of particles, together with the striped pattern, can be seen in
Fig. 10a.
Fig. 12. Absolute angular velocity distribution at t = 2.48 s in the case of silo
(in s1).
 T.J. Goda, F. Ebert / Powder Technology 158 (2005) 58 – 68 65

Fig. 13. Distributions of normal wall forces and mean wall pressures acting on the walls of the silo at the end of the filling process (at t = 1.5 s).

At t = 1.5 s, the silo was opened to initiate the silo
discharge. The motion of particles and the deformation of
the interfaces between strata during discharge are shown in
Fig. 10. Similar to the previous case, the interfaces deform
across the whole cross section of the silo and have a
characteristic ‘‘V’’ shape within a section located parallel to
the y – z plane. Since the absolute velocity distribution has a
similar tendency as shown in Fig. 5 for the hopper flow, the
v y velocity component at t = 2.48 s is presented here (see
Fig. 11). The absolute angular velocity is represented in Fig.
12. The figures make clear that the flow is the most
intensive near and above the outlet.
In the case of silo, the mean wall pressures are plotted
horizontally for both the vertical-sided and the hopper
sections. As Fig. 13 shows, at the end of the filling process,
the wall pressures in the hopper section are slightly above
those predicted from the extension of the near linear
dependence, which was obtained for the vertical-sided
section of the silo. The sum of the normal forces acting
on the horizontal wall of Fig. 13 is ~F bottom = 5.19 N. For
the walls of vertical and hopper sections of the silo located
on the right side, the sums of normal forces are as follows:
~F right vertical = 19.49 N, ~F right hopper = 19.42 N. For the left
side of Fig. 13, these forces are ~F left vertical = 19.61 N,
~F left hopper = 19.95 N.
As it can be seen from Fig. 14, the hopper bottom has a
strong effect on the wall pressures that develop during
discharge. The mean wall pressure increases with increasing
depth measured from the free surface of the granular
material. At the transition from the vertical-sided section
to the hopper bottom, a pressure peak occurs, which is a
consequence of the sudden change in wall slope. For a
similar reason, the pressures are greater below the transition
than above it. The peak value in the hopper section is about
two times greater than the one in the vertical-sided section.
The pressure distributions obtained in this study are in
accordance with both analytical [8] and two-dimensional
numerical [9] calculations.
3.3. Flat-bottomed silo
Similar to the previous case, the regularly generated
particles (Fig. 15a) have a small, random initial velocity at
the beginning of simulation. The transition from a regular,
loosely packed configuration to an irregular, densely packed
one is illustrated in Fig. 15. The snapshots of the settling
process show the position of particles at different time steps.
In Fig. 15b– d, the lower part of the flat-bottomed silo is
enlarged. The static, densely packed configuration has a
height of 0.3 m.

Fig. 14. Mean wall pressure distribution along the walls in the case of silo outflow (at t = 1.8 s).
66 T.J. Goda, F. Ebert / Powder Technology 158 (2005) 58 – 68

Fig. 15. Position of particles during settling in the case of flat-bottomed silo: (a) at t = 0 s (initial configuration), (b) at t = 0.15 s, (c) at t = 0.2 s, and (d) at t = 1 s.

Fig. 16. Flow patterns at different simulation times in the case of flat-bottomed silo: (a) at t = 1.5 s (static state at the end of the filling process), (b) at t = 2 s, and
(c) at t = 3.25 s.

Fig. 17. Absolute velocity distribution at t = 4.47 s in the case of flat- Fig. 18. Absolute angular velocity distribution at t = 4.47 s in the case of
bottomed silo (in m/s). flat-bottomed silo (in s1).
 T.J. Goda, F. Ebert / Powder Technology 158 (2005) 58 – 68 67

Fig. 19. Distributions of normal wall forces and mean wall pressures acting on the walls of the flat-bottomed silo at the end of the filling process (at t = 1.5 s).

Comparing Fig. 16 with Figs. 4 and 10, we found that
the flow pattern is more concentrated in this case than in
the other two cases. The flow is most intensive near and
above the outlet, and the deformation of interfaces between
strata is strongly localized. From the velocity distributions,
shown in Figs. 17 and 18, similar conclusions can be
drawn.
In the case of static state, the mean wall pressure
distribution as a function of height is represented in Fig.
19. There is no significant deviation from a linear pressure
dependence due to the small size of the model measured in
vertical direction (z-direction). The sum of the normal
forces acting on the horizontal wall of Fig. 19 is
~F bottom = 48.88 N. For the right and left walls of Fig.
19, these values are as follows: ~F right = 37.82 N,
~F left = 37.55 N. As it was anticipated, in the case of
flat-bottomed silo, mostly the horizontal wall carries the
weight of the granular material if the system is at rest. Fig.
20 shows that the wall pressures developed during
discharge as a function of height. The greatest pressures
appear in the region located near the bottom of the silo.
The averaging method was the same as in the former two
cases. The results agree qualitatively with the wall pressure
distributions of Sanad et al. [4] obtained by two-
dimensional DE analysis.
4. General discussion
Comparing the motion of particles in the hoppers and
silos studied, the following conclusions may be drawn on
the basis of our simulation results.
In the cases of hopper and silo with a hopper bottom, all
the materials are in flowing condition during the discharging
process. According to this, the flow type for these cases is
mass flow. The vertical velocity of particles in the central
part of the material is larger than at the sidewalls.
Immediately above the outlet, the particles are in free fall.
In the case of silo, the uppermost layer of dark particles at
t = 3.25 s is not deformed, which means that the particles
located within this region move with the same vertical
velocity.
In the case of flat-bottomed silo, the funnel flow
phenomenon can be recognized on the basis of the flow
patterns. At the beginning of the discharging process,
similar to the previous cases, five layers of dark-colored
particles can be distinguished (see Fig. 16a). As it can be
seen in Fig. 16c, at t = 3.25 s, all of these layers are
recognizable near the sidewalls. It follows from this that in
the bottom part of the silo, near the silo walls, the particles
flow very slowly or are stagnant during the discharging
process. The regions where the velocities of the particles are

Fig. 20. Mean wall pressure distribution along the walls in the case of outflow from the flat-bottomed silo (at t = 1.8 s).
68 T.J. Goda, F. Ebert / Powder Technology 158 (2005) 58 – 68
nearly equal to zero are the so-called stagnation or dead
zones. In spite of the fact that the friction coefficient used in
the present study is rather low, the stagnation zones have
been appeared unambiguously during discharging. Appear-
ance of stagnant zones reduces the flow performance and
makes it difficult to unload the silo. In contrast to funnel
flow, the mass flow is an ideal flowing type for the
unloading of a silo. In the upper section of the silo, the
magnitudes of vertical velocities are small and almost
constant across its whole cross section at t = 3.25 s. The
uppermost layer of dark particles is not deformed, which
confirms our statement for the constant velocity profile. In
the middle part of the silo, the magnitudes of velocities near
the sidewalls are smaller than those in the central region.
The static wall pressure distributions obtained for the
silos show no significant deviation from a linear pressure
dependence. The reason for this is that the systems modeled
are not sufficiently high to show the asymptotic behavior of
wall pressures. It must be noted that an additional simulation
for a slender flat-bottomed silo, which is double as large in
height and half as large in width than the one used in the
present study, has shown already the asymptotic behavior as
expected, for example, by the Janssen formula.
Apart from advantages of our simulation program
mentioned in the Introduction, the originality of our study
lies in the following: the simulations have been carried out
in three dimensions, the hoppers and silos studied have a
square cross section, and a realistic contact model has been
used to model the contact behavior at particle level.
5. Conclusions
In the present study, three-dimensional DE simulations
have been carried out to investigate the static and dynamic
behaviors of granular material in different containers. Both
the flow patterns and the distributions of the normal wall
forces and mean wall pressures were investigated and
discussed. The pressure distributions calculated during the
discharge were in accordance with both analytical and two-
dimensional simulation results. In the case of silo flow, a
pressure peak has appeared at the transition from the
vertical-sided section to the hopper bottom. The peak value
in the hopper section was about two times greater than the
one in the vertical-sided section. Since the models were not
sufficiently high in vertical direction, the static wall
pressures did not reach a constant, height-independent
value. Instead, they showed a linear tendency.
As it has been proved in the present study, the DEM is
able to capture the container flow dynamics to a large
extent. At the same time, it must be noted that the flow
behavior of granular materials is strongly affected, for
example, by the interparticle cohesion and the particle
shape. In order to take into account these factors, the
simulation program developed will be extended in the near
future.
References
[1] P.A Cundall, O.D.L. Strack, Geotechnique 29 (1) (1979) 47 – 65.
[2] G.H. Ristow, H.J. Herrmann, Physica. A 213 (1995) 474 – 481.
[3] S. Masson, J. Martinez, Powder Technology 109 (2000) 164 – 178.
[4] A.M. Sanad, J.Y. Ooi, J.M. Holst, J.M. Rotter, Journal of Engineering
Mechanics 127 (10) (2001) 1033 – 1043.
[5] V. Dürr, G. Dau, F. Ebert, Chemie Ingenieur Technik 76 (1 – 2) (2004)
51 – 55.
[6] P.A. Langston, U. Tüzün, D.M. Heyes, Chemical Engineering Science
50 (6) (1995) 967 – 987.
[7] J. Schäfer, S. Dippel, D.E. Wolf, Journal de Physique. I 6 (5) (1996)
5 – 20.
[8] J.K. Walters, Chemical Engineering Science 28 (1973) 779 – 789.
[9] S.-C. Yang, S.-S. Hsiau, Powder Technology 120 (2001) 244 – 255.