You are on page 1of 6

S

oybeans and many other agricultural products are


often handled as bulk materials. The dynamic
behavior of bulk materials such as soybeans must
be understood to properly design handling and
storage equipment. A project has been started as an activity
of the National Science Foundations Engineering Research
Center at the University of Florida to develop a discrete
element model of particle flow. A discrete element model
exists for flow of elastic spheres (Walton and Braun, 1986).
However, most commercially important particles are not
spheres and their mechanical properties are not perfectly
elastic. The long term goal of the modeling effort is to
develop a model that can be used to predict flow
characteristics of irregularly shaped particles. The
coefficient of restitution is an important physical property
needed for the discrete element model. As a first step in
this process, the algorithms required to model the bouncing
behavior of a single soybean were developed and the
coefficients of restitution of soybeans for several conditions
were measured. A soybean seed is approximately an
ellipsoid with mechanical properties that are inelastic. The
coefficient of restitution characterizes the energy loss that
occurs when particles collide with each other or with wall
surfaces. This study measured the coefficient of restitution of
impacts between soybeans and glass, aluminum and acrylic
surfaces at two moisture contents.
BACKGROUND
Two methods can be used to develop a computer
simulation of the motion of particles. One is to use a
continuum model of the moving particles using a technique
such as the Finite Element Method (Bishara et al., 1981).
The other is to use a discrete model that considers the
motion of single particles (Liu et al., 1995; Walton, 1993a).
Simulation of the motion of particles using the Discrete
Element Method (DEM) is based on dynamics. An accurate
simulation requires evaluations of the interactive forces
between the particles. For particles such as soybeans,
interactive forces can be calculated using contact
mechanics (Johnson, 1985; Mindlin and Deresiewicz,
1953). However, using contact mechanics theory to directly
compute forces is not computationally practical, especially
when simulating a system with a large number of particles.
Therefore, it is important to use simplified models based on
contact mechanics to compute the contact forces between
the particles efficiently. Walton and Braun (1986) have
proposed using a semi-latched spring force model in the
normal direction of contact and a Mindlin like force model
(Mindlin and Deresiewicz, 1953) for the force in the
tangential direction of contact (Walton, 1993a, 1993b).
Both the theoretical solutions of contact mechanics and
the simplified force models require mechanical properties
of the material to calculate the forces. Some of the
mechanical properties of soybeans required for the discrete
element simulation, such as Youngs modulus, Poissons
SOYBEAN IMPACTS: EXPERIMENTS AND DYNAMIC SIMULATIONS
G. J. LoCurto, X. Zhang, V. Zakirov, R. A. Bucklin, L. Vu-Quoc, D. M. Hanes, O. R. Walton
ABSTRACT. The coefficient of restitution of soybeans impacting aluminum, glass, and acrylic was measured for drop
heights of 15.1 cm, 29.2 cm, and 51.1 cm and at moisture contents of 10.7% and 15.5% (db). Statistically significant
changes in the coefficient of restitution were produced by changes in material surface, drop height and moisture content.
Changes in moisture content from 10.7% to 15.5% (db) reduced the average coefficient of restitution from 0.69 to 0.59, a
decrease of approximately 14%. Over the range of impact velocities tested (1.7 to 3.2 m/s produced by drops from heights
from 15.1 cm to 51.1 cm) the coefficient of restitution was relatively insensitive to velocity, decreasing by less than 5%
from 0.65 to 0.62. The measured values of coefficient of restitution were used as inputs to algorithms describing the
behavior of a single bouncing soybean. The algorithm predicted reasonable results when compared to the observed
behavior of dropped soybeans. Keywords. Soybeans, Coefficient of restitution.
Article was submitted for publication in October 1996; reviewed and
approved for publication by the Structures & Environment Div. of ASAE
in April 1997.
Approved for publication as Florida Experiment Station Journal
Series No. R-05559
The authors would like to acknowledge the financial support of the
National Science Foundations Engineering Research Center on Particle
Science and Technology at the University of Florida, The National
Science Foundation (NSF) grant #EEC-94-02989, and the Industrial
Partners of the ERC.
The authors are Greg J. LoCurto, Graduate Research Assistant,
Department of Agricultural and Biological Engineering, Xiang Zhang,
Graduate Research Assistant, Department of Aerospace Engineering
Mechanics and Engineering Science, Vadim Zakirov, Graduate Research
Assistant, Department of Aerospace Engineering Mechanics and
Engineering Science, Ray A. Bucklin, ASAE Member Engineer,
Professor, Department of Agricultural and Biological Engineering, Loc
Vu-Quoc, Associate Professor, Department of Aerospace Engineering
Mechanics and Engineering Science, and Dan M. Hanes, Associate
Professor, Department of Coastal and Oceanographic Engineering,
University of Florida, Gainesville, Fla.; and Otis R. Walton, Physicist,
Lawrence Livermore National Laboratory, Livermore, Calif.
Corresponding author: Ray Bucklin, Dept. of Agricultural and
Biological Engineering, University of Florida, Gainesville, FL 32611-
0570; tel.: (352) 392-7728; fax: (352) 392-4092; e-mail:
<bucklin@agen.ufl.edu>.
Transactions of the ASAE
1997 American Society of Agricultural Engineers 0001-2351 / 97 / 4003-789 789 VOL. 40(3):789-794
ratio, and viscoelastic parameters can be obtained from the
literature (Liu et al., 1990). Those mechanical properties
not available in the literature, but necessary for computer
simulation such as the coefficient of restitution for soybean
collisions, must be measured experimentally.
The coefficient of restitution is among the important
parameters needed to predict post-collision motion of
particles. There are several different definitions for the
coefficient of restitution (Smith and Liu, 1992). Yang and
Schrock, (1994) performed drop tests of soybeans and
analyzed the images of soybean trajectories to measure the
coefficient of restitution of soybeans. Their methodology
was similar to that used in this work; however, they defined
the coefficient of restitution in a different manner. They
defined the coefficient of restitution as the ratio of the
resultant velocity after impact to the resultant velocity
before impact of the soybean in the drop test. They used
this definition for both the case when rotation of the
particle was involved and when rotation was not involved.
Therefore, their coefficient of restitution is not exactly a
measure of energy loss in a collision. The coefficient of
restitution measured in this study is defined in terms of the
square root of the total kinetic energy before and after
collisions that did not involve tangential frictional losses.
Only those soybeans that fell without significant rotation
and whose rebound trajectories were almost vertical were
used in this study. In that case, the coefficient of restitution
was defined as the ratio of the separation velocity to the
approach velocity. Therefore, the results obtained were
different from Yang and Schrock (1994), but more suitable
for the force model necessary for development of a discrete
element model. The present definition is a measure of the
energy lost during a collision, as given by the following
definition of the coefficient of restitution:
e = (KE
reb
/KE
init
)
1/2
= v/v (1)
where
KE
reb
= total kinetic energy of a particle after collision
KE
init
= total kinetic energy of a particle before
collision
v = velocity of particle before collision
v = velocity of particle after contact
e = coefficient of restitution.
SIMULATION AND MODELING
Algorithms for simulating the motion of non-spherical
particles typical of agricultural and industrial material
handling processes are being developed. Clusters of spheres
are used to model the geometrical shape of non-spherical
particles in order to simplify contact detection and to
evaluate contact forces. When simulating the motion of
soybeans, the dynamics of motion used are those associated
with ellipsoids. On the other hand, contact detection and the
contact force-displacement law are based on clusters of
spheres (Walton and Braun, 1993; Zhang et al., 1995). To
this end, the ellipsoidal shape of a soybean is approximated
by a cluster of four spheres of equal radius, with the centers
lying on a plane (fig. 1) The distance between the centers of
the spheres is adjusted so to match the measured statistical
characteristics of the geometry of soybeans.
Measurements of the particle stiffness of soybeans
indicate that the empirical force-displacement law
proposed by Walton and Braun (1986) is a good first
approximation (Walton, 1993a, 1993b). In this model, the
normal contact force, P, between two particles is calculated
by:
P = K
1
for loading (2a)
and
P = K
2
(
0
) for unloading (2b)
where
P = normal contact force between two particles
K
1
= loading normal stiffness of contact
K
2
= unloading normal stiffness of contact
= normal displacement relative to the centers of
the two particles

0
= residual displacement
This model can be represented by a partially-latching
spring system (fig. 2) consisting of two springs with
different stiffnesses and connected so that the compressing
stiffness K
1
and separating stiffness K
2
are different.
790 TRANSACTIONS OF THE ASAE
Figure 1Three-dimensional view of cluster of four spheres.
Figure 2Partially-latching spring force model.
The area between the loading and unloading paths
represents the energy lost to plastic deformation in the
normal direction during contact as shown in figure 3.
Assuming that there is a particle of mass m colliding with a
rigid plane, the normal force-displacement law is given by
equation 2 as shown in figure 3. To avoid complexity,
rotation and tangential forces were not considered.
Comparing conditions before and after the collision, the
energy balances are expressed as:
K
1

2
=
1
2mv
2
before collision (3a)
K
2
(
0
)
2
=
1
2mv
2
after collision (3b)
where v is the velocity of the particle before collision and
v is the velocity of the particle after collision. Note that at
the point of highest force of the contact shown in figure 3:
K
1
= K
2
(
0
) (4)
Substituting equation 4 into equation 3 gives the following:
v
2
/v
2
= K
1
/K
2
(5)
Thus, from definition of the coefficient of restitution
(eq. 1), the ratio of normal loading stiffness, K
1
, to normal
unloading stiffness, K
2
, can be determined by the
coefficient of restitution as (Walton, 1993a):
e = v/v = (K
1
/K
2
)
1/2
(6)
In this work, the normal loading stiffness, K
1
, was
measured directly from soybean compression tests. The
value of K
2
required to carry out the simulation was
calculated from experimentally measured values of the
coefficient of restitution, e.
It should be noted that even though this article does not
report experimental results from collisions involving rotations
and significant tangential forces, the simulation model does
include calculations of tangential friction force and its effect
on rebound behavior including rigid body rotation and
frictional sliding losses. Quantitative details of friction and
rotation effects will be the subject of another article.
By using a partially-latching spring system to model the
contact force-displacement behavior in the normal
direction and connecting the initial stiffness of contact in
the tangential direction, K
0
, with the stiffness in the normal
direction, K
1
, as described in Johnson (1985) and Mindlin
and Deresiewicz (1953), the initial tangential stiffness, K
0
,
can be calculated as:
K
0
= K
1
2(1 )/(2 ) (7)
where is Poissons ratio of soybeans from Liu et al. (1990).
Using this method, the stiffness is reduced in the tangential
direction when plastic flow occurs in the contact region.
EXPERIMENTS
The coefficient of restitution was measured using drop
tests from three heights (15.1 cm, 29.2 cm, and 51.1 cm)
onto solid blocks of either aluminum (modulus of elasticity
of 70 GPa), glass (modulus of elasticity of 72 GPa) or
acrylic (modulus of elasticity of 3 GPa) at two soybean
moisture contents. One sample of soybeans had a moisture
content of 10.7% (db) and a bulk density of 0.876 g/cm
3
.
The other sample of soybeans had a moisture content of
15.5% (db) and a bulk density of 0.850g/cm
3
. The
coefficient of restitution was calculated from the initial
height of drop (H
init
) and the height of rebound (H
reb
). To
ensure the energy was translated to the height of rebound,
soybeans which translated the energy optimumly to the
vertical direction were selected. Only soybeans that
rebounded with minimal rotation and a trajectory within
90 1.6% to the plate were selected. About 400 soybeans
were dropped for each set of conditions. Assuming no loss
of energy except during contact:
H
reb
KE
reb
rebound kinetic
energy immediately after the contact (8a)
H
init
KE
init
initial kinetic
energy just before the contact (8b)
From equation 1:
e (H
reb
/H
init
)
1/2
(KE
reb
/KE
init
)
1/2
. (9)
The apparatus used was a rigid tube fitted with a screw
mechanism so that the height of drop could be varied as
depicted in figure 4 and figure 5. The soybeans were held
in place by vacuum. The drop and rebound and the image
from a mirror were recorded using a video camera with a
speed of 1,000 frames/s. The tape was reviewed to select
soybeans which translated their energy solely to the
vertical component. A regular grid of known dimensions
was placed directly behind the falling beans to help ensure
accurate measurement of the height of rebound.
Experimental results are listed in table 1.
791 VOL. 40(3):789-794
Figure 3Normal contact force-displacement model.
A statistical analysis of the restitution data was
conducted using the General Linear Models routine of the
SAS statistical analysis package (SAS, 1988). The
difference between the mean coefficient of restitution for
the aluminum (0.65) and glass (0.64) surfaces was shown
to be statistically significant with a p-value of 0.002. This
indicates that the probability of the means being the same
is 0.002. The differences between the mean coefficient of
restitution for a moisture content of 10.7% (0.69) and a
moisture content of 15.5% (0.59) was shown to be
statistically significant with a p-value of 0.0001. The
difference between the coefficient of restitution for the
heights of 15.1 cm (0.66), 29.2 cm (0.64), and 51.1 cm
(0.63) was shown to be statistically significant between
15.1 cm and 51.1 cm with a p-value of 0.0016. The
decrease in e with larger drop height is consistent with
greater plastic deformation at higher impact velocity.
SIMULATION RESULTS
The behavior of a single bean was simulated using the
granular flow simulation code with geometric and
mechanical parameters measured experimentally and from
the literature. This simulation served as a benchmark test to
show that using the measured coefficient of restitution the
DEM model could correctly predict soybean behavior
bouncing up after being dropped. Two cases of the soybean
drop tests were simulated. The first case was a soybean
bouncing vertically from a planar base. Because there is no
rotation, the energy lost during the contact can be easily
evaluated by comparing the bounce height to initial
dropping height. In the simulation, when a soybean (cluster
of spheres) bounces without rotation, the energy loss was
calculated as:
Input: Free drop height H
init
= 29.17 cm
Coefficient of restitution e = 0.70
Result: Bounce height H
reb
= 14.15 cm
Thus:
(H
reb
/H
init
)
1/2
= 0.70 (10)
The second case is a soybean bouncing in a parabolic
trajectory with rotation around its center of mass. This case
occurs when the soybean drops and collides with the floor
plane at an angle. In this simulation, the contact angle
(the angle between the longest axis of the soybean and the
floor plane) was set to be 45. Other input parameters were
the same as for the first case. Based on these assumptions:
(K
reb
/(TE
init
) = 0.64 (11)
where TE
init
is the total potential energy of the soybean at
the initial position of the drop.
The total kinetic energy of the soybean after the
collision with the floor plane is:
TE
reb
= KE
vert
+ KE
horiz
+ KE
rot
(12)
792 TRANSACTIONS OF THE ASAE
Figure 4Apparatus for soybean drop tests.
Figure 5Apparatus for soybean drop tests (top view).
Table 1. Coefficients of restitution from drop tests
Surface Type
and Soybean
Moisture Impact No. of
Content Velocity Obser- e e e
(db) (m/s) vations Low High Avg SD
Aluminum 1.7 6 0.666 0.741 0.71 0.028
(10.7% mc) 2.4 6 0.605 0.729 0.70 0.048
3.2 7 0.671 0.711 0.69 0.005
Glass 1.7 9 0.633 0.787 0.70 0.055
(10.7% mc) 2.4 5 0.661 0.693 0.68 0.015
3.2 7 0.651 0.698 0.67 0.016
Acrylic 1.7 7 0.633 0.699 0.68 0.024
(10.7% mc) 2.4 7 0.599 0.719 0.67 0.041
3.2 6 0.638 0.696 0.66 0.025
Aluminum 1.7 8 0.592 0.666 0.62 0.022
(15.5% mc) 2.4 5 0.593 0.611 0.60 0.009
3.2 6 0.571 0.594 0.58 0.009
Glass 1.7 6 0.566 0.636 0.60 0.025
(15.5% mc) 2.4 5 0.564 0.581 0.57 0.007
3.2 6 0.559 0.591 0.57 0.011
Acrylic 1.7 8 0.532 0.681 0.59 0.047
(15.5% mc) 2.4 7 0.522 0.612 0.57 0.032
3.2 6 0.524 0.596 0.56 0.024
where
TE
reb
= total kinetic energy after collision
KE
vert
= vertical component of kinetic energy after
collision
KE
horiz
= horizontal component of kinetic energy after
collision
KE
rot
= rotational component of kinetic energy after
collision
More energy was lost during the collision than in the first
case. This is because when the soybean collides with the
plane at an angle, the tangential friction forces also produce
a horizontal component to the rebound trajectory, and
frictional microslip occurs resulting in greater energy loss.
The simulation for the second case (fig. 6) can be
compared with the results of a similar experiment recorded
by a high-speed camera (fig. 7 and fig. 8). Comparison of
these figures demonstrates that a simulation utilizing a
partially latching spring normal force model, coupled with
a tangential friction force, produces a total translational and
rotational response that is quite close to the experimental
behavior when rebound with rotation occurs.
CONCLUSIONS
The coefficient of restitution of soybeans between
aluminum, glass and acrylic was measured for drop heights
of 15.1 cm, 29.2 cm, and 51.1 cm and at moisture contents
of 10.7% and 15.5% (db). Statistically significant changes
in the coefficient of restitution were produced by changes
in material surface, drop height and moisture content.
Changes in moisture content from 10.7% to 15.5% (db)
reduced the average coefficient of restitution from 0.69 to
0.59, a decrease of approximately 14%. Over the range of
impact velocities tested (1.7 to 3.2 m/s produced by drops
from heights from 15.1 cm to 51.1 cm) the coefficient of
restitution was relatively insensitive to velocity, decreasing
by less than 5% from 0.65 to 0.62.
The measured values of coefficient of restitution were
used as inputs to algorithms describing the behavior of a
single bouncing soybean. The algorithm predicted
reasonable results when compared to the observed behavior
of dropped soybeans. The results from this study are being
incorporated into a continuing effort to develop a discrete
element model of the flow of granular materials.
REFERENCES
Bishara, A. G., S. S. El-Azazy and T. D. Huang. 1981. Practical
analysis of cylindrical farm silos based on finite element
solutions. ACI J. 78:456-462.
Johnson, K. L. 1985. 2nd Ed. Contact Mechanics. New York,
N.Y.: Cambridge University Press.
Liu, M., K. Haghighi, R. Stroshine and E. Ting. 1990. Mechanical
properties of the soybean cotyledon and failure strength of
soybean kernels. Transactions of the ASAE 33(2):559-566.
Liu, Z., S. C. Negi and J. D. Jofriet. 1995. Ahybrid FEM and
DEM model for numerical analysis of granular material flow.
ASAE Paper No 95-4450. St. Joseph, Mich.: ASAE.
Mindlin, R. D. and H. Deresiewicz. 1953. Elastic spheres in
contact under varying oblique forces. Transactions of the
ASME (Series E) 75:327-344.
SAS. 1988. SAS Version 6.03 Users Guide. Cary, N.C.: SAS
Institute Inc.
793 VOL. 40(3):789-794
Figure 6Visualization of the computer simulation.
100%
lines for position only
Figure 7Snapshot from high-speed video showing soybean dropping.
100%
lines for position only
Figure 8Snapshot from high-speed video showing soybean
rebounding after contact.
100%
lines for position only
Smith, C. E. and P. P. Liu. 1992. Coefficients of restitution. ASME
J. Appl. Mechanics 59:963-969.
Walton, O. R. 1993a. Numerical simulation of inclined chute flows
of monodisperse inelastic frictional spheres. Mechanics of
Materials 16:239-247.
Walton, O. R. 1993b. Numerical simulation of inelastic, frictional
particle-particle interactions. In Particulate Two-Phase Flow,
884-911, ed. M. C. Roco. Stoneman, Mass.: Butterworth-
Heinemann.
Walton, O. R. and R. L. Braun. 1986. Viscosity, granular-
temperature, and stress calculations for shearing assemblies of
inelastic, frictional disks. J. Rheol. 30(5):949-980.
. 1993. Simulation of rotary-drum and repose tests for
frictional spheres and rigid sphere clusters. Technical Report.
Livermore, Calif.: Lawrence Livermore National Laboratory.
Yang, Y. and M. D. Schrock. 1994. Analysis of grain kernel
rebound motion. Transactions of the ASAE 37(1):27-31.
Zhang, X., L. Vu-Quoc, O. Walton, Y. Cao and B. Vemuri. 1995.
Modeling and simulation of dry soybean flow. In SES 95
Society of Engineering Science 32nd Annual Technical
Meeting, 641-642, 29 Oct-2 Nov, New Orleans, La. Society of
Engineering Science.
794 TRANSACTIONS OF THE ASAE

You might also like