Theories and Applications of CFD-DEM Coupling Approach For Granular Flow: A Review

Archives of Computational Methods in Engineering
https://doi.org/10.1007/s11831-021-09568-9
ORIGINAL PAPER
Theories and Applications of CFD–DEM Coupling Approach

for Granular Flow: A Review
Mahmoud A. El‑Emam1,2 · Ling Zhou1 · Weidong Shi3 · Chen Han1 · Ling Bai1 · Ramesh Agarwal4
Received: 20 December 2019 / Accepted: 2 March 2021

© CIMNE, Barcelona, Spain 2021
Abstract
Bio-particulate matter includes grains, cereal crops, and biomass that are considered discrete materials with irregular size
and shape. Although the flow of these particles can behave like a continuum fluid at times, their discontinuous behavior
cannot be simulated with traditional continuum-based modeling. The Discrete Element Method (DEM), coupled with
Computational Fluid Dynamics (CFD), is considered a promising numerical method that can model discrete particles by
tracking the motion of each particle in fluid flow. DEM has been extensively used in the field of engineering, where its
application is starting to achieve the popularity in agricultural processing. While CFD has been able to simulate the complex
fluid flows with a quantitative and qualitative description of the temporal and spatial change of the flow field. This paper
reviews the recent strategies and the existing applications of the CFD–DEM coupling approach in aerodynamic systems of
bio-particles. It mainly represents four principal aspects: the definition of aerodynamic systems with its principals, modeling
of particle motion including interaction forces of particle–particle and particle–fluid in the system, CFD–DEM coupling
methodologies, and drag correlation models with theoretical developments, and the applications of aerodynamic systems
related to the agricultural field. The existing published literature indicates that CFD–DEM is a promising approach to study
the bio-particulate matter behavior immersed in fluid flow, and it could be benefiting from developing and optimizing the
device’s geometry and the operations. The main findings are discussed and summarized as a part of the review, where future
developments and challenges are highlighted.
Abbreviations di Particle diameter

Ai Projected area of particle divol Particle diameter with the same volume of the
a Separation radius between particles actual one
Cd Drag coefficient Ei Young’s modulus of particle
Cij Convection Ew Young’s modulus of wall boundary
DT,ij Turbulent diffusion Fnet Net force acting on a particle
DL,ij Molecular diffusion FijC Total contact force acting on a particle
D Domain diameter
d Normal distance to the wall Fi Fluid-particle interaction force
f
Fijn Normal contact force on a particle

Fijt Tangential contact force on a particle
* Ling Zhou
lingzhou@ujs.edu.cn Fit,T Tangential contact force at current time
Fit,(T−ΔT) Tangential contact force at previous time
1
Research Center of Fluid Machinery Engineering Fin,T Normal contact force at current time
and Technology, Jiangsu University, Zhenjiang 212013,
China Fin,(T−ΔT) Normal contact force at previous time
2
Fin,T Normal adhesive force at current time
Department of Agricultural and Biosystems Engineering, adh
Alexandria University, Shatby 21526, Egypt Fij Force production by system rotation

3
School of Mechanical Engineering, Nantong University,
Fi Interaction force of the fluid on particle
f
Nantong 226019, China
fi Gravity force
g
4
Department of Mechanical Engineering and Materials
Science, Washington University in Saint Louis, fadh Adhesive force fraction
St. Louis 63130, USA Gk Kinetic energy generation due to the velocity
13
Vol.:(0123456789)
M. A. El‑Emam et al.
Gb Kinetic energy generation due to buoyancy ΔT Timestep

Gij Buoyancy production rate of u′i u′j Ti Response time of particle
g Gravity acceleration 𝜌f Fluid density
TK Kolmogorov time scale
Te Turnover time of large eddy 𝜌f ui uj Reynolds stress tensor
′ ′
T Time 𝜌a Air density

u Root mean square of fluid velocity 𝜌i Particle density
uf Average velocity component of fluid 𝜇f Fluid turbulent viscosity
Va Air velocity 𝜇t Turbulent viscosity
VT Terminal velocity 𝜇 Friction coefficient
Vf Average velocity of the fluid 𝜇a Dynamic viscosity of the air
Vi and Vj Particles velocity before interaction 𝜇feff Effective turbulent viscosity of a fluid
Vi and Vj Particles velocity after interaction
′ ′
𝛼f Fluid volume fraction

Voli Particle volume 𝛼k and 𝛼𝜖 Inverse effective Prandtl number for k and 𝜖
Vol Volume occupied by particles and fluid 𝛼i Particle volume fraction of
YM Dilatation contribution over dissipation rate 𝜀 Coefficient of restitution
Z Particle size or geometry 𝜎k and 𝜎𝜖 Turbulent Prandtl numbers for k and 𝜖
𝜙ij,1 Slow pressure-strain Γ Surface energy
𝜙ij,2 Rapid pressure-strain 𝜖ij Dissipation
𝜙ij,3 Wall-reflection term 𝛿adh Adhesive distance
𝜙ij Pressure strain 𝜏f Viscous stress tensor
∅i Particle sphericity 𝜐f Kinematic viscosity of the fluid
𝜂 Damping ratio
𝜂 t Tangential damping ratio
S Relative
d t
tangential velocity at the contact
K1 Stokes’ shape factor dT
K2 Newton’s shape factor

ka Coefficient of aerodynamic resistance
Kun Unloading contact stiffnesses
1 Introduction
Kln Loading contact stiffnesses
l Length scale of the energy-containing eddies
1.1 Bio‑particulate Matter
m∗ Equivalent mass of particles
mi and mj Mass of particles i and j
Agriculture crops are the primary source of various bio-
N Number of particles
particulate matter include cereal grains, seeds, and biomass,
nk Component of xk normal to the wall
which are differences in shape, size, density, and other physi-
n̂c Unit vector in the normal direction
cal characteristics. These materials, which are considered the
n and t Normal and tangential coordinates
primary bio-particulate matter in this investigation, are used
p Pressure shared by two phases
as a raw material or grinded for energy (biomass) production
Pmax Maximum liquid bridge tensile force
or food to provide calories and protein resource for human
Pij Stress production rate of ui uj
′ ′
and animal diet. It sometimes contains a heterogeneous mass

Rei Particle Reynolds number of foreign materials such as sands, stones, sticks, chaff, glass
R𝜖 Dissipation of swirl and rotational effect fragments, weed, and other crop seeds. These contaminants
Re Reynolds number are mainly introduced during harvesting and post-harvesting
Ri and Rj Radii of particles i and j process [1]. Therefore, before being processed, separation
stmax Maximum relative tangential displacement and purification of cereal grains and seeds from impurities
st Tangential relative displacement at contact is an essential process in food processing and oilseed trad-
Sn Normal overlapping ing [2]. This process can be carried out by more techno-
ΔSn Change in the normal overlapping logical procedures depending on their physio-mechanical
ΔSt Change in the tangential overlapping characteristics. In this way, a part of impurities is cast away
Sn,T Normal overlap value at the current time depending on their size (width, length, thickness); others
Sn,T−ΔT Normal overlap value at the previous time are separated depending on their drag coefficient [3]. Given
Sk and S𝜖 Constant source terms for user-defined that the principle of separation and the type of impurities, a
Suser Constant source term for user-defined wide range of technical equipment and installations are used
13
Theories and Applications of CFD–DEM Coupling Approach for Granular Flow: A Review
for carrying out the separation of impurities such as aero- the aerodynamic properties seems necessary in the proper
dynamic separators, screening devices, magnetic separators, design of aerodynamic equipment. The drag coefficients of
vibrating separators, etc. [4]. Herein, the current work will a wide range of bio-particles material have been measured
focus only on the aerodynamic systems and processing used experimentally by finding the suspension velocities of the
for bio-particles such as cereal grains and seeds. particles in an airstream [19, 20]. To correlate the drag coef-
ficient, the Reynolds number has been used by grouping the
1.2 Aerodynamic Principals particles within the limits of a sphere and a cylinder shape
[21]. The need to correlate the drag coefficient is the differ-
Generally, aerodynamic devices are commonly used in the ence between the experiment and the calculated terminal
separation process of bio-particulate matter. It can be defined velocity of grains. As mentioned above, the aerodynamic
as a process of using air generated by natural or mechanical properties can be mainly described by two major aspects:
fan to remove light, chaffy, and dusty materials out of the critical or terminal velocity and drag coefficient [22]. The
bio-particles. It is considered one of the critical methods first parameter, which is considered the main representation
which can be used lonely or usually in conjunction with of the aerodynamic properties of particles, can be mainly
another method to separate and clean various heterogeneous used as their distinguishing feature in such aerodynamic
agricultural grain mixtures [5–7]. The advantages of this applications and optimizations [8, 20, 23]. Terminal velocity
technique can be summarized in the following points: allow is the velocity of the vertical air stream which keeps the par-
for cleaning and segregation of seeds particles simultane- ticle to be suspended in the air as a result of balancing their
ously without changing the separation parameters, obtain weight with the force of the air stream [24–26]. The values
the product without damage, no change in the physical and of this parameter are affected by some secondary parameters
biological properties of the particles during the separation like density, shape, and cross-sectional area [22] that can
process, and characterized by a high efficiency without be formulated according to Eq. (1) [27, 28]. The greater
demand much power [8]. Found that the most effective fac- the particle mass, the greater the terminal velocity, and the
tors to improve the quality of the aerodynamic system are: larger the particle area, the lower the velocity.
the aerodynamic properties of the particulate matter, feeding √ ( )
√
load ratio of granular flow, speed and angle of feed grain mi g 4di 𝜌i − 𝜌a g
VT = or ≅ (1)
material, airflow properties, particles properties, and the ka Ai 𝜌a 3Cd 𝜌a
system inlet geometry [9–12].
In this process, the aerodynamic properties of particles, where VT is the terminal velocity, mi is the mass of the par-
such as their critical or terminal velocity and drag coeffi- ticle, g is the gravity acceleration, ka is the dimensionless
cient, are mainly used [7, 13]. The large the difference values coefficient of aerodynamic resistance, Ai is the projected or
of terminal velocities in the mixture component, the better lifting area of the particle,di is the particle diameter, 𝜌i and
and more efficient is the aerodynamic separation [14, 15]. 𝜌a are the particle and air density,Cd is the drag coefficient
Commonly, the aerodynamic systems can be classified into and can be calculated, as mentioned in Sect. 6. The drag
three significant aspects based on the direction of the air coefficient is a dimensionless parameter of the particle that
stream in the system: vertical, horizontal, and oblique.it can is used to quantify how easy or difficult it is for a particle to
also classified depending on the conditions of the airflow or be entrained in the fluid flow [29]. It is related to the particle
location of the fan in the system (vacuum, overpressure, and Reynolds number, which is given by Eq. (2)
hybrid) [15, 16]. Herein, the value of the redistribution of
product components between fractions should be taken into 𝜌a di ||va − vi ||max
Rei = (2)
account to evaluate the quality of the aerodynamic separa- 𝜇a
tion process, mainly when the initial product is divided into
where Rei is the particle Reynolds number and 𝜇a is the
three or more fractions [17].
dynamic viscosity of the air. Since the particle Reynolds
The expediency of using forced air technology is deter-
number depends on terminal velocity, the calculation of
mined by the presence in the initial material of components
terminal velocity is an iterative procedure. The following
that differ from the main product by aerodynamic properties.
three flow regimes can be characterized based on particle
If there are no such differences, other methods will be used.
Reynolds number: Stokes’ regime when Rei < 0.2, interme-
Up to 70% of the impurities in the original grain material
diate regime when 0.2 < Rei < 1000, and Newton regime
that comes to post-harvest processing are separated from
when Rei > 1000 that will be described in detail in Sect. 6.
the main crop by aerodynamic characteristics [17]. The effi-
Typically, the value of the drag coefficient Cd of spheri-
ciency of any air-using machine depends on the weight and
cal and cylindrical particles can be assumed 0.44 and 1.0,
aerodynamic drag of the particles, which in turn affects their
terminal velocity of fall [18]. Consequently, awareness of
13
respectively, in the range of Reynolds number of Re = 1000 the current increase of computing power and technology,
to 200,000 [22, 30]. computational simulations are showing high adaptability to
Experimentally, the suspension or terminal velocity of simulate the multiphase flow accurately. Numerical models
particles can be determined by the suspension apparatus test, have been used with some relevant modifications in recent
and it can be slightly differing from experiment to another and previous studies to describe the particulates matter and
depending on the grain material and the study condition [24, gas flow field inside the aerodynamic systems [35–42].
31–34]. Since the bulk bio-particulate matter mass introduced
When applying aerodynamic principles to cereal grain to the structure domain, it is considered discrete elements,
crops, a method for expressing the projected area of the par- where each particle is an element, having an interaction with
ticle must be chosen because of the irregular shape of the neighboring elements. Discrete Element Method (DEM) is
particles. So, in most cases, the particles are assumed to be a common and powerful numerical simulation technique for
spherical, where the diameter of the particle could be taken predicting and modeling this discrete media flow basis on
as the diameter of a sphere of the same volume as the origi- Newton’s laws of motion. The extensive usage of bio-par-
nal size. In cases where the volume of the object is difficult ticulate matter applications in industrial processing methods
to evaluate, the geometric mean diameter is a good approxi- leads to several DEM articles development [43–51]. Com-
mation of the diameter of the equivalent sphere, provided puters with significant memories and a high-performance
that the shape factor is close to unity [26]. Theoretically, the processor make it possible to simulate a large number of par-
net force acting on a particle moving in a fluid is defined by ticles with a straight-forward approach. DEM can dynami-
the difference between the drag and the gravitational forces, cally simulate each particle trajectory of dense phase flow
as shown in Eq. (3) [21]. by tracking its rotational and translational movement in the
domain, taking into consideration the contact forces and the
1
Fnet = C A 𝜌 V 2 − mi g (3) interaction couples of particle–particle and particle-bound-
2 d i a a
ary [35, 52, 53].
where Fnet is the net force acting on the particle, Va is the Practically, the particulate matters are introduced to the
velocity of the air. aerodynamic systems with air or gas flow; hence, it can be
regarded as multiphase flow. Therefore, Computational Fluid
Dynamics (CFD) models are mainly used to capture the air
1.3 Modeling Representations or fluid flow characteristics successfully with essential data
in predicting the aerodynamic behavior [54–57]. For gas-
Nowadays, improving the design and the performance of particle flows, CFD multiphase models cannot be used to
harvesting and post-harvesting devices such as combine estimate the particulate matter phases; due to many restric-
harvester, grain dryers, spouted bed, aerodynamic separa- tions happen when the dense flow is applied, those restric-
tors, etc. becomes an urgent necessity for industrial needs. tions can be exceeded by coupling CFD with DEM [58].
Experimental studies, mathematical models, and computer This coupled process increases the computational time but
simulations are essential tools to solve this problem. With enables a resolved particle–particle interaction combined
Fig. 1 Number of published 700

works obtained from Science
Direct using keywords: “CFD– 600
Number of publicaons
DEM” (Accessed in December

2019) 500
400
300
200
100
0
2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019
Year
13
with gas to solid and solid to gas modeled interactions as 2 Theoretical Background of DEM
well.
The DEM is considered one of the numerical modeling
1.4 Scope of this Work technique, which is first published by Cundall and Strack
[60]. It is mainly developed to complement the finite ele-
In this decade, as illustrated in (Fig. 1), CFD–DEM cou- ment method to solve the complicated problems in engi-
pling has become a hot research area for fluid-particles flow neering and applied science and investigating natural phe-
modeling, especially with the increase of computer tech- nomena of granular flow of differently shaped particles that
nology progress and theoretical developments. However, display sporadic conduct [61]. DEM simulations can also
previous reviews related to the bio-particulate matter field provide dynamic information which is extremely difficult or
concentrated on DEM modeling [51, 59] or CFD technique impossible to obtain by traditional experimentations, such
[57] only; no reviews can be found on CFD–DEM coupling as trajectories of, and transient forces acting on, individual
basics on modeling the bio-particulate matter in aerody- particles, and analyze multiple, interacting, deformable, dis-
namic systems. However, these reviews did not concentrate continuous, or fractured elements undergoing rotations and
on the force models related to fluid-particle interactions or large displacements [59, 62]. With advanced computers, the
the coupling ways between CFD and DEM for the bio-par- simulations software reiterates the calculations million times
ticulate matter. This paper is the first try aiming to provide a during a second for each element of the system. Currently,
summary of the investigates based on computational simula- the most advanced software and hardware allow simulations
tions in the past two decades or so to open the way for the of processes with over a million particles, which facilitate
researcher to know the last working in these field. This infor- the practicality of the DEM approach [63].
mation is important in the application of grain modeling in In DEM modeling, particle–particle interaction is treated
particle–fluid multiphase flows, and therefore, an extensive as a dynamic process and solved by the explicitly numeri-
review is necessary. cal scheme. The contact forces, rotations, momentum, and
The objective of this paper is to comprehensively displacements of a stressed particle are described by tracking
review and collect the existing published research that used each contact of the individual particles [64]. The primary
the coupling of CFD–DEM as a numerical technique for assumption in the DEM is that every discrete element has
simulating the aerodynamic particulate systems related to different boundaries that physically separate it from every
the agricultural field. Firstly, this review summarizes the other element in the analysis. During the element move-
representations of the aerodynamic system in the agricul- ment in the domain, the translating and rotating behavior
tural process and industries, including some experimental equations are applied on an element-by-element separately
works. Secondly, it represents the equations of particle and the elements contacting through boundary forces. The
motion and contacts force models between particles and deformation resulting from the collision can then be decou-
interaction forces between particles and fluid in the aerody- pled from the mean motion and written as the sum of the
namic systems as apart of DEM and CFD theories. Thirdly, bodies’ normal modes, which in turn gives a newly derived
drafts the CFD–DEM coupling ways, including theoretical set of decoupled modal equations. These new equations are
developments and applications with a brief introduction of then solved by an explicit central difference scheme, where
drag correlation models. Fourthly, introduces the applica- the final solution is obtained through modal superposition
tions of CFD–DEM modeling in aerodynamic systems and [59, 65].
particle–fluid flow systems where the significant findings Recently, the Discrete element method (DEM) applica-
were discussed in some detail. Finally, the main challenges tion is gaining popularity in the harvest and postharvest
of CFD–DEM currently faces, and the exciting research processing of grain and highly utilized for many industrial
topics for future studies are discussed. We conclude that processes, such as granular mixing, drying food, pharma-
CFD–DEM is an effective method for particle–fluid scale ceutical industries, mining, and agricultural processing. The
research in the aerodynamic systems related to the agricul- widely spread is because of its close and real characteri-
ture field. It also has a bright future as it offers a new way zation of actual conditions in predicting various processes
of understanding industrial processes and/or phenomena [66]. Unlike the field of mining and the chemical industry,
with practical particle shapes. However, there are still some however, DEM is not widely applied in the agricultural pro-
aspects that are needed to be investigated in the future. cesses because of various particle property issues arising
from the biological origins of grain and food products.
In the agricultural industry, the DEM in agricultural
and food processing applications have been considerably
developed to describe; the flow in chutes and dryers [37,
13
67–69], the aerodynamic conveying systems [70, 71], the each collision. Therefore, this approach is probably suited
discharging silo [72], the transportation and compression of for rapid granular flow simulations, where the particulate
seeds [73], the dynamics discharge of particles in hoppers material is partially or entirely fluidized [92].
[74], the shaking separation of paddy and brown rice [75], In contrast to the soft-sphere approach, the overlapping at
the grain–straw separation screening in combine harvester the contact points is allowed, and the contact forces between
[76, 77], the fluidization and spray drying of food powders particles are strongly considered [62, 91]. Herein, it is essen-
[78], the grain transportation and aerodynamic separation tial in this approach to model the forces of elastic restitution
[79], the grains segregation and seed motion in seed feed- and friction that happened during the collision of the parti-
ing device [80, 81], the screening separation of crop seeds cles [95]. Because of the nature of the iterative schemes used
[82, 83], and the mixing process in a bucket elevator [84]. in the soft-sphere approach, it is susceptible to the time step.
DEM has also been useful in the modeling of interactions Therefore, the time step interval is assumed constant during
between soft materials: fruits such as grape [85], apple [86], simulation [93] and should be smaller than the duration of
or biomass [87]. A complete description and good over a contact for stability considerations [96]. Usually, in most
review of the DEM can be found in much previous literature soft- sphere models, the duration of the contact is artificially
[51, 59, 65, 88, 89]. increased by allowing softer interaction and hence reducing
the required CPU time. The obtained deformation due to
2.1 Approaches in DEM Modeling the collision can be considered similar to that occurs at the
contacts among bio-particles. These deformations are used
In general, the granular dynamics modeling through DEM to calculate elastic, plastic, and frictional forces between
techniques can be classified into hard-sphere and soft-sphere particles [59]. Hence, this approach highly used in particu-
approaches depending on the particle deformation during late mechanics models related to agricultural cereals and
contact or collision [90]. The first concept of the “hard crops. The motion of particles in this approach is described
sphere” that comes to mind to simplify the collision mod- through numerical integration of the Newtonian equations
eling is to treat the particles as hard spheres; however, it is of motion. Apart from external forces acting on the system,
not necessary to describe the collision as entirely elastic. The the interparticle forces are of crucial importance to these
word hard simply means that there is no interpenetration or models [38, 97]. The magnitude of the overlap can be related
deformation during the collision. In contrast, the word “soft” to the interparticle contact force, as discussed subsequently.
revolves around the deformation of spheres and the duration The significant advantage of soft-sphere models is that
in contact. Figure 2 schematically indicates the main dif- they can model the dense-phase bulk granular materi-
ferences between these two modeling approaches [91, 92]. als regarding the multiple particle contacts [98], which is
In the hard-sphere modeling approach, the effect of over- essential when modeling quasi-static systems [62]. These
lapping and deformation or interpenetration between par- models start by solving the equations governing the linear
ticles during contact is assumed to be neglected [93]. So, translation and rotational motions of particles [92]. While
this contact can be named as a non-smooth DEM where the in hard-sphere models, the interaction forces are assumed to
particle’s motion and the energy loss during a collision can be impulsive, and hence the particle modeling strategy starts
be modeled through shock laws and the coefficients of resti- from the equations governing momentum exchanges [93].
tution [94]. Collisions are also assumed to happen in a short However, hard-sphere approaches are computationally
time and may be assumed to be instantaneous; because the cheap, fast, and preferred for non-dense flow; it can give a
collision is limited between two particles only, one collision limited description of the dense material’s response involv-
at a time, and does not consider multiple contacts [58]. So, ing multiple simultaneous contacts [99]. Therefore, given
different time steps will be applied, and the time step inter- the advantages of soft-sphere approaches mentioned above
val for the numerical solution varies with the time between for describing the bulk material physics, where it is most
Fig. 2 Illustration of hard-
sphere and soft-sphere
approaches [91, 92]
13
commonly used in the grain and food processing industries, The contact between two particles is not taken place at a
no further consideration of the hard-sphere algorithms is single point but on a specific area, which can lead to making
given in this study. Thus, soft-sphere modeling is called the particles slightly overlapping on the normal and tangen-
DEM (Discrete Element Method or Distinct Element tial coordinates (n and t) [62]. Particle overlapping cannot be
Method) [100], which is the scope of this review. neglected because it can significantly affect the simulation
result [105]. So, the total contact force acting on the particle at
2.2 Governing Equations of Motion the contact plan can be analyzed on n and t coordinates which
can be mathematically calculated according to Eq. (6). In addi-
In soft-sphere DEM modeling, the motion of each particle can tion, (Fig. 3) indicates a schematic illustration of the typical
be tracked by calculating the contact forces and displacements contact force vectors with overlapping and torques implicated
using an explicit numerical scheme and a very small time step in a soft-sphere DEM modeling.
[60]. The particle movements can be divided into two parts:
the translation motion calculated by the basic’s equations of FijC = Fijn + Fijt (6)
Newton’s second law as shown in Eq. (4), and rotation motion
where Fijn and Fijt are the normal and tangential contact forces
described by Euler’s law as formulated in Eq. (5) [101–103].
In general, the spaces between the particles are filled with on the particle i by contacting particle j , respectively.
interstitial fluid, usually air in aerodynamic systems. When
that fluid forces the particles, the effect of this interstitial fluid 2.3 Modeling of Contact Force
should be taken into account. Nevertheless, if the particles are
considered to be large and heavy, they are exempt from the In any granular flow, forces are mainly generated by inter-
effects of the interstitial fluid [98, 104]. particle contacts that can be mechanically described by how
the particles “see” one another. The philosophy of DEM is to
nci nci
d2 ∑ g f
∑ use rigid springs, acting both normal and tangentially to the
mi 2 xi = Fijc + Fi + Fi + Fiknc (4) contact plane to calculate the interparticle forces. Although
dT j=1 k=1
DEM simulations are worked efficiently in predicting many
processes in the agricultural industry, it requires a detailed
nci
d2 ∑ understanding of the interactions between particles under
Ii 2 Φi = Mij + Mrij (5) various loads.
dT j=1
Precise values of physical and mechanical parameters
where mi is the particle mass, xi and Φi are the position and of granular materials that have biological origins are con-
the orientation in the space of particle i , respectively,FijC the sidered critical parameters for modeling contact models of
DEM [106]; due to the high moisture content of these parti-
total contact force, Mij is the momentum acting on particle i
cles, which makes the deformability to be high during con-
by particle j or wall, fi is the gravitational force, Fi is the
g f
tact [51]. As such, different contact models can be generated
fluid-particle interaction force on the particle i , Mrij the
and developed based on elastic, viscous, plastic, dry friction,
momentum generated by rolling friction when the particle
and adhesive interactions.
rotates in a fluid which will be represented in Sect. 5, Fiknc is
Usually, the main assumptions of the theoretical contact
a term of non- contact force acting on particle i by particle
models are considered the particles to be isotropic and elas-
k or other sources, and Ii is the moment of inertia of particle
tic, where the contact is completely smooth. It is a beyond
i , and T is the time.
contradiction that the bio-particulate matters are varied in
Fig. 3 Contact forces analysis

during the collision of the dis-
crete particles considering the
overlapping [35, 105]
13
geometry and shape. Most of them have a convex shape, where E∗ and R∗ are Young’s modulus equivalent and radius,
while others are spherical in shape, but none of them are Ei , Ej , and 𝜈i , 𝜈j are Young’s moduli and Poisson’s ratios of
purely elastic bodies [51]. Therefore, the application and the two contacting spheres of radii Ri , Rj,Sn is the normal
the development of contact models for biomaterials related overlapping, 𝜂 is called a damping ratio, which is a dimen-
to the agricultural field face many restrictions. sionless parameter related to the restitution coefficient that
The contact forces in any DEM code consist of forces can be calculated as expressed in Eq. (10) [44, 111].
normal-to-contact plane and force tangent-to-contact plane. √
For spherical particles, the contact plane is perpendicular to 𝜂 = − ln 𝜀
5
(10)
the line that connects the centers of two spheres. For non- ln2 𝜀 + 𝜋 2
round particles, the methodology for determining a contact
where 𝜀 is the coefficient of restitution defined by DEM user
plane is more complicated. It can be determined by the clos-
according to the studying case and m∗ is the equivalent mass
est points of two particles or the closest points of a particle
for the particles in contact and can be calculated according
and triangle or the two points with the maximum overlap
to Eq. (11).
distance in the case of a physical contact [107]. This chapter
will introduce the standard contact models and the interac- 1 1 1
= + (11)
tion properties, including coefficients of restitution, static m∗ mi mj
and rolling friction, plastic or viscous damping, and adhe-
sion used by investigators for the bio-particulate matter.
2.3.1.2 Elastic–Plastic Contact Model This model was first
introduced by Walton and Braun [112] and named as a Hys-
2.3.1 Normal Force Modeling teretic Linear Spring Model [113, 114]. This elastic–plastic
or repulsive-dissipative normal contact model allows simu-
The primary requirements to model the normal contact force lation of the plastic energy dissipation on contact without
are that the force should have to be repulsive, and the model introducing the overhead of long simulation times. Besides,
has to allow significant energy dissipation. For this purpose, the energy dissipation is not considered the velocities of the
several investigators developed many models to describe the neighboring particles and loading rate that makes the energy
behavior contact of the granular medium systems. dissipation insensitive to other contacts. Furthermore, the
coefficient of restitution is velocity independent, and it can
2.3.1.1 Hertzian Contact Model Following the work of be accurately modeled the compressible materials with cor-
Hertz in the nineteenth century and the theory of elastic- rectly simulating the non-adhesive particles flow [35, 52, 53,
ity [108, 109], the normal contact force is modeled by con- 115]. The following equations starting from Eqs. (12–17)
sidering the response of two elastic spheres in contact over draw and describe mathematically one by one the linear hys-
a small circular contact area. This model has been widely teresis model used in DEM code [29, 107, 116–118]
used to determine the modulus of elasticity for various agri- � �
cultural grains that have a convex shape [110]. It is usually ⎧ n,(T−ΔT)
⎪ min Fi + Kun ΔSn , Kln Sn,T , if ΔSn ≥ 0
called as a Hertzian spring dashpot model where the normal Fin,T =⎨ � �
n,(T−ΔT)
contact force is modeled as the sum of an elastic force and ⎪ max Fi + Kun ΔSn , 0.001Kln Sn,T , if ΔSn < 0
⎩
a damping force. When considering the overlapping during (12)
contact, the model can be expressed mathematically in an
incremental way, as shown in Eqs. (7, 8, and 9). ΔSn = Sn,T − Sn,T−ΔT (13)
where and
Fin,T are the normal elastic–plastic con-
Fin,(T−ΔT)
� �1
n 4 ∗ √ ∗ n3 4 ∗ ∗ √ ∗ n ∕2 d n (7) tact forces acting on the particle i at the current time T and
Fij = E R S + 𝜂 m E R S S
3 3 dT at the previous time T − ΔT respectively, where ΔT is the
timestep, ΔSn is the change in the overlapping during the
( )
( ) time step, Sn,T , Sn,T−ΔT are the normal overlap values at the
1 − 𝜈i2 1 − 𝜈j2
1
= + (8) current and at the previous time, respectively, Kun and Kln are
E∗ Ei Ej the unloading and loading contact stiffnesses, respectively.
Note that the unloading force is limited by the value of
Ri Rj 0.001Kln Sn,T to ensure that the normal force becomes zero
R∗ = (9) with zero overlappings. A schematic sketch, as shown in
Ri + Rj
(Fig. 4) can demonstrate the overlapping circle (A–C) of load-
ing and unloading response due to the normal force effect
13
Fig. 5 The relationship between damping ratio and the restitution

coefficient [107, 120]
Fig. 4 Normal force–overlapping response for the hysteretic linear
spring model [35, 107]
is that it takes less time and machine resources to process
where the hatched area is represented the energy dissipation but provides less accurate results compared to the hysteretic
during the particles contact. Besides,Kln and Kun are the slope linear spring model [107]. When Visco-elastic particles are
of loading and unloading lines, respectively, and the plastic in contact, the normal contact force is given by the sum of
deformation only happens during the contact of the particles elastic repulsive force and a damping force proportional to
i and j ; after that, the residual deformation neglected. the velocity as formulated in Eq. (18) [51, 119].
The coefficient of restitution can be determined by the
user, which represents the energy dissipation due to parti- d n
Fin = Kln Sn + Cn S (18)
cle–particle or particle–wall contact; also, the loading and dt
unloading stiffness represented in Eqs. (14 and 15) can be where Cn is the normal damping coefficient and defined as
taken into account by considering the valve of Young’s follows in Eq. (19).
modulus and the particle geometry. √
Cn = 2𝜂 m∗ Kln (19)
⎧ 1 1
1 ⎪ Klin
+ Kljn
Particle - Particle
n =⎨ 1 1 (14) where the relationship between the damping ratio and the
Kl ⎪ + Particle - wall
⎩
Klin n
Klw restitution coefficient can be described in Eq. (20) [120].
In detail, (Fig. 5) describes the relationship between the
damping ratio and restitution coefficient given by Eq. 19.
Kln The graph is showing the increase of damping ratio when
Kun = (15)
𝜀2 the restitution coefficient decreased; because of the inverse
relationship between the two parameters.
whereas:
� � √ ��
Klin = Ei Z (16) ⎧ 𝜂 2𝜂 1−𝜂 2
⎪ exp − √ 𝜋 − arctan 1−2𝜂 2
if 0 ≤ 𝜂 < √1
⎪ � 1−𝜂 2 � √ �� 2
2𝜂 1−𝜂 2
𝜀 = ⎨ exp − √ 𝜂 2 arctan 1−2𝜂2 if √1 ≤ 𝜂 ≤ 1
Klkn = Ew Z (17) ⎪ � 1−𝜂 � √ �� 2
2𝜂 1−𝜂 2
⎪ exp − √ 𝜂2 ln 1−2𝜂2 if 𝜂 > 1
where Ei and Ew are Young’s modulus of particle and bound- ⎩ 𝜂 −1
ary, respectively, and Z is representing the particle size (20)

or geometry. This model should be applied to adequately
selected individual cases of the behavior of dry and hard For soft biological materials such as apple, potato,
seeds with the elastic behavior in the normal direction, such and tomato, etc. in addition to the soft and high moisture
as the application of discharging silo with rapeseeds and the content of seeds, Visco-elastic contact model is efficient
aerodynamic systems of seeds and foreign materials separa- and appropriate to predict the energy dissipation and the
tion [35, 72]. impact reaction [121, 122]. While for dry and hard seeds
such as rapeseed, the Elastic–plastic contact model works
2.3.1.3 Visco‑Elastic Contact Model The simple model to well to estimate the contact and impact behavior [72].
describe the Visco-elastic behavior of the particles is called
a linear elastic, viscous model. The advantage of this model
13
2.3.2 Tangential Force Modeling

i
2.3.2.1 Linear Spring Coulomb Model This model named by
the Linear Elastic-frictional model; because of its elastic behav- Normal Dash-pot
Normal
Normal Spring
force
ior before the friction commencement, and it can regenerate
Coulomb frictional contact behavior for static and dynamic
Tangenal Spring
frictional values. During the tangential contact, the particle
sliding plays a vital role in determining the total tangential con-
tact force, where Eq. (21). can express this effect [29, 53, 118]. j
Tangenal
( ) force
Slider
Fit,T = min Fit,(T−ΔT) + Kln ΔSt , 𝜇Fin,T (21)
Tangenal Dash-pot
where Fit,T and Fit,(T−ΔT) are the tangential contact forces
at the current and previous time of the simulation, respec- Fig. 6 Hertz-Mindlin contact model with viscous damping and fric-
tively,ΔSt is the change in the tangential overlapping during tional slider in the tangential direction [35]
the time step, and 𝜇 is the friction coefficient which can be
defined as a static and dynamic coefficient ( 𝜇s , 𝜇d ) depend- 1
ing on the tangential sliding contact, as shown in Eq. (22) 𝜂 t = − ln 𝜀 √ (24)
[29, 117]. ln 𝜀 + 𝜋 2
2
{
𝜇s if no sliding at the contact ( )−1
𝜇= (22) 1 − 𝜈i 1 − 𝜈j
𝜇d if sliding act at the contact stmax = 𝜇 + Sn (25)
2 − 𝜈i 2 − 𝜈j
When the tangential force exceeds the limit of 𝜇s Fin,T The
sliding is considered to be taking place on the contact. Once The Mindlin–Deresiewicz model does not work appropri-
that force falls below the value of this value, the contact is ately with rolling resistance models for non-spherical par-
considered non-sliding again. This model applied effectively ticles. Therefore, when the Mindlin-Deresiewicz model is
and shoe a high validity to separate jojoba seeds from its leaves selected with non-round particles, the rolling resistance must
using cyclonic separation process [35] and for vertical pneu- be neglected [107]. This model was improved to be suitable
matic separation of sugarcane bagasse particles [29]. in simulating soybeans flowing down an inclined chute [68].
In this simulation, the soybeans (non-spherical particles) are
2.3.2.2 Mindlin‑Deresiewicz Model This model is called the represented by clusters of spheres to simplify contact detec-
elastic-frictional contact model, which is first derived from tion, and the modeling showed a good agreement with the
the mathematical analytic of elastic spheres in contact under experimental data. The frictional contact model of spherical
varying oblique forces [123]. It is mainly consisting of simple particles was also developed based on the Mindlin-Deresie-
loading histories of varying normal and tangential forces and wicz model to predict the particle–particle slip and the slid-
a minimal range of tangential displacement between the con- ing of stainless steel beads, pea, and rapeseed [126].
tacting spheres [124, 125]. The tangential force in this model
is given by Eq. (23) [107]:
⎡ � � t t � �3∕2 ⎤ � � � t �� ∕2 �
1
⎢ min �s �, s ⎥ t 6𝜇m ∗ n
F �s � �
max s i d t (23)
Fit,T = 𝜇Fin ⎢1 − 1 − ⎥ + 𝜂 t
1 − ̂
n × S × n̂c
⎥ �s �
c
⎢ stmax t stmax stmax dT
⎣ ⎦
where st is the tangential relative displacement at the con- 2.4 Tangent Stiffness Model
tact, n̂c is the unit vector in the normal direction at the con-
tact, 𝜂 t is the tangential damping ratio and is estimated using Hertz-Mindlin contact model is usually used to repre-
Eq. (24), stmax is the maximum relative tangential displace- sent the tangent stiffness model, which is applied in DEM
ment at which particles begin to slide as can be formulated applications as the primary law to calculate particle inter-
in Eq. (25), and dT S is the relative tangential velocity at the
d t
action forces in the tangent directions [44, 123, 125]. It is
contact. also named as a linear spring-dashpot model that includes
springs, dashpots, and frictional sliders, as shown in
13
(Fig. 6) [127]. This model combining Hertz’s theory [109] JKR adhesive model force can be calculated according to
to describe normal stiffness and Mindlin and Deresiewicz’s Eq. (26) [107]:
theory [123] to describe tangential stiffness. These stiff-
nesses were calculated as mentioned in the above sections by E ∗ a3
Fin,T = (26)
assuming the presence of elastic springs and dashpots in the
adh 6R∗
normal and tangential directions. The spring stiffness, dash- where Fin,T is the normal adhesive contact force at the current
pot coefficient, and friction sliders are expressed through adh
time, a is the separation distance of contact between parti-
Young’s modulus parameter, friction coefficient, and coef- cles or between a particle and a boundary, as shown in
ficient of restitution, respectively [39, 116, 128]. (Fig. 7), where it is related to the normal overlap using the
This model is applied intensively to model and predict expression introduced in Eq. (27) [51, 107]:
the interaction of bio-particulate matter, including seeds
and grains. The calibrations and validations showed high ( )1
a2 𝜋Γa ∕2 (27)
reliability of this model, which is presented a high predic- Sn = −
4R∗ E∗
tion of the particle’s behavior. For an instant, simulating
the motion of rice grain and short straw in air-and-screen where Γ is the surface energy and can be used to calculate
cleaning device [39], simulating the flow of corn-shaped the maximum liquid bridge tensile force that occurs between
particles in spouted bed [40], simulating of rapeseed and two particles as formulated in Eq. (28) [119]:
wheat grain motion in seed feeding device [81], predicting
Pmax = −2𝜋ΓE∗ (28)
the corn grain flow behavior in a commercial screw auger
[129], simulating the motion of rice grain and rice straw in where Pmax is the maximum liquid bridge tensile force since
a variable-amplitude screen box [76], and optimizing the the effect of its behavior can be observed through the DEM
air-blowing maize precision seed-metering device [42], etc. simulation of the particle’s free-flowing and the discharge
rate from hoppers of different inclinations [74, 136]. Another
study was performed on agricultural material and reported
2.5 Adhesion Contact Model that the maximum liquid bridge tensile force between wet
rapeseeds is approximately tenfold of the seed weight, which
In order to separate bodies in intimate contact; mechanical means that cohesive behavior can occur in the case of some
work must be expended to overcome the adhesive forces. millimeters of particles’ size [51].
They have appeared in beddings of biological materials that The JKR model has a strong theoretical basis and is
have an amount of moisture content, and its effect is shown widely accepted for adhesive elastic sphere particles. Since
in agglomeration of powders and small size particles that the surface energy can be measured experimentally, this
may create problems for storage and handling [130, 131]. model could be used without calibration, in principle, for
In order to capture this behavior numerically, the normal simulating perfect spheres [107].
repulsive force has to be supplemented by the attractive On the other hand, when adhesion between particles
normal force to predict its flow characteristics accurately is due to liquid bridge forces (big attraction between the
[107]. Adhesion between biological particles not only acts polar molecules of water), another adhesive model was
because of moisture content but van der Waals forces, elec- used, namely the constant adhesive force model [29,
trostatic forces, or liquid bridges effects as well [132]. Liq- 127]. This model depends on two significant aspects: the
uid bridge force plays a vital role during modeling particles
more significant than a fraction of millimeter-scale, which
arises from the capillary pressure in the bulk of the liquid
Liquid bridge force
and the sum of the surface tension of the liquid [119, 133]. Sn
a
This force value depended on some significant aspects, such
as space area between particles, the surface tension of the i j i j i j
liquid, geometry and the volume of the bridge, and contact

Liquid bridge force
angle between the liquid and solid surface [134]. during enduring contact (A)
Liquid bridge force during
(B) parcle separaon
The most common and popular model for representing (overlapping parcles)
the adhesive behavior is developed by Johnson-Kendall-
(C)
Roberts (JKR) [135]. This model introduces the effect of
adhesion into the Hertzian contact model, where the contact Separaon Distance
area between two particles is slightly larger than the one
predicted by the Hertzian theory. This surface energy term Fig. 7 Schematic diagram of the liquid bridge force as a function of
is added to the total energy of the system [107]. The normal the separation distance between two particles [74]
13
minimum distance between particles or particles and a Real particles are almost non-spherical, so they are rarely
boundary before the adhesive forces applied and the force used in simulation studies due to difficulties in representing
fraction expressed as a function of the particle gravity these shapes in simulations’ software [137]. Early, particles
force. Hence, it means that since the force fraction reached in DEM simulations were modeled as discs and spheres;
1, the adhesive force will be equal to the gravity force because the direction of the normal contact forces is always
applied to the particle. In the case of contact between two toward the center, and the particles’ rotation is only affected
particles of different masses, the smaller mass is consid- by tangential contact forces [142]. Athanassiadis et al.[143]
ered for the calculation of gravitational force [107], as reported that the particle shape has a strong influence on the
described in Eq. (29): mechanical response and the interaction forces of granular
{ n
material, particularly on stiffness and yielding parameters.
Fin,T =
0 ( ) if − ΔSn > 𝛿adh (29) They obtained this quote based on an experimental work
adh fadh × g × min mi , mj if − ΔS < 𝛿adh carried out on fourteen particle geometries where all are in
three dimensions (3-D) printed from the same material. The
where fadh and 𝛿adh are the force fraction and the adhesive
work includes several shapes, such as sphere, tetrahedron,
distance, respectively. Note that, when ΔSn ≥ 0 the parti-
cube, octahedron, dodecahedron, icosahedron truncated, tet-
cles start adhering to each other, but if ΔSn < 0 the particles
rahedron, triangular bipyramid, tetrahedral frame, and jacks
separate [127]. This model is applied effectively to represent
with arm length. Not only the particle size, shape, physical
the adhesion force of sugarcane bagasse particles during the
and mechanical properties are regarded as a big challenge for
pneumatic separation [29].
operating the DEM simulation modeling but the computing
power, time-consuming, and the number of discrete particles
in the system as well [59, 106, 143].
3 Modeling of Particles
Complex shapes or non-spherical particles which are very
common in the bio-particulate matter are often modeled on
The granular material is introduced in many fields of indus-
DEM as the multi-sphere approach [124, 137, 144, 145].
tries such as chemistry, agricultural processing, civil engi-
In this method, the particle shape is made by assembling a
neering, but their behavior is not yet well understood. Even
different number of spherical particles together, which in
though a pile of particles will behave like a porous solid, it
turn increases the computational speed and increasing the
will flow like a liquid if it is put on an inclined plane. Never-
accuracy to determine the interaction between spheres [145].
theless, if it subjected to significant vibrations, it would have
The multi-sphere approach was ideally used to model the
a gas-like behavior. These behaviors of granular solids were
shape of bio-particles such as soybean [106, 124], corn [40],
first described in an old proverb (ca. 98–55 B.C.) as “One
rice [146], wheat [69], pea, and bean [147]. The Accuracy of
can scoop up poppy seeds with a ladle as easily as if they
the particle behavior prediction is influenced by the number
were water and, when dipping the ladle, the seeds flow in a
of sub-spheres combined. In the case of a small number of
continuous stream” [95]. To model granular materials cor-
sub-spheres ( ≤ 9 ), the particles interpreted as interlocked, if
rectly, the shape is not only required but physical properties,
higher than that it may be simply classified as a decrease of
which include size, density, moisture content, etc. should
roughness when the number of sub-spheres increases [148].
be taken into consideration and determined individually for
Advanced approaches were developed to model more
each particular DEM modeling application as well [106].
complex particle shape in DEM using 3-D shapes, such
Generally, realistic modeling of the particle shape in
as ellipsoidal [149–151], super-ellipsoid [152–154],
simulations plays an essential role in emulating real pro-
sphero-cylinders and capsules [41, 155, 156], superquadric
cesses [137–140]. The Bio-granular materials, such as grain,
[157–159], bonded particles [160–163], polyhedral [35, 164,
seeds, tubers, and soft fruits have a considerable difference
165] and polygonal [166–168]. Applying these advanced
of shapes, from spherical (e.g., rapeseed), ellipsoidal (e.g.,
approaches needs a precise determination of the contact
rice) and diamond with round (e.g., corn) or sharp-edged
point, and normal and tangential contact vectors [51, 169].
(e.g., buckwheat), to very complex, non-convex shapes of
Table 1 summarized the developed particle shape models
fruits and tubers [51]. The dimensions of these materials
used for predicting the behavior of most common agricul-
can vary from below an mm to tens of cm, which gives it a
tural cereal grains. As presented, the multi-sphere approach
significant consideration during modeling. Although meas-
is commonly used where the best number of spheres to rep-
uring the three dimensions: length, width, and thickness of
resent a single grain, or a seed kernel has not been confirmed
the particles is very important for modeling, the roundness
yet. This leakage is because of the complicated shape of
of edges and a ratio of the dimensions should also be deter-
agricultural grains, and the presence of the surface crease
mined especially to describe the complex shapes [106, 141].
makes it challenging to develop a particle with identical
shapes. So, the number of spheres should be approximated
13
Table 1 Grain models used for DEM simulation

Grain type DEM model/Reference Particle shape Research representations and investigations
Soybeans 2-D single-circular disc model [82] Circular disc Simulate the separation of soybeans and mustard
seeds using a sieve device under the condition of
feeding rate and bed depth
3-D 4-sphere model [170] Ellipsoid Simulate the behavior of a single soybean kernel
bouncing in aluminum, glass, and acrylic
surfaces to measure the coefficient of restitution
with high accuracy
3-D 4-sphere model [124] Ellipsoid Predict the dry flow of soybean in a chute device
comparing with experimental data
3-D 4-sphere model [68] Ellipsoid Simulating soybeans flowing down on an inclined
chute with a bumpy bottom to predict the flow
velocity and force statistics of the soybean dur-
ing flow
Comparison of 3-D 1-sphere, 2-sphere, Sphere & ellipsoid Developing a particle model with appropriate
3-sphere, and 4-sphere model [106] parameters of coefficients of restitution, static
friction, rolling friction, particle size distri-
bution, and particle shear modulus that best
matched the property values available in the
literature. The best model is a single-sphere par-
ticle model that can correctly simulate soybean
kernels in the bulk property tests
Comparison of 3-D 5-sphere, 9-sphere, Ellipsoid Developing a particle model to analyze the
13- sphere model [171] discharging process of a silo. The result of
the 5-sphere model was the best for modeling
the soybean seed particle, which verified the
feasibility and validity of the modeling methods
of soybean seeds
Rice 2-D single circular model [75] Circular element Modeling paddy and brown rice kernels in the
shaking separation process comparing with an
experiment on the same scale
3-D triaxial particle model [142] Ellipsoid Simulating the impact behavior against the grain
loss sensor. The DEM results were similar to
experimental data, which provide evidence for
optimizing the grain loss sensors design for
accurately identifying rice seed impact signals
3-D 11-sphere model [146] Ellipsoid Simulating the filling and discharge flow and
piling of the rice kernels with rolling friction
coefficients of zero and 0.3. Considering the
rolling friction, the pile shape is similar to the
actual pile. While, without rolling friction, parti-
cle mobility was higher, resulting in a spread of
particles rather than a pile
3-D 7-sphere model [172] Ellipsoid Predicting the impact behavior of rice kernels on
the impact board of an inclined elevator head
with high accuracy
3-D single-spheroid model [39] Spheroid Simulating the separation process of rice kernels
and straws using a coupled DEM and CFD
model in an air and screen cleaning device
13
Table 1 (continued)
Corn Comparison of 3-D 1-sphere, 2-sphere, Sphere & cluster shape Studying the influence of corn shape and the
5-sphere, and 13-sphere mode [129] interaction effects of DEM parameters such as
coefficients of static and rolling friction. The
developed models were validated and calibrated
by comparing the flow rate and corn pile with
experimental tests of commercial screw auger.
It is showed that the 2-sphere model is the best
DEM corn shape model
2-D 2-disc clump model [173] Disc Modeling the internal friction angle and stiffness
of grains where the calibration process validated
by modeling silo discharge and bucket filling
Bonded particle model (BPM) [42] Spheres Simulating and optimizing the air-seeds flow in
the aerodynamic seed-metering device using
CFD–DEM. The model is reliable as a tool for
understanding the physical phenomenon of seed
movement in the gas phase and can be applied
successfully to analyze the sources of difference
3-D 4-sphere model [174] Cluster autonomous shape Simulating the influence of the gravitational field
on the bulk responses of the corn seeds under
three different loadings
3-D 6-sphere model [175] Cluster autonomous shape Studying the pressure distributions, bulk density
distributions, and flow properties during filling
and emptying of silos
Comparison of 3-D single-sphere, Cluster autonomous shape Developing the most accurate modeling shape
4-sphere, 6-sphere 8-sphere, and considering the rolling friction effect using the
12-sphere model [176] calibration of silo discharging. Depending on the
required accuracy, the approach of 6 sub-spheres
could be sufficient to represent the shape of the
grain of maize
13
Rapeseed 3-D single-sphere model [81] Sphere Simulating rapeseed and wheat motion using a
coupled of DEM and CFD model in seed feed-
ing device at different air velocities. The model
can correctly predict the physical phenomenon
of seeds movement in the airflow field
3-D single-sphere model [72] Sphere Predicting the rapeseed flow through a horizon-
tal orifice using different contact models. The
result showed that contact models give the same
behavior for slow particle flow comparing to
experimental results but needed some improve-
ment, including dissipation for higher particle
flow rates
3-D single-sphere model [73] Sphere Simulating the deformation under bulk compres-
sive loading of rapeseeds that there was a little
deviation in the initial particle positions between
the experiment and simulation
3-D single-sphere model [177] Sphere Investigating the influence of the moisture content
of rapeseeds on the physical properties of grain
bedding during uniaxial compression testing.
Both the simulations and experiments showed
differences in the elasticity and the stress
transmission at various grain moisture contents.
Besides, the mechanical response is significantly
affected by the moisture content of kernels
3-D single-sphere model [178] Sphere Simulating the free fall and impact of rapeseeds
against a flat surface using different contact
models and moisture content
2-D single circular disc model [179] Circular disc Modeling the bulk behavior of rapeseed during
a direct shear test to determine the influence of
three different levels of the coefficient of inter-
particle friction
13
Wheat 3-D 3-sphere model [67] Cluster shape Predicting the velocity distribution of wheat ker-
nels in a mixed-flow dryer where the obtained
velocity was nearly accurate to the real
Comparison of 3-D single-sphere, Cluster autonomous shape Comparing the performance of different particle
4-sphere, and 8-sphere model [180] models in predicting the angle of internal fric-
tion and cohesion of wheat using both linear
and nonlinear contact models. The single-sphere
particle model behaved better in the simulations
than the multi-sphere models. Also, 3-D particle
models have higher accuracy in predicting the
bulk behavior of wheat than a 2-D approach
2-D single circular disc model [181, 182] Circular disc Investigating the wheat velocity, moisture content,
and residence time distributions in a mixed-flow
dryer. The DEM can adequately predict the
main features of particle flow comparing to the
experimental results
2-D 5-disc ellipsoidal clump model [69] Ellipsoid Evaluating the traditional designs of mixed-flow
dryers with different air ducts by coupling CFD–
DEM to discover design deficits. The model can
predict the real flow pattern, but not produce the
expected dynamic angle of repose that typically
formed under the air ducts
Comparison of 3-D single-sphere, multi- Sphere & kernel shape Modeling the wheat in the first break milling pro-
sphere model [183] cess at different moisture content. The sphere-
shaped model had a substantial deviation from
the experimental milling results. However, the
two shaped models were able to predict the trend
observed for change in particle size of the first
break stream with moisture content
Other
Jojoba & Spherical and custom polyhedron shape [35] Modeling Jojoba seeds and its leaves as a spheri-
leaves cal and custom polyhedron shape to investigate
the separation process using cyclone separator
and applying the coupling of CFD–DEM. The
model can ideally use to optimize cyclone geom-
etry resulting in improving the performance
Bagasse Rounded cylinder shape [29, 52] Investigating theoretically the bagasse particle
particles separation process in a pneumatic cylinder
device using CFD–DEM in order to optimize the
machine used for that purpose
by trials depending on the computation time and prediction phenomena equations of mass, momentum, and energy [54,
accuracy required. 185]. It has been widely used in industry to optimize pro-
cesses, reduce the energy costs, create new designs without
wasting resources by carrying out experiments, and predict
4 Theoretical Background of CFD fluid flow in several processes such as mixing, drying, sepa-
ration, and fluidized bed granulators [186].
At present, Computational Fluid Dynamics (CFD) is fre- Usually, to drive the continuum flow pattern in the pres-
quently used in the simulation of the internal or external flow ence of a secondary particulate phase, the system domain
field of several fluid systems [184]. It is continued to expand is discretized in small grid cells where the volume-aver-
the scope of application and offers new ways to solve prac- aged Navier–Stokes–Equations governed the fluid phase
tical engineering and agricultural problems. It can predict motion [53, 102, 103, 187]. These 3-D governing Eqs. (30
and provide a qualitative and sometimes even quantitative and 31) are typically referred to as continuity and momen-
prediction of fluid flows via mathematical modeling (par- tum conservation equations. Recently, the solution of these
tial differential equations), numerical methods such as finite equations is solved computationally utilizing finite volume
volume method, and software tools by solving the transport method in a CFD code, which can be written as:
13
( )
𝜕 𝜌f 𝛼f ( ) and temperature profiles of the fluid over the given domain
+ ∇. 𝜌f 𝛼f Vf = 0 (30) [54]. While it is impossible to state the best turbulent model
𝜕t
for a specific application categorically, so in this section, we
( ) will draft a small brief about the most critical and popular
𝜕 𝜌f 𝛼f vf ( ) f turbulence models used in the aerodynamic systems related
+ ∇. 𝜌f 𝛼f Vf Vf = − ∝f ∇p + ∇.(∝f 𝜏f ) + ∝f 𝜌f g − Fi
𝜕T to the agricultural application.
(31)
where 𝛼f is the fluid volume fraction in each cell,𝜌f is the
fluid density,vf the average velocity of a fluid cell, p is the 4.1 Standard k − ε Model
pressure shared by two phases, 𝜏f is the viscous stress tensor,
and Fi is the interaction force of the fluid on the particle i .
f The standard k − 𝜖 model was firstly proposed by Launder
The numerical settings of CFD should be selected appro- and Spalding [190, 191], which is regarded as a semi-empir-
priately to make the results more reliable. Therefore, a series ical model based on modeling transport equations for the
of numerical calculations for aerodynamic systems were per- turbulence kinetic energy ( k ) and its dissipation rate ( 𝜖 ) as
formed using different grid numbers and turbulence models presented in Eqs. (32 and 33). Authenticity, validity, econ-
[184]. The accuracy of the solution generally increases with omy, and realism for a wide range of turbulent flows give
an increase in cells in the region but using a smaller grid size an explanation of its popularity in industrial flow and heat
to discretize the region results in longer computational time transfer simulations. The assumption of driving this model is
[186, 188]. Although the internal flow of most agricultural that the flow is fully turbulent with high-Reynolds-number,
aerodynamic systems is considered a turbulent flow, it is and the effects of molecular viscosity are negligible [29,
189].
[( ) ]
𝜕( ) 𝜕( ) 𝜕 𝜇t 𝜕k
(32)
𝜌k + 𝜌 ku = 𝜇f + + Gk + Gb − 𝜌𝜖 − YM + Sk
𝜕t f 𝜕x f f 𝜕x 𝜎k 𝜕x
[( ) ]
𝜕( ) 𝜕( ) 𝜕 𝜇t 𝜕𝜖 𝜖( ) 𝜖2 (33)
𝜌f 𝜖 + 𝜌f 𝜖uf = 𝜇f + + C1 Gk + C3 Gb − C2 𝜌f + S𝜖
𝜕t 𝜕x 𝜕x 𝜎𝜖 𝜕x k k
an unfortunate fact to accept a single turbulence model to where uf is the average velocity component of fluid at the
be generally used as a solver for all classes of problems. position x , 𝜇f is the turbulent fluid viscosity,Gk is the genera-
Numerically, the choice of turbulence model mainly depends tion of turbulence kinetic energy due to the average velocity
on some parameters such as physics encompassed in the gradients, Gb is the generation of turbulence kinetic energy
flow, established technique for a specific class of problem, due to buoyancy, YM is the contribution of the fluctuating
level of accuracy required, the availability of computational dilatation incompressible turbulence to the overall dissipa-
resources, and the time available for the simulation [189]. tion rate, 𝜎k and 𝜎𝜖 is the turbulent Prandtl numbers for k and
Currently, the CFD technique with fast and advanced 𝜖 , respectively, Sk and S𝜖 are user-defined source terms, R𝜖
computational processing equipment makes the complex represents the dissipation of swirl and rotational effect on
turbulent fluid systems possible to model. Furthermore, most turbulence, where C1 , C2 , and C3 are constants.
commercial CFD software such as STAR-CCM + from CD- Although the standard k − 𝜖 model considered famous
Adapco, COMSOL Multiphysics, SIMULIA PowerFlow, and popular for modeling the full turbulence flow, it provides
ANSYS Fluent, and other open or licensed sources software an inaccurate rate result for high internal velocity with high
provide parallel computing options to speed up the computa- swirling and turbulence (anisotropic flow). These results
tional processes. A large number of turbulence models start- make it useless and weak for modeling real-flow in the cha-
ing by k − 𝜀 model and renormalization group RNG k − 𝜀 otic aerodynamic systems, mainly cyclone separators [35,
model to the more complicated as Reynolds stress models 49, 54, 192, 193].
(RSM) can be applied smoothly in many CFD software
codes. These models can provide a real velocity, pressure,
13
4.2 RNG k − ε Model contributions to the pressure-strain correlation considered

below, 𝜖ij is the dissipation, Fij is the production by system
To overcome the shortages of the standard k − 𝜖 model, the rotation, and Suser is the user-defined source term.Whereas:
RNG-based k − 𝜀 model has been successfully developed. It ( )
𝜕
is derived from the instantaneous Navier–Stokes equations
� �
Cij =
𝜕x
𝜌f u k u i u j (37)
of average mass and momentum conservation, using a math- k
ematical technique, namely “Renormalization Group”(RNG) [ ( )]

methods [194]. This method has some powerful features, 𝜕 𝜕 � �
DL,ij = 𝜇 uu (38)
such as the effect of swirl on turbulence, an additional term 𝜕xk 𝜕xk i j
in the dissipation rate equation that significantly improves the
accuracy for rapidly strained flows, an analytically derived ( )
𝜕uj 𝜕u
differential formula for effective viscosity that accounts for + uj uk i (39)
� � � �
Pij = −𝜌f ui uk
𝜕xk 𝜕xk
low-Reynolds-number effects. These characteristics make the
models more accurate and reliable for a broader category of ( )
flows than the standard k − 𝜖 model [189, 195, 196]. Consid- � �
Fij = −2𝜌f Ωk uj u�m 𝜖ikm + ui u�m 𝜖jkm (40)
ering these refinements in the standard k − 𝜖 model equations
mentioned above, the model can be mathematically described
well as formulated in Eqs. (34 and 35): ⎛ � �
⎞
𝜕 ⎜ 𝜇t 𝜕ui uj ⎟
[ ] DT,ij =− (41)
𝜕( ) 𝜕( ) 𝜕 𝜕k 𝜕xk ⎜ 𝜎k 𝜕xk ⎟
𝜌f k + 𝜌f kuf = 𝛼k 𝜇feff + Gk + Gb − 𝜌𝜀 − YM + Sk ⎝ ⎠
𝜕t 𝜕x 𝜕x 𝜕x
(34)
[ ] ( )
𝜕( ) 𝜕( ) 𝜕 𝜕𝜀 𝜕uj � � 𝜕ui
(42)
� �
𝜌f 𝜀 + 𝜌f 𝜀uf = 𝛼𝜀 𝜇eff Pij = 𝜌f ui uk + uj uk
𝜕t 𝜕x 𝜕x 𝜕x 𝜕xk 𝜕xk
𝜀( ) 𝜀2 (35)
+ C1 Gk + C3 Gb − C2 𝜌f − R𝜀 + S𝜀
k k �
�
𝜕ui 𝜕uj
where 𝜇feff is the fluid effective turbulent viscosity, 𝛼k and 𝛼𝜖 𝜖ij = 2𝜇t (43)
𝜕xk 𝜕xk
are the inverse Prandtl numbers for k and 𝜖 , respectively. For
more details and the sufficient background theory about the
RNG k − 𝜀 model, it can be found in the following biogra- 𝜙ij = 𝜙ij,1 + 𝜙ij,2 + 𝜙ij,3 (44)
phies [197–199].
where 𝜙ij,1 is the slow pressure-strain which representing a
return to isotropy by purely turbulent interactions [209], 𝜙ij,2
4.3 Reynolds Stress Model
is the rapid pressure-strain taking account of the return to
isotropy by the interaction between the turbulence and the
The Reynolds Stress Model (RSM) is mainly constructed for
mean velocity gradients [201], and 𝜙ij,3 is the wall-reflection
the effects of streamline curvature, swirl, rotation, and rapid
term which is responsible for the damping of the velocity
changes in strain rate in a more accurate way than one-equation
fluctuation component close and perpendicular to the wall
and two-equation models [200, 201]. So, this model has more
that is included by default in the Reynolds Stress Model.
significant potential to give accurate predictions for complex
Each component of the pressure strain can be formulated as
flows, especially when the flow has heavy swirls and aniso-
following Eqs. (45–47):
tropic turbulence inside aerodynamic systems such as cyclone
devices [197, 202–205]. The algebraic transport equation of [ ]
𝜖 � � 2
the Reynolds stress 𝜌ui uj tensor can be given in Eq. (36), while
′ ′ 𝜙ij,1 = −C1 𝜌f ui uj − 𝛿ij k (45)
k 3
the description can be drafted in Eqs. (37–44) [206–208]:
[ ( )]
( ) 2 � � 𝜕ui
𝜕 � � 𝜙ij,2 = −C2 Pij − 𝛿ij −ui uj (46)
𝜌f ui uj + Cij = DT,ij + DL,ij + Pij + Gij + 𝜙ij + 𝜖ij + Fij + Suser 3 𝜕xj
𝜕T
(36)
where the first term from the left side is the local time deriv- ( ) C k 32
ative of Reynolds stress tensor, Cij is the convection; DT,ij is 𝜖 � �
� 3 � � 3 � �
𝜙ij,3 = C1 uk um nk nm 𝛿ij − ui uk nk nj − uj uk nk ni 1
the turbulent diffusion, DL,ij is the molecular diffusion, Pij k 2 2 𝜀d
( ) C k3∕2
and Gij are the stress and buoyancy production rate of ui uj , 3 3
′ ′ �
+ C2 𝜙km,2 nk nm 𝛿ij − 𝜙ik,2 nk nj − 𝜙jk,2 nk ni 1
respectively, 𝜙ij is the pressure strain consists of three 2 2 𝜀d
(47)
13
where nk is the xk component of the unit normal3∕4to the wall,d (| |)

𝜔i
𝜎 = di | |
1
C
is the normal distance to the wall, and C1 = 𝜇k . According | | (51)
2 |vi |
to some applications, four terms can be neglected, such as
the convection, the molecular diffusion, and the production The Saffman lift force FS which generated in the normal
by system rotation, where the buoyancy production can be direction due to fluid shearing field around the spherical
equals zero for isothermal flow. particle can be proposed by Eq. (52) [214–217]:
| 𝜕v | 2
1
( )
( )( )1 2 | a| 𝜕va
5 CFD–DEM Coupling FS = 1.615 va − vi 𝜌a 𝜇a di CLS |
2
| Sign (52)
| 𝜕nij | 𝜕nij
| |
A CFD simulation has to be integrated with a DEM simu- where CLM represented the Saffman shear lift coefficient and
lation to consider the effect of aerodynamic forces, which can be calculated from Eq. (53) [81, 216]:
play an essential role in particle behavior. When a spheri- ��
cal particle is moving through the air, it is usually modeled CLM =
1 − 0.3314𝛾 0.5 exp −0.1Rei + 0.3314𝛾 0.5 Rei ≤ 40
√
by the sophisticated methodology of Maxey-Riley [210], 0.0524 𝛾Rei Rei > 40
which considers the flow field around the particle and the (53)
stresses obtained from the background flow. Several inves- where 𝛾 is the ratio between transitional and rotational Reyn-
tigators such as Casas et al. [211], Ren et al. [40], and olds number as defined in Eq. (54):
Sturm et al. [70] simplified these equations by combining
the interaction of particle’s Collision forces, Drag force, Resh
𝛾= (54)
Magnus lift force, Saffman lift force, and Gravitation 2Rei
force. So, the translation motion of a particle through the
where Resh is the particle Reynolds number of shear flow and
airflow field was governed by the force balance equation,
can be expressed by Eq. (55):
as indicated in Eq. (48) [70]:
dv 𝜌f di2 𝜔a
mi i = FijC + FD + FM + FS + FGb (48) Resh = (55)
dt 𝜇a
where FijC contact forces due to particle–particle and parti- Note that the value of ( FM + FS ) is so-called the fluid-par-
ticle interaction force Fi . The net gravitational force of the
f
cle–wall contacts and can be calculated as mentioned before
in Eq. (3), FD is the drag force resulting from particle–fluid particle immersed in the fluid Fgb was defined as the differ-
interaction (fluid damping force), and can be calculated as ence between particle gravity and buoyancy, as indicated in
mentioned before in Eq. (3), FM is the Magnus lift force Eq. (56):
generated by the relative motion between the rotating parti- ( )
𝜌f
cle and the air which can be mathematically expressed by FGb = mi 1 − g (56)
𝜌i
Eq. (49) [211, 212]:
( ) Later, Guillermo et al. [211] concluded that the parti-
1 𝜔i × vi | |
FM = 𝜌a Ai CLM v
| 𝜔i | | i | (49) cle motion in an aerodynamic system could be numeri-
2 | | cally described through three non-dimensional groups by
where vi and 𝜔i are the relative velocity and the relative rota- assuming the characteristics of length scale Ls and velocity
tional angular velocity between air and particle, respectively, scale Us for the time scale. These non-dimensional groups
and CLM represented the Magnus lift coefficient due to the as formulated in Eqs. (57–59) are mainly depended on the
rotation of the particle and can be expressed by the empirical non-dimensional parameters such as spin parameter due to
formula given by Eq. (50) [40, 212, 213]: particle rotation 𝜎 , Remolds number of particles Rei , particle
Stokes number Sti , and gravitational parameter.
⎧ 2𝜎 Rei ≤ 1 ( )
⎪ Cd Rei Rei
CLM = ⎨ 2𝜎(0.178 + 0.822Re−0.522
i
1 < Rei < 1000 𝜋1 = (57)
⎪ 0.45 + (2𝜎 − 0.45)exp[−0.75𝜎 0.04 Re0.7 ] 10 < Rei < 140 24Sti
⎩ i
(50) ( )
CLM 𝜎, Rei Rei
where 𝜎 is a non-dimensional spin parameter which repre- 𝜋2 = (58)
sents the particle lift due to its rotation and can be calculated 24Sti
according to Eq. (51).
13
Ls scheme to understand the main algorithm of CFD–DEM

𝜋3 = g (59) coupling.
Us2
First, the granular flow was assigned by DEM code to set
where group 𝜋1 represents the drag forces over particle iner- the properties of the particles where each particle placed
tia, group 𝜋2 represents the Magnus lift forces over particle into the grid cell. Second, the fluid flow field characteristics
inertia, and group 𝜋3 represents the gravity over particle were set and resolved by the CFD algorithm to reach the
inertia. convergence. Afterward, the velocity, pressure, density, and
Recently, the rotational motion of the particle in a flow viscosity of the flow at each position in the computational
field has been investigated considering the rolling friction domain are transferred to DEM code in order to generate the
when the particle rotates in a fluid that can be given by coupling interface to compute the forces acting on each par-
Eq. (60) [41, 104, 218, 219]: ticle. Thus, the CFD–DEM coupling interface took the par-
ticle translational and rotational motion data from the DEM
1
Mrij = 𝜌 d5 C 𝜔 |𝜔 | (60) solver and computed the volume fractions and momentum
64 i i R ir | ir |
exchange in the mesh cell of CFD. Consequently, the CFD
where 𝜔ir is the angular velocity of the particle relative to and DEM solvers entered into the cycles of the time steps
the fluid and CR is the rotational resistance factor and can be until the simulation time ended.
calculated according to Eq. (61) Generally, there are three main methods for simulating
the interaction between particles and the surrounding flu-
5.32 37.2
⎧ Re0.5
+ Rer
Rer < 20 ids, depending on the turbulence effect [58, 221]. The first
⎪ r
64𝜋 method simulates how fluids change the flow of the particles
CR ⎨ Rer
20 ≤ Rer < 32 (61)
⎪ 12.9 128.4 with which they come into contact. In contrast, the second
⎩ + 32 ≤ Rer < 1000
Re0.5
r Rer approach simulates how the fluid flow changes due to the par-
ticle movement. Thus, the particle movement is not affected
where Rer is the rotational Reynolds number and can be
by the flow, but the flow around the particle is affected by
defined in Eq. (62)
the presence of particles. The last one considers the fluid
𝜌a di2 ||𝜔ir || streamlines compressed between particles during the simula-
Rer = (62) tion. In this review, we give a little brief about these coupling
4𝜇a
techniques where a complete and very comprehensive review
Casas et al. [211] also employed another reasonably of different CFD–DEM approaches can be found in [62].
approximated equation to describe the rotational motion of Figure 9 illustrates the different regimes of coupling
the spherical particle in the aerodynamic systems, as indi- techniques on a diagram using dimensionless coordinates
cated in Eq. (63): considering the interaction between the particles and tur-
bulence [221, 222]. Nowadays, various configurations of
d𝜔i � √ �
Ii = −𝜇a di3 𝜔i 2.01 + 0.40401 Rer (63) commercial or open-source software packages have been
dT developed for CFD‐DEM coupling simulation, such as
As mentioned before, Coupling Discrete Element Rocky® (DEM package, can be coupled to Ansys Fluent®),
Method (DEM) with Computational Fluid Dynamics (CFD) EDEM® (DEM package can be coupled to Ansys Fluent®),
is considered a powerful and reliable technique for under- PFC™ (DEM package with a built‐in fluid coupling pack-
standing the physical phenomenon of particle movement in age), SAMADII© (DEM for single and multiple GPUs),
the fluid flow field, resulting in optimizing and designing the CFDEM® ( CFD‐DEM code in C + + ; OpenFOAM®
granular matter systems [35, 58, 97]. The CFD–DEM cou- for CFD and LIGGGGHTS for DEM), LIGGGHTS® (
pling approach was firstly proposed by Tsuji et al. [44], and DEM code in C + + on the LAMMPS platform), Yade (In
then followed and developed by many others [47, 93, 96]. C + + for DEM simulations), and MFIX‐DEM (CFD‐DEM
Generally, the granular flow can be defined as a group of code in FORTRAN for fluid‐solid flows).
discrete solid, and whenever it interacts with itself or fluid In the flow regimes diagram, the terminology of symbols
flow, part of the energy will be lost [95]. The granular matter can be defined as: 𝛼i is the volume fraction of particles;
[ 𝛼i = Vol i , N number of particles, Voli particle volume, Vol
N×Vol
has a wide range of industrial applications and is frequently
modeled by coupling CFD–DEM approach [220]. CFD code volume occupied by particles and fluid], Ti is the response time
𝜌d
can provide an interface to generate a set of fluid flow data of particle; [ Ti = 18𝜌i i𝜐 , 𝜐f kinematic viscosity of the fluid], TK
f f √
𝜐
files, which can include air, water, and other gases and fluids; is the Kolmogorov time scale; [ TK = 𝜀f ], Te is the turnover
thus, DEM code characterized the granular matter existing time of large eddy; [ Te = l∕u , l length scale of the energy-con-
in that fluid to predict its behavior. Figure 8 provides a basic taining eddies, u root mean square of fluid velocity].
13
Fig. 8 Schematic diagram of CFD–DEM coupling technique
Ti / Te Ti / TK and the interaction between the particles and turbulence is

104 Parcles Enhance
named as one-way coupling. This phenomenon means that
102
Producon the distribution of the particles depends on the state of fluid
1 102
Negligible Effect
on Turbulence
turbulence and the momentum transfer from particles to tur-
bulence. This phenomenon has an insignificant effect on the
Parcles Enhance
10-2 1 Dissipaon flow due to the low discrete particle concentrations. Con-
sequently, the movement of the particle is only influenced
10-4 10-2 αi
by the fluid flow, where its velocity is considered to be the
10-7 10-5 10-3 10-1 1
ONE-WAY TWO-WAY FOUR-WAY
same as the surrounding fluid [221]. Numerically, the solu-
COUPLING COUPLING COUPLING tion methodology starts with CFD calculations of the veloc-
Fluid Parcle Fluid Parcle Fluid Parcle Parcle
ity and pressure distribution profiles generated by the fluid
Dilute Flow Dense Flow
flow through the domain. Since the effect of the particles
on the fluid is neglected, the only steady-state simulation
Fig. 9 Diagram of flow regimes depending on the interaction between needs to be performed [223]. At the end of CFD simulation,
particles and turbulence [221, 222]
or once a steady state has reached the flow field, the CFD
obtained data is then exported out and imported into DEM
5.1 One‑Way Coupling code. Thus, DEM calculates how the fluid flow would affect
the particle flow by applying the external forces [35]. This
In the first interval of the flow regime, when the value of method is particularly useful and ideally recommended in
the particle volume fraction becomes very low ( 𝛼i ≤ 10–6) many investigations, especially for large particle separation
the particles have infinitesimally effect on fluid turbulence, process and unconfined homogeneous flows that has differ-
ent densities of particles [186, 221, 224].
13
Potapov et al. [225] simulated the multiphase flow of fraction, when the diameters of the particles are small with
gas and iron particles around the transfer chute using mov- the same material and fluid viscosity lowering ( Ti is low), the
ing porosity technique with DEM Rocky® and CFD Ansys surface area of the particulate phase increases resulting in
Fluent® one-way coupling. The authors mainly used this increasing the dissipation rate of turbulence energy. On the
technique because of the relatively large and dense particles other hand, increasing Ti for the same value of 𝛼i (inclined
(+ 4 mm). The approving comparison of the model predic- line), the particle Reynolds number increases, and the vortex
tion against experimental results shows the validity of this shedding takes place resulting in promoting the turbulence
model for the prediction of airflow around the industrial size energy production.
of transfer chutes. Fonte et al. [53] flowed the strategy of The simulation scheme used in this coupling technique
one-way coupling of DEM Rocky® and CFD Ansys Fluent® is a transient state simulation that allows both solvers of
to separate light from heavy particles in the wind-shifter CFD and DEM to run in a parallel way that can significantly
device using different drag laws at a fixed value of airflow. decrease the simulation time [223]. This coupling method-
Course particles with different geometries of paper (flat), ology is suitable for dealing with a large number of parti-
stone (sphere), metal (rod), and wood (briquette), were used cles through a CFD cell and not for dealing with particles
to determine the suitable drag low to use while simulating larger than CFD cells. Therefore, it is recommended that the
the separation process. The simulation showed that Ganser domain mesh size should be larger than the largest particle
drag model [226] gives more accurate results because it con- size through the simulated domain [58, 227]. The significant
siders the particle geometry. advantage of the two-way coupling simulation is that particle
Almeida et al. [29, 52] also used the one-way coupling agglomeration can alter airflow trajectories and velocities
method of DEM Rocky® and CFD Ansys Fluent® to put a and vice versa. Thus, the interaction between solids and gas
theoretical algorism for separating the sugarcane bagasse is considerably closer to actual phenomena [29].
into different groups of particles. Different air velocities Zhong et al. [156] carried out a three-dimensional simu-
through an aerodynamic conveying system were used con- lation of cylinder-shaped particles in a gas–solid fluidized
sidering the contact forces between particle–particle and bed using a two-way approach considering drag force, con-
particle–wall. The result showed that this coupling tech- tact force, and gravitational force. The predictive behavior
nique perfectly describes the separation process with rea- of particles and pressure drops at different superficial gas
sonable accuracy. Since the main recommendation of this velocities were in good agreement with the experimental
work is that the separation process will be more efficient results. Ren et al. [40] study the flow behavior of corn-
and real when using coarse and large particles. El-Emam shaped particles in the cylindrical spouted bed with a coni-
et al. [35] perfectly applied the above recommendations of cal base device using this coupling technique. The obtained
DEM Rocky® and CFD Ansys Fluent® codes to simulate results demonstrate that the coupling model methodology is
and optimize the separation process of real-particles of valid and useful in describing the hydrodynamic behavior of
Jojoba seeds and its leaves. Five geometries of gas-cyclones the spouted bed.
were used, and the particle interaction was considered. The Fonte et al. [53] set a fluidized bed simulation case to
simulation can successfully predict the separation efficiency apply the two-way coupling technique of DEM Rocky®
and the behavior of the particles that help to optimize the and CFD Ansys Fluent® using Huilin and Gidaspow drag
cyclone geometry and performance. law. The agreement between the pressure drop predicted by
Ergun’s correlation and the pressure drop obtained in the
5.2 Two‑Way Coupling simulation is relatively good until fluidization takes place,
which states the capability of the fluid being affected by
When the value of the particle volume fraction is located a reactive force. Devarampally [186] developed a two-way
between 10–6 ≤ 𝛼i ≤ 10–3 and the particle loading volume is coupling of CFD–DEM using Simcenter STAR-CCM+ code
large enough to affect the turbulence structure, the term two- for a top spray fluidized bed granulator to study the effects
way coupling used. In this regime, the exchange of mass, of inlet air flow rate, the temperature on the particle dynam-
momentum, and velocity information between the solid ics and the residence time in the spray zone. The developed
phase and the fluid phase occurred, where the meaningful model was able to predict the changes in particle velocities,
feedback is noticed in turbulence damping due to buoyancy temperatures, collision dynamics, and particle transfer from
effects [186]. So, it can say that the particles are part of the one compartment to the other as the inlet velocity and the
fluid flow and will affect it in a two-way interaction. This temperature of the air changes. This model can be recom-
phenomenon means that the particle movement is affected mended to determine the agglomeration rates in a granula-
by the interaction with other particles and the fluid around tion process, develop hybrid CFD-DEM-PBM that describes
it, while the flow is also affected by the particle presence. As the rate processes in granulation, and understand the effect
described in (Fig. 9), at a specific value of particle volume of process parameters on the product quality attributes.
13
Carlos Varas et al. [101] employed a CFD–DEM two- is appropriate to simulate aerodynamic conveying of spherical
way model to simulate the co-current gas-particle flow solid particles in a horizontal channel device with the dense
in a pseudo 2-D dilute flow riser. The model reasonably flow. Kuang et al. [231] presented a 3-D numerical study of
predicted the core-annulus flow and other hydrodynamic vertical aerodynamic conveying by combining a four-way
parameters such as cluster frequency, internal solids vol- CFD–DEM approach for dilute and dense gas–particles flow
ume fraction, and aspect ratio that showed close agreement regimes. The numerical method can satisfactorily capture the
with experimental data. Baran et al. [155] applied a fully key flow behaviors observed, such as particle flow pattern and
coupled CFD-DEM two-way approach to study the effect of gas pressure characteristics. In the dense-phase flow regime,
particle shape on fluidization in a cylindrical fluidized bed the simulation results confirmed that the particle–particle and
using Simcenter STAR-CCM+ software. The observations particle–wall force increases with the increase of the parti-
are consistent with the similar simulation and experimen- cle–fluid force, but in contrast for dilute-phase flow where it
tal results reported in the literature [228, 229], which gives is small, if not negligible.
strong evidence for applying this technique to predict the Sturm et al. [70] extended the previous simulation to
dynamic behavior of different particles shape in the fluidized also show the applicability of the presented simulation
bed. Almeida et al. [29] simulated the dry bagasse separa- model to the different flow regimes: a dense vertical aero-
tion in a pneumatic device using DEM Rocky® and CFD dynamic conveying system and dilute conveying in a hori-
Ansys Fluent® coupling approach. The comparison with the zontal-to-vertical pipe bend. To track the flow behavior, the
experiment showed that the developed model predicted the Euler–Lagrange approach is presented, and an in-house soft
particle separation phenomena with high accuracy. sphere DEM code is coupled to Ansys Fluent®. It is shown
that the different aerodynamic conveying regimes can be
5.3 Four‑Way Coupling behaved well comparing to the experimental data. Traoré
et al. [232] also derived a complete four-way coupling model
Herein, the four-way coupling term is developed due to the of a gas–solid 3-D fluidized bed by taking into considera-
interactions between particle-turbulence, particle–fluid- tion particle–particle, fluid-particle, and particle–wall inter-
particle, particle–particle, and frictions of particles., In this actions. The simulation constructed on a high dense flow of
case, the fluid streamlines between particles are compressed 2.7 million particles in two different ways: pulsed jet and
where the flow is considered to be dense [221]. This means bubbly regime. The output data perfectly pointed out the
the particle loading is very high and concentrated where the existence of complex structures such as the worm-like shape
value of particle volume fraction is assumed to be higher one, similar to those that have already been described in
than 10–3. Due to the complexities of flows in this regime, other literature. Considering the above results, it is shown
most experimental and numerical studies considered the that a definite proof to use this coupling scheme for predict-
flow to be dilute. As can be seen in (Fig. 9), the line which ing sensitive phenomena in the large dense granular systems.
separates the two-way and four-way couplings are inclined
as the particle–particle collision occurs at higher values 6 Drag Correlation Models
of Ti ∕Te , thus transforming the two-way to four-way cou-
pling technique. Note that, when particle volume fraction During the run of CFD–DEM coupling in an aerodynamic
approaches to l, a total granular flow is presented in which system, the drag correlation model shows vital importance
there is no fluid will be obtained between particles. In this besides pressure gradient forces, especially for describing
coupling approach, the simulation technique runs on a par- and dealing with the interaction between particle shape and
allel implementation of CFD and DEM, which can signifi- concentration and the surrounding fluid [139, 233, 234]. For
cantly stimulate the two-phase fluid–solid phase. successful coupling, it is crucial to choose a suitable drag
Laín and Sommerfeld [230] predicted the behavior and correlation model [53]. There are several drag correlation
the pressure drop in a particle-laden horizontal channel with models developed and available in the literature. Most of
rectangular cross-section by the Euler–Lagrange approach them are mainly dependent on the Reynolds number, where
accounting for two-way and four-way coupling. Different others need other properties, such as fluid volume fraction
diameters and mass loading ratios of particles were considered and particle sphericity. Here, in this review, we will con-
where the particle–wall interactions and inter-particle colli- centrate the attention on the most common correlations that
sions have also been taken into account. The numerical simula- can be used in modeling bio-particles based on the flow
tion technique has reasonably reproduced the studied trends; regimes. The attempts by investigators and the developments
however, the numerical predictions of wall-bounded flows can- for models listed in Table 2 illustrate the applicability of this
not be appropriately applied when neglecting the wall rough- approach.
ness effects. Comparing the obtained data with the experiment
result, it can be somewhat indicated that the considered model
13
Table 2 Drag correlation models used in CFD–DEM coupling

Flow regime Models and Reference Equations Notes
Dilute flow Schiller and Naumann 24 � �

For spherical particles
⎧ 1 + 0.15R0.687 Rei ≤ 1
Rei �
��
ei
(1933) [53, 100, 205, ⎪ 24 �
235–237] C D = ⎨ max R 1 + 0.15Rei 0.687
1 < Rei ≤ 1000
⎪ ei
⎩ 0.44 Rei > 1
� �
Dallavalle (1948) [238] C = 0.63 + √4.8 R < 3000 For spherical particles
D ei Rei
Haider and Levenspiel 24 ( ) C For isometric non-spherical particles and

CD = 1 + ARBei + Rei < 2.6 × 105
1+ RD
(1989) [28] Rei
ei non-isometric non-spherical particles
For spherical particles:
A = 0.1806 , B = 0.6459 , C = 0.4251 , D = 6880.95
For non-spherical particles:
( )
A = exp 2.3288 − 64581�i + 2.4486�2i
B = 0.0964 + 0.5565�
( )
C = exp 4.905 − 13.8944�i + 18.4222�2i − 10.2599�3i
( )
D = exp 1.4681 + 12.2584�i − 20.7322�2i + 15.8855�3i
Ganser (1993) [29, 52, [ ( )0.6567 ]
Cd 24 0.4305 Valid when Rei K1 K2 < 105 Used when the
= 1 + 0.1118 Rei K1 K2 +
53, 226, 239] K2 Rei K1 K2 1 + R 3305 particles in the domain differ in shape,
ei K1 K2 concentration, and alignment
( )−1
di 2 −1 di
K1 = + �i 2 − 2.25 vol
3divol 3 D
0.5743
K2 = 101.8148(−log10 �i )
{ [ ( 0.687 )0.687 ] −1.65
Dense flow Wen & Yu (1966) 24
1 + 0.15 𝛼f Rei 𝛼f 𝛼f Rei < 1000 Relatively for low particle concentra-
[240, 241] CD = 𝛼f Rei tion:𝛼i < 0.2
0.44𝛼f−1.65 𝛼f Rei ≥ 1000
Ergun (1958) [100, 𝛼

CD = 200 𝛼 �2pR + 7 For high particle concentration:𝛼i ≥ 0.2
139, 242, 243] f i ei 3�i
Huilin & Gidaspow CD = 𝜑CD,Ergun + (1 − 𝜑)CD,Wen&Yu For particle concentration:0.2 > 𝛼i ≥ 0.2
(2003) [244] [ ( )]
arc tan 262.5 0.8 − 𝛼f
𝜑= + 0.5
𝜋
Di Felice (1994) [29, CD = 𝛼f2−𝛽 CD,singleparticle 10−2 ≤ Rei ≤ 104

100, 223, 243, 245] { [ ( )]2 }
1.5 − log10 𝛼f Rei
𝛽 = 3.7 − 0.65exp −
2
7 Applications to construct the separation, cleaning, sorting, grading, and

density classification processes in one machine. Recently,
To model the granular flow in the agricultural aerody- with the increase of agricultural industry applications con-
namic systems, various mathematical and numerical mod- currently with the high power of computer technology,
els include theoretical approaches, differential equations CFD–DEM coupling approach is frequently used to model,
of motion, incremental technique, and computer program- predict the behavior, and provide the solutions of the air-
ming were used. For instance, by MATLAB modeling, particle flow problems [80, 246].
Adewumi [79] started to simulate the flow of cowpea
grains through the transport and separation process in a 7.1 Air‑Screening System
horizontal airstream. He developed a 2-D Modeling using
the aerodynamic principles of drag and gravitational forces The process of separation and cleaning grain materials from
to provide a clear vision about the air-particle dynamics impurities such as spikelet, leaves, straw remnants, husks,
in such a system. The obtained results showed the ability and small weed seeds, etc. is essential in agricultural grains
13
(a) 2.5
(b) 3.5
Cleaning loss rao Fing curve of cleaning loss rao
Grain impurity rao Fing curve of Grain impurity rao
2.0 3.0
Cleaning performance, %
Cleaning performance, %
1.5 2.5
1.0 2.0
0.5 1.5
0.0 1
-0.5 0.5
0 1 2 3 4 1.0 1.1 1.2 1.3 1.4 1.5
Throughput, Kg.s-1 Fan speed, 1000 rpm
Fig. 10 Polynomial diagram of (a) throughput and (b) fan speed with cleaning performance [249]
production. In modern harvesting machines such as com- results as described in (Fig. 10a,b) showed that the fan speed
bine harvesters, the separation and cleaning processing are have a great effect on cleaning performance by affecting the
mainly implemented by air-screening or air-sieving systems overall airflow velocities in the cleaning device where the
[247]. In such a system, the modeling process can be numer- throughput has a sharp effect on grain impurity ratio.
ically described as a complex system of two-phase flow. One Furthermore, Lenaerts et al. [77] modeled the separa-
phase is the airflow generated by a centrifugal fan, and the tion chamber in the combine harvester through a cubic box
other is solid particles dispersed in the system by the airflow to predict the separation profile of the grain-straw using
[248]. As mentioned before, the effect of air velocity has DEM simulations. The grains were assumed to be spheri-
significant influence in such a separation process that the cal, while the straw particles are modeled as a segmented
airflow entrains the light particles due to the aerodynamic hollow cylinder that can be bent in all directions using
drag force. At the same time, the heavy ones are mainly DEMeter + + software. The main prediction is that separa-
separated by the force of gravity and the inertia of sieves. tion speed is influenced by particle properties, but the main
Numerically, Li et al. [39] carried out a 3-D coupled of effects can be related to straw coverage and grain diameter.
CFD–DEM approach in an air-and-screen cleaning unit to Ma et al. [76] developed another variable-amplitude screen
study and analyze the influence of inlet airflow velocity on box to analyze the traditional reciprocating screen in the
rice grain and short straw taking into consideration the lon- combine harvester. The particles used in their study were
gitudinal velocity, vertical height, and cleaning loss. The rice divided into two groups: rice grain and rice straw that are
grains were modeled as a spheroid particle, and the short modeled as ellipsoidal and cylindrical, respectively, by a
straws introduced as a cylinder shape. The particle materials multi-sphere approach. Based on the Hertz-Mindlin contact
were modeled and generated in EDEM software and allowed model in EDEM software, the simulation showed that the
to fall on an inclined vibrating screen. A Eulerian–Eule- variable-amplitude screen could make the particles at the
rian model was used in simulating the airflow using Ansys front of the screen be quickly thrown up and moved back
Fluent® as a part of a coupling, where the Hertz-Mindlin but could not favor their segregation and separation at the
contact model was applied to simulate particle–particle and screen front.
particle-screen (wall) collisions. The simulation results On the other hand, [250] studied the grain behavior in an
reported that the length of the screen could be shortened if inertia separation chamber of the combine harvester based
material straws and foreign materials are not too much since on CFD–DEM simulation. The grains were modeled as ellip-
the higher the inlet airflow velocity, the faster the backward soidal, while the straw modeled as cylindrical shape through
velocity of the materials on the screen surface and the more the multi-sphere approach. The k − 𝜖 turbulence model was
the cleaning loss. This work is extended by Li et al. [249] to used in Ansys Fluent® software to reach the convergence
simulate the effects of throughput and operating parameters state of the airflow while the Euler-Lagrangian method in
on cleaning performance in the same device. Ansys Fluent®, EDEM code was applied to model the particles. The simu-
as an auxiliary investigation software, was used to analyze lations showed that the grains were effectively separated
the movement behavior of the internal airflow. The stand- from the short straw through the airflow. Besides, the per-
ard k − 𝜖 turbulence model is proposed for this simulation formance of separating and cleaning grains was improved
study to describe the real-physical processes at the different with the increase of airflow velocity. However, turbulence
airflow velocities above and below the vibrating screen. The
13
Fig. 11 The particle trajectories of the rice-threshed mixture during the simulation process [218]
Air velocity 8.64 m.s-1 Sieve diameter 250mm Rotaon speed 21rpm
80 Air velocity 9.67 m.s-1 60 Sieve diameter 300mm Rotaon speed 26rpm
70
-1
70 Air velocity 11.34 m.s Sieve diameter 300mm Rotaon speed 31rpm
50 60
60
50
Grains, %
Grains, %
40
50
40
40 30
30
30
20
20 20
10 10
10
0 0 0
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
Collect areas Collect areas Collect areas
Fig. 12 The mass distribution of the grain during the screening process [218]
and pressure drop were severely increased at the end of the screening efficiency tendency when the cylindrical sieve
separation chamber. diameter increased. At the same time, the rotating speed
Yuan et al. [218] investigated the screening process of has a significant effect on the average axial velocity of the
rice-threshed mixture components (grains, shriveled grains, short stalks, screening loss rate, and screening efficiency.
and short stalks) through a cylindrical sieve device using Figure 12 showed the mass distribution in every collect area
CFD–DEM coupling. The influence of inlet airflow veloc- (from 1 to 5) as a relation of inlet airflow velocity, cylindri-
ity, the diameter of, and the rotating speed of the cylindrical sieve diameter, and its rotational speed.
cal sieve on the separation process was studied. First, the As shown in the investigations, the technique and the
particles were scanned by a 3-D laser scanning system then, methodology of the numerical simulations used for mod-
modeled by multi-balls fast filling method (API) in EDEM eling the air-screening systems can provide a useful ref-
software and allowed to move smoothly through the cylin- erence for predicting the flow behavior and solving the
der sieve. The particle phase is considered to be in a con- separation problems of biomaterials. Because CFD–DEM
tinuous quasi-fluid phase and solved through the average coupling models had a reasonable predict comparing to the
Navier–Stokes equation, where the airflow was simulated experimental results in the above investigations, it is strongly
through the existing Eulerian–Eulerian model in the Ansys recommended to be used for improving and optimizing the
Fluent® code. To track the separation process (Fig. 11), vis- design of air-screening devices in the combined harvesters.
ualized the simulation of particle trajectories with the time
though the device, where the yellow, green, and blue colors
represent grains, shriveled grains, and short stalks, respec-
tively. The numerical simulation showed an increase in the
13
Velocity m.s-1 Velocity m.s-1 Velocity m.s-1 Velocity m.s-1

0.820 1.070 1.110 1.390
0.656 0.859 0.890 1.120
0.492 0.644 0.667 0.837
0.328 0.430 0.445 0.558
0.164 0.215 0.223 0.280
0.003 0.003 0.003 0.003

End of seed-filling process Start of seed-cleaning process End of seed-cleaning process Seed state during pressing process
Fig. 13 Seed motion behavior of the seed-cleaning process [42]
7.2 Seed‑Metering Device the position is lower, the optimized lateral hole, which has
a width of 1.5 mm and an average arc length of 10 mm, had
Precision agriculture made agricultural production and agri- an excellent working performance, with less pressure loss,
cultural management more accurate and useful [251]. Pre- superior drag force, and differential pressure. The perfor-
cision seeding is an essential part of precision agriculture, mance of the optimized seed-metering device in the experi-
which has always attracted the attention of researchers [252]. ments reached 93% when the working pressure was above
It can be defined as the placement of single seeds in the 5.5 kPa with a working speed lower than 10 km.h−1.
soil at the desired plant spacing [253]. The Seed-metering Lei et al. [81] studied and analyzed the effects of the
device considers the main component of the planter machine throat’s structure, airflow inlet velocity, and seed feed rate
to represent adequate seeding performance. Although there in seed feeding device (pneumatic Venturi tube) in terms of
are various types of seed-metering devices, each with unique air and seed behavior using CFD–DEM coupling approach.
individual working principles, aerodynamic seed-metering This feeding device is usually used in the pneumatic seed-
devices are used more widely because of their high accu- ing planter machine. Rapeseed and wheat were used and
racy, preservation of seed integrity and fitness for high-speed modeled as spheres and multi-sphere, respectively, using
operation [254, 255]. Recently, the numerical simulation EDEM software. Hertz-Mindlin model was chosen as par-
approach has effectively applied for modeling the air-seeds ticle modeling while the gas-phase was solved with the
flow in the aerodynamic seed-metering devices to promote standard k − 𝜖 two equations turbulence model. The cou-
effectiveness and performance. pling interface was implemented by the Lagrangian coupling
For instance, Dandan et al. [256] applied the coupling method. Simulation results showed that the throat’s area and
method of CFD–DEM to simulate the effect of different air airflow inlet velocity mainly affected airflow outlet velocity,
nozzle installation positions on the working performance while seed movement was slightly affected by throat length
of inside-filling air-blowing precision seed metering device. and seed feed rate. The increase in the throat area resulted
They concluded that CFD–DEM simulation could detect in a decrease in seed’s velocity and pressure loss in a cer-
and predict the maize seed movement behavior in a single tain range. When the inlet air velocity increased, resultant
seed meter. Han et al. [42] extended their work to simulate force and seed velocity developed. Figure 14a,b showed the
and optimize the air-maize flow in the same device using simulation flow patterns of rapeseed and wheat seed at the
the same numerical methodology. The effects of positions, designed air velocities. It can be concluded that the appropri-
width, and average arc length of the lateral hole were stud- ate airflow inlet velocity was 16–20 m.s−1 and 20–24 m.s−1
ied in terms of the gas field and seed movement. The maize for rapeseed and wheat, respectively.
particles were regarded as a hard-sphere model and created Although Venturi tube was widely used in industry, lim-
by EDEM software using the bonded particle model (BPM) ited numerical studies are available in the literature, espe-
technique, which allows the particles treated as a soft sphere cially for particle motion in aerodynamic seeding systems.
model. While the standard k − 𝜖 two-equation turbulence Lei et al. [257] extended their study to characterize the
model was used to solve the gas phase and the Eulerian two- rapeseed and wheat migration trajectory and predict the
fluid coupling approach was applied. distribution behavior of a pneumatic distribution head in
According to the numerical simulations, the seed motion the air-assisted centralized seed-metering device (Planter).
behavior of the seed-cleaning process can be correctly pre- The streamlined angle, the radius of baseline, outlet diam-
dicted, as illustrated in (Fig. 13). The simulation results also eter of streamlined angle, lid angle, and installation loca-
indicated that the area of the lateral hole seriously affected tion in the distribution head were the main parameters to
the movement and airflow field of seeds in the hole. When investigate in this study. The particles were regarded as
13
(a) Velocity m.s-1 Velocity m.s-1 Velocity m.s-1 Velocity m.s-1

7.8661
2.5072 4.2427 7.2348
2.0064 3.4023 5.7960 6.2932
1.5056 2.5619 4.3572 4.7203
1.0048 1.7216 2.9184 3.1474
0.504 0.8812 1.4796 1.5745
0.0032 0.0408 0.0408 0.0016

Velocity m.s-1
(b) Velocity m.s-1
2.5222
Velocity m.s-1
4.244
Velocity m.s-1
5.5765 6.1515
2.0245 3.3994 4.4694 4.9254
1.5266 2.5549 3.3623 3.6994
1.7103 2.2551 2.4733

1.0286
0.5306 0.8657 1.148 1.2473
0.0327 0.0212 0.0408 0.0212

Air flow 8 ms-1 Air flow 12 ms-1 Air flow 16 ms-1 Air flow 20 ms-1 Air flow 24 ms-1
Fig. 14 Flow patterns of (a) rapeseed and (b) wheat seed at different air velocity [81]
(a)
Velocity m.s-1 Velocity m.s-1

11.543 11.543
9.234 9.234
6.925 6.925
4.617 4.617
2.308 2.308
3.004e-05 0.004
streamlined angle 60º Radius of baseline 40 mm

(b)
Velocity m.s-1 Velocity m.s-1 Velocity m.s-1 Velocity m.s-1

9.262 10.569 11.131 13.154
7.41 8.455 8.905 10.526
5.557 6.341 6.679 7.898
3.705 4.227 4.452 5.271
1.852 2.113 2.226 2.643
3.834e-06 3.161e-06 3.04e-06 0.0163
Air velocity 24 m.s-1 Air velocity 28 m.s-1 Air velocity 22 m.s-1 Air velocity 24 m.s-1
(Wheat) (Rapeseeds)
Fig. 15 Snapshots of wheat and rapeseeds simulation in the distribution head [257]
a hard-sphere model where the Hertz-Mindlin (no-slip) Furthermore, the resultant seed force and resultant veloc-
model was applied as the particle contact model, and the ity increased with the increase of airflow velocity. Second, as
seed feed rating was adjusted by EDEM software. The showed in (Fig. 15a), the distribution of the seed was more
rapeseeds were modeled as the spherical shape where uniform when the streamlined angle is 60°, the radius of the
the wheat seeds were generated as ellipsoidal shape by a baseline is 40 mm, and the lid angle is 120°. Also, the dis-
multi-sphere approach. To model the turbulence in the air- tribution uniformity and power consumption improved at the
flow, the standard k − 𝜖 turbulence model and the airflow airflow velocity of 20–24 m.s−1 for rapeseed and 24–28 m.
velocity were set in the Ansys Fluent® software. Simula- s−1 for wheat of different seed feed rates, as illustrated in
tion results first predicted the variation coefficient of seeds (Fig. 15b).
distribution in the head. With the increase of streamlined The obtained result approved that the CFD–DEM cou-
angle from 10° to 50°, the variation decreased initially pling approach is a reliable and feasible tool for simulating
and then increased for an inlet diameter of 20 mm and the physical phenomenon of seed movement in the airflow
an airflow velocity of 20 m.s−1. Also, when the radius of field. Also, It can help improve distribution uniformity,
baseline was less than 20 mm, the air pressure loss and explain the seed distribution mechanism, assist the per-
resultant seed force visibly increased. formance, and optimizing the design of air blowing seed
devices.
13
(a) (b)
(d)
2
1.8
Average parcle velocity, m.s-1

1.6
1.4
1.2
(c)
1
0.8
0.6
0.4 Wedge - 40
Wedge - 60
0.2 Wedge - 80
Circular
0
0 100 200 300 400 500 600
Distance from dryer module wall, mm
Fig. 16 Simulated velocity differences around the different air ducts (right): a wedge duct, b circular duct, c velocity distribution of the wheat
(red higher, blue lower velocities), and (d) velocity differences from the dryer wall [67]
7.3 Drying Systems design imperfection. The dryer geometry and different air
duct arrangements (horizontal and diagonal) were studied.
Drying technology is widely applied in agriculture to pre- Wheat particles were chosen and modeled in two geom-
serve large mass flows of food grains, seeds, kernels, and etries as an ellipsoidal clump and spherical by Particle
other macroscopic particulate matter after harvesting in Flow Code (DEM-PFC 2-D) software to compare the par-
farms [258, 259]. For continuous drying, which is common ticle flow simulation. The airflow domain was discretized
in the industry, mixed-flow or cross-flow dryers (MFDs) are by generating a finite volume grid employing the software
widely used [260]. The moisture is removed from the grain ANSYS ICEM. Then, the airflow was regarded as a turbu-
by forced dry air convection through the grain bed using lence flow and modeled by ANSYS CFX® software using
air ducts installed in the bed, in order to prevent microbial the two-equation shear stress transport (SST) model. The
activity and spoilage. simulation data successfully captured the particle-air flow
Keppler et al. [67] analyzed the influence of different distribution inside the dryer, comparing to the experi-
constructional modifications of air duct shape and wall fric- mental result. It was found that the drying is not perfect
tion on the velocity distribution difference of wheat particle in the traditional mixed-flow dryers because of air duct
within the MFD module through the EDEM code package. arrangements, the appearance of dead zones in airflow, and
Two types of modified air duct shapes were used: wedge irregular distribution of airflow due to the wall frictions.
and circular ducts, as shown in (Fig. 16a,b). The particles In this context, appropriate air duct arrangements should
were modeled by a multi-sphere approach as a clump of be fixed to adjust the higher and lower air velocities in
three spheres, where the Hertz-Mindlin (no-slip) contact the middle part and near to sidewalls of the dryer, respec-
model was used to model the interaction forces. The sim- tively. Consequently, the current authors recommended a
ulation data represented substantial velocity differences new dryer model to adjust the airflow distribution; conse-
when using the circular air ducts, so it is not appropriate in quently the particle flow velocities that the sidewalls of
wheat drying usage. Also, significant velocities difference the dryer should be inclined, and the half air ducts should
observed depending on particle place in the module. While, be removed.
the near-wall is much lower than in the middle, as shown in So, Weigler et al. [261] worked in the previous recom-
(Fig. 16c,d). So, using smooth structure materials lead to mended model to develop a new geometry of MFDs that
decreasing the particle–wall friction ratio resulting in more results in the homogeneous particle drying, higher energy
even velocities distribution. By this, it would be easy to get efficiency, and increase product quality. The wheat flow was
an excellent homogeneous drying for the particles. tracked by the different grain colors through a transparent
To avoid extra drying costs and improve the prod- sidewall using DEM-PFC2-D. Based on the experiments and
uct quality, Weigler et al. [69] evaluated the traditional numerical simulation observations, the new design meets
designs of MFDs employing CFD–DEM to discover most requirements of the particles drying.
13
Fig. 17 Particles flow patterns in MFD: a particles after DEM gravitational generation, b DEM sphere particles, c DEM ellipsoidal clamp, d
experimental result [258]
(a) (b) Velocity m.s-1

0.2
0.16
0.12
0.08
0.04
0.0025
Fig. 18 a comparison of soya bean discharge for simulation (left frame) and experiment (right frame) at discharging hole 20 mm, and (b) veloc-
ity field distribution [37]
On the other hand, Weigler and Mellmann [258] discov- leads to a non-uniform moisture distribution at the dryer
ered without airflow the dry wheat kernels behavior through outlet. Consequently, developing a new design that can
the traditional MFD of Weigler et al. [69] using DEM-PFC2- adequately evanesce the difference in residence time of the
D where the results were compared with the experiments. A particles is still need. Moreover, the main recommendation
multi-sphere approach shaped the wheat kernels as an ellip- is that the 2-D DEM model does not accurately predict the
soidal clump and its flow behavior compared to the spherical absolute values of the grain mass flow pattern and veloc-
shape. Alternately, the particles were colored black and blue ity. Therefore, a three-dimensional simulation should be
to observe the mass flow behavior in the 2-D view, as shown applied for dryer modeling.
in (Fig. 17a–d). As illustrated, the DEM model based on Hence, Khatchatourian et al. [37] carried out the first
the ellipsoidal particle can effectively predict the real-grain trial to develop a 3-D model of soya beans flow apply-
flow behavior that the colored grain layer in the experiment ing DEM-Yade software package to predict the behavior
showed the same profile as in the numerical simulation. In and motion of particles during the discharging process
contrast, when spherical particles were used, the bed angle in MFDs. The particles were modeled as a 3-D single
that was typically formed under the air ducts did not appear. spherical geometry considering the physical properties.
The simulation findings which are entirely agreed with An experiment was carried out to validate the simulation
the experimental results promoted the main flow features data that showed a satisfactory agreement between their
of particle flow: the higher flow velocities were through results, as shown in (Fig. 18a).
the center of the dryer and had lower friction effects than By Applying the method of colored particles for continu-
those flowing through the regions near the dryer walls. ous flow in the MFD, the differences of soya bean velocities
These particles traveled with lower vertical velocities due successfully observed. Near the wall regions, seed velocities
to the higher frictional effects, which made the grains have were lower than in the center, as illustrated in (Fig. 18b).
different residence times in the dryer. This phenomenon Thus, the seed velocity field is nonuniform, leading to
13
Parcle diameter mm 0s 50 s 100s 150 s 200 s

(a) 3.5 Relave air velocity
(b)
3.47
Shrinkage, %
3.44
3.41
3.38
3.35
150 0C
3.32
Time, s
Fig. 19 a Snapshot of particle size change throughout the drying process, b effect of inlet air velocity on particle size reduction [263]
different residence times of the particles. This result is typi- with the increase of inlet air temperature but increases
cally agreed with the previous investigation, as mentioned slightly with increasing inlet air velocity.
above. The current authors suggested thoughtfully combin- Azmir et al. [264] continued their research to study the
ing the mass and heat transfer model with the air-particle influence of the grain properties such as size, density, and
flow for better simulation results that lead to optimizing the initial moisture content on the drying process using the same
flow in the dryers. methodology of CFD–DEM coupling. Millet particles were
As a result of that, Scaar et al. [262] extended the work mainly used and compared with the five same genera that
on the developed MFD model recommended by Weigler all of them modeled in shape as spherical particles. Three
et al. [69] using the heat and mass transfer models. Their different inlet air temperatures (50, 150, and 200 °C) are set
study aims to predict the airflow distribution and investi- to predict the effect of different heat transfer modes. The
gate the effect of air duct arrangements. Based on 3-D CFD numerical simulations proved that the convection and con-
ANSYS CFX® code, two different bed materials were used: duction heat transfer has a significant influence on drying,
wheat and rapeseed and considered as a porous medium. while the contribution of the radiation mode has a negligible
The turbulence of fluid flow is modeled by an eddy viscosity effect. With decreasing the grain size, the drying rate expo-
approach using SST (Shear Stress Transport) k − 𝜔 model. nentially increases while the size distribution and the uni-
The numerical observations were validated by experimental formity of moisture content decreases. Besides, decreasing
data which showed that the air duct arrangement has a strong the initial moisture content leads to low drying rate and make
influence on the airflow distribution. In addition, the hori- the moisture content to be homogeneous. Finally, increasing
zontal air duct arrangement showed more homogeneous air- grain density occurs at a lower drying rate with longer dry-
flow pattern as compared to the diagonal one. Furthermore, ing time and a lower uniformity of moisture content.
the numerical model well calculated the influence of the bed A sequel, the main advantage of using computational
material with its different particle characteristics which gives modeling, especially CFD–DEM coupling techniques in
a good support to use this model in future investigations. grain drying, is the ability to study the grain and air veloc-
On the other hand, Azmir et al. [263] established an in- ity distribution through the dryer sections, which affected by
house CFD–DEM drying model for food grains considering geometrical modifications. Based on the previous results, the
the shrinkage characteristics to investigate the influence of coupling approach with mass and heat transfer models will
inlet air temperature and velocity on drying process in the be beneficial in a realistic-industrial scale because under-
fluidized bed. Wheat particles with spherical shape are con- standing grain behavior within the dryer allows analysis of
sidered as a model grain where a 2-D CFD–3-D DEM are drying without requiring an expensive prototype.
coupled by assuming only one control volume in the thick-
ness direction, and the source in the bed thickness direction 7.4 Pneumatic Separation Systems
is negligible. Hot air (100º–180º) is used to dry the grains
at a given velocity (1.2–2 m s−1), which is introduced from The separation process is widely used in process engineer-
the two corners of the bed. The developed drying model can ing, for example, to obtain more homogeneous fractions of
successfully predict the overall particle shrinkage rate and material in terms of size or even in chemical characteristics.
the diameter of each particle during drying the simulation Nowadays, studies have been conducted with the simulation
time, as shown in (Fig. 19a,b). At the studied range of air of fluidization of mineral particles and medicines for equip-
velocity, the particle shrinkage rate increases significantly ment development and process improvement. However, with
the bio-particulate matter field, it still has little studies.
13
14 Air velocity, 2 m.s-1 14 Air velocity, 2 m.s-1 14 Air velocity, 3.5 m.s-1 14 Air velocity, 3.5 m.s-1
(a)
12 12 12 12
Parcle mass, g
Parcle mass, g
Parcle mass, g
Parcle mass, g
10 10 10 10
8 8 8 8
6 6 6 6
4 4 4 4
2 2 2 2
0 0 0 0
Parcle at the boom Parcle at the top Parcle at the boom Parcle at the top
(b) Parcle Group
Group 7
Group 6
Group 5
Group 4
Group 3
Group 2
Fig. 20 a particle mass distributions, b snapshot during the simulation at the bottom and top of the riser device with different air velocities [52]
Almeida et al. [52] established a new computational air velocities form the two coupling approaches provided the
model to separate the sugarcane bagasse particles pneumati- right prediction and showed high feasibility of the separation
cally by a riser device based on CFD Ansys Fluent® and process comparing to the experiment result. The utilized
DEM Rocky® tools. One-way coupling was used referred parameters and the obtained results of this simulation give
to as the Lagrangian–Eulerian method, while the linear hys- a confidant for these coupling models to be extended in sev-
teresis model was applied to model the interaction forces eral processes such as drying, pyrolysis, fluidization, and
between particles and boundaries. For air turbulence flow, gasification. It will also result in studying and developing
k − 𝜖 turbulent model was chosen. The particles were classi- the ideal configurations of these devices.
fied into seven groups based on its characteristics and aspect In this context, El-Emam et al. [35] also applied the one-
ratio: group 1 as pith (fine particles), groups 2, 3, and 4 as way coupling approach of CFD Ansys Fluent® and DEM
fibers (coarse particles), and groups 5, 6, and 7 as the rind Rocky® codes to observe the separation behavior of Jojoba
(coarse particles). The fiber and the rind particles are mod- seeds from its leaves using cyclone separator. The highly
eled as cylinders using Rocky® software. The pith particles curved streamlines and the chaotic turbulence of the contin-
will be neglected in simulation; because the air will entrain uum airflow in the cyclone separator were modeled through
all of them due to the chosen velocities; besides, a lot of RNG k − 𝜖 turbulent model. The particles were modeled in
computational power and time will be consumed, making shape as a sphere and custom polyhedron for the seeds and
the simulations insignificant. leaves, respectively. Also, the interaction forces of parti-
The chosen air velocities were set at 2.0 m s−1 and cle–particle and particle-boundary in the normal and tangen-
3.5 m s−1 based on the calculated terminal velocities of the tial directions were modeled by the Hysteretic linear spring
particles. During the simulation process, the particle mass model and Linear spring coulomb limit model, respectively.
distributions at the top and the bottom of the riser device Four geometrical modifications of the cyclone inlet height
with different air velocities are analyzed and visualized, as were used compared to the original one to optimize the per-
shown in (Fig. 20a,b). As a result of these data, it can be formance. First, to make sure that the modeling methodol-
proved that the coarse particles can be easily tracked during ogy is correct, the performance validation of the original
separation even though the particles have a similar property. model has been carried out that showed a good between the
The current authors continued their investigation using numerical and the experimental data as shown in (Fig. 21a).
one-way, and two-way coupling approaches for dry and wet Second, by tracking the number of separated seeds and the
particles to put a theoretical basis and understand the separa- bottom of the cyclones and counting the number of leaves
tion phenomena resulting in developing and improving the dragged by air at the top outlet numerically, the performance
separation equipment used for that process [29]. The same of the cyclone can be obtained as shown in (Fig. 21b).
methodology of the particle modeling was followed from the As illustrated in (Fig. 21c), the snapshots discovered the
previous study. The mass distribution of dry and wet bagasse disadvantages of the original cyclone that a considerable
collected at the top and bottom outlets of the riser at different number of leaves concentrated in the top portion of the
13
Fig. 21 a validation of the original model, b optimized cyclones performance, and (c) snapshot of particle separation at the end of the simulation
time [35]
cyclone lead to decreasing the performance. So, by modify- being perfect enough to deal with real-particulate systems
ing the cyclone inlet height, the performance significantly in nature or industry accurately. Besides, it is still focusing
improved by increasing the separation efficiency, cleaning on laboratory-scale systems. On the other hand, develop-
efficiency, and effectiveness. This data was useful for con- ing an advanced process for understanding and determining
sidering cyclone (2) as the most suitable modified and opti- contact models of irregular shapes of grains opens up the
mized design geometry for harvesting jojoba seeds with the possibility of computing particle behavior (translation and
lowest operation cost and highest performance. rotation) in 3-D modulation and hence of obtaining high-
resolution impact.
As indicated above, many models have been used for
8 Challenges and Future Issues describing bio-particle shapes and perceiving particle con-
tacts. However, each of these models has its advantages and
Considerable efforts mentioned above have proved that disadvantages regarding the accuracy, flexibility, numerical
CFD–DEM coupling is a high-potential method to inves- stability, and efficiency, considerable challenges still exist
tigate the bio-particulate matter by capturing its behav- in the bio-particles flow models especially when applying
ior characteristics in the aerodynamic systems. However, in the real world. Usually, to compute the contact force of
since these systems are significantly varied and complex, the non-spherical grains, which are common, models that
there are still a significant shortage and many problems in are initially constructed for the ideal spherical particles
terms of determining accuracy, efficiency, and appropriate are directly used. As indicated, the last decade introduced
model of CFD–DEM coupling for simulations. Insufficient new and advanced models, namely bounded particle model
understanding and description of the interaction between and software such as Rocky®, for simulating the irregular
the particle–particle, particle–fluid, and particle-boundaries particles by its true and real shape. By this methodology,
give misunderstanding of the way of CFD–DEM coupling the simulations proved to be useful for different modeling
resulting in impairing the simulation modeling. Addition- problems in the agricultural process. Therefore, to find out
ally, due to the limitation of computational resources and and develop more efficient, accurate, and trustable models
complex shape structures of grains, most of the current to represent different particle shapes and detect their contact
CFD–DEM simulations for bio-particles are a little far from forces is still one of the urgent needs and most interesting
13
areas to focus on in the near future. As mentioned in this contact points requiring force and deformation calculation
review, since the current CFD–DEM coupling methods can at each contact point. To avoid this excess computation time
be briefly divided into three categories: namely one-way, problem, most researchers have used single-sphere models
two-way, four-way, most of the research related to agricul- and had reasonable success in predictions.
tural field concentrate on one-way only, while other ways are Depending on the software used in the simulation case,
neglected which are considered more precise and accurate both linear and nonlinear Hertz-Mindlin contact models
for representing real contacts. have been used effectively to study the aerodynamic opera-
One of the main bottlenecks faces CFD–DEM simulation tions such as separation and cleaning of crop grains. In
is the massive computational cost when simulating intensive general, coupling DEM with CFD has adequately simu-
particulate systems, particularly the multiphase flow system. lated the separation of air-grain flow and recommended
With typical PC hardware or workstation, three-dimensional for better predictions. Although this coupling has been
CFD–DEM simulations are usually limited to systems with increasingly used to study particle processes, it has not
limited particle numbers, since in realistic systems, the been widely applied in agricultural grains. The consid-
number of particles often exceeded millions of particles. erable variation in grains characteristics such as size,
With recent developments in computer technology, parallel shape, surface roughness, density, friction coefficients,
computing has shown great potential in accelerating com- composition, and other factors could be limiting the use
puter simulations using multi-core CPU and GPU (Graphics of CFD–DEM. Computational cost also limits CFD–DEM
Processing Unit). To date, using CPU and GPU platforms application; specifically, most of the cereal crops are
in the coupling CFD–DEM algorithm is still new technol- smaller, which need long computation time for simula-
ogy compared to the traditional CPU that needs adequate tion. The development of precision particle models could
development to reduce the time-consuming, especially when help spur the adoption of this numerical modeling concept
simulating agricultural crop grains. and optimize process and equipment design in the grain
Another significant problem still poses a severe challenge handling and processing industry.
is simulating large ranges in grain shapes and sizes in some To date, up to the authors’ investigating and searching,
aerodynamic systems; also, the presence of the smallest few studies have been made on the gas–solid flow in aerody-
particles in large systems requires an efficient treatment to namic separation devices employing CFD–DEM approach.
model and simulate. Overall, modeling the bio-particulate One-way coupling approach is a proper technique for pre-
matter process in disperse systems is a particularly open dicting the behavior of macroscopic particles, where it
area of research that is very important for understanding commonly applied in agriculture systems that consume less
agricultural processes such as precise planting, harvesting, power for simulation. In contrast, two-way and four-way are
post-harvesting, industrial, and agrophysical systems. At the commonly used in describing the particles flow in fluidized
same time, until now, little investigations have been done on bed related to chemical or pharmaceutical field. However,
these systems. However, the vital evaluating factors such as limited research used these techniques in agriculture, but it
the validity, accuracy, speed, and versatility of the numerical still a big gap for simulation in many aerodynamic systems
technique need to be improved in the near future. related to the agriculture field. Future developments for this
coupling include implementation of non-drag laws, thermal
fluid-particle coupling, and mass exchange between fluid and
9 Summary and Conclusions particles should be considered.
Finally, research effort should also focus on seeking far
Existing literature that used CFD–DEM coupling approach more efficient methods for multi-phase and multi-physics
to simulate aerodynamic granular systems, limited to bio- problems, which are often complexly coupled together. The
particulate matter, has been reviewed. Usually, the soft- comparison between simulations and real experimental
sphere approach of DEM was commonly used to develop results showed that the CFD-DEM model is predictive and
these grain simulations process. The advantage of soft can be extended to several aerodynamic processes related to
sphere models was their capability to represent multiple the agricultural field. Using this coupling technique and sim-
particle contacts, which are of importance when modeling ulation methods will make it possible to improve the aerody-
bulk grain systems. Particle models varied with the type namic systems of agricultural machines without high costs
of grain. For near-spherical grains such as soybean and and efforts. The obtained characteristics from the above
rapeseed, single-sphere particle models predicted particle investigations can seriously provide a basis for developing
behavior with high accuracy. For non-spherical grains such and optimizing the structural and technological parameters
as rice, wheat, and corn, particle, multi-sphere approach of the aerodynamic systems of agricultural machines.
was used, resulting in increasing the simulation time and
computational cost. That is because of the higher number of
13
Acknowledgements This work was supported by the National Natural 16. Hellevang KJ (1985) Pneumatic grain conveyors. Fargo, North
Science Foundation of China (Grant Nos. 52079058 and 51979138), Dakota
Nature Science Foundation for Excellent Young Scholars of Jiangsu 17. Piven VV (2018) Determination of the extent of fraction in air
Province (Grant No. BK20190101), National Key Research and Devel- separation of grain material. J Phys Conf Ser. https://doi.org/10.
opment Project (Grant No. 2020YFC1512404). 1088/1742-6596/1059/1/012001
18. Clarke B (1985) Cleaning seeds by fluidization. J Agric Eng Res
31:231–242. https://doi.org/10.1016/0021-8634(85)90090-3
Declarations 19. Gorial BY, O’Callaghan JR (1990) Aerodynamic properties of
grain/straw materials. J Agric Eng Res 46:275–290. https://doi.
Conflict of interest The authors are not aware of any biases that might org/10.1016/S0021-8634(05)80132-5
be perceived as affecting the objectivity of this review. The authors 20. Zewdu AD (2007) Aerodynamic properties of tef grain and straw
also declare no conflict of interest. material. Biosyst Eng 98:304–309. https://doi.org/10.1016/j.
biosystemseng.2007.08.003
21. Samantaray SK, Mohapatra SS, Munshi B (2018) Experimental
findings and analysis of terminal velocity and drag coefficient of
Raschig Ring in vertical and inclined channel. Powder Technol
References 340:440–448. https://doi.org/10.1016/j.powtec.2018.09.030
22. Zhang L, Honaker R, Liu W et al (2015) Calculation of terminal
1. Simonyan KJ, Yiljep YD (2008) Investigating grain separation velocity in transitional flow for spherical particle. Int J Min Sci
and cleaning efficiency distribution of a conventional stationary Technol 25:311–317. https://d oi.o rg/1 0.1 016/j.i jmst.2 015.0 2.0 22
Rasp-bar Sorghum Thresher. Agric Eng Int CIGR J 10:1–13 23. Czernik Z (1983) Studies of aerodynamical features of seeds of
2. Wang YJ, Chung DS, Spillman CK et al (1994) Evaluation of Scotch pine, Norway spruce and European larch (Badania wlas-
laboratory grain cleaning and separating equipment: part I. Trans ciwosci aerodynamicznych nasion sosny zwyczajnej, swierka
ASAE 37:507–513 pospolitego i modrzewia europejskieg). Sylwan 9:31–40
3. Sokhansanj S, Fang R, Opoku A (1999) Influence of physical 24. Ghamari S, Borghei AM, Rabbani H et al (2010) Modeling
properties and operating conditions on particle segregation on the terminal velocity of agricultural seeds with artificial neural
gravity table. Appl Eng Agric. 15(5):495–499. https://doi.org/ networks. Afr J Agric Res 5:389–398. https://doi.org/10.5897/
10.13031/2013.5809 AJAR09.626
4. Pouliot Y, Conway V, Leclerc P (2014) Separation and concentra- 25. Khodabakhshian R, Emadi B, Khojastehpour M, Golzarian MR
tion technologies in food processing. In: Clark S, Jung S, Lamsal (2018) Aerodynamic separation and cleaning of pomegranate
B (eds) Food processing: principles and applications, second. arils from rind and white segments (locular septa). J Saudi Soc
John Wiley and Sons Ltd, Québec, pp 33–60 Agric Sci 17:61–68. https://doi.org/10.1016/j.jssas.2016.01.003
5. Sosulski F (1987) Yield and functional properties of air classified 26. Mohsenin NN (1986) Physical properties of plant and animal
protein and starch fractions from eight legume flours. J Am Chem materials. Gordon and Breach Science, New York
Soc 6:363–370 27. Tylek P, Walczyk J (2004) Effectiveness of the pneumatic separa-
6. Uhl JB, Lamp BJ (1966) Pneumatic separation of grain and straw tion of Norway spruce Picea abies (L.) Karst seeds. Dendrobiol-
mixtures. Trans ASAE 61–135:244–246 ogy 51:101–104
7. Hollatz B, Quick GR (2003) Combine tailings returns, part 1: the 28. Levenspiel O, Haider A (1989) Drag coefficient and terminal
effects of combine performance and settings on tailings. In: Inter- velocity of spherical and nonspherical particles. Powder Technol
national conference on crop harvesting and processing. ASABE, 58:63–70. https://doi.org/10.1016/0032-59109(89)80008-7
St. Joseph, Michigan, Louisville, Kentucky, USA, p 22 29. De AE, Spogis N, Taranto OP, Silva MA (2019) Theoretical
8. Tylek P, Walczyk J (2003) Critical air velocity as a separation study of pneumatic separation of sugarcane bagasse particles.
feature in nuts of european beech (Fagus sylvatica L.). Electron Biomass Bioenerg 127:105256. https://doi.org/10.1016/j.biomb
J Polish Agric Univ, 6 ioe.2019.105256
9. IAEE (2008) Develop technical and technological bases improve 30. Lapple CE (1951) Fluid and particle mechanics, 1st edn. Univer-
separation of grain and seed set of physical and mechanical prop- sity of Delaware, Newark
erties. Glevaha 31. Hauhouot-O’Hara M, Criner BR, Brusewitz GH, Solie JB (2000)
10. Kotov BI, Stepanenko SP, Swiftly VA, Smith JG (2009) Design, Selected physical characteristics and aerodynamic properties of
production and operation of agricultural machinery. A Natl Inter- cheat seed for separation from wheat. J Sci Res Dev, 2
dep Sci Tech Collect 39:209–214 32. Afonso Júnior PC, Corrêa PC, Pinto FAC, Queiroz DM (2007)
11. Kotov BI (2009) The theory of separation of grains in the air Aerodynamic properties of coffee cherries and beans. Biosyst
stream. A Natl Interdep Sci Tech Collect 39:209–214 Eng 98:39–46. https://doi.org/10.1016/j.biosystemseng.2007.03.
12. Stepanenko S (2017) Research pneumatic gravity separation 010
grain materials. Scientific proceedings V international scien- 33. Shellard JE, Macmillan RH (1978) Aerodynamic properties of
tific-technical conference “agricultural machinery.” Glevakha, threshed wheat materials. J Agric Eng Res 23:273–281. https://
Ukraine, pp 143–145 doi.org/10.1016/0021-8634(78)90101-4
13. Reichert RD (1982) Air classification of peas (Pisum sativum) 34. Shahbazi F (2013) Aerodynamic properties of wild mustard
varying widely in protein content. J Food Sci 47:1263–1267 (Sinapis arvensis L.) seed for separation from canola. J Sci Food
14. Emami S, Tabil LG, Tyler RT, Crerar WJ (2007) Starch-protein Agric 93:1466–1470. https://doi.org/10.1002/jsfa.5916
separation from chickpea flour using a hydrocyclone. J Food Eng 35. El-Emam MA, Shi W, Zhou L (2019) CFD-DEM simulation and
82:460–465. https://doi.org/10.1016/j.jfoodeng.2007.03.002 optimization of gas-cyclone performance with realistic macro-
15. Panasiewicz M, Sobczak P, Mazur J et al (2012) The technique scopic particulate matter. Adv Powder Technol 30:2686–2702.
and analysis of the process of separation and cleaning grain mate- https://doi.org/10.1016/j.apt.2019.08.015
rials. J Food Eng 109:603–608. https://doi.org/10.1016/j.jfood 36. Binelo MO, de Lima RF, Khatchatourian OA, Stránský J (2019)
eng.2011.10.010 Modelling of the drag force of agricultural seeds applied to the
13
discrete element method. Biosyst Eng 178:168–175. https://doi. 56. Wang Y, Williams K, Jones M, Chen B (2017) CFD simulation
org/10.1016/j.biosystemseng.2018.11.013 methodology for gas-solid flow in bypass pneumatic conveying: a
37. Khatchatourian OA, Binelo MO, de Lima RF (2014) Simulation review. Appl Therm Eng 125:185–208. https://d oi.o rg/1 0.1 016/j.
of soya bean flow in mixed-flow dryers using DEM. Biosyst Eng applthermaleng.2017.05.063
123:68–76. https://d oi.o rg/1 0.1 016/j.b iosys temse ng.2 014.0 5.0 03 57. Ullah A, Hong K, Gao Y et al (2019) An overview of Eulerian
38. Guo Y, Curtis JS (2015) Discrete element method simulations for CFD modeling and simulation of non-spherical biomass parti-
complex granular flows. Annu Rev Fluid Mech 47:21–46. https:// cles. Renew Energy 141:1054–1066. https://doi.org/10.1016/j.
doi.org/10.1146/annurev-fluid-010814-014644 renene.2019.04.074
39. Li H, Li Y, Gao F et al (2012) CFD-DEM simulation of material 58. Norouzi HR, Zarghami R, Sotudeh-gharebagh R, Mostoufi N
motion in air-and-screen cleaning device. Comput Electron Agric (2016) Coupled CFD-DEM modeling : formulation, implementa-
88:111–119. https://doi.org/10.1016/j.compag.2012.07.006 tion and application to multiphase flows. Wiley, New York
40. Ren B, Zhong W, Chen Y et al (2012) CFD-DEM simulation 59. Boac JM, Ambrose RPK, Casada ME et al (2014) Applications
of spouting of corn-shaped particles. Particuology 10:562–572. of discrete element method in modeling of grain postharvest
https://doi.org/10.1016/j.partic.2012.03.011 operations. Food Eng Rev 6:128–149. https://doi.org/10.1007/
41. Liu X, Gan J, Zhong W, Yu A (2019) Particle shape effects on s12393-014-9090-y
dynamic behaviors in a spouted bed: CFD-DEM study. Powder 60. Cundall PA, Strack ODL (1979) A discrete numerical model for
Technol. https://doi.org/10.1016/j.powtec.2019.07.099 granular assemblies. Géotechnique 29:47–65. https://doi.org/10.
42. Han D, Zhang D, Jing H et al (2018) DEM-CFD coupling simula- 1680/geot.1979.29.1.47
tion and optimization of an inside-filling air-blowing maize pre- 61. Dewicki G (2003) Bulk material handling and processing:
cision seed-metering device. Comput Electron Agric 150:426– numerical techniques and simulation of granular material. Bulk
438. https://doi.org/10.1016/j.compag.2018.05.006 Solids Handl Int J Storing Handl Bulk Mater 23:110–113
43. Tsuji Y (2007) Multi-scale modeling of dense phase gas–particle 62. Zhu HP, Zhou ZY, Yang RY, Yu AB (2007) Discrete particle
flow. Chem Eng Sci 62:3410–3418 simulation of particulate systems: theoretical developments.
44. Tsuji Y, Tanaka T, Ishida T (1992) Lagrangian numerical-simu- Chem Eng Sci 62:3378–3396. https://d oi.o rg/1 0.1 016/j.c es.2 006.
lation of plug flow of cohesionless particles in ahorizontal pipe. 12.089
Powder Technol 71:239–250 63. Berger R, Kloss C, Kohlmeyer A, Pirker S (2015) Hybrid paral-
45. Zhou ZY, Kuang SB, Chu KW, Yu AB (2010) Assessments of lelization of the LIGGGHTS open-source DEM code. Powder
CFD–DEM models in particle–fluid flow modelling. J Fluid Technol 278:234–247. https://d oi.o rg/1 0.1 016/j.p owtec.2 015.0 3.
Mech 661:482–510 019
46. Wang B, Chu KW, Yu AB (2007) Numerical study of particle– 64. Theuerkauf J, Dhodapkar S, Jacob K (2007) Modeling granular
fluid flow in a hydrocyclone. Ind Eng Chem Res 46:4695–4705 flow using discrete element method—from theory to practice.
47. Xu BH, Yu AB (1997) Numerical simulation of the gas–solid Chem Eng 114:154–162
flow in a fluidized bed by combining discrete particle method 65. Williams JR, Hocking G, Mustoe G (1985) The theoretical basis
with computational fluid dynamics. Chem Eng Sci 52:2785–2809 of the discrete element method. In: NUMETA ’85 conference.
48. Chu KW, Wang B, Yu AB, Vince A (2009) CFD–DEM model- Balkema, Rotterdam, pp 897–906
ling of multiphase flow in dense medium cyclones. PowderTech- 66. Zhu HP, Zhou ZY, Yang RY, Yu AB (2008) Discrete particle
nology 193:235–247 simulation of particulate systems: a review of major applications
49. Cortés C, Gil A (2007) Modeling the gas and particle flow inside and findings. Chem Eng Sci 63:5728–5770. https://doi.org/10.
cyclone separators. Prog Energy Combust Sci 33:409–452. 1016/j.ces.2008.08.006
https://doi.org/10.1016/j.pecs.2007.02.001 67. Keppler I, Kocsis L, Oldal I et al (2012) Grain velocity distribu-
50. Derksen JJ, Sundaresan S, van den Akker HEA (2006) Simu- tion in a mixed flow dryer. Adv Powder Technol 23:824–832.
lation of mass-loading effects in gas-solid cyclone separators. https://doi.org/10.1016/j.apt.2011.11.003
Powder Technol 163:59–68. https://d oi.o rg/1 0.1 016/B 978-0 0804 68. Zhang X, Vu-Quoc L (2000) Simulation of chute flow of soy-
4544-1/50089-3 beans using an improved tangential force-displacement model.
51. Horabik J, Molenda M (2016) Parameters and contact models for Mech Mater 32:115–129. https:// d oi. o rg/ 1 0. 1 016/ S 0167-
DEM simulations of agricultural granular materials: a review. 6636(99)00043-5
Biosyst Eng 147:206–225. https://doi.org/10.1016/j.biosystems 69. Weigler F, Scaar H, Mellmann J (2012) Investigation of particle
eng.2016.02.017 and air flows in a mixed-flow dryer. Dry Technol 30:1730–1741.
52. Almeida E, Spogis N, Silva MA (2016) Computational study of https://doi.org/10.1080/07373937.2012.703742
the pneumatic separation of sugarcane bagasse. In: Nzihou A 70. Sturm M, Wirtz S, Scherer V, Denecke J (2010) Coupled DEM-
(ed), 6th international conference on engineering for waste and CFD simulation of pneumatically conveyed granular media.
biomass valorisation. Albi : Mines Albi, 2016, Albi, France, pp Chem Eng Technol 33:1184–1192. https://doi.org/10.1002/ceat.
1–16 201000162
53. Fonte CB, Jr JAAO, Almeida LC De (2015) Dem-Cfd coupling : 71. Lim EWC, Wang CH, Yu AB (2006) Discrete element simu-
mathematical modelling and case studies using Rocky-Dem ® lation for pneumatic conveying of granular material. AIChE J
and Ansys Fluent ®. In: Eleventh international conference on 52:496–509. https://doi.org/10.1002/aic.10645
CFD in the minerals and process industries. CSIRO Organiza- 72. Parafiniuk P, Molenda M, Horabik J (2013) Discharge of rape-
tion, Melbourne, Australia, pp 1–7 seeds from a model silo: physical testing and discrete element
54. Chu KW, Wang B, Xu DL et al (2011) CFD-DEM simulation of method simulations. Comput Electron Agric 97:40–46. https://
the gas-solid flow in a cyclone separator. Chem Eng Sci 66:834– doi.org/10.1016/j.compag.2013.06.008
847. https://doi.org/10.1016/j.ces.2010.11.026 73. Raji AO, Favier JF (2004) Model for the deformation in agricul-
55. Romaní Fernández X, Nirschl H (2013) Simulation of particles tural and food particulate materials under bulk compressive load-
and sediment behaviour in centrifugal field by coupling CFD ing using discrete element method. II: Compression of oilseeds. J
and DEM. Chem Eng Sci 94:7–19. https://doi.org/10.1016/j.ces. Food Eng 64:373–380. https://doi.org/10.1016/j.jfoodeng.2003.
2013.02.039 11.005
13
74. Anand A, Curtis JS, Wassgren CR et al (2009) Predicting dis- 90. Kosinski P, Hoffmann AC (2010) An extension of the hard-
charge dynamics of wet cohesive particles from a rectangular sphere particle-particle collision model to study agglomeration.
hopper using the discrete element method (DEM). Chem Eng Chem Eng Sci 65:3231–3239. https://d oi.o rg/1 0.1 016/j.c es.2 010.
Sci 64:5268–5275. https://doi.org/10.1016/j.ces.2009.09.001 02.012
75. Sakaguchi E, Suzuki M, Favier JF, Kawakami S (2001) Numeri- 91. O’Sullivan C (2011) Particle-based discrete element modeling:
cal simulation of the shaking separation of Paddy and Brown rice geomechanics perspective. Int J Geomech 11:449–464. https://
using the discrete element method. J Agric Eng Res 79:307–315. doi.org/10.1061/(asce)gm.1943-5622.0000024
https://doi.org/10.1006/jaer.2001.0706 92. O’Sullivan C (2011) Particulate discrete element modelling : a
76. Ma Z, Li Y, Xu L (2015) Discrete-element method simulation of geomechanics perspective, 1st edn. Spon Press, New York
agricultural particles’ motion in variable-amplitude screen box. 93. Hoomans BPB, Kuipers JAM, Briels WJ, Van Swaaij WPM
Comput Electron Agric 118:92–99. https://doi.org/10.1016/j. (1996) Discrete particle simulation of bubble and slug forma-
compag.2015.08.030 tion in a two-dimensional gas-fluidized bed: a hard-sphere
77. Lenaerts B, Aertsen T, Tijskens E et al (2014) Simulation of approch. Chem Eng Sci 51:99–118. https://doi.org/10.1016/
grain-straw separation by discrete element modeling with bend- 0009-2509(95)00271-5
able straw particles. Comput Electron Agric 101:24–33. https:// 94. Fortin J, Millet O, de Saxcé G (2005) Numerical simulation of
doi.org/10.1016/j.compag.2013.12.002 granular materials by an improved discrete element method. Int
78. Woo MW, Daud WRW, Mujumdar AS et al (2010) Role of rheo- J Numer Methods Eng 62:639–663. https://d oi.o rg/1 0.1 002/n me.
logical characteristics in amorphous food particle-wall collisions 1209
in spray drying. Powder Technol 198:251–257. https://doi.org/ 95. Duran J (2000) Sands, powders, and grains: an introduction to
10.1016/j.powtec.2009.11.015 the physics of granular materials. Springer, New York
79. Adewumi BA (2008) 2D modeling of grain transport and separa- 96. Tsuji Y, Kawaguchi T, Tanaka T (1993) Discrete particle simula-
tion in the horizontal air stream. In: Food processing automation tion of two-dimensional fluidized bed. Powder Technol 77:79–
conference CD-Rom, pp 1–10 87. https://doi.org/10.1016/0032-5910(93)85010-7
80. Guo Y, Wu CY, Kafui KD, Thornton C (2011) 3D DEM/CFD 97. Zhao J, Shan T (2013) Coupled CFD-DEM simulation of fluid-
analysis of size-induced segregation during die filling. Powder particle interaction in geomechanics. Powder Technol 239:248–
Technol 206:177–188. https://doi.org/10.1016/j.powtec.2010. 258. https://doi.org/10.1016/j.powtec.2013.02.003
05.029 98. Campbell CS (2006) Granular material flows: an overview. Pow-
81. Lei X, Liao Y, Liao Q (2016) Simulation of seed motion in der Technol 162:208–229. https://d oi.o rg/1 0.1 016/j.p owtec.2 005.
seed feeding device with DEM-CFD coupling approach for 12.008
rapeseed and wheat. Comput Electron Agric 131:29–39. 99. Delaney GW, Inagaki S, Aste T (2007) Fine tuning DEM simu-
https://doi.org/10.1016/j.compag.2016.11.006 lations to perform virtual experiments with three-dimensional
82. Li J, Webb C, Pandiella SS, Campbell GM (2002) A numerical granular packings. In: Tomaso A, Di MT, Tordesillas A (eds)
simulation of separation of crop seeds by screening: effect of Granular and complex materials. World Scientific, Singapore,
particle bed depth. Food Bioprod Process Trans Inst Chem Eng pp 141–168
Part C 80:109–117. https://doi.org/10.1205/096030802529387 100. Crowe CT, Schwarzkopf JD, Sommerfeld M, Tsuji Y (2011)
44 Multiphase flows with droplets and particles, 2nd edn. CRC
83. Li J, Webb C, Pandiella SS, Campbell GM (2003) Discrete Press, Taylor and Francis Group, Boca Raton
particle motion on sieves: a numerical study using the DEM 101. Carlos Varas AE, Peters EAJF, Kuipers JAM (2017) CFD-DEM
simulation. Powder Technol 133:190–202. https://doi.org/10. simulations and experimental validation of clustering phenomena
1016/S0032-5910(03)00092-5 and riser hydrodynamics. Chem Eng Sci 169:246–258. https://
84. Boac JM, Casada ME, Maghirang RG, Harner JP (2012) 3-D doi.org/10.1016/j.ces.2016.08.030
and quasi-2-d discrete element modeling of grain commingling 102. He Y, Bayly AE, Hassanpour A (2018) Coupling CFD-DEM with
in a bucket elevator boot system. Trans ASABE 55:659–672 dynamic meshing: a new approach for fluid-structure interaction
85. González-Montellano C, Baguena EM, Ramírez-Gómez Á, in particle-fluid flows. Powder Technol 325:620–631. https://d oi.
Barreiro P (2014) Discrete element analysis for the assessment org/10.1016/j.powtec.2017.11.045
of the accuracy of load cell-based dynamic weighing systems 103. Neuwirth J, Antonyuk S, Heinrich S, Jacob M (2013) CFD-DEM
in grape harvesters under different ground conditions. Comput study and direct measurement of the granular flow in a rotor
Electron Agric 100:13–23. https://doi.org/10.1016/j.compag. granulator. Chem Eng Sci 86:151–163. https://d oi.o rg/1 0.1 016/j.
2013.10.008 ces.2012.07.005
86. Van Zeebroeck M, Tijskens E, Dintwa E et al (2006) The discrete 104. Kuang SB, Yu AB, Zou ZS (2009) Computational study of flow
element method (DEM) to simulate fruit impact damage during regimes in vertical pneumatic conveying. Ind Eng Chem Res
transport and handling: model building and validation of DEM 48:6846–6858
to predict bruise damage of apples. Postharvest Biol Technol 105. Akhshik S, Behzad M, Rajabi M (2015) CFD-DEM approach to
41:85–91. https://doi.org/10.1016/j.postharvbio.2006.02.007 investigate the effect of drill pipe rotation on cuttings transport
87. Ramírez-Gómez Á, Gallego E, Fuentes JM et al (2014) Values behavior. J Pet Sci Eng 127:229–244. https://doi.org/10.1016/j.
for particle-scale properties of biomass briquettes made from petrol.2015.01.017
agroforestry residues. Particuology 12:100–106. https://doi.org/ 106. Boac JM, Casada ME, Maghirang RG (2010) Material and inter-
10.1016/j.partic.2013.05.007 action properties of selected grains and oilseeds for modeling
88. Cundall PA (1988) Detect and represent contacts in a system discrete particles. Trans ASABE. 53:1201–1216. https://doi.org/
composed of many polyhedral blocks. Int J Rock Mech Min Sci 10.13031/2013.32577
Geomech 25:107–116 107. ESSS-Rocky (2018) Rocky-Dem technical manual. In: Rocky E
89. Hart R, Cundall PA, Lemos J (1988) Formulation of a three- (ed) Models used in Rocky, 4.13. ESSS, Rio de Janeiro, Brazi,
dimensional distinct element model-part II. Mechanical calcula- pp 3–11
tions for motion and interaction of a system composed of many 108. Cundall PA (1988) Computer simulations of dense sphere assem-
polyhedral blocks. Int J Rock Mech Min Sci 25:117–125. https:// blies. Micromech Granul Mater 20:113–123. https://doi.org/10.
doi.org/10.1016/0148-9062(88)92294-2 1016/B978-0-444-70523-5.50021-7
13
109. Timoshenko SP, Goodier JN (1970) Theory of elasticity, 3rd edn. 129. Mousaviraad M, Tekeste MZ, Rosentrater KA (2017) Calibra-
McGraw-Hill, New York tion and validation of a discrete element model of corn using
110. ASAE S368.4 (2001) Compression test of food materials of con- grain flow simulation in a commercial screw grain auger. Trans
vex shape, revised. ASAE, Michigan ASABE 60:1403–1415. https://doi.org/10.13031/trans.12200
111. Antypov D, Elliott JA (2011) On an analytical solution for the 130. Chan EL, Washino K (2018) Coarse grain model for DEM sim-
damped Hertzian spring. Europhys Lett. https://doi.org/10.1209/ ulation of dense and dynamic particle flow with liquid bridge
0295-5075/94/50004 forces. Chem Eng Res Des 132:1060–1069. https://doi.org/10.
112. Walton OR, Braun RL (1986) Viscosity, granular-temperature, 1016/j.cherd.2017.12.033
and stress calculations for shearing assemblies of inelastic, fric- 131. Olaleye AK, Shardt O, Walker GM, Van Den AH (2019) Pneu-
tional disks. J Rheol (NY) 30:949–980. https://d oi.o rg/1 0.1 122/1. matic conveying of cohesive dairy powder: experiments and
549893 CFD-DEM simulations. Powder Technol. https://doi.org/10.
113. Tomas J (2000) Particle adhesion fundamentals and bulk powder 1016/j.powtec.2019.09.046
consolidation. KONA Powder Part J 18:157–169. https://d oi.o rg/ 132. Sakai M, Takahashi H, Pain CC et al (2012) Study on a large-
10.14356/kona.2000022 scale discrete element model for fine particles in a fluidized bed.
114. Luding S (2008) Cohesive, frictional powders: contact models Adv Powder Technol 23:673–681. https://doi.org/10.1016/j.apt.
for tension. Granul Matter 10:235–246. https://doi.org/10.1007/ 2011.08.006
s10035-008-0099-x 133. Sorace CM, Louge MY, Crozier MD, Law VHC (2009) High
115. Freireich B, Litster J, Wassgren C (2009) Using the discrete ele- apparent adhesion energy in the breakdown of normal restitution
ment method to predict collision-scale behavior: a sensitivity for binary impacts of small spheres at low speed. Mech Res Com-
analysis. Chem Eng Sci 64:3407–3416. https://d oi.o rg/1 0.1 016/j. mun 36:364–368. https://d oi.o rg/1 0.1 016/j.m
echre scom.2 008.1 0.
ces.2009.04.019 009
116. Walton OR (1993) Numerical simulation of inelastic, frictional 134. Althaus TO, Windhab EJ, Scheuble N (2012) Effect of pendu-
particle-particle interactions. In: Roco MC (ed) Particulate two- lar liquid bridges on the flow behavior of wet powders. Powder
phase flow. Butterworth-Heinemann, Adivision of Reed Publish- Technol 217:599–606. https://d oi.o rg/1 0.1 016/j.p owtec.2 011.1 1.
ing (USA) Inc., Reed, pp 855–947 026
117. Bartsch P, Zunft S (2019) Granular flow around the horizontal 135. Johnson KL, Kendall K, Roberts AD (1971) Surface charge and
tubes of a particle heat exchanger : DEM-simulation and experi- the contact of elastic solids. Proceeding R Soc London 324:301–
mental validation. Sol Energy 182:48–56. https://doi.org/10. 313. https://doi.org/10.1088/0022-3727/6/9/304
1016/j.solener.2019.01.086 136. Nase ST, Vargas WL, Abatan AA, McCarthy JJ (2001) Discrete
118. Vivacqua V, López A, Hammond R et al (2019) DEM analysis characterization tools for cohesive granular material. Powder
of the effect of particle shape, cohesion and strain rate on pow- Technol 116:214–223. https://doi.org/10.1016/S0032-5910(00)
der rheometry. Powder Technol 342:653–663. https://doi.org/10. 00398-3
1016/j.powtec.2018.10.034 137. Favier JF, Abbaspour-Fard MH, Kremmer M, Raji AO (1999)
119. Seville JPK, Willett CD, Knight PC (2000) Interparticle forces Shape representation of axisymmetrical, non-spherical particles
in fluidisation.pdf. Powder Technol 113:261–268. https://d oi.o rg/ in discrete element simulation using multi-element model par-
10.1016/S0032-5910(00)00309-0 ticles. Eng Comput 16:467–480. https://doi.org/10.1108/02644
120. Schwager T, Pöschel T (2007) Coefficient of restitution and lin- 409910271894
ear-dashpot model revisited. Granul Matter 9:465–469. https:// 138. Hogu C (1998) Shape representation and contact detection for
doi.org/10.1007/s10035-007-0065-z discrete element simulations of arbitrary geometries. Eng Com-
121. Van Zeebroeck M, Tijskens E, Van Liedekerke P et al (2003) put 15:374–390
Determination of the dynamical behaviour of biological mate- 139. Zhong W, Yu A, Liu X et al (2016) DEM/CFD-DEM modelling
rials during impact using a pendulum device. J Sound Vib of non-spherical particulate systems: theoretical developments
266:465–480. https://doi.org/10.1016/S0022-460X(03)00579-0 and applications. Powder Technol 302:108–152. https://doi.org/
122. Stropek Z, Gołacki K (2015) A new method for measuring impact 10.1016/j.powtec.2016.07.010
related bruises in fruits. Postharvest Biol Technol 110:131–139. 140. Kureck H, Govender N, Siegmann E et al (2019) Industrial scale
https://doi.org/10.1016/j.postharvbio.2015.07.005 simulations of tablet coating using GPU based DEM : a valida-
123. Deresiewicz H, Mindlin RD (1953) Elastic spheres in contact tion study. Chem Eng Sci 202:462–480. https://d oi.o rg/1 0.1 016/j.
under varying oblique forces. J Appl Mech 20:327–344 ces.2019.03.029
124. Vu-Quoc L, Zhang X, Walton OR (2000) A 3-D discrete-element 141. ISO 3435 (1977) Continuous mechanical handling equipment
method for dry granular flows of ellipsoidal particles. Comput classification and symbolization of bulk materials. Geneva: Inter-
Methods Appl Mech Eng 187:483–528. https://doi.org/10.1016/ national Organization for Standardization
S0045-7825(99)00337-0 142. Zhan Z, Yaoming L, Zhenwei L, Zhiqiang G (2013) DEM simu-
125. Di Renzo A, Di Maio FP (2004) Comparison of contact-force lation and physical testing of rice seed impact against a grain loss
models for the simulation of collisions in DEM-based granular sensor. Biosyst Eng 116:410–419. https://d oi.o rg/1 0.1 016/j.b iosy
flow codes. Chem Eng Sci 59:525–541. https://d oi.o rg/1 0.1 016/j. stemseng.2013.10.002
ces.2003.09.037 143. Athanassiadis AG, Miskin MZ, Kaplan P et al (2014) Particle
126. Łukaszuk J, Molenda M, Horabik J, Wiącek J (2009) Method of shape effects on the stress response of granular packings. Soft
measurement of coefficient of friction between pairs of metallic Matter 10:48–59. https://doi.org/10.1039/c3sm52047a
and organic objects. Acta Agrophys 13:407–418 144. Kruggel-Emden H, Rickelt S, Wirtz S, Scherer V (2008) A study
127. Nyembwe AM, Cromarty RD (2019) Simulation of the pressure on the validity of the multi-sphere discrete element method. Pow-
drop across granulated mixtures using a coupled DEM–CFD der Technol 188:153–165. https://d oi.o rg/1 0.1 016/j.p owtec.2 008.
model. Adv Powder Technol 30:85–97. https://d oi.o rg/1 0.1 016/j. 04.037
apt.2018.10.010 145. Saeed MK, Siraj MS (2019) Mixing study of non-spherical par-
128. Sun J, Battaglia F, Subramaniam S (2006) Dynamics and struc- ticles using DEM. Powder Technol 344:617–627. https://d oi.o rg/
tures of segregation in a dense, vibrating granular bed. Phys Rev 10.1016/j.powtec.2018.12.057
E 74:061307. https://doi.org/10.1103/PhysRevE.74.061307
13
146. Markauskas D, Kačianauskas R (2011) Investigation of rice behavior of transversely isotropic rock. Int J Rock Mech Min Sci
grain flow by multi-sphere particle model with rolling resist- 110:120–132. https://doi.org/10.1016/j.ijrmms.2018.07.018
ance. Granul Matter 13:143–148. https://d oi.o rg/1 0.1 007/ 164. Govender N, Wilke DN, Pizette P, Abriak N (2018) A study of
s10035-010-0196-5 shape non-uniformity and poly-dispersity in hopper discharge of
147. Wiacek J, Molenda M, Horabik J, Ooi JY (2012) Influence of spherical and polyhedral particle systems using the Blaze-DEM
grain shape and intergranular friction on material behavior in GPU code. Appl Math Comput 319:318–336. https://doi.org/10.
uniaxial compression: experimental and DEM modeling. Powder 1016/j.amc.2017.03.037
Technol 217:435–442. https://d oi.o rg/1 0.1 016/j.p owtec.2 011.1 0. 165. Nezami EG, Hashash YMA, Zhao D, Ghaboussi J (2007) Simu-
060 lation of front end loader bucket–soil interaction using discrete
148. Markauskas D, Kačianauskas R, Džiugys A, Navakas R (2010) element method. Int J Numer Anal Methods Geomech 31:1147–
Investigation of adequacy of multi-sphere approximation of ellip- 1162. https://doi.org/10.1002/nag.594
tical particles for DEM simulations. Granul Matter 12:107–123. 166. Li X, Wang F, Zhang D et al (2019) Fluid-solid interaction simu-
https://doi.org/10.1007/s10035-009-0158-y lation for particles and walls of arbitrary polygonal shapes with
149. Ma H, Zhao Y (2017) Modeling of the flow of ellipsoidal par- a coupled LBM-IMB-DEM method. Powder Technol 356:177–
ticles in a horizontal rotating drum based on DEM simulation. 192. https://doi.org/10.1016/j.powtec.2019.08.006
Chem Eng Sci 172:636–651. https://doi.org/10.1016/j.ces.2017. 167. Peña AA, Lind PG, Herrmann HJ (2008) Modeling slow defor-
07.017 mation of polygonal particles using DEM. Particuology 6:506–
150. Zhou Z, Zou R, Pinson D, Yu A (2011) Dynamic simulation of 514. https://doi.org/10.1016/j.partic.2008.07.009
the packing of ellipsoidal particles. Ind Eng Chem Res 50:9787– 168. Fu X, Yao Z, Zhang X (2017) Numerical simulation of polygonal
9798. https://doi.org/10.1021/ie200862n particles moving in incompressible viscous fluids. Particuology
151. Luo X, Zhao L, Zhang S et al (2018) Experimental and DEM 31:140–151. https://doi.org/10.1016/j.partic.2016.05.016
studies on the distribution of voidage in the random filling bed 169. Džiugys A, Peters B (2001) An approach to simulate the motion
of ellipsoidal particles. Powder Technol 340:400–410. https:// of spherical and non-spherical fuel particles in combustion cham-
doi.org/10.1016/j.powtec.2018.09.042 bers. Granul Matter 3:231–265
152. Ma H, Zhao Y, Cheng Y (2019) CFD-DEM modeling of rod- 170. Locurto GJ, Zhang X, Zakirov V et al (1997) Soybean
like particles in a fl uidized bed with complex geometry. Powder impacts: experiments and dynamic simulations. Trans ASAE
Technol 344:673–683. https://d oi.o rg/1 0.1 016/j.p owtec.2 018.1 2. 40:789–794
066 171. Xu T, Yu J, Yu Y, Wang Y (2018) A modelling and verification
153. Chen H, Zhao S, Zhou X (2019) DEM investigation of angle of approach for soybean seed particles using the discrete element
repose for super-ellipsoidal particles. Particuology. https://doi. method. Adv Powder Technol 29:3274–3290. https://doi.org/
org/10.1016/j.partic.2019.05.005 10.1016/j.apt.2018.09.006
154. Zhao Y, Xu L, Umbanhowar PB, Lueptow RM (2019) Discrete 172. Jiang G, Qiu B (2011) Discrete element method simulation
element simulation of cylindrical particles using super-ellipsoids. of impact-based measurement of grain mass flow. In: 2011
Particuology 46:55–66. https://doi.org/10.1016/j.partic.2018.04. international conference on computer distributed control and
007 intelligent environmental monitoring. IEEE, Changsha, China,
155. Baran O, Eppinger T, Han K (2018) DEM simulation of cylinders pp 419–422
and capsules in a fluidized bed. In: 8th world congress on particle 173. Coetzee CJ, Els DNJ (2009) Calibration of discrete element
technology, Orlando, pp 232–242 parameters and the modelling of silo discharge and bucket filling.
156. Zhong WQ, Zhang Y, Jin B, Zhang M (2009) Discrete element Comput Electron Agric 65:198–212. https://doi.org/10.1016/j.
method simulation of cylinder-shaped particle flow in a gas-solid compag.2008.10.002
fluidized bed. Chem Eng Technol 32:386–391. https://d oi.o rg/1 0. 174. Chung YC, Ooi JY (2008) A study of influence of gravity on bulk
1002/ceat.200800516 behaviour of particulate solid. Particuology 6:467–474. https://
157. Wu M, Peters B, Rosemann T, Kruggel-emden H (2019) A forc- doi.org/10.1016/j.partic.2008.07.017
ing fictitious domain method to simulate fluid-particle interaction 175. González-Montellano C, Ramírez A, Fuentes JM, Ayuga F
of particles with super-quadric shape. Powder Technol. https:// (2012) Numerical effects derived from en masse filling of agri-
doi.org/10.1016/j.powtec.2019.09.088 cultural silos in DEM simulations. Comput Electron Agric
158. Lu G, Third JR (2012) Critical assessment of two approaches 81:113–123. https://doi.org/10.1016/j.compag.2011.11.013
for evaluating contacts between super-quadric shaped particles 176. Markauskas D, Ramírez-Gómez Á, Kačianauskas R,

in DEM simulations. Chem Eng Sci 78:226–235. https://d oi.o rg/ Zdancevičius E (2015) Maize grain shape approaches for DEM
10.1016/j.ces.2012.05.041 modelling. Comput Electron Agric 118:247–258. https://d oi.o rg/
159. Wang S, Fan Y, Ji S (2018) Interaction between super-quadric 10.1016/j.compag.2015.09.004
particles and triangular elements and its application to hopper 177. Wiacek J, Molenda M (2011) Moisture-dependent physical prop-
discharge. Powder Technol 339:534–549. https://doi.org/10. erties of rapeseed: experimental and DEM modeling. Int Agro-
1016/j.powtec.2018.08.026 phys 25:59–65
160. Metzger MJ, Glasser BJ (2013) Simulation of the breakage of 178. Wojtkowski M, Pecen J, Horabik J, Molenda M (2010) Rape-
bonded agglomerates in a ball mill. Powder Technol 237:286– seed impact against a flat surface: physical testing and DEM
302. https://doi.org/10.1016/j.powtec.2012.12.006 simulation with two contact models. Powder Technol 198:61–68.
161. Potyondy DO, Cundall PA (2004) A bonded-particle model for https://doi.org/10.1016/j.powtec.2009.10.015
rock. Int J Rock Mech Min Sci 41:1329–1364. https://d oi.o rg/1 0. 179. Molenda M, Horabik J, Łukaszuk J, Wiacek J (2011) Variability
1016/j.ijrmms.2004.09.011 of intergranular friction and its role in DEM simulation of direct
162. Spettl A, Bachstein S, Dosta M et al (2016) Bonded-particle shear of an assembly of rapeseeds. Int Agrophys 25:361–368
extraction and stochastic modeling of internal agglomerate struc- 180. Sarnavi HJ, Noor Mohammadi A, Modares Motlagh A, Rahmani
tures. Adv Powder Technol 27:1761–1774. https://doi.org/10. Didar A (2013) Dem model of wheat grains in storage consider-
1016/j.apt.2016.06.007 ing the effect of moisture content in direct shear test. Res J Appl
163. Park B, Min K, Thompson N, Horsrud P (2018) Three-dimen- Sci Eng Technol 5:829–841. https://doi.org/10.19026/rjaset.5.
sional bonded-particle discrete element modeling of mechanical 5029
13
181. Iroba KL, Mellmann J, Weigler F et al (2011) Particle veloc- 199. Ma L, Ingham DB, Wen X (2000) Numerical modelling of the
ity profiles and residence time distribution in mixed-flow grain fluid and particle penetration through small sampling cyclones.
dryers. Granul Matter 13:159–168. https://doi.org/10.1007/ J Aerosol Sci 31:1097–1119. https://doi.org/10.1016/S0021-
s10035-010-0222-7 8502(00)00016-1
182. Mellmann J, Iroba KL, Metzger T et al (2011) Moisture content 200. Gibson MM, Launder BE (1978) Ground effects on pressure
and residence time distributions in mixed-flow grain dryers. Bio- fluctuations in the atmospheric boundary layer. J Fluid Mech
syst Eng 109:297–307. https://doi.org/10.1016/j.biosystemseng. 86:491–511. https://doi.org/10.1017/S0022112078001251
2011.04.010 201. Launder BE, Reece GJ, Rodi W (1975) Progress in the devel-
183. Patwa A, Ambrose RPK, Casada ME (2016) Discrete element opment of a Reynolds-stress turbulence closure. J Fluid Mech
method as an approach to model the wheat milling process. Pow- 68:537–566. https://doi.org/10.1017/S0022112075001814
der Technol 302:350–356. https://d oi.o rg/1 0.1 016/j.p owtec.2 016. 202. Wan G, Sun G, Xue X, Shi M (2008) Solids concentration simu-
08.052 lation of different size particles in a cyclone separator. Powder
184. He X, Jiao W, Wang C, Cao W (2019) Influence of surface rough- Technol 183:94–104. https://doi.org/10.1016/j.powtec.2007.11.
ness on the pump performance based on computational fluid 019
dynamics. IEEE Access 7:105331–105341. https://doi.org/10. 203. Xiang RB, Lee KW (2005) Numerical study of flow field in
1109/access.2019.2932021 cyclones of different height. Chem Eng Process 44:877–883.
185. Zhou L, Deshpande K, Zhang X, Agarwal RK (2020) Process https://doi.org/10.1016/j.cep.2004.09.006
simulation of chemical looping combustion using ASPEN plus 204. Gronald G, Derksen JJ (2011) Simulating turbulent swirling flow
for a mixture of biomass and coal with various oxygen carri- in a gas cyclone: a comparison of various modeling approaches.
ers. Energy 195:116955. https://doi.org/10.1016/j.energy.2020. Powder Technol 205:160–171. https://doi.org/10.1016/j.powtec.
116955 2010.09.007
186. Devarampally DR (2017) Development of two-way coupled 205. Parvaz F, Hosseini SH, Elsayed K, Ahmadi G (2020) Influence
CFD–DEM model for top spray fluid bed granular using star of the dipleg shape on the performance of gas cyclones. Sep
CCM +. The State University of New Jersey, Rutgers Purif Technol 233:116000. https://d oi.o rg/1 0.1 016/j.s eppur.2 019.
187. Drew DA (1983) Mathematical modeling of two-phase flow. 116000
Annu Rev Fluid Mech 15:261–291. https://doi.org/10.1146/ 206. Inc. AF (2001) Modeling Turbulence. In: ANSYS I, ANSYS
annurev.fl.15.010183.001401 Europe L (eds) ANSYS Fluent Guide, 12th ed. SAS IP, Inc.,
188. Yang Y, Zhou L, Shi W et al (2020) Interstage difference Canonsburg, pp 1–102
of pressure pulsation in a three-stage electrical submersible 207. Houben JJH, Weiss C, Brunnmair E, Pirker S (2016) CFD simu-
pump. J Pet Sci Eng. https://d oi.o rg/1 0.1 016/j.p etrol.2 020. lations of pressure drop and velocity field in a cyclone sepa-
107653 rator with central vortex stabilization rod. J Appl Fluid Mech
189. ANSYS Fluent Inc. (2009) Turbulence. In: ANSYS I, ANSYS 9:487–499
Europe L (eds) Ansys Fluent 12.0 Theory Guide, 12th ed. SAS 208. Talbi K (2011) An experimental study and a numerical simula-
IP, Inc., Canonsburg, pp 1–58 tion of the turbulent flow under the vortex finder of a cyclone
190. Launder BE (1972) Lectures in mathematical models of turbu- separator an experimental study and a numerical simulation of
lence. Academic Press, London, New York the turbulent flow under the vortex finder of a cyclone separator.
191. Launder BE, Spalding DB (1974) The numerical computation J Appl Fluid Mech 4:69–75
of turbulent flows. Comput Methods Appl Mech Eng 3:269– 209. Rotta JC (1962) Turbulent boundary layers in incompressible
289. https://doi.org/10.1016/0045-7825(74)90029-2 flow. Prog Aerosp Sci 2:1–95. https://doi.org/10.1016/0376-
192. Boysan F, Ayers WH, Swithenbank J (1982) A fundamental 0421(62)90014-3
mathematical modelling approach to cyclone design. Chem 210. Maxey MR, Riley JJ (1983) Equation of motion for a small rigid
Eng Res Des 60:222–230 sphere in a nonuniform flow. Phys Fluids 26:883–889. https://
193. Narasimha M, Brennan MS, Holtham PN (2006) Numeri- doi.org/10.1063/1.864230
cal simulation of magnetite segregation in a dense medium 211. Casas G, Mukherjee D, Celigueta MA et al (2017) A modular,
cyclone. Miner Eng 19:1034–1047. https://doi.org/10.1016/j. partitioned, discrete element framework for industrial grain dis-
mineng.2006.03.013 tribution systems with rotating machinery. Comput Part Mech
194. Orszag SA, Yakhot V, Flannery WS, Boysan F (1993) Renor- 4:181–198. https://doi.org/10.1007/s40571-015-0089-9
malization group modeling and turbulence simulations. In: 212. Lun CKK (2000) Numerical simulation of dilute turbulent gas-
International conference, Near-wall turbulent flows. Elsevier, solid flows. Int J Multiph Flow 26:1707–1736. https://d oi.o rg/1 0.
Tempe, Arizona, p 1031 1016/S0301-9322(99)00106-8
195. Azadi M, Azadi M, Mohebbi A (2010) A CFD study of the 213. Oesterlé B, Bui Dinh T (1998) Experiments on the lift of a spin-
effect of cyclone size on its performance parameters. J Hazard ning sphere in a range of intermediate Reynolds numbers. Exp
Mater 182:835–841. https://doi.org/10.1016/j.jhazmat.2010. Fluids 25:16–22. https://doi.org/10.1007/s003480050203
06.115 214. Saffman PG (1965) The lift on a small sphere in a slow shear
196. Soe TM, Khaing SY (2017) Comparison of turbulence models for flow. J Fluid Mech 22:385–400. https://doi.org/10.1017/S0022
computational fluid dynamics simulation of wind flow on cluster 112065000824
of buildings in mandalay. Int J Sci Res Publ 7:337–350 215. Saffman PG (1968) The lift on a small sphere in a slow shear
197. Gimbun J, Chuah TG, Fakhru’l-Razi A, Choong TSY (2005) The flow: corrigendum. J Fluid Mech 31:624–624. https://doi.org/
influence of temperature and inlet velocity on cyclone pressure 10.1017/S0022112068999990
drop: a CFD study. Chem Eng Process 44:7–12. https://doi.org/ 216. Mei R (1992) An approximate expression for the shear lift force
10.1016/j.cep.2004.03.005 on a spherical particle at finite reynolds number. Int J Mul-
198. Griffiths WD, Boysan F (1996) Computational fluid dynamics tiph Flow 18:145–147. https://doi.org/10.1016/0301-9322(92)
(CFD) and empirical modelling of the performance of a number 90012-6
of cyclone samplers. J Aerosol Sci 27:281–304. https://doi.org/ 217. Ren B, Zhong W, Chen Y et al (2012) Particuology CFD-DEM
10.1016/0021-8502(95)00549-8 simulation of spouting of corn-shaped particles. Particuology
10:562–572. https://doi.org/10.1016/j.partic.2012.03.011
13
218. Yuan J, Wu C, Li H et al (2018) Movement rules and screen- 237. Liu R, Zhou Z, Xiao R, Yu A (2019) CFD-DEM modelling of
ing characteristics of rice-threshed mixture separation through a mixing of granular materials in multiple jets fluidized beds. Pow-
cylinder sieve. Comput Electron Agric 154:320–329. https://doi. der Technol. https://doi.org/10.1016/j.powtec.2019.08.002
org/10.1016/j.compag.2018.09.012 238. Dallavalle JM (1948) Micromeritics: the technology of fine par-
219. Karimi H, Dehkordi AM (2015) Prediction of equilibrium mix- ticles, 2nd edn. Pitman Publishing Corp, New York
ing state in binary particle spouted beds: effects of solids density 239. ESSS-Rocky (2016) Particle drag laws implemented in the Rocky
and diameter differences, gas velocity, and bed aspect ratio. Adv DEM–CFD oneway and twoway coupling modules. In: Rocky E
Powder Technol 26:1371–1382. https://doi.org/10.1016/j.apt. (ed) Rocky-Dem Technical Manual, 4.13. Rocky DEM, Inc., Rio
2015.07.013 de Janeiro, Brazi, pp 1–9
220. Wang Z, Teng Y, Liu M (2019) A semi-resolved CFD–DEM 240. Wen CY, Yu YH (1966) Mechanics of fluidization. Chem Eng
approach for particulate flows with kernel based approxima- Prog Symp Ser 162:100–111
tion and Hilbert curve based searching strategy. J Comput Phys 241. Benzarti S, Mhiri H, Bournot H (2012) Drag models for simula-
384:151–169. https://doi.org/10.1016/j.jcp.2019.01.017 tion gas-solid flow in the bubbling fluidized bed of FCC particles.
221. Elghobashi S (1994) On predicting particle-laden turbulent flows. Int J Chem Mol Eng 61:1138–1143
Appl Sci Res 52:309–329. https://doi.org/10.1007/BF00936835 242. Ergun S (1952) Fluid flow through packed columns. Chem Eng
222. Elghobashi S (1991) Particle-laden turbulent flows: direct simu- Prog 48:89–94
lation and closure models. Appl Sci Res 48:301–314. https://d oi. 243. Knight C, O’Sullivan C, van Wachem B, Dini D (2020) Comput-
org/10.1007/BF02008202 ing drag and interactions between fluid and polydisperse particles
223. Zhao H, Zhao Y (2019) CFD–DEM simulation of pneumatic in saturated granular materials. Comput Geotech 117:103210.
conveying in a horizontal channel. Int J Multiph Flow 118:64–74. https://doi.org/10.1016/j.compgeo.2019.103210
https://doi.org/10.1016/j.ijmultiphaseflow.2019.06.003 244. Huilin L, Gidaspow D (2003) Hydrodynamics of binary fluidiza-
224. Elsayed K, Lacor C (2014) Analysis and optimisation of cyclone tion in a riser: CFD simulation using two granular temperatures.
separators geometry using RANS and LES methodologies. Tur- Chem Eng Sci 58:3777–3792. https://doi.org/10.1016/S0009-
bul Interact Notes Numer Fluid Mech Multidiscip Des 125:65– 2509(03)00238-0
74. https://doi.org/10.1007/978-3-662-43489-5 245. Di Felice R (1994) The voidage function for fluid-particle inter-
225. Potapov A, Chen X, Donohue T, Wheeler C (2013) Computer action systems. Int J Multiph Flow 20:153–159. https://doi.org/
simulation of airflow around transfer chutes via linked discrete 10.1016/0301-9322(94)90011-6
element method: computational fluid dynamics approach. In: 246. Wang Q, Melaaen MC, De Silva SR (2001) Investigation
11th international congress on bulk materials storage, Han- and simulation of a cross-flow air classifier. Powder Technol
dling and transportation 120:273–280. https://doi.org/10.1016/S0032-5910(01)00282-0
226. Ganser GH (1993) A rational approach to drag prediction non- 247. Badretdinov I, Mudarisov S, Lukmanov R et al (2019) Math-
spherical particles. Powder Technol 77:143–152. https://doi. ematical modeling and research of the work of the grain combine
org/10.1016/0032-59109(93)80051-b harvester cleaning system. Comput Electron Agric 165:104966.
227. ESSS-Rocky (2015) Rocky User Manual, 4.0. Rocky DEM, https://doi.org/10.1016/j.compag.2019.104966
Inc., Rio de Janeiro, Brazi 248. Mudarisov S, Khasanov E, Rakhimov Z et al (2017) Specifying
228. Mahajan VV, Nijssen TMJ, Fitzgerald BW, et al (2017) Flu- two-phase flow in modeling pneumatic systems performance of
idization of spherocylindrical particles. In: 8th international farm machines specifying two-phase flow in modeling pneumatic
conference on micromechanics on granular media systems performance of farm machines. J Mech Eng Res Dev
229. Buist KA, Jayaprakash P, Kuipers JAM et al (2017) Magnetic 40:706–715. https://doi.org/10.7508/jmerd.2017.04.018
particle tracking for nonspherical particles in a cylindrical flu- 249. Li Y, Xu L, Zhou Y et al (2018) Effects of throughput and oper-
idized bed. AIChE J 63:5335–5342. https://doi.org/10.1002/ ating parameters on cleaning performance in air-and-screen
aic.15854 cleaning unit: A computational and experimental study. Comput
230. Laín S, Sommerfeld M (2008) Euler/Lagrange computations Electron Agric 152:141–148. https://doi.org/10.1016/j.compag.
of pneumatic conveying in a horizontal channel with different 2018.07.019
wall roughness. Powder Technol 184:76–88. https://doi.org/10. 250. Enchen J, Zhanfeng S, Zhiyang P, Lijun W (2014) Numerical
1016/j.powtec.2007.08.013 simulation based on CFD-DEM and experiment of grain moving
231. Kuang SB, Yu AB, Zou ZS (2009) Flow regimes in vertical laws in Inrtia separation chamber. J Agric Mach 45:117–122.
pneumatic conveying. Ind Eng Chem Res 1145:1005–1008. https://doi.org/10.6041/j.issn.1000-1298
https://doi.org/10.1063/1.3179812 251. Balafoutis A, Beck B, Fountas S et al (2017) Precision agri-
232. Traoré P, Laurentie JC, Dascalescu L (2015) An efficient 4 culture technologies positively contributing to GHG emissions
way coupling CFD-DEM model for dense gas-solid particulate sustainability precision agriculture technologies positively con-
flows simulations. Comput Fluids 113:65–76. https://doi.org/ tributing to GHG emissions mitigation. Farm Prod Econ Sustain
10.1016/j.compfl uid.2014.07.017 9:1339. https://doi.org/10.3390/su9081339
233. Carruthers AC, Filippone A (2005) Aerodynamic drag of 252. Minfeng J, Yongqian D, Hongfeng Y et al (2018) Optimal struc-
streamers and flags. J Aircr 42:976–982. https://doi.org/10. ture design and performance tests of seed metering device with
2514/1.9754 fluted rollers for precision wheat seeding machine. IFAC Pap
234. Gent SP, Gent SP (2009) Computational modeling of multiphase OnLine 17:509–514. https://d oi.o rg/1 0.1 016/j.i facol.2 018.0 8.1 58
fibrous flows for simulation based engineering. Iowa State 253. Singh RC, Singh G, Saraswat DC (2005) Optimisation of design
University and operational parameters of a pneumatic seed metering device
235. Schiller L, Naumann AZ (1933) Ueber die grundlegenden for planting cottonseeds. Biosyst Eng 92:429–438. https://doi.
Berechnungen bei der Schwerkraftaufbereitung. Ver Deut Ing org/10.1016/j.biosystemseng.2005.07.002
77:317–320 254. Jianbo Z, Junfang X, Yong Z, Shun Z (2014) Design and experi-
236. Fox RW, Mcdonald’s AL, (2011) Introdution to fluid mechanics, mental study of the control system for precision seed-metering
8th edn. John Wiley and Sons, INC., New York device. Int J Agric Biol Eng 7:13–18. https://doi.org/10.3965/j.
ijabe.20140703.002
13
255. Kamgar S, Eslami M, Maharlouie M (2013) Design, development 261. Weigler F, Mellmann J, Franke G, Scaar H (2013) Experimental
and evaluation of a mechatronic transmission system to improve studies on a newly developed mixed-flow dryer. Dry Technol
the performance of a conventional row crop planter. Intl J Agron 31:1736–1743. https://doi.org/10.1080/07373937.2013.806929
Plant Prod 4:480–487 262. Scaar H, Franke G, Weigler F et al (2016) Experimental and
256. Dandan H, Dongxing Z, Li Y et al (2017) EDEM-CFD simula- numerical study of the airflow distribution in mixed-flow grain
tion and experiment of working performance of inside-filling dryers. Dry Technol 34:595–607. https://doi.org/10.1080/07373
air-blowing seed metering device in maize. Trans Chin Soc Agric 937.2015.1064946
Eng 33:23–31 263. Azmir J, Hou Q, Yu A (2019) CFD-DEM simulation of drying
257. Lei X, Liao Y, Zhang Q et al (2018) Numerical simulation of of food grains with particle shrinkage. Powder Technol 343:792–
seed motion characteristics of distribution head for rapeseed and 802. https://doi.org/10.1016/j.powtec.2018.11.097
wheat. Comput Electron Agric 150:98–109. https://doi.org/10. 264. Azmir J, Hou Q, Yu A (2019) CFD-DEM study of the effects of
1016/j.compag.2018.04.009 food grain properties on drying and shrinkage in a fluidised bed.
258. Weigler F, Mellmann J (2014) Investigation of grain mass flow Powder Technol. https://doi.org/10.1016/j.powtec.2019.10.021
in a mixed flow dryer. Particuology 12:33–39. https://doi.org/10.
1016/j.partic.2013.04.004 Publisher’s Note Springer Nature remains neutral with regard to
259. Bacelos MS, Almeida PIF (2011) Modelling of drying kinetic jurisdictional claims in published maps and institutional affiliations.
of potatoes taking into account shrinkage. Procedia Food Sci
1:713–721. https://doi.org/10.1016/j.profoo.2011.09.108
260. Oksanen T (2018) Controlling air flow in recirculating mixed
flow batch dryer with double bed mode. Comput Electron Agric
149:133–138. https://doi.org/10.1016/j.compag.2017.07.018
13

Theories and Applications of CFD-DEM Coupling Approach For Granular Flow: A Review

Uploaded by

Copyright:

Available Formats

Theories and Applications of CFD-DEM Coupling Approach For Granular Flow: A Review

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Theories and Applications of CFD-DEM Coupling Approach For Granular Flow: A Review

Uploaded by

Copyright:

Available Formats

Archives of Computational Methods in Engineering

Theories and Applications of CFD–DEM Coupling Approach

Received: 20 December 2019 / Accepted: 2 March 2021

Abbreviations di Particle diameter

Fijn Normal contact force on a particle

Alexandria University, Shatby 21526, Egypt Fij Force production by system rotation

Gb Kinetic energy generation due to buoyancy ΔT Timestep

T Time 𝜌a Air density

𝛼f Fluid volume fraction

K2 Newton’s shape factor

and animal diet. It sometimes contains a heterogeneous mass

Fig. 1 Number of published 700

DEM” (Accessed in December

Fig. 3 Contact forces analysis

Fig. 5 The relationship between damping ratio and the restitution

ary, respectively, and Z is representing the particle size (20)

2.3.2 Tangential Force Modeling

liquid, geometry and the volume of the bridge, and contact

Table 1 Grain models used for DEM simulation

4.2 RNG k − ε Model contributions to the pressure-strain correlation considered

ematical technique, namely “Renormalization Group”(RNG) [ ( )]

where nk is the xk component of the unit normal3∕4to the wall,d (| |)

Ls scheme to understand the main algorithm of CFD–DEM

Fig. 8 Schematic diagram of CFD–DEM coupling technique

Ti / Te Ti / TK and the interaction between the particles and turbulence is

Table 2 Drag correlation models used in CFD–DEM coupling

Dilute flow Schiller and Naumann 24 � �

Haider and Levenspiel 24 ( ) C For isometric non-spherical particles and

Ergun (1958) [100, 𝛼

Di Felice (1994) [29, CD = 𝛼f2−𝛽 CD,singleparticle 10−2 ≤ Rei ≤ 104

7 Applications to construct the separation, cleaning, sorting, grading, and

Velocity m.s-1 Velocity m.s-1 Velocity m.s-1 Velocity m.s-1

0.656 0.859 0.890 1.120

0.492 0.644 0.667 0.837

0.328 0.430 0.445 0.558

0.164 0.215 0.223 0.280

0.003 0.003 0.003 0.003

Fig. 13 Seed motion behavior of the seed-cleaning process [42]

(a) Velocity m.s-1 Velocity m.s-1 Velocity m.s-1 Velocity m.s-1

2.0064 3.4023 5.7960 6.2932

1.5056 2.5619 4.3572 4.7203

1.0048 1.7216 2.9184 3.1474

0.504 0.8812 1.4796 1.5745

0.0032 0.0408 0.0408 0.0016

2.0245 3.3994 4.4694 4.9254

1.5266 2.5549 3.3623 3.6994

1.7103 2.2551 2.4733

0.5306 0.8657 1.148 1.2473

0.0327 0.0212 0.0408 0.0212

Velocity m.s-1 Velocity m.s-1

streamlined angle 60º Radius of baseline 40 mm

Velocity m.s-1 Velocity m.s-1 Velocity m.s-1 Velocity m.s-1

7.41 8.455 8.905 10.526

5.557 6.341 6.679 7.898

3.705 4.227 4.452 5.271

1.852 2.113 2.226 2.643

3.834e-06 3.161e-06 3.04e-06 0.0163

Fig. 15 Snapshots of wheat and rapeseeds simulation in the distribution head [257]

Average parcle velocity, m.s-1

Abbreviations di Particle diameter

Fijn Normal contact force on a particle

Alexandria University, Shatby 21526, Egypt Fij Force production by system rotation

Gb Kinetic energy generation due to buoyancy ΔT Timestep

T Time 𝜌a Air density

𝛼f Fluid volume fraction

K2 Newton’s shape factor

Fig. 1 Number of published 700

Fig. 3 Contact forces analysis

Fig. 5 The relationship between damping ratio and the restitution

2.3.2 Tangential Force Modeling

Table 1 Grain models used for DEM simulation

4.2 RNG k − ε Model contributions to the pressure-strain correlation considered

Fig. 8 Schematic diagram of CFD–DEM coupling technique

Table 2 Drag correlation models used in CFD–DEM coupling

7 Applications to construct the separation, cleaning, sorting, grading, and

Fig. 13 Seed motion behavior of the seed-cleaning process [42]

Fig. 15 Snapshots of wheat and rapeseeds simulation in the distribution head [257]

Average parcle velocity, m.s-1

Parcle diameter mm 0s 50 s 100s 150 s 200 s