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

net/publication/330970285

CFD-DEM modelling and simulation of pneumatic conveying – A review

Article  in  Powder Technology · April 2020

DOI: 10.1016/j.powtec.2019.02.011

CITATIONS READS

59 1,373

3 authors:

Shibo Kuang Mengmeng Zhou

Monash University (Australia) Monash University (Australia)
89 PUBLICATIONS   1,998 CITATIONS    7 PUBLICATIONS   94 CITATIONS   

SEE PROFILE SEE PROFILE

Aibing Yu
UNSW Sydney
826 PUBLICATIONS   30,832 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

china policz View project

Effect of particle Shape in Blast Furnaces View project

All content following this page was uploaded by Shibo Kuang on 18 May 2020.

The user has requested enhancement of the downloaded file.

 Powder Technology 365 (2020) 186–207

Contents lists available at ScienceDirect

Powder Technology

journal homepage: www.elsevier.com/locate/powtec

CFD-DEM modelling and simulation of pneumatic conveying: A review

Shibo Kuang ⁎, Mengmeng Zhou, Aibing Yu ⁎
ARC Research Hub for Computational Particle Technology, Department of Chemical Engineering, Monash University, Clayton, Victoria 3800, Australia

a r t i c l e i n f o a b s t r a c t

Article history: Pneumatic conveying is an important operation used in many industries for the transportation of granular mate-
Received 25 August 2018 rials from one place to another. In recent years, many efforts have been made to study this transport process by
Received in revised form 18 January 2019 means of the numerical approach of combing discrete element method (DEM) for particles and computational
Accepted 8 February 2019
fluid dynamics (CFD) for gas. Such studies have been focused on understanding and predicting the gas-solid
Available online 12 February 2019
flow characteristics and process performance over a broad range of conditions, aiming to achieve optimal system
Keywords:
design and control. This paper reviews the progress in this area, covering the formulation, validation and appli-
Pneumatic conveying cation of the mathematical models proposed. The emphasis is given to the modelling of key phenomena such as
CFD-DEM flow regimes and transition, pipe wear, particle attrition, and electrostatics. The needs for future research are also
Flow regimes and transition discussed.
Pipe wear © 2019 Elsevier B.V. All rights reserved.
Particle attrition
Electrostatic charge

1. Introduction conveying of light and free-flowing granular products. Based on the

Pneumatic conveying has been widely used in various industries to
transport granular materials from one place to another. For example,
the solids processing industry estimates that the pneumatic conveying
system market will grow to $30 billion by 2025 [1]. Some advantages as-
sociated with this method of solid transportation include relatively high
levels of safety, low operational costs, flexibility of layout, ease of auto-
mation and installation, and low maintenance requirements. On the
other hand, pneumatic transport has some disadvantages with respect
to equipment wear, product degradation, electrostatics, and high
power consumption [2]. Therefore, how to minimize these disadvan-
tages is the key for the design, control, and optimization of pneumatic
conveying. This raises a need for better understanding and prediction
of pneumatic conveying characteristics.
In general, the pipelines of pneumatic conveying consist of a series of
straight sections and bends, which may be arranged horizontally or ver-
tically, or in an inclined matter. Also, the transport distance can vary
from a few meters up to a few miles, along which the compressible
gas somewhat expands as the pressure declines, leading to the decrease
of gas density and the increase of gas velocities. Depending on material
properties, geometries and operational conditions, different flow re-
gimes and their transition may take place along conveying pipelines.
For example, slug flow, unstable transition flow, stratified flow, and sus-
pension flow can in turn be observed when superficial gas velocity is in-
creased for a given solid flow rate in the horizontal pneumatic
⁎ Corresponding authors.
E-mail addresses: shibo.kuang@monash.edu (S. Kuang), aibing.yu@monash.edu
(A. Yu).
flow regimes involved, pneumatic conveying can be classified into two
modes: dense-flow and dilute-flow modes. In the dense-phase mode
like slug flow and stratified flow, the pressure drop decreases with the
increase of superficial gas velocity. However, this does not apply to the
transition flow occurring at gas velocities between those for slug and
stratified flows, and the pressure of the transition flow may increase
to be extraordinarily high, leading to very unstable solids transportation
and even pipeline blockage [3,4]. On the other hand, in the dilute mode
like suspension flow, the pressure drop increases as the superficial gas
velocity increases. Overall, dense-phase pneumatic conveying has less
power consumption, pipe wear and product damage compared than di-
lute-phase one which is, however, suited for a broader range of mate-
rials. These complicated features bring difficulties for developing a
general and reliable method for the design and control of a pneumatic
conveying system. In fact, in spite of the fact that there are many pneu-
matic conveying systems operating reliably throughout the world, the
lack of a comprehensive theory is still of concern [5].
In recent years, in line with the development of computational tech-
nology, numerical models have been widely used to reveal fundamen-
tals underlying pneumatic conveying and increasingly used to assess
process performance. The models may be continuum- or discrete-
based with respect to particles or the solid phase. The former is typically
represented by the so-called Two-Fluid Model (TFM). Because of its
computational convenience and efficiency, the TFM approach has been
used to study pneumatic conveying under different conditions (see,
[6–15], for example). However, the effective use of a continuum
model is heavily dependent of constitutive relations that are employed
to close the governing equations of the model. At present, the TFM stud-
ies have been applied mainly to pneumatic transport of powders or fine

https://doi.org/10.1016/j.powtec.2019.02.011
0032-5910/© 2019 Elsevier B.V. All rights reserved.
 S. Kuang et al. / Powder Technology 365 (2020) 186–207
particles [6–12], although mixtures of fine and coarse particles are
widely observed in practice. Moreover, it is yet a challenge to develop
a TFM model to reproduce all the typical flow regimes and transition.
The discrete-based approach is typically represented by the com-
bined approach of Computational Fluid Dynamics and Discrete Element
Method (CFD-DEM) or Lagrangian Particle Tracking method (CFD-LPT).
The latter, which is also named as CFD-DPM (Discrete particle Method)
in literature, can be thought as a simplified CFD-DEM model ignoring par-
ticle-particle interactions. Thus, the CFD-LPT approach is valid only to
pneumatic conveying conditions where particle-particle collision does
not play an important role [16–18]. Unlike the TFM approach, the CFD-
DEM approach does not need the complex constitutive relations between
the stress and strain tensors for discrete particles under different flow
conditions, and thus can apply to a wide range of flow systems. Moreover,
it can generate microscopic information such as the trajectories of and
forces acting on individual atoms or particles, which is important to un-
derstand the mechanisms underlying particle-fluid flows [19,20].
The CFD-DEM approach was firstly proposed by Tsuji et al. for the
studies of slug-flow pneumatic conveying [21] in 1992 and fluidization
[22] in 1993, and then followed by many others. This approach, after ra-
tionalized by Xu and Yu [23,24], has been widely accepted as an effective
tool to study particle-fluid systems [19,20]. Pneumatic conveying is one
of the major areas tackled by the CFD-DEM approach. Fig. 1 shows that
publications in this area have a rapid increase in recent years. Generally,
the earlier CFD-DEM models for pneumatic conveying have been based
mainly on some simplified models, where three-dimensional (3D) par-
ticle flow was sometimes treated as 2D and gas flow was 2D or 1D, lead-
ing to different models resulting from different combinations, such as 1D
CFD-3D DEM [21,25–29], 2D CFD-2D DEM [30–43], and 2D CFD-3D DEM
[44–46]. In recent years, efforts have been increasingly made to develop
and apply full 3D CFD-DEM models [47–73] and consider the effect of
particle shape [53,58,60–63,69,70,73]. These efforts are necessary to
generate realistic simulation results. In particular, at least 70% of raw
materials consist of non-spherical particles in modern industries and
particle shape plays a key role in determining the behaviors of bulk
solids [74]. Along with the model development, various key phenomena
in pneumatic transport have been modelled and analysed, covering flow
regimes especially slug flow [21,26–29,31,33,34,41,44,45,50,51,54–
56,59,71,72], flow transition [25,36,37,39,42,50,53–55,59], pipe wear
[48,58,64,70,75,76], particle attrition [30,35,52,69,73], electrostatics
[37–39,42,43,67,77], and heat transfer [32].
This paper reviews the development and application of CFD-DEM
models to pneumatic conveying, with reference to the formulation, val-
idation, and application of the models. It is structured as follows. In
Section 2, the model framework is outlined. In Section 3, the numerical
studies of various key phenomena are discussed. The emphasis is given
to the efforts in modelling flow regimes, flow transition, pipe wear, elec-
trostatic charge, particle attrition. In Section 4, the development of
50
CFD-DEM models
1D CFD-3D DEM
40
Number of journal papers
2D CFD-2D DEM
2D CFD-3D DEM
30
3D CFD-3D DEM
Particle shape modelling
20
All CFD-DEM models
10
0 tribution with a zero mean and standard deviation (2 k/3)1/2 [86]. To
1992 1997 2002 2007 2012
Year
Fig. 1. Number of publications related to CFD-DEM modelling and simulation of pneumatic
conveying since 1992, obtained from Web of Science.
187
large-scale simulations is discussed in handling industrial applications.
Section 5 provides a summary and discusses some needs for future re-
search and developments.
2. Framework for mathematical modelling
2.1. Governing equations for particle flow
The solid phase is treated as a discrete phase and described by the
DEM method originally proposed by Cundall and Starck [78]. In the
DEM approach, a particle has two types of motion: translational and ro-
tational, both obeying Newton's second law of motion. During its move-
ment, the particle may collide with its neighboring particles or with
walls and interact with the surrounding fluid, through which momen-
tum is exchanged. At any time t, the equations governing the transla-
tional and rotational motions of particle i in this flow system are:
kX
i þkw 
dvi 
mi ¼ f p‐ f ;i þ f c;ij þ f d;ij þ mi g ð1Þ
dt j¼1
and
kX
i þkw 
dωi 
Ii ¼ Tt;ij þ Tr;ij þ Tls;i ð2Þ
dt j¼1
where mi, Ii, vi and ωi are the mass, moment of inertia, translational and
angular velocities of particle i, respectively; fp-f,i represents the particle-
fluid force; fc,ij and fd,i are, respectively, the elastic and viscous contact
damping forces between particle i and particle/wall j; mig is the gravita-
tional force, and Tij is the torque acting on particle i due to particle/wall j.
For a particle undergoing multiple interactions, the forces and torques
are summed over the ki particles and the kw walls in contact with parti-
cle i. Different contact force models have been developed to calculate
the particle-particle and particle-wall interaction forces, as summarised
elsewhere [19,79]. Usually, the damping coefficient in a contact force
model is linked to the restitution coefficient that describes the loss of ki-
netic energy during a particle-particle/wall collision. Restitution coeffi-
cient may vary with impact velocity and particle/wall material
properties [79,80]. Its dependence on impact velocity is conditional in
DEM-based simulations, up to the contact force model involved [79].
For example, a linear spring-dashpot model provides a constant restitu-
tion coefficient, which can also occur to non-linear models under certain
conditions [81]. Nevertheless, Table 1 lists the commonly used non-lin-
ear models [82–85].
The particle-fluid force fp-f,i, as listed in Table 2, is the sum of all types
of particle-fluid interaction forces acting on individual particles by fluid.
The modelling of pneumatic conveying includes mainly the fluid drag
force fdrag,i and pressure gradient force f∇p,i, and in some occasion, vis-
cous force f∇⋅τ,i, virtual mass force fvm,i, and lift forces such as the
Saffman force fls,i and Magnus force flm,i. When the gas flow is turbulent,
a particle experiences at its position an instantaneous fluid velocity,
which is stochastic. The effect of this stochastic flow behavior on the
particle, namely, turbulent dispersion effect, should be taken into ac-
count. The instantaneous fluid velocity u ~ consists of the mean velocity
u and the fluctuating velocity u′. Different models may be used to obtain
instantaneous fluid velocities. For example, when the gas flow is solved
using the Reynolds-Averaged Navier–Stokes (RANS) equations, the
mean velocity is obtained by the solution of RANS equations, while
the fluctuating velocity has to be reconstructed from the turbulent ki-
netic energy given by a gas turbulence model according to a Gauss dis-
2017
estimate turbulent dispersion effect in a CFD-DEM simulation, fluid
drag forces are calculated using instantaneous gas velocities directly
other than by mean velocities as done usually. Note that in the CFD-
LPT approach, a particle is assumed to interact with the fluid phase
188 S. Kuang et al. / Powder Technology 365 (2020) 186–207

Table 1
Non-linear contact forces and torques acting on particle i.

Forces and torques Symbol Equations

pffiffiffiffiffi
Normal forces Contact fcn,ij − 43 Y  R δ3=2
n n
^
Damping fdn,ij 
pffiffiffiffiffiffiffiffiffiffi 1=2
−γ n ð6mij Y R δn Þ vn;ij
Tangential forces Contact fct,ij −μ s jf cn;ij j½1−ð1− minfjδt;ij j; δt;ij; max g=δt;ij; max Þ3=2 δ ^t
Damping fdt,ij pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 1=2
‐γ t ð6μ s mij j f cn;ij j 1−δt;ij =δt;ij; max =δt;ij; max Þ vt;ij ðδt;ij bδt;ij; max Þ
Torque by tangential forces Tt,ij Rij × (fct, ij + fdt, ij)
_
Rolling friction torque Tr,ij μ r;ij di j f n;ij jωt;ij
_
where, R1 ¼ jR1i j þ jR1j j, Y  ¼ 2ð1−v
Y ^ Ri ω t;ij ^ δt 2−v ^ ^
2 Þ, n ¼ jR j, ωt;ij ¼ jω j, δt ¼ jδ j, δt;ij;max ¼ μ s 2ð1−vÞ δn , vij = vj − vi+ωj × Rj − ωi × Ri, v n;ij ¼ ðv ij  nÞn, vt, ij = vij − vn, ij
i t;ij t

eddy over the smaller of the eddy lifetime and the eddy crossing time
[87]. Thus, when this time is reached, a new value of the instantaneous
velocity needs to be reconstructed even within a given time step. This
situation essentially requires a variable time step, which is yet difficult
to implement in the CFD-DEM approach where a fixed time step is
often used. In addition, to be general, the drag correlation used needs
to cover a wide range of porosities and flow regimes. On account of
this, Ergun-Wen-Yu's correlation [88,89] and De Felice's correlation
[90] are often used, which have been formulated based on extensive ex-
perimental data. Recently, various numerical simulations have also been
conducted on a sub-particle level to establish more general drag corre-
lations for considering, for example, the effects of particle shape [91,92],
particle size [93,94], porosity [95], and non-Newtonian fluid rheology
[96–98].
The torque acting on particle i due to particle j includes two compo-
nents. One arises from the tangential forces given by Tt,ij = Ri,j × (fct,ij +
fdt,ij), where Ri,j is a vector from the centre of mass to the contact point,
_
and another is the rolling friction torque given by Tr;ij ¼ μ r;ij di jf n;ij jωt;ij ,
where μr,ij is the (dimensionless) rolling friction coefficient and di is par-
ticle diameter [85]. The second torque is attributed to the elastic hyster-
esis loss and viscous dissipation in relation to particle-particle or
particle-wall contacts, and it causes the decay in the relative rotational
motion of particles. For viscoelastic material, it has been reported that
μr,ij together with other parameters of DEM simulation can be evaluated
as a function of material properties [99,100]. Also, because of the inter-
action between particle and viscous fluid, a torque Tls,i acting on the par-
ticle by the fluid may be considered [101], which can be determined by
the correlation listed in Table 2.
2.2. Governing equations for gas flow
The gas flow is treated as a continuous phase and modelled in a sim-
ilar way to the one in the conventional two-fluid modelling. Thus, the
governing equations are the conservation of mass and momentum in
terms of local mean variables over a computational cell. Generally,
three model formulations are available [51,102]: an original format
Set I, and subsequent derivations of Set II and Set III respectively corre-
sponding to the so-called Model A and Model B in literature. Sets I and II
result from different treatments of fluid stress tensor term in the origi-
nal equation, but they are essentially the same. Set III is simplified
from Set I by adopting certain assumptions, e.g. steady and uniform
flow when calculating the pressure gradient and viscous forces. Corre-
spondingly, the mass and momentum equations of gas phase are re-
spectively given as follows:
 
∂ ρf εf  
þ ∇  ρf εf u ¼ 0 ð3Þ
∂t

Table 2
Particle-fluid forces and torques acting on particle i.

Forces/torques Equations Refs

Fluid drag force, fd,i f d;i ¼

2 2−βðε;h ReiÞ [93,156]
8 C D0 πρ f αðdi ; xi ; εÞyi hdi ju−vi jðu−vi Þε f
1

2
βðε f ; h ReiÞ ¼ 2:65ðε f þ 1Þ−ð5:3−3:5ε f Þε2f exp½− ð1:5− logh
2
ReiÞ

0:5ð1þε f Þ
αðdi ; xi ; ε f Þ ¼ X x þ 0:5yi þ 0:5ð1−ε f Þyi 2
i

i
yi 2
−1 di ε f ju−vi jhdi
hdi ¼ ½∑ xi =di  , yi ¼ hdi , h Rei ¼ ν
i
Pressure gradient force, f∇p,i − ∇ PVp [157]
Viscous force, f∇⋅τ,i εs ∇ ⋅ τ [157]
Virtual mass force, fvm,i f vm;i ¼ 12 ρ f V p ð∂u þ uð∇  uÞ− dv i
Þ [158]
∂t dt
Saffman lift force, fls,i 2 pffiffiffiffiffiffiffiffi ðu−vi ÞD [159,160]
f ls;i ¼ 8 6:46C ls di ρ f μ pffiffiffiffi
1
D∥
−1
D = (∇u) + (∇u)
 −0:1 Re pffiffiffiffiffiffi −0:1 Re
e p
p −0:3314
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi β ls ðe
p
þ 1Þ Rep ≤40
C ls ¼
0:0524 βls Rep Rep N40
pffiffiffiffi
Dd
βls ¼ 0:5 ju−v∥ i pj 0:005 b β ls b 0:4
Magnus lift force, flm,i 3 Rep [161]
f ls;i ¼ π8 di ρ f Rer C lm ½Ω  ðu−vi Þ
Ω = 0.5 ∇ × u − ωiRer = ρfd2i |Ω|/μRep, i = ρfdi|u − vi|/μ Rep, i ≤ 2000
0:4
Re0:3
C lm ¼ 0:45 þ ð Repr −0:45Þe−0:05684 Rer
Re p

Torque due to particle-fluid interaction, Tls,i ρf 3 5 [101]

Tls;i ¼ d ðdi Þ C lmt jΩjΩ
2
( i 2
64π
Rer Rer ≤32
C lmt ¼
Rer þ Rer b Rer ≤1000
12:9 128:4
32
Fluid drag force obtained using instantaneous gas velocity, fd,t,i ~ −vi Þ
f d;t;i ¼ f d;i ðu [86]
~ ¼ u þ u0
u
 S. Kuang et al. / Powder Technology 365 (2020) 186–207 189

and 2.3.1. Property mapping from computational cell center to individual

  particle
∂ ρf εf u   In a CFD cell, there are many particles which may reside at any point
Set I : þ ∇  ρ f ε f uu ¼ −∇P− FSet I
p− f þ ∇  τ þ ρ f ε f g ð4aÞ of the cell, whereas properties of the cell, such as gas velocity and pres-
∂t
sure drop, are usually stored in the center of the cell. To correctly calcu-
Pkc } late the particle-fluid force acting on a particle, a mapping technique
where FSet I
p− f ¼ ΔV
1
i¼1 ðf drag;i þ f ∇p;i þ f ∇τ;i þ f i Þ and fp−f, i = fdrag, i
that transforms a cell-based property into a particle-based property is
+ f∇p, i + f∇⋅τ, i + fi"
required. The least-square interpolation [104] can be used, which ex-
  presses a gas property at a particle position as
∂ ρf ε f u  
Set II : þ ∇  ρ f ε f uu ¼ −ε f ∇P−FSet II
p− f þ ε f ∇  τ φp ¼ φcell þ ∇φcell  Δr ð6Þ
∂t
þ ρf εf g ð4bÞ
where φp and φcell are the gas properties at the particle position and the
Pkc }
where FSet II
p− f ¼ ΔV
cell center, respectively; and Δr is the distance vector directed from the
i¼1 ðf drag;i þ f i Þ and fp−f, i = fdrag, i + f∇p, i + f∇⋅τ, i
1

"
+ fi
 
∂ ρf εf u  
Set III : þ ∇  ρ f ε f uu ¼ −∇P−FSet III
p− f þ ∇  τ þ ρ f ε f g ð4cÞ
∂t
Pkc } Pkc
where FSet III
p− f ¼ ε f ΔV
1
i¼1 ðf drag;i þ f i Þ− ΔV
1
i¼1 ðρ f V p;i gÞ
"
(fdrag, i + fi )/εf − ρfVp, ig
In Eqs. (4a), (4c), ρf, u, P and τ are the fluid density, fluid velocity and
pressure, and fluid viscous stress tensor, respectively, f"i = fvm, i + fls, i
+ flm, i, and τ is given by an expression analogous to that for a Newto-
nian fluid. That is
 h i
τ ¼ ηlaminar þ ηturbulent ð∇uÞ þ ð∇uÞ−1
where ηlaminar is fluid molecular viscosity, and ηturbulent is the turbulent
viscosity which can be obtained using turbulence models depending on
applications, such as Spalart-Allmaras, k-epsilon, k-omega, Transition k-
kl-omega, Transition SST, Reynolds Stress, Detached Eddy Simulation, or
Large Eddy Simulation (LES), or direct numerical simulation [103].
In theory, both Model A and Model B come from the original Model B.
The difference between original Model B and Model A is small, but Model
B adopts certain assumptions, e.g. steady and uniform flow when calcu-
lating the PGF and viscous force. Not surprisingly, it may have limitations.
These limitations do not cause serious problems for simple flow systems,
e.g., gas fluidization or pneumatic conveying where the drag and pressure
gradient forces are largely in the same direction and the acceleration of
pure fluid flow is negligible. However, for the complex flows, e.g., in
hydrocyclones where the magnitude of the pressure gradient force is sig-
nificant, and its direction is completely different from that of the drag
force, Model B predicts different results from Model A and original
Model B. To overcome this problem, Set I formulations (or the original
Model B) and Model A should be used. Moreover, as Model B has different
eigenvalues compared with Model A, the use of the original Model B
could solve the ill-posed problem associated with Model A.
2.3. Coupling schemes between particle and fluid flows
The modelling of particle flow by DEM is at an individual particle
scale, whereas the gas flow by CFD is at a computational cell scale.
Their coupling can be numerically achieved as follows according to
the work of Xu and Yu [23]. At each time step, the DEM provides infor-
mation, such as the positions and velocities of individual particles, for
the evaluation of porosity and volumetric particle-fluid force in each
computational cell. The CFD model then uses this information to com-
pute the gas flow field, which is used to obtain the particle-fluid forces
acting on individual particles. Incorporation of the resulting forces into
the DEM enables the determination of the motion of individual particles
for the next time step. The particle-fluid forces acting on individual par-
ticles will react on the fluid phase so that Newton's third law of motion
is satisfied. Some techniques useful to its implementation in a simula-
tion are given below.
cell center to the particle position. The gradient of the gas property
∇φcell at the cell center is evaluated by a weighted least-square method
[105].
2.3.2. Calculation of porosity and source terms
Fluid drag force and governing equations for gas phase are closely
and fp−f, i = associated with porosity that is usually calculated at a computational
cell level. By definition, the porosity value can be determined by 1.0
minus the ratio of the total volume of all particles in a computational
cell to that of the cell. To be computationally efficient in calculating po-
rosity, a particle is regarded to be completely in a computational cell as
long as its center is located within the space enclosed by the boundaries
of the cell, while errors arising from particles lying only partially within
ð5Þ a cell have largely been neglected. Neighboring cells can be included in
the calculation to minimize the errors [36,106]. To be more accurate,
only the partial volumes of particles that are included in a computa-
tional cell can be taken into account in porosity calculation when the
particles are located over the cell boundaries. In the past, various
methods have been proposed to calculate the partial volumes of parti-
cles based on, for example, analytical methods [107] and particle
meshing methods [108,109]. Also, some other approaches based on
local averaging theory [44], two grids [110], or diffusion-based averag-
ing theory [111,112] have been developed for handling complex geom-
etries meshed with various shapes and sizes of CFD cells. The diffusion-
based averaging method is established for spherical particles and its ex-
tension to non-spherical particles is not clear at present. The two-grid
method needs the mapping procedure between extra Cartesian grid
and original grid, causing accuracy loss. These problems do not occur
to the local averaging theory-based method, which can consistently
treat source terms imposed by particles on fluid. In this method, a prop-
erty at any point is calculated using a spherical cell, following the ap-
proach used in the local averaging of granular materials [113].
Therefore, the local porosity ε corresponding to a particle, usually de-
fined as the porosity value at the center of the computational cell
where the particle is residing, can be expressed as
PNp;sphere
asphere;i V p;i
ε ¼ 1− i¼1
ð7Þ
ΔV sphere
where ΔVsphere is the volume of a spherical cell; Np,sphere and asphere,i are,
respectively, the number of particles and the fractional volume of parti-
cle i residing in the spherical cell. Correspondingly, the volumetric par-
ticle-fluid interaction force in the considered cell is calculated by
PNp;sphere
ashpere;i f p− f ;i
Fp− f ¼ i¼1
ð8Þ
ΔV shpere
Eqs. (7) and (8) also apply to spherical cells close to wall boundary,
where the region out of the computational domain is not considered. At
this stage of development, the size of such a spherical cell has to be de-
termined case by case [19,113]. The size of spherical cell can be simply
set to 3 particle diameters in many cases [50,54–56,59,72].
190 S. Kuang et al. / Powder Technology 365 (2020) 186–207
3. Modelling and simulation of key pneumatic conveying
characteristics
3.1. Slug-flow pneumatic conveying
Slug-flow conveying mode is one of the most important recent de-
velopments in the field of pneumatic conveying. In such an operation,
particles are conveyed in the form of slugs that fill the entire cross sec-
tion. The slugs or plugs, separated by gas pockets, move gently along
the pipeline. This type of flow has the advantage over others that less
gas is needed to transport a given mass of solids. Moreover, pipe wear,
dust generation and product attrition generated by dilute-phase con-
veying can be significantly reduced [114]. However, the behaviors of
slug flow are very complicated and difficult to be described reliably by
a theoretical or empirical method, which restricts industrial applica-
tions. In particular, slug flow may have a high solid concentration,
approaching packing density within a slug. Also, particles may experi-
ence quasi-static flow, rapid flow, and transitional flow that lies be-
tween. These features raise challenges for the CFD-LPT and TFM
models, where constitutive relations have not been established for de-
scribing solid stresses governing slug behaviors together with particle-
fluid forces. The CFD-DEM approach does not need the global assump-
tions introduced into CFD-LPT and TFM models and thus is theoretically
more rational. Therefore, it can confidently predict complicated slug
flow. As such, since 1992, the CFD-DEM approach has been widely
used to understand the behaviors of slug flow in feeder [31,72], horizon-
tal pipe [21,25,26,28,29,33,34,36,41,44,45,51,54,56,59,71], vertical pipe
[26,27,36,37,50,71,72], inclined pipe [37,55] or bend [71]. Below dem-
onstrate the distinct features of CFD-DEM model in studying slug flow
from four aspects.
All in all, the CFD-DEM approach can realistically reproduce the
complicated flow patterns of slug flow under different conditions,
which has never been achieved by any other numerical models, to the
best of our knowledge. Fig. 2 presents such a typical example where dif-
ferent slug behaviors in the vertical pipe of a newly developed feeder
have been obtained by the high-speed CCD camera and 3D CFD-DEM
model in the transport of 6-mm glass beads. It is shown that under
the condition considered, slugs with different lengths are formed.
Also, two slugs can be merged into one slug, while some slugs may be
pierced through by the air flow and collapse. Overall, the experimental
and numerical results are in reasonable agreement.
Fig. 3 shows the process of slug formation in horizontal pneumatic
conveying, where the behaviors of particles are more complicated com-
pared than those of a macroscopically stable slug. It can be seen that
Fig. 2. Comparison of slug flow patterns at Ug = 5.79 m/s in the experiment and simulation [72]: (1) slugs with random length, (2) downward merging and (3) slug collapse.
particles go through four stages to finally form a stable slug after
they are loaded into a horizontal pipe from the top of the pipe, in-
cluding particle accumulation in the loading zone (being highlighted
by the solid red line) (Fig. 3a), formulation of the settled layer in the
loading zone and the downstream region (Fig. 3b), formation of a
relatively short slug (Fig. 3c–d), and growth in length into a stable
slug (Fig. 3e).
Fig.4 shows the predicted slug formation in a newly developed
feeder. It can be observed that particles, loaded into a draft tube type
feeder, are entrained into the vertical pipe by the gas-particle flow
through particle-fluid and particle-particle interactions. These particles
initially form small particle clouds or clusters at the entrance of the ver-
tical pipe, and are then merged into small slugs, which join together to
grow into a stable slug as they move upwards.
Here, it should be pointed out that feeder is the important part of a
pneumatic conveying system, however, the relevant research is not as
much as might be expected. Moreover, there are various types of
feeders, featured with complex geometries, and in some situations, mo-
bile parts. It would be interesting to apply the CFD-DEM approach to un-
derstand the complicated gas and solid behaviors within feeders. The
results will, ultimately, help reduce the dependence of feeder design
and control on empirical experience.
Second, the CFD-DEM approach can be used to quantify the process
parameters of slug-flow conveying, such as pressure drop, slug length,
slug velocity, and settled layer thickness etc. For example, Fig. 5 gives
the predictions of slug velocity and settled layer thickness, which are
two main parameters widely used to characterize horizontal slug flow.
As seen from this figure, the slug velocity increases linearly with gas ve-
locity in both the predicted results and experimental measurements,
and they are in a good agreement with each other. It also shows that
the relations between the cross-sectional area ratio of settled layer to
pipe and the average particle velocity in a slug, predicted respectively
by the CFD-DEM simulations and the analytical model, reasonably
agree with each other.
Third, the CFD-DEM approach can provide some insights into the
particle and gas behaviors, which are not easy to obtain experimentally.
Fig. 6 gives an example showing the prediction of the characteristics of a
stable slug. As a slug moves, it picks up particles in the front settled layer
and leaves behind a new settled layer, and thus particles in the slug keep
updating with time (Fig. 6a–c). Also, particles in the center of a settled
layer move into the upper part of a slug while particles in the lower
part of a settled layer move into the lower area of the slug. As expected,
it is observed that the solid concentration in a slug is, to some degree,
non-uniform in radial and axial directions, however, its average value
 S. Kuang et al. / Powder Technology 365 (2020) 186–207 191

Fig. 3. Slug formulation in the slug flow regime: (left) particle flow patterns, and (right) gas flow fields [56].

Fig. 4. Predicted slug formation in a newly developed feeder [72].

Fig. 5. Comparisons of 3D CFD-DEM predictions [50] of horizontal slug flow with the experimental work of Tsuji et al. [21] and the analytical model of Konrad and Davidson [162]: slug
velocity (left) and settled layer thickness (right).
192 S. Kuang et al. / Powder Technology 365 (2020) 186–207

over the cross-sectional pipe area fluctuates around a constant (Fig. 6d).
Also, the overall pressure drop in the pipe is mainly contributed by the
pressure difference across the slug.
Forth, CFD-DEM simulations can generate particle-scale forces sta-
tistics that govern particle behaviors, including particle-particle/wall
forces and particle-fluid forces. The analysis of such forces is useful to
elucidate complicated slug behaviors. Fig. 7 shows the predicted parti-
cle-particle and particle-fluid forces inside a slug for horizontal pneu-
matic conveying. The result suggests that the axial particle-fluid forces
inside a slug are much bigger than the radial ones. The axial particle-
fluid and particle-wall interactions macroscopically control the move-
ment of a slug, whereas the particle-particle interaction microscopically
causes a slug to sweep up particles in a settled layer. Model studies re-
veal that the magnitudes of these interaction forces increase with the
increase of gas or solid flow rate [44].
The main problem associated with the pressure drop prediction for
slug flow is due to the lack of reliable and general constitutive relations
for describing solid stresses, especially the stress transmission ratio
from axial to radial directions [115,116]. CFD-DEM simulations are
promising to overcome this problem, as detailed follows. By use of an
averaging procedure, a discrete particle system can be transformed
into a corresponding continuum system, leading to the establishment
of balance equations of mass, linear momentum and angular momen-
tum for continuum-based modelling. Such a procedure can be based
on volume, time-volume or weighted time-volume averaging method,
as reviewed by Zhu et al. [19]. Under the support of a proper averaging
theory, the macroscopic properties of solid flows such as mass density,
velocity, angular velocity, stress tensor, and couple stress tensor can
be investigated based on the microscopic information generated by
DEM simulations [117,118]. Based on such studies, constitutive

Fig. 6. Snapshots showing the spatial distributions of flow properties when the number of particles = 11,000 and gas velocity = 2.09 m/s [44]: (a)-(c), particle configurations at 14.236 s,
14.498 s and 14.760 s; (d), porosity; (e), axial particle velocity; (f) radial particle velocity; (g), gas velocity profile; (h), gas streamlines; and (i), axial pressure drop; (d)-(i) are
corresponding to (b).
 S. Kuang et al. / Powder Technology 365 (2020) 186–207 193

Fig. 7. Snapshots showing spatial distributions of: (a), axial particle-fluid force; (b), radial particle-fluid force; and (c), network of normal contact forces [44].

relations can be developed, as demonstrated by Zhou et al. [119] in their
study of ironmaking blast furnaces. Using an averaging approach,
Stratton and Wensrich [45] have revealed the stress state of a slug, as
shown in Fig. 8. The result indicates the regions of passive stress at the
top-front and rear of the slug and the active stress at the bottom-front
of the slug. It is found that the stress transmission ratio has strong var-
iations in all available directions, which is in contrast to the constant
ratio assumption of Konrad [115], although Konrad's correlation and
its variants have been widely used to predict the pressure drop of a
slug flow.
Because of the challenge of establishing reliable continuum theories/
models for granular flow [19,20,120,121], at present the stress
transmission ratio needed for predicting the pressure drop of pneumatic
conveying cannot be described theoretically. To overcome this problem,
a more general averaging theory has to be established, especially in the
proper consideration of the wall effect, which plays an important role in
the process of pneumatic conveying. On the other hand, various exper-
imental devices have been developed in the past to measure the interac-
tion forces between slug and wall [122–126]. These devices can be used
to validate and improve CFD-DEM approach and averaging theory or
vice versus. This would be useful for tackling not only the specific chal-
lenge in predicting the pressure drop of slug-flow pneumatic conveying,
but also those encountered in the development of constitutive relations
for granular flow.

Fig. 8. Two-dimensional axial normal material stress field, lateral normal material stress field, stress transmission ratio field, and maximum material shear stress field. Stress values are
calculated as averages over time interval 10.00 to 14.99 s sampled at 100 Hz [45].
194 S. Kuang et al. / Powder Technology 365 (2020) 186–207

Fig. 9. Snapshots showing representative particle flow patterns at different gas velocities in vertical pneumatic conveying [50]: (a) Ug = 10.0 m/s, (b), Ug = 13.0 m/s, (c), Ug = 18.5 m/s,
(d) Ug = 30.5 m/s, and (e), Ug = 30.5 m/s, corresponding to different times.

3.2. Flow transition
Depending on material properties, pipe geometries, and operational
conditions, different flow regimes and their transition take place along
conveying pipelines. The determination of flow transition mode and
further a suitable flow regime represents two major tasks in the design
of a pneumatic conveying system [5]. The former is often based on ma-
terials properties, in particular, those properties which involve particle-
gas interaction (e.g., permeability, air retention and de-aeration) [127].
The latter is achieved by determining the boundaries between which a
flow regime occurs [4,128]. However, how to develop a general and
cost-effective way of completing these two tasks has been a challenge.
The CFD-DEM approach is promising in solving this problem.
3.2.1. Prediction of flow transition
The CFD-DEM approach has been proved capable of predicting typi-
cal flow regimes and their transition when some materials are
transported via a specific pipeline system at varying gas velocities and
a fixed solid flow rate. Fig. 9 demonstrates this aspect using the

Fig. 10. Snapshots showing representative particle flow patterns at different gas velocities in horizontal pneumatic conveying [54].
 S. Kuang et al. / Powder Technology 365 (2020) 186–207 195

Fig. 11. (a) Flow-transition-modes shown in pressure drop-gas velocity diagram, and (b) their relation with friction and restitution coefficients [59].

predictions of vertical pneumatic conveying. As seen from this figure,
the slug flow regime is expectedly obtained numerically at the lowest
gas velocity. With increasing gas velocity, the slug length decreases
until slugs cannot be observed at relatively high velocities at which
the dispersed flow regime is developed.
In horizontal pneumatic conveying, flow regimes exhibit more com-
plicated behaviors compared to those in vertical one because the gravity
is perpendicular to the conveying direction here. The typical flow re-
gimes including slug flow regime, transition flow regime between slug
and stratified flow regimes, and stratified flow regime and dispersed
flow regime, can successfully be reproduced by the CFD-DEM model,
as shown in Fig. 10. These flow regimes have also been obtained in the
modelling of 45° inclined pneumatic conveying [55], which show simi-
lar features as shown in Fig. 10.
Pressure drop-gas velocity diagram is a common way used to de-
scribe flow transition, as shown in Fig. 11a. This diagram shows how dif-
ferent flow regimes transit from one to another when superficial gas
velocity is increased at a given solid flow rate for specific materials
and conveying system. Generally, three transition modes can be identi-
fied from a pressure drop-gas velocity diagram: Mode I for dilute-phase
only, Mode II for unsmooth transition from dilute-phase, to unstable-
zone, finally to slug-flow, and Mode III for smooth transition from di-
lute-phase to slug-flow [127]. As shown in Fig. 11, these three flow tran-
sition modes are obtained by CFD-DEM simulations where variables are
particle friction and restitution coefficients. A flow transition map has
been established to predict different transition modes resulting from
the change of friction or restitution coefficient, as given by Fig. 11b. It
is shown that Mode III is suitable for particles of large friction and resti-
tution coefficients; Mode I for particles of relatively low friction coeffi-
cient; and Mode II for particles with properties not covered by Modes
I and III.
Based on CFD-DEM simulations, some other properties, other than
gas velocity and pressure drop, have also been proposed to describe
the flow transition in horizontal conveying. Fig. 12a is an ellipticity-
pressure gradient phase diagram used to consider the effect of particle
shape on the flow transition in a horizontal square pipe. It can be seen
that cuboidal particles show very similar behavior to high aspect ratio
ellipsoidal particles. As the pressure gradient increases, cuboidal parti-
cles transit from fixed bed flow (A) through sheared bed flow (i.e. strat-
ified flow) (B) directly to dilute flow (C) with no slug formation.
Although useful to understand the effect of particle shape, this elliptic-
ity-pressure gradient phase diagram, unlike the conventional gas

Fig. 12. (a) Ellipticity-pressure drop [53], and (b) solid flow rate - gas velocity phase diagrams [36].
196 S. Kuang et al. / Powder Technology 365 (2020) 186–207

velocity-pressure drop diagram, can hardly be used to determine a) at the pipe scale:
suitable solid and gas flow rates for applications. This problem does  8 2 0 139
 =
not occur to Fig. 12b, which uses solid and gas flow rates to describe  1 X< 1 Np;X section kX
i þkw
 4 @ f cn;ij þ f dn;ij þ f ct;ij þ f dt;ij A5 
the flow transition, but misses the pressure drop which is one of the Δt
 sp Δt sp :V pipe i¼1 j¼1
;
main industrial concerns in designing and controlling pneumatic Π1;pipe ¼ " !# Π2;pipe
1 X 1 X
Np; section
conveying.
mg i
Δt sp Δt V pipe
 "
sp i¼1
!#

 1 X 1 Np;X section

3.2.2. New phase diagram in terms of forces acting on particles  f drag;i þ f pgf ;i 
Δt V
To better understand the flow transition, the pressure drop, which is  sp Δt sp pipe i¼1 
often used together with gas velocity to demarcate the dense-phase and ¼ " !#
1 X 1 X
Np; section
dilute-phase flows, can be linked to the microscopic forces by introduc- mg i
ing the concept of volumetric force, for example, see Fig. 13. Such anal- Δt sp Δt V pipe i¼1
sp

ysis of CFD-DEM results has been applied to horizontal, vertical and ð9Þ
inclined pneumatic conveying [50,54–56,59]. It assesses the flow transi-
tion using the forces that contribute the pressure drop. As seen from Fig. b) at the particle scale:
13, the pressure drop can be well represented by the volumetric parti- 8 2 0 139
cle-fluid force and fluid-wall friction force. The former balances with 1 X < 1 4 p;X =
N section kX i þkw  
@  f cn;ij þ f dn;ij þ f ct;ij þ f dt;ij A5
the gravity and particle-wall friction force in the flow direction. The Δt sp Δt :V pipe ;
i¼1 j¼1
roles of different forces vary with flow regime and pipe orientation. Π1;particle ¼
sp
" !# Π2;particle
1 X 1 X
N p; section
Overall, the pressure drop is dominated by the particle-wall friction
mg i
force and the gravitational force in the slug flow regime, but by the grav- Δt sp Δt V pipe i¼1
" !#
sp
itational force, particle-wall friction force and gas-wall friction force in
1 X 1
N p;X section
 
the stratified and dispersed flow regimes. In the dilute-phase inclined  f drag;i þ f pgf ;i 
pneumatic conveying, the increase of pressure drop at relatively low in- Δt sp Δt V pipe i¼1
sp
¼ " !#
clination angles is attributed to the increase of gravitational force, while X X
N p; section
1 1
the decrease of pressure drop at relatively high inclination angles is due mgi
Δt sp Δt V pipe i¼1
to the decrease of friction force. sp

It should be pointed out that the pressure drop−gas velocity ð10Þ

phase diagram focuses on pressure drop that is related to the gas
phase only. It indirectly reflects how a gas-solid flow behaves in pipe-
lines. Consequently, such a diagram may not correspond well to a vi-
sualized particle flow pattern and has limitations in its application. To
overcome this problem, attempts have been made to establish a new
phase diagram in terms of the forces governing the particle flow
based on CFD-DEM simulations results. For such a purpose, the forces
on a particle or a particle assembly are classified into three categories.
These are the particle-fluid force which is the sum of the fluid drag
and pressure gradient forces, the particle-particle and particle-wall
forces (the two are grouped together because they behave similarly
and are of the same nature), and the gravity force. This classification
allows producing two dimensionless numbers, П 1 and П 2 [50,54],
given by:
A new phase diagram can hence be established in terms of the two
dimensionless number, as given in Fig. 14. It is shown that in the new
diagram, there are two regions: the particle-particle or particle-wall in-
teraction is almost constant at the smallest value in the left region, but
increases with the increase of П 2 in the right region, and the boundary
of the two regions is located at the minimum of П 1. The curves for hor-
izontal and vertical pneumatic conveying are different in position due to
the different effects of gravity force on flow regimes, which are not
taken into account by the dimensionless numbers used [50,54]. Further
efforts are needed to consider the effect of pipe inclination angle, so that
a more general phase diagram in terms of forces can be established.
Compared to the pressure drop−gas velocity phase diagrams, the
new phase diagram not only has a profile independent of the conditions

Fig. 13. Average volumetric fluid drag force, pressure gradient force and gas-wall friction force, gravitational force and particle-wall friction force as a function of gas velocity in 45° inclined
pneumatic conveying [55].
 S. Kuang et al. / Powder Technology 365 (2020) 186–207 197

Fig. 14. Pressure drop-gas velocity phase diagram (left) and new phase diagram in terms of the forces on particles at a particle scale (right) for horizontal pneumatic conveying [54].

considered, but also can better represent the underlying fundamentals.
For example, it highlights an important characteristic of a dilute-phase
flow that the particle-particle and particle-wall forces are small, if not
negligible, compared to those in a dense-phase flow. Here, the so-called
dilute-phase flow is a flow where П 1 almost keeps constant and small,
corresponding to the dispersed flow regime and the stratified flow re-
gime, and the dense-phase flow is a flow where П2 increases with the
increase of П1, corresponding to the slug flow regime and the transition
flow regime. The concepts are similar to those used in the pressure drop
−gas velocity phase diagram. However, physically, they are different. In
the old phase diagram, the division between the dilute- and dense-
phase flows is based on the decrease and increase of pressure drop. It
is case-dependent. This problem can be overcome in the new phase di-
agram as noted above. This is because the new diagram is based directly
on the forces governing the behaviors of particles, and hence is funda-
mentally more reliable. On the other hand, visual observations give
more flow regimes. Some of these flow regimes (stratified flow and
transition flow) cannot be recognized in the new phase diagram.
While the need for their full reorganization is questionable, it is cer-
tainly useful to better understand the transition flow regime that can
be identified in the pressure drop-gas velocity diagram. This is an inter-
esting topic needed to be investigated further.
3.2.3. Energy dissipation
Energy efficiency is of great importance to pneumatic conveying.
Taking advantage of the CFD-DEM approach, the information about en-
ergy can be obtained macroscopically and microscopically. In this flow
system, the energy loss of gas flow is mainly used to replenish the en-
ergy dissipations as a result of the interactions between particles,
particles and wall, particles and fluid, fluids, and fluid and wall, as well
as the energy for accelerating gas and solid phases and lifting particles
in the gravitational direction. Basic equations have been established to
determine energy dissipations as a result of different interactions
among particles, fluid, and wall based on CFD-DEM simulation outputs
[55], as given in Table 3. The validity of these equations has been further
confirmed by applying them to the fluidization process [129].
Fig. 15 shows how energy is dissipated through the interactions
among particle, wall and fluid. It is found that for steady-state pneu-
matic conveying, the energy dissipation macroscopically calculated
from pressure drop and gas flow rate mainly consists of the particle-
fluid energy dissipation, gravitational potential energy, collision energy
dissipation, and fluid-wall energy dissipation at a microscopic level. The
most significant energy dissipation is the collision energy dissipation in
the slug flow regime, the particle-fluid energy dissipation in the strati-
fied flow regime, and the fluid-wall viscous energy dissipation in the
dispersed flow regime. The collision energy dissipation is mainly con-
tributed by the particle-wall friction energy dissipation in all the flow
regimes considered. However, it is to some degree by the particle-parti-
cle friction energy dissipation, impact energy dissipation and rolling
friction dissipation in the slug flow regime. The gravitational potential
energy in the 45° inclined pneumatic conveying is nearly uniform
with the varying gas velocity.
3.3. Pipe wear
Wear represents one of the major concerns in bulk solids handling. It
increases equipment maintenance costs and environmental burden and
causes loss of productivity and a requirement to replace damaged

Table 3
Energy dissipations of solid and gas phases in the considered pipe section [55].

Energy Correlations

Gravitational potential energy (eg) PNp; section

i¼1 ðmi g  vi ΔtÞ
Change of transitional kinetic energy (ek,v) PNp; section
i¼1 ½ð12 mi jvi j2 ÞtþΔt −ð12 mi jvi 2 jÞt 
Change of rotational kinetic energy (ek,ω) PNp; section 1
i¼1 ½ð2 Ii jωi j2 ÞtþΔt −ð12 Ii jωi 2 jÞt 
Particle-fluid energy dissipation (ep-f) PNp; section
½ f drag;i  ðui −vi ÞΔt
Pi¼1
q
Elastic potential energy between particles or between particles and wall (ec,p-p/ec,p-w) ½f
l¼1 cn;ij
 vcn;ij Δt þ f ct;ij  vct;ij Δtl
Pq
Impact energy dissipation between particles or between particles and wall (ed,p-p/ed,p-w) ½f
l¼1 dn;ij
 vdn;ij Δt þ f dt;ij  vdt;ij Δtl
Pq
Friction energy dissipation between particles or between particles and wall (es,p-p/es,p-w) ½ f  vt;ij Δtl
l¼1 s;ij
Pq
Rolling energy dissipation between particles or between particles and wall (er,p-p/er,p-w) l¼1
½Tr;ij  ωij Δtl
Viscous energy dissipation between fluids or between fluid and wall (ef-f/ef-w) PNcel
i¼1 ðτ : ∇uÞΔV i Δt
Collision energy (ecollision) ec, +ed +es, +er,

where n and Ncel are, respectively, the number of particles and the number of CFD cells in the considered pipe section, and q is the number of contact points between particles or particle
and wall.
198 S. Kuang et al. / Powder Technology 365 (2020) 186–207
Fig. 15. Energy dissipations as a function of gas velocity in 45° inclined pneumatic conveying: major energy dissipations (left), and energy dissipations due to particle-particle and particle-
wall collisions (right) [55].
components. For example, in the oil and gas industry, the erosion of
down-hole equipment and surface facilities costs billions of dollars for
annual maintenance [130]. In essence, the pipeline wear is caused by
the surface abrasion (erosion) resulting from the dynamic action of
moving particles on the pipe wall. The abrasion includes (i) deformation
(impact) wear caused by the normal impact of solid particles, and (ii)
cutting (sliding) wear caused by the oblique impact of solid particles.
The quantification of particle-wall interaction will provide useful infor-
mation to estimate the surface abrasion. With this realization, the so-
called time-averaged collision intensity (TACI) has been used to reflect
pipeline wear qualitatively in CFD-DEM studies of pneumatic bend
[48,58]:
Pkw  !
1 X 
i¼1 f cn;ij þ f dn;ij þ f ct;ij þ f dt;ij
TACI ¼
Δt sp Δt Asp
sp
where Asp is the surface area of a sample wall, Δtsp is the simulation time
or sampling time period.
Fig. 16 shows the distribution of TACI between particles and wall.
The result reveals that the most intensive particle-wall interaction posi-
tion depends on not only the particle velocity but also on the particle
concentration. The concept of TACI can be used to locate the pipe wall
position that experiences the highest particle-wall interaction and can
be first worn out, as reflected from the comparison between the numer-
ical result and experimental observation. This concept has been con-
firmed by comparing the TACI distribution with the wear rate
obtained in the CFD-DEM study of dense medium cyclone [131].
To be quantitative in predicting pipeline wear, a numerical model for
gas-solid flow is often combined with a wear equation. Among the most
popular wear equations is the one proposed by Finnie [132,133]:
8
separators [140,141].
>
> mv2impact 6
>
< sin2α− sin2 α α ≤18:5°
8p κ 3.4. Particle attrition and electrification
Q erosion ¼ ð12Þ
> 2
> mvimpact
>
: 2
cos α α ≥18:5 °
24p
where Qerosion is the volume of material removed by the impact of a sin-
gle particle; p is the plastic flow pressure of erosion surface, m is the par-
ticle mass, vimpact is the particle impact velocity, κ is the ratio of vertical
to horizontal force components on the particle, and α is the particle im-
pact angle. v and α in Eq. (9) are provided by numerical simulations.
In the past, extensive numerical studies have been conducted to
study pipe wear in pneumatic conveying, where other wear equations
are also used in addition to Finnie's equation. The majority of such stud-
ies have been based on the CFD-LPT-Wear model, focusing on bends
which are most likely to suffer erosion, as reviewed by Parsi et al.
[134]. Because of the nature of LPT model, the CFD-LPT-Wear model
cannot explicitly consider the effect of particle-particle collision on
pipe erosion. To overcome this problem, a few efforts have been made
recently to apply a CFD-DEM-wear model to study pneumatic pipe
wear [64,70,75,76]. In particular, in the work of Xu et al. [64], CFD-
DEM the model is combined with the recent wear equation developed
specifically for the DEM modelling [135]:
Eshear
Q erosion ¼ ð13Þ
4:0p
ð11Þ where Eshear = − ∫tt01fct, ij ∙ vt, ij dt, when fct, ij ∙ vt, ij b 0 and fct, ij ∙ vi b 0.
Using this model, Xu et al. [64] found that the particle concentration
plays a dominant role on the erosion of the elbow, as shown in Fig. 17.
With an increase in the particle flow rate, the wear rate on the inner
wall of the extrados apparently increases in the range of 70–90°. It is
of interest to note that such a model can also be applied to other partic-
ulate systems such as mill and fluidized bed [136–138].
Clearly, wear equation plays a critical role for numerical models to
predict wear rate. There are about 182 wear equations according to
the review of Meng and Ludema in 1995 [139], who however suggested
none of them can be taken for general and practical use. Finite element
method or the like may be used to establish a more general wear equa-
tion with the support of theoretical and experimental analysis, as done
elsewhere for developing theoretically sound particle-particle/wall con-
tact models [99,100]. Also, to facilitate industrial applications, a predic-
tion wear model should be formulated based on systematic numerical
studies. This framework has been proved feasible in the study of cyclone
3.4.1. Electrostatics
Electrostatics has had a dramatic and sometimes catastrophic effect
on industry over the years, especially in the flour and grain handling in-
dustries [1]. Of all the solids processing operations, pneumatic convey-
ing is the one that produces the largest amount of electrostatics in the
system. The interactions of the particles with themselves and the
system's pipeline are many, and each has the ability to transfer an elec-
trostatic charge, depending on the material type, shape, humidity of the
operation and the character of the pipeline. Because of electrostatics,
 S. Kuang et al. / Powder Technology 365 (2020) 186–207 199

Fig. 16. Spatial distribution of particle-wall TACI [48]: (a) front view; (b) on the central plane; (c) at the outer wall of the bend; (d) an experimental bend of solid clear acrylic, worn by the
passage of alumina particles.

particle patterns show some unique features and further affect pneu- where Qe is the constant charge assumed to be carried by all particles, ε0
matic conveying performance. As such, understanding electrostatics is the permittivity of free space, rij is the distance between particles i and
has attracted increasing experimental research interests [1,142]. Mean- ^ is the unit normal vector in the direction of the line joining the
j, and n
while, a few CFD-DEM studies have also been conducted to describe two particle centers.
electrostatics and/or their effects on solids transportation [37– The average electric field strength near a wall of the pneumatic con-
39,42,43,67,77]. In the CFD-DEM modelling, the electrostatic force act- veying pipe is estimated by assuming the pipe to be a flat plat infinitely
ing on particle i is given as following expression [37,42]: long in the axial direction in the works of Lim et al. [37,42].The electric
field strength is given as:
f e;i ¼ f ep;i þ f ew;i ð14Þ
Z Z π
1 dqe λ 2 λ
Ee ¼ ∙ ¼ cosθdθ ¼ ð16Þ
where fep,i is the electrostatic forces due to other charged particles act- 4πε0 r 2 2πε0 d 0 2πε 0 d
ing on particle i and few,i is the electrostatic forces due to the pipe
walls acting on particle i. The electrostatic force arising from charges where Ee is the electric field strength, qe is the equilibrium charge on the
carried by other particles may be calculated by assuming each particle pipe wall, ε0 is the permittivity of free space, r is the average distance of
to be a constant point charge. particles to the pipe wall, and λ is the linear charge density along the
pipe wall. The electrostatic force on a charged particle i located near
X
N
Q 2e the pipe wall may then be calculated as:
f ep;i ¼ ^
n ð15Þ
4πε0 r 2ij
j¼1 f ew;i ¼ Ee ∙Q e ð17Þ
j≠i
200 S. Kuang et al. / Powder Technology 365 (2020) 186–207

Fig. 17. Effect of solid flow rate on bend wear rate [64].

In a similar framework, Korevaar et al. [43] have modelled the case CFD-DEM model and particle attrition model, as demonstrated by dif-
where the wall is grounded and conducting by the method of images. ferent investigators [30,35,52,69,73].
In such a method, an image charge is placed at the same distance but Generally, three different types of breakage models have been inte-
at the opposite side of a conducting planar wall and with a charge of grated into the CFD-DEM approach to study pneumatic conveying re-
same magnitude but opposite sign. Different from the use of constant lated to particle attrition. One is based on an empirical model. For
Q in Eq. (15) or (17), Grosshans et al. [67] have further developed the example, in the work of Han et al. [30], Ghadiri's attrition model has
electrostatic force model to consider the charge exchange ΔQ when a been used, which accounts for two different mechanisms of attrition,
particle collides with another particle or pipe wall. namely chipping and fragmentation, with the former occurring at im-
Fig. 18 shows the representative results obtained in the CFD-DEM pact velocities between 4 m/s and 13 m/s and the latter at N13 m/s.
study of electrostatics by Lim et al. [37]. In the simulations, the ratio of The respective equations for calculating particle sizes produced by the
the electrostatic force arising from the charged pipe wall to the gravita- two mechanisms are as follows:
tional force exerted on each particle (Λ) is varied to consider the scenario For chipping:
that the upper walls experience a larger amount of collisions from parti-
4 !1=3
cles and thus are charged to a larger extent as the pipe inclination angle ρp Hv2impact ds0
increases. It can be seen that as Λ increases, the particle flow patterns ds ¼ αf 4 m=s b vimpact b 13 m=s ð18Þ
K 2c
change significantly. In fact, this CFD-DEM model can successfully pre-
dict the experimentally observed formation of a ring structure along For fragmentation:
the pipe walls [37,42]. Nevertheless, further efforts are still needed to de-
velop models considering the effect of particle shape and study flow re-
gimes related to electrostatics, especially those unique ones reported in
literature [1]. Additionally, it should be pointed out that bulk solids han-
dling in terms of pneumatic transport may have moisture, likely having a
big impact on conveying performance [143]. In the framework of CFD-
DEM approach, the modelling of the moisture is in principle similar to
the modelling of electrostatics. However, this modelling effort is lacking
for pneumatic conveying, though it can be found for other particulate
systems [144,145]. Research in this area is therefore called.
3.4.2. Particle attrition
Attrition and breakage of solid particles are unavoidable phenomena
in pneumatic conveying, especially in a dilute-phase mode. It affects
both the conveying characteristics and the quality of particulate mate-
rials. Many different particle attrition test apparatus exist, there remain
a significant number of gaps in the research to be explored, especially
with respect to linking the method of characterization to the process
under consideration [146]. A successful particle attrition modelling ap-
proach will need to capture both the material and process functions of
the phenomenon in order to achieve meaningful results. An early effort
in this respect is the work of Frye and Peukert [73], who used CFD-LPT
model to determine impact conditions, and on this base, employed ex-
periments to obtain material function to evaluate particle attrition. At
present, one of the most promising approaches is those combining the
 
  ds
3−ζ 0 ln ¼ lnλ0
ds0 !
0
ρp Hv2impact ds0
þ ζ ln vimpact N 13 m=s ð19Þ
K 2c
where αf is a proportionality factor, ds is the diameter of the smaller part
of a particle after impact, ds0 is the diameter of the mother particle be-
fore impact, H is the material hardness, vimpact is the particle impact ve-
locity, λ' and ζ’′ are the parameters that are determined experimentally,
and Kc is the critical stress intensity factor.
The second approach uses the bonded particle model in the DEM
modelling, which has been originally developed by Potyondy and
Cundall [147]. The concept of this modelling is based on bonding or glu-
ing a packed distribution of spheres together forming a breakable body
via the introduction of an elastic beam between two particles, as illus-
trated in Fig. 19. The particles bonded together can either be of mono
size or have a size distribution. The breakage is determined by comput-
ing the normal and shear stresses and checking if the resulting values
exceed the pre-set critical ones. Such a model is applicable to the break-
age of brittle rocks, such as coal, granite, sandstone, and concrete. It has
been applied to study the particle breakage in pneumatic conveying of
coarse coal by Zhou et al. [69,73]. Fig. 19 shows the typical numerical re-
sults, which demonstrate how a cluster is broken when it passes a bend.
The third approach integrates a comminution model with the CFD-
DEM approach. Fig. 20a illustrates the framework of the CFD-DEM-
 S. Kuang et al. / Powder Technology 365 (2020) 186–207 201

Fig. 18. Pneumatic conveying through a pipe inclined at 45° to the horizontal with an inlet gas velocity of 3 m·s−1, α = 0.16 (1000 particles) and: (a) Λ = 0.0, (b) Λ = 0.5, (c) Λ = 1.0 and
(d) Λ = 2.0 [37]. The range for the scale is −3 m·s−1 (black) to 3 m·s−1 (white). The insets to (c) and (d) show the enlarged image of the respective sections enclosed in dashed boxes.

Fig. 19. Predicted breakage of particle cluster within a bend by the so called parallel bond model [73].
202 S. Kuang et al. / Powder Technology 365 (2020) 186–207

Fig. 20. (a) Comminution model framework for DEM–CFD simulations [148], and (b) thus predicted averaged particles breakage velocities as function of the conveying air velocities and
number of cycles [52].

comminution model proposed by Kalman et al. [148]. According to this
approach, each particle in the simulation has its own strength given by
the strength distribution function. As the simulation progresses, the
particles undergo impact events. For each impact event/collision, the se-
lection function is called to decide whether the particle breaks or not. If
the particle breaks, the breakage function is called in order to generate
the new fragments at strength according to the strength distribution.
If the particle is strong enough to survive, the impact the particle will
endure fatigue, which will dictate its strength after the impact. It should
be pointed out that this attrition model describes the compression
strength of particles, where the selection function is derived from im-
pact experiments. Therefore, the implementation of this model requires
the equivalence function between impact and compression strength
[149]. Brosh et al. [52] have extended such a model to study particle at-
trition in dilute-phase pneumatic conveying. The parametric study on
the effect of the impact velocity and the number of conveying cycles
on the particle size distribution has been conducted. It is found that a
higher conveying air velocity leads to higher impact velocities and
therefore to larger breakage ratios, as shown Fig. 20b. Moreover, the av-
erage particle breakage velocity increases with the number of cycles due
to the reduction in the particle size. Using this CFD-DEM based
particle attrition model, a TFM-based one can be developed to facilitate
industrial applications [150,151]. Also, it should be pointed out that
at present how to explicitly model the effect of a non-spherical
particles in an attrition model is not clear. Future efforts are needed
for this topic.
4. Large-scale simulation
CFD-DEM studies of pneumatic conveying can be conducted in the
two ways. One considers a long pipeline consisting of start-up and
established flow sections (see Fig. 21). It requires high computational
efforts, and at this stage of development, cannot be applied to industrial
pneumatic conveying where the pipeline length is quite long, involving
either coarse or fine particles. Another uses periodic boundary condition
(PBC) in a simulation to consider a short pipe that represents the
established-flow section and can be scaled up straightforwardly accord-
ing to the length. The latter is computationally much more efficient.
However, to use a CFD-DEM model with PBC to solve practical pneu-
matic conveying, it is necessary to establish methods to exclude the

Fig. 21. Application of periodic boundary condition (PBC) to CFD-DEM simulation [56].
 S. Kuang et al. / Powder Technology 365 (2020) 186–207 203

Table 4
Correlations developed for the use of periodic boundary condition [56].

Relation on between particle number and solid flow rate Np; section ¼ kw
wp ρ f

V pipe
f ρp þwp ρ f Vp
w 0:17 w 0:464
k¼ 0:684ð 2:5 p 0:5 Þ ð 2:5 f 0:5 Þ
dp ρp g dp ρp g
Start-up section length d 0:167 ρ 0:167 w 0:164 w 0:83
L
D ¼ 3:222ð Dp Þ ðρp Þ ðρ g0:5pD2:5 Þ ðρ g0:5fD2:5 Þ
f f f

Characteristic length of slug flow Lc ¼ 145:457ðdgÞ

0:5
ð
wp
Þ
−0:643 w
ð 2:5 f 0:5 Þ
0:264
ð0:605U g −0:236Þ
2:5
dp ρp g 0:5 d p ρp g
Average particle velocity vp ¼ ½0:684ð
wp 0:17 w
Þ ð 2:5 f 0:5 Þ
0:464
u
w
þ ρ pA
2:5
dp ρp g 0:5 dp ρp g p

Fig. 22. Comparison of flow patterns from the simulations of non-PBC pipe (left) and PBC pipe (right) at: (a) Ug = 3 m/s, (b) Ug = 15 m/s, and (c) Ug = 23 m/s [56]. The length of non-PBC
pipe is 14 m for stratified and dispersed flow regimes and 6 m for slug flow regime, which are respectively shorten to 1 m and 1.2 m for the PBC pipe.

start-up section from the pipeline simulated, determine length of PBC
pipe, and control the solid flow rate simulated as constant as done in a
common operation in practice. To tackle these issues, semi-theoretical
correlations, as listed in Table 4, have been formulated based on CFD-
DEM simulation data [56]. Moreover, to be accurate in this PBC applica-
tion, a numerical iterative method has also been proposed to control the
solid flow rate simulated to achieve a given value.
Fig. 22 compares the predicted typical particle flow patterns from
PBC and non-PBC pipes. In this figure, the results from the non-PBC
pipe were taken from the established flow section. It can be seen from
Fig. 22 that the particle flow patterns predicted using two different
boundary conditions satisfactorily match each other for the three typical
flow regimes: the slug, stratified and dispersed flow regimes. This good
agreement can also be observed in the axial variation of the average
pressure drop, as shown in Fig. 23. This result suggests that the PBC
treatment is reliable to reproduce the flows inside a long pipe by a
much shorter pipe. This CFD-DEM model with PBC has been applied to
different studies to pneumatic conveying [55,56,59]. In principle, the
model is capable of predicting different flow regimes and their transi-
tion as well as pipe wear. Moreover, such a model is not suitable for par-
ticle attrition and electrification, where particle-particle interactions
play essential roles and have to be described precisely to generate real-
istic results. As such, it would be preferred to develop various scale-up
and scale-down theories to speed up the simulations of practical pneu-
matic conveying.
Additionally, with the rapid improvement of computer hardware
performance, GPU (Graphic Process Unit) has emerged to not only ren-
der 3D scene but also support more and more complicated computa-
tions. In the DEM simulation of granular flow, the overall speed-up
ratio of GPU parallel codes to its serial CPU counterparts is reported to
3.0
Ug (m/s) PBC pipe Non-PBC pipe
2.5 23
15
3
Pressure drop (kPa)

CFD-DEM with PBC can be used to simulate a realistic pneumatic con- 2.0
veying system by diving pipelines into start-up sections and developed
sections and conducting the simulations sequentially. Furthers studies 1.5
are needed to test this concept.
Another potential method to speed up CFD-DEM simulations for
large-scale industrial applications is to use a “scaled” method in simula- 1.0
tions. The so called coarse-grain models are this kind of method, where
a particle assembly is represented by a number of original particles and
0.5
a scaling law is derived according to the force balance between a coarse-
grain or parcel particle and corresponding original particles [29,152]. By
use of such a model, slug flow pattern has been reasonably reproduced 0.0
0.0 0.2 0.4 0.6 0.8 1.0 1.2
[29]. It should be pointed out that a coarse-grain model needs to esti-
Axial position (m)
mate the physical properties of a coarse-grain particle and there is no
theoretical base to do so. Furthermore, the number of original particles
Fig. 23. Comparison of average (a) pressure drop and (b) particle velocity obtained from
considered by a parcel particle has to be determined according to spe-
the simulation of non-PBC and PBC [56]. For the non-PBC pipes, the results are
cific applications to balance prediction accuracy and computational effi- measured within the pipe section between x = 12 and 13 m for the stratified and
ciency [153]. Because of these problems, it is not clear if a coarse-grain dispersed flow regimes, and x = 4.6 and 5.8 m for the slug flow regime.
204 S. Kuang et al. / Powder Technology 365 (2020) 186–207
be from several times to tens of times. More importantly, by the use of
multiple GPUs, simulating systems with millions of particles is becom-
ing realistic for engineering applications [154,155]. However, this tech-
nology has not been applied to the studies of pneumatic conveying and
deserves specific efforts in the future.
5. Concluding remarks
In the process of pneumatic conveying, various flow regimes and
their transition may take place depending on conveying conditions,
which affect, to a large extent, system performance with regards to en-
ergy consumption, production degradation, equipment erosion, and op-
eration reliability. This raises the needs of developing a better
understanding of these complicated flow behaviors and process perfor-
mance toward the achievement of more reliable, general, and cost-ef-
fective applications. Numerical simulation and modelling play an
increasingly important role in this area. With the rapid development
of computer technology, the CFD-DEM approach is thus far the most
promising in generally tackling various complicated phenomena. In re-
cent years, this approach has been rapidly developed and applied to
study pneumatic conveying, covering from flow behaviors to process
performance, which are summarised as follows:
(1) Various CFD-DEM models have been developed and applied to
study pneumatic conveying in the aspects of flow, pipe wear,
particle attrition, and electrostatics. This approach has been
proved valid to reproduce complicated flow regimes and their
transition over a wide range of conditions. On the contrary, its
applications to pipe wear, particle attrition, and electrostatics
are less established. The effectiveness of such CFD-DEM models
is, to a large extent, dependent on the theoretical and semi-the-
oretical correlations used to describe the process phenomena
considered. This necessitates the fundamental research to ex-
plore the underlying mechanisms and develop more general cor-
relations. In this direction, sub-particle simulations based on FEM
or the like, in addition to particle-scale CFD-DEM simulations,
will play a useful role. These studies of this kind will represent
an important research area in the future. Also, future efforts are
needed to model particle-turbulence interaction. Besides, the ef-
fect of particle shape has been increasingly considered in the
CFD-DEM studies of pneumatic conveying but focused on flow
and pipe wear. Considering that at least 70% of raw materials
consist of non-spherical particles in modern industries, further
efforts are needed for CFD-DEM model development and appli-
cation related to pneumatic conveying of irregular particles.
(2) Microdynamic analysis, based on the trajectories and velocities
of, and forces acting particles, has played an important role in
depicting the fundamentals governing the complicated gas and
particle behaviors in different flow regimes. The new phase dia-
grams established in terms of particle-particle and particle-fluid
forces show a great potential in describing flow regimes and
flow transition generally and better representing underlying fun-
damentals, although further efforts are needed in order to cover
different conditions in the same diagrams. Furthermore, based
on the particle-scale results generated by CFD-DEM simulations,
macroscopic energy loss can be linked with microscopic energy
dissipations among particle, wall, and fluid. Also, with the sup-
port of the averaging theory, solid stresses can be analysed.
How to generalise this analysis and apply it to solve different
problems associated with the development of predictive models
represents a challenge in the future.
(3) CFD-DEM simulations are computationally very demanding. The
use of periodic boundary conditions makes it possible to apply
these simulations to industrial applications. Because of wide
ranges of transport distances, particle sizes, and particle shapes
encountered in pneumatic conveying, besides the use of periodic
boundary conditions, it is necessary in the future to not only em-
ploy advanced GPU computing technology but also develop gen-
eral scaling theories to reduce the computational effort.
Acknowledgments
The authors are grateful to the Australian Research Council (ARC)
(IH140100035 and LP160100819) for the financial support of this work.
References
[1] G.E. Klinzing, A review of pneumatic conveying status, advances and projections,
Powder Technol. 333 (2018) 78–90.
[2] O. Molerus, Overview: pneumatic transport of solids, Powder Technol. 88 (1996)
309–321.
[3] R. Pan, Material properties and flow modes in pneumatic conveying, Powder
Technol. 104 (1999) 157–163.
[4] P.W. Wypych, J.L. Yi, Minimum transport boundary for horizontal dense-phase
pneumatic conveying of granular materials, Powder Technol. 129 (2003) 111–121.
[5] G.E. Klinzing, F. Rizk, R. Marcus, L.S. Leung, Pneuamtic Conveying of Solids: A The-
oretical and Pratical Approach, Springer, New York, 2010.
[6] P. Patro, S.K. Dash, Two-fluid modeling of turbulent particle–gas suspensions in
vertical pipes, Powder Technol. 264 (2014) 320–331.
[7] N. Behera, V.K. Agarwal, M.G. Jones, K.C. Williams, Modeling and analysis of dilute
phase pneumatic conveying of fine particles, Powder Technol. 249 (2013) 196–204.
[8] A. Rao, J.S. Curtis, B.C. Hancock, C. Wassgren, Numerical simulation of dilute turbu-
lent gas-particle flow with turbulence modulation, AICHE J. 58 (2012) 1381–1396.
[9] W. Pu, C. Zhao, Y. Xiong, C. Liang, X. Chen, P. Lu, C. Fan, Numerical simulation on
dense phase pneumatic conveying of pulverized coal in horizontal pipe at high
pressure, Chem. Eng. Sci. 65 (2010) 2500–2512.
[10] K. Hadinoto, J.S. Curtis, Reynolds number dependence of gas-phase turbulence in
particle-laden flows: effects of particle inertia and particle loading, Powder
Technol. 195 (2009) 119–127.
[11] D. Mcglinchey, A. Cowell, E.A. Knight, J.R. Pugh, A. Mason, B. Foster, Bend pressure
drop predictions using the Euler-Euler model in dense phase pneumatic convey-
ing, Part. Sci. Technol. 25 (2007) 495–506.
[12] K.H. Henthorn, K. Park, J.S. Curtis, Measurement and prediction of pressure drop in
pneumatic conveying: effect of particle characteristics, mass loading, and Reynolds
number, Ind. Eng. Chem. Res. 44 (2005) 5090–5098.
[13] K.W. Zhu, C.K. Wong, S.M. Rao, C.H. Wang, Pneumatic conveying of granular solids
in horizontal and inclined pipes, AICHE J. 50 (2004) 1729–1745.
[14] L.Y. Lee, T.Y. Quek, R.S. Deng, M.B. Ray, C.H. Wang, Pneumatic transport of granular
materials through a 90 degrees bend, Chem. Eng. Sci. 59 (2004) 4637–4651.
[15] A. Levy, Two-fluid approach for plug flow simulations in horizontal pneumatic
conveying, Powder Technol. 112 (2000) 263–272.
[16] B. Kuan, W. Yang, M. Schwarz, Dilute gas–solid two-phase flows in a curved 90
duct bend: CFD simulation with experimental validation, Chem. Eng. Sci. 62
(2007) 2068–2088.
[17] A. Yilmaz, E.K. Levy, Formation and dispersion of ropes in pneumatic conveying,
Powder Technol. 114 (2001) 168–185.
[18] E.V.P.J. Manjula, W.K.H. Ariyaratne, C. Ratnayake, M.C. Melaaen, A review of CFD
modelling studies on pneumatic conveying and challenges in modelling offshore
drill cuttings transport, Powder Technol. 305 (2017) 782–793.
[19] H.P. Zhu, Z.Y. Zhou, R.Y. Yang, A.B. Yu, Discrete particle simulation of particulate
systems: theoretical developments, Chem. Eng. Sci. 62 (2007) 3378–3396.
[20] H.P. Zhu, Z.Y. Zhou, R.Y. Yang, A.B. Yu, Discrete particle simulation of particulate
systems: a review of major applications and findings, Chem. Eng. Sci. 63 (2008)
5728–5770.
[21] Y. Tsuji, T. Tanaka, T. Ishida, Lagrangian numerical-simulation of plug flow of cohe-
sionless particles in a horizontal pipe, Powder Technol. 71 (1992) 239–250.
[22] Y. Tsuji, T. Kawaguchi, T. Tanaka, Discrete particle simulation of two-dimensional
fluidized bed, Powder Technol. 77 (1993) 79–87.
[23] B.H. Xu, A.B. Yu, Numerical simulation of the gas-solid flow in a fluidized bed by
combining discrete particle method with computational fluid dynamics, Chem.
Eng. Sci. 52 (1997) 2785–2809.
[24] B.H. Xu, A.B. Yu, Comments on the paper numerical simulation of the gas–solid
flow in a fluidized bed by combining discrete particle method with computational
fluid dynamics, Chem. Eng. Sci. 53 (1998) 2645–2646.
[25] F.Y. Fraige, P.A. Langston, Horizontal pneumatic conveying: a 3D distinct element
model, Granul. Matter 8 (2006) 67–80.
[26] S. Mcnamara, M. Strauβ, F. Zeller, H.J. Herrmann, Simulations of dense-phase pneu-
matic conveying, Behav. Granular Media (2006) 17–23.
[27] M. Strauβ, S. Mcnamara, H.J. Herrmann, G. Niederreiter, K. Sommer, Plug conveying
in a vertical tube, Powder Technol. 162 (2006) 16–26.
[28] M. Strauβ, S. Mcnamara, H.J. Herrmann, , Plug conveying in a horizontal tube
Granul. Matter 9 (2007) 35–48.
[29] M. Sakai, S. Koshizuka, Large-scale discrete element modeling in pneumatic con-
veying, Chem. Eng. Sci. 64 (2009) 533–539.
[30] T. Han, A. Levy, H. Kalman, DEM simulation for attrition of salt during dilute-phase
pneumatic conveying, Powder Technol. 129 (2003) 92–100.
[31] J. Li, S.S. Pandiella, C. Webb, T. Dyakowski, M.G. Jones, Analysis of gas-solids feeding
and slug formation in low-velocity pneumatic conveying, Part. Sci. Technol. 21
(2003) 57–73.
 S. Kuang et al. / Powder Technology 365 (2020) 186–207
[32] J.T. Li, D.J. Mason, A.S. Mujumdar, A numerical study of heat transfer mechanisms
in gas-solids flows through pipes using a coupled CFD and DEM model, Dry.
Technol. 21 (2003) 1839–1866.
[33] J. Li, C. Webb, S.S. Pandiella, G.M. Campbell, T. Dyakowski, A. Cowell, D. Mcglinchey,
Solids deposition in low-velocity slug flow pneumatic conveying, Chem. Eng. Pro-
cess. 44 (2005) 167–173.
[34] D. Mcglinchey, A. Cowell, J.R. Pugh, E.A. Knight, J. Xiang, J. Li, Axial and radial pres-
sure drops of dense phase plugs, Part. Sci. Technol. 23 (2005) 215–227.
[35] E.W.C. Lim, C.H. Wang, Diffusion modeling of bulk granular attrition, Ind. Eng.
Chem. Res. 45 (2006) 2077–2083.
[36] E.W.C. Lim, C.H. Wang, A.B. Yu, Discrete element simulation for pneumatic convey-
ing of granular material, AICHE J. 52 (2006) 496–509.
[37] E.W.C. Lim, Y. Zhang, C.H. Wang, Effects of an electrostatic field in pneumatic con-
veying of granular materials through inclined and vertical pipes, Chem. Eng. Sci. 61
(2006) 7889–7908.
[38] S. Watano, Mechanism and control of electrification in pneumatic conveying of
powders, Chem. Eng. Sci. 61 (2006) 2271–2278.
[39] Y. Zhang, E.W.C. Lim, C.H. Wang, Pneumatic transport of granular materials in an
inclined conveying pipe: comparison of computational fluid dynamics-discrete el-
ement method (CFD-DEM), electrical capacitance tomography (ECT), and particle
image Velocimetry (PIV) results, Ind. Eng. Chem. Res. 46 (2007) 6066–6083.
[40] Y. Zhang, E.P. Reuterfors, B.S. Mclaury, S.A. Shirazi, E.F. Rybicki, Comparison of com-
puted and measured particle velocities and erosion in water and air flows, Wear
263 (2007) 330–338.
[41] J. Xiang, D. Mcglinchey, J.-P. Latham, An investigation of segregation and mixing in
dense phase pneumatic conveying, Granul. Matter 12 (2010) 345–355.
[42] E.W.C. Lim, J. Yao, Y. Zhao, Pneumatic transport of granular materials with electro-
static effects, AICHE J. 58 (2011) 1040–1059.
[43] M.W. Korevaar, J.T. Padding, M.A. Van Der Hoef, J.a.M. Kuipers, Integrated DEM–
CFD modeling of the contact charging of pneumatically conveyed powders, Pow-
der Technol. 258 (2014) 144–156.
[44] S.B. Kuang, K.W. Chu, A.B. Yu, Z.S. Zou, Y.Q. Feng, Computational investigation of
horizontal slug flow in pneumatic conveying, Ind. Eng. Chem. Res. 47 (2008)
470–480.
[45] R.E. Stratton, C.M. Wensrich, Horizontal slug flow pneumatic conveying: numerical
simulation and analysis of a thin slice approximation, Powder Technol. 214 (2011)
477–490.
[46] F. Zhou, S. Hu, Y. Liu, C. Liu, T. Xia, CFD–DEM simulation of the pneumatic convey-
ing of fine particles through a horizontal slit, Particuology 16 (2014) 196–205.
[47] Y.Q. Xiong, M.Y. Zhang, Z.L. Yuan, Three-dimensional numerical simulation method
for gas-solid injector, Powder Technol. 160 (2005) 180–189.
[48] K.W. Chu, A.B. Yu, Numerical simulation of the gas-solid flow in three-dimensional
pneumatic conveying bends, Ind. Eng. Chem. Res. 47 (2008) 7058–7071.
[49] S.B. Kuang, A.B. Yu, Z.S. Zou, A new point-locating algorithm under three-dimen-
sional hybrid meshes, Int. J. Multiphase Flow 34 (2008) 1023–1030.
[50] S.B. Kuang, A.B. Yu, Z.S. Zou, Computational study of flow regimes in vertical pneu-
matic conveying, Ind. Eng. Chem. Res. 48 (2009) 6846–6858.
[51] Z.Y. Zhou, S.B. Kuang, K.W. Chu, A.B. Yu, Discrete particle simulation of particle-
fluid flow: model formulations and their applicability, J. Fluid Mech. 661 (2010)
482–510.
[52] T. Brosh, H. Kalman, A. Levy, DEM simulation of particle attrition in dilute-phase
pneumatic conveying, Granul. Matter 13 (2011) 175–181.
[53] J.E. Hilton, P.W. Cleary, The influence of particle shape on flow modes in pneumatic
conveying, Chem. Eng. Sci. 66 (2011) 231–240.
[54] S.B. Kuang, A.B. Yu, Micromechanic modeling and analysis of the flow regimes in
horizontal pneumatic conveying, AICHE J. 57 (2011) 2708–2725.
[55] S.B. Kuang, R.P. Zou, R.H. Pan, A.B. Yu, Gas solid flow and energy dissipation in in-
clined pneumatic conveying, Ind. Eng. Chem. Res. 51 (2012) 14289–14302.
[56] S.B. Kuang, K. Li, R.P. Zou, R.H. Pan, A.B. Yu, Application of periodic boundary con-
dition to CFD-DEM simulation of gas-solid flow in pneumatic conveying, Chem.
Eng. Sci. 93 (2013) 214–228.
[57] M. Ebrahimi, M. Crapper, J.Y. Ooi, Experimental and simulation studies of dilute
horizontal pneumatic conveying, Part. Sci. Technol. 32 (2014) 206–213.
[58] H. Kruggel-Emden, T. Oschmann, Numerical study of rope formation and disper-
sion of non-spherical particles during pneumatic conveying in a pipe bend, Powder
Technol. 268 (2014) 219–236.
[59] K. Li, S.B. Kuang, R.H. Pan, A.B. Yu, Numerical study of horizontal pneumatic con-
veying: effect of material properties, Powder Technol. 251 (2014) 15–24.
[60] T. Oschmann, K. Vollmari, H. Kruggel-Emden, S. Wirtz, Numerical investigation of
the mixing of non-spherical particles in fluidized beds and during pneumatic con-
veying, Procedia Eng. 102 (2015) 976–985.
[61] B. Van Wachem, M. Zastawny, F. Zhao, G. Mallouppas, Modelling of gas–solid tur-
bulent channel flow with non-spherical particles with large stokes numbers, Int. J.
Multiphase Flow 68 (2015) 80–92.
[62] F. Zhao, W.K. George, B. Van Wachem, Four-way coupled simulations of small par-
ticles in turbulent channel flow: the effects of particle shape and stokes number,
Phys. Fluids 27 (2015) 37.
[63] M. Ebrahimi, M. Crapper, J.Y. Ooi, Numerical and experimental study of horizontal
pneumatic transportation of spherical and low-aspect-ratio cylindrical particles,
Powder Technol. 293 (2016) 48–59.
[64] L. Xu, Q. Zhang, J. Zheng, Y. Zhao, Numerical prediction of erosion in elbow based
on CFD-DEM simulation, Powder Technol. 302 (2016) 236–246.
[65] W. Chen, J.Y. Zhang, T. Donohue, K. Williams, R.X. Cheng, M. Jones, B. Zhou, Effect of
particle degradation on electrostatic sensor measurements and flow characteristics
in dilute pneumatic conveying, Particuology 33 (2017) 73–79.
205
[66] M. Ebrahimi, M. Crapper, CFD-DEM simulation of turbulence modulation in hori-
zontal pneumatic conveying, Particuology 31 (2017) 15–24.
[67] H. Grosshans, M.V. Papalexandris, Numerical study of the influence of the powder
and pipe properties on electrical charging during pneumatic conveying, Powder
Technol. 315 (2017) 227–235.
[68] Y. Lu, Z. Tong, D.H. Glass, W.J. Easson, M. Ye, Experimental and numerical study of
particle velocity distribution in the vertical pipe after a 90° elbow, Powder Technol.
314 (2017) 500–509.
[69] J.W. Zhou, Y. Liu, C.L. Du, S.Y. Liu, Effect of the particle shape and swirling intensity
on the breakage of lump coal particle in pneumatic conveying, Powder Technol.
317 (2017) 438–448.
[70] J.W. Zhou, Y. Liu, S.Y. Liu, C.L. Du, J.P. Li, Effects of particle shape and swirling inten-
sity on elbow erosion in dilute-phase pneumatic conveying, Wear 380-381 (2017)
66–77.
[71] S. Sun, Z. Yuan, Z. Peng, B. Moghtaderi, E. Doroodchi, Computational investigation
of particle flow characteristics in pressurised dense phase pneumatic conveying
systems, Powder Technol. 329 (2018) 241–251.
[72] H. Zhang, T.J. Li, Z.Y. Huang, S.B. Kuang, A.B. Yu, Investigation on vertical plug for-
mation of coarse particles in a non-mechanical feeder by CFD-DEM coupling
method, Powder Technol. 332 (2018) 79–89.
[73] J.W. Zhou, Y. Liu, C. Du, S. Liu, J. Li, Numerical study of coarse coal particle breakage
in pneumatic conveying, Particuology 38 (2018) 204–214.
[74] W. Zhong, A. Yu, X. Liu, Z. Tong, H. Zhang, DEM/CFD-DEM modelling of non-spher-
ical particulate systems: theoretical developments and applications, Powder
Technol. 302 (2016) 108–152.
[75] A. Uzi, Y. Ben Ami, A. Levy, Erosion prediction of industrial conveying pipelines,
Powder Technol. 309 (2017) 49–60.
[76] D. Zeng, E. Zhang, Y. Ding, Y. Yi, Q. Xian, G. Yao, H. Zhu, T. Shi, Investigation of ero-
sion behaviors of sulfur-particle-laden gas flow in an elbow via a CFD-DEM cou-
pling method, Powder Technol. 329 (2018) 115–128.
[77] S. Watano, S. Saito, T. Suzuki, Numerical simulation of electrostatic charge in pow-
der pneumatic conveying process, Powder Technol. 135 (2003) 112–117.
[78] P.A. Cundall, O.D.L. Strack, A discrete numerical model for granular assemblies,
Geotechnique 29 (1979) 47–65.
[79] H. Kruggel-Emden, E. Simsek, S. Rickelt, S. Wirtz, V. Scherer, Review and extension
of normal force models for the discrete element method, Powder Technol. 171
(2007) 157–173.
[80] A. Uzi, A. Levy, Energy absorption by the particle and the surface during impact,
Wear 404–405 (2018) 92–110.
[81] R. Ramírez, T. Pöschel, N.V. Brilliantov, T. Schwager, Coefficient of restitution of col-
liding viscoelastic spheres, Phys. Rev. E 60 (1999) 4465–4472.
[82] P.A. Langston, U. Tuzun, D.M. Heyes, Discrete element simulation of granular flow
in 2D and 3D hoppers—dependence of discharge rate and wall stress on particle in-
teractions, Chem. Eng. Sci. 50 (1995) 967–987.
[83] P.A. Langston, U. Tuzun, D.M. Heyes, Discrete element simulation of internal-stress
and flow-fields in funnel flow hoppers, Powder Technol. 85 (1995) 153–169.
[84] P.A. Langston, U. Tuzun, D.M. Heyes, Continuous potential discrete particle simula-
tions of stress and velocity-fields in hoppers—transition from fluid to granular
flow, Chem. Eng. Sci. 49 (1994) 1259–1275.
[85] Y.C. Zhou, B.D. Wright, R.Y. Yang, B.H. Xu, A.B. Yu, Rolling friction in the dynamic
simulation of sandpile formation, Physica A 269 (1999) 536–553.
[86] A.D. Gosman, E. Ioannides, Aspects of computer-simulation of liquid-fueled com-
bustors, J. Energ. 7 (1983) 482–490.
[87] C.T. Crowe, T.R. Troutt, J.N. Chung, Numerical models for two-phase turbulent
flows, Annu. Rev. Fluid Mech. 28 (1996) 11–43.
[88] S. Ergun, Fluid flow through packed columns, Chem. Eng. Prog. 48 (1952) 89–94.
[89] C.Y. Wen, Y.H. Yu, Mechanics of fluidization, Chem. Eng. Prog. Symp. Ser. 162
(1966) 100–111.
[90] R. Di Felice, The voidage function for fluid particle interaction systems, Int. J. Mul-
tiphase Flow 20 (1994) 153–159.
[91] L.W. Rong, Z.Y. Zhou, A.B. Yu, Lattice–Boltzmann simulation of fluid flow through
packed beds of uniform ellipsoids, Powder Technol. 285 (2015) 146–156.
[92] A. Hölzer, M. Sommerfeld, New simple correlation formula for the drag coefficient
of non-spherical particles, Powder Technol. 184 (2008) 361–365.
[93] L.W. Rong, K.J. Dong, A.B. Yu, Lattice-Boltzmann simulation of fluid flow through
packed beds of spheres: effect of particle size distribution, Chem. Eng. Sci. 116
(2014) 508–523.
[94] R. Beetstra, M.A. Van Der Hoef, J.a.M. Kuipers, Drag force of intermediate Reynolds
number flow past mono- and bidisperse arrays of spheres, AICHE J. 53 (2007)
489–501.
[95] L.W. Rong, K.J. Dong, A.B. Yu, Lattice-Boltzmann simulation of fluid flow through
packed beds of uniform spheres: effect of porosity, Chem. Eng. Sci. 99 (2013)
44–58.
[96] S.D. Dhole, R.P. Chhabra, V. Eswaran, Flow of power-law fluids past a sphere at in-
termediate Reynolds numbers, Ind. Eng. Chem. Res. 45 (2006) 4773–4781.
[97] Z. Qi, S.B. Kuang, L.W. Rong, A. Yu, Lattice boltzmann investigation of the wake ef-
fect on the interaction between particle and power-law fluid flow, Powder
Technol. 326 (2018) 208–221.
[98] Z. Qi, S.B. Kuang, A.B. Yu, Lattice boltzmann investigation of non-Newtonian fluid
flow through a packed bed of uniform spheres, Powder Technol. 343 (2019)
225–236.
[99] Q.J. Zheng, H.P. Zhu, A.B. Yu, Finite element analysis of the rolling friction of a vis-
cous particle on a rigid plane, Powder Technol. 207 (2011) 401–406.
[100] Q.J. Zheng, H.P. Zhu, A.B. Yu, Finite element analysis of the contact forces between a
viscoelastic sphere and rigid plane, Powder Technol. 226 (2012) 130–142.
206 S. Kuang et al. / Powder Technology 365 (2020) 186–207
[101] S.I. Rubinow, J.B. Keller, The transverse force on a spinning sphere moving in a vis-
cous fluid, J. Fluid Mech. 11 (2006).
[102] K.W. Chu, S.B. Kuang, Z.Y. Zhou, A.B. Yu, Model A vs. model B in the modelling of
particle-fluid flow, Powder Technol. 329 (2018) 47–54.
[103] J. Ferziger, M. Peric, Computational Methods for Fluid Dynamics, 3rd ed. Springer,
New York, 2002.
[104] K. Imaichi, K. Ohmi, Numerical processing of flow-visualization pictures measure-
ment of two-dimensional vortex flow, J. Fluid Mech. 129 (1983) 283–311.
[105] P.A. Jayantha, I.W. Turner, A comparison of gradient approximations for use in fi-
nite-volume computational models for two-dimensional diffusion equations,
Numer. Heat Transf. Part B-Fundam. 40 (2001) 367–390.
[106] J. Su, Z. Gu, C. Chen, X.Y. Xu, A two-layer mesh method for discrete element simu-
lation of gas-particle systems with arbitrarily polyhedral mesh, Int. J. Numer.
Methods Eng. 103 (2015) 759–780.
[107] C.L. Wu, J.M. Zhan, Y.S. Li, K.S. Lam, A.S. Berrouk, Accurate void fraction calculation
for three-dimensional discrete particle model on unstructured mesh, Chem. Eng.
Sci. 64 (2009) 1260–1266.
[108] J.E. Hilton, L.R. Mason, P.W. Cleary, Dynamics of gas–solid fluidised beds with non-
spherical particle geometry, Chem. Eng. Sci. 65 (2010) 1584–1596.
[109] Z. Peng, B. Moghtaderi, E. Doroodchi, A modified direct method for void fraction
calculation in CFD–DEM simulations, Adv. Powder Technol. 27 (2016) 19–32.
[110] S. Deb, D.K. Tafti, A novel two-grid formulation for fluid–particle systems using the
discrete element method, Powder Technol. 246 (2013) 601–616.
[111] R. Sun, H. Xiao, Diffusion-based coarse graining in hybrid continuum–discrete
solvers: theoretical formulation and a priori tests, Int. J. Multiphase Flow 77
(2015) 142–157.
[112] N. Gui, X. Yang, J. Tu, S. Jiang, A fine LES-DEM coupled simulation of gas-large par-
ticle motion in spouted bed using a conservative virtual volume fraction method,
Powder Technol. 330 (2018) 174–189.
[113] H.P. Zhu, A.B. Yu, Averaging method of granular materials, Phys. Rev. E 66 (2002).
[114] I. Lecreps, O. Orozovic, M.G. Jones, K. Sommer, Application of the principles of gas
permeability and stochastic particle agitation to predict the pressure loss in slug
flow pneumatic conveying systems, Powder Technol. 254 (2014) 508–516.
[115] K. Konrad, Dense-phase pneumatic conveying: a review, Powder Technol. 49
(1986) 1–35.
[116] B. Mi, P.W. Wypch, Pressure-drop prediction in low-velocity pneumatic conveying,
Powder Technol. 81 (1994) 125–137.
[117] H.P. Zhu, A.B. Yu, Averaging method of granular materials, Phys. Rev. E 66 (2002)
021302.
[118] H.P. Zhu, Z.Y. Zhou, A.B. Yu, P. Zulli, Stress fields of solid flow in a model blast fur-
nace, Granul. Matter 11 (2009) 269–280.
[119] Z.Y. Zhou, H.P. Zhu, B. Wright, A.B. Yu, P. Zulli, Gas–solid flow in an ironmaking
blast furnace-II: discrete particle simulation, Powder Technol. 208 (2011) 72–85.
[120] S.B. Kuang, Z.Y. Li, A.B. Yu, Review on modeling and simulation of blast furnace,
Steel Res. Int. 89 (2018) (1700071-1/25).
[121] Q.F. Hou, D.Y. E, S.B. Kuang, Z.Y. Li, A.B. Yu, DEM-based virtual experimental blast
furnace: a quasi-steady state model, Powder Technol. 314 (2017) 557–566.
[122] I. Lecreps, K. Sommer, K. Wolz, Stress states and porosity within horizontal slug by
dense-phase pneumatic conveying, Part. Sci. Technol. 27 (2009) 297–313.
[123] G. Niederreiter, K. Sommer, Modeling and experimental validation of pressure
drop for pneumatic plug conveying, Granul. Matter 6 (2004) 179–183.
[124] T. Krull, M.G. Jones, S. Keys, Stress-field modeling and pressure drop prediction for
slug-flow pneumatic conveying in an aerated radial stress chamber, Part. Sci.
Technol. 22 (2004) 129–138.
[125] N. Vasquez, L. Sanchez, G.E. Klinzing, S. Dhodapkar, Friction measurement in dense
phase plug flow analysis, Powder Technol. 137 (2003) 167–183.
[126] S. Shaul, H. Kalman, Friction forces of particulate plugs moving in vertical and hor-
izontal pipes, Powder Technol. 256 (2014) 310–323.
[127] R. Pan, Material properties and flow modes in pneumatic conveying, Powder
Technol. 104 (1999) 157–163.
[128] S.S. Mallick, P.W. Wypych, Minimum transport boundaries for pneumatic convey-
ing of powders, Powder Technol. 194 (2009) 181–186.
[129] Q.F. Hou, S.B. Kuang, A.B. Yu, A DEM-based approach for analyzing energy transi-
tions in granular and particle-fluid flows, Chem. Eng. Sci. 161 (2017) 67–79.
[130] A. Acock, T. Orourke, D. Shirmboh, J. Alexander, G. Andersen, T. Kaneko, A.
Venkitaraman, J. López-De, M. Cárdenas, M. Nishi, K. Numasawa, A. Yoshioka, A.
Roy, A. Wilson, Twynam, Practical approaches to sand management, Oilfield Rev.
16 (2004) 10–27.
[131] K.W. Chu, S.B. Kuang, A.B. Yu, A. Vince, G.D. Barnett, P.J. Barnett, Prediction of wear
and its effect on the multiphase flow and separation performance of dense me-
dium cyclone, Miner. Eng. 56 (2014) 91–101.
[132] I. Finnie, Erosion of surfaces by solid particles, Wear 3 (1960) 87–103.
[133] I. Finnie, Some observations on the erosion of ductile metals, Wear 19 (1972)
81–90.
[134] M. Parsi, K. Najmi, F. Najafifard, S. Hassani, B.S. Mclaury, S.A. Shirazi, A comprehen-
sive review of solid particle erosion modeling for oil and gas wells and pipelines
applications, J. Nat. Gas Sci. Eng. 21 (2014) 850–873.
[135] Y. Zhao, H. Ma, L. Xu, J. Zheng, An erosion model for the discrete element method,
Particuology 34 (2017) 81–88.
[136] L. Xu, K. Luo, Y. Zhao, Numerical prediction of wear in SAG mills based on DEM
simulations, Powder Technol. 329 (2018) 353–363.
[137] L. Xu, K. Luo, Y. Zhao, J. Fan, K. Cen, Multiscale investigation of tube erosion in flu-
idized bed based on CFD-DEM simulation, Chem. Eng. Sci. 183 (2018) 60–74.
[138] Y. Zhao, L. Xu, J. Zheng, CFD-DEM simulation of tube erosion in a fluidized bed,
AICHE J. 63 (2017) 418–437.
[139] H.C. Meng, K.C. Ludema, Wear models and predictive equations: their form and
content, Wear 181 (1995) 443–457.
[140] J. Chen, K.W. Chu, R.P. Zou, A.B. Yu, A. Vince, Prediction of the performance of dense
medium cyclones in coal preparation, Miner. Eng. 31 (2012) 59–70.
[141] L. Ji, J. Chen, S.B. Kuang, Z. Qi, K.W. Chu, A.B. Yu, Prediction of separation perfor-
mance of hydrocyclones by a PC-based model, Sep. Purif. Technol. 211 (2019)
141–150.
[142] G.E. Klinzing, Review of unusual and some unexplained phenomena in pneumatic
conveying, Part. Sci. Technol. 34 (2016) 424–431.
[143] H.J. Zhou, Y.Q. Xiong, Y. Pei, Effect of moisture content on dense-phase pneumatic
conveying of pulverized lignite under high pressure, Powder Technol. 287 (2016)
355–363.
[144] P.Y. Liu, R.Y. Yang, A.B. Yu, DEM study of the transverse mixing of wet particles in
rotating drums, Chem. Eng. Sci. 86 (2013) 99–107.
[145] M. Wang, W.B. Zhu, Q.Q. Sun, X.B. Zhang, A DEM simulation of dry and wet parti-
cles flow behaviors in riser, Powder Technol. Submitted 267 (2014) 221–233.
[146] B.A. Kotzur, R.J. Berry, S. Zigan, P. García-Triñanes, M.S.A. Bradley, Particle attrition
mechanisms, their characterisation, and application to horizontal lean phase pneu-
matic conveying systems: a review, Powder Technol. 334 (2018) 76–105.
[147] D.O. Potyondy, P.A. Cundall, A bonded-particle model for rock, Int. J. Rock Mech.
Min. Sci. 41 (2004) 1329–1364.
[148] H. Kalman, V. Rodnianski, M. Haim, A new method to implement comminution
functions into DEM simulation of a size reduction system due to particle-wall col-
lisions, Granul. Matter 11 (2009) 253–266.
[149] Y. Rozenblat, A. Levy, H. Kalman, J. Tomas, Impact velocity and compression force
relationship — equivalence function, Powder Technol. 235 (2013) 756–763.
[150] A. Uzi, H. Kalman, A. Levy, A novel particle attrition model for conveying systems,
Powder Technol. 298 (2016) 30–41.
[151] A. Uzi, A. Levy, Collision characteristics for predicting particle attrition and pipe
wear in conveying systems, Chem. Eng. Technol. 41 (2018) 1511–1523.
[152] K.W. Chu, J. Chen, A.B. Yu, Applicability of a coarse-grained CFD–DEM model on
dense medium cyclone, Miner. Eng. 90 (2016) 43–54.
[153] L. Ji, K. Chu, S. Kuang, J. Chen, A. Yu, Modelling the multiphase flow in
hydrocyclones using coarse-grained VOF-DEM and mixture-DEM approaches,
Ind. Eng. Chem. Res. 57 (2018) 9641–9655.
[154] J.Q. Gan, Z.Y. Zhou, A.B. Yu, A GPU-based DEM approach for modelling of particu-
late systems, Powder Technol. 301 (2016) 1172–1182.
[155] J. Xu, H. Qi, X. Fang, L. Lu, W. Ge, X. Wang, M. Xu, F. Chen, X. He, J. Li, Quasi-real-
time simulation of rotating drum using discrete element method with parallel
GPU computing, Particuology 9 (2011) 446–450.
[156] L.W. Rong, K.J. Dong, A.B. Yu, Lattice-Boltzmann simulation of fluid flow through
packed beds of uniform spheres: effect of porosity, Chem. Eng. Sci. 99 (2013)
44–58.
[157] C. Crowe, M. Sommerfeld, Y. Tsuji, Multiphase Flows with Droplets and Particles,
CRC Press, Boca Raton, 1998.
[158] E.E. Michaelides, A. Roig, A reinterpretation of the Odar and Hamilton data on the
unsteady equation of motion of particles, AICHE J. 57 (2011) 2997–3002.
[159] P.G. Saffman, The lift on a small sphere in a slow shear flow, J. Fluid Mech. 22
(1965) 385–400.
[160] R. Mei, An approximate expression for the shear lift force on a spherical particle at
finite Reynolds number, Int. J. Multiphase Flow 18 (1992) 145–147.
[161] B. Oesterle, T.B. Dinh, Experiments on the lift of a spinning sphere in a range of in-
termediate Reynolds numbers, Exp. Fluids 25 (1998) 16–22.
[162] K. Konrad, J.F. Davidson, The gas-liquid analogy in horizontal dense-phase pneu-
matic conveying, Powder Technol. 39 (1984) 191–198.
Nomenclature
A: Pipe cross-section area, m2
Asp: Surface area of a sample wall, m2
asphere,i: Fractional volume of particle i residing in the spherical cell
Cd0: Fluid drag coefficient on a isolated particle
Clm: Coefficient of Magnus lift force
Cls: Coefficient of Saffman lift force
dp: Particle diameter, m
ds: Diameter of the smaller part of a particle after impact, m
ds0: Diameter of the mother particle before impact
Dpipe: Diameter of pipe, m
D: Shear rate
E: Time-averaged energy, J
Eshear: Shear impact energy
Ee: Electric field strength
e: Energy, J
f: Particle scale forces, N
F: Volumetric force, N⋅m−3
g: Gravitational acceleration, m⋅s−2
H: Material hardness
I: Moment of inertia of particle, kg⋅m2
kc: Number of particles in a considered computational cell
ke: Number of particles exerting electrostatics forces on particle i
ki: Number of particles in contact with particle i
kw: Number of walls in contact with particle i
Kc: Critical stress intensity factor
L: Start-up section length, m
Lc: Characteristic length in the slug flow regime, m
 S. Kuang et al. / Powder Technology 365 (2020) 186–207 207

T: Torque, N⋅m εf: Local porosity

TACI: Time-averaged collision intensity, kg⋅m−1⋅s−2 εs: Local solid volume fraction
m: Mass of particle, kg η: Gas viscosity, kg⋅m−3⋅s−1
^ : Unit normal vector, m
n κ: Ratio of vertical to horizontal force components on the particle
N: Particle number λ: Charge density
Ncel: Number of CFD cells in the considered pipe section λ': Model parameter in in Eq. (19)
Np,section: Particle number in a pipe section μ: Friction coefficient
Np,sphere: Number of particles in the spherical cell ρf: Gas density, kg⋅m−3
Qerosion: Volume of material removed by the impact of a single particle, m3 ρp: Particle density, kg⋅m−3
Qe: Charge τ: Fluid viscous stress tensor, kg⋅m−1⋅s−2
qe: Equilibrium charge on the pipe wall ω: Particle angular velocity, s−1
_ _
q: Number of contact points between particles or particle and wall ω: Unit vector defined by ω¼ ω= j ω j
p: Plastic flow pressure, Pa ζ’′: Model parameter in Eq. (16)
P: Pressure, Pa
R: Vector from the mass centre of the particle to the contact point, m Subscripts
Rep: Reynolds number of particle
Rer: Reynolds number of particle rotation c: Contact
t: Time, s Cell: Computational cell
Δt: Time step, s d: Damping
Δtc: Characteristic time in the slug flow regime, s Drag: Fluid drag force
Δtsp: Sampling time period, s e: Electrostatic force
ΔVshpere: Volume of the spherical cell ep: Electrostatic forces due to other charged particles
v: Translational velocities of particle, m⋅s−1 ew: Electrostatic forces due to the pipe walls
vs: Slug velocity, m⋅s−1 Erosion: Wall erosion
vp: Average particle velocity, m⋅s−1 f: Fluid
vimpact: Particle impact velocity, m⋅s−1 i: Particle i
Vp: Volume of a particle, m3 ij: Between particle i and particle/wall j
Vpipe: Volume of a pipe section, m3 Impact: Impact of a particle on a wall
ΔV: Volume of a computational cell, m3 j: Particle/wall j
u: Gas velocity, m⋅s−1 Laminar: Laminar flow
~ : Instantaneous gas velocity, m⋅s−1
u lm: Magnus lift force
ui: Component of the velocity vector u ls: Saffman lift force
Ug: Superficial velocity, m⋅s−1 n: Normal component
wp: Solid mass flowrate, kg⋅s−1 p: Particle
wf: Gas mass flowrate, kg⋅s−1 Particle: Particle scale
Y: Young's Modulus, Pa p-f: Particle-fluid
pipe: Pipe section
Greek r: Rolling friction
s: Sliding
Π: Dimensionless number sp: Sampling
Ω: Relative angular velocity of gas to solid phases t: Tangential component
α: Particle impact angle Turbulent: Turbulent flow
αf: Proportionality factor vm: Virtual mass force
δt: Vector of the accumulated tangential displacement between particles i and j. ∇p: Pressure gradient force
ε0: Permittivity of free space ∇⋅τ: Viscous force

View publication stats