Download as pdf or txt
Download as pdf or txt
You are on page 1of 14

Advanced Powder Technology 30 (2019) 2338–2351

Contents lists available at ScienceDirect

Advanced Powder Technology


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

Original Research Paper

Numerical study of particle mixing in a tilted three-dimensional tumbler


and a new particle-size mixing index
Ziwei Zhang a, Nan Gui b,⇑, Liang Ge b, Zhenlin Li a
a
Department of Thermal Energy Engineering, College of Mechanical and Transportation Engineering, China University of Petroleum, Beijing 102249, PR China
b
Institute of Nuclear and New Energy Technology, Collaborative Innovation Center of Advanced Nuclear Energy Technology, Key Laboratory of Advanced Reactor Engineering
and Safety of Ministry of Education, Tsinghua University, Beijing 100084, PR China

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

Article history: A new mixing index is proposed, which is an improved Lacey index based on coordination number frac-
Received 23 March 2019 tion. The differences and similarities among many mixing indices are compared, including the new mix-
Received in revised form 16 July 2019 ing index, the information entropy based on coordination number fraction, the Lacey index based on local
Accepted 18 July 2019
concentration, and the information entropy based on local concentration. The first two indices are micro-
Available online 30 July 2019
scopic since the coordination number fraction is on particle-scale, whereas the latter two are mesoscopic
as the local concentration is mesoscopic scale. The newly proposed mixing evaluation indices does not
Keywords:
include inauthentic temporal oscillations. Moreover, using mixing index, the mixing characteristics of
Discrete element method
Particle
particles in a tilted tumbler are studied by discrete element method (DEM). The tumbler’s angle of tilt
Mixing a = 0°, 10°, 20°, 30°, 40°, 50°, 60° and 70°, at five rotating velocities x = 0.175, 0.35, 0.5, 0.6, 0.7 and
Tilted tumbler 1.4 rad/s corresponding to Froude number Fr = 0.0025, 0.001, 0.002, 0.003, 0.004, 0.016 respectively are
Mixing flow simulated. It is found that both increasing the tilt angle and the rotating speed have negative effects
on the particle mixing within the scope of this study.
Ó 2019 The Society of Powder Technology Japan. Published by Elsevier B.V. and The Society of Powder
Technology Japan. All rights reserved.

1. Introduction zone were investigated. It was found that the particle size and the
loading method have a considerable effect on this area, whereas
Granule mixing is a common and important unit operation in the mixer speed and shape do not act as critical parameters. Chou
many industries, such as food [1], chemical [2] and pharmaceutical et al. [6] measured the dynamic properties of particles in a rotating
industry [3]. There are many studies on the phenomenon of parti- drum by experimental methods. The control variables were the
cle mixing in horizontal tumblers. For example, Chandratilleke drum speed, particle surface structure and sidewall surface. The
et al. [3] studied the mixing of cohesive particles in a mixer with effects of internal friction of particles in both mono-disperse and
rotating blades to examine the cohesion and rake angle on flow bi-disperse systems were quantified. They found that the dynamic
and mixing behavior of fine particles. Finnie et al. [4] studied the properties reduce as the drum speed increases, and increase as the
longitudinal and transverse mixing phenomenon of particles in a internal friction of particles increases in a mono-disperse system,
rotary kiln, in which the size of particles exhibited a normal distri- while the dynamic properties of particles are more complicated
bution in a narrow range. It was shown that the mixing in the lon- in the bi-disperse systems. Arntz et al. [1] studied the rich phase
gitudinal direction can be described by a diffusion equation and behavior of the bi-disperse particle system in a rotating horizontal
the diffusion coefficient increased linearly with rotational speed, drum by discrete element method. The focus was on how drum
while the influence of the filling degree was relatively small. More- speed and filling level affect flow regime, mixing and segregation
over, the transverse mixing speed reduces as the rotational speed process.
and the filling degree increase with the same number of revolu- However, few studies have been contributed to three-
tions. Soni et al. [5] simulated the formation of the dead zone in dimensional tilted tumblers. Ward and Hourigan [2] focused on
a three-dimensional drum. The effects of particle sizes and packing the segregation of a bi-disperse mixture of granular material (glass
arrangements, the mixer speed and shape on the formation of dead beads) in a titled cylindrical drum. It was observed that smaller
particles form a single-segregated band at a fixed axial location.
⇑ Corresponding author. Moreover, the segregation would appear obviously for high Froude
E-mail address: guinan@mail.tsinghua.edu.cn (N. Gui). and high tilt angles. Similarly, the segregation dynamics in slurries

https://doi.org/10.1016/j.apt.2019.07.017
0921-8831/Ó 2019 The Society of Powder Technology Japan. Published by Elsevier B.V. and The Society of Powder Technology Japan. All rights reserved.
Z. Zhang et al. / Advanced Powder Technology 30 (2019) 2338–2351 2339

Nomenclature

Scalars v velocity
D particle diameter W evaluation functions, either information entropy or
Ek kinetic energy Lacey index
Fr Froude number x angular velocity rotating speeds
Lc cell side length
k stiffness coefficient vectors
m particle mass F force
Mc Lacey index based on coordination number fraction Fc contact force
Ml Lacey index based on local concentration Fg gravity force
n average cell size I moment of inertia
n0p average sample size M torque
N number of cells bn inter-particle displacements in the tangential and
Na;i N b;inumber of type-‘A’ particle and type-‘B’ particle in cell i, normal direction, respectively
respectively v velocity
NA NB total number of type-‘A’ particle and type-‘B’ particle, x angular velocity
respectively
N0p total number of type-‘p’ particle superscript
N0p;j number of type-‘p’ particle in sample j n normal direction
S information entropy t tangential direction
S0 information entropy of completely mixed system ‘0 ’ parameter in functions of the mixing indices based on
Sc information entropy based on coordination number coordination number fraction instead of these based
fraction on local concentration
Sl information entropy based on local concentration
VAR20 VAR2R VAR2 variance of fully demixed system, random
subscript
mixed system and fraction of type-‘B’ particle (or type-
a type-‘A’ particle
‘A’ particle) in each cell, respectively
b type-‘B’ particle
xi local concentration of type-‘A’ particles in cell i
c coordination number fraction
x0k;j local concentration of type-‘k’ particles in sample j
i cell i of W based on local concentration
x0k;m overall proportion of type-‘k’ particle
j sample j of W based on coordination number fraction
xm overall proportion of type-‘B’ particle in the system
l local concentration
a inclination of tumbler
p type of particle
c restitution coefficient
ea;j eb;j concentration of type-‘A’ and type-‘B’ particle in the cell
i,respectively.

of a tilted tumbler that contains water as the interstitial fluid and ence function were proposed to study mixing in a two-
bi-disperse glass beads was studied experimentally [7]. It was dimensional drum. Among them, the information entropy based
found that the local band and core correspond to a comparable fill on local concentration has the most wider applications since its
level in a horizontal cylinder. computing time is moderate. But it is still limited because of its
According to those researches on the horizontal drum, the drum dependence on grids. The information entropy based on radial dis-
speed is an important factor affecting the movement of particles tribution is also limited by its definition, since it can only evaluate
[4,6]. There is very little research on the mixing of particles in the radial mixing degree of particles. The information entropy
inclined drum. As a result, we will study the impact of rotating based on coordination number fraction is suitable for evaluating
speed and inclination on particle mixing in this work. the system with few particles, even the particle-scale system,
In order to quantify the degree of particle mixing, researchers and does not depend on the grid scales. Nevertheless, it is compu-
have put forward a number of mixed indices, such as the average tationally expensive for systems with multiple particles.
height method [8], the nearest neighbor method [9], the neighbor The original form of the Lacey index was presented by Lacey
distance method [9], the Lacey index [10-12], the information [10]. It is a method based on the variance of local concentrations.
entropy [13-16], the coordination number method [9,17], the Siiria It has been very widely used, but it cannot be applied to evaluate
method [18], the PSMI method [17], and the concentration differ- multi-particle systems and it relies on grids. Similar to the method
ence method [19]. Among them, the information entropy and of information entropy based on the coordination number fraction
Lacey index have the most extensive applications. [14], we propose an new mixing index based on the coordination
The information entropy [13] was firstly presented by Shannon number fraction herein, and use it to quantify the mixing of parti-
in 1949, whose initial definition is the expectation of information. cles in the tilting tumblers.
It is a concept to measure information and similar to thermody- To say conclusively, although the impact of rotating speed and
namic entropy. The information entropy has a wide range of appli- inclination on particle mixing in tilted tumblers is significant, it
cations. Based on different functions, information entropy also has rarely been investigated, particularly by using meso- and
shows different physical properties. The information entropy can particle-scale methods. To continue the earlier work further, a
be divided into three categories according to different functions, new mixing index based on the coordination number fraction is
i.e. the local concentration [1,4,20-25], the radial distribution func- proposed here to study the mixing characteristics in tilted 3D drum
tions [16] and the coordination number fraction [14]. In the earlier quantitatively. In essence, it is an improved Lacey index based on
work [16], a radial distribution function-based information coordination number fraction, rather than on the local concentra-
entropy function and a dimensionless local concentration differ- tion as in the conventional Lacey index. Since the coordination
2340 Z. Zhang et al. / Advanced Powder Technology 30 (2019) 2338–2351

number is particle-scale, the new mixing index can evaluate micro- 2.2. Numerical setup
scopic mixing degree, while the local concentration is limited to
mesoscopic scale. Although different particle mixing indices can The inclination of the tilted tumbler is defined as the angle
be used to evaluate particle-mixing degree, each particle-mixing between the axis of the tumbler and the horizontal plane. The tum-
index has its own properties and scopes of application. Therefore, bler with a tilt angle of 0°is horizontal. For display convenience, the
we compare the information entropy based on the coordination XOY plane is established on the front wall of the tumbler and the Z-
number fraction, the information entropy based on the local con- axis direction is the length direction of the tumbler. Eight angles of
centration, the Lacey index based on the coordination number frac- tilt a = 0°, 10°, 20°, 30°, 40°, 50°, 60°, 70° at six rotating velocities
tion and the lacey index based on the local concentration to select x = 0.175, 0.35, 0.5, 0.6, 0.7, 1.4 rad/s are simulated.
the optimal mixing index, and to provide a reference for future First, all the particles are pre-filled in a horizontal tumbler.
research work. Then, we show the respective characteristics and Then, the tumbler is tilted to the predetermined angle. After all
the applicable conditions of them. the particles are stationary, a fictitious plane of x = 0 is used to
divide the particles into two categories, i.e. type-‘A’ particle and
type-‘B’ particle. Then, the tumbler starts to rotate counterclock-
2. Numerical approach wise at a predetermined speed (Fig. 1a). The simulation parameters
and cases are presented in Table 1. The values of them are deter-
2.1. Discrete element method mined mainly according to our empirical experience in this field,
including the restitution coefficient, the friction coefficient, and
In this work, the DEM method is used to simulate the particle the poison’s ratio which are within the ranges widely encountered
motion in tilted tumblers. The DEM method has two basic assump- and applied in engineering. The friction coefficient is small since
tions:(1). Direct interaction: within a time step, there are only par- our mostly interested particles in pebble bed reactor are made of
ticles in direct contact possibly affect the motion of each other, and graphite, which has a fairly small friction coefficient. Finally, simi-
other particles that are not in contact have no effect on the parti- lar DEM simulations have been conducted to study particle mixing
cles. Therefore, the collision disturbance can only propagate to in bladed mixer, in accompany with the comparison of existing
the particles in direct contact. (2). Instantaneously decoupled many mixing indices. Therefore, the model has been well validated
motion and collision: The velocities and accelerations of particles in the earlier work [26].
remain unchanged during the computation of collision, whereas
the motions are solved immediately after each collision 2.3. Mixing index
computation.
The limitation is that the hard-sphere approach can only be Two kinds of concentrations are defined, which are local con-
used to simulate dilute flows, whereas the soft-sphere approach centration and coordination number fraction. The former one is
can be used to simulate dense flows. There are two reasons causing based on the cell divided by Cartesian grids, whereas the latter
large amount of computation. Firstly, the time step needs to be one is composed of a core particle and its coordination particles.
fairly small to restrict the effects of instantaneous decomposition The mixing indices, the information entropy and the Lacey index,
of motion and collision. Secondly, the contact collision determina- based on the local concentration are mesoscopic scale since all
tion/retrieval is always necessary in each time step. Dense flow is information smaller than the mesoscopic scale is averaged to com-
computationally intensive and have hardware limitations. How- pute the concentration. On the contrary, the mixing indices based
ever, as the particle flow in rotating drum is dense, the soft- on the coordination number fraction is particle-scale since it is
sphere approach of discrete element method is still adopted in this directly determined by the most microscopic information sur-
study. rounding a core particle.
In the soft-sphere approach, the normal and tangential force
models are governed [25], respectively, by: 2.3.1. Information entropy
The information entropy was first introduced by Shannon [13].
n _ 1=4
F n ¼ k n3=2  cn nn ð1Þ Its initial definition is an expectancy of information, like the con-
ception of thermal entropy. The information entropy has such a
F t ¼ k b  ct b_ mathematical property: the greater the uncertainty of the input
t

information, the greater the value, and vice versa. As for informa-
where F, k, c, n and b are the force, the stiffness coefficient, the tion entropies based on appropriate mixing information of parti-
damping coefficient, the inter-particle displacements, respectively. cles, its mathematical property is: the better the mixing degree,
The superscripts ‘n’ and ‘t’ denote the normal and tangential direc- the greater the corresponding value, and vice versa. The informa-
tion, respectively. tion entropy based on the local concentration, where the space is
According to Newton’s second law, the governing equations divided by Cartesian grids into some cubes as cells, has been
include the translational motion and the rotational motion of any widely used to evaluate the mixing degree. In a system with two
particle: kinds of particle, e.g. type-‘A’ particle and type-‘B’ particle, the local
concentration in cell i is defined as:
dv i X c 8
mi ¼ F ij þ F gi ð2Þ < ea;i ¼ Na;i
Na;i þN b;i
dt j ð4Þ
:e ¼
N b;i
b;i N a;i þNb;i

dxi X where Na,i and Nb,i are the number of type-‘A’ particle and type-‘B’
Ii ¼ M ij ð3Þ
dt j
particle in cell i, respectively. The information entropy in the cell
can be given by:
where mi, v, Fc, Fg, I, x and M are the particle mass, the velocity, the Si ¼ ea;i log2 ðea;i Þ  eb;i log2 ðeb;i Þ ð5Þ
contact force, the gravity force, the moment of inertia, the angular
velocity and the torque respectively. Finally, the total information entropy is:
Z. Zhang et al. / Advanced Powder Technology 30 (2019) 2338–2351 2341

Fig. 1. Snapshots of particle distribution in the tumbler under x = 0.35 rad/s for a = 40° at t = 6 (a), 12 (b), 18 (c), 24 (d) and 30 s (e) respectively.

1 X N a;i þ N b;i type-‘B’ particle respectively. Divided by S0, the final entropy
Sl ¼ Si ð6Þ
S0 j N A þ N B become a dimensionless variable. Moreover, its value equals 1 in
completely mixed system and equals 0 in fully de-mixed system.
where subscript ‘l’ of Sl denotes that information entropy is based Ayeni et al. [14] presented a new entropy, which is based on the
on local concentration, S0 denotes the information entropy of com- coordination number fraction. The coordination number of a parti-
pletely mixed system, NA and NB is total number of type-‘A’ and cle is the number of particles in contact with the core particle. The
2342 Z. Zhang et al. / Advanced Powder Technology 30 (2019) 2338–2351

Table 1 in which VAR20, VAR2R and VAR2 respectively present the variance of
Parameters used in the tumbler. fully demixed system, the randomly mixed system and the fraction
Radius of gyration, R0 (m) 0.8 of type-‘B’ particle (or type-‘A’ particle. In this work, we choose
Length of tumbler, L (m) 0.6
type-‘B’ particle) in each cell, which is divided by Cartesian grids.
Number of particle 18,000 Ml denotes the Lacey index based on local concentration and the
Particle diameter,D(m) 0.030 subscript ‘l’ means local concentration. These variables can be
Particle density, (kg/m3) 1600 obtained as follows:
Restitution coefficient,e 0.9
Friction coefficient, c 0.3 1 XN
Poisson’s ratio of particle, kp 0.35 VAR2 ¼ ðxi  xm Þ2 ð12Þ
Poisson’s ratio of wall, kw 0.23 N i¼1
Young modulus of particles, Ep (GPa) 0.2
Young modulus of wall, Ew (GPa) 10
Rotating velocity, x (rad/s) 0.175, 0.35, 0.5, 0.6, 0.7, 1.4
VAR20 ¼ xm ð1  xm Þ ð13Þ
Simulation time step, s 5.0106
Total simulation time, Ts (s) 80,120 xm ð1  xm Þ
Cases x = 0.175 rad/s, a = 10°, 40° VAR2R ¼ ð14Þ
n
x = 0.35 rad/s, a = 0°, 10°, 20°, 30°,
40°, 50°, 60°, 70° where xi is the local concentration of type-‘A’ particles in cell i, and
x = 0.5 rad/s, a = 10°, 40° xm is the overall proportion of type-‘B’ particle in the system. n is the
x = 0.6 rad/s, a = 10°, 40° average number in a cell and N is the number of cells with existing
x = 0.7 rad/s, a = 0°, 10°, 20°, 30°,
40°, 50°, 60°, 70°
particles. xi, xm, and n can be calculated through:
x = 1.4 rad/s, a = 0°, 10°, 20°, 30°, xi ¼ Nb;i =ðNb;i þ Na;i Þ ð15Þ
40°, 50°, 60°, 70°

xm ¼ Nb =ðNa þ Nb Þ ð16Þ

contact is defined that the distance between the surfaces of two 1X N


n¼ ðNb;i þ Na;i Þ ð17Þ
particles is less than 10%D, where D is diameter of particle in a N i¼1
mono-system. The sample of a certain particle is composed by
the particle and its contact particles. The concentration in Ayeni’s The definition of Na,i, Nb,i, Na and Nb are the same as in the for-
research is volume concentration [14], because they studied bi- mer section.
dispersed system. In this paper, we simplify it as the ratio of parti- According to the definition of information entropy based on the
cle numbers to deal with the mono-dispersed systems. coordination number fraction, we propose a new mixing index
In sample j, when the core particle is type-‘A’ particle, its con- here, which is a kind of improved Lacey index based on the coordi-
centration function is defined as: nation number fraction rather than on local concentration. The def-
8 inition of the sample is the same as that of information entropy
> N 0 a;j þ 1
< e0 a;j ¼ N0 a;j þN0 b;j þ 1
based on the coordination number fraction. A sample should
ð7Þ include a core particle and several particles in contact with it.
>
: e0 b;j ¼ N 0 b;j
N 0 a;j þN 0 b;j þ 1 The core particle can be any kind of particle.
To evaluate the system with multi-kinds of particle, the new
When the core particle is type-‘B’ particle, its concentration mixing index is defined as follow:
function is defined as:
8 X N0 p VAR02 02
p  VARp;0
>
< e0 a;j ¼
N 0 a;j Mc ¼ P 0 M c;p ; Mc;p ¼ ð18Þ
N0 a;j þN0 b;j þ 1 p Np VAR02 02
p;R  VARp;0
ð8Þ
>
: e0 b;j ¼ N0 b;j þ1
N 0 a;j þN 0 b;j þ 1 in which the subscript ‘c’ means the index is based on the coordina-
tion number fraction and ‘p’ means the type of particle. The index
in which N0 a,j and N0 b,j are the type-‘A’ and type-‘B’ particle contact can evaluate mixing for a system with multi-kinds of particles
numbers with the core particle, respectively, e0 a,j and e0 b,j are the through the weighted average index of all kinds of particles. The
concentrations of type-‘A’ and type-‘B’ particle in the sample, subscripts ‘p’ can be 2, 3 or larger numbers.
respectively.
The definitions of variancesVAR02 02 02
p , VARp;0 and VARp;R are similar
The information entropy based on the coordination number
as that in Ml:
fraction is defined as:
0
1 X
Np
S0j ¼ e0a;j log2 ðe0a;j Þ  e0b;j log2 ðe0b;j Þ ð9Þ
VAR02
p ¼ 0 ðx0 p;j  x0 p;m Þ ð19Þ
N p j¼1
1 X n a;j þ n b;j 0
0 0
Sc ¼ Sj ð10Þ
S0 0 j N A þ N b VAR02 0 0
p;0 ¼ xp;m ð1  xp;m Þ ð20Þ
0
in which S 0 denotes the information entropy of completely mixed
x0 p;m ð1  x0 p;m Þ
system, subscript ‘c’ of Sc denotes that information entropy is based VAR02
p;R ¼ ð21Þ
on the coordination number fraction. n0 p
where x0 p,j is the concentration of type-‘p’ particles in sample j. x0 p,m
2.3.2. Lacey index is the overall proportion of type-‘p’ particle in the system. n0 p is the
Lacey [10] introduced the Lacey index, which is based on the average sample size and N0 p is the number of samples where type-
local concentration. It is defined as follows: ‘p’ particle is the core particle. In order to distinguish two kinds of
Lacey indices, these variances are added superscript ‘0 ’ and subscript
VAR2  VAR20
Ml ¼ ð11Þ ‘p’. Corresponding to the sample based on the coordination number
VAR2R  VAR20 fraction, x0 p,j, x0 p,m, n0 p and N0 p can be defined as follows:
Z. Zhang et al. / Advanced Powder Technology 30 (2019) 2338–2351 2343

X 0
x0p;j ¼ ðN0p;j þ 1Þ=ð N p;j þ 1Þ ð22Þ tical direction and the gravity vector in Fig. 2), more and more par-
p ticles accumulate on the rear wall of the drum, and the maximum
X heights of the particle in the y-directions are increasing. As the tilt
x0p;m ¼ Np 0 = Np 0 ð23Þ angle increases, the helical de-mix streaks become more and more
p obscured. The contact area of the front wall with the particle
becomes reduced. When the inclination increases to a = 60° and
0
1 X
Np X 70° (Fig. 2g,h), the particles leave the front wall almost completely,
n0p ¼ 0 ð N0 p;j þ 1Þ ð24Þ and the helical streaks disappear since they are highly suppressed
N p i¼1 p
within a short range in z-direction (length-direction) by the large
where N0 p,j is the number of type-‘p’ particle in sample j. ‘10 means component of gravity toward the rear wall (see the insets of
the number of core particle itself.
In currently proposed index, it regards type-‘p’ particle as the
core particle. The concentration of the type-‘p’ (p = 1, 2, . . .) particle
in the sample is calculated and used to replace the local concentra-
tion in the conventional Lacey method. Then, the index of type-‘p’
particle based on the concentration is calculated. Finally, the
weighted average indices for all kinds of particles are calculated
to obtain the new index (Eq. (18)). The benefit of current new index
is that it can evaluate mixing of a system with three or more kind
of particles, i.e. it can evaluate multiple particle systems.

2.3.3. Kinetic energy


The total kinetic energy of particles is computed here to help
analyze the mixing behavior. The total kinetic energy is transferred
from the rotating tumbler to particles, which is the direct driving
force for the movement and mixing of the particle system. The for-
mulation of kinetic energy is:
X1
Ek ¼ mv 2 ð25Þ
2
in which Ek represents the total kinetic energy of all particles at a
certain moment, m is the mass of a particle, v represents the speed
of the particle.

3. Results and discussions

In this work, the four kinds of mixing indices are used to


evaluate mixing and show the advantages and disadvantages of
them, as well as their characteristics and application conditions.
The evolution of mixing processes, the effects of rotating speeds
and the angles of tilt are compared through the snapshots of parti-
cle distributions and the evolutions of mixing indices.

3.1. Mixing process

In this section, the evolution of mixing processes, the effects of


rotating speeds x and inclinations a are compared through the
representative snapshots of particle distributions.

3.1.1. Temporal evolution of mixing and mixing states at various


angles of tilt a
The mixing processes of particles in the tumbler with a rotation
speed of 0.35 rad/s at an angle of tilt of 40° are shown in Fig. 1a–e.
As the number of revolution increases, the mixing in the inclined
tumbler is ongoing. The mixing interface is inclined to the horizon-
tal plane and the two kinds of particles form helical streaks, which
are different from the horizontal rotating tumbler. To say specifi-
cally, the helical structure of the red and blue parts is composed
of 1 helix within 0.6 m in z-direction at t = 6 s (Fig. 1a), whereas
it is about 1.5 helixes within the same 0.6 m in z-direction at
t = 12 s (Fig. 1b) and 2 helixes at t = 18, 24, and 30 s (Fig. 1c–e).
On the other hand, the snapshots of particles in the tumbler at
the rotation speed of 0.35 rad/s with angles of tilt of a = 0°, 10°, 20°, Fig. 2. Snapshots of particle distribution in the tumbler under x = 0.35 rad/s at
30°, 40°, 50°, 60° and 70° at t = 30 s are shown in Fig. (2a-h). We t = 30 s for a = 0° (a), 10° (b), 20° (c), 30° (d), 40° (e), 50° (f), 60° (g) and 70° (h)
find that as the tilt angle increases (see the angles between the ver- respectively.
2344 Z. Zhang et al. / Advanced Powder Technology 30 (2019) 2338–2351

Fig. 2g and h). Especially for Fig. 2h at a = 70°, the helical structure fed into a downwardly moving surface region at all angles of the
is degraded into a completely circular mixing structure in the XOY drum. Opposite processes take place in the bottom half of the
plane and with a certain thickness in z-direction. However, we can- slope”. The phenomena of particle flows shown in Figs. 1–3 are
not intuitively observe how the degree of mixing varies with incli- consistent with this definition of the rolling flow regime.
nations. Therefore, we need to evaluate the particle mixing in
following sections by using the new mixing index, i.e. the Lacey 3.1.2. Mixing states at various rotating speeds x
index based on the coordination number fraction. The snapshots of particle mixing in the tumbler with the angle
In addition, the particle flow regimes are within the rolling of tilt a = 40° and rotation speeds of x = 0.175, 0.35, 0.5, 0.6, 0.7
regime. Referring to Mellmann’s summary [28] on rolling flow and 1.4 rad/s at revolution number = 2 are shown in Fig. 3. At this
regime definition: ‘‘the free surface is flat with a steady feed of par- moment, as the rotation speed increases, the maximum height of
ticles to the top of the surface. In the upper half, material is being the particles in the y-direction increases because the centrifugal

Fig. 3. Snapshots of particle distribution in the tumbler for inclined angle a = 40° and x = 0.175 (a), 0.35 (b), 0.5 (c), 0.6 (d), 0.7 (e) and 1.4 rad/s (f) respectively at revolution
number = 2.
Z. Zhang et al. / Advanced Powder Technology 30 (2019) 2338–2351 2345

force may increase caused by the high-speed rotation. Meanwhile, rioration with increasing inclination. As we described before,
similar helical streaks exist in the snapshots of particle mixing in the increasing inclination causes a higher gravity component
the tumbler with the rotation speeds of x = 0.175, 0.35, 0.5, 0.6 in the axial direction of the drum and a lower gravity com-
and 0.7 rad/s. At the rotational speed of x = 1.4 rad/s, the particle ponent in the coordinate-y direction. It also causes particles
mixing is worse than other cases. Therefore, in the inclined tum- to leave the front wall at high inclinations (Fig. 1), which
bler, high rotation speed may be a negative factor affecting particle leads to structural degradation of the mixing interface. As
mixing. the inclination increases, we speculate that, both of these
factors result in the reduction of kinetic energy transferred
3.2. Quantitative analysis from the drum to particles, i.e., the direct power for particle
mixing. In addition, as the tilt angle a increases, the differ-
The kinetic energy and the new mixing index are analyzed ence between adjacent average kinetic energy increases first
herein in order to quantitatively study the influences of tumbler (x = 0°-50°) and then decreases (x = 60°-70°), which vali-
rotation speed and angle of tilt on the particle mixing in the dates the degradation of mixing interface structure. To say
inclined tumbler. In general, the rate of increase of mixing index conclusively, we can argue that the increase of tilted angle
indicates the grow rate of mixing, whereas the value of mixing has a negative effect on particle mixing.
index indicates the absolute degree of particle mixing.
3.2.2. Effect of rotation speed
3.2.1. Effect of the angle of tilt Fig. 5 shows the evolutions of the new mixing index and the
The mixing index and kinetic energy at rotation speeds of kinetic energy for various rotating speeds at angles of tilt a = 10°
x = 0.35, 0.7 and 1.4 rad/s are shown in Fig. 4. It can be clearly seen and 40°. It can be clearly seen:
from Fig. 4a-c that:
1) For the cases of inclinations of a = 10° (Fig. 5a), the rotating
1) As tumbler rotation progresses, the trends of new mixing speed has few effects on the mixing rate and the final mixing
index Mc, or the degree of mixing, are continuously increas- state. The slope of Mc under x = 1.4 rad/s, i.e., the mixing
ing. With the increase of the inclination angle, the slopes of rate, is slightly smaller than that under other rotation
Mc, i.e. the rate of mixing development, are getting worse. At speeds. For the case at inclination angle a = 40°, the mixing
a fixed speed of rotation, the mixing degrees with tilt angles rate decreases as the revolution number increases. Because,
of a = 0° and 10° are very close at any time, whereas at an at the same revolution number, the mixing time may be less
inclination of a = 70°, the mixing of particles is kept on a and the mixing may be worse when the rotation speed
low level shortly after the start of rotation. That is to say, becomes greater. The increasing of inclination angle a
when the tilt angle is increased to a certain extent, the mix- enlarges the difference of mixing rate variation caused by
ing of the particles may be suppressed by the axial compo- the increase of rotation speed. It also reflects the negative
nent of gravity in z-direction. Recalling the observations in impact of tilt angle a on mixing from another perspective.
Fig. 1, the deterioration of mixing at large angle of tilt is In combination with Section 3.2.1, it can be determined that
mainly caused by the degradation of mixing interface from the effect of tilt angle a on the particle mixing is greater than
helical structure to circular planar structure. the effect of rotational speed x.
2) As shown in Fig. 4c, in the cases of rotation speed 2) In the cases of a = 10° and the rotating speed x = 1.4, 0.7,
x = 1.4 rad/s and tilt angle a = 0°-50°, the mixing degrees 0.6 rad/s, the final values of the new mixing index Mc succes-
of particles successively reach the same ultimate state with sively reach the same ultimate state of particle mixing with
the rotation of the tumbler. While the degree of particle mix- the rotation of the tumbler. The conclusions in cases of
ing at inclination of a = 60° continues to rise and is lower a = 40° and the rotating speed x = 1.4, 0.7 rad/s are similar.
than the level of mixing at the former angles of inclination. It is because their mixing structures and flow regimes are
Therefore, the increase of tilt angle can slow down the rate same, respectively.
of development of mixing and make more time to reach 3) Similar to the case of Fig. 4d-f, ignoring the initial distur-
the ultimate mixing state. It may be because that the bance, the kinetic energies of cases are basically unchanged
increasing inclination causes a higher gravity component with time in this section. As the rotational speed increases,
in the axial direction of the drum (i.e., the z-axis or the coor- the kinetic energy provided by the tumbler to particles is
dinate axis) and a lower gravity component in the y direc- increased. As described in Eq. (25), the kinetic energy Ek is
tion, which suppresses the mixing speed. When particles proportional to the square of the velocity v2. However, as
do not completely leave the front wall (a = 0°-50°), the ulti- the angular velocity x = v/r in which r is a constant radius,
mate state of evolution is basically the same since the degra- Ek is proportional to the square of angular velocity x2. To
dation of mixing structure do not occur and systems are all construct Fig. 5c and Fig. 5d, the average kinetic energy dur-
in the rolling flow regime. Otherwise, the ultimate states ing the last 20 s is calculated as the dependent variable, and
may be deteriorated when the angle of inclination is high the square of angular velocity is calculated as the indepen-
(a = 60°-70°). Although in some cases of x = 0.35 rad/s and dent variable. In Fig. 5g, we find that Ek  x2 shows almost
x = 0.7 rad/s, the ultimate mixing degree are not achieved, a straight line at rotation speeds x = 0.175, 0.35, 0.5, 0.6
it can be seen that the general trends are similar with that and 0.7 rad/s. That is to say, the average kinetic energy Ek
in the case of x = 1.4 rad/s. Therefore, similar results can and x2 are linearly related, whose radio is a constant. Nev-
be obtained. ertheless, the point under x = 1.4 rad/s is lower than the fit-
3) Moreover, from the kinetic energy shown in Fig. 4d–f, we ted line. Under the high speed x = 1.4 rad/s, the kinetic
can see that ignoring the initial perturbation, the kinetic energy provided by the tumbler for particles is lower than
energies basically do not change with time. Therefore, we the product of this ratio and the square of rotation speed
calculated the average kinetic energy during the last 20 s x. Therefore, the mixing rate of the case under x = 1.4 rad/
and showed them in Fig. 4g. As the inclination angle a s is lower than that of other cases. To sum up, high rotation
increases, the average kinetic energies provided by the tum- speed is a negative factor affecting the mixing rate of parti-
bler to particles are reduced, which leads to the mixing dete- cles, but its influence is lower than the tilt angle.
2346 Z. Zhang et al. / Advanced Powder Technology 30 (2019) 2338–2351

Fig. 4. Lacey index based on coordination number fraction (a–c), kinetic energy (d–f) and average kinetic energy during the last 20 s (g) at x = 0.35, 0.7 and 1.4 rad/s
respectively.
Z. Zhang et al. / Advanced Powder Technology 30 (2019) 2338–2351 2347

3.3. Performance comparison among mixing evaluation functions centration Ml and the coordination number fraction Mc, as well as
the information entropies based on local concentration Sl and coor-
The performance of a number of evaluation functions are dination number fraction Sc. It should be emphasized that, the
compared in this section, i.e. the mixing indices based on local con- evaluation functions, either mixing index or information entropy,
are differentiated from whether they evaluate the degree of mixing
from the mesoscopic or the microscopic perspective. To say specif-
ically, the functions based on the local concentration (either Ml or
Sl) are mesoscopic, whereas the functions based on the coordina-
tion number fraction (either Mc or Sc) are microscopic.
Additionally, only the mixing indices with the tilt angles of
a = 20°, 50° and 70° at rotating speed of x = 0.7 rad/s are analyzed
in the following section for case study. In fact, the performances of
mixing indices in other cases are basically similar.

3.3.1. Comparison between mixing index m and information entropy s


Fig. 6a shows the information entropy Sl and the Lacey index Ml
based on local concentration at x = 0.7 rad/s, which evaluate the
degree of particle mixing at a mesoscopic scale. For the tilt angle
a = 20°, the upright auxiliary red lines are on the location of the
peaks and the troughs of Sl and Ml. It can be observed that the ris-
ing and falling trends of Sl and Ml, as well as the positions of
troughs and valleys of Sl and Ml, are almost located at the same
time points. Their specific values are also similar, though not all
the same. This coincident occurrence time of peaks and troughs
of Sl and Ml also applies to other cases of inclinations a = 50° and
70° in Fig. 6a.

Fig. 6. (a) information entropy based on local concentration Sl and Lacey index
based on local concentration Ml at x = 0.7 rad/s; (b) information entropy based on
Fig. 5. Lacey index based on coordination number fraction (a, b) and average Ek- x
2
coordination number fraction Sc and Lacey index based on coordination number
(c, d) at a = 10° and 40° respectively. fraction Mc at x = 0.7 rad/s.
2348 Z. Zhang et al. / Advanced Powder Technology 30 (2019) 2338–2351

On the other hand, the information entropy based on the on the local concentration (Wl) rise with time in oscillating and
coordination number fraction Sc and the Lacey index based on periodical patterns.
the coordination number fraction Mc at rotating velocity Why do Sl and Ml ascend in an oscillating pattern with time?
x = 0.7 rad/s are shown in Fig. 6b to evaluate the microcospic fea- Why do Sc and Mc not oscillate? The oscillation of Sl and Ml means
ture of particle mixing. The rising trend and the trend to reach the that the particle mixing level can go down sometimes, e.g. from
ultimate state of Sc and Mc are similar under the same tilt angle a. t = 10 s to 12 s in Fig. 7a. However, Sl and Ml do not oscillate during
However, the values of them are different, because the information t = 10–12 s in Fig. 7b. Therefore, we must refer to the mixing states
entropy and the Lacey index evaluate mixing degree differently. to determine which one is right. Fig. 8a-c shows the mixing states
The local concentration weakens the difference, whereas the coor- at t = 10 s, 11 s, and 12 s respectively. It seems that Sl or Ml reaches
dination number fraction enhances it. a local peak when the main blue part is above the main red part
In Fig. 6, from physical point of view, the trends of mixing (Fig. 8a), whereas they get local minimum values when the main
indices (either (Sc and Mc) or (Sl and Ml)) based on the same scale red part is above the main blue part (Fig. 8c). Remember that the
(either mesoscopic or microscopic) should be unique for the same definitions of Sl or Ml are both based on the local concentration.
case. This confirms the correctness of the new mixing indices. The local concentration is computed by fixed meshes to discretize
or cover the mixing interface. Therefore, the values of local concen-
tration will be certainly related to the orientation or relative posi-
3.3.2. Comparison between mesoscopic (‘l’) and microscopic (‘c’)
tion of the mixing interface. Thus, such oscillating patterns of Sl or
mixing
Ml do not indicate the real improvement or deterioration of mixing
Using the same evaluation functions of information entropy
levels. In other words, the new mixing indices or information
(Fig. 7a) and Lacey mixing index (Fig. 7b) at x = 0.7 rad/s, the
entropy (Sc and Mc) defined based on the coordination number
mesoscopic (‘l’) and microscopic (‘c’) mixing characteristics are
fraction are different from that defined based on the local concen-
analyzed by comparing them defined based on the local concentra-
tration (Sl and Ml). The oscillation of Wl is basically caused by the
tion and the coordination number fraction. Although both (Wl) and
division of cells when the mesoscopic fixed cells are applied to
(Wc) (W denotes evaluation functions, i.e. W = S for information
compute evaluation function W.
entropy or W = M for mixing index) indicate the tendencies of
However, the existence of oscillation is irrelevant to the cell
degree of mixing which are increasing with rotation, the evaluation
size. In the calculation of local concentration, we used a 3D Carte-
functions based on the coordination number fraction (Wc) rise
sian grid to divide cells, which are cubes with a certain side length.
monotonically with time, while the evaluation functions based
This method has been widely used in previous studies [1,19,23,27].
In Fig. 9, six cell side lengths Lc = 1.5D, 2D, 3D, 4D, 5D, and 10D are
chosen and the corresponding indices Wl are shown. The corre-
sponding numbers of cells are 36  36  14, 27  27  10,
18  18  7, 14  14  5, 12  12  4, 6  6  2, and particle num-
ber in each cell is about 3, 7, 21, 48, 87, and 514, respectively.
It is found that the choice of cell side lengths Lc does not elim-
inate the oscillation. Moreover, as the length increases, the unreal-
istic oscillation amplitude increases. Notice that the mixing index
of particle system is a weighted average of mixing indices of every
sampled cell. In a segregation-dominated particle system, most of
the samples are in segregated states, and only a small portion of
the samples are in mixed states. As the cell side length increases,
the proportion of samples in the mixing state increases. Conse-
quently, the value of mixing index increases, and vice versa. In
other words, when the cell side length is too large, the details of
the particle system structure should be ignored. Like a sieve, if
the mesh size is larger, more detailed information will be screened
out.
Moreover, both in Fig. 9a and Fig. 9b, with a large cell side
length Lc = 10D, it causes the index value in segregation state to
be larger than that in mixing state, which violates the physical
meaning of mixing indices itself, i.e., truly reflecting the mixing
state. With a small cell side length Lc = 1.5D, the average particle
number in a cell is only about 3, which means the sample size is
too small to constitute one as a basic calculation unit cell, proved
by the abnormal situation in the case of Lc = 1.5D (Fig. 9b). In sum-
mary, we choose the cell of Lc = 2D as the basic unit for calculating
local concentration. The corresponding variable n in Eq. (14) is
about 7.
On the contrary, the evaluation functions Wc based on the coor-
dination number fraction is computed from the microscopic rela-
tive positions and movements of particles themselves, which
coincides better with the feature of periodically movement of par-
ticles with the rotation of tumbler. Therefore, Wc does not have
inauthentic oscillations during the increasing variation of W,
Fig. 7. (a): Information entropy based on local concentration and coordination
whereas the oscillating variations of Wl are intrinsic features which
number fraction for x = 0.7 rad/s; (b): The Lacey index based on local concentration cannot be eliminated. Similar oscillations in mixing devices were
and coordination number fraction for x = 0.7 rad/s. observed by MAI et al.[19,23]. But this oscillation does not neces-
Z. Zhang et al. / Advanced Powder Technology 30 (2019) 2338–2351 2349

Fig. 9. (a) Lacey index based on local concentration Ml for x = 0.7 rad/s and a = 20°
at Lc = 1.5D, 2D, 3D, 4D, 5D, 10D respectively; (b) information entropy based on local
concentration fraction Sl for x = 0.7 rad/s a = 20° at Lc = 1.5D, 2D, 3D, 4D, 5D, 10D
respectively.

sarily mean a wrong evaluation of the homogenization index. It


also has some benefits to indicating the intrinsic features of the
tumbler. For example, it reflects to some extent the periodic pat-
tern of particle motion. Like in Fig. 6a, both Sl and Ml of a = 70°
exhibit periodic oscillations. It means that, at high inclination
angles, the particles are almost stationary relative to the tumbler,
which will be verified in Section 3.3.3 (Fig. 10).

3.3.3. Comparison of information entropy in non-rotating and rotating


grids
In the former section, the mixing indices based on local concen-
trations Wl (Sl and Ml) are oscillating due to the non-rotating grid.
However, such oscillations do not exist in the mixing index based
on the coordination scores Wc (Sc and Mc). The difference between
these two functions is only the choice of samples. The former sam-
ple is a computational unit divided by a non-rotating Cartesian
grid, and the latter sample is a computational unit divided by a
central particle and surrounding particles.
We hypothesized that the oscillation of the evaluation function
W based on local concentration is caused by cell division which is
fixed (Cartesian grid is non-rotating) while the particle mixing
interface is rotating following the cyclical movement of the tum-
Fig. 8. Mixing states at t = 10 s (a), 11 s (b), and 12 s (c) at x = 0.7 rad/s and a = 20°
corresponding to the falling range of Sl and Ml in Fig. 7a.
bler. Therefore, we calculate the information entropy based on
the local concentrations within rotating Cartesian grids to check
2350 Z. Zhang et al. / Advanced Powder Technology 30 (2019) 2338–2351

given case, the trends of evaluation functions computed following


the same manner (either based on the local concentration or the
coordination number fraction) should be consistent.
Therefore, we have come to the conclusion that the rotating grid
cannot completely eliminate the vibration of information entropy
based on the local concentration. It is because that the particle sys-
tem is affected not only by the force of the tumbler, but also by
gravity. Therefore, the grid division relative to drum static and
the grid division relative to ground static both cannot completely
eliminate the oscillation. In addition, the information entropy
based on the local concentration in the rotating grids is not a good
evaluation function to evaluate the degree of mixing.

4. Conclusions

For quantitative analysis, we proposed an improved Lacey


mixing index based on the coordination number fraction and per-
formed the comparative analyses of mixing evaluation functions
through following perspectives:

a) The Lacey mixing index and the information entropy are dif-
ferent mixing evaluation functions. It is indicated by the
evaluation functions that: in the rolling flow regime, the
rotational speed has a positive effect on the particle mixing.
The mixing speed increases with increasing of rotating
speed. The inclination angle has a negative effect on the par-
ticle mixing. With a certain speed, the rate of development
of mixing degree slows down with increasing inclination,
since the component of gravity toward the rear wall sup-
presses or degrades the development of mixing interface in
the axial direction. With low angles of tilt at a rotation speed
of x = 1.4 rad/s, the ultimate states of mixing are similar
since the mixing interface is totally degraded from three-
dimensional helical structure to two-dimensional circular
planar structure. Therefore, the final degrees of mixing at
Fig. 10. (a): information entropy based on local concentration in the rotating and
non-rotating grids for x = 0.7 rad/s; (b): information entropy based on local high tilt angles are far less than that at low tilt angles. The
concentration in rotating grids and the information entropy based on coordination new mixing index (the Lacey index based on coordination
number fraction in non-rotating grids for x = 0.7 rad/s. number fraction) has such excellent properties: ①it can
evaluate mixing of a system with three or more kind of par-
ticles, because it uses every particle as the core particle and
whether it can eliminate the oscillation of W. The center of rotating the mixing of each particle is considered; ②it is not affected
grids on the axis of the tumbler and the rotation speed is the same by meshing, and does not have unrealistic oscillation; ③it
as the tumbler rotation speed x, which ensures the simultaneous evaluates the particle system from the particle scale, allow-
movements between the rotating grids and the tumbler. ing for more accurate evaluation of particle-level systems.
Fig. 10 shows the information entropy based on the local con- b) mesoscopic and microscopic definitions: The mesoscopic
centration Sl in rotating and non-rotating grids (Fig. 10a) and the perspective refers to mixing evaluation function based on
information entropy based on the local concentration in rotating local concentrations, and microscopic perspective refers to
grids Sl and the information entropy based on the coordination that based on the coordination number fraction. Through
number fraction in non-rotating grids Sc (Fig. 10b) at x = 0.7 rad/ microscopic viewpoint, the evaluation functions based on
s. It can be found that Sl in rotating grids does not completely elim- the coordination number fraction do not oscillate with time.
inate the oscillations. Especially, the oscillation of information In contrast, the evaluation functions defined based on the
entropy based on the local concentration in the rotating grids for mesoscopic local concentration may vary with time in an
x = 0.7 rad/s and a = 70° is completely eliminated, because the oscillating pattern. The oscillation of Wl does not imply the
particle system is almost stationary relative to the sidewall of tum- real oscillation of mixing levels. It is preferred to consider
bler at high inclination angle. However, the oscillations of the other that the oscillation of Wl is caused by inconsistency between
two cases for x = 0.7 rad/s and a = 20° and 50° have not been elim- the rotation of mixing interface and the fixed non-rotating
inated, and the new oscillations have no obvious periodicity. This is grids. Although this oscillation is somewhat unreasonable
because the particle system is not completely stationary relative to compared to the mixing indices based on the coordination
the tumbler at low and middle inclination angles. The new oscilla- number fraction Wc (Sc and Mc) (Section 3.3.2, Figs. 6 and
tions are different from Sl in non-rotating grids (Fig. 9a). The ampli- 7), it reflects to some extent the periodic pattern of particle
tude of Sl in rotating grids is smaller than Sl in non-rotating grids motion in the tumbler. But, both of the mesoscopic Wl and
(Fig. 9a) and larger than the amplitude of Sc in non-rotating grids microscopic Wc evaluation functions can indicate the basi-
(Fig. 9b). The values and trends of either couple of (Sl in non- cally similar characteristics of mixing variation. Based on
rotating grids .vs. Sl in rotating grids) or (Sl in rotating grids .vs. the same scale, the values and trends of the information
Sc in non-rotating grids) are not similar. As aforementioned, for a entropy and the Lacey index are very similar.
Z. Zhang et al. / Advanced Powder Technology 30 (2019) 2338–2351 2351

c) Rotating and non-rotating grids. The oscillation of informa- [9] Y. Wen, M. Liu, B. Liu, et al., Comparative study on the characterization method
of particle mixing index using DEM method, Procedia Eng. 102 (2015) 1630–
tion entropy based on local concentration is caused by
1642.
non-rotating grids in computation which shows obvious [10] P.M.C. Lacey, Developments in the theory of particle mixing, J. Appl. Chem. 4
periodicity. In contrast, the information entropy based on (1954) 257–268.
local concentration computed in rotating grids simultane- [11] Z. Gu, J.J.J. Chen, A probabilistic analysis of some selected mixing indices,
Chem. Eng. Res. Des. 93 (2015) 293–303.
ously with the rotation of the tumbler does not show obvi- [12] A.N. Huang, H.P. Kuo, A study of the three-dimensional particle size
ous periodicities. Moreover, computed from rotating grids, segregation structure in a rotating drum, AIChE J. 58 (4) (2012) 1076–1083.
the information entropies based on the local concentration [13] Kapur JN., Kevasan HK, Entropy Optimization Principles with Applications.
1992
and the coordination number fraction do not show similar [14] O.O. Ayeni, C.L. Wu, J.B. Joshi, et al., A discrete element method study of
trends as that computed from the non-rotating grids. It is granular segregation in non-circular rotating drums, Powder Technol. 283
caused by the inconsistency between the paces of rotation (2015) 549–560.
[15] Z. Gu, J.J.J. Chen, An analysis of the entropy of mixing for granular materials,
of particle mixing and tumbler. Powder Technol. 266 (266) (2014) 90–95.
[16] N. Gui, J.R. Fan, K. Cen, A macroscopic and microscopic study of particle mixing
in a rotating tumbler, Chem. Eng. Sci. 65 (10) (2010) 3034–3041.
Acknowledgement [17] G.R. Chandratilleke, A.B. Yu, J. Bridgwater, et al., A particle-scale index in the
quantification of mixing of particles, AIChE J. 58 (4) (2012) 1099–1118.
The authors are grateful for the support of this research by the [18] S. Siiriä, J. Yliruusi, Determining a value for mixing: Mixing degree, Powder
Technol. 196 (3) (2009) 309–317.
National Natural Science Foundations of China (Grant No. [19] M.A.I. Schutyser, F.J. Weber, W.J. Briels, et al., Three-dimensional simulation of
51576211), the National High Technology Research and Develop- grain mixing in three different rotating drum designs for solid-state
ment Program of China (863) (2014AA052701), and the Foundation fermentation, Biotechnol. Bioeng. 79 (3) (2002) 284–294.
[20] N. Gui, J.R. Fan, J.S. Gao, X.T. Yang, Particle mixing study in rotating wavy wall
for the Author of National Excellent Doctoral Dissertation of P.R. tumblers by discrete element method simulation, Ind. Eng. Chem. Res. 53
China (FANEDD, Grant No. 201438). (2014) 13087–13097.
[21] A. Guida, A.W. Nienow, M. Barigou, Shannon entropy for local and global
description of mixing by Lagrangian particle tracking, Chem. Eng. Sci. 65 (10)
References (2010) 2865–2883.
[22] S. Masiuk, R. Rakoczy, The entropy criterion for the homogenisation process in
[1] M. Arntz, W.K. den Otter, W.J. Briels, et al., Granular mixing and segregation in a multi-ribbon blender, Chem. Eng. Proc.: Proc. Intensific. 45 (6) (2006) 500–
a horizontal rotating drum: a simulation study on the impact of rotational 506.
speed and fill level, AIChE J. 54 (12) (2008) 3133–3146. [23] M.A.I. Schutyser, J.T. Padding, F.J. Weber, et al., Discrete particle simulations
[2] T. Ward, W. Hourigan, Granular segregation in a tilted-rotating drum, Powder predicting mixing behavior of solid substrate particles in a rotating drum
Technol. 215 (2012) 227–234. fermenter, Biotechnol. Bioeng. 75 (6) (2001) 666–675.
[3] R. Chandratilleke, A.B. Yu, J. Bridgwater, et al., Flow and Mixing of Cohesive [24] F.S. Lai, L.T. Fan, Application of a discrete mixing model to the study of mixing
Particles in a Vertical Bladed Mixer, Ind. Eng. Chem. Res. 53 (2014) 4119–4130. of multicomponent solid particles, Ind. Eng. Chem. Proc. Des. Develop. 14 (4)
[4] G.J. Finnie, N.P. Kruyt, M. Ye, et al., Longitudinal and transverse mixing in (1975) 403–411.
rotary kilns: A discrete element method approach, Chem. Eng. Sci. 60 (15) [25] H. Kruggel-Emden, E. Simsek, S. Rickelt, et al., Review and extension of normal
(2005) 4083–4091. force models for the discrete element method, Powder Technol. 171 (3) (2007)
[5] R.K. Soni, R. Mohanty, S. Mohanty, et al., Numerical analysis of mixing of 157–173.
particles in drum mixers using DEM, Adv. Powder Technol. 27 (2) (2016) 531– [26] Wu. Hao, Nan Gui, Xingtuan Yang, Tu. Jiyuan, Shengyao Jiang, Effects of
540. particle size and region width on the mixing and dispersion of pebbles in two-
[6] S.H. Chou, H.J. Hu, S.S. Hsiau, Investigation of friction effect on granular region pebble bed, Granular Matter 18 (2016) 76.
dynamic behavior in a rotating drum, Adv. Powder Technol. 27 (5) (2016) [27] M. Alian, F. Ein-Mozaffari, S.R. Upreti, et al., Using Discrete element method to
1912–1921. analyze the mixing of the solid particles in a slant cone mixer, Chem. Eng. Res.
[7] T. Arndt, T. Siegmann-Hegerfeld, S.J. Fiedor, et al., Dynamics of granular band Des 93 (2015) 318–329.
formation: Long-term behavior in slurries, parameter space, and tilted [28] J. Mellmann, The transverse motion of solids in rotating cylinders—forms of
cylinders, Phys. Rev. E 71 (1) (2005) 011306. motion and transition behavior, Powder Technol. 118 (3) (2001) 251–270.
[8] W. Godlieb, N.G. Deen, J.A.M. Kuipers, A discrete particle simulation study of
solids mixing in a pressurized fluidized bed, International Conference on
Fluidization-new Horizons in Fluidization Engineering, 2007.

You might also like