Sie sind auf Seite 1von 17

Powder Technology 278 (2015) 94110

Contents lists available at ScienceDirect

Powder Technology

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

CFD prediction of scale-up effect on the hydrodynamic behaviors of a


pilot-plant uidized bed reactor and preliminary exploration of its
application for non-pelletizing polyethylene process
Yu Che a, Zhou Tian b,, Zhen Liu a, Rui Zhang c, Yuxin Gao c, Enguang Zou c, Sihan Wang c, Boping Liu a,
a
State Key Laboratory of Chemical Engineering, East China University of Science and Technology, Meilong Road 130, Shanghai 200237, PR China
b
Key Laboratory of Advanced Control and Optimization for Chemical Processes, Ministry of Education, East China University of Science and Technology, Shanghai 200237, PR China
c
Daqing Petrochemical Research Center, Petrochemical Research Institute of PetroChina, Chengxiang Road 2, Daqing City, Heilongjiang Province 163714, PR China

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

Article history: This work aims to explore the scale-up effect on the hydrodynamic natures in a pilot-plant gas phase ethylene
Received 11 July 2014 polymerization uidized bed reactor (FBR) based on two-dimensional (2D) transient Eulerian model integrating
Received in revised form 20 November 2014 the kinetic theory of granular ow (KTGF). Unlike lab scale FBR, signicant differences in the coreannulus
Accepted 14 February 2015
structures caused by the scale-up effect are revealed. The core region in pilot-plant scale FBR is much broader,
Available online 6 March 2015
and the distribution is nearly at for two phases. The annulus structure is thin, and exhibits a very small particle
Keywords:
velocity uctuation. The bed expanding section could reduce the solid velocity and improve the ow patterns of
Computational uid dynamic (CFD) gas and polymer particles. Simulations are also performed to assess the effects of polymer particle size and gas
Non-pelletizing polyethylene process (NPPP) velocity on the hydrodynamics for the non-pelletizing polyethylene process. With the increase of polymer
Pilot-plant uidized bed reactor particle diameter (446 m to 1338 m), the supercial gas velocity should be increased from 0.60 m/s to
Coreannulus structure 0.90 m/s to achieve better steady uidized state. The results are helpful for understanding how the local mean
Scale-up effect two phases' velocities and volume fractions vary in pilot-plant reactor for PE production caused by scale-up
effect.
2015 Elsevier B.V. All rights reserved.

1. Introduction phases as interpenetrating continua with the kinetic theory of


granular ow (KTGF) needs fairly less computational cost [8,
Polyethylene (PE) is one of the most widespread resins for its versa- 1214]. Based on the EulerianEulerian model, Dudukovic compre-
tile physical and chemical properties, and can be produced by various hensively reviewed the CFD models validated by experimental data
technologies [1,2]. Gas phase uidized bed reactor (FBR) is commonly [15], and Taghipour et al. compared the effect of drag models
used for polyolen production due to its simple construction, excellent (Gidaspow, WenYu and SyamlalO'Brien) on the ow patterns
solid mixing and effective heat and mass transfer [3]. An efcient and hydrodynamics of two phases in a lab scale FBR [16].
reacting gas and solid particle ow and mixing are of prime importance Considerable attention has been devoted to modeling and simu-
for FBR operation [4]. Yet, the two phases' ow behaviors exhibit intri- lating the hydrodynamic characteristics of ethylene polymerization
cate hydrodynamic natures at different scales for the interactions FBR. Most of the studies mainly focused on the prediction of gas
between the gas and solid phases via viscous drag or particle collisions. and solid holdups, bubble behaviors, etc. in lab scale FBRs [8,
A better understanding of the complicated ow behaviors is necessary 1723]. The discussions concerning hydrodynamic parameters in
for improving the reactor performance. pilot-plant FBRs for PE production are relatively scarce. Fan et al.
Two different categories of computational uid dynamics (CFD) used the EulerianEulerian model to predict the hot spots in a
model are often applied to simulate the gassolid ow in FBRs, pilot-plant ethylene polymerization FBR [24,25]. More recently,
namely the EulerianLagrangian model and the EulerianEulerian Rokkam et al. developed a CFD model incorporating the effect of
model [3,510]. The Lagrangian model solves the equation of each electrostatics on particle dynamics, and it was applied to investigate
particle motion, taking into account the particleparticle collisions the particles segregation in pilot-plant FBR [26]. However, the effect
and forces. Because of high numerical effort to calculate individual of scale-up on the hydrodynamic behaviors in pilot-plant beds has
particles, the model is limited to lab scale FBRs [5,9,11]. On the not been concerned exclusively. Furthermore, the restitution coef-
contrary, the EulerianEulerian model dealing the gas and solid cients, specularity coefcient, and interphase drag coefcients for
modeling of pilot-plant FBRs should also be veried in detail.
Corresponding authors. Tel.: +86 21 64253627. Non-pelletizing polyethylene process (NPPP) has acquired rapid
E-mail addresses: tianzhou@ecust.edu.cn (Z. Tian), boping@ecust.edu.cn (B. Liu). development and wide application. Terano et al. developed a new

http://dx.doi.org/10.1016/j.powtec.2015.02.022
0032-5910/ 2015 Elsevier B.V. All rights reserved.
Y. Che et al. / Powder Technology 278 (2015) 94110 95

highly active catalyst with larger particle size and in-situ stabilization a single inlet at the bottom of the plenum, and dispersed uniformly
technologies to omit the pelletizing step in polypropylene production through the gas distributor. The pressure gauges and thermometric
[27]. Spherilene technology is another example, and it is a gas-phase instruments are all connected to the Distributed Control Systems (DCS).
process for full density range of PE production [28]. The polymer particle Table 1 presents the preferred process operating conditions. The
size is a critical factor that affects the ow and mixing for NPPP. operating parameters are also selected in simulations to investigate
However, there are no open literatures regarding the applications of the effects of scaling-up and applying NPPP on the hydrodynamic
NPPP to pilot-plant FBR, especially for the hydrodynamic natures. behaviors.
The main objective of this work is to provide a deep insight into In ethylene polymerization, each catalyst particle will eventually
predicting the effect of scale-up on the hydrodynamic characteristics become a single polymer particle that acts as a polymerization micro-
of the gas and PE particles in pilot-plant FBR, and to anticipate how reactor based on the Reactor Granule Technology (RGT) [29]. There
different parameters respond to the changes in different operating are two types of applications for RGT: development of new polymers,
conditions for applications. An EulerianEulerian model combining and economic polyolen processes [29,30]. For the latter, it aims to
KTGF was applied to describe the two phases' ow behaviors, and vali- reduce energy consumption and waste handling [28,3032]. It is
dated by industrial data. The effects of scale-up and the bed expanding known that the pelletizing step could consume more than 40% of total
section were discussed in terms of uidization features, volume fraction production energy [27]. Fig. 2 shows the differences between TPPP
proles, and axial solid velocity. Then, the impacts of particle size and (the traditional pelletizing polyethylene process) and NPPP, and both
operating velocity on the reactor performance for NPPP were also the pelletizing and compounding steps have been skipped. Two crucial
evaluated. technologies are required for NPPP: the large particle size catalyst
technology, and in-reactor stabilization technology [27]. Plenty of
2. Pilot-plant reactor description experimental efforts have been paid to investigation the in-reactor
stabilization technology [27,33]. At present, the study of large particle
In this work, a pilot-plant ethylene polymerization FBR in a Chinese size catalyst related technology is only limited to the development of
chemical plant shown in Fig. 1a is studied. The FBR fabricated from catalysts [34]. There have been few open reports on exploring the effects
special stainless steels consists of three zones given in Fig. 1b, a uidized of large catalyst and polymer particle on the ow eld and operating
zone, a transition zone and a disengagement zone. The whole bed is performance in reactor for NPPP.
4.347 m in height, and the uidized zone and disengagement zone are
0.7 m and 1.4 m in diameter, respectively. A stainless steel perforated 3. Mathematical modeling
gas distributor containing sixty-one 10-mm-diameter holes is installed
between the inlet plenum and bed chamber. The holes are approxi- 3.1. EulerianEulerian two-uid model equations
mately equally spaced in triangular manner to ensure evenly gas distri-
bution and low-pressure drop. The open area ratio of the gas distributor The CFD model used in this study is based on the EulerianEulerian
is 2.1%. The reacting gas components are introduced into the system by two-uid model. The hydrodynamic model uses the conservation of

Fig. 1. Sketch of the pilot-plant uidized bed reactor (a. the reactor photo; b. the equipment drawing; c. the computational domain).
96 Y. Che et al. / Powder Technology 278 (2015) 94110

Table 1 3.2. Kinetic theory of granular ow


The pilot-plant FBR operating conditions in the ethylene polymeriza-
tion plant.
Closure of the solid phase momentum equation requires a descrip-
Description Value tion for the solid phase stress. When the particle motion is dominated
Operating pressure 2 106 Pa by collisional interactions, concepts from the gas kinetic theory can be
Operating temperature 361 K used to describe the effective stresses in the solid phase resulting from
Inlet gas velocity 0.61 m/s particles streaming (kinetic contribution) and direct collisions (colli-
Initial bed height 1.137 m
sional contribution) [36]. The kinetic theory of granular ow (KTGF) is
Bulk density 320 kg/m3
proposed based on similarities between the ow of granular material
and gas molecules. It is possible to describe the behavior of molecules
mass and momentum of each phase, and is represented by correspond- or particles with well-dened properties and interactions using KTGF.
ing conservation equations. The interaction force between two phases is It introduces a granular temperature (), which represents the uctua-
expressed as additional source terms added to the conservation tion of the particle velocity. Analogous to the thermodynamic tempera-
equations which can be written as [35], ture for gases, the granular temperature of solid phase is dened as one
third of the mean square velocity of particle's random motion [17],
Continuity equations
1 02
  
!
 v 7
g g  g g v g 0 1 3 s
t
where vs is the uctuating particle velocity.
 
! The granular temperature is proportional to the granular energy. The
s s  s s v s 0 2
t solid phase granular energy equation must be solved to calculate the
granular temperature. Since the solid phase stress depends on the mag-
Momentum equations nitude of these particle velocity uctuations, a balance of the granular
 
energy 32 associated with these particle velocity uctuations is
 !
 
!!

v  g g v g v g required to supplement the continuity and momentum conservation
t g g g  
! ! for both phases. This balance equation is given as [37],
g p  g K gs v s v g g g g 3
    
3 ! !
s s  s s v s ps I s v s
 !
 
!!

2 t
s s v s  s s v s v s s pps  s
t    ks s gs 8
! !
K gs v g v s s s g 4
where the rst term on the right side represents the generation of uc-
tuating energy due to shear in the solid phase, the second term is the
The gas and solid phases' stress tensors are given by, diffusion of uctuating energy along gradients in , s denotes the
dissipation for inelastic particle and particle collisions, and gs accounts
  2   for the dissipation or creation of granular energy resulting from the
! !T !
g g g v g v g g g   v g I 5 working of the uctuating force exerted by the gas through the uctuat-
3
ing velocity of the particles [36].
      The diffusion coefcient for granular energy (ks), is given by Syamlal
! !T 2 !
s s s v s v s s s s   v s I 6 et al. [38],
3
p 
15s ds s 12 2 16
ks 1 43 s g 0 4133 s g 0 9
where s is the solid shear viscosity and s is the solid bulk viscosity. 4133 5 15

Fig. 2. Schematic diagram of TPPP and NPPP for PE production.


Y. Che et al. / Powder Technology 278 (2015) 94110 97

where for the investigated system for its solid volume fraction limitation
(b 20%). Thus, the drag correlations of Gidaspow and SyamlalO'Brien
1=21 es 10 are investigated in this study.

The collision dissipation of energy (s), is modeled using the correla- 3.4.1. Gidaspow model
tion by Lun et al. [39], Gidaspow model combines Wen and Yu model (1966) for gas
  volume fraction larger than 0.8 with the Ergun equation (1952) for
12 1e2s g 0 2 1:5 gas volume fraction lower than 0.8, and cover the whole range of void
s p s s 11
ds fraction,

s g ! !
2
gs 3K gs 12 Ergun s g
K gs 150 1:75 v s v g ; g b 0:8 21
g d2s ds
Rather than solving the complete granular energy balance given in
the equation, some researchers assume that the granular energy is in a WenYu 3 s g g ! ! 2:65
K gs C v s v g g ; g b 0:8 22
steady state and dissipated locally, and neglect the convection and diffu- 4 D ds
sion terms. Thus, the equation can simplify to an algebraic expression,
  where
!
0 psI s v s s 13 8   0:687
<
24
1 0:15 g Res ; Res b1000
C D g Res 23
:
0:44; Res N1000
3.3. Solid phase stress tensor
g ds ! !
Constitutive relations for the solid phase stress based on the kinetic Res v s v g 24
g
theory concepts have been derived by Lun et al. [39]. The solids pressure
represents the normal solid phase forces for particle and particle inter-
To resolve the inherent discontinuity of the model, Gidaspow
actions, and it can be calculated by,
considered a switch function for rapid transition from one regime to
2 the other [42],
ps s s 2s g 0 s 1 es 14
1
tan 150  1:750:2 s 
The solids bulk viscosity describes the resistance of the particle gs 0:5 25

suspension against compression, and it can be written as [39],
r Accordingly, the following model was proposed,
4 2
s s s ds g0 1 es 15  
3 Ergun WenYu
K gs 1gs K gs gs K gs 26
where

1 3.4.2. Syamlal and O'Brien model


g0 v
! 16
u
u s
1 t
3

s;max
! !
3 C D g v s v g
K gs g s 27
4 v2r;s ds
The models for concerning the solid phase dynamic viscosity in this
work are listed as follows [40,41], 0 12

s s;col s;kin s; f r 17 B 4:8 C


C D @0:63 qA 28
Res
vr;s
where

r ! !
g ds v s v g
4 Res 29
s;col s s ds g 0 1 es 18 g
5
p  2
10ds s 4 The model applied an analytical formula for the terminal velocity
s;kin 1 1 es s g 0 19
96 s 1 es g 0 5 correlation for solid phase (vr,s) from the velocityvoidage correlation
proposed by Garside and Aldibouni as follows,
p sin 
s; f r sp 20 q
2 I2D vr;s 0:5 A0:06Res 0:06Res 2 0:12Res 2BA A2 30

3.4. Drag models where

4:14
A g 31
In the EulerianEulerian model, the transfer of forces between the
 
! !
gas and particle phases is done by the drag forceK gs v g v s . Different 8
>
<
C
g 1 ; g 0:85
drag models are reported for calculating the drag coefcient Kgs, and the
most widely used drag models in gassolid FBRs are Gidaspow model, B 1:28
C 2 g ;
g b0:85 32
>
:
SyamlalO'Brien model and WenYu model. WenYu model is not t C 1 2:65 and C 2 0:8
98 Y. Che et al. / Powder Technology 278 (2015) 94110

3.5. Turbulence model top of cylindrical riser. The 2D physical models and their meshes are
both constructed in Gambit 2.4.6 (Ansys Inc., USA), and the computa-
Disperse phase RNG k turbulence model is used to solve the trans- tional domain used in the simulation is shown in Fig. 1c. Mesh indepen-
port equations for k and . It is based on the Tchen theory of dispersion dence tests had been performed, and the simulation results showed that
of discrete particles by homogeneous turbulence [43]. The turbulence increasing the cell numbers from 237,500 to 314,000 resulted in less
equations are solved for the gas phase, and the turbulence parameters obvious changes in solid volume fraction proles. Thus, 237,500 quadri-
for the solid phase are deducted using dispersion rules. According to lateral cells could provide mesh independent results.
Hartge et al. [44], the disperse phase RNG k turbulence model
shows better agreement with the experimental data than per-phase 4.1. Simulation parameters
turbulence model in the FBRs modeling.
The transport equation associated to the turbulence parameters k 4.1.1. Physical properties of gas and solid phases
and are shown as follows [35,44], Because the growth rate of PE particles is very slow, the particle size
mainly depends on the residence time in the reactor [12]. Thus, when
 
m k  m vm k  k t;m k Gk;m m 33 the stable state situation is achieved, the integration of the gas and
t solid ow system with ethylene polymerization in the FBR is simulated.
  The properties of the gas and solids used in the simulations at any
particular polymerization time are listed in Tables 2 and 3.
 m vm  t;m
t m
 
C 1 Gk;m C 2 m R 34 4.1.2. Boundary conditions and model parameters
k
The minimum uidization velocity (Umf) and particle terminal
where the mixture properties are described as, velocity (UT) can be estimated by using Wen and Yu's equations [12,52],

m g g s s 35 82   30:5 9
> >
g < ds 3 g s g g =
Um f 433:7 0:0434
2 5 33:7 43
g g vg s s vs ds g > g 2 >
vm 36 : ;
g g s s

2   30:5
The eddy viscosity for the mixture phase is calculated as, 3:1 s g gds
UT 4 5 44
k2 g
t;m m C 37

ds g U T
The production of the turbulent kinetic energy (Gk,m) is calculated Re 45
using the following relation, g

h i
T
Gk;m t;m vm vm : vm 38 The reactor must be operated between the values of Umf and UT, and
the supercial gas velocity can vary from 3 to 7 times of Umf [53].
The terms k and are the inverse effective Prandtl numbers, which According to the industrial operating conditions, the incoming gas
are obtained by, velocity in rst simulation part is 0.61 m/s. When NPPP is applied, the
gas velocity will be set to 57 times of Umf to optimize the operating
k 39 conditions. Initially, the solid particle velocity is set as zero, and the
gas inlet velocity is assumed to have the same value everywhere in
k 1:3929 0:6321 k 2:3929 0:3679 the bed cross section. Table 4 lists the values of model parameters and
m 40
1:3929 2:3929 t;m related computational conditions that applied in the simulations.
0 0

0 1 41 4.2. Solution procedure

C 1 1:42 and C 2 1:68 42 EulerianEulerian two-uid model has been used to simulate and
investigate the ow characteristics in the FBR. The RNG k model is
applied to take into account the turbulence, meanwhile, KTGF is used
4. Simulation scheme to close the momentum balance equation for the solid phase. The simu-
lation is implemented by the commercial CFD software Fluent 14.0
In this work, a two dimensional physical model of the reactor system (Ansys Inc., USA), in double precision mode. The phase-coupled
must be available to study the pilot-plant FBR. Although Reuge et al. [45] SIMPLE algorithm is used to couple the pressure and velocity. The
point out the differences between 2D and 3D simulated void fractions, rst-order upwind is applied as the spatial discretization scheme for
2D model is still recommended to reduce the cost of calculation while its better convergence and less computational effort [40]. For transient
maintaining accuracy [40,46]. In addition, 2D simulation has always
been applied for much cheaper numerical cost and less computational
time [4750]. Herein, a Cartesian coordinate system is adopted for the Table 2
Gas composition and solid size distribution.
2D mesh generation, and applied to simulate and predict the hydrody-
namic natures. The comparisons between 2D and 3D simulations have Description Items
been performed in this work. As mentioned [25,26], gas distributor Gas Ethylene 1-Butylene Hydrogen Nitrogen
can be excluded in the computational region for the complicated struc- Mass fraction 0.35 0.14 0.065 0.445
ture and enormous computation cost, and the ow parameters can be Solid particle size 74 99 165 250 590 830
accurately simulated compared with experimental data [40]. Therefore, (m)
Percentage 2.57 5.26 5.03 48.59 38.09 0.46
the simulated domain is selected between the gas distributor and the
Y. Che et al. / Powder Technology 278 (2015) 94110 99

Table 3 pressure drop, one of the few data directly measured from the plant, it
Properties of the gas and solid phases. was found that the results were almost insensitive to most of model
Description Value Unit Comment parameters, except the drag models. Hence, the only results from the
Gas density, g 20.2 kg/m3 Experimental measurement
drag model investigation are demonstrated below.
Gas viscosity, ug 1.72 105 Pas Experimental measurement
Solids density, s 918 kg/m3 Experimental measurement 5. Results and discussion
Mean particle 446, 892a, 1338b m Calculated [51]
diameter, ds
a
5.1. Scale-up effect on the hydrodynamics of gas and solid particles
Data from considering the effect of large particle size for NPPP (446 2).
b
Data form considering the effect of large particle size for NPPP (446 3).
5.1.1. Instantaneous solid volume fraction
In the pilot-plant FBR, hydrodynamic features are related to the
presence of smaller bubbles, which are mainly induced by the scaling-
formulation, the rst order implicit scheme is used as it is usually more up effect. Fig. 3 illustrates the bubble formation, development and
stable for solving stiff equations [10]. breakup with representation of the instantaneous solid volume fraction
The simulations are carried out with 2D Cartesian frameworks, and in a vertical plane versus the ow time at two types of drag models. In
the computational domain was discretized by 237,500 quadrilateral contrast to lab scale FBRs [16], the observed axisymmetry gives way to
cells. In the radial direction of each parts shown in Fig. 1c except the chaotic transient generation of bubble formation after only 1 s in the
bed expanding section where the mesh spacing is increasing to stress initial uidization stages. But for the lab scale FBR, the axisymmetric
the structure effect, the mesh spacing was distributed uniformly. ow pattern changes more than 2 s due to the relative simple ow
However, non-uniform mesh spacing was used in the axial direction structure. Then promptly, the bubbles start to coalesce when they
for capturing the complex ow behavior in this large region. The size move upward generating bigger bubbles. Finally, they become
of the cells used in the CFD simulations for different parts of the bed stretched, deformed and broken because of the effects between the
are summarized in Table 5. bed and the wall, and the interactions with other bubbles. In addition,
In this work, the convergence criterion is set as 1 103 according the visual results indicate that the bubbles formed at the bottom of
to literatures [40,41,58], and the time step is set from 1 10 4 s to the reactor are relatively small and concentrated, and they grow up
1 10 3 s to ensure the numerical stability. All the simulations are slowly until they rise up to the top surface with coalescence. Because
performed for a period of 60 s of real simulation time. The time- of the existence of bed expanding section, the rising granules entrained
averaged parametric values are taken from 10 s (by which a dynamic in the bubble wake drop back to the wall and collide with other granules
steady state is achieved) to 60 s of the simulation results. The CFD or bubbles, and it leads to relative uniform distribution of solid volume
simulations of the pilot-plant FBR are executed using a Supermicro fraction in the bed.
server platform with two Intel Xeon E5-2600 series CPUs and 32 GB of By comparison of Fig. 3a and b, it is found that the bed expansion
RAM on the High Performance Computing Cluster at East China Univer- heights at two drag models increase equally before 1 s, and the generat-
sity of Science and Technology. ed bubbles can accelerate the motions of other bubbles. Therefore, the
To determine the sensitivity of the simulation data to the parame- total uidized velocity is fairly fast, and it gives rise to stable ow eld
ters, a large series of calculations were carried out with the related in the reactor after a very short ow time [12]. Initially, the bed expan-
model parameters for the mean solid particle size. Compared with the sion height continuously increases until it levels off at a steady-state bed
height after 5 s, and the time-averaged simulations are taken from 10 s
to 60 s.
Fig. 3 also presents obvious differences between Gidaspow model
Table 4
and SyamlalO'Brien model to predict the bed expansion height and
Model parameters and computational conditions used in the simulations.
solid volume fraction for t = 60 s. For the solid volume fraction
Description Value Comment contours, it can be easily seen that Gidaspow model shows a relatively
Turbulence model k (RNG, Shi et al. [12] stable bed surface and uniform solid volume fraction, and it also holds
dispersed) a smooth and steady bed expansion height.
Granular viscosity Gidaspow Gidaspow [42]
Granular bulk viscosity Lun Lun et al. [54]
Frictional viscosity Schaeffer Shi et al. [12] 5.1.2. Mean volume fraction of gas and solid
Angle of internal friction 30 Fixed value The variations of mean gas and solid volume fractions as a function
Granular temperature Algebraic Johnson et al. of the axial and radial positions in the FBR can be represented by
[55]
the contour plots shown in Fig. 4. Fig. 4a and b depicts the comparison
Drag law model Gidaspow Vejahati et al.
[56] of time-averaged gas volume fraction in the radial direction at different
Coefcient of restitution for 0.9 Fixed value height of reaction zone with different drag models. The coreannulus
particleparticle collisions ow structure is apparently predicted by Gidaspow and Syamlal
Solid pressure Lun et al. Lun et al. [54]
O'Brien models. It consists of a central region in which the polymer
Radial distribution Lun et al. Lun et al. [54]
Inlet boundary condition Velocity Fan [24]
Outlet boundary condition Pressure outlet Fan [24]
Wall boundary condition No slip for gas Fan [24]
phase Table 5
Free slip for solid Mesh sizes used for the 2D CFD simulations.
phase
Parts Max interval size Min interval size Total nodes
Initial bed height 1.137 m Calculated value
[57] a 3.5 3.5 201
Initial volume fraction of the solid phase 0.63 Calculated value b 3.5 3.5 451
[57] c Radial 6.46 4.94 301
Outlet pressure 1.01325 105 Pa Wei et al. [58] Axial 1.33 1.02 201
Maximum iterations 40 Specied d 14.11 8.49 61
Convergence criteria 1 103 Specied e 7 7 201
Time step 1 103 s Specied The FBR 237,500
100 Y. Che et al. / Powder Technology 278 (2015) 94110

Fig. 3. Instantaneous volume fraction proles of the PE particles as a function of time at different drag models (Ug = 0.61 m/s, ds = 446 m).

particles travel upwards by the reacting gas and an annular area where indeed have changed [49]. The structure is distinct and orderly in the
the particles fall downwards along the wall. case of Gidaspow model, and the changes in different height of the reac-
Compared with lab scale FBR [59], due to the effects of scale-up and tion zone from the wall to the core region seem to be physically reason-
bed expanding section, the coreannulus ow structures and shapes able. The solid volume fraction decreases from the wall to the center,
present apparent differences. Firstly, the core region in large scale FBR and then remains horizontal in the core region. Yet, there is still an
is rather broad and distinct, and the gas volume fraction distribution is obvious change at the height of 2 m; a more likely reason is that the
almost at not concave. Secondly, in the wall area, the gas holdup has ow pattern has been clearly changed because of the impact of the
an apparent change rather than a small peak. To demonstrate quantita- bed expanding section on the solid velocity. In contrast, the case of
tively, the ratio of core region and annulus area has been used to repre- applying SyamlalO'Brien model seems to be disorderly and unsystem-
sent the scale up effect. From the lab scale to the pilot-plant scale FBR, atic. Compared with the coreannulus pattern of gas and solids ow in
the ratio changes from 1/1 to 5/2, which means that the coreannulus lab scale reactor [35,49], by reason of the scale-up effect, it is shown a
ow structures have undergone a through transformation. The possible thin dense annular region in the wall, and relative thick dense core
reason is that the gas and solid particles could ow and mix uniformly region.
with intensive interactions of numerous small bubbles in large scale
FBR. 5.1.3. Mean velocity of solid
In comparison, the proles of mean gas volume fractions predicted Fig. 5a and b illustrates the comparisons of time-averaged axial solid
by Gidaspow model are relatively at and stable throughout the radial velocity proles along radial direction at different height using two drag
direction at each height of the reaction zone in the core region. On the models. The velocity varies from negative at wall to positive at central
contrary, the proles calculated by SyamlalO'Brien model are almost region, and two regions are easy to be distinguished in the gure. The
chaotic, especially at the height of 2 m where the bed expanding section coreannulus ow structures are clearly predicted by Gidaspow
is equipped. One can easily conjecture that the ow proles and model and SyamlalO'Brien model. Compared with lab scale FBR [35],
patterns have clearly changed, and it can be visually depicted in Fig. 3. the coreannulus ow structures and shapes have distinct changes for
The mean solid volume fractions with different drag models are the effects of scaling-up and bed expanding section. Firstly, the core
plotted in Fig. 4c and d, and it is shown the opposite behavior as the region in large scale FBR is rather thick, and the axial solid velocity has
gas holdup. The coreannulus structure of the ow is also existed, and a smooth and stable variation. Secondly, in the wall region, the annulus
it is clearly predicted by both drag models. Just as mentioned before, structure is rather thin and a very small axial velocity change can be
the ratio of core region and annulus area in pilot-plant FBR also shows clearly observed. In order to present quantitatively, the ratio of core
an obvious difference compared to lab scale bed, the value changes region and annulus area can still be used to represent the scale up effect.
from 5/2 to 1/1, which conrms that the coreannulus ow structures For the lab scale FBR, the ratio is 1/1, instead, in the pilot-plant scale bed,
Y. Che et al. / Powder Technology 278 (2015) 94110 101

Fig. 4. Time-averaged gas and solid volume fraction variation along the radial direction at different height of reaction zone (t = 60 s).

the ratio changes to a relative large value of 5/2, which means that the drop in FBR is a signicant parameter in process scaling-up, controlling
ow structures inside the large scale reactor have changed thoroughly and operating. Fig. 6 shows the difference of pressure drop in the bed
compared to lab scale reactor. height direction between the simulated results and experimental data.
The preferable reason is that the reacting gas and polymer particles It is shown that the pressure drop distribution is nearly linear in the
fully undergo ow and mixing for the interactions of bubbles and bed axial height. The simulated result is in good agreement with the
particles. However, the results show obvious difference between measured pressure drop data. The errors incurred in the simulation
(a) and (b) in the gure. The axial velocity distribution in the case of derive from several sources, such as ow patterns, numerical treatment
SyamlalO'Brien model is relative disorderly, especially near the wall of initial and boundary conditions.
area, and the value of velocity has a large change at each height of The velocity distribution proles of the gas and solid phases in the
reaction zone. It will lead to a bad ow and mixing behaviors of two reactor play a vital role in the overall pressure drop, and they are
phases. closely related to the existing ow patterns and behaviors in FBRs.
The mean solid velocity proles demonstrated in Fig. 5c and d show Table 6 shows the overall pressure drop in the pilot-plant reactor
a very similar situation. Due to the existence of bed expanding section, at different operation conditions, and it will be discussed in detail
signicant differences are depicted in the height of 2 m of reaction below.
zone between (c) and (d) in the gure. It indicates that Gidaspow
model shows outstanding performance in modeling and predicting
the hydrodynamic characteristics of the pilot-plant FBR. 5.2.2. Solid velocity vector
From the comparisons of the instantaneous solid volume fraction, In Fig. 7, vector plots of time-averaged solid velocity and contour of
mean gas and solid holdups and solid velocity, it can be proven that solid velocity are shown together at an initial ow velocity of 0.61 m/s.
the CFD model applying Gidaspow drag model is feasible and reliable The circular upow in the bed is apparently depicted as a result of the
to investigate the hydrodynamic characteristics caused by the scale-up enormous bubbles and polymer particles motions. The granules are
effect in the pilot-plant bed. Therefore, in the following section, lifted up from the middle position and come back along with the break-
Gidaspow drag model is chosen as drag model in the CFD model. up of the bubbles due to gravitation and effect of bed expanding section.
Visibly, in the core region, the gases carry upwards granules along
5.2. Model evaluation the vertical axis; meanwhile, the granules can ow down along the
wall. It can lead to the formation of bigger circular upows. At the
5.2.1. Comparison of the simulated and the plant data bottom of the bed, because of the intensive interactions of gas and
Owing to the difculty of data measurement in industrial plant, the polymer particles, numerous small circulations are easily formed. The
pressure drop distribution is the only reliable and favorable approach vortexes appearing among the circulations have a great impact on the
for the model verication and evaluation. It is known that the pressure hydrodynamics.
102 Y. Che et al. / Powder Technology 278 (2015) 94110

Fig. 5. Time-averaged axial velocity of solid (a, b) and mean solid velocity (c, d) variation along the radial direction at different height of reaction zone (t = 60 s).

Compared to the lab scale FBR [12], the action of the circulations and 5.2.3. Comparison of monodisperse and polydisperse cases for 2D and 3D
the motion of vortexes lead to a good ow and mixing in the pilot-plant simulations
reactor, and it is very helpful in providing more uniform temperature Fig. 8 demonstrates the comparisons of time-averaged particles
and concentration distributions in PE production. In addition, the bed volume fraction in the bed height direction for the cases (2D monodis-
expanding section in the pilot-plant FBR is an essential part to operate perse, 2D polydisperse, 3D monodisperse and 3D polydisperse). As
and run the reactor properly and easily. The wall of bed expanding shown in the gure, the coreannulus ow structure can be clearly
section can clearly reduce the solid velocity and collect the polymer observed with different values of mean solid holdups in the core area
particles, and the gure clearly shows the granules motion state and and wall region. A more detailed description of the ow structure in
velocity magnitude. the FBR is provided with the axial and radial proles of solids volume
fraction in Fig. 9.
In Fig. 9a, the quantitative comparisons of the mean solid volume
fraction along the bed height direction are presented from 2D to 3D
simulations for monodisperse and polydisperse cases. Both 2D and 3D
simulated results demonstrate very good solid holdup distribution
behaviors. However, the proles indicate that PE particles near the
inlet region of the bed in 2D and 3D monodisperse simulations show
clear oscillation created due to the particles inlet effect. Instead, the
2D and 3D polydisperse simulations predict slightly different and
uniform proles for their accurate capture of the quantitative ow

Table 6
The pilot-plant FBR overall pressure drop at different operation conditions [12,60].

Description Items Pressure drop (Pa)

0.61 m/s 446 m 6585


892 m 6622
1338 m 6660
1338 m 0.75 m/s 11,503
0.90 m/s 12,637
Fig. 6. Comparison of time-averaged pressure drop versus FBR bed height between simu-
1.05 m/s 14,219
lated and experimental data at a gas velocity of 0.61 m/s.
Y. Che et al. / Powder Technology 278 (2015) 94110 103

Fig. 7. Time-averaged solid velocity vector and velocity magnitude distribution inside the computational domain.

characteristics in the FBR. The trends of distribution are nearly main- It should be noted that the main objective of this work is to explore
tained, in the middle-upper region of the reactor, all the cases demon- the scale up effect on the coreannulus ow structures and hydrody-
strate relatively uniform distribution proles. Fig. 9b shows the mean namic features between the large and lab scale FBRs. The above discus-
solid volume fractions along the radial direction in the bed height of sions show that the 2D monodisperse is acceptable and reliable to
1 m. The coreannulus ow structures are clearly presented for all predict the effect of scale up in the pilot-plant scale reactor for
cases with the nearly identical thickness of each sections (0.5 m for its lower numerical cost and less computational time. It must be
core region, 0.2 m for annulus area), which means that the effect of mentioned that the 2D and 3D polydisperse cases are being performed
scale up on the coreannulus ow structure can be investigated by to study the effects of ethylene polymerization and particle kinetics
any of the cases. (particle growth, aggregation and breakage) on PE particle behaviors

Fig. 8. The visual representations of mean solid volume fraction along the bed height direction in a (2D monodisperse case), b (2D polydisperse case), c (3D monodisperse case) and d (3D
polydisperse case).
104 Y. Che et al. / Powder Technology 278 (2015) 94110

Fig. 9. The mean volume fractions of PE particles along the axial (a) and radial (b) direction in A (2D monodisperse case), B (2D polydisperse case), C (3D monodisperse case) and D
(3D polydisperse case).

and to evaluate the operating performance of the large FBR under TPPP The solid holdups distribution trends predicted in a(2), b(2) and
and NPPP in our research group now. c(2) of Fig. 11 are very similar, but the distribution varies clearly as a
result of the effect of polymer particle size on the ow pattern and struc-
5.3. Model application to NPPP ture. For the radial proles, the distribution behaviors are almost identi-
cal at the center area, and signicant differences are easily found near
5.3.1. Effect of the polymer particle size the wall region in the reaction zone. These different phenomena could
Fig. 10 illustrates the solid volume fraction proles for different PE be explained by the ow behaviors described in Fig. 10, and the changes
particle sizes (446 m, 892 m and 1338 m) under the gas velocity of of overall pressure drop under different particle sizes shown in Table 6.
0.61 m/s. During the uidization process, an obvious increase of solid
holdups can be observed with the increase of polymer particle size, 5.3.2. Effect of operation conditions
and then decreases with the uidization proceeding. The bubble In this section, the hydrodynamic features under NPPP will be
behaviors are also depicted in the gure, it is shown that with the explored, preliminarily. When the polymer particle size increases to
increase of polymer particle size, the number of bubbles becomes less 1338 m, the gas velocity can vary from 3 to 7 times of Umf [53]. Three
and the size gets larger. That may cause abnormal operation of the different gas inlet velocities of 0.75 m/s (= 5Umf), 0.90 m/s (= 6Umf)
bed. Correspondingly, the bed expansion ratio gets smaller. It is indicat- and 1.05 m/s (= 7Umf) are selected in the simulations, and the corre-
ed that the gas velocity (0.61 m/s) is only applicable for describing a sponding simulated results are shown in Figs. 12 and 13.
specic range of PE particle size (446 m). Moreover, it takes less time Fig. 12a, b and c depicts the solid volume fraction proles at different
to arrive at the steady uidization state for smaller particles. Once the operating gas velocities. As indicated in the gure, it is clear that an
particle size becomes larger by applying NPPP, the uidization and escalation of gas velocity leads to rising trend of the bed expansion
operation state will get worse, and it is extremely difcult to optimize height and increase of the bubble size. The bubbles grow as they rise
and control of the reactor. According to the correlation of Gidaspow to the top surface with coalescence, and the elongation of the bubbles
drag model, the particle size shows a dominant factor to affect the is induced by the bed wall effects and interactions with other bubbles
drag force. For the given gas velocity, with the increase of the polymer [16]. The most likely reason is that with the increase of the gas inlet
particle size, the drag force acting on the solid particles decreases. velocity, the particle movement in the bed is sharply strengthened. It
Conversely, it is easier to generate bubbles, the smaller particles ow will result in the increase of the forces between the gas and solid phases
towards the middle-upper region of the bed with less uidization time (including drag force) acting on the particles. At the lower gas velocity
and higher bed expansion height. The gure also shows the solid (0.75 m/s), the particles are accumulated in the lower portion of the
volume fraction distribution at t = 60 s for three different PE particle bed. As the gas velocity raised, the solid volume fraction at the bottom
sizes. When the bed arrives at steady state, with the increase of PE of the bed declines gradually. Thus, the bubble size and bed expansion
particle size, the solid holdups present together near the wall and height apparently increase. However, excessively high gas velocity
bottom regions, and the FBR expansion height decreases obviously. (N7Umf) may lead to the continuous coming out of PE particles from
The results also prove that the polymer particles are extensively gath- the top of the bed, and it could destroy the normal operation of the
ered near the bottom of the bed, it could result in the occurrence of reactor. Besides, the increasing gas inlet velocity can induce to occur
the deuidization phenomenon. The most likely reason is that the su- and enhance entrainment of particles. Thus, it can be concluded that
percial gas velocity is too small to make larger polymer particles fully the operating gas inlet velocity for 1338 m is 0.90 m/s, and further
uidized. quantitative validation will be shown in Fig. 13.
Gas and solid velocities are important parameters to affect the ow The greater variation of the time-averaged solid volume fraction
behavior in FBR, and it can highly inuence the solid volume fraction near the core region in the reaction zone can be observed in Fig. 13,
distribution. In Fig. 11, mean solid velocity and volume fraction varia- and the calculated mean solid velocity proles are all depicted. The
tion along the radial direction under several PE particle sizes are trends of solid velocities in b(1) and c(1) are very similar, yet, because
depicted at different height of reaction zone. As mentioned above, of the effects of gas velocity and bed expanding section, the solid veloc-
a(1) and a(2) demonstrates a relative uniform and favorable ow ities in c(1) have changed denitely in the wall and core regions near
pattern and structure, and it is conceivable that the operation in the the height of 2 m. It could be explained by the contour obtained in
reactor is stable and reliable. At the same axial direction in reaction Fig. 12c. For Fig. 13a(1), the gas inlet velocity is too low to uidize the
zone, the solid volume fractions near the wall are much larger than particles fully, the formed bubbles cause the uctuation of the bed.
those values at the center region, and it indicates that the radial solid The velocity of solid phase is controlled simultaneously by lots of factors,
volume fraction exhibits a typical coreannulus ow structure shape. such as the interactions between two phases, granules gravitation,
Y. Che et al. / Powder Technology 278 (2015) 94110 105

Fig. 10. The visual representations of solid volume fraction at different polymer particle sizes (ug = 0.61 m/s): (a) 446 m, (b) 892 m, and (c) 1338 m.

particle friction and collision. All of the factors contribute to the solid ve- to 1338 m, the supercial gas velocity should increase from 0.61 m/s to
locity proles described in a(1), b(1) and c(1). As depicted in 0.90 m/s to achieve fully uidization state in the bed. However, an
Fig. 13b(2), the solid volume fractions become uniformly distributed apparent change in the average bubble size induced by the increase of
in the core region across the bed, and signicant differences are found gas velocity will lead to enhancement of the particles mixing along the
near the wall region in the reactor. It means that after the gas carries bed height, and it may give rise to unstable operation state. In addition,
the granules to the top of the bed, it jets out and the polymer particles as shown in Table 6, with the increase of gas inlet velocity, the overall
are circulated back down along the walls for the impact of bed pressure drop will increase remarkably, and the pressure drop is
expanding section. Lower mean solid volume fraction in the core region directly related to the reactor operating state. Thus, the higher overall
can be observed in c(2), and it could cause the entrainment of particles. pressure drop results in higher energy consumption, and it is not
For the given solid particles and reactors, the entrainment is very sensi- considered economically sustainable.
tive to the gas inlet velocity. For industrial operation, when NPPP (1338 m) is adopted, the
The comparison and analysis of the hydrodynamic natures in Figs. 12 operating gas velocity must be increased to a specic value (0.90 m/s)
and 13 show that, when the polymer particle size changes from 446 m to attain and sustain a more stable and better uidized state in the
106 Y. Che et al. / Powder Technology 278 (2015) 94110

Fig. 11. Time-averaged solid velocity (1) and solid volume fraction (2) variation along the radial direction at different particle size (a. 446 m, b. 892 m, c. 1338 m) (ug = 0.61 m/s, t = 60 s).

bed, rapidly. As mentioned above, an apparent increase of the average the operating state of the FBR under TPPP and NPPP in our research
bubble size will result in increase of the polymer particles ow and group now.
mixing along the bed axial and radial direction, but it leads to an
unstable operating state. Fig. 12b shows obvious changes of the bubble 6. Conclusion
behaviors and solid holdups in exploring the hydrodynamics and
uidized state. The reactor is becoming more sensitive to operating In the present paper, the scale-up effect on the hydrodynamic
conditions, and the bubbles become unstable, even in the optimized characteristics of gas and PE particles in a pilot-plant FBR was predicted
gas velocity. The conclusion can be safely drawn that, when the polymer through CFD modeling and simulating. A 2D EulerianEulerian CFD
particle size increased under NPPP, as a critical operating parameter, the model integrating KTGF was established, and validated by the industrial
supercial gas velocity will make the bed operation become more pressure drop data.
sensitive. It indeed presents a great challenge to operate and run the Due to the effect of the scale-up, the hydrodynamic issues show
pilot-plant scale FBR, which needs further investigation on the reactor remarkable differences. Firstly, during the initial uidization process,
performance. It should be noted that more detailed investigations are the observed axisymmetry turns to chaotic transient uidization state
being implemented to explore the effects of ethylene polymerization after only 1 s. Secondly, signicant differences in the coreannulus
and particle kinetics on the polymer particle behaviors and to assess structures and shapes of both gas and solid volume fractions are
Y. Che et al. / Powder Technology 278 (2015) 94110 107

Fig. 12. The visual representations of solid volume fraction at different operating gas velocities (ds = 1338 m): (a) 0.75 m/s, (b) 0.90 m/s, and (c) 1.05 m/s.

found. The core region in large scale FBR is much broader, and the distri- the local two phases' mean velocities and volume fractions vary in the
bution is nearly at for both of the phases. Meanwhile, an apparent pilot-plant reactor for PE production caused by scale-up effect.
change rather than a small peak for gas phase and a thin dense solid To explore the operating stability of applying NPPP preliminarily,
annular region in the wall is revealed, the ratio of core region and simulations are performed to assess the effects of particle sizes and
annulus area in pilot-plant FBR is 5/2. For the axial solid velocity, the operating conditions. The results indicate that the gas velocity pro-
core area is relatively thick, and the velocity varies slightly. The annulus vided by the plant is only applicable for PE particle size conned to
structure is thin and exhibits a very small particle velocity uctuation, approximately 446 m. With the increase of polymer particle diam-
the ratio has the same value with solid holdups. Finally, the bed eter (from 446 m to 1338 m), the supercial gas velocity should
expanding section could reduce the solid velocity and improve the be increased from 0.61 m/s to 0.90 m/s to achieve better steady u-
ow pattern of gas and polymer particles. From the comparisons of 2D idized state and maintain good ow eld distribution. Meanwhile, the
and 3D simulations, the scale up effect on the ow structure can be gas velocity will make the bed operation become more sensitive, and
observed with the same distribution situation in both monodisperse it presents a great challenge to operate and run the pilot-plant scale
and polydisperse cases. The results are helpful for understanding how FBR under NPPP.
108 Y. Che et al. / Powder Technology 278 (2015) 94110

Fig. 13. Time-averaged solid velocity (1) and solid volume fraction (2) variation along the radial direction at different gas velocity (a. 0.75 m/s, b. 0.90 m/s, c. 1.05 m/s) (ds = 1338 m, t =
60 s).

Nomenclature Res particle Reynolds number


Cd drag coefcient Re Reynolds number
ds particle diameter, m t ow time, s
D diameter, m Umf minimum uidization velocity, m/s
es particleparticle restitution coefcient UT particle terminal velocity, m/s
ew particlewall restitution coefcient Ug gas inlet velocity, m/s
!
g gravitational acceleration, m/s2 vg gas velocity, m/s
!
h bed height, m vs solid velocity, m/s
I identity matrix vr,s terminal velocity correlation for the solid, m/s
I2D second invariant of the deviatoric stress tensor vs the uctuating particle velocity, m/s
Kgs momentum exchange coefcient, kgm2/s
k turbulence kinetic energy tensor Greek symbols
p pressure, Pa g volume fraction of the gas phase
ps solid phase pressure, Pa s volume fraction of the solid phase
Pg gas phase pressure, Pa s,max maximum volume fraction of the solid phase
Y. Che et al. / Powder Technology 278 (2015) 94110 109

granular temperature, m2/s2 [19] G. Price, B. Chandrasekaran, I. Hulme, A. Kantzas, Comparison of 2-D and 3-D CFD
simulations of bubbling uidized beds with X-ray uoroscopy and imaging experi-
g stressstrain tensor of gas phase, N/m2 ments, AIChE Ann. Meet. (2004) 84478473.
s stressstrain tensor of solid phase, N/m2 [20] I. Hulme, E. Clavelle, L. van der Lee, A. Kantzas, CFD modeling and validation of
g viscosity of the gas phase, Pas bubble properties for a bubbling uidized bed, Ind. Eng. Chem. Res. 44 (2005)
42544266.
s solids shear viscosity, Pas [21] B.K. Chandrasekaran, L. van der Lee, I. Hulme, A. Kantzas, A simulation and experi-
s,col solids collisional viscosity, Pas mental study of the hydrodynamics of a bubbling uidized bed of linear low density
s,kin solids kinetic viscosity, Pas polyethylene using bubble properties and pressure uctuations, Macromol. Mater.
Eng. 290 (2005) 592609.
s,fr solids frictional viscosity, Pas [22] L. van der Lee, B.K. Chandrasekaran, I. Hulme, A. Kantzas, A non-invasive hydrody-
s solid bulk viscosity, Pas namic study of gassolid uidised bed of linear low density polyethylene, Can. J.
g gas density, kg/m3 Chem. Eng. 83 (2005) 119126.
[23] M.A. Dehnavi, S. Shahhosseini, S.H. Hashemabadi, S.M. Ghafelebashi, CFD based
s solid density, kg/m3
evaluation of polymer particles heat transfer coefcient in gas phase polymerization
turbulence dissipation rate, m2/s3 reactors, Int. J. Heat Mass Transfer 35 (2008) 13751379.
voidage [24] R. Fan, Computational Fluid Dynamics Simulation of Fluidized Bed Polymerization
s the collision dissipation of energy, m2/s2 ReactorsChem Eng-New York Iowa State University, Ames, Iowa, 2006.
[25] R. Fan, R.O. Fox, Segregation in polydisperse uidized beds: validation of a multi-
gs the dissipation or creation of granular energy, m2/s2 uid model, Chem. Eng. Sci. 63 (2008) 272285.
[26] R.G. Rokkam, R.O. Fox, M.E. Muhle, Computational uid dynamics and electrostatic
modeling of polymerization uidized-bed reactors, Powder Technol. 203 (2010)
Acknowledgments 109124.
[27] M. Terano, K. Suehiro, T. Sagae, E. Tobita, 2 Japanese national project for the
innovation of industrial polypropylene process technology, Stud. Surf. Sci. Catal.
This work is nancially supported by the National High Technology (2006) 712.
Research and Development Program 863 (No. 2012AA040306), [28] P. Galli, G. Vecellio, Technology: driving force behind innovation and growth of
polyolens, Prog. Polym. Sci. 26 (2001) 12871336.
the National Natural Science Foundation of China (21406061), and the [29] P. Galli, The reactor granule technology: a revolutionary approach to polymer blends
Shanghai Municipal Natural Science Foundation (14ZR1410600). and alloys, Macromol. Symp. 78 (1994) 269284.
The authors also express thanks to Prof. Huanxin Lai for his valuable [30] P. Galli, G. Collina, P. Sgarzi, G. Baruzzi, E. Marchetti, Combining ZieglerNatta and
mettalocene catalysis: new heterophasic propylene copolymers from the novel
suggestions and help.
multicatalyst reactor granule technology, J. Appl. Polym. Sci. 66 (1997) 18311837.
[31] Z. Li, Y. Ma, W. Yang, A facile, green, versatile protocol to prepare polypropylene-
g-poly (methyl methacrylate) copolymer by watersolid phase suspension
References grafting polymerization using the surface of reactor granule technology
polypropylene granules as reaction loci, J. Appl. Polym. Sci. 129 (2013)
[1] J.T. McCoy, J.B.P. Soares, R. Rawatlal, Analysis of slurry-phase co-polymerization of 31703177.
ethylene and 1-butene by ZieglerNatta catalysts part 1: experimental activity [32] Z. Fu, X. Wang, N. Li, Z. Fan, Synthesis of high toughness and stiffness polyolen
proles, Macromol. React. Eng. 7 (2013) 350361. In-reactor alloys by reactor granule technology, J. Polym. Mater. 22 (2005) 153158.
[2] T. Xie, K.B. McAuley, J.C.C. Hsu, D.W. Bacon, Gas phase ethylene polymerization: [33] B. Rotzinger, M. Brunner, In-reactor stabilization of polypropene, Polym. Degrad.
production processes, polymer properties, and reactor modeling, Ind. Eng. Chem. Stab. 93 (2008) 316320.
Res. 33 (1994) 449479. [34] L. Wu, S.E. Wanke, MgCl2-supported TiCl4 catalysts for production of morphology-
[3] M.J.H. Khan, M.A. Hussain, Z. Mansourpour, N. Mostou, N.M. Ghasem, E.C. Abdullah, controlled polyethylene, Handbook of Transition Metal Polymerization Catalysts,
CFD simulation of uidized bed reactors for polyolen production a review, J. Ind. John Wiley & Sons, Inc., Hoboken, New Jersey, 2010, pp. 231259.
Eng. Chem. 20 (2014) 39193946. [35] C. Loha, H. Chattopadhyay, P.K. Chatterjee, Assessment of drag models in simulating
[4] G.N. Ahuja, A.W. Patwardhan, CFD and experimental studies of solids hold-up bubbling uidized bed hydrodynamics, Chem. Eng. Sci. 75 (2012) 400407.
distribution and circulation patterns in gassolid uidized beds, Chem. Eng. J. 143 [36] B.G.M. van Wachem, J.C. Schouten, C.M. van den Bleek, R. Krishna, J.L. Sinclair,
(2008) 147160. Comparative analysis of CFD models of dense gassolid systems, AIChE J. 47
[5] Z. Mansourpour, S. Karimi, R. Zarghami, N. Mostou, R. Sotudeh-Gharebagh, Insights (2001) 10351051.
in hydrodynamics of bubbling uidized beds at elevated pressure by DEM-CFD [37] J. Ding, D. Gidaspow, A bubbling uidization model using kinetic theory of granular
approach, Particuology 8 (2010) 407414. ow, AIChE J. 36 (1990) 523538.
[6] M.A. van der Hoef, M. van Sint Annaland, J.A.M. Kuipers, Computational uid [38] M. Syamlal, W. Rogers, T.J. O'Brien, MFIX documentation: theory guide, Technical
dynamics for dense gassolid uidized beds: a multi-scale modeling strategy, Note, DOE/METC-94/1004, NTIS/DE94000087, National Technical Information
Chem. Eng. Sci. 59 (2004) 51575165. Service, Springeld, VA, 1993.
[7] M.A. van der Hoef, M. van Sint Annaland, N.G. Deen, J.A.M. Kuipers, Numerical [39] C.K.K. Lun, S.B. Savage, D.J. Jeffrey, N. Chepurniy, Kinetic theories for granular ow:
simulation of dense gassolid uidized beds: a multiscale modeling strategy, inelastic particles in Couette ow and slightly inelastic particles in a general ow
Annu. Rev. Fluid Mech. 40 (2008) 4770. eld, J. Fluid Mech. 140 (1984) 223256.
[8] J. Sun, Y. Zhou, C. Ren, J. Wang, Y. Yang, CFD simulation and experiments of dynamic [40] J. Sun, J. Wang, Y. Yang, CFD investigation of particle uctuation characteristics of
parameters in gassolid uidized bed, Chem. Eng. Sci. 66 (2011) 49724982. bidisperse mixture in a gassolid uidized bed, Chem. Eng. Sci. 82 (2012) 285298.
[9] S. Karimi, Z. Mansourpour, N. Mostou, R. Sotudeh-Gharebagh, CFD-DEM study of [41] W.-C. Yan, Z.-H. Luo, Y.-H. Lu, X.-D. Chen, A CFD-PBM-PMLM integrated model for
temperature and concentration distribution in a polyethylene uidized bed reactor, the gassolid ow elds in uidized bed polymerization reactors, AIChE J. 58
Part. Sci. Technol. 29 (2011) 163178. (2012) 17171732.
[10] J.J. Wang, Y. Han, X.P. Gu, L.F. Feng, G.H. Hu, Effect of agitation on the uidization [42] D. Gidaspow, Multiphase Flow and Fluidization, Continuum and Kinetic Theory
behavior of a gassolid uidized bed with a frame impeller, AIChE J. 59 (2013) Descriptions with Application, Academic Press, Inc., California, 1994.
10661074. [43] J.O. Hinze, Turbulence, McGraw-Hill College, New York, 1975.
[11] Y. Kaneko, T. Shiojima, M. Horio, DEM simulation of uidized beds for gas-phase [44] E.-U. Hartge, L. Ratschow, R. Wischnewski, J. Werther, CFD-simulation of a
olen polymerization, Chem. Eng. Sci. 54 (1999) 58095821. circulating uidized bed riser, Particuology 7 (2009) 283296.
[12] D.-P. Shi, Z.-H. Luo, A.-Y. Guo, Numerical simulation of the gassolid ow in [45] N. Reuge, L. Cadoret, C. Coufort-Saudejaud, S. Pannala, M. Syamlal, B. Caussat,
uidized-bed polymerization reactors, Ind. Eng. Chem. Res. 49 (2010) 40704079. Multiuid Eulerian modeling of dense gassolids uidized bed hydrodynamics:
[13] C. Amornsirirat, B. Chalermsinsuwan, L. Mekasut, P. Kuchonthara, P. Piumsomboon, inuence of the dissipation parameters, Chem. Eng. Sci. 63 (2008) 55405551.
Experiment and 3D simulation of slugging regime in a circulating uidized bed, [46] L. Cammarata, P. Lettieri, G.D. Micale, D. Colman, 2D and 3D CFD simulations of
Korean J. Chem. Eng. 28 (2011) 686696. bubbling uidized beds using EulerianEulerian models, Int. J. Chem. React. Eng. 1
[14] O. Ashra, N. Mostou, R. Sotudeh-Gharebagh, Two phase steady-state particle size (2003) A48.
distribution in a gas-phase uidized bed ethylene polymerization reactor, Chem. [47] H.P. Kuo, H.Z. Zhang, R.C. Hsu, CFD modelling of a countercurrent staged uidised
Eng. Sci. 73 (2012) 17. bed, Chem. Eng. J. 137 (2008) 664676.
[15] M.P. Dudukovic, Opaque multiphase ows: experiments and modeling, Exp. [48] A. Busciglio, G. Vella, G. Micale, L. Rizzuti, Analysis of the bubbling behaviour of 2D
Thermal Fluid Sci. 26 (2002) 747761. gas solid uidized beds: part II. Comparison between experiments and numerical
[16] F. Taghipour, N. Ellis, C. Wong, Experimental and computational study of gassolid simulations via Digital Image Analysis Technique, Chem. Eng. J. 148 (2009)
uidized bed hydrodynamics, Chem. Eng. Sci. 60 (2005) 68576867. 145163.
[17] M.A. Dehnavi, S. Shahhosseini, S.H. Hashemabadi, S.M. Ghafelebashi, CFD simulation [49] S.H. Hosseini, G. Ahmadi, R. Rahimi, M. Zivdar, M.N. Esfahany, CFD studies of solids
of hydrodynamics and heat transfer in gas phase ethylene polymerization reactors, hold-up distribution and circulation patterns in gassolid uidized beds, Powder
Int. J. Heat Mass Transf. 37 (2010) 437442. Technol. 200 (2010) 202215.
[18] I. Hulme, A. Kantzas, Determination of bubble diameter and axial velocity for a [50] T. Li, C. Guenther, MFIX-DEM simulations of change of volumetric ow in uidized
polyethylene uidized bed using X-ray uoroscopy, Powder Technol. 147 (2004) beds due to chemical reactions, Powder Technol. 220 (2012) 7078.
2033. [51] L.-S. Fan, C. Zhu, Principles of GasSolid Flows, Cambridge University Press, 2005.
110 Y. Che et al. / Powder Technology 278 (2015) 94110

[52] C.Y. Wen, Y.H. Yu, Mechanics of uidization, Chem. Eng. Prog. Symp. Ser. 62 (1966). [57] D. Kunii, O. Levenspiel, Fluidization Engineering, Butterworth-Heinemann, Boston,
[53] G. Dompazis, V. Kanellopoulos, V. Touloupides, C. Kiparissides, Development of a 1991.
multi-scale, multi-phase, multi-zone dynamic model for the prediction of particle [58] L.-H. Wei, W.-C. Yan, Z.-H. Luo, A preliminary CFD study of the gassolid ow elds
segregation in catalytic olen polymerization FBRs, Chem. Eng. Sci. 63 (2008) in multizone circulating polymerization reactors, Powder Technol. 214 (2011)
47354753. 143154.
[54] N. Chepurniy, Kinetic theories for granular ow: inelastic particles in Couette ow [59] J. Min, J.B. Drake, T.J. Heindel, R.O. Fox, Experimental validation of CFD simulations of
and slightly inelastic particles in a general oweld, J. Fluid Mech. 140 (1984) a lab.-scale uidized-bed reactor with and without side-gas injection, AIChE J. 56
223256. (2010) 14341446.
[55] P. Johnson, R. Jackson, Friction akollisional constitutive relations for granular [60] Y.-C. Li, Introduction to Fluidization Process Engineering, Science Press, Beijing,
materials, with application to plane shearing, J. Fluid Mech. 176 (1987) 6793. 2008.
[56] F. Vejahati, N. Mahinpey, N. Ellis, M.B. Nikoo, CFD simulation of gassolid bubbling
uidized bed: a new method for adjusting drag law, Can. J. Chem. Eng. 87 (2009)
1930.

Das könnte Ihnen auch gefallen