Chemical Engineering Science 109 (2014) 8597

Contents lists available at ScienceDirect

Chemical Engineering Science

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

Modelling and experimental investigation of the full-loop gassolid

ow in a circulating uidized bed with six cyclone separators
Yu Jiang a,b, Guizhi Qiu a,b, Haigang Wang a,n
a
b

Institute of Engineering Thermophysics, Chinese Academy of Sciences, PO Box 2706, Beijing 100190, China
University of Chinese Academy of Sciences, PO Box 2706, Beijing 100190, China

H I G H L I G H T S

CPFD model is used to simulation the full-loop of a circulating uidized bed with six cyclone separators.

Simulation results are compared with Electrical capacitance tomography and pressure measurement.

The CPFD model combined with ECT technology provides a possibility to optimize the design for large scale circulating uidized bed.

art ic l e i nf o

a b s t r a c t

Article history:
Received 22 November 2013
Received in revised form
15 January 2014
Accepted 25 January 2014
Available online 31 January 2014

In the literature, there are few reports on the full-loop gassolid ow in a circulating uidized bed (CFB) with
large scale and complex cyclone arrangement. In this paper, a new approach based on computational particle
uid dynamic (CPFD) method combined with electrical capacitance tomography (ECT) is used to investigate
the hydrodynamic behavior of gassolid ow in a CFB with six cyclone separators in order to improve the
design and performance of a large scale CFB boiler. The full-loop CFB system for the simulation includes the
CFB riser, cyclone, standpipe and U-loop. Two types of cyclone arrangement, i.e. axis and point based
symmetric arrangement, are used for the CPFD simulation and ECT measurements. To validate the CPFD
simulation, ECT is applied to measure the solids concentration in the standpipe with eight electrodes
mounted on the outside of the standpipe. Key parameters including pressure, solids recirculation ux and
velocity prole along different positions based on the CPFD simulation are analyzed and compared with
experimental results. The CPFD simulation shows that the gassolid ow is non-uniform among the six
parallel cyclones. The solids concentration of four cyclones at the corner of the riser is higher than that of the
others. The location of cyclone as well as the inlet angle of the cyclone needs to be optimized. The study
shows that the presented approach based on CPFD simulation and ECT measurements can be used to
optimize the arrangement of cyclone separators in a supercritical pressure circulating uidized bed system.
& 2014 Elsevier Ltd. All rights reserved.

Keywords:
Circulating uidized bed
CPFD
Cyclone Separator
Full-loop
Process tomography

1. Introduction
Circulating uidized bed (CFB) is one kind of clean coal
combustion technologies which plays an important role in the
power generation and coal gasication industry (Reh, 2003). Large
thermal capacity and high steam pressure is a tendency for the
development of CFB boilers (Lv et al., 2007, Fan et al., 2008, Chen
et al., 2008). To meet the demands for high steam parameter and
large thermal capacity, high efciency of gassolid separation is a
key to achieving high combustion efciency, reducing limestone
consumption and NOx emission (Koornneef et al., 2007).

Corresponding author. Tel.: 0086 10 8254 3140.

E-mail address: wanghaig@hotmail.com (H. Wang).

http://dx.doi.org/10.1016/j.ces.2014.01.029
0009-2509 & 2014 Elsevier Ltd. All rights reserved.

With the scaling up of a CFB boiler, the dimension of the

cyclone is increased accordingly and the separation efciency
decreases due to a reduction in the centrifugal force. To overcome
the above issue, a large cyclone is replaced by numbers of smaller
cyclone with the increase of boiler size to reduce the cyclone size.
Different arrangement of cyclones on the top of the CFB riser is
provided and patented (Hack et al., 2008). Experimental research
has been carried out and methods related with cyclone arrangement have been patented (Armistead et al., 2002, Lv et al., 2007,
Zhou et al., 2012). However, there is a non-uniform solids mass
ux distribution among cyclones with a maximum difference of
17% (Morin, 2003, Chen et al., 2008, Zhou et al., 2012). For a CFB
boiler with multi-cyclone separators, it is important to investigate
the gassolid ow in the whole loop including the CFB riser,
cyclones as well as standpipe and U-loop.

86

Y. Jiang et al. / Chemical Engineering Science 109 (2014) 8597

Computational uid Dynamics (CFD) simulation provides detailed

information for the investigation of uidization characteristics with
large-scale CFB boilers (Reh, 2003). Research has been reported using
the CFD approach to investigate the three dimensions gassolid ow
in a CFB boiler (Zhang et al., 2008, 2010, Ahuja and Patwardhan,
2008, Hartge et al., 2009). In dealing with gassolid uidized beds,
two approaches, i.e. the Eulerian-Eulerian two-uid model (TFM)
(Gidaspow and Ettehadieh, 1983, Lun et al., 1984) and the EulerianLagrangian discrete particle method (DPM), are commonly used. The
TFM model treats solids as a continuous phase which interacts with
the gas phase by momentum exchange. Conservation equations for
each phase have similar terms and are solved together with a set of
constitutive equation derived by experiment. The TFM model has
been widely used in multiphase ows simulation. However, it has
limitations, such as not applicable to particle size distribution and
inter-particle forces (Makkawi et al., 2006). The DPM method
describes the discrete phase by tracking numerous particles trajectories which exchange mass, momentum and energy with the gas
phase all through the whole simulation eld. The DPM model takes
into account the particle size distribution as well as particleparticle
interactions. However, it is difcult to simulate dense gassolid ow
with solids volume fraction above 5% due to large amount of particle
numbers. In general, the particle number is under the order of 2 
105 in the DPM model and it is often applied to two-dimensional
simulation. Recently, an EulerianLagrangian model called CPFD
(computational particle uid dynamics) has been used to model
the gassolid ow in a uidized bed (Abbasi et al., 2011, Chen et al.,
2013). This methodology incorporates the multi-phase-particle-incell (MP-PIC) method for calculating a dense gassolid ow
(Andrews and O'Rourke, 1996, Snider, 2001). In the CPFD approach,
the gas phase is modeled as a continuous uid and particles as a
discrete phase which can handle particle size distribution. Particles
are classed into numerous of computational parcels. Each parcel
represents a number of physical particles which have a same velocity
and material property in the computational domain. With this
scheme, billions of particles can be simulated much more efciently.
To validate the CFD simulation results in a CFB boiler, it is
necessary to verify the results with measurements. Electrical
capacitance tomography (ECT) provides an option to investigate
the gassolid ow in a uidized bed due to its no-intrusive and
no-invasive nature (Dyakowski et al., 1997, 1999 , Makkawi and
Wright, 2004, 2006, Du et al., 2005, Wang et al., 2006).
To understand the hydrodynamic behavior of a gassolid
circulating uidized bed, a cold CFB test facility with six cyclone
separators in the top of the riser and rectangular shape combustion chamber has been built in the Institute of Engineering
Thermophysics, Chinese Academy of Sciences. CPFD is used to
simulate the whole circulating loop. CPFD simulation results are
compared with experimental results by ECT and pressure measurements. Two different arrangements of cyclones are used
to compare the gassolid uid hydrodynamic behavior in the
multi-cyclone CFB system. The objective of the research is to
evaluate the applicability of CPFD method for the gassolid ow
simulation in a whole-scale CFB boiler and compare with experiment results. The CPFD simulation is based on a commercial code
BARRACUDA. ECT was used to measure the solids concentration in
the cross sections of the standpipe with eight electrodes mounted
outside of the pipe and to validate the CPFD simulation results.

2. CPFD mathematical model

described as a continuous phase with strong coupling to the discrete

solids phase in mass and momentum equations. As the gas and solids
phases are isothermal and the gas phase is incompressible, no
volume averaged uid energy equations are needed. In the CPFD
scheme, a concept of numerical particle is introduced, which is a
numerical approximation similar to the numerical control volume
within which the uid has a common property. The solids phase is
modeled as numerical parcels each containing quite a number of
physical particles with same properties (species, density, size, etc) in
the same location. The ow elds of gas and solids phase are
calculated by separated governing equations. For the gas phase, the
governing equations are

g g g vg Sg
t g

g vg g g vg vg P g g g g g  F
t g

where g represents the volume fraction of gas, g and vg stands for

density and velocity of gas respectively, Sg is a source term, g
represents the gas stress tensor, p stands for the pressure of gas, g
is the acceleration of gravity, F is the rate of momentum exchange per
volume between the gas and solids phases. The momentum equation
presented here neglects the viscous molecular diffusion in the uid
but retains the viscous drag between particles and uid through an
interphase drag force, F, which is
!
P
F f m Dp vg vp 
dmdv
3

where Dp is the drag function, vp and p represents particle velocity

and density respectively, f is the probability distribution function
which is calculated from Liouville equation as


f
d
f vp vp f vp 0
4
t
dt
d
vp is the particle acceleration, which is obtained by
where dt
calculating all forces on the particles and is given by MP-PIC method
(Andrews and O'Rourke, 1996, Snider, 2001) as following:

d
P p
vp Dp vg  vp 

g
dt
p p p

where p is inter-particle normal stress, p represents volume

fraction of particles. The trajectory of a particle is solved by
dxp
vp
dt

where x is the location of the tracing particle.

2.2. Drag model
The Wen-Yu drag model is applicable to gassolid ow with
solids volume fraction lower than 0.61 while the Ergun drag model
covers the range of 0.470.7. As the volume fraction of solids in the
present study is less than 0.65 at close packing limit, the interphase drag function is dened by Wen-Yu model (Wen and Yu,
1966)
Dp C d

3 g jvg  vp j
8 p 3V p =4 1=3

where Cd is the drag coefcient. It depends on the Reynolds

number, i.e..

2.1. Governing equations

Cd
The CPFD methodology takes an Eulerian-Lagrangian approach to
describe the gassolid ow in three dimensions. The gas phase is

24
 2:65
1 0:15Re0:687 g
for Re o 1000
Re

C d 0:44g

 2:65

for Re Z 1000

Y. Jiang et al. / Chemical Engineering Science 109 (2014) 8597

particle volume to the grid is dened as follow

The Reynolds number is given by



2g jvg  vp j 3V p 1=3
Re
g
4

where g is the gas density, g is the gas viscosity, vg and vp

represent the gas and particle velocity respectively, and V p is the
volume of a particle.

2.3. Solids stress model

In the CPFD scheme, unlike the DEM approach which models
the particle-particle interaction forces by a spring-damper model,
the inter-particle collision forces are calculated as a spatial
gradient. Because it is difcult to calculate particle stress gradient
for each particle in a dense ow, CPFD calculates particle forces as
a gradient on the grid and maps it back to particles. The present
study adopts the particle normal stress model by Harris and
Crighton (1994).

P S p

max cp  p ; 1  p 

87

10

where is a constant suggested to be on the order of 10  7, cp is

the solids volume fraction at close packing. Ps has pressure units
and is supposed to be 2 r r5 (Auzerais et al., 1988).
2.4. CPFD simulation procedure
In the CPFD approach, each cell contains numbers of numerical
parcels. The solids volume fraction in the cell (i, j, k) from mapping

i;j;k

1
V i;j;k

Np

V p np Si;j;k

11

where Vi,j,k is the volume of cell (i, j, k), Np is the total number of
numerical parcels in the cell, Vp is the particle volume, np is the
number of real particles in a numerical parcel, and Si,j,k is the
interpolation operator. The interpolation operator in x direction to
the cell i is given by
 x Zx Zx
Sxi xp 01 i  1 xp p xi i 1
12
The interpolation operators in y and z directions have a similar
form. With particle volume fraction obtained, the gas volume
fraction can be updated and it is used to solve uid continuity and
momentum equations in the next time step.
The mass and momentum equations are approximated and
solved by nite volumes with staggered scalar and momentum
nodes. The numerical particle velocity at the following time step is
updated by
h
i
vnp t Dp vnf ;p 1  1 pnp 1  1p np 1  g
p
p
n1
vp
13
1 tDp
where, vnp 1 is the interpolated particle velocity, vnf ;p 1 ,pnp 1 ,np 1
represent the gas phase velocity, the pressure gradient and the
solid stress gradient interpolated at the particle location
respectively.
Based on Eq. (13), the particle location at the following time
step is obtained by
xnp 1 xnp vnp 1 t

Fig. 1. CFB model (units by mm). (a) Schematics of simulation geometry, (b) axis-symmetric, (c) point-symmetric.

14

88

Y. Jiang et al. / Chemical Engineering Science 109 (2014) 8597

The inter-phase momentum transfer at momentum cell (i, j, k) is

then
"
#
1 Np
1
n1
n1
n1
n1
Dp vg;p vp  pp
15
F i;j;k
n p mp
S
V i;j;k 1 i;j;k
p
1
is the interpolated rate of
where mp is the particle mass and F ni;j;k
inter-phase momentum exchange per volume.

3. Circulating uidized bed model and ECT sensors

The schematic diagram of a cold CFB model as shown in Fig. 1 is
scaled down from a 600 MWe super-critical pressure circulating
uidized bed following the scaling rules proposed by Glicksman,
Hyre and Woloshun (1993). The main dimensions and parameters
of the model are given in Table 1. Two types of arrangement
for cyclones are considered in the CPFD simulation and shown in
Fig. 1(b) and (c), they are named axis and point based
symmetric in this research. As it is recommended by Barracuda's
manual, CPFD requires hundreds of thousands of grids for experimental scale apparatus. In the present study, the total number of
grids is about 500,000 for two arrangements and the mesh is
shown in Fig. 2.
The particle size distribution given in Fig.3 is the same with
experiment and analyzed using a Malvern particle size analyzer.
Gas phase and particle properties used in the simulation are listed
in Table 2. The simulation parameters are listed in Table 3 and two
simulation conditions are given in Table 4.
In the CPFD scheme, the interpolation operators which are both
locally and globally conservative are used for the simulation, and

the sub-grid model for modeling the particle normal stress applied
to discrete particles is implemented in a robust and fast algorithm.
Meanwhile, particles are implicitly coupled to the uid phase, and
the uid momentum and pressure equations are implicitly solved,
which gives a robust solution. In the present study, the residual of
velocity in each time step is less than 10  8 while the residual of
pressure is 10  9.
The inlet boundary for the gas phase is at the base of the riser and
the bed material is packed in the bottom of the CFB riser before the
simulation start. As the air distributor used in experiment is a porous
plate and difcult to model directly, it is not included in the CPFD
simulation. Instead, a simplied inlet boundary with pressure drop of
820 Pa is used in the simulation which is in accordance with the
experimental measurement. The initial packed bed height is 0.5 m and
the total weight is 200 kg. An air ow rate of 0.6 m/s is introduced
from the bottom of the U-loop for delivering the recycled solids back

Table 1
Dimensions of CFB boiler and cyclone separator.
Parameter

Units (m)

CFB boiler

Width
Depth
Height

0.42
0.92
5.8

Cyclone separator

Vortex tube diameter

Inlet port width
Inlet port height
Diameter of cylindrical part of cyclone
Cylindrical part length
Conical part diameter in the bottom
Conical part length

0.14
0.19
0.30
0.31
0.39
0.07
0.20

Fig. 3. Particle size distributions.

Table 2
Gas and solid properties.
Parameter

Value
3

Solid density, s (kg/m )

Gas density, g (kg/m3)
Gas viscosity, vg (Pa s)

Fig. 2. Mesh of the CFB. (a) Axis-symmetric, (b) point-symmetric.

2620
1.205
1.85E-05

Y. Jiang et al. / Chemical Engineering Science 109 (2014) 8597

89

Table 3
Input parameters in the CPFD simulation.
Particle-to-wall interaction

Normal retention coefcient, en

Tangential retention coefcient, et
Diffuse bounce, Df

0.89
0.68
0

Solver settings

Time step, t
Total time
gravitational acceleration, g
Maximum volume iteration
Volume residual
Maximum pressure iterations
Pressure residual
Maximum velocity iterations
Velocity residual
Maximum momentum redirection from collision
close pack fraction limit

4E-4 s
32 s
 9.81
1
1.00E-05
2000
1.00E-08
50
1.00E-07
40%
0.65

Table 4
Simulation conditions.
Case

U (m/s)

Bed material (kg)

priser (kPa)

Separators layout

A
B

4
4

200
200

5
5

Axis-symmetric
Point-symmetric

into the CFB riser. The top of the cyclones is set as the out ow
condition with a pressure 1000 Pa lower than the atmospheric
pressure, which is derived from the time-averaged pressure drop of
the six cyclones in the experiment. The x-axis is along the front-toback wall direction, the y-axis is along the side-to-side wall direction,
and the z-axis is against the gravity direction.
Fig. 4 shows the ECT sensor used in the research. The frames
of the sensors are the wall of the standpipes. The sensor is
enclosed by copper shielding to eliminate external interference.
Each electrode has a length of 5 cm and width 2.5 cm. The
measurement region encompasses a height between 50 and
55 cm above the U-loop distributor. Therefore, each pixel in the
ECT imaging area represents an axial average over this circular
measurement volume. The measurement system is an AC-based
ECT system with a data acquisition rate up to 250 frames/s for an
8-electrode ECT sensor (Yang and York, 1999). Using a Linear Back
Projection (LBP) algorithm, images can be reconstructed online
(Xie et al., 1992). To improve the image quality and accuracy, the
Landweber iteration is used in the research (Yang and Peng, 2003).

4. Results analysis for CPFD modeling

Fig. 4. ECT sensor design.

4.1. Solids concentration and velocity vectors distribution

In the CPFD simulation, the rst 10 s were neglected to
disregard the effect of the impulsive initialization effect. The real
simulation times for both cases are longer than 20 s to ensure a
complete circulation of particles. Figs. 5 and 6 give a snapshot of
the particle ow eld for case A and B respectively. From Fig. 5(b),
it can be seen that the cyclone labeled as E in Fig. 1 has higher
solids volume fraction closed to the inner wall of the inlet, which
will negatively affect the efciency of separation. In both cases, it
can be nd that cyclones in the corner (cyclones labeled as A,
C, D and F) have a similar distribution which particles tend to
pass by the outer wall of the inlets, which can improve the
efciency of separation. However, in the middle ones (cyclones
labeled as B and E), the solids moves through the middle of the
inlet, which decreases the efciency.
Fig. 7 shows the velocity eld of gas phase at the top and the
bottom of the CFB riser. From Fig. 7(a) and (b), it can be found out that

the gas ow changes its direction from vertical to horizontal as it

approaches the inlet of cyclones. The gas ow has a strong rotation in
the center of cyclone, which can also be observed in Fig. 7(e) and (f).
The gas ow tends to ow downward along the wall of cyclone and
then turns around joining the central upward ow right below the
outlet of cyclone until it ows through the outlet. In the bottom of the
CFB riser, the gas ow is not uniform with both upward and downward ows as shown in Fig. 7(c) and (d). From Fig. 7(c) and (d), it can
also be observed that the air ow used to uidize the particles in the
U-loop ows back into the CFB riser. Thus, particles in the standpipe
are dominated by gravitational force.
4.2. Comparison of pressure drop
In the experiment, the measured points for pressure are located
at ve layers in the vertical direction and each layer includes

90

Y. Jiang et al. / Chemical Engineering Science 109 (2014) 8597

Fig. 5. Contours of particle volume fraction at t 32.93 s for case A. (a) Particle tracks in the full-loop of CFB, (b) contour of cross-section at Z 5.5 m, (c) contour of the
bottom.

6 points at different locations in the same cross section in the CFB

riser. Thus, pressure drops in the dense region, middle region,
dilute region and cyclones are recorded after the ow becomes
steady. The signals are connected to a PC with Agilent Data
acquisition system. The area-weighted pressures taken from
different heights along the CFB riser obtained by simulation are
compared with measurements. Fig. 8 shows the axial prole of
time-averaged pressure drop for case A and B. The average
pressure in the top of the riser is about 100,520 Pa, which
conforms to the requirement of slightly negative pressure by
experiment. The total pressure drop in the CFB riser is about
4 kPa, while in the experiment it is 5 kPa.
Tables 5 and 6 compare the pressure drops obtained from the
CPFD simulation and measurements for case A. From Table 5 it can
be found out that CPFD has a lower pressure drop prediction for
the dense region, which will result in the underestimation of
solids volume fraction in this region. The maximum relative error
in the dense region is 128.23% and much higher than the predicted
results by the TFM method combined with the EMMS drag model
(Zhang et al., 2008). In the dilute region, the CPFD result has a
good agreement with measurements. The calculated pressure drop
across the cyclones by CPFD is smaller than the experiment results.
The main reason for the discrepancy is due to the inaccurate
estimation of drag force both in the dense and dilute region for
the CPFD simulation. The dense gassolid ow in a CFB riser

is heterogeneity and the CPFD model should account for this

intrinsic characteristics. Therefore, it is necessary to consider a
hybrid model to estimate the drag force for the whole-loop CPFD
simulation. The EMMS drag model is one of the options for the
dense gassolid ow CPFD simulation (Zhang et al., 2008, Chen
et al., 2013). However, the current CPFD code does not provide the
user dened functions (Chen et al., 2013). Another reasons for the
above discrepancy is due to the simplied inlet boundary used in
the CPFD simulation. In the experiment, one type of tuyere gas
distributor is used to uidize the particles. It is well known that
the uidization is affected by the performance of the gas distributor (Lombardi et al., 1997). However, it is difcult to simulate
the real air distributor in the CPFD approach due to the limitation
of computer capacity. Therefore, the simplied inlet boundary will
also introduce predicted errors.
4.3. Particle re-circulation ux
Particle re-circulation ux into the riser from the U-loop is an
important parameter to indicate the performance of a CFB boiler.
As the particle mass ow rate at the inlet of the cyclones uctuates
rather intensely over time, a time-averaged estimation of the ux
is taken after the gassolid ow become steady. Fig. 9 shows
the particle ux distribution among the cyclones from the CPFD
simulation.

Y. Jiang et al. / Chemical Engineering Science 109 (2014) 8597

91

Fig. 6. Contours of particle volume fraction at t 21.07 s for case B. (a) Particle tracks in the full-loop of CFB, (b) contour of cross-section at Z 5.5 m, (c) contour of the
bottom.

From the results it can be seen that the cyclone in the middle
has a lower particle mass ow rate than the rest in both cases for A
and B. It reveals that more particles go through cyclones located at
the corner. The total mass ow rates in both sides are almost the
same, which means that the distribution of particles in the two
sides of the riser is quite uniform. It also can be seen that the
cyclone arrangement for axis symmetry is better than point
arrangement in terms of uniform solids distribution.

4.4. Distribution of solid concentration and velocity in the CFB riser

The solids concentration in the boiler can be derived from the
measured pressure drop and dened as following (Kunii and
Levenspiel, 1991).
1  p=p g l

16

where, is the area-averaged solids volume fraction in the cross

section and l is the distance between two measured points.
Fig. 10 gives the time-averaged solids concentration both from
CPFD simulations and pressure drop predictions in the CFB riser.
As can be seen from Fig. 10, CPFD predicts a bottom-dense and
upper-dilute structure for the solids concentration. In both cases,
the CPFD results are in well agreement with experimental data in
the region of z41 m. However, the CPFD predictions for solids
concentration are much smaller than the experiment data in the
region below z 1 m, which means more particles are brought
upwards and thus the predictions for upper region are denser than
the experiment. The results are quite contrary to the CPFD
simulation for a CFB riser with single cyclone separator (Chen
et al., 2013). This is mainly due to the inaccurate estimation of drag
force in the dense region.
Fig. 11 shows the horizontal distribution of particle vertical
velocity at different heights along the riser. The CPFD simulation

92

Y. Jiang et al. / Chemical Engineering Science 109 (2014) 8597

Fig. 7. Comparison of vector eld obtained in two cases by CPFD. (a) Top of the riser at t 32.92 s (case A), (b) top of the riser at t 21.07 s (case B), (c) bottom of the riser at
t 32.92 s (case A), (d) bottom of the riser at t 21.07 s (case B), (e) Z 5 m at t 32.92s (case A), (f) Z5 m at t 21.07 s (case B).

result shows that the particle velocity is higher in center. In

contrast to the core upward ow, the particles form a downward
ow near the wall. This is the core-annulus structure or crownlike shape concentration which is congruent with the results
reported from the TEM simulations (Pain et al., 2002 Malcus et al.,
2002, Wang et al., 2006). In the upper region of the riser, two
peaks instead of one are observed in the distribution curve at

z5.5 m for case A, which may be caused by the existence of

horizontal ows through cyclone inlets. In both cases, horizontal
particle velocity prole at z 2.5 m and z 3.5 m exhibits a steeper
peak than the rest areas.
Fig. 12 depicts the horizontal solids concentration prole in x
direction at different heights. It is observed from the curves that
the solids concentration decreases slightly with the increases in

Y. Jiang et al. / Chemical Engineering Science 109 (2014) 8597

93

performance of the distribution of solids concentration in the

upper region at z4.5 m and z 5.5 m.
4.5. Distribution of solid ux on the cross-section of cyclone inlets
Fig. 13 shows the horizontal distribution of particle mass ow
rate at the inlets of different cyclones. The x/L coordinate represents the dimensionless distance from the outer wall of the
inlet of cyclone. As can be seen, all the cyclone separators except
B have more particles passing by the outer wall of the inlet (x/
L  0).
From the above CPFD simulation results, it can be concluded
that the axis based symmetric arrangement for cyclone is better
than point arrangement in term of solids distribution as well as
ow elds in the CFB riser. However, further CFPD simulations as
well as ECT measurements are necessary to investigate the effect
of cyclone geometry shape and cyclone inlets angle on the gas
solid ow characteristics.

Fig. 8. Axial pressure distribution calculated by CPFD.

4.6. Distribution of solids in the standpipe

Table 5
Pressure drop of the riser (case A).

CPFD (Pa)
Experiment (Pa)
Relative error (%)

Dense region

Middle region

Dilute region

1689.8
3856.58
128.23

1273.62
522.1
59.0

1069.13
483.08
54.82

Table 6
Pressure drop of cyclones (case A).
Cyclone

CPFD (Pa)
Experiment (Pa)
Relative error (%)

585.39
1078.11
84.17

580.01
828.12
42.78

589.57
918.94
55.87

591.33
833.19
40.90

583.54
740.46
26.89

588.95
958.84
62.81

Figs. 14 and 15 give the reconstructed images in the cross

sections in the standpipe with different uidization ow rate from
ECT measurements. The solids concentration is in the range of
02%, 23%, 36% when the ow rate is 4500 m3/h, 5500 m3/h and
6500 m3/h respectively. As can be seen from those images, the
solid ow in the standpipe is a typical annular ow due to the
strong swirl effect in the cyclone separator. The solids concentration increases with the increase in uidization air through the
U-loop.
Fig. 16 shows the distribution of time-averaged solids concentration in the cross sections calculated by CPFD simulation
and ECT measurements. The simulation condition is 5500 m3/h
in both cases. As a result of un-symmetrical structure of the
arrangement of multi-cyclone separators, ow deviation in
cyclones is inevitable. Table 7 summarized the solids concentration in the six standpipes from the CPFD simulation and ECT
measurements. From Table 7 and Fig. 16, it can be seen that
particle ow has a non-uniform distribution. The maximum
relative error is 28.55% and 24.97% for case A and B respectively.
However, there is no obvious difference from the reconstructed
image for case A and B.
From the above simulation results analysis, it can be seen that
CPFD simulation can provide details of the gassolid hydrodynamics behavior in the full-loop of a circulating uidized bed
with six cyclone separators and predict the effect of operation
parameters on the circulating process. The results indicate
that the presented CPFD model combined with ECT measurements
can be applied to the design and optimization for a circulating
uidized bed. Furthermore, the ECT measurement results indicate
that ECT can not only provide the instantaneous parameter
distribution in a cross section, but also provide accurate timeaveraged solids concentration in well agreement with the CPFD
simulation.

5. Discussions and conclusions

Fig. 9. Time-averaged particle ux distribution in cyclones.

height. In the dense region of the riser, solids concentration near

the wall is higher than that near the center, exhibiting a coreannulus ow as mentioned above. However, there is no clear

The gassolid hydrodynamic behavior is one of the key issues

to scale up a circulating uidized bed (CFB) from small scale
to large scale with multi-cyclone separators. To investigate the
gassolid ow in a large scale CFB with six cyclone separators,
Barracuda CPFD is used to simulate the full-loop of gassolid ow
in a super-critical pressure circulating uidized bed with sixcyclone separators and rectangular shape of combustion chamber.

94

Y. Jiang et al. / Chemical Engineering Science 109 (2014) 8597

Fig. 10. Comparison of axial solids concentration prole. (a) Case A and (b) case B.

Fig. 11. Horizontal distribution of particle vertical velocity. (a) Case A and (b) case B.

Fig. 12. Horizontal distribution of solids concentration. (a) Case A and (b) case B.

A circular ECT sensor has been designed and used to measure the
solids concentration in the cross sections in the standpipe. The
main conclusions are as follows.
(1) From the CPFD simulation, the gassolid ow is non-uniform
among the six parallel cyclones. The solids concentration of
four cyclones at the corner of the chamber is higher than that

of the others. The location of cyclone as well as the inlet angle

of the cyclone need to be optimized based on the CPFD
simulation and ECT measurements.
(2) The ECT results show that the solids distribution in the cold
CFB model with different arrangement of cyclones can be
measured in different bed inventories and supercial velocities. The measurement results provide valuable information

Y. Jiang et al. / Chemical Engineering Science 109 (2014) 8597

95

Fig. 13. Solid ux along the x direction in the inlet of cyclone. (a) Case A and (b) case B.

4500 m3/h

5500 m3/h

6500 m3/h

No.1

No.2

No.3

No.4

No.5

No.6

Fig. 14. Solids concentration in the cross-section in the standpipe (case A).

for the scaling up of a super-critical pressure circulating

uidized bed with multi-cyclone separators.
(3) The gas phase is relatively uniform in the CFB riser as well as in
the cyclones. The pressure drops of cyclones accounts for a
large proportion of the whole boiler pressure drop. The solids
concentration in the CFB riser from the CPFD simulation agrees
well with pressure drop measurements except in the bottom
dense region.
(4) In terms of uniform distribution of solids ux in the cyclone
and ow elds in the CFB riser, axis based symmetric
arrangement for cyclone is better than point arrangement.

Circulating uidized bed with multi-cyclone separators is a

complex gassolid ow system, the design as well as scaling up is
extremely difcult and several key issues need to be addressed
before the commercial success of the system, including the
arrangement of the cyclone, the geometrical shape of the riser
and the dimensions of cyclone. Further work is necessary to
provide valuable information for the design and process optimization based on CFPD simulations and ECT measurements. In
additional, it is necessary to consider a hybrid model to estimate
the drag force for the whole-loop CPFD simulation to improve the
prediction error.

96

Y. Jiang et al. / Chemical Engineering Science 109 (2014) 8597

4500 m3/h

5500 m3/h

6500 m3/h

No.1

No.2

No.3

No.4

No.5

No.6

Fig. 15. Solids concentration in the cross-section in the standpipe (case B).

Fig. 16. Average solids concentration in the cross sections of standpipes by CPFD. (a) Case A and (b) case B.

Table 7
Solids concentration in the standpipes.
Cyclone

A
B
C
D
E
F

Case A

Case B

CPFD (%)

ECT (%)

Error (%)

CFPD (%)

ECT (%)

Error (%)

3.05
2.95
2.92
3.08
3.05
2.99

2.76
2.54
2.30
2.26
2.18
2.68

9.51
13.89
21.18
26.57
28.55
10.37

4.215
3.885
4.063
4.193
3.989
4.458

4.17
4.58
3.93
4.22
3.19
4.03

11.03
15.08
3.46
7.10
24.97
10.59

Y. Jiang et al. / Chemical Engineering Science 109 (2014) 8597

Nomenclature
Cd
Dp
F
f
g
l
m
Np
np
p
PS
Re
S
Sg
t
V
v
xp
x; y; z

Drag coefcient ()
Inter-phase drag function (s1)
Inter-phase momentum exchange rate per volume
(N=m3 )
Probability distribution function ()
Gravitational acceleration (m=s2 )
Location of pressure sensor point (m)
Particle mass (kg)
Total number of numerical particles in a cell ()
Total number of real particles in a numerical particle ()
Pressure (Pa)
Pressure constant (Pa)
Reynolds number ()
Interpolation operator ()
Gas source term (kg=m3 s)
Time (s)
Volume (m3)
Velocity (m=s)
Particle displacement (m)
Orthogonal directions (m)

Greek symbols

Density (kg=m3 )
Volume fraction ()
Viscous stress tensor (N2 =m2 )
Gas viscosity (kg=ms)
Constant number ()
Average voidage ()
Constant number ()
Pi ()

Subscripts
cp
g
i, j, k
n
p

Close packing limit

Gas phase
Coordinate direction
Time step
Particle phase

Acknowledgment
The authors would like to thank the National Natural Science
Foundation of China (No. 61072001) and Strategic Priority
Research Program Demonstration of Key Technologies for Clean
and Efcient Utilization of Low-rank Coal (No. XDA07030100) from
the Chinese Academy of Sciences for nancially supporting this
research.
References
Ahuja, G.N., Patwardhan, A.W., 2008. CFD and experimental studies of solids holdup distribution and circulation patterns in gassolid udized beds. Chem. Eng. J.
143, 147160.
Abbasi, A., Ege, P.E.G., de Lasa, H.I., 2011. CPFD simulation of a fast uidized bed
steam coal gasier feeding section. Chem. Eng. J. 174, 341350.
Andrews, M.J., O'Rourke, P.J., 1996. The multiphase particle-in-cell (MP-PIC) method
for dense particulate ows. Int. J. Multiph. Flow 22, 379402.

97

Armistead, T.F., Schwieger, R., Reina, P., 2002. Boiler technology pursues goals of
efciency and lower emissions-circulating uidized-bed equipment pushes size
limits. Power Ind. Powerpl., 2002
Auzerais, F.M., Jackson, R., Russel, W.B., 1988. The resolution of shocks and the effects of
compressible sediments in transient settling. J. Fluid Mech. 195, 437462.
Chen, C., Werther, J., Heinrich, S., Qi, H.Y., Hartge, E.U., 2013. CPFD simulation of
circulating uidized bed risers. Powder Technol. 235, 238247.
Chen, Y.M., Zhou, X.L., Zheng, C.H., Wang, L.H., Fang, M.X., Shi, Z.L., Luo, Z.Y., Cen, K.
F., 2008. Development of large-scale circulating uidized bed boiler. J. Power
Eng. 28, 817826.
Du, B., Warsito, W., Fan, L.S., 2005. ECT studies of gassolid uidized beds of
difference diameters. Ind. Eng. Chem. Res. 44, 50205030.
Dyakowski, T., Edwards, R.B., Xie, C.G., Williams, R.A., 1997. Application of
capacitance tomography to gassolid ows. Chem. Eng. J. 52, 20992110.
Dyakowski, T., Luke, S.P., Ostrowski, K.L., Williams, R.A., 1999. On-line monitoring of
dense phase ow using real time dielectric imaging. Powder Technol. 104, 287295.
Fan, Z., Goidich, S., Robertson, A., 2008. 800 MWe circulating uidized bed boiler
with 1300 1C superhigh steam. In: Proceedings of the 33rd International
Technical Conference on Coal Utilization and Fuel Systems June 15, 2008,
Clearwater, FL, USA.
Gidaspow, D., Ettehadieh, B., 1983. Fluidization in two dimensional beds with a jet:
Part II. Hydrodynamic Modeling Ind. Eng. Chem. Fund. 22, 193201.
Glicksman, l.R., Hyre, M., Woloshun, K., 1993. Simplied scaling relationships for
uidized-beds. Powder Technol. 77, 177199.
Hack, H., Hotta, A., Kettunen, A., Eriksson, T., 2008. Ultra-Supercritical CFB
Technology to Meet the Challenge of Climate Change, Power_Gen International
2008, Orlando, Florida, USA, December 2-4, 2008, pp 111.
Harris, S.E., Crighton, D.G., 1994. Solitons, solitary waves and voidage disturbances
in gas-uidized beds. J. Fluid Mech. 266, 243276.
Hartge, E.U., Ratschow, L., Wischnewski, R., Werther, J., 2009. CFD-simulation of a
circulating uidzied bed riser. Particuology 7, 283296.
Koornneef, J., Junginger, M., Faaij, A., 2007. Development of uidized bed combustion
an overview of trends, performance and cost. Prog. Energy Combust. Sci. 33, 1955.
Kunii, D., Levenspiel, O., 1991. Fluidization Engineering. 1991. ButterworthHeinemann, Boston p. 69
Lombardi, G., pagliuso, J.D., Jr. Goldstein, L., 1997. Performance of a tuyere gas
distributor. Powder Technol. 94, 514.
Lun, C.K.K., Savage, S.B., Jeffrey, D.J., Chepurniy, N., 1984. Kinetic theories for
granular ow: inelastic particles in coquette ow and slightly inelastic particles
in a general ow eld. J. Fluid Mech. 140, 223256.
Lv, J.F., Yu, L., Zhang, Y.J., Yue, G.X., Li, Z.Y., Wu, Y.X., 2007. A 600 MW super-high
circulating uidized bed boiler. J. Power Eng. 27, 497501.
Makkawi, Y.T., Wright, P.C., 2004. Electrical capacitance tomography for conventional uidized bed measurements-remarks on the measuring technique.
Powder Technol. 148, 42157.
Makkawi, Y.T., Wright, P.C., Ocone, R., 2006. The effect of friction and inter-particle
cohesive forces on the hydrodynamics of gassolid ow: a comparative analysis
of theoretical predictions and experiments. Powder Technol. 163, 6979.
Malcus, S., Cruz, E., Rowe, C., Pugsley, T.S., 2002. Radial solid mass ux proles in a
high-suspension density circulating uidized bed. Powder Technol. 125, 59.
Morin, J.X., 2003. Recent Alstom power large CFB and scale up aspects including
steps to superhigh. In: Proceedings of the 47th International Energy Agency
Workshop on Large Scale CFB. Zlotnicki, Poland.
Pain, C.C., Mansoorzadeh, S., Gomes, J.L.M., de Oliveira Cr, E., 2002. A numerical
investigation of bubbling gas solid uidized bed dynamics in 2-D geometries.
Powder Technol. 128, 5677.
Reh, L., 2003. Development potentials and research needs in circulating uidized
bed combustion. Particulology 1, 185200.
Snider, D.M., 2001. An incompressible three-dimensional multiphase particle-incell method for dense particle ows. J, Comput, Phys, 170, 523549.
Wang, H.G., Yang, W.Q., Dyakowski, T., Liu, S., 2006. Study of bubbling and slugging
uidized beds by simulation and ECT. AIChE J. 52, 30783087.
Wen, C.Y., Yu, Y.H., 1966. Mechanics of uidization. Chem. Eng. Prog. Symp. Ser. 62,
100111.
Xie, C.G., Huang, S.M., Hoyle, B.S., Thorn, R., Lenn, C., Snowden, D., Beck, M.S., 1992.
Electrical capacitance tomography for ow imaging: system model for development of image reconstruction algorithms and design of primary sensors. IEE
Proc. GCircuits Dev. Syst. 139, 8998.
Yang, W.Q., York, T.A., 1999. New AC-based capacitance tomography system.
IEE Proc.Sci. Meas. Technol. 146, 4753.
Yang, W.Q., Peng, L.H., 2003. Image reconstruction algorithms for electrical
capacitance tomography. Meas. Sci. Technol. 14, R1R13.
Zhang, N.B.L., Wang, W., Li, J.H., 2010. 3D CFD simulation of hydrodynamics of a
150 Mwe circulating uidized bed boiler. Chem. Eng. J. 162, 821828.
Zhang, N., Lu, B.N., Wang, W., Li, J.H., 2008. Virtual experimentation through 3D
full-loop simulation of a circulating uidized bed. Particuology 6, 529539.
Zhou, X.L., Cheng, L.M., Wang, Q.H., Luo, Z.Y., Cen, K.F., 2012. Non-uniform
distribution of gassolid ow through six parallel cyclones in a CFB system:
an experimental study. Particuology 10, 170175.