Fuel 192 (2017) 187–200

Contents lists available at ScienceDirect

Fuel
journal homepage: www.elsevier.com/locate/fuel

Full Length Article

Numerical investigation of particle transport hydrodynamics and coal

combustion in an industrial-scale circulating fluidized bed combustor:
Effects of coal feeder positions and coal feeding rates
Massoud Massoudi Farid a, Hyo Jae Jeong b, Keun Ho Kim c, Jongmin Lee d, Dongwon Kim d, Jungho Hwang a,⇑
a
Mechanical Engineering Department, Yonsei University, 50 Yonsei-ro, Seodaemun-gu, Seoul 03722, South Korea
b
Corporate R&D Institute, Doosan Heavy Industries & Construction, 10 Suji-ro 112beon-gil, Suji-gu, Yongin 16858, South Korea
c
Yeosu Power Plant, Korea South-East Power Corporation, 727 Yeosusandan-ro, Yeosu 59613, South Korea
d
Green Power Generation Laboratory, Korea Electric Power Corporation Research Institute, 105 Munji-ro, Yeoseong-gu, Daejeon 34056, South Korea

h i g h l i g h t s

Particle transport was simulated in an industrial scale CFB combustor in 3D.

Coal combustion was simulated in an industrial scale CFB combustor in 3D.

Dense discrete phase model (DDPM) with several UDFs was used for the simulation.

Two different coal feeder positions and coal feeding rates were investigated.

Solid volume fraction, pressure, gas temperature, and gas species were reported.

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

Article history: This study investigates particle transport hydrodynamics and coal combustion in an industrial-scale cir-
Received 26 September 2016 culating fluidized bed (CFB) combustor using the dense discrete phase model (DDPM). DDPM is an exten-
Received in revised form 6 December 2016 sion of the discrete phase model (DPM); however, unlike the standard formulation of DPM, DDPM
Accepted 9 December 2016
considers the solid volume fraction when solving the Navier–Stokes equations for the gas phase. In the
Available online 22 December 2016
DDPM, the kinetic theory of granular flows is used to calculate the particle interaction in the Eulerian
frame of reference. This interaction is then mapped to the particles in the Lagrangian frame of reference.
Keywords:
In this study, user defined functions (UDFs) were used to extend the ANSYS FLUENT original code. These
Circulating fluidized bed
Coal combustion
UDFs were used to reinject particles into to the combustor (cyclones were not modeled), calculate the
3D Computational fluid dynamics pressure drop, circulation rate, and combustor mass load control. Various operation indexes such as dis-
Dense discrete phase model tributions of gas temperature, solid volume fraction, pressure, and mass fractions of combustion products
Gas-particle hydrodynamics were displayed, and the selected indexes were compared with operating data obtained from a 340 MWe
CFB combustor located in Yeosu, South Korea. The effects of both coal feeder positions and coal feeding
rates on operation indexes were investigated.
Ó 2016 Elsevier Ltd. All rights reserved.

1. Introduction try is to produce energy using low-grade solid fuels in circulating

While worldwide interest in utilizing fluidized beds in various
industrial processes such as coal combustion, biomass gasification,
drying systems, granulation, catalysis, and coating has increased in
recent decades, a complete and comprehensive knowledge of the
phenomena occurring in fluidized beds is becoming increasingly
necessary. One of the main applications of fluidized beds in indus-
⇑ Corresponding author.
E-mail address: hwangjh@yonsei.ac.kr (J. Hwang).
fluidized beds (CFBs). A CFB, in its simplest form, consists of a riser
(in which solid particles are fluidized), a cyclone (by which parti-
cles are separated from gas), and a return leg (through which par-
ticles return to the riser). In addition to considerable combustion
efficiency, capturing sulfur emissions in a combustor using sor-
bents, and low nitrogen oxide emissions owing to low combustion
temperature are the main reasons for the dramatic growth in the
use of CFBs since the first CFB combustors were developed in the
late 1970s [1]. Moreover, CFBs are flexible to use various low-
grade fuels, including biomass and waste-derived fuels. Large-
scale CFB units with new combustion processes that utilize various

http://dx.doi.org/10.1016/j.fuel.2016.12.025
0016-2361/Ó 2016 Elsevier Ltd. All rights reserved.
188 M. Massoudi Farid et al. / Fuel 192 (2017) 187–200
fuels have been developed recently. Therefore, an effective model-
ing tool is required to study solid–gas hydrodynamics and the mix-
ing of different species in a combustor in order to optimize the
design of the combustor layout for performance and emissions.
Owing to the non-uniform feeding of fuel and air into the combus-
tor, fluidization and combustion are spatially non-uniform pro-
cesses, and hence need to be modeled three-dimensionally.
Considering the factors mentioned above, a numerical simulation
is an effective tool for determining different aspects of solid–gas
hydrodynamics and fuel combustion in CFBs three-dimensionally.
There are two approaches for modeling gas-particle interaction
numerically in dense beds: the Eulerian two fluid model (TFM)
approach and the Eulerian–Lagrangian approach. In the TFM
approach, solid and gas phases are both treated as interpenetrating
continua, and the kinetic theory of granular flows (KTGF) approach
is used to calculate particle interaction by solving an additional
equation for solids fluctuating energy or granular temperature.
The main disadvantage of the TFM approach is the long calculation
time required for evaluating the time-averaged solution. The con-
sideration of particle size distribution can significantly increase
this time, as an additional dispersed phase for each particle size
must be used [2]. In the Eulerian–Lagrangian approach, the gas
phase is treated as a continuum while the solid phase is solved
by tracking a large number of particles through the flow field cal-
culated from the Navier–Stokes equations for the gas phase. All the
forces applied on the particles are calculated to find their new posi-
tions and velocities. The solid phase can exchange momentum,
mass, and energy with the gas phase.
In the discrete phase model (DPM), one of the models for the
Eulerian–Lagrangian approach, a fundamental assumption is that
the solid phase occupies a very low volume fraction, which is
ignored in the Navier–Stokes equations for the gas phase. DPM
can be used as either one way (fluid affects the particle motion
only) or two ways coupled (particle motion affects the fluid motion
as well) [3]. Additionally, In DPM the interaction between particles
is neglected when solving the equation of motion for a particle.
According to Elghobashi [4] a four way coupling is required already
at even a low volume fraction of solids. Therefore, for modeling
systems with particles of a high volume fraction such as those
found in fluidized beds, DPM is inappropriate and instead a four
ways coupled approach (fluid affects particle motion, particle
affects the fluid motion, and particles affect each other [3]) is
needed.
To consider the interaction between particles DPM can be
extended for use in fluidized beds by considering the volume frac-
tion of the solid phase in the Navier–Stokes equations for the gas
phase. The volume fraction and velocity field of the solid phase
come from the Lagrangian tracking solution, and the interaction
between the solid particles is calculated at the Eulerian grid using
solid stress–strain tensor, which is calculated using the KTGF
approach. This method is considered under dense discrete phase
model (DDPM) in ANSYS FLUENT [5]. In general, DDPM has com-
mon roots with the multiphase particle in cell (MP-PIC) method.
The MP-PIC method is described in details in Refs. [6–10].
Several numerical simulations of reacting flow in fluidized beds
are found in the literature. However, most of them employed 2D
simulation, which cannot illustrate all features of the flow. Most
of the 3D studies have simulated lab scale or pilot scale units
and only a few have modeled industrial scale fluidized beds. Jung
and Gamwo [11] developed a 2D TFM in-house code that included
a reaction kinetics model for mimicking reactive flow behavior in a
bubbling fluidized bed fuel reactor of a chemical looping combus-
tor. Deng et al. [12] simulated fuel reactor flows and reactions in a
bubbling fluidized bed reactor of a chemical looping combustor
using the 2D TFM introduced in FLUENT. Papadikis et al. [13] stud-
ied the effects of different drag models on the fast pyrolysis of
wood in a 2D simulation of a lab-scale bubbling fluidized bed reac-
tor. The TFM in FLUENT was used to model the bubbling behavior
of sand, which was treated as continuum, and the DPM was used to
model discrete wood particles. Zhou et al. [14] applied a 2D TFM
with a drag coefficient correction based on the extended energy-
minimization multiscale (EMMS/matrix) model to simulate the
air-coal two-phase flow and combustion characteristics of a
50 kW CFB combustor (riser). Reactions during coal combustion
included moisture evaporation, devolatilization, volatile gas com-
bustion, char combustion, and char gasification. The model pre-
dicted the main features of complex gas–solid flow, including
cluster formation along the walls, and flow structures of upward
flow in the core and downward flow in the annular region. FLUENT
software was used for that simulation. Zhou et al. [15] utilized
their model to predict coal combustion characteristics in an
oxygen-fired CFB riser with wet flue gas recycling. Cornejo and
Farías [16] simulated the coal gasification process inside a pilot
fluidized-bed reactor (riser) using a 3D TFM, taking into account
drying, volatilization, combustion, and gasification processes. Xie
et al. [17] modeled the coal gasification process using a Lagrangian
approach for both sand and coal particles in the same geometry
used by Cornejo and Farías [16]. Adamczyk et al. [18] employed
DDPM with KTGF in simulating particle transport and combustion
in a pilot scale CFB, and showed that DDPM with KTGF can be used
for predicting particle transport in fluidized beds working under
the oxy-fuel combustion regime. Klimanek et al. [19] also used
DDPM with KTGF to model a small-scale experimental facility in
which coal was gasified in air and air/steam mixture. Adamczyk
et al. [2] applied DDPM with KTGF to three-dimensionally model
the dense gas–solid flow combined with a combustion process in
a large-scale industrial CFB combustor. Adamczyk et al. [20] also
modeled an industrial compact CFB combustor using DDPM-
with-KTGF. The impact of geometrical model simplification on pre-
dicted mass distribution and temperature profiles was investi-
gated. Moreover, some additional calculations were carried out to
check the influence of radiative heat transfer on predicted temper-
ature profiles within the CFB combustor. Distributions of species
mass fractions were not illustrated in either Adamczyk et al.
[2,20] studies.
In this study, a model was developed to investigate the tran-
sient 3D numerical simulation of air and sand particle hydrody-
namics, along with coal combustion, in an industrial-scale CFB
combustor. This study used the DDPM-with-KTGF approach to
model the interactions of both sand (as an inert bed material)
and coal particles with the gas phase. The ultimate goal was to
investigate the effects of coal feeder positions and coal feeding
rates, on particle transport hydrodynamics and coal combustion
in an existing 340 MWe CFB combustor that is being operated by
Korea South-East Power Corporation (KOSEP).
2. CFB combustor and coal
The 340 MWe CFB combustor of interest in this study is located
in Yeosu, South Korea. The CFB combustor has four cyclones, as
shown in Fig. 1(a). A mixture of hot sand particles, ash particles,
and limestone particles are fluidized by primary air, which enters
the combustor through nozzles located at the bottom of the com-
bustor. Secondary air enters through side nozzles to improve flu-
idization, promote the combustion of coal particles which are
injected into the combustor by coal feeders, and generate staged
combustion in order to reduce NOx emissions. Particles, which
are carried by a mixture of air and combustion products to the
top of the combustor, enter the cyclones and are separated from
the combustion products. The sand particles then return to the
combustor through return legs, while the combustion products
 M. Massoudi Farid et al. / Fuel 192 (2017) 187–200 189

Fig. 1. CFB located at the Yeosu power plant: (a) 3D view, (b) side view, (c) top view.

leave the cyclones through exit ports. Steam flows into combustor
membrane walls, evaporator wing wall tubes, superheater wing
wall tubes, and division wall tubes, and receives heat from the
gas-solid mixture inside the combustor.
Half of the combustor was simulated owing to computational
barriers. The simulated domain had a height of 42 m and a rect-
angular cross section of 8 m  17 m (see Fig. 1(b and c)). Thirteen
out of 27 superheaters, 2 out of 4 evaporator wing wall tubes, and
1 out of 2 division wall tubes were modeled as a total of 16 plates
inside the combustor. The coal silo, feeder pipes, and gas aeration
nozzles at the bottom of the combustor were not considered in
the simulation. Twenty-nine nozzles (used for secondary air
injection) were positioned around the lower part of the combus-
tor in two rows, and 5 nozzles were considered as coal feeders.
Instead of cyclones, a set of UDFs were used to model particle
recirculation.
190 M. Massoudi Farid et al. / Fuel 192 (2017) 187–200

Table 1 mum, maximum, mean diameters, and the spread parameter for
Proximate and ultimate analyses of coal. coal particles were 1 mm, 16 mm, 3.09 mm, and 1.120114, respec-
Proximate analysis wt.% Ultimate analysis wt.% tively, and for sand, they were 0.05 mm, 2 mm, 1 mm, and 1.98931,
Ash 10.15 Ash 11.86 respectively. The density of sand was constant as 2600 kg/m3. The
Fixed Carbon 39.34 Carbon 69.4 density of coal was initially 2200 kg/m3, and was assumed as con-
Moisture 14.42 Hydrogen 5.19 stant during evaporation in FLUENT. However, during devolatiliza-
Volatiles 36.09 Nitrogen 1.53 tion, FLUENT assumes that the diameter of coal particles does not
Sulfur 1.11 change. Since the mass of the particles is decreasing, the effective
Oxygen 10.91 density decreases. Data for the coal and sand particle size distribu-
Total 100 Total 100 tions were obtained from the Yeosu power plant. For coal particles,
the size range of 1–16 mm was divided to 10 characteristic diam-
eters while the size range of 0.05–2 mm was divided to 46 charac-
Indonesian Trafigura coal was used as fuel. Table 1 shows the teristic diameters for sand particles. For each characteristic
results of the ultimate and proximate analyses. Nitrogen and sulfur diameter, a group of particles with the same diameter was injected
were not considered in the simulation. The Rosin-Rammler mini- on the Lagrangian frame of reference in each time step. For coal,

Table 2
Chemical reactions.

Gas-phase reaction Reaction rate v 0i;r v 00i;r A [–] B [–] Ref.

CH2.66O0.35 + 1.49O2 ? Eddy-dissipation rate Minimum of 1 for 1 for CO2 4.0 0.5 [5]


CO2 + 1.33H2O CH2.66O0.35 and
Ri;r ¼ v 0i;r M w;i Aq ke minRe v 0 YM Re
Re;r w;Re and 1.49 for 1.33 for
P
O2 H2O
and Ri;r ¼ v 0i;r M w;i ABq ke PN P00
YP

j
v j;r Mw;j
Chemical reaction Ar [1/K/s] br [–] E [J/kmol] g0r;j [–] g00r;j [–]
rate

N
ðg0j;r þg00j;r Þ 2.119  1011 1 2.027  108 1 1
Ri;r ¼ ðv 00
i;r v 0
i;r Þ kf ;r P ½C j;r 
j¼1
br Er =RT
where kf ;r ¼ Ar T e

Char reaction C1 [m3/s/K0.75] C2 [kg/s] E [J/kmol]

C(s) + O2 ? CO2 Kinetic/diffusion dmp
¼ Ap pox DD00þR
R 9  1010 0.02 7.9  107 —
dt
limited-rate ½ðT þT Þ=2 0:75
where D0 ¼ C 1 p dgp and
R ¼ C 2 eðE=RT p Þ

Fig. 2. (a) Simulated geometry and (b) mesh.

 M. Massoudi Farid et al. / Fuel 192 (2017) 187–200 191

the flow rate of each group of particles having the same diameter Table 3
was calculated based on coal size distribution and flow rate of each Boundary conditions.

coal feeder. For sand, the flow rate of each group of particles having Hydrodynamic boundary conditions
the same diameter was calculated based on sand size distribution Primary air inlet 101 kg/s
and total required mass of sand (which should be injected during Secondary air inlet 54.2 kg/s
Return leg inlet 0.361 kg/s
the first 30 s of the simulation). Total coal feeding rate 73 t/h
Outlet pressure 91.5 Pa
Wall B.C Gas No-slip
3. Numerical methods Solid Reflect
Heat transfer boundary conditions
This study used a parallelized CFD software, ANSYS FLUENT, for Primary air temperature 482.15 K
the simulation. The simulation was done in an unsteady state. The Secondary air temperature 498.15 K
devolatilization process was simulated using a constant rate model Return Leg air temperature 1073.15 K
Wall temperature 632.15 K
(details are described in supplementary material). Particle swelling
Refractory walls Adiabatic
was taken into account during devolatilization. Volatile matter Division wall temperature 632.15 K
produced during devolatilization was assumed to be a CHO com-
Wing wall tube temperature 1st super heater 686.65 K
pound, the composition of which was calculated from ultimate 2nd super heater 766.65 K
and proximate analyses results as CH2.66O0.35. CH2.66O0.35 was then
reacted with O2 (see the first reaction presented in Table 2). Char
and gas phase reactions were modeled using the kinetic/diffusion except at the roof and at the conic section of the combustor (both
and the finite-rate/eddy-dissipation models, respectively. The are made of refractory materials). At each refractory wall, adiabatic
finite-rate/eddy-dissipation model compares both the finite rate boundary condition was applied. The details of boundary condi-
and the eddy-dissipation rate, and the slower rate is selected for tions can be found in Table 3. The time step used for this simula-
computing continuous phase reactions. Table 2 also shows the char tion was 0.1 s. Using such a large time step in the DDPM was
reaction used in this simulation. The kinetic parameters for the
char reaction were iteratively derived during the simulation in
the way that the exit O2 concentration was close to the real mea-
surement datum in Yeosu power plant. Then obtained kinetic
parameters were used to simulate the ‘after design’ condition. An
almost similar approach was used by Adamczyk et al. [2]. However,
they assumed that the injected coal particles were burned before
exiting the computational domain.
The Discrete Ordinate (DO) radiation model, together with the
cell-based Weighted-Sum-of-Gray-Gases Model (WSGGM) was
applied for the radiative properties of the gases. Particle-to-
particle radiation was considered in the simulation. The standard
k–e model was applied to capture turbulence flow, and gravita-
tional force was considered in the simulation. Heat transfers
between gas and solid particles and walls were obtained by solving
the energy equation. For information regarding governing equa-
tions, please see supplementary material.
For pressure-velocity coupling, the phased coupled SIMPLE
scheme was used. The Green-Gauss node-based method was used
to evaluate the gradient terms. Second-order upwind and QUICK
schemes were used for momentum and volume fraction discretiza-
tion, respectively, and a first-order upwind scheme was used for
the remaining equations. The stochastic tracking method was used
to consider the turbulent dispersion effect of particles.
Fig. 2 shows the geometry and mesh used for the simulation.
Owing to the high computational cost of calculating chemical reac-
tions, a mesh independency study was performed for non-reacting
flow at a constant temperature of 850 °C. Based on this mesh inde-
pendency study, around 1.6 million cells were used. This number
of cells was higher than that used by Adamczyk et al. [2], who car-
ried out a combustion simulation in an industrial-scale CFB with a
size almost similar to that used in this study. Average cell size was
around 0.15 m. Since attention was required to generate the mesh
with DDPM, a UDF was written to control mass balance between
the Lagrangian and Eulerian frames of reference. For more informa-
tion regarding this mesh and UDF, please see supplementary
material.
For all inlets, velocity inlet boundary condition was applied. For
all outlets, pressure outlet boundary condition was applied. No-slip
and reflect boundary conditions were applied at the walls for the
gas phase and solid phase, respectively. Furthermore, constant Fig. 3. Coal feeder positions and rates for the half-combustor: (a) before design
temperature boundary condition was applied at all the walls, change and (b) after design change.
192 M. Massoudi Farid et al. / Fuel 192 (2017) 187–200

made possible by applying a node-based averaging procedure for

the solid phase, as explained by Adamczyk et al. [2]. A set of UDFs
was used for the reinjection of particles into the combustor (note
that the cyclones were not modeled), and for the calculation of
pressure drop and circulation rate. Measurements taken in the
Yeosu power plant indicated that the temperature of recirculating
solids exiting the combustor was around 800 °C, the same temper-
ature as that of the exiting combustion products. Hence, the tem-
perature of reinjected solids was fixed at 800 °C in reinjection
UDFs. For more information regarding these UDFs, please see sup-
plementary material.

4. Problems in the Yeosu power plant

Fig. 4. Pressure drop.
Fig. 5. (a) General behavior of circulating fluidized bed [22]; (b) solid volume
fraction, X–Z plane; (c) solid vertical velocity (m/s), X–Z plane; (d) axial profile of
cross-sectionally averaged and time-averaged solid volume fraction.
The Yeosu power plant reported two main problems when coal
feeding rates and feeder positions were arranged as shown in
Fig. 3a (the feeding rates were 9, 10, 20, 20, and 14 tons/h for coal
feeders a, b, c, d, and e, respectively). The first problem was that
erosion occurred at the roof of the combustor near the side walls.
The second problem was that the heat absorption to wing wall
tubes was low. To solve these problems, coal feeder positions
and feeding rates were changed to those shown in Fig. 3b. To avoid
roof erosion near the side walls, coal feeder ‘‘a” was moved to the
middle of the combustor, as were the secondary air nozzles beside
coal feeder ‘‘a”. To obtain more uniform coal feeding rates through
the cross-section of the combustor, and to obtain better heat
absorption by wing wall tubes, coal feeding rates were changed
to 16, 16, 16, 9, and 16 tons/h for coal feeders a, b, c, d, and e,
respectively. Numerical simulation was performed to investigate
the effect of these changes on the combustor’s operating
conditions.

B efore des ign change A fter de sig n ch ange

Z=6m

Z = 20 m

Fig. 6. Solid volume fraction and Z velocity (m/s) graphs at different heights.
 M. Massoudi Farid et al. / Fuel 192 (2017) 187–200
5. Results and discussion
Numerical simulation was carried out until t = 600 s, when the
flow condition was in a stable situation. Injections of coal and air
were started at t = 0 s. The sand particles were gradually injected
into the combustor during the first 30 s of the simulation. To obtain
a semi-steady-state result, the last 60 s (540 6 t 6 600 s) of the
simulation were used for time averaging. All instantaneous results
were presented at t = 600 s. Because the symmetry boundary con-
dition was used for the middle of the combustor, the results
obtained for one half of the combustor were mirrored to the sym-
metry interface to obtain the full combustor contours. The compu-
tational time for each case was approximately 5 months on a
computer with 8 dual core i7 processors, and 64 GB operating
memory.
5.1. Gas–solid hydrodynamics
Fig. 4 shows that the change in coal feeder positions and feeding
rates did not affect the pressure distribution. Measurement data
were only available for the before-design-change condition, and
it was found that the numerically calculated pressure values were
generally in good agreement with the measured pressure values.
The calculated values were the averaged values for 540 6 t 6 600 s.
The pressure gradient was high at the bottom, and comparatively
low at the top of the combustor. This trend qualitatively agrees
with that observed in Zhang et al. [21].
(a)
Before design
change
After design
change
Fig. 7. Solid vertical velocity contours: (a) central X–Z plane, (b) Y–Z planes, (c) X–Y planes.
193
Fig. 5a shows the schematic of solid particles behavior in a CFB
bed, which is explained by Grace et al. [22]. This schematic illus-
trates the presence of a dense bottom coexisting with a dilute
top in the combustor chamber, an annulus near the walls in which
the solid-volume fraction is high, a core in which the solid volume
fraction is low, and some particle clusters in the middle of the
domain. Solid particles flow downward in the annulus region and
upward in the core zone, implying that dense particle clusters
are mainly formed near the walls of the combustor and move
downwards, while dilute suspensions mainly flow upwards in
the core zone of the combustor.Numerical simulation was carried
out for the configurations shown in Fig. 3a (the before-design-
change case). Fig. 5b shows our calculation results for the
solid-volume fraction distribution on four vertical (X–Z plane)
cross-sections. The results were obtained at t = 600 s. All of the
solid particle behaviors described in Fig. 5a are clearly seen in
Fig. 5b. Fig. 5c shows the distribution of the solid vertical velocity
on the vertical (X–Z plane) cross-sections. Comparing Fig. 5b and c
indicates that the solid vertical velocity was mainly positive in
regions of low volume fraction (see region I) and negative in
regions of high volume fraction (see region II). Hence, the negative
velocity was due to particle clustering. Generally, the solid volume
fraction in a combustor is high at the bottom and low at the top
(the so-called ‘‘S-Shaped Profile” [23,24]). Fig. 5d shows the
cross-sectionally averaged values of the solid volume fraction at
different heights. The results were averaged over 540 6 t 6 600 s.
The results show that the solid volume fraction was high (20%)
(b) (c)
194 M. Massoudi Farid et al. / Fuel 192 (2017) 187–200

(a) (b) (c)

Before design

change

After design

change

Fig. 8. Gas vertical velocity contours: (a) central X–Z plane, (b) Y–Z planes, (c) X–Y planes.

Table 4 at the bottom and drastically decreased at heights of 1–3 m. At a

Gas velocity. height of around 6 m, the steep slope evened out and a relatively
H (m) Averaged velocity (m/s) constant solids concentration was attained.Fig. 6 shows volume
fraction contours for the X–Y plane at two different heights (Z = 6
Before design change After design change
and 20 m). Fig. 6 also shows plots of solid volume fraction and solid
0.4 6.2 6.7
Z velocity along two horizontal lines (Y = 3.9 and 12.25 m) in the X
3 5.7 6.4
6 5.1 5.4
direction at two different heights (Z = 6 and 20 m). The results
10 4.9 5.5 were averaged over 540 6 t 6 600 s. The results show the so-
20 4.9 5.4 called ‘‘core–annulus structure” (higher solids concentration near
35 4.1 4.1 the walls than in the center), resulting in solid back-mixing [24].
After the design change, at Z = 6 m, in the entire core region, no sig-
nificant change was observed in solid volume fraction plots along
Table 5
Solid circulation rate. the Y = 3.9 m and Y = 12.25 m lines. In the annulus region (near
X = 4.2 m (front wall) and X = 4.2 m (rear wall)) along the
Cyclone Solid circulation rate (kg/s)
Y = 3.9 m line, solid volume fractions were higher for the after-
A B design-change case (0.0097 for X = 4.2 m and 0.013 for
Before design change 376 344 X = 4.2 m) than for the before-design-change case (0.0092 for
After design change 381 390 X = 4.2 m and 0.0099 for X = 4.2 m). In the annulus region along

Table 6
Heat absorption.

Design Before design Error After design Improvement

value change (%) change (%)
Evaporator tubes (including evaporator wing wall tubes, division walls, and combustor 147 163 +11 185 13.5
walls) (MWt)
Secondary super-heaters wing wall tubes (MWt) 41 41 0 52 26.8
Final super-heaters wing wall tubes (MWt) 52 32 38 46 44
Total (MWt) 240 237 1.25 284 20
 M. Massoudi Farid et al. / Fuel 192 (2017) 187–200 195

the Y = 12.25 m line, solid volume fraction was lower (0.0094) for After the design change, at Z = 20 m, in the entire core region,
the region near X = 4.2 m in the after-design-change case than also no significant change was observed in the solid volume frac-
the solid volume fraction (0.011) in the before-design-change case. tion plots along the Y = 3.9 m and Y = 12.25 m lines. In the annulus
However, no significant change was observed in the region near region along the Y = 3.9 m line, solid volume fractions were higher
X = 4.2 m. for the after-design-change case (0.0044 for X = 4.2 m and 0.0058
for X = 4.2 m) than for the before-design-change case (0.0032 for
X = 4.2 m and 0.0042 for X = 4.2 m). Also, in the annulus region
along the Y = 12.25 m line, solid volume fractions were higher for
the after-design-change case (0.0048 for X = 4.2 m and 0.0051
for X = 4.2 m) than for the before-design-change case (0.0032 for
X = 4.2 m and 0.0043 for X = 4.2 m).
From the contours inserted into the background of plots of
Before solid-volume fraction and solid Z-velocities, the core-annulus
structure was more apparent in the X–Y plane at Z = 6 m (below
design
the wing wall tube level) than in the X–Y plane at Z = 20 m (at
change the wing wall tube level). At Z = 20 m, the core-annulus structure
was affected slightly by the wing wall tubes, but maintained its
general structure. Moreover, a local core-annulus structure was
observed in between the wing wall tubes, implying that the solids
concentration was higher near the wing wall tubes than in the

After

design

change

Fig. 9. Solid concentrations on the roof of the combustor. Fig. 11. Axial distribution of gas temperature averaged at each cross section.

0 < dp < 200 µm 200 < dp < 400 µm 400 < dp < 600 µm 600 < dp < 1000 µm 1000 < dp < 2000 µm

Before
design
change

After
design
change

Fig. 10. Sand particle positions colored by volume fraction.

196 M. Massoudi Farid et al. / Fuel 192 (2017) 187–200
regions between them. In addition, solids concentration was high
near the division wall. Comparing the before-design-change and
after-design-change results at Z = 6 m, it is obvious that the
solid-volume fraction was lower for the after-design-change case
in the middle of the combustor, near the front wall. However, the
solid volume fraction was higher near the side walls. This is due
to the movement of coal feeder ‘‘a” and the secondary air nozzles
(located beside coal feeder ‘‘a”) to the middle of the combustor,
which resulted in the washing of sand particles from the middle
of the combustor into the side walls (see Fig. 3). At Z = 20 m, a more
intense local core-annulus structure was observed between wing
wall tubes for the after-design-change case. In addition, the solid
volume fraction was higher in the middle of the combustor.
The core–annulus structure can be confirmed by Z-velocity
curves showing a downward movement of solids in the near-wall
region, while solids rise in the center, as reported in previous
works [21,23–31]. At Z = 6 m, in the entire core region, the solid
Z velocity along the line Y = 3.9 m was lower for the after-design-
change case than for the before-design-change case, especially
for X > 0. Along the Y = 12.25 m line, however, the opposite trend
was observed. The solid Z velocity in the after-design-change case
was higher in the core region, especially for X > 0. At this level
(Z = 6 m) in the annulus region, the falling velocities along the
Y = 3.9 m line, were higher for the after-design-change case
(5 m/s for X = 4.2 m and 4.6 m/s for X = 4.2 m) than the falling
velocities for the before-design-change case (4 m/s for
X = 4.2 m and 4.2 m/s for X = 4.2 m). In the annulus region along
the Y = 12.25 m line, no significant change was observed for the
region near X = 4.2 m; however, the falling velocity was higher
(a)
Before design
change
After design
change
Fig. 12. Gas temperature contours: (a) central X–Z plane, (b) Y–Z planes, (c) X–Y planes.
near X = 4.2 m for the after-design-change case (3.6 m/s) than
that for the before-design-change case (4.2 m/s).
In the core region at Z = 20 m, the solid Z velocity along the line
Y = 3.9 m was lower for X > 0 and higher for X < 0. However, along
the Y = 12.25 m line, the opposite trend was observed; the solid Z
velocity was higher for X > 0 and lower for X < 0. In the annulus
region along the Y = 3.9 m line, the falling velocity was higher
(4.2 m/s) for the after-design-change case than the falling veloc-
ity (2 m/s) for the before-design-change case for the near-wall
region (X = 4.2 m). However, no significant change was observed
for the near-wall region (X = 4.2 m) in the after-design-change case
compared to the before-design-change case. In addition, no signif-
icant change was observed in the annulus region along the
Y = 12.25 m line in the after-design-change case compared to the
before-design-change case.
Fig. 7 shows contours of 60 s time-averaged solid Z (vertical)
velocity. Negative velocities were observed near the walls and pos-
itive velocities were observed in the middle of the combustor. After
changing the coal feeding rates and coal feeder positions (the after-
design-change case), the solid vertical velocities decreased near the
side walls and increased in the middle of the combustor. These
conditions led to higher heat transfer to the upper part of the com-
bustor (see Table 6 and Fig. 16). Fig. 8 shows contours of 60 s time-
averaged gas vertical velocity. In the after-design-change case, gas
velocities also increased in the middle of the combustor. Further-
more, the magnitude of the average velocity at each horizontal
(X–Y plane) cross-section also increased (see also Table 4), result-
ing in higher heat transfer to the upper part of the combustor (see
also Fig. 16).
(b) (c)
 M. Massoudi Farid et al. / Fuel 192 (2017) 187–200
(a)
Before design
change
After design
change
Fig. 13. O2 mass fraction contours: (a) central X–Z plane, (b) Y–Z planes, (c) X–Y planes.
Table 5 shows 60 s time-averaged values of solid circulation
rates for cyclones A and B. Changing the coal feeder positions to
the middle of the combustor dramatically increased the solid circu-
lation rate of cyclone B, because more coal was fed to the middle of
the combustor and combustion zone was therefore located in the
middle of the combustor. These conditions also improved heat
transfer to the upper part of the combustor, because higher circu-
lation rate means that more particles go to the upper part of the
combustor, which results in higher heat transfer from these parti-
cles to that combustor region (see Fig. 16).
As stated in Section 4, one of the problems in the Yeosu power
plant was that erosion occurred at the roof of the combustor near
the side walls. Fig. 9 shows contours of 60 s time-averaged solid
concentrations on the roof of the combustor. In the after-design-
change case, the solid concentrations decreased on the roof of
the combustor, near the side walls, which could be helpful for
decreasing erosion near the side walls.
Fig. 10 shows sand particle positions for different size ranges,
colored by size range volume fraction. An additional UDF was
developed to calculate the volume fraction of each size range. As
expected, larger particles remained at the bottom of the combus-
tor, while smaller particles were fluidized and entered the cyclone.
The maximum volume fraction for all size ranges was observed at
the bottom of the combustor.
197
(b) (c)
5.2. Combustion
As coal particles are injected into the combustor, heat and
momentum are transferred from the hot fluidized sand particles
to the coal particles. Coal devolatilization is then started, followed
by combustion.
Fig. 11 shows the effects of design change on axial temperature
distribution (averaged at each cross-section). Before the design
change, the temperature was 1193 K at a location just above the
primary air entrance (Z = 0.4 m), decreased because of heat trans-
fer to the combustor wall tubes, and reached 1159 K at Z = 10 m.
At positions above Z = 10 m, the temperature decreased along the
axial distance because of heat transfer from the bulk gas to the
wing wall tubes and combustor wall tubes, and reached 1062 K
at Z = 35 m. Fig. 11 also shows the measured data for the before-
design-change condition at the bottom and top of the combustor.
The simulation results at these two points were very close to the
measured data, and the error at the top of the combustor was
below 2%, which confirms the accuracy of the simulation. After
the design change, the combustion zone was moved to the center
of the combustor and the temperature increased, which can also
be seen in Fig. 12, which shows contours of 60 s time-averaged
gas temperature. It is interesting that after the design change,
the temperature drop increased at positions higher than Z = 17 m
198 M. Massoudi Farid et al. / Fuel 192 (2017) 187–200

(a) (b) (c)

Before design

change

After design

change

Fig. 14. CO2 mass fraction contours: (a) central X–Z plane, (b) Y–Z planes, (c) X–Y planes.

Fig. 15. Axial distributions of O2 and CO2 mass fractions averaged at each cross section.

(the height where the wing wall tubes are located), implying that
that more heat was transferred above this level.
Fig. 13 shows mass fraction contours for O2 species. The con-
centration of O2 was high at the bottom of the combustor and
decreased in the axial direction owing to the conversion to CO2
by oxidation. After the design change, the combustion zone was
moved to the central region of the combustor, where most of
the O2 was consumed. Hence, the amount of CO2 increased in
the central region of the combustor (see Fig. 14). CO2 concentra-
tions in both cases were high both near the coal feeders and in
the upper portion of the combustor. Fig. 15 shows axial distribu-
tions of O2 and CO2 mass fractions. In both cases, the O2 mass frac-
tion decreased and CO2 mass fraction increased with the
increasing height of the combustor. Although the combustion zone
was moved to the central region of the combustor in the after
design change case, the overall distribution of O2 and CO2 along
the axial direction was similar to the distribution observed in
the before-design-change case. The O2 concentration at the exit
 M. Massoudi Farid et al. / Fuel 192 (2017) 187–200
Before design
change
After design
change
Fig. 16. Details of heat absorption for different parts of the combustor.
of the combustor for ‘before design change’ was 2.5% and for ‘after
design change’ was 1.8%. The CO2 concentration at this location for
‘before design change’ was 25% and for ‘after design change’ was
26%.
As stated in Section 4, another problem in the Yeosu power
plant was that the heat absorption by wing wall tubes was low.
Fig. 16 shows details of the heat transfer to different parts of the
combustor. After the design change, total heat absorption
increased by 20% (see Table 6). Heat absorption to the evaporator
tubes (including the evaporator wing wall tubes, division walls,
and combustor walls) improved by 13.5%, and heat absorption to
the secondary and final superheater wing wall tubes improved
by 26.8% and 44%, respectively (see Table 6). Table 6 also compares
the calculation values for the before-design-change case with the
design values. Overall, the numerical results are in good agreement
with the design values. The maximum error occurred at the final
superheater wing wall tubes.
199
6. Conclusions
Three-dimensional numerical simulation of air and sand parti-
cle hydrodynamics and coal combustion was investigated in an
industrial-scale CFB combustor using DDPM for two different coal
feeder positions and rates (before-design-change and after-design-
change cases) and the following results were obtained:
(1) The distributions of gas temperature, solid volume fraction,
pressure, and mass fractions of combustion products were
successfully predicted, and both the pressure drop along
the axial direction and the bottom and exit gas temperatures
for the before-design-change case agreed with real operating
data.
(2) After the design change, solid concentrations on the roof of
the combustor and near the side walls decreased, which
could be helpful for decreasing erosion in these regions.
200 M. Massoudi Farid et al. / Fuel 192 (2017) 187–200
(3) After the design change, heat transfer to the upper part of
the combustor (including wing wall tubes) improved.
Acknowledgment
This work was supported by the Korea Institute of Energy Tech-
nology Evaluation and Planning (KETEP) and the Ministry of Trade,
Industry & Energy (MOTIE) of the Republic of Korea (No.
20153010102030).
Appendix A. Supplementary material
Supplementary data associated with this article can be found, in
the online version, at http://dx.doi.org/10.1016/j.fuel.2016.12.025.
References
[1] Koornneef J, Junginger M, Faaij A. Development of fluidized bed combustion—
an overview of trends, performance and cost. Prog Energy Combust Sci
2007;33:19–55.
[2] Adamczyk WP, We˛cel G, Klajny M, Kozołub P, Klimanek A, Białecki RA.
Modeling of particle transport and combustion phenomena in a large-scale
circulating fluidized bed boiler using a hybrid Euler-Lagrange approach.
Particuology 2014;16:29–40.
[3] Loth E, Tryggvason G, Tsuji Y, Elghobashi S, Crowe CT, Berlemont A, et al. 13.1
overview of multiphase modeling; 2006.
[4] Elghobashi S. An updated classification map of particle-laden turbulent flows.
In: IUTAM symposium on computational approaches to multiphase
flow. Springer; 2006. p. 3–10.
[5] Fluent A. Version 14.0. 0; Finite volumes software for the calculation of flow
fields through the simulation within a virtual environment. Canonsburg (PA,
USA): ANSYS Inc; 2011.
[6] Weng M, Nies M, Plackmeyer J. Computer-aided optimisation of gas-particle
flow and combustion at the Duisburg circulating fluidised bed furnace. VGB
Powertech 2011;91.
[7] Snider D, O’Rourke P, Andrews M. Sediment flow in inclined vessels calculated
using a multiphase particle-in-cell model for dense particle flows. Int J Multiph
Flow 1998;24:1359–82.
[8] Snider D. An incompressible three-dimensional multiphase particle-in-cell
model for dense particle flows. J Comput Phys 2001;170:523–49.
[9] Li F, Song F, Benyahia S, Wang W, Li J. MP-PIC simulation of CFB riser with
EMMS-based drag model. Chem Eng Sci 2012;82:104–13.
[10] Andrews M, O’rourke P. The multiphase particle-in-cell (MP-PIC) method for
dense particulate flows. Int J Multiph Flow 1996;22:379–402.
[11] Jung J, Gamwo IK. Multiphase CFD-based models for chemical looping
combustion process: fuel reactor modeling. Powder Technol 2008;183:401–9.
[12] Deng Z, Xiao R, Jin B, Song Q. Numerical simulation of chemical looping
combustion process with CaSO4 oxygen carrier. Int J Greenhouse Gas Control
2009;3:368–75.
[13] Papadikis K, Gu S, Fivga A, Bridgwater AV. Numerical comparison of the drag
models of granular flows applied to the fast pyrolysis of biomass. Energy Fuels
2010;24:2133–45.
[14] Zhou W, Zhao C, Duan L, Qu C, Chen X. Two-dimensional computational fluid
dynamics simulation of coal combustion in a circulating fluidized bed
combustor. Chem Eng J 2011;166:306–14.
[15] Zhou W, Zhao C, Duan L, Liu D, Chen X. CFD modeling of oxy-coal combustion
in circulating fluidized bed. Int J Greenhouse Gas Control 2011;5:1489–97.
[16] Cornejo P, Farías O. Mathematical modeling of coal gasification in a fluidized
bed reactor using a Eulerian granular description. Int J Chem React Eng 2011;9.
[17] Xie J, Zhong W, Jin B, Shao Y, Huang Y. Eulerian-Lagrangian method for three-
dimensional simulation of fluidized bed coal gasification. Adv Powder Technol
2013;24:382–92.
[18] Adamczyk WP, Kozołub P, We˛cel G, Klimanek A, Białecki RA, Czakiert T.
Modeling oxy-fuel combustion in a 3D circulating fluidized bed using the
hybrid Euler-Lagrange approach. Appl Therm Eng 2014;71:266–75.
[19] Klimanek A, Adamczyk W, Katelbach-Woźniak A, We˛cel G, Szle˛k A. Towards a
hybrid Eulerian-Lagrangian CFD modeling of coal gasification in a circulating
fluidized bed reactor. Fuel 2015;152:131–7.
[20] Adamczyk WP, Kozołub P, Klimanek A, Białecki RA, Andrzejczyk M, Klajny M.
Numerical simulations of the industrial circulating fluidized bed boiler under
air-and oxy-fuel combustion. Appl Therm Eng 2015;87:127–36.
[21] Zhang N, Lu B, Wang W, Li J. 3D CFD simulation of hydrodynamics of a
150 MWe circulating fluidized bed boiler. Chem Eng J 2010;162:821–8.
[22] Grace J, Avidan A, Knowlton T. Circulating fluidized beds. Chapman & Hall
(UK): Blackie Academic & Professional; 1997.
[23] Rhodes M, Sollaart M, Wang X. Flow structure in a fast fluid bed. Powder
Technol 1998;99:194–200.
[24] Chu K, Yu A. Numerical simulation of complex particle–fluid flows. Powder
Technol 2008;179:104–14.
[25] Zhang W, Johnsson F, Leckner B. Fluid-dynamic boundary layers in CFB boilers.
Chem Eng Sci 1995;50:201–10.
[26] Zhang M, Qian Z, Yu H, Wei F. The solid flow structure in a circulating fluidized
bed riser/downer of 0.42-m diameter. Powder Technol 2003;129:46–52.
[27] Zhang H, Huang WX, Zhu JX. Gas-solids flow behavior: CFB riser vs. downer.
AIChE J 2001;47:2000–11.
[28] Kim SW, Kirbas G, Bi H, Lim CJ, Grace JR. Flow structure and thickness of
annular downflow layer in a circulating fluidized bed riser. Powder Technol
2004;142:48–58.
[29] Hartge E-U, Ratschow L, Wischnewski R, Werther J. CFD-simulation of a
circulating fluidized bed riser. Particuology 2009;7:283–96.
[30] Grace JR, Issangya AS, Bai D, Bi H, Zhu J. Situating the high-density circulating
fluidized bed. AIChE J 1999;45:2108–16.
[31] Chen Y, Williams C. FCC technology-recent advances and new challenges. In:
8th int conference on circulating fluidized beds, Hangzhou, China. p. 26–40.