You are on page 1of 11

Fuel 90 (2011) 728–738

Contents lists available at ScienceDirect

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

Three-dimensional modelling of in-furnace coal/coke combustion in a blast furnace


Y.S. Shen a, B.Y. Guo a, A.B. Yu a,⇑, P.R. Austin b, P. Zulli b
a
Lab for Simulation and Modelling of Particulate Systems, School of Materials Science and Engineering, University of New South Wales, Sydney, NSW 2052, Australia
b
BlueScope Steel Research, P.O. Box 202, Port Kembla, NSW 2505, Australia

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

Article history: A three-dimensional mathematical model of the combustion of pulverized coal and coke is developed.
Received 18 January 2010 The model is applied to the region of lance-blowpipe-tuyere-raceway-coke bed to simulate in-furnace
Received in revised form 19 August 2010 phenomena of pulverized coal injection in an ironmaking blast furnace. The model integrates not only
Accepted 23 August 2010
pulverized coal combustion model in the blowpipe-tuyere-raceway-coke bed but also coke combustion
Available online 12 October 2010
model in the coke bed. The model is validated against the measurements under different conditions.
The comprehensive in-furnace phenomena are investigated in the raceway and coke bed, in terms of flow,
Keywords:
temperature, gas composition, and coal burning characteristics. The underlying mechanisms for the
Pulverized coal injection
Mathematical modelling
in-furnace phenomena are also analysed. The simulation results indicate that it is important to include
Blast furnace lower zone recirculation region in the raceway and the coke bed reactions for better understanding in-furnace
In-furnace phenomena. The model provides a cost-effective tool for understanding and optimizing the in-furnace
flow-thermo-chemical characteristics of the PCI operation in full-scale blast furnaces.
Ó 2010 Elsevier Ltd. All rights reserved.

1. Introduction lower part of a BF, especially in terms of coal burnout and gas spe-
cies distributions.
Blast furnace (BF) is the most widely used technology to pro- In a real BF, actual measurements of such in-furnace phenomena,
duce iron, where coke is the dominant fuel in the process. One of especially the LBTRC region, are extremely difficult due to the severe
the most significant recent improvements in BF operation is pul- practical environment (high temperature, high pressure etc.). The
verized coal injection (PCI), where pulverized coal works as an laboratory experimental studies via replicating the in-furnace
auxiliary fuel to partly replace relatively expensive coke. In this phenomena related to PCI process are laborious and very expensive.
process, pulverized coal is injected via tuyere and then combusts As a result, only a few such attempts have been reported in the liter-
in the raceway cavity and the surrounding coke bed (Fig. 1). The ature [6–9] and they are mainly on gas compositions along tuyere
PCI rate has increased due to economic, operational and environ- axis in actual BFs [7,8] or pilot-scale experiments [6,9].
mental benefits. This has resulted in reduced coke consumption As an alternative, a mathematical approach provides an efficient
and thus production costs; enhanced furnace stability; greater flex- method in investigating the in-furnace phenomena of PCI opera-
ibility in BF operation; and reduced overall emissions from steel tion [5,10,11]. A brief summary of early PCI models (one- and
plants [1,2]. In a modern BF, over 200 kg of coal per tonne of hot two-dimensions) can be found elsewhere [1]. These models em-
metal (kg/t-HM) can be injected; and over 300 kg/t-HM is under ployed numerous assumptions and some important operational
test [3]. However, as the PCI rate increases to a higher level, the features were not included, especially relating to the spatial repre-
coal combustibility will be reduced to a certain degree. As a result, sentation of the process. Three-dimensional (3D) modelling is
this could have an adverse effect on the in-furnace operation. Be- more reliable for practical problems. In this connection, some 3D
cause more unburnt char particles are swept into the coke bed, numerical studies of coal combustion were reported [2,12,13],
the permeability of the surrounding coke bed could deteriorate without including the coke bed region. To date, few 3D integrated
resulting in the improper distribution of gas flow and composition models of coal/coke combustion of PCI operation were found in the
(Fig. 1) [4]. Higher coal burnout and proper gas species distribution literature [14–16]. Gu et al. [15] described a 3D Eulerian–Eulerian
is necessary for furnace stability and cost reduction [5,6]. There- model of pulverized coal combustion for the tuyere-raceway-
fore, it is important to understand the in-furnace aerodynamic coke bed region. But coke reactions were not considered. More-
and physicochemical behaviours for the whole region of lance- over, only one coal size group was considered. In practice, the
blowpipe-tuyere-raceway-surrounding coke bed (LBTRC) in the coal size range is large, typically 5–300 lm. The Eulerian–Eulerian
approach needs multiple phases and then multiple sets of govern-
⇑ Corresponding author. Fax: +61 2 93855956. ing equations for the multiple size groups, leading to expensive
E-mail address: a.yu@unsw.edu.au (A.B. Yu). computation. Nogami et al. [16] reported a 3D transient-state

0016-2361/$ - see front matter Ó 2010 Elsevier Ltd. All rights reserved.
doi:10.1016/j.fuel.2010.08.030
Y.S. Shen et al. / Fuel 90 (2011) 728–738 729

Nomenclature

A1, A2 pre-exponential factors of devolatilization reactions, s1 np particle number per unit volume, m3
Ac pre-exponential factors in Gibb (m s1 K1)/Field Nu Nusselt number
(kg m2 s1) models p pressure, Pa
Ap particle projected area, m2 P local pressure
As constant in Gibb model, 0.0004 PA atmospheric pressure
B coal burnout q heat transfer from a particle, W
C0 mass of raw coal, kg R resistance to flow in porous media (R = 0 for cavity)
C1, C2 turbulent model constants rp particle radius, m
CD drag coefficient ri reaction rate of gas species i, mol m3 s1
Cp particle heat capacity, J kg1 k1 Re Reynolds number
D external diffusion coefficient of oxygen in Gibb model, SU source terms due to the interaction between the gas
m2 s1 phase and the particle phase/coke bed
d particle diameter, m T temperature, K
d0 particle diameter at the start of devolatilization, m Tc activation energy in Gibb/Field model, K
Dref dynamic diffusivity in Field/Gibb model, 1.8  105 Tref reference temperature, 293 K
kg m1 s1 Ts constant in Gibb model, 6240 K
e void fraction of char particles U mean (true) velocity of gas, m s1
E1, E2 activation energy of devolatilization reactions, K u, v, w gas velocity components, m s1
fD drag force from a particle, N vi stoichiometric coefficient of species i.
H enthalpy, J kg1 Wi reaction rate of species i (per unit volume), kg m3 s1
hg heat transfer coefficient Yi mass fraction of species i
Hreac reaction heat, J kg1
I radiation intensity, W m2 s1 Greek letters
[i] molar fraction of reactant species i a volume/internal surface area ratio in Gibb model
k turbulent kinetic energy, m2 s2 a1, a2 volatile yield
k1, k2 devolatilization rate constant, s1 e turbulent dissipation rate, m2 s3
k1 rate of external diffusion in Gibb model, s1 ep particle emissivity
i2 rate of surface reaction rate in Gibb model, s1 k thermal conductivity, W m1 k1
k3 rate of internal diffusion and surface reaction in Gibb rk, r turbulence model constant
model, s1 rB Stefan–Boltzmann constant, 5.67  108 W m2 k4
kc char oxidation rate in Gibb model, m s1 / mechanism factor in Gibb model
kc oxidation rate of coke reactions in Field model, q density, kg m3
kg m2 s1 c volume porosity (c = 1 for cavity)
kd diffusion rate of coke reactions in Field model, kg m2 l dynamic viscosity, Pa s
s1 lt turbulent viscosity, Pa s
_
m mass transfer rate from a particle, kg s1 Ci molecular diffusivity of species i, kg m1 s1
ma ash mass fraction
ma,0 original ash mass fraction Subscripts
mc mass of char, kg c char
Mc molecular weight of carbon coke coke
Mo2 molecular weight of oxygen molecule g gas
mrc rate of change of mass of the raw coal, kg s1 p particle
mrc,0 mass of the raw coal at the start of the devolatilization, kg

method (DEM) was used for coke movement so that the raceway
Active Coke Zone/
Cohesive
Dripping Zone
structure could be predicted directly. This approach is generally
Blast Zone Zone of PCI
difficult to apply to a practical system where the number of par-
Influence
ticles is huge, and thus the computation will be extremely expen-
PC lance
sive. Therefore, these models are not suitable for simulating the
Char particles
in-furnace phenomena of PCI operation based on the numerical
swept into the
coke bed results.
To overcome these deficiencies, in this study, a comprehensive
Tuyere 3D integrated model is developed for simulating the coal/coke
combustion in the LBTRC region under practical conditions. This
Deadman
Raceway model is considered superior to previous models for its three
dimensionality and inclusion of coke bed and its reactions. The
model is validated against the measurements under different con-
Fig. 1. Schematic of pulverized coal injection in the lower part (lance-blowpipe-
tuyere-raceway-coke bed) of a BF [6].
ditions. The comprehensive in-furnace phenomena of the LBTRC
region are then simulated and analysed, in terms of flow and
model for a laboratory-scale test rig, where the reactions of both combustion characteristics. The underlying mechanisms of the
coal and coke were considered. The so-called discrete element in-furnace phenomena are also examined.
730 Y.S. Shen et al. / Fuel 90 (2011) 728–738

2. Model formulation clude mass (m), momentum (u, v, w), turbulence kinetic energy
(k), turbulence dissipation rate (e), enthalpy (H) and a number of
In the present model, one single computational domain covers species (Yi), including O2, CO2, CO, H2, H2O and volatiles, as summa-
the lance, blowpipe, tuyere, raceway and coke bed, so that the ef- rized in Table 1. Coal particles are treated as a dispersed phase and
fects of operational conditions and coke bed properties on coal modelled using the Lagrangian method. Particle behaviours are
combustion could be directly evaluated in real time. The blow- tracked along the discrete particle trajectories without considering
pipe-tuyere-raceway region is treated as a cavity. The coke bed is interaction between coal particles. Newton’s second law of motion
treated as a porous media. The model includes the following phys- is used to calculate their movements. The drag force (fD) and turbu-
ical and chemical processes: (1) turbulent gas-particle flow; (2) lence dispersion are considered. The change of particle tempera-
coal combustion (devolatilization, volatile combustion, and char ture is determined by three heat transfer modes: convective heat
reactions); (3) coke combustion and gasification; (4) heat transfers transfer, latent heat transfer associated with mass transfer, and
in the considered gas-particle-coke bed. On the other hand, radiative heat transfer. Full coupling of mass, momentum and en-
assumptions are made in this model for simplicity: (a) coal and ergy of particles with the gaseous phase is carried out.
coke particles are spherical; (b) there is no break-up or coalescence
of particles; (c) liquid flow in the raceway-coke bed region is not 2.1.2. Coke bed
considered; (d) unburnt char particle accumulation at the raceway The coke bed is treated as an isotropic porous media for compu-
boundary is not considered; (e) raceway shape is assumed and no tational efficiency. This is because based on the experiment obser-
coke falling into the raceway; and (f) gas dissociation under high vation [14] and previous DEM modelling [16], only a limited
temperature and reactions of secondary species such as Si and S number of coke particles are moving, which would not affect the
are not considered. The model is described below. coal combustion much. The general form of the governing equa-
tions of gas flow in the porous media is
2.1. Governing equations for gas-particle flow in cavity and porous
media r  ðqcUUÞ  r  ðCeff crUÞ ¼ cSU ð1Þ

The momentum source through the coke bed is formulated using


The model formulation of gas-particle flow and coal combustion
Ergun equation,
has been detailed elsewhere [2,12]. They are outlined below for
completeness. The new features relating to coke bed are described 150lð1  cÞ2 1:75qð1  cÞ
in detail. rP ¼ Uþ jUjU ð2Þ
c2 d2p cdp
2.1.1. Gas-particle flow In real BFs, the temperature of coke bed is much lower than the
The gas phase is described by a set of 3D, steady-state Reynolds raceway due to the effects of various complicated phenomena,
averaged Navier–Stokes equations closed by the standard k–e tur- such as FeO-Coke reaction (highly endothermic), solid-liquid heat
bulence model equations, as used before [12,13]. Specifically, the transfer, Si and metalloids reactions, and convection of coke solid
variables in the governing equations solved for the gas phase in- flow. At present, these phenomena are difficult to be included in

Table 1
Governing equations for the gas and particle phases.

For the gas phase X


Mass r  ðqUÞ ¼ _
m
np

  X
Momentum 2
r  ðqUUÞ  r  ððl þ lt ÞðrU þ ðrUÞT ÞÞ ¼ r p þ qk þ fD
3 np

    X
Energy k l
r  qUH  þ t rH ¼ q
C p rH np

   
Gas species i lt
r  qUY i  Ci þ rY i ¼ W i
rY i
   
Turbulent kinetic energy lt
r  qUk  l þ rk ¼ ðP k  qeÞ
rk
   
Turbulent dissipation rate lt e
r  qU e  l þ re ¼ ðC 1 Pk  C 2 qeÞ
re k

For a particle in the particle phase


Mass dmp
_
¼ m
dt
Momentum dUp
mp ¼ f D
dt
1 2
f D ¼ pdp qC D jU  Up jðU  Up Þ
8
Energy dT p
mp C p ¼ q
dt
X dmp
q ¼ pdp kNuðT g  T p Þ þ Hreac þ Ap ep ðpI  rB T 4p Þ
dt

2
where lt ¼ C l q ke ; Pk ¼ ðl þ lt ÞrU  ðrU þ ðrUÞT Þ; C D ¼ maxð24ð1 þ 0:15Re0:687 Þ=Re; 0:44Þ; i = O2, CO2, CO, VM, H2, H2O.
Y.S. Shen et al. / Fuel 90 (2011) 728–738 731

the coal combustion model to avoid the complexity. For example, two-competing-reactions model [19]. A pair of first-order reactions
in a previous study [17], the coke bed temperature was simply as- (R1, R2) with different rate parameters (k1, k2) and volatile yields
sumed as 0.8 Tg. In this study, in order to account for these effects, (a1, a2), compete to pyrolyse the raw coal. The Gibb model [20]
a heat sink is used, allowing for the consideration of more coke bed is used for char oxidation and gasification, where the diffusion rate
properties. of reacting gas within the pores of a char particle (k3) is considered
as well as the external diffusion rate (k1) and surface reaction rate
sourceT;coke ¼ hg Acoke ðT g  T 0 Þ ð3Þ (k2). Their equations are respectively given in Table 2.
The important coke reactions are considered in the porous media
6ð1  eÞ kg Nug region where the consumption of coke is refilled continuously to
Acoke ¼ ;h ¼ ; T ¼ maxð0:75T g ; 1773½KÞ ð4Þ
udcoke g udcoke 0 give an unchanged simulation domain. The Field model [21] is used
for coke reactions in the coke bed, including coke solution loss and
The value of hg is estimated to be 128.2 W m2 K1 based on the
coke combustion. The overall rate at the surface of a coke particle is
data from a BF, where kg is equal to 0.117 W m1 K1, / is equal to
determined by a combination of chemical reaction and diffusion of
0.85 and dcoke is assumed to be 0.03 m. Nu number is calculated
reacting gas, as shown in Table 2. This is controlled by the smaller
based on the Wakao equation [18].
one of the rates kd and kc [i] is the molar fraction of the reacting
Nug ¼ 2 þ 1:1Re0:6
coke Pr
1=3
ð5Þ gas specie i (i = O2 for coke combustion, and CO2 for coke solution
loss reaction).
In connection with our previous studies [2,12,13], the so called
2.2. Chemical reactions of coal and coke eddy dissipation model (EDM) [22,23] is used to model the gas com-
bustion. Compared to several new gas combustion models such as
The various reactions considered in this model and their model- USM (unified second-order moment) model [24], laminar flamelet
ling methods are listed in Table 2. The coal reaction model assumes model [25], and PDF (probability density function) model [26], the
coal consists of volatiles, char and ash, and the coke reaction model EDM model may have limitations in some applications, especially
assumes a pure carbon. The model has been successfully used in for gas only combustion, due to neglecting chemistry rate [24].
various investigations [2,12,13]. However, it is certainly applicable to the combustion of pulverised
The coal reaction model has been described in our previous coal under blast furnace conditions where the temperature is
work [12]. The devolatilization process is modelled by the so-called extremely high (2800 K) and the fast chemistry assumption is

Table 2
Reactions of coal and coke considered and their rates expressions.

Reactions Models Reaction rates Rate constants


Coal reactions
Coal = VM + Char Two-competing– k1 A1 = 3.7  105 s1 E1 = 18000 K
α1VM1 + (1 − α1 )char1 (R1, Low temperature)
reactions model raw coal A2 = 1.46  1013 s1 E2 = 30189 K
k2 α 2VM2 + (1 − α 2 )char2 (R2, High temperature)
dVM
¼ ða1 k1 þ a2 k2 ÞC O
dt
k ¼ A expðE=T p Þ

a1 = VM (daf.); a2 = 1.25a21 + 0.92a1


 
VM + O2 = CO2 + H2O Eddy dissipation model
e ½i
CA = 4.0
ri ¼ C A min
j m ’i
Char + O2 = CO + CO2 Gibb model dmc 3/ M C q1 1 Ac = 14 m s1 K1 Tc = 21580 K
¼ ðk þ ðk2 þ k3 Þ1 Þ1 mC
dt 1  e M O 2 qc 1
 
2ð/  1Þ Ts
¼ As exp  ;
2/ Tp
Char + CO2 = 2CO Gibb model Ac = 20230 m s1 K1 Tc = 39743 K
D
k1 ¼ ;
r2p
 0:75
Dref Tp þ Tg
D¼ ;
q 2T ref
kc
k2 ¼ ð1  eÞ ;
Char + H2O = CO + H2 Gibb model rp Ac = 606.9 m s1 K1 Tc = 32406 K
2
k3 ¼ kc T p ðb coth b  1Þ=b a;

kc ¼ Ac Tp expðT c =T p Þ;
 0:5
kc
b¼R
DP ea

Coke reactions
Coke + O2 = CO Field model dmcoke 1 1 P Ac = 3.26  106 kg m2 s1 Tc = 10855 K
¼ ðkd þ kc Þ1 ½i4pr2coke
dt PA
 0:75
Dref T coke þ T g PA
kd ¼ ;
rcoke 2T ref P
Coke + CO2 = 2CO Field model   Ac = 4.71  109 kg m2 s1 Tc = 29018 K
Tc
kc ¼ Ac exp 
T coke
732 Y.S. Shen et al. / Fuel 90 (2011) 728–738

correct [27]. In fact, EDM has been widely accepted in this area Table 3
[1,2,11–13,16,17]. As a model sensitivity check, we tested the so Boundary conditions and operating conditions.

called two-step (EDM – finite chemistry) model, available in Operating conditions Proximate analysis (ad.)
ANSYS-CFX, and found that this more complicated model yields Working volume 2749 m3
Moisture, % 3.2
almost the same results in terms of gas composition and coal burn- Productivity 2.4 tHM/m3 day Volatile matter, % 32.5
out. Therefore, the EDM offers a good compromise between accu- Tuyere number 28 Ash, % 9.8
racy and computational effort and is considered to be suitable for Reference 461.0 kPa Fixed carbon, % 54.5
pressure
the present work. Boundary conditions Gross specific 30.08
The composition of gas species (O2, CO, CO2, H2, H2O and N2) is energy, MJ/kg
the consequence of coal and coke reactions at respective reaction O2 enrichment 6000 Nm3/h Ultimate analysis (daf.)
rates. The rate constants listed in Table 2 also come from the pub- in blast
Blast (22.9% O2) 300,000 Nm3/h 1200 °C C, % 83.5
lished literature and are experiment-based. The gas compositions
Cooling gas 5000 Nm3/h 327 °C H, % 5.3
in the LBTRC region are obtained by solving the governing equa- (100% O2)
tions of each gas species. Conveying gas 1317 Nm3/h 45 °C N, % 1.95
(100% N2)
Coal 35 t/h 45 °C S, % 0.6
3. Simulation conditions (127.3 kg/tHM)
O (by diff), % 8.6
The model geometry is based on a commercial BF in BlueScope
Steel. The main dimensions of the model geometry are shown in
Fig. 2. Note that the lance, blowpipe and tuyere are in actual
dimensions, the raceway is designed in the shape of a ‘balloon’, angle of 10 degrees with its tip on the centreline. Three gas streams
rather than a divergent ‘tube’, as used elsewhere [13], and the coke (conveying gas, cooling gas and hot blast) are introduced into the
bed is assumed as a packed bed. In the previous studies, the shape domain. The lance details are shown in Fig. 2(d), where an internal
of raceway was determined by means of experiments [14], contin- tube diameter for the conveying gas stream is 17 mm and an outer
uum approach [10,28], and DEM coupled with computational fluid diameter of the lance shell is 35 mm. The mesh structure for blow-
dynamics (CFD) [16,29,30]. Under the present conditions (Table 3), pipe-tuyere-raceway-coke bed region is refined at the raceway,
a constant raceway profile is assumed as shown in Fig. 2, deter- especially in front of the lance tip. The whole simulation domain
mined based on the CFD-DEM simulations using a model similar is divided into four zones according to porosity, that is, the
to Feng et al. [29] and previous experimental/practical observa- porosities for raceway (Zone 0), deadman (Zone 1), dripping zone
tions [31]. Note that different operational conditions may give dif- (Zone 2) and cohesive zone (Zone 3) are 1.0, 0.25, 0.5 and 0.4,
ferent raceway shapes. This aspect will be studied in the future. respectively, based on the actual measurements [8], as shown in
The co-axial lance is introduced into the blowpipe at an inclination Fig. 2(b). A gradient transition zone of porosity is assumed at the

Fig. 2. Geometry of the model: (a), the whole model; (b), porosity distribution; (c), blowpipe and raceway; and (d), lance tip. The detailed dimensions are, (1) for blowpipe,
radius: 90 mm, and length: 800 mm; (2) for tuyere, radius: 75/90 mm, and length: 135 mm; (3) for raceway, depth: 1600 mm, height: 1000 mm (925 + 75), and width:
710 mm; and (4) for coke bed, depth: 3700 mm, height: 4500 mm, and width: 1000 mm.
Y.S. Shen et al. / Fuel 90 (2011) 728–738 733

100
(a) 16
measurements [31]

Velocity, m/s
prediction
80
Cumulative volume, % .

60 0

O2 concentration, % Temperature, °C
40 1600

1200
20
800
20
0
1 10 100 1000
Particle size, μm
10
Fig. 3. Particle size distribution of the pulverized coal considered.

0
raceway boundary, where porosity varies from 1.0 (raceway cav- -0.4 -0.2 0.0 0.2 0.4
ity) to 0.4 (surrounding coke bed). Distance to axis (m)
The mathematic model outlined can be used to study the per-
formance of different types of coal. In this work, in order to test (b) 40
O2,calculated
the applicability of the model, we are focused on one typical coal CO2,calculated
used in the BlueScope Steel. The operating conditions of the BF CO,calculated

Gas mole fraction, % .


O2,measured
and the proximate/ultimate analyses of the coal considered are 30 CO2,measured
summarized in Table 3. The size distribution of the coal is obtained CO,measured
from laser diffraction measurement (Fig. 3), covering a size range
of 5–250 lm. In the current simulation, 49 particle size classes 20
are sampled in the range of 5–250 lm and a total of over 6000 rep-
resentative particles are tracked.
10

4. Results and discussion


0
4.1. Model validation -0.6 -0.3 0.0 0.3 0.6 0.9 1.2 1.5
Distance from tuyere, m
The model is validated against the measurements from a labo-
Fig. 4. Model validation in terms of: (a) gas velocity, temperature and O2
ratory scale experiment in terms of gas velocity, temperature and concentration against measurements in laboratory experiments [32]; and (b) gas
concentration) [32]; and more importantly in a real BF in terms compositions against measurements in a real BF [31].
of gas composition [31], respectively. Fig. 4(a) shows the gas
velocity, temperature and concentration. It is clear that the predic-
tions are comparable to the measurements. Fig. 4(b) shows that the
flow starts a large-scale recirculation above the main gas flow jet
predicted gas composition in the region of raceway and coke bed is
in the raceway. In addition, as shown in Fig. 5(b), the streamlines
in good agreement with the measurements in a real BF (Newcastle
of the central jet gradually disperse at the end of raceway and to-
No. 4 BF in Australia) [31]. The model validity of coal burnout has
ward the side of raceway, i.e., in the direction of Z. That is, inside
been confirmed elsewhere [2,12]. Note that the measurements of
the raceway, the flow pattern can be divided into two parts: a high-
in-furnace thermal and chemical properties are extremely difficult.
speed jet and a large-scale recirculation. On the other hand, in the
On the other hand, coal burnout and gas composition are the most
coke bed, as shown in Fig. 5(b), gas velocities in the coke bed de-
important concerns in BF practice. Their distributions reflect the
crease rapidly to <5 m/s within a very short distance once the gas
collected effects of gas-particle flow, interphase heat transfer and
flow exits the raceway cavity and enters the surrounding coke
various reactions of coal combustion.
bed. Subsequently, the behaviours of low-speed gas flow differ in
In the following, the flow and associated thermal-chemical
the coke bed: the velocity is extremely low in the deadman;
behaviours in the LBTRC region will be analysed in detail, aiming
whereas it is relatively high in the dripping zone due to its high
to establish a clear picture of in-furnace phenomena about PCI
porosity.
operation in BF ironmaking.
It is shown in Fig. 5(c) that, corresponding to the gas flow, the
coal particle trajectories inside the raceway have two different
4.2. Flow field flow patterns: (i) an inclined main coal plume located along the
lower part of the raceway, where fine particles are observed at
Fig. 5(a) shows the gas velocity vectors in the raceway cavity. It the upper part of the plume initially and then leave the main coal
is shown that inside the raceway, the blast stream, together with plume before reaching the end of the raceway; and (ii) a large-scale
the inclined low-speed gas flow of conveying stream and cooling recirculation of the fine particles of up to 70 lm around the race-
stream from the inclined lance, is accelerated through the tuyere. way centre. Subsequently, the coke bed also shows two flow pat-
As a result, a high-speed jet of up to 220 m/s forms along the terns of coal particles accordingly (Fig. 5(c)): (i) the main coal
tuyere axis. Subsequently, after reaching the raceway boundary, plume (relatively large particles of around 100 lm) penetrates into
the gas flow is reduced to 20–30 m/s; at the same time, the gas the deadman zone; (ii) the recirculating fine particles exit mainly
734 Y.S. Shen et al. / Fuel 90 (2011) 728–738

Fig. 5. Flow pattern of gas-particle flow: (a), vectors of gas phase in the raceway; (b), streamlines of gas flow; (c), particle trajectories coloured by particle mean size; and (d),
particle trajectories coloured by particle travelling time.

from the top of the raceway and then move upward into the drip-
ping zone. This is because, the large particles tend to maintain their
initial momentum and the fine particles are easier to be affected by
the turbulence and then dispersed more widely. Fig. 5(d) shows the
particle trajectories coloured by travelling time. The residence time
of coal particles along the main coal plume is around 10–50 ms be-
fore reaching the end of the raceway, while the recirculating coal
particles may be up to 0.9 s in the raceway. On the other hand,
compared with the raceway in the coke bed, the travelling time
of the particles penetrating the coke bed is quite long, around
1.0 s. The travelling time is even longer in the deadman compared
to the dripping zone.

4.3. Temperature and gas species

Fig. 6 shows the temperature contours of the raceway and sur-


Fig. 6. Temperature contours along the symmetry plane.
rounding coke bed. The main coal plume and recirculation region
show a great difference in temperature. Along the main coal plume,
a high temperature field of up to 2900 K forms at the downstream the other hand in the coke bed, the temperature is decreased to
of the coal plume and the nearby coke bed. In particular, an annu- around 1800 K when heating the surrounding coke bed.
lar high-temperature zone, the so-called flame front, is observed at Fig. 7 shows the distributions of gas species in the raceway cav-
the surface of coal plume in front of tuyere. In the recirculation re- ity. Inside the raceway, the main coal plume and recirculation re-
gion, the temperature is decreased to 2000 K. This temperature dif- gion show different distributions. Fig. 7(a) shows the contour of
ference results from the gas–solid flow in the raceway and its VM (fuel gas) in the raceway cavity. After exiting the tuyere, the
subsequent heat releases from chemical reactions, as show in the coal starts to release VM beyond the distance of 0.4 m and forms
following sessions. Compared with the previous models [12,13], a VM-rich core at 0.4–0.9 m. Beyond this region, the VM is nearly
the present model predicts a lower raceway temperature due to consumed completely, especially in the recirculation region. This
the recirculated cold gas flow from the surrounding coke bed. On is because, the VM core is surrounded by the flame-front region
Y.S. Shen et al. / Fuel 90 (2011) 728–738 735

Fig. 7. Contours of gas species inside the raceway: (a), VM; (b), CO2; (c), CO, and (d), O2.

of high temperature 2800 K, resulting in the rapid generation and in the deadman region (shown in Fig. 8(c) in logarithmic scale).
build-up of VM. Other gas species (CO2, CO, O2) change signifi- Along the Y direction, more qualitative comparisons are made
cantly along the coal plume and are then found relatively uniform along the tuyere axis as a function of distance from the lance tip
in the recirculation region (Fig. 7(b–d)): Along the coal plume after (Fig. 8(d)). The conversion rate of CO2 to CO along Y direction is
exiting the tuyere, CO2 is quickly increased to 0.22 (mass fraction) much faster compared to X direction. The variation of gas species
at the location of VM core, i.e., 0.4 m from the lance tip. CO con- along the tuyere axis can be divided into three stages (Fig. 8(d)):
centration is generally low in the whole raceway, where the main In Stage 1, three gas species keep almost constant, where coal par-
coal plume is even lower than the recirculation region, resulting ticles are heated up before 0.4 m. In Stage 2, O2 is converted to CO2
from the gas recirculation from the surrounding coke bed. Con- quickly, raising CO2 up to 0.2 in mass fraction which is mainly due
tours of O2 are shown in logarithmic scale for better clarification. to the strong VM combustion. In Stage 3, i.e., beyond the end of
A very fast decrease to zero is shown along the coal plume due raceway, CO2 is converted to CO by coke reactions in the fuel
to the strong VM combustion, whereas a relatively high concentra- (coke)-rich region. Beyond the end of raceway, there is little O2 left.
tion is shown near the surface of the plume. This explains the exis-
tence of the flame-front temperature. To sum up, after the
4.4. Coal combustion characteristics
introduction of blast, O2 is evolved into CO2 rapidly in front of
tuyere due to the strong VM combustion, whereas in the upper
The burnout is defined according to the ash balance,
part of the raceway (i.e. recirculation region), the conversion of
O2 to CO2 is slow, controlled by slow char reactions. Note that,  
ma;0
the concentration of CO in the raceway can be predicted using this Burnout ¼ 1 =ð1  ma;0 Þ ð6Þ
ma
model. This cannot be achieved using the previous models [12,13],
especially in the recirculation region. The previous models did not It represents the total weight loss of the coal due to devolatiliza-
consider coke bed and its reactions, and therefore cannot or greatly tion and char reactions. Fig. 9(a) shows the particle trajectories col-
under-predict the CO concentration in the recirculation region of oured by burnout in the raceway cavity. The recirculation region
raceway cavity, e.g. 0.01 (mass fraction) in Ref. [12]. This problem shows a higher burnout than the main coal plume, 85% vs.
is solved in this model after including the recirculation region and 60%, resulting from the different residence time (Fig. 5(d)) and
the surrounding coke bed. oxygen distribution (Fig. 7(c)). Specifically, in the following sec-
The gas species show more complex patterns in the coke bed tions, the detailed combustion characteristics are investigated
(Fig. 8). CO2 is completely converted to CO in the coke bed due along the tuyere axis and over the raceway surface, respectively.
to the strong coke reactions outside the raceway cavity. Their dis- Fig. 9(b) shows the evolutions of coal combustion characteris-
tributions vary in the two directions: X and Y. Along the X direc- tics along the tuyere axis. The burnout evolution can be divided
tion: CO2 is gradually converted to CO after the gas flow exits the into three parts: Part I, covering 0–0.4 m, where burnout does not
top of raceway. O2 content is quite low in the coke bed, even lower rise; Part II, covering 0.4–1.0 m, where burnout increases rapidly;
736 Y.S. Shen et al. / Fuel 90 (2011) 728–738

Fig. 8. Distributions of gas species (a), CO2; (b), CO, (c), O2 along the symmetry plane, and (d), along the tuyere axis.

and Part III, covering the distance beyond 1.0 m, where burnout burnout prediction along the tuyere axis. It is true that the recircu-
keeps increasing very slowly. Comparing evolution of burnout with lation region does exist in real BFs [1,31], therefore the present
the particle temperature and VM content, it is found that in Part I, model could truly simulate the in-furnace phenomena related to
coal particles are heated, with almost no burnout rise and no VM PCI operation. On the other hand, this comparison also indicates
content drop. The particle temperature increases slowly from 300 that it is important to include recirculation region in the future
to 800 K in this part after heating. In Part II, burnout increases very numerical studies of PCI operation for better understanding of in-
rapidly up to 55%, and VM content starts to drop from 0.4 m at a furnace phenomena.
very fast rate to nearly zero. The particle temperature in Part II is con- The three combustion characteristics are further analysed for
nected with Part I via a dent transition at about 0.4 m, resulting from particles of different sizes (Fig. 10). Six particle size groups are
the strong devolatilization (a strong endothermic reaction). Then the investigated. The burnout evolution curve for each size group can
particle temperature increases rapidly to 2400 K. In Part III, the also be divided into three parts, Part I (constant), Part II (increase
burnout plateaus at around 57%. The particle temperature keeps rapidly), Part III (plateau), similar to the overall performance of
increasing to 2600 K. The results indicate that Part I corresponds to these characteristics. However, the coal combustion performance
the preheating stage of raw coal particles, the burnout is mainly con- varies greatly for different size groups. Specifically, in Part I, the coal
trolled by devolatilization in Part II, and then followed by slow char particles less than 20 lm burn earlier at a distance of around 0.4 m,
oxidation in Part III. Moreover, compared to part III (slow char while the large coal particles of over 200 lm only start to burn
reactions control), Part II (fast devolatilization controls) plays a dom- beyond the distance of 0.7 m. In Part II, where devolatilization con-
inant role in determining final burnout along the tuyere axis. trols, the burnout of the fine particles increases faster than the large
In addition, it is noted that the current model predicts a lower ones. After the transitions, in Part III, where slow char reactions are
final burnout along the tuyere axis by 5% (absolute) compared prevalent, although the increasing rates for all the size groups are
with the previous models. In the previous studies [12,13] raceway slowed down, the fine particles still increase faster compared with
was assumed as a tube without considering recirculation region. In the large ones. These are found to be consistent with the evolutions
this study, when using the present model, the recirculation region of VM for each size group (Fig. 10(b)), where at 0.4 m, particles
is included and as a result, the fine particles of higher burnout below 30 lm have nearly completed devolatilization at the faster
leave the main coal plume at the downstream, leading to a lower rates, whereas those above 200 lm have hardly started beyond
Y.S. Shen et al. / Fuel 90 (2011) 728–738 737

(a) (a)
80

Burnout, % .
60

40
1-10
10-20
20-30
20 30-50
50-80
80-120
120-200
0
0 0.2 0.4 0.6 0.8 1 1.2
100 40 3000 Distance from lance tip, m
(b)

Particle temperature, K
Burnout
VM 2500
80 40
Tp 30 (b)
Burnout, %

1-10
60 2000 10-20
VM, %

20 20-30
40
1500
30 30-50
50-80
10 1000 80-120
20
120-200

VM,%
500
0 0 20
0.0 0.2 0.4 0.6 0.8 1.0 1.2
Distance from lance tip, m

10

(c)
0
0 0.2 0.4 0.6 0.8 1 1.2
Distance from lance tip, m

(c)
2800

2300
Particle temperature,K

1800

1300 1-10
10-20
Fig. 9. Combustion characteristics of coal: (a), along particle trajectories in the 20-30
raceway; (b), along the tuyere axis in the raceway; and (c), along particle 30-50
800 50-80
trajectories in the coke bed.
80-120
120-200
300
0.8 m. With respect to particle temperature (Fig. 10(c)), fine parti- 0 0.2 0.4 0.6 0.8 1 1.2
cles increase their temperature faster and always have a higher Distance from lance tip, m
particle temperature. To sum up, the fine particle group releases
Fig. 10. Combustion characteristics along the tuyere axis inside the raceway for
VM earlier and always gives a higher particle temperature and a
difference size group: (a), burnout; (b), VM; and (c), particle temperature.
higher burnout.
It is worth noting that the burnout at one final point of the tuyere
axis near the end of the raceway has been widely used in various burnout over the raceway bottom is about 82%; the burnout over
experimental and numerical studies to represent the coal combus- the entire raceway surface is about 92%.
tion efficiency. This method is not representative. In this study, the Moreover, Fig. 9(c) shows the coal burnout distribution along
average burnout over the raceway surface is predicable. The average the particle trajectories in the coke bed. It is shown in the simula-
burnout of the coal particles exiting the entire raceway surface is tion that the burnouts vary greatly in different zones of the coke
considered significant in providing more reliable information on bed, nearly 100% in the dripping zone, and around 75% in the dead-
the amount of unburnt char entering the coke bed. This information man zone. This results from the local differences in oxidants
is not predicted by the previous models which do not consider the (O2 + CO2 + H2O) and particle specific surface area. The dripping
raceway. However, due to the inclusion of more complex settings zone shows relatively higher oxidant concentration (Fig. 8(a,c))
such as recirculation region of raceway and surrounding coke bed, and the fine particle size distribution (Fig. 5(c)). As a result, the
this information can be predicted using the present model. Specifi- stronger char oxidation could occur there, yielding a higher coal
cally, under the conditions in the current study, the burnout at the burnout, compared to the deadman zone. Note that the interac-
raceway end (at 1.3 m from lance tip) along tuyere axis is 57%; the tions between the coal particles (treated as a dilute phase) and
738 Y.S. Shen et al. / Fuel 90 (2011) 728–738

the coke bed (treated as a packed bed) are not included in this (University of Newcastle, formerly BHPBilliton’s Newcastle Tech-
model. This simplification could shorten the residence time of coal nology Centre) is acknowledged for providing the experimental
particles in the coke bed and subsequently may underestimate the data used in the model validation.
coal burnout in the coke bed. How to address this issue needs fur-
ther work in the future. References

[1] Ishii K. Advanced pulverised coal injection technology and blast furnace
5. Conclusions operation. Oxford, UK: Elsevier; 2000.
[2] Shen YS, Guo BY, YuZulli P AB, Zulli P. A three-dimensional numerical study of
the combustion of coal blends in blast furnace. Fuel 2009;88:255–63.
A three dimensional integrated model of coal/coke combustion is [3] Vamvuka D, Schwanekamp G, Gudenau HW. Combustion of pulverized coal
developed and then applied to the lance-blowpipe-tuyere-raceway- with additives under conditions simulating blast furnace injection. Fuel
1996;75:1145–50.
coke bed region in a BF. One single computational domain is used, [4] Yagi J. Mathematical modeling of the flow of four fluids in a packed bed. ISIJ Int
where the blowpipe-tuyere-raceway is treated as a cavity and coke 1993;33:619–39.
bed as a porous media. The model is validated against measure- [5] Hutny WP, Lee GK, Price JT. Fundamentals of coal combustion during injection
into a blast-furnace. Prog Energy Combust Sci 1991;17:373–95.
ments from laboratory experiments and real blast furnace, respec- [6] Mathieson JG, Truelove JS, Rogers H. Toward an understanding of coal
tively. On this basis, the typical in-furnace phenomena of the combustion in blast furnace tuyere injection. Fuel 2005;84:1229–37.
raceway and surrounding coke bed are investigated. (1) Inside the [7] Jamaluddin AS, Wall TF, Truelove JS. Combustion of pulverised coal as a tuyere-
injectant to the blast furnace. In: 21st Symposium (International) on
raceway: main coal plume and recirculation region show different Combustion. The Combustion Institute. 1986. p. 575–84.
patterns. Compared to the recirculation region, the main coal plume [8] Nakajima R, Kishimoto S, Htta H, Ishii K, Sakuai M, Itagaki S. Behavior of coke
shows considerably higher gas velocity, shorter particle residence degradation in a blast furnace. NKK Tech Rev 1990;60:1–8.
[9] Ariyama T, Sato M, Yamakawa Y, Yamada Y, Suzuki M. Combustion behavior of
time, higher gas temperature and significantly higher CO2 and VM
pulverized coal in tuyere zone of blast-furnace and influence of injection lance
concentrations. The coal is heated and does not start burning until arrangement on combustibility. ISIJ Int 1994;34:476–83.
exiting the tuyere; subsequently the burnout increases rapidly at [10] Burgess JM. Fuel combustion in the blast-furnace raceway zone. Prog Energy
the upstream of the raceway due to strong devolatilization and then Combust Sci 1985;11:61–82.
[11] Shen YS, Guo BY, Yu AB, Zulli P. Model study of the effects of coal properties
plateaus at the downstream of the raceway due to slow char reac- and blast conditions on pulverized coal combustion. ISIJ Int 2009;49:819–26.
tions. (2) In the coke bed: compared with the dripping zone, the [12] Shen YS, Guo BY, Yu AB, Maldonado D, Austin P, Zulli P. Three-dimensional
deadman shows slightly lower gas velocity and longer residence modelling of coal combustion in blast furnace. ISIJ Int 2008;48:777–86.
[13] Shen YS, Maldonado D, Guo BY, Yu AB, Austin P, Zulli P. Computational fluid
time with large particle size, significantly higher CO, lower O2 con- dynamics study of pulverized coal combustion in blast furnace raceway. Ind
centrations, and a lower burnout. (3) Coal devolatilization is the Eng Chem Res 2009;48:10314–23.
main contributor to raise the final burnout level in the raceway. [14] Goto K, Murai R, Murao A, Sato M, Asanuma M, Ariyama T. Massive
combustion technology of solid fuel injected into blast furnace. In:
Local coal burnout mainly depends on particle size, residence time International Blast Furnace Lower Zone Symposium. Australasian Institute of
and oxygen availability in both the raceway and coke bed. Mining and Metallurgy (AusIMM). Wollongong, Australia. 2002. p. 1.
Compared with the previous 3D models [12,13], the present [15] Gu M, Zhang M, Selvarasu NKC, Zhao Y, Zhou CQ. Numerical analysis of
pulverized coal combustion inside tuyere and raceway. Steel Res Int
model includes more complex settings, such as recirculation region 2008;79:17–24.
of raceway, surrounding coke bed and its reactions etc. When com- [16] Nogami H, Yamaoka H, Takatani K. Raceway design for the innovative blast
paring the simulations using the present model with the previous furnace. ISIJ Int 2004;44:2150–8.
[17] Takeda K, Lockwood FC. Integrated mathematical model of pulverised coal
models, (a), in this study along the main coal plume, the dispersion
combustion in a blast furnace. ISIJ Int 1997;37:432–40.
of the coal plume is limited within the tuyere but more significant [18] Wakao N, Kaguei S, Funazkri T. Effect of fluid dispersion coefficients on
in the raceway. Then a large-scale recirculation of fine coal parti- particle-to-fluid heat-transfer coefficients in packed-beds - correlation of
cles forms in the upper part of raceway; (b), the present model pre- Nusselt numbers. Chem Eng Sci 1979;34:325–36.
[19] Ubhayakar SK, Stickler DB, von Rosenberg CW, Ganon RE. Rapid
dicts a lower temperature in the raceway; (c), the previous models devolatilization of pulverized coal in hot combustion gases. In: 16th
were found to greatly under-predict the CO concentration in the Symposium (International) on Combustion. The Combustion Institute.
raceway; and (d), the current model predicts a lower final burnout Pittsburgh. 1976. p. 427–36.
[20] Gibb J. Combustion of residual char remaining after devolatilization. In:
along the tuyere axis by 5% (absolute). The inclusion of recircula- Lecture at Course of Pulverised Coal Combustion. London: Imperial College;
tion region in the raceway plays an important role in determining 1985.
burnout. These comparisons also indicate that it is important to in- [21] Field M, Gill D, Morgan B, Hawksley P. The combustion of pulverised
coal. London: British Coal Utilisation Research Association; 1967.
clude recirculation region in the future numerical studies of PCI [22] Magnussen BF, Hjertager BW. On mathematical modelling of turbulent
operation for better understanding in-furnace phenomena. More combustion with special emphasis on soot formation and combustion. In:
importantly, compared with final burnout at one point predicted 16th Symposium (International) on Combustion. The Combustion Institute.
Pittsburgh. 1977. p. 719–29.
in previous models, this model could give a more reliable burnout [23] ANSYS-CFX-Documentation.
prediction over the raceway surface, which could better represent [24] Zhou LX, Qiao L, Chen XL, Zhang J. A USM turbulence-chemistry model for
the amount of unburnt char entering the coke bed. The average simulating NOx formation in turbulent combustion. Fuel 2002;81:1703–9.
[25] Peters N. Multiscale combustion and turbulence. Proc Combust Inst
burnout over the raceway surface is around 90% in the base case.
2009;32:1–25.
To our knowledge, this model is the first 3D model to simulate [26] Pope SB. PDF methods for turbulent reactive flows. Prog Energy Combust Sci
the in-furnace phenomena of PCI operation in the literature. 1985;11:119–92.
Considering the high cost for physical experiments or BF in situ [27] Williams A, Pourkashanian M, Jones JM, Skorupska N. Combustion and
gasification of coal. New York: Taylor & Francis; 2000.
measurements, this model provides a cost-effective tool for under- [28] Mondal SS, Som SK, Dash SK. Numerical predictions on the influences of the air
standing the in-furnace flow-thermo-chemical behaviours of the blast velocity, initial bed porosity and bed height on the shape and size of
PCI operation, and then optimizing operating conditions in full- raceway zone in a blast furnace. J Phys D-Appl Phys 2005;38:1301–7.
[29] Feng YQ, Pinson D, Yu AB, Chew SJ, Zulli P. Numerical study of gas-solid flow in
scale BFs in the future. the raceway of a blast furnace. Steel Res Int 2003;74:523–30.
[30] Xu BH, Yu AB, Chew SJ, Zulli P. Numerical simulation of the gas-solid flow in a
bed with lateral gas blasting. Powder Technol 2000;109:13–26.
Acknowledgements [31] Jamaluddin AS, Combustion of pulverised coal as a tuyere-injectant to blast
furnace. 1985, University of Newcastle: Newcastle.
[32] Truelove JS, Holcombe D. Measurement and modelling of coal flame stability
The authors wish to thank the Australian Research Council and in a pilot-scale combustor. Symposium (International) on Combustion
BlueScope Steel for their support of this project. Dr Harold Rogers 1991;23:963–71.

You might also like