Applied Thermal Engineering 66 (2014) 298e308

Contents lists available at ScienceDirect

Applied Thermal Engineering

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

CFD modeling of multiphase reacting flow in blast furnace shaft with

layered burden
Dong Fu a, Yan Chen a, Yongfu Zhao b, John D’Alessio c, Kyle J. Ferron c, Chenn Q. Zhou a, *
a
Center for Innovation through Visualization and Simulation, Purdue University Calumet, 2200 169th Street, Hammond, IN 46323, USA
b
United States Steel Corporation, Research and Technology Center, 800 East Waterfront Drive, Munhall, PA 15120, USA
c
U.S. Steel Canada, Hamilton Works, 530 Gage Avenue North, Canada

h i g h l i g h t s g r a p h i c a l a b s t r a c t

A novel methodology is proposed to

efficiently model the blast furnace
shaft with layered burden.

The effects of layered burden on flow,
heat transfer, and chemical reactions
are considered in the model.

The shape and location of the cohe-
sive zone is determined by an itera-
tive method.

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

Article history: The ironmaking blast furnace is a counter-current chemical reactor which includes the ascending gas
Received 15 November 2013 flow and the counter-current descending porous bed (burden). A Computational Fluid Dynamics (CFD)
Accepted 28 January 2014 model has been developed to simulate the multiphase reacting flow in blast furnace shaft. The gas flow
Available online 13 February 2014
dynamics, burden movement, chemical reactions, heat and mass transfer between the gas phase and
burden phase are included in the CFD model. The blast furnace burden consists of alternative layers of
Keywords:
iron ore and coke. A novel methodology is proposed to efficiently model the effects of alternative burden
Blast furnace
layer structure on gas flow, heat transfer, mass transfer and chemical reactions. Different reactions and
Shaft
Layered structure
heat transfer characteristics are applied for difference types of layer. In addition, the layered CFD model
Burden accurately predicts the Cohesive Zone (CZ) shape where the melting of solid burden taking place. The
CFD shape and location of the CZ are determined by an iterative method based on the ore temperature
distribution. The theoretical formation and the methodology of the CFD model are presented and the
model is applied to simulate industry blast furnaces. The proposed method can be applied to investigate
the blast furnace shaft process and other moving bed system with periodic burden structure
configuration.
Ó 2014 Elsevier Ltd. All rights reserved.

1. Introduction

The ironmaking blast furnace is regarded as the largest energy

* Corresponding author. Tel.: þ1 219 989 2665.
E-mail addresses: dfu@purdue.edu, fudong1985@gmail.com (D. Fu), chen180@
purdue.edu (Y. Chen), yzhao@uss.com (Y. Zhao), JDAlessio@uss.com (J. D’Alessio),
KJFerron@uss.com (K.J. Ferron), czhou@purduecal.edu (C.Q. Zhou).
consumption metallurgical reactors to produce iron. The blast
furnace is a counter-current moving bed reactor where the iron ore
and coke are charged from the top to form the burden and the hot
gas is generated at lower combustion zone. The blast furnace

http://dx.doi.org/10.1016/j.applthermaleng.2014.01.065
1359-4311/Ó 2014 Elsevier Ltd. All rights reserved.
 D. Fu et al. / Applied Thermal Engineering 66 (2014) 298e308
efficiency is directly depended on the heat and mass transfer to the
ascending gas and descending burden. Due to the complicated
phenomenon and difficulties indirect measurement, numerical
models have been applied to simulate and understand the multi-
phase reacting flow inside the blast furnace [1e15].
In general, discrete and continuum models are the two major
approaches to model the blast furnace moving bed. The Discrete
Element Method (DEM) is applied to describe each individual ore
and coke particle of the layered burden [14,15]. With coupling with
the CFD for the gas phase, such simulation reveals the detail of the
gas and particle behavior. However, it is computationally expensive
due to the fact that blast furnace is a dense moving bed and the
transient simulation is needed for DEM model [15]. On the other
hand, the continuum approach assumes that the solid burden
behave like a continuum medium. The existing blast furnace con-
tinuum models [3e8,10,12,13] treat the layered burden as a ho-
mogenous mixture of coke and ore. Therefore, the effects of layer
thickness may not be reflected in the homogenous treatment for
the burden. Kuwabara et al. [11] developed a 1D model that takes
consideration of the layer structure and showed the effects of layer
structure on gas distribution. The model neglects the variation in
the furnace radius direction. In addition, the model assumed the
coke temperature and ore temperature is the same therefore used
one energy equation for the combination of coke and ore.
The coke and ore are alternately charged into the blast furnace
to form a layered burden. The layer thickness is ranged from 0.2 to
0.8 m. The layer structure of the burden maintains up to the CZ and
liquid iron and coke co-exists below the CZ. Usually there are about
30e60 layers of alternative coke layer and iron ore layer above the
cohesive in blast furnace. The layer structure of the burden is
important for furnace operation in the following aspects. First, the
permeability differs significantly in the coke layer and ore layer.
Second, different reactions are taken place for each type of layer,
i.e., the reduction (FeyOx þ CO / Fe þ CO2) takes place in iron ore
layer and the coke gasification (C þ CO2 / 2CO) takes place in coke
layer. The reaction rates are directly related to the temperature of
the individual layer. Pervious model assumed the ore and coke
share the same temperature. The heat transfer and reaction heat
are also differs from coke layer and ore layer. At last, the ore layer
melts and becomes impermeable in the CZ so the coke layer is the
only path that the gas can flow through. Dong et al. [2] concluded
that the layered CZ treatment may provide a better overall picture
of steady blast furnace operation in terms of the permeability
prediction.
2. Model description
2.1. Model assumptions and simplifications
The following assumptions are made to derive the mathematic
model.
1. The quasi-steady state approximation is assumed for the
ascending gas phase due to the relative slow motion of the
burden to the gas. It takes about 5 h for the burden to decent
from the top to the bottom of the furnace but 2e3 s for the gas to
flow from the bottom to the top [16].
2. The interface between the ore layer and coke layer is assumed
thin so that the mix layer is neglected.
3. The layer structures remain until reaches the CZ lower
boundary.
4. The layer volume maintains constant during burden
descending.
5. The particles are connected in point so that the heat conduction
between burden particles is neglected.
299
2.2. Governing equations
In the mathematical model, the ascending gas is described as
the gas phase and the descending burden is treated as the burden
phase (ore and coke). The burden phase is defined as the gross bed,
i.e. the combination of the solid and liquid [1,10]. As state in
assumption 1, the governing equations for the gas phase are quasi-
steady state. The general transport equation for gas phase is given
as Eq. (1).

  
!
V$ rg u g Bg ¼ V$ Gg VBg þ Sg (1)
For the solid phase, the diffusion term is zero because the
transport of the solid phase properties is solely by convection (the
movement of the burden). The general transport equation for
burden phase is given as Eq. (2). The detailed conservation equa-
tions for gas phase and burden phase are expressed in Table 1. Two
individual energy equations are used for the ore and coke due to the
heat transfer and reactions are different in coke and ore layer.
!
V$ðrs u s Bs Þ ¼ Ss (2)
The gas flow through granular burden is described by Ergun
type equation [17]. The potential flow model is adopted to describe
the burden movement [1,4,9,10]. The mass conservation equation
and momentum equation for the gas phase are combined to solve
the pressure in the Poisson type equation as Eq. (3). Then the ve-
locity is obtained by substituting the pressure into the momentum
equation. Similarly, the same treatment is applied for the burden
phase as Eq. (4). The coefficient Ks is the relative speed factor
ranging from zero to unity.


V$ rg Kg Vp ¼ mog
_ þ mcg
_ (3)
V$ðrs Ks Fs Þ ¼ mog
_  mcg
_ (4)
Because of the reaction heat and convective heat transfer differ
in the coke layer and ore layer, two energy conservation equations
are used, for coke and ore, respectively. The consideration of the
burden structure on gas flow, heat transfer and chemical reactions
are detailed in Section 2.4.
The boundary conditions are schematically shown in Fig. 1. At
the furnace top, the top gas pressure Ptop is set as a fixed value. The
burden velocity distribution vs(r) along the radius is specified. The
gradient of the gas enthalpy Hg and species mass fraction Yi are
assumed to be zero at the top outlet. The enthalpy of the ore Ho,top
and the enthalpy of the coke Hc,top are set according to the charging
material temperature. The concentrations of the burden Zo,j,top and
Zc,j,top along the radius are obtained by combining the burden dis-
tribution and total charge rate. At the furnace bottom, the gas ve-
locity, species mass fraction and gas enthalpy along the radius are
mapped from the raceway model results [18]. The gradient of the
coke enthalpy, ore enthalpy and burden composition is set as zero.
At the furnace side wall, the heat loss from the furnace shell is
specified via the convection boundary condition where Tg is the gas
temperature and TN is the cooling water temperature (300 K). The
effective heat transfer coefficient hg includes the convection be-
tween the hot gas and the refractory, the conduction within the
refractory, the conduction within the stave cooler, and the con-
vection between the stave material and the cooling water. The
actual value of hg is difficult to determine analytically due to the
complicated gas flow pattern and the erosion profile of the re-
fractory. Therefore, hg value is adjusted until the calculated total
heat flux matches with the measured value. The gradient of all
300 D. Fu et al. / Applied Thermal Engineering 66 (2014) 298e308

Table 1
Governing equations for gas and burden phase.

Gas Burden

Ore Coke
! !
Mass V$ðrg ug Þ ¼ mog _
_ þ mcg V$ðrs us Þ ¼ mog
_  mcg
_
! !
Momentum Kg Vp ¼ ug Ks VFs ¼ us


r ju j 1
Kg ¼ ma þ C2 g 2 g
150ðjj dj Þ2 ε3j 3:5ð1ε Þ
a ¼ C2 ¼ ðj d Þε3j
ð1εj Þ2 j j j

 ! !
Energy ! k
V$ðrg ug Hg Þ ¼ V$ g;Cpeff VHg þ SgH V$ðro us Ho Þ ¼ SoH V$ðrc us Hc Þ ¼ ScH
! ! !
Chemical species V$ðrg ug Yi Þ ¼ V$ðDi; eff VYi Þ þ SgY;i V$ð us Zo;j Þ ¼ SoZ;j V$ð us Zc;j Þ ¼ ScZ;j

variables on the central axis are set to zero for the axisymmetric
configuration.
2.3. Chemical reactions and interphase exchange
The major chemical reactions in an ironmaking blast furnace are
mainly gasesolid reactions as listed in Table 2. The following re-
actions, including nine gasesolid reactions and two phase trans-
formations are considered in the model. Two models are included
in the CFD code for the indirect reduction reactions (1e6). The first
mode is the unreacted shrinking core (URC) model as the expres-
sion of single-pellet reaction kinetics [19]. The reduction of iron ore
proceeds via the successive steps Fe2O3 / Fe3O4 / FeO / Fe. The
URC model assumes that the reduction of an iron oxide sphere
occurs at the surface with the formation of an outer shell of metallic
iron. The shell grows in the radial direction until the entire oxide is
completely reduced to iron.
Blast furnace coke undergoes three main reactions in the shaft.
These are the Boudouard reaction (7), the wateregas reaction (8)
and the direct reduction reaction (C þ FeO / Fe þ CO) which is,
kinetically, a combination of reaction (3) and reaction (7) or a
combination of reaction (6) and reaction (8). The Boudouard reac-
tion going to the right is also known as the endothermic gasifica-
tion of carbon or the solution-loss reaction. The unreacted core
model [20] is utilized to model the coke reaction (7) and (8). Due to
the high porosity of the coke particle, the diffusion resistance is
neglected. The flux decomposition reaction (8) is greatly dependent
on the decomposition pressure. The unreacted core model is also
applied to the flux decomposition reaction [21].
The heat transfer between the gas and burden is described as Eq.
(6) and the heat transfer coefficient proposed by Wakao et al. [22]
are used to account the porous bed effects on the heat transfer
coefficient of a single particle. Rej is the particle Reynolds number
based on particle diameter and superficial gas velocity.
 
SgH ¼ n_ j hj Aj Tj  Tg (5)
hj dj 1=3
¼ 2 þ 1:1Re0:6
j Prg valid for 3 < Rej < 104 (6)
kg

2.4. Burden layer structure consideration

The ore and coke is charged alternatively to form the layered

burden. The new layer will be charged as the burden slowly de-
scends due to the melting of reduced ore and consumption of the
coke in the lower part of the blast furnace. As the burden descents
and materials been recharged from the top. The layer structure
changes periodically during each charging cycle. At any fixed
location of the furnace, the descending of the burden follows the
periodical cycle from coke to ore then to coke again. This cycle
repeats in blast furnace because the top layer profile is the same
under a specific charging condition. Thus, the physical variable such
as temperature, composition and gas species at any specific loca-
tion will be also periodically changing with time. The average of the
transient periodical change for each physical variable in the blast
furnace shaft process can be obtained by averaging of the results

Table 2
Chemical reactions considered in the model.

No. Reaction Rate Ref.

Indirect reduction 1 3Fe2O3(s) þ CO(g) / 2Fe3O4(s) þ CO2(g) R1 [19]

of iron oxide 2 Fe3O4(s) þ CO(g) / 3FeO(s) þ CO2(g) R2 [19]
by CO 3 FeO(s) þ CO(g) / Fe(s) þ CO2(g) R3 [19]
Indirect reduction 4 3Fe2O3(s) þ H2(g) / 2Fe3O4(s) þ H2O(g) R4 [19]
of iron oxide 5 Fe3O4(s) þ H2(g) / 3FeO(s) þ H2O(g) R5 [19]
by H2 6 FeO(s) þ H2 (g) / Fe(s) þ H2O(g) R6 [19]
Coke reaction 7 CO2 þ C / 2CO R7 [20]
8 H2O(g) þ C / CO þ H2 R8 [20]
Flux 9 MeCO3(s) / MeO(s) þ CO2(g) R9 [21]
decomposition (Me ¼ Ca, Mg)
Fig. 1. Boundary conditions.
 D. Fu et al. / Applied Thermal Engineering 66 (2014) 298e308 301

from a number of steady state cases. The detail development of the Table 3
method is reported previously [23]. For each steady state case, the Source term Sg configuration for gas phase.

layer structure of coke and ore is prescribed beforehand of the CFD Bg Ore cell Coke cell
calculation. The layer structure is predicted by the burden distri- Gas momentum
!
ug ; p do,jo,eo dc,jc,ec
bution model published elsewhere [24]. The CFD cell is divided into Gas enthalpy Hg SgH ¼ n_ o ho Ao ðTo  Tg Þ SgH ¼ n_ c hc Ac ðTc  Tg Þ
four types depending upon the layer structure as shown in Fig. 2. If Gas Species Yi SgS; CO ¼ n_ o ðR1 þ R2 þ R3 Þ SgS;CO ¼ 2n_ c R7 þ n_ c R8
the space of the grid is occupied by the coke layer, the type of grid is SgS; CO2 ¼ n_ o ðR1 þ R2 þ R3 Þ SgS; CO2 ¼ n_ c R7

“coke cell”. Same rule is applied to “ore cell”. The “free space cell” þn_ f R9

type is assigned for the space above the top burden layer to
calculate the mixing of the top gas. The cells between two different
types of layer are “interface cell”. A certain percentage of ore and
coke and ore are co-exists in the “interface cell” as the layer
boundary passes through the cell. The interface between the free
space cell and the bed cell is not considered because only one layer
of such interface exists in a furnace. The error caused by the
simplification is considered to be negligible.
The conservation equations for the gas and burden phase are
solved in the cell system with the type of cell prescribed. The effects
of the layer structure are reflected in the source term settings of
conservation equations for different cell type. The detail source
term settings are given in Table 3. Different layer properties such as
particle size, shape factor and porosity are used in the Ergun type
momentum equation for ore cell and coke cell respectively. Simi-
larly, the source terms associated with gas enthalpy and species will
be also cell type dependent. The species source term associated
with iron ore reactions is only activated in the “ore cell” and those
with coke reaction only activated in the “coke cell”. The source term
of the burden enthalpy and composition for different types of grid
is set accordingly for different governing equations. An example for
such configuration is given in Table 4. The treatment for source
term is applied both chemically and thermally. The top boundary is
set as the inlet for burden phase with the time averaged value for
burden composition. The bottom boundary is set as the gas inlet
with the velocity and composition from raceway. Essentially, the
steady state conservation equations for both gas and burden are
solved. The heat and mass exchanges between the two phases are
determined by the type of cell that forms the layer structure. In this
manner, the effects of the layer structure are explicated considered
in this model.
The grid is set as a fixed layer configuration for the steady state
case. In order to consider the layer descending effects, the changing
cycle has been equally divided into a number of snaps. Case 1 and
Case 2 equally divide the cycle into two parts. Case 3 and Case 4
further divide the cycle into four parts. The corresponding burden
configurations of the four cases are shown in Fig. 3. More cases will
be added by the similar principle. The source term configuration
changes correspondingly to the burden structure of each case.
Average value of the variables distributions of the periodical
SgS; H2 ¼ n_ o ðR4 þ R5 þ R6 Þ SgS; H2 ¼ n_ c R8
SgS; H2 O ¼ n_ o ðR4 þ R5 þ R6 Þ SgS; H2 O ¼ n_ c R8
process can be approximated by averaging multiple steady state
cases with fixed burden configurations [23]. The equation for the
average value is given in Eq. (7) where n is the total number of
cases. Similar averaging method has also been used in blast furnace
shaft modeling [2]. The average of two cases (Case 1 and 2) will be
compared with the average of four cases (Case 1, 2, 3 and 4). If the
difference is larger than the defined criteria, divide one charging
cycle into more snaps (run Case 5, 6 then averaging Case 1e6 and so
on) until a case number independent solution is obtained.
1 Xn
B ¼  B n ¼ 2; 4; 6; 8; . (7)
n i¼1 i
B ¼ p; u; v; Tg ; To ; Tc ; Zj ; Yi ; etc:
2.5. Calculation procedures
The calculation flowchart is shown in Fig. 4. The shape and
position of the CZ is updated with each iteration, i.e., constant
temperature for upper boundary and local liquidus temperature for
the lower boundary, until the previous and current positions of
both boundaries are within the difference of less than a mesh size.
3. Results and discussion
3.1. Simulation condition
The baseline case used geometry from an industrial blast
furnace. Operating data is listed in Table 5. The properties for the
coke layer and ore layer are given in Table 6. The effective heat
transfer coefficient for wall heat loss hg is set to be 100 W/(m2 s). The
relative speed factor Ks is assumed to be linearly changing from 0.8
to 1.0 from the furnace center to the wall due to the fact the coke is
mostly consumed in the raceway which is closer to the furnace wall.

Fig. 2. Types of CFD grid.

302 D. Fu et al. / Applied Thermal Engineering 66 (2014) 298e308

Table 4
Source term Ss configuration for solid burden phase.

Bs Ore cell Coke cell

Ore enthalpy Ho SoH ¼ n_ o ho Ao ðTg  To Þ  n_ o ðR1 H1 þ R2 H2 þ R3 H3 þ R4 H4 þ R5 H5 þ R6 H6 þ R9 H9 Þ SoH ¼ 0

Coke enthalpy Hc ScH ¼ 0 ScH ¼ n_ c hc Ac ðTg  Tc Þ  n_ c R7 H7  n_ c R8 H8
Burden composition Zj SsZ; Fe2 O3 ¼ n_ o ðR1 þ R4 Þ SsZ; Fe2 O3 ¼ 0
SsZ; Fe3 O4 ¼ n_ o ðR2 þ R5  R1  R4 Þ SsZ; Fe3 O4 ¼ 0
SsZ; FeO ¼ n_ o ðR3 þ R6 Þ SsZ,FeO ¼ 0
SsZ,c ¼ 0 SsZ; c ¼ n_ c R7  n_ c R8

Fig. 3. Layered burden configurations.

3.2. Results of shaft CFD simulation

The simulation was conducted using the in-house developed

Fig. 4. Flowchart of the solution procedure.
code by FORTRAN language. The cases in the paper are 2D
axisymmetric since the furnace condition is assumed symmetric
about the central axis. The gas streamline in Fig. 5(a) illustrations
the zigzag flow pattern caused by the difference in porosity of the
coke and ore layer. The ore layered inside CZ possesses a much
lower porosity due to the fusion of the iron ore burden. The ore
layer inside CZ has a porosity of 0.1 and it is observed that the CZ
the ore layers become virtually impermeable thus forcing the
ascending gas to pass through the coke slits present between the
ore layers. The gas temperature with its isothermal lines in Fig. 5(b)
is similar to the coke temperature distribution in Fig. 5(c). The ore
temperature with its isothermal lines is shown in Fig. 5(d). Because
only coke exists in the center of the furnace, the ore temperature in
the center is not available. The isothermal lines of the coke and ore
temperature are presented ranging from 400 K to 1200 K. The low
temperature thermal reserve zone (TRZ) is observed from 800 K to
1000 K.
Fig. 6 shows the volume fractions of the major gas species with
the isothermal lines of the gas temperature. In Fig. 6(a), the CO
volume fraction is high around the CZ due to the direct reduction,
solution loss and the water gas reaction. As the gas ascends, the
indirect reduction occurs and CO is gradually consumed. On the
other hand, the distribution of CO2 is correspondingly followed the
distribution of CO as a product of the CO reactions as shown in
Fig. 6(b). In Fig. 6(c), the H2 is mostly consumed in the high tem-
perature region near the vicinity of CZ upper boundary due to the
fact that H2 is more active to for FeO reduction when the temper-
ature is higher than 1100 K. The H2O vapor also increases drastically
 D. Fu et al. / Applied Thermal Engineering 66 (2014) 298e308 303

Table 5 in the furnace top due to the water evaporation from the ore and
Operation data. coke.
Productivity (MTHM/day) 4993 Fig. 9 shows the stepwise reduction process of the iron oxides to
PCI rate (kg/MTHM) 67 hot metal. The isothermal line of the ore temperature is presented
Ore rate (kg/MTHM) 1501 ranging from 400 K to 1200 K (CZ upper boundary) with an interval
Flux rate (kg/MTHM)* 112
Top absolute pressure (Pa) 1.70  105
of 100 K. The CZ lower boundary liquidus temperature is found to
Averaging bosh gas temperature (K) 2300 be approximately the value of 1673 K. The reduction of hematite
(Fe2O3) took place at the upper part of the BF and it is completed
transformed to magnetite (Fe3O4) above 800 K. The wustite (FeO) is
starting to generate from 800 K and it is not reduced until near the
CZ. Below the CZ, almost all the wustite has been converted to
melted iron (Fig. 7).
Table 6
The detailed gas and burden information in the middle of the
Layer properties.
furnace (location is indicated in Fig. 9) is select to illustrate the
Coke layer dc ¼ 0.050 m features of layered burden treatment. The gas temperature and
ec ¼ 0.36
streamline are shown in Fig. 8(a) and (b) for Case 1 and Case 2,
jc ¼ 0.8
Ore layer do ¼ 0.012 m
respectively. The figures reflect the realistic gas temperature when
eo ¼ 0.45, eo ¼ 0.10 (inside CZ) the burden moves to such position because of the relative slow
jo ¼ 1.0 burden descending speed. For the gas phase, the heat transfer takes
place in both coke layer and ore layer. A continues distribution of
temperature for the gas is observe. In Fig. 8(c) and (d), the coke
temperature gradually increases throughout the coke layer. How-
ever, the isothermal lines are almost vertical inside the ore layer,

Fig. 5. Results of case 1: (a) gas streamline, (b) gas temperature, (c) coke temperature, (d) ore temperature.

Fig. 6. Gas species distributions of case 1: (a) CO, (b) CO2, (c) H2, (d) H2O.
304 D. Fu et al. / Applied Thermal Engineering 66 (2014) 298e308

Fig. 7. Burden composition: (a) Fe2O3, (b) Fe3O4, (c) FeO, (d) Fe.

indicating that the value maintains unchanged from the boundary
between ore layer and coke layer until the next coke layer.
Conversely, the ore temperature increases during the ore layer and
maintains during the coke layer as shown in Fig. 8(e) and Fig. 8(f).
Fig. 9(a) is colored by the gas temperature and a vertical line at the
3 m radius position is selected as the location for plotting data. As
shown in Fig. 9(b), the effects of the layered burden on temperature
are further evidenced in the gas, ore and coke temperature plotted
along the vertical line. The ore temperature reaches the gas tem-
perature very quickly due to the small size. Notable temperature
difference between the coke and gas is observed. The difference is
increased as the coke descending to the lower level of the furnace.

Fig. 8. Temperature distributions: (a) Case 1, gas, (b) Case 2, gas, (c) Case 1, coke, (d) Case 2, coke, (e) Case 1, ore, (f) Case 2, ore.
 D. Fu et al. / Applied Thermal Engineering 66 (2014) 298e308 305

Fig. 9. Case 1 result: (a) gas temperature distribution and the location of the vertical line (b) Plot of gas, ore, and coke temperature along the vertical line.

Fig. 10. Distribution along the vertical line: (a) ore temperature (b) average ore temperature (c) CO (d) average CO.

In reality, any specific location inside coke layer or ore layer distribution in blast furnace shaft. The ore temperature with
should has continuous temperature change as it is descending. opposite burden configurations (Case 1 and Case 2) and their
Averaging of multiple cases result with different burden structure average is plotted in Fig. 10(a). Case 1 and Case 2 with another two
in one charging cycle is adopted to obtain the average variable cases (Case 3 and 4) equally divided one charging cycle into four

Fig. 11. Distribution in the CZ region: (a) CO (b) CO2.

306 D. Fu et al. / Applied Thermal Engineering 66 (2014) 298e308
Fig. 12. CO distribution along vertical line at radius 3 m: (a) in whole shaft (b) near CZ region.
snaps. Fig. 10(b) shows that the average of Case 1 and Case 2 is
comparable with the average of Case 3 and Case 4 and also the
average of all four cases. The CO is consumed in ore layer but
slightly increased in coke layer due to mixing as shown in Fig. 10(c).
Similar to the average of temperatures, the same conclusion is
reached for averaging the CO distribution provided by Fig. 10(d).
The distributions of the CO and CO2 near the CZ are examined
closely in Fig. 11. The CO from the raceway combustion is consumed
in the melting ore layer and regenerated in coke layer by the solution
loss and water gas reaction due to the high temperature. The CO
distribution at 3 m radius along the entire blast furnace shaft is
shown in Fig. 12(a). The magnitude of the oscillation of CO volume
fraction inside the CZ is much larger than that above the CZ. The
oscillation inside the CZ is detailed in Fig. 12(b). The average CO
distribution above CZ is smooth comparing to the distribution inside
CZ as shown in Fig. 13(a). Fig. 13(b) shows the oscillation inside CZ is
smoothed by involving more cases in the average. Each case reflects
the quasi steady state distribution for gas phase including the spe-
cies, velocity, and pressure and temperature distribution under that
specific burden structure. The average results from multiple steady
state cases represent the averaged value of the periodical change for
both the gas phase and burden phase. Despite the difference in layer
structure of the four cases which causes the local distribution differs
from case to case, the overall CO distributions are similar for all the
four cases as shown in Fig. 14. The top gas composition, total coke
rate, total pressure drop is also similar as listed in Table 7. Good
agreements have been found by comparing the predicated value
with the plant measurement. It has been found that each steady
state case also predicts the global blast furnace performance due to
the number of layers in a blast furnace is large.
Fig. 13. Average CO distribution along vertical line at radius 3 m: (a) in whole shaft (b) near CZ region.
4. Conclusion
A CFD model has been developed to simulate the ascending gas
flow and the counter-current descending porous bed for iron-
making blast furnace shaft by considering the layer structure of the
burden thermally and chemically. The model proposed to consider
the layer structure of the burden is to solve the conservation
equations for the gas and burden phase in the cell system with the
type of cell prescribed. The effects of the layer structure are re-
flected in the source term settings of conservation equations for
different cell type. Essentially, the steady state conservation equa-
tions for both gas and burden are solved. The heat and mass ex-
changes between the two phases are determined by the type of cell
that forms the layer structure. In this manner, the effects of the
layer structure are explicated considered in this model.
It has been concluded that on one hand, each steady state case
reflects the quasi steady state distribution for gas phase under that
specify layer structure. In addition, each steady state case also
predicts the global blast furnace performance such as top gas
composition, pressure drop and coke rate due to the number of
layers in a blast furnace is large. On the other hand, the descending
cycle of the layer is considered by the averaging of multiple cases
with different layer structure in one charging cycle. Therefore, the
averaged local distribution over time can be obtained by averaging
multiple such cases. The model was applied to simulate an industry
blast furnace. Good agreements have been found by comparing the
predicated value with the plant measurement. The proposed
method can be applied to investigate the blast furnace shaft process
and other moving bed system with periodic burden structure
configuration.
 D. Fu et al. / Applied Thermal Engineering 66 (2014) 298e308 307

Fig. 14. Distribution in the CZ region: (a) Case 1 (b) Case 2 (a) Case 3 (b) Case 4.

R9 single particle flux decomposition, mol CO2/s

Sg source term of general transport equation for the gas
Table 7
phase, variety
Comparison with plant measurement.
SgC source term of continuity equation for the gas phase,
Measurement Case 1 Case 2 Case 3 Case 4 kg/(m3 bed s)
Top gas CO2% 22.18 22.46 22.91 22.51 22.86 Sgu, Sgv source term of momentum equation for the gas phase,
Top gas CO% 23.88 24.23 24.07 24.41 24.24 N/m3
Top gas H2% 6.33 6.50 6.36 6.44 6.30 SgY,i source term of species equation for the gas phase,
Top gas N2% 47.62 46.81 46.66 46.65 46.60
CO gas utilization 48.15 48.11 48.77 47.97 48.54
kg/(m3 bed s)
Pressure drop (Pa) 162,716 175,127 178,574 175,127 168,922 Ss Source term of general transport equation for the solid
Coke rate (kg/MTHM) 412 408 411 411 413 phase, variety
SgH source term of energy equation for the gas phase, W/m3
SoH source term of energy equation for the solid iron ore,
Acknowledgements W/m3
ScH source term of energy equation for the solid coke, W/m3
This research was supported by the American Iron and Steel SsZ,j source term of composition equation for the solid phase,
Institute (AISI) and the U.S. Department of Energy (DOE) under j ¼ o (ore), j ¼ c (coke), mol/(m3 bed s)
Award No. DE-FG36-07GO17041. The authors would like to thank Tj burden temperature, j ¼ o (ore), j ¼ c (coke), K
the support from the collaborators at ArcelorMittal-USA, Arce- TOB type of burden, e
lorMittal-Dofasco, SeverStal, US Steel, US Steel e Stelco Inc., and Yi gas species mass fraction,e
Union Gas. The authors would like to thank all the help from the Zj burden composition concentration, j ¼ o (ore), j ¼ c
students and staffs of CIVS. (coke), mol/m3 bed
dj particle diameter, j ¼ o (ore), j ¼ c (coke), m
hj Heat transfer coefficient, j ¼ o (ore), j ¼ c (coke), w/(m2 K)
Nomenclature hg Effective heat transfer coefficient for wall heat loss,
Aj surface area of a single particle, j ¼ o (ore), j ¼ c (coke), m2 w/(m2 K)
Cp gas heat capacity, J/(kg K) k gas thermal conductivity, W/(m K)
Di gas species diffusivity, m2/s n_ j number density of burden, j ¼ o (ore), j ¼ c (coke),
Hg gas enthalpy, J/kg 1/m3 bed
Ho ore enthalpy, J/kg tcyc the time for the burden to move in one periodic cycle of
Hc coke enthalpy, J/kg layers, s.
!
H1, H4 heat of reaction of Fe2O3 reduction by CO and H2, J/mol ug gas superficial velocity vector, m/s
H2, H5 heat of reaction of Fe3O4 reduction by CO and H2, J/mol
H3, H6 heat of reaction of FeO reduction by CO and H2, J/mol
H7, H8 heat of reaction of coke gasification by CO2 and H2O, J/mol Greek
H9 heat of reaction of flux decomposition, J/mol rg gas density, kg/m3 bed
Ks relative speed factor for burden descending, e ro ore bulk density, kg/m3 bed
R1, R4 single particle Fe2O3 reduction rate by CO and H2, mol rc coke bulk density, kg/m3 bed
CO/s or mol H2/s Bg general variable for gas phase, variety
R2, R5 single particle Fe3O4 reduction rate by CO and H2, mol Bs general variable for solid phase, variety
CO/s or mol H2/s εj burden porosity, j ¼ o (ore), j ¼ c (coke), e
R3, R6 single particle FeO reduction rate by CO and H2, mol CO/s Gg diffusion coefficient of general transport equation for the
or mol H2/s gas phase, variety
R7, R8 single particle coke gasification by CO2 and H2O, mol m gas viscosity, kg/(s m)
CO2/s or mol H2/s jj particle shape factor, j ¼ o (ore), j ¼ c (coke), e
308 D. Fu et al. / Applied Thermal Engineering 66 (2014) 298e308
Subscripts
g gas
s solid burden
References
[1] K. Yang, S. Choi, J. Chung, J.I. Yagi, Numerical modeling of reaction and flow
characteristics in a blast furnace with consideration of layered burden, ISIJ Int.
50 (7) (2010) 972e980.
[2] X.F. Dong, A.B. Yu, S.J. Chew, P. Zulli, Modeling of blast furnace with layered
CZ, Metall. Mater. Trans. B 41 (2) (2010) 330e349.
[3] M. Chu, H. Nogami, J.I. Yagi, Numerical analysis on charging carbon composite
agglomerates into blast furnace, ISIJ Int. 44 (3) (2004) 510e517.
[4] G. Danloy, J. Mignon, R. Munnix, G. Dauwels, L. Bonte, A blast furnace model to
optimize the burden distribution, in: 60th Ironmaking Conference Pro-
ceedings, Baltimore, vol. 60, 2001, pp. 37e48.
[5] J.A. de Castro, H. Nogami, J.I. Yagi, Transient mathematical model of blast
furnace based on multi-fluid concept, with application to high PCI operation,
ISIJ Int. 40 (7) (2000) 637e646.
[6] K. Takatani, T. Inada, Y. Ujisawa, Three-dimensional dynamic simulator for
blast furnace, ISIJ Int. 39 (1999) 15e22.
[7] P.R. Austin, H. Nogami, J.I. Yagi, A mathematical model of four phase motion
and heat transfer in the blast furnace, ISIJ Int. 37 (5) (1997) 458e467.
[8] X. Bi, K. Torssell, O. Wijk, Simulation of the blast furnace process by a math-
ematical model, ISIJ Int. 32 (4) (1992) 470e480.
[9] Y. Sawa, K. Takeda, S. Taguchi, Mathematical modeling of blast furnace
characterized by the precise layer structure in stock column, in: Ironmaking
Conference Proceedings, vol. 50, 1991, pp. 417e423.
[10] P.D. Burke, J.M. Burgess, A coupled gas and solid flow, heat transfer and
chemical reaction rate model for the ironmaking blast furnace, Ironmaking
Proc. Conf. 48 (1989) 773e781.
[11] M. Kuwabara, I. Muchi, Mathematical model for blast furnace operation with
inclined layers of burdens, Tetsu-to-Hagane 61 (6) (1975) 787e796.
[12] J. Yagi, Theoretical investigations of the blast furnace operations with the aid
of mathematical models, Tetsu-to-Hagane 54 (1968) 1019.
[13] I. Muchi, Mathematical model of blast furnace, Trans. Iron Steel Inst. Jpn. 7 (5)
(1967).
[14] Y. Yu, H. Saxén, Effect of DEM parameters on the simulated inter-particle
percolation of pellets into coke during Burden descent in the blast furnace,
ISIJ Int. 52 (5) (2012) 788e796.
[15] S. Yuu, T. Umekage, S. Matsuzaki, M. Kadowaki, K. Kunitomo, Large scale
simulation of coke and iron ore particle motions and air flow in actual blast
furnace, ISIJ Int. 50 (2010) 962e971.
[16] C. Van der Vliet, Modern Blast Furnace Ironmaking: An Introduction, IOS Press,
2009.
[17] S. Ergun, Pressure drop in blast furnace and in cupola, Ind. Eng. Chem. 45 (2)
(1953) 477e485.
[18] D. Fu, D. Zheng, C.Q. Zhou, J. D’Alessio, K.J. Ferron, Y. Zhao, Parametric studies
on PCI performances, in: Proceedings of the ASME/JSME 2011 8th Thermal
Engineering Joint Conference, March 13e17, 2011. Honolulu, Hawaii, USA.
[19] Y. Xie, M. Qin, A mathematical model for reduction of iron ore in moving bed,
Acta Metall. Sin. Ser. B 20 (1984) 249e260.
[20] I. Yuji, T. Kouji, Mathematical model analysis for oxidation of coke at high
temperature, ISIJ Int. 29 (l) (1989) 43e48.
[21] X. Huang, Principle of Steel and Iron Metallurgy, Metallurgy Industry Press,
Beijing, 2008, pp. 239e275.
[22] N. Wakao, S. Kaguei, T. Funazkri, Effect of fluid dispersion coefficients on
particle-to-fluid heat Transfer coefficients in packed beds: correlation of
Nusselt numbers, Chem. Eng. Sci. 34 (3) (1979) 325e336.
[23] D. Fu, Y. Chen, C.Q. Zhou, Numerical methods for simulating the reduction of
iron ore in blast furnace shaft, J. Therm. Sci. Eng. Appl. 6 (2) (2014), 021014.
[24] D. Fu, F. Huang, F. Tian, C.Q. Zhou, Burden descending and redistribution in a
blast furnace, in: 2010 AISTech e Iron and Steel Technology Conference
Proceedings, Pittsburgh, PA, vol. 1, 2010, pp. 693e700.