Applied Thermal Engineering 25 (2005) 2398–2411

www.elsevier.com/locate/apthermeng

Radiative models for the furnace side of a

bottom-fired reformer
F. Farhadi *, M. Bahrami Babaheidari, M.M.Y. Motamed Hashemi
Chemical and Petroleum Engineering Department, Sharif University of Technology, Tehran 11365-9465, Iran

Received 7 December 2003; accepted 16 December 2004

Available online 23 February 2005

Abstract

Two different groups of radiative models are used to simulate a Midrex reformer. The modeling includes
the furnace-side as well as the reactor-side equations. The simultaneous solution of governing equations
provides the flue gas and tube wall temperature profiles. These are compared with literature and plant data.
It was observed that the Flux model, applied in this work on the furnace of a bottom-fired reformer, shows
a good agreement with observed plant data. The well-stirred model is still satisfactory but the long-furnace
model is far away to merit an attention.

2005 Elsevier Ltd. All rights reserved.

Keywords: Radiation; Heat transfer; Modeling; Zone method; Flux method; Simulation

1. Introduction

In a Midrex direct reduction plant, natural gas is used to manufacture reducing gas via a
reforming process. The Midrex reformer is a bottom-fired box-type furnace with the structure
of a straight-flow, co-current type heat exchanger (Fig. 1). Up-fired burners are interposed be-
tween rows of vertical reactor tubes. The reactor tubes are filled with catalysts made of dispersed
nickel on an alumina or magnesium spinel carrier in the shape of Raschig or ribbed rings.

*
Corresponding author. Fax: +98 21 6022853.
E-mail address: farhadi@sharif.edu (F. Farhadi).

1359-4311/$ - see front matter
2005 Elsevier Ltd. All rights reserved.
doi:10.1016/j.applthermaleng.2004.12.017
 F. Farhadi et al. / Applied Thermal Engineering 25 (2005) 2398–2411 2399

Nomenclature

A surface area, m2
Ar half of refractory surface area per unit free volume, m2/m3
Aref surface of refractory in each zone, m2
At half tube surface area per unit free volume, m2/m3
At,o total external surface of all tubes in each zone, m2
C specific heat at constant pressure, J/kg K
E black body emission power, W/m2
f heat released fraction, 1/m
H energy flux, W/m2
I radiation intensity, W/m2
Ib blackbody radiation intensity, rT 4g =p, W/m2
K absorption coefficient, 1/m
L length, m
Lb mean beam length, m
m mass flow, kg/m2 s
q heat flux, W/m2
qlos heat loss, W/m2
qrad net radiative flux between flue gas and tube surface, W/m2
qrel heat released along the flame length, W/m2
q+ radiant flux density in increasing z-direction, W/m2
q
radiant flux density in decreasing z-direction, W/m2
T temperature, K
U overall heat transfer coefficient, W/m2 K
z axial distance, m

Greek symbols
r Stefan–Boltzmann constant (5.667 · 10
8 W/m2 K4)
e emissivity
X unit vector in spatial coordinate
w band radiation fraction
s window radiation fraction

Subscripts
B band radiation
W window radiation
f flame
fg flue gas
pg process gas
r refractory
t tube
t,o tube-out
2400 F. Farhadi et al. / Applied Thermal Engineering 25 (2005) 2398–2411

Syngas collector

Flue gas
to stack

Reactor tubes
Burners
Fuel/air

Fig. 1. Simplified diagram of a Midrex reformer.

Catalysts loading involve the provision of an inert zone filled by non-active materials in the bot-
tom of tubes. In this zone, the temperature of process gas increases above the upper limit of Bou-
douard reaction and lower limits of reforming reactions. The burners provide the required heat of
highly endothermic reforming reactions by combustion of fuel and air.
Besides of feed flow direction and firing location, the Midrex reformer differs from the conven-
tional steam reformers in several ways:

• A typical steam reformer operates at pressures of 20–40 atm, but the Midrex reformer operates
at pressures of 2–3 atm. Shorter tube length and larger tube diameter cause less pressure drop
compared with the conventional steam reformers.
• There is no steam generation unit in a Midrex plant. The required steam content of feed gas is
provided only by the steam content of recycled off gas stream from the shaft furnace. The recy-
cled off gas contains carbon dioxide and water vapor and is blended with fresh natural gas. This
feed stream enters the reactor tubes where it is heated and experiences chemical reactions.
• Steam reformers normally operate at a high stoichiometric ratio of oxidant (H2O) to carbon of
2–4. At this ratio, the reducing gas quality will not be suitable, so the gas must be quenched to
remove the excess steam and then heated up again to proper reduction temperature. The Mid-
rex reformer operates at an oxidant/carbon ratio (CO2 and H2O to hydrocarbons) lower than
conventional steam reformers, so the steam content of reformed gas is not very high.
• The carbon dioxide content of process feed gas in Midrex reformer is higher than most of con-
ventional steam reformers. This specification and low content of steam entail a greater risk of
carbon deposits, which must be balanced by selecting appropriate operating reaction parame-
ters and catalysts.
• Midrex reformer can tolerate sulfur content of the feed gas. Sulfur blocks the catalyst sites
against carbon nucleation and decreases the risk of carbon formation.

The production of syngas for reducing iron ores in direct reduction plants has been investigated
less than conventional steam reformers commonly used in chemical and petrochemical plants. In
 F. Farhadi et al. / Applied Thermal Engineering 25 (2005) 2398–2411 2401

this work an industrial Midrex reformer is modeled and simulated. Through the modeling, differ-
ent radiative models have been applied for furnace-side description. The various radiative heat
transfer models of fired heaters together with their advantages and disadvantages have been re-
viewed by Scholand [1], Rhine and Tucker [2] and Lihou [3].
Zone and flux methods are the two main groups of radiative methods, which have been used in
heat transfer modeling of fired heaters. In the zone method, the radiating enclosure is divided into
isothermal volume and surface zones. The well-stirred model (WSM), long-furnace model (LFM),
and multi-dimensional zone model are different forms of the zone method. Singh and Saraf [4],
Soliman et al. [5] and Rajesh et al. [6] have applied the well-stirred model (WSM) for simulating
furnaces of side-fired steam reformers. Plehiers and Froment [7] and Stehlik et al. [8] have applied
the multi-dimensional zone model on side and top-fired furnaces respectively. The long-furnace
model (LFM) has not been applied on steam reformer furnaces yet.
In flux methods the space around each point is divided into a number of equal solid angles, in
which the angular intensity of radiation is uniform. The energy balance for an infinitely small ele-
ment within the flue gas is (Truelove [9]):
ðX  rÞI ¼
KI þ KI b ð1Þ
where X is a unit vector in the direction of radiation, I is the total radiation intensity, Ib is the
blackbody radiation intensity at the gas temperature, and K is the absorption coefficient of the
flue gas. With the assumption of constant radiation intensity at each solid angle, the above equa-
tion turns to a finite number of ordinary differential equations (ODEs). Depending on the number
of solid angles, the number of ODEs will change. The Roesler [10] flux model (RFM) and its
extension, i.e. extended two-flux model (EFM) are the most commonly used forms of flux meth-
ods [11,12].
McGreavy and Newmann [13] used the Roesler flux model (RFM) for a top-fired steam-
reforming furnace. Murty and Murthy [14] applied the extended two-flux model (EFM) for a simi-
lar furnace. Soliman et al. [5] and Elnashaie et al. [15] used an improved version of the Roesler flux
model (RFM) on top-fired furnaces.
In this work, the thermal behavior of an industrial Midrex reformer is investigated by means of
four models, well-stirred model (WSM), long-furnace model (LFM), Roesler flux model (RFM),
and extended two-flux model (EFM). These models are coupled with a one-dimensional model for
reactor-side. Numerical solution of the resulting equations is compared with data from an indus-
trial plant.

2. Furnace-side modeling

In modeling the furnace-side heat transfer by different models, the following assumptions have
been made:

• The refractory walls are radiatively adiabatic under steady-state conditions. In other words, the
heat-loss to the atmosphere is assumed to be equal to the heat gained by convection and con-
duction from flue gas.
• All surfaces are grey and diffuse.
2402 F. Farhadi et al. / Applied Thermal Engineering 25 (2005) 2398–2411

• The effects of conductive and convective heat transfer on radiation field are negligible.
• In long-furnace model (LFM), Roesler flux model (RFM), and extended two-flux model
(EFM) a plug flow pattern is assumed for combustion products.

2.1. Zone methods

The well-stirred model (WSM) is the simplest model and implies a uniform profile for the flue
gas temperature and concentration through the furnace chamber. Generally, the outside wall tem-
perature is considered to be uniform, but the accurate determination of process gas composition
needs more realistic values of tubes wall temperature. Therefore, in the present study the tube wall
is considered to have different temperatures at different zones. Tube wall temperature is calculated
through an energy balance in each zone. The view factor of each tube element to any point of
refractory walls is assumed to be the same. Hence, there is no distinction between different ele-
ments of reformer tube surface due to their location.
The net radiative flux between flue gas and tube surface in each element, qrad, is calculated as
[6]:
r  ðAt;o þ Aref Þ  efg  et
qrad ¼ ðT 4
T 4t;o Þ ð2Þ
ðAt;o þ Aref Þ  efg þ At;o  ð1
efg Þ  et fg

where Tfg and Tt,o are the flue gas and outside tube wall temperatures respectively and At,o and
Aref are the total external surface area of all tubes and surface area of refractory in each zone,
respectively. et and eg are the emissivity of the tubes and furnace gas respectively and r is the Ste-
fan–Boltzmann constant.
The tube skin temperature in each zone can be calculated through an energy balance on the
tubes as:
qrad ¼ U  ðT t;o
T pg Þ ð3Þ

The flue gas temperature is initially unknown and the solution procedure starts with assigning an
initial guessed value for it. Convergence is achieved whenever the difference between the 99% of
the total heat input and flue gas enthalpy is less than the convergence criterion. Heat loss through
the walls is assumed to be 1% of the total heat input.
The long-furnace model (LFM) is a one-dimensional model using longitudinal series of con-
nected well-stirred zones, where the enthalpy of the combustion products leaving one zone is
the enthalpy input to the next adjacent zone [2]. Since in the well-stirred model (WSM) it is as-
sumed that the flue gas has a mean radiating temperature, this model cannot provide accurate
maximum tube-wall temperature and its location. The high alloy reformer tubes are very expen-
sive and any tube failure could result in plant shutdown for a long period. The tube material selec-
tion is very sensitive to variations in the maximum tube-wall temperature. A slight increase of this
temperature (10 K) will result in reduction of the tube life by 30% [16]. In addition, the well-stirred
model (WSM) cannot account for the heat release pattern through the flame length. The long-fur-
nace model (LFM) is a model that accounts for these factors. The net radiative flux in the flow
direction is assumed to be negligible in long-furnace model (LFM).
 F. Farhadi et al. / Applied Thermal Engineering 25 (2005) 2398–2411 2403

For this model, the energy balance on each zone of flue gas may be written as:
mfg C fg dT fg ¼ qrel
qlos
qrad ð4Þ
where qlos is the heat loss and qrel is the heat released along the flame length due to combustion in
each zone. The net radiative heat flux of each zone, qrad, is calculated similar to the well-stirred
model (WSM). The heat release pattern along the flame can be described by exponential [2], par-
abolic [10] or simple forms. In the simple form it is assumed that the fractional heat released in
each zone is constant along the flame length. In the present study, the RoeslerÕs heat release pat-
tern, which is a parabolic form, is applied. So the fractional heat released per unit furnace height,
f, can be calculated as:
 
6 z z2
f ¼
for 0 6 z < Lf ð5Þ
Lf Lf L2f

f ¼ 0 for Lf 6 z 6 L ð6Þ
Flame length, Lf, is the length at which the combustion is completed from the tip of the burner.
Eq. (4) can be numerically integrated to obtain temperature profile in the flow direction. The
value of Tfg at the bottom of the furnace is assumed to be equal to the mix temperature of fuel
and combustion air.

2.2. Flux methods

The Roesler flux model (RFM) is a simple two-flux method, which can be easily adopted for heat
transfer problems within an enclosure containing a real gas. In this model, the radiation field
around each point is divided into two hemispheres and the radiation heat transfer is considered
both along and perpendicular to the direction of the temperature gradient [11]. Roesler [10] used
a mixture of one-clear gas plus one-grey gas to predict the real gas behavior. The emissivity of
such a gas mixture, e, is expressed as:
e ¼ wð1
e
KLb Þ ð7Þ
where Lb is the mean beam length of the radiation field and w is the fraction of any incident radiation
which interacts with the gas. This fraction is known as band radiation. The remaining fraction of the
incident radiation s (s = 1
w) passes unaltered through the gas and is known as window radiation.
The radiative flux is divided into band and window radiation in positive and negative direction of z,
using (+) and (
) for these directions and subscripts B and W to designate band and window fluxes
respectively. The radiant energy balances on each of these fluxes can be written as:
 
dqþ wer Ar wer Ar
wer Ar þ wer Ar

B
¼ 2wKEfg þ wet At Et þ
2K
er Ar
et At qþ
B þ qB þ qW þ qW
dz 2 2 2 2
ð8Þ
 
dq
wer Ar wer Ar þ wer Ar þ wer Ar


B
¼ 2wKEfg þ wet At Et þ
2K
er Ar
et At q

B þ qB þ qW þ qW
dz 2 2 2 2
ð9Þ
2404 F. Farhadi et al. / Applied Thermal Engineering 25 (2005) 2398–2411
 
dqþ ser Ar ser Ar
ser Ar þ ser Ar

W
¼ set At Et þ
er Ar
et At qþ
W þ q þ q þ q ð10Þ
dz 2 2 W 2 B 2 B
 
dq
ser Ar ser Ar þ ser Ar þ ser Ar


W
¼ set At Et þ
er Ar
et At q

W þ q þ q þ q ð11Þ
dz 2 2 W 2 B 2 B
In the above equations, At and Ar are the half tube and refractory surface area per unit free vol-
ume of furnace, respectively, et and er are the emissivity of tube and refractory, respectively.
Efg ð¼ rT 4fg Þ and ðEt ¼ rT 4t;o Þ are the black body emissive power of flue gas and tubes. All of
the quantities designated by q are based on furnace free cross-section area. The tube surface tem-
perature, Tt,o, can be found from the local flux balance equation:
U ðT t;o
T pg Þ ¼ et ððqþ
þ

B þ qB þ qW þ qW Þ=2
E t Þ ð12Þ
The axial variation of the energy flux of flue gas can be determined by an energy balance as:
dH fg
¼
4KEfg þ 2Kðqþ

B þ qB Þ þ qrel
qlos ð13Þ
dz
where Hfg is the energy flux of flue gas. The radiation boundary conditions at the bottom of fur-
nace (z = 0) are obtained from the radiation balance at an adiabatic gray surface:
qþ



B ¼ wer ðqB þ qW Þ þ ð1
er ÞqB ð14Þ

qþ



W ¼ ser ðqB þ qW Þ þ ð1
er ÞqW ð15Þ
Similar conditions can be written at the top of the furnace.
The extended two-flux model (EFM) is a simplification of Roesler flux model (RFM) through
the following assumptions:

• Absorbing/emitting medium may be represented by a gray gas.

• Adiabatic refractory is assumed to emit/reflect whatever radiation it has absorbed.

With these assumptions, the flux equations are reduced as follows:

dqþ
¼ 2KEfg
ð2K þ et At Þqþ þ et At Et ð16Þ
dz
dq


¼ 2KEfg
ð2K þ et At Þq
þ et At Et ð17Þ
dz
dH fg
¼
4KEfg þ 2Kðqþ þ q
Þ þ qrel
qlos ð18Þ
dz
 þ 
q þ q

U ðT t;o
T pg Þ ¼ et
Et ð19Þ
2
with the boundary condition:
qþ ¼ q
at z ¼ 0; L ð20Þ
 F. Farhadi et al. / Applied Thermal Engineering 25 (2005) 2398–2411 2405

Simultaneous solution of the flux method equations is performed using the method suggested by
Selçuk et al. [12].

3. Model validation and results

In this section, the furnace-side heat transfer models will be coupled with a one-dimensional
reactor-side model [17] to simulate the reactor conditions and thermal behavior of an industrial
Midrex reformer, Mobarakeh direct reduction plant, located at Esfahan, Iran. All ODEÕs describ-
ing conservation laws for material, energy and radiation are written for a differential height of
tubes (see Fig. 2). The temperature of outer surface of tubes calculated by radiative models is ap-
plied to solve the reactor-side equations in each step. It is assumed that the heat transferred to the
outer surface of the each tube is conducted through it to the inner surface, in the radial direction.
The inner surface, in turn, transfers the heat to the process gas enclosed within the differential sec-
tion. The calculated exit conditions are evaluated using the plant data, where the furnace gas and
tube wall temperature profiles are compared with available experimental data for a similar Midrex
reformer [18]. The physical and operating conditions of Mobarakeh reformer are shown in Tables
1 and 2.

Catalyst bed

heat
from
furnace dz

length, z

Flue gas feed gas

Fig. 2. Schematic diagram of a reformer tube.

Table 1
Mobarakeh plant furnace data
Total number of tubes 432
Reformer tubes inside diameter, m 0.200
Reformer tubes outside diameter, m 0.224
Heated length of reformer tubes, m 7.9
Total number of burners 168
Emissivity of tubes 0.9
Flame length, m 4
2406 F. Farhadi et al. / Applied Thermal Engineering 25 (2005) 2398–2411

Table 2
Mobarakeh plant operating conditions
Fuel flow rate, N m3/h 45,198
Combustion air flow rate, N m3/h 126,390
Fuel inlet temperature, K 310
Combustion air inlet temperature, K 873
Fuel and air mixture inlet temperature, K 697
Feed gas flow rate, N m3/h 107,122
Feed gas inlet temperature, K 673
Inlet pressure of feed gas, kPa 246
Mole% CH4 C2H6 C3H8 C4H10 H2 H2O CO CO2 N2
Feed 14.99 1.4 0.53 0.19 35.02 13.64 18.95 14.24 1.03
Fuel 6.79 0.44 0.16 0.06 43.95 6.11 23.77 17.76 0.99

Table 3
Models results for reformer exit conditions and furnace duty
Zone methods Flux methods Plant data
WSM LFM RFM EFM
Furnace gas temperature, K 1393 1275 1338 1335 1393
Process gas temperature, K 1190 1198 1172 1173 1198
Process gas pressure, kPa 212 214 214 214 212
Process mole% CH4 1.88 1.25 1.89 1.93 1.90
Process mole% CO2 2.39 2.13 2.50 2.49 2.37
Process mole% CO 35.16 35.58 35.06 35.04 35.2
Process mole% H2 54.62 55.54 54.69 54.64 54.5
Process mole% H2O 5.18 4.75 5.10 5.13 5.21

The reformer exit conditions are reported in Table 3. Among these models, the well-stirred
model (WSM) predicts the exit conditions better than the others do. Although the Roesler flux
model (RFM) and the extended two-flux model (EFM) can well predict the syngas composition
and pressure, but their calculated values for syngas and flue gas temperatures are relatively poor.
Furthermore, the results show that reformer exit conditions obtained from long-furnace model
(LFM) are not very satisfactory. The underestimated values for the furnace exit conditions by dif-
ferential models may lie in the one-dimensional assumption for reactor side. In one-dimensional
model a mean radial temperature is assumed for process gas in each section. This will lead to a
lower furnace gas temperature at a constant heat flux. Pedernera et al. [19] developed a heteroge-
neous two-dimensional reactor model for primary steam reformers and compared the results with
one-dimensional reactor model. Their results are in agreement with present work and show that
the one-dimensional model underestimates the outlet gas temperature.
The furnace gas temperature profiles calculated by different models are compared with available
experimental data in Fig. 3. In well-stirred model (WSM), the temperature profile is a horizontal
line. In other models, the profiles are almost similar, with a steep rise in the first quarter of the
furnace where heat is released more rapidly than absorbed by the sink. After the combustion is
completed, the temperature falls continuously till the furnace exit.
 F. Farhadi et al. / Applied Thermal Engineering 25 (2005) 2398–2411 2407

1700

flue gas temperature , (K)

1400

1100 EFM

LFM

RFM
800
WSM

EXPERIMENTAL (Farnell [18])

500
0 1 2 3 4 5 6 7 8
tube length , (m)

Fig. 3. Furnace gas temperature profiles.

The temperature gradient in the long-furnace model (LFM) is more than flux models. This
model also predicts a higher maximum temperature of flue gas than the flux models. This is
mainly due to the long-furnace model (LFM) assumption, i.e. ignoring the radiative flux in the
flow direction. The second reason lies in the plug flow assumption for the flue gas. In an actual
furnace, an internal recirculation occurs at the neighborhood of the burners, which replaces high
temperature material from further downstream. This recirculation increases the average gas tem-
perature near the floor. The higher maximum furnace gas temperature results in higher heat
adsorption, and therefore less exit flue gas temperature.
Roesler [10] states that most of free volume is occupied by slow recirculation of combustion
products near the floor. The state of this gas cannot be far from equilibrium with the local radi-
ation field. Hence a more precise estimate of combustion products temperature at the bottom of
the furnace can be found:
 þ 0:25
qB þ q
B
T fg ¼ at z ¼ 0 ð21Þ
2rw

1700
flue gas temperature , (K)

1400

1100

ORIGINAL RFM
800
CORRECTED RFM

EXPERIMENTAL (Farnell [18])

500
0 1 2 3 4 5 6 7 8
tube length , (m)

Fig. 4. Corrected furnace gas temperature near the furnace floor.

2408 F. Farhadi et al. / Applied Thermal Engineering 25 (2005) 2398–2411

The value obtained by Eq. (21) is closer to the real and actual temperature of the furnace gas near
the floor. The resultant profile involving above correction on Roesler flux model (RFM) is shown
in Fig. 4.
The tube wall temperature profiles resulting from different models are presented in Fig. 5. The
long-furnace model (LFM) predicts a maximum tube temperature at the end of the flame pene-
tration, while the well-stirred model (WSM) predicts that the maximum tube temperature will oc-
cur at the top of the furnace. Two maxima are predicted by the flux methods, one at the top of the
furnace, similar to well-stirred model (WSM) and other at the same region of the long-furnace
model (LFM). The maximum tube temperature at the end of the flame is higher by 40 K in Roes-
ler flux model (RFM), but in extended two-flux model (EFM) the difference is only 9 K. Although
the well-stirred model (WSM) prediction for metal temperature distribution is poor, but its pre-
dicted value for maximum tube temperature is in good agreement with absolute maximum tube
temperature predicted in flux methods. The long-furnace model (LFM) prediction is poor and cal-
culated value of the maximum tube temperature by this model is more than maximum allowable
working temperature of tubes, i.e. 1100 C. A summary of models predictions is presented in
Table 4.
Heat flux distributions on tube wall are shown in Fig. 6. In the well-stirred model (WSM), the
furnace gas temperature is constant and the tubes wall temperature increases continuously, there-
fore the flux intensity decreases gradually up to the furnace exit. In the flux methods, the heat flux

1600
tube-wall temperature , (K)

1400

1200

EFM
1000 RFM

LFM
800
WSM

EXPERIMENTAL (Farnell [18])

600
0 1 2 3 4 5 6 7 8
tube length , (m)

Fig. 5. Outside tube wall temperature profiles.

Table 4
Model predictions for maximum tube wall temperaturea
WSM LFM RFMb RFMc EFMb EFMc
Max. tube wall temperature, K 1326 1490 1326 1286 1303 1294
Distance from furnace floor, m 7.9 3.56 2.98 7.9 3.28 7.9
a
Maximum allowable temperature is 1373 K.
b
First maximum.
c
Second maximum.
 F. Farhadi et al. / Applied Thermal Engineering 25 (2005) 2398–2411 2409

0/5
EFM
RFM
0/4 LFM

furnace duty , (mW)

WSM

0/3

0/2

0/1

0
0 1 2 3 4 5 6 7 8
tube length , (m)

Fig. 6. Heat flux distribution profiles.

1400
process gas temperature , (K)

1200

1000

EFM
800
RFM
LFM
WSM
600
0 1 2 3 4 5 6 7 8
tube length , (m)

Fig. 7. Process gas temperature profiles inside the tubes.

distribution is similar to the well-stirred model (WSM). In the long-furnace model (LFM), the
heat flux on tubes wall increases up to the end of the flame penetration region where combustion
is completed. Both heat flux and the furnace gas temperature profile pass through a maximum and
then decrease continuously.
The process gas temperature profiles resulting from of all models are shown in Fig. 7. Although
the well-stirred model (WSM) is simpler than the flux methods, it can predict the distribution of
process gas temperature as well as the flux models. The long-furnace model (LFM) is not in agree-
ment with other methods especially in the entrance of reformer tubes.

4. Conclusion

The long-furnace model (LFM) is not suitable for modeling of such furnaces. The reason lies in
its assumptions in which the inter-zone radiation transfer and recirculation of combustion
2410 F. Farhadi et al. / Applied Thermal Engineering 25 (2005) 2398–2411

products inside the furnace are not considered. Because of these two effects, the flue gas has a
nearly uniform temperature distribution inside the furnace in a real reformer.
The flux methods are satisfactory in nearly all of the predictions. They well estimate tube wall
temperature distribution, process gas temperature profile, and process gas stream composition,
whereas exit furnace gas temperature prediction is not accurate.
A comparison of calculated results confirms the fact that the well-stirred model (WSM) predic-
tions are very close to the actual ones. Although the well-stirred model (WSM) is simpler than the
other models, but its prediction for exit conditions is quite satisfactory.

Acknowledgements

The authors are grateful to the NISCO Mobarakeh direct reduction plant for providing the ac-
tual data and also for supporting the energy audit project in the plant. The authors would also like
to thank Mr. Peter Farnell for providing the experimental data of furnace gas and tube wall tem-
perature profiles.

References

[1] E. Scholand, Modern procedures for the calculation of radiant heat transfer in direct-fired tubed furnaces, Int.
Chem. Eng. 23 (1983) 600–610.
[2] J.M. Rhine, R.J. Tucker, Modeling of Gas-fired Furnaces and Boilers, first ed., British Gas Publication, Berkshire,
1991.
[3] D.A. Lihou, Review of furnace design methods, Trans. IChemE. 55 (1977) 225–242.
[4] C.P.P. Singh, D.N. Saraf, Simulation of side-fired hydrocarbon reformers, Ind. Eng. Chem. Process. Des. Dev. 18
(1979) 1–7.
[5] M.A. Soliman, S.S.E.H. Elnashaie, A.S. Al-Ubaid, A. Adris, Simulation of steam reformers for methane, Chem.
Eng. Sci. 43 (1988) 1801–1806.
[6] J.K. Rajesh, S.K. Gupta, A.K. Ray, Multiobjective optimization of steam reformer performance using genetic
algorithm, Ind. Eng. Chem. Res. 39 (2000) 706–717.
[7] P.M. Plehiers, G.F. Froment, Coupled simulation of heat transfer and reaction in a steam reforming furnace,
Chem. Eng. Technol. 12 (1989) 20.
[8] P. Stehlik, J. Sika, L. Bebar, L. Raus, Contribution to the research and development of radiation chambers in
steam reforming, Collect, Czech. Chem. Commun. 54 (1989) 2357.
[9] J.S. Truelove, in: E.W. Schlander (Ed.), Heat Exchanger Design Handbook, Hemisphere Publication Co, 1983.
[10] F.C. Roesler, Theory of radiative heat transfer in co-current tube furnaces, Chem. Eng. Sci. 22 (1967) 1325–1336.
[11] R.G. Siddal, Flux methods for the analysis of radiant heat transfer, J. Inst. Fuel 47 (1974) 101–109.
[12] N. Selçuk, R.G. Siddal, J.M. Beer, Prediction of the effect of flame length on temperature and radiative heat flux
distributions in a process fluid heater, J. Inst. Fuel 48 (1975) 89–96.
[13] C. McGreavy, M. Newmann, in: Inst. Electrical Engineers Conference on the Industrial Applications of Dynamic
Modeling, Durham, 1969.
[14] C.V.S. Murty, M.V.K. Murthy, Modeling and simulation of a top-fired reformer, Ind. Eng. Chem. Res. 27 (1988)
1832–1840.
[15] S.S.E.H. Elnashaie, M.A. Soliman, A.S. Al-Ubaid, A. Adris, Digital simulation of industrial steam reformers for
natural gas using heterogeneous models, Can. J. Chem. Eng. 70 (1992) 786–793.
[16] J.R. Rostrup-Nielsen, in: Catalysis Science Technology, vol. 4, Springer, Berlin, 1984.
 F. Farhadi et al. / Applied Thermal Engineering 25 (2005) 2398–2411 2411

[17] F. Farhadi, M.Y. Motamed Hashemi, M. Bahrami Babaheidari, Modeling and simulation of syngas unit in large
scale direct reduction plant, Ironmaking Steelmaking 30 (2003) 35–41.
[18] P.W. Farnell, Private Communication, Available from: <WWW.Synetix.com>, 2002.
[19] M.N. Pedernera, J. Pina, D.O. Borio, Verńica Bucala, Use of a heterogeneous two-dimensional model to improve
the primary steam reformer performance, Chem. Eng. J. 94 (2003) 29–40.