World Academy of Science, Engineering and Technology 34 2007

Simulation Study of Radial Heat and Mass

Transfer inside a Fixed Bed Catalytic Reactor
K. Vakhshouri1, M.M. Y. Motamed Hashemi2

Abstract—A rigorous two-dimensional model is developed for primarily iron oxide, is contacted with a syngas stream
simulating the operation of a less-investigated type steam reformer through a moving bed reactor to remove oxygen from the raw
having a considerably lower operating Reynolds number, higher tube iron in the following manner:
diameter, and non-availability of extra steam in the feed compared Fe2O3 + 3(H2/CO) = 2Fe + 3(H2O/CO2) (8)
with conventional steam reformers. Simulation results show that Direct reduced iron (DRI) known as sponge iron can then be
reasonable predictions can only be achieved when certain correlations
for wall to fluid heat transfer equations are applied. Due to severe
treated to produce desired steel products.
operating conditions, in all cases, strong radial temperature gradients
inside the reformer tubes have been found. Furthermore, the results II. EXPLANATION OF OPERATIONAL SEVERE
show how a certain catalyst loading profile will affect the operation CONDITIONS
of the reformer.
A Midrex® reformer differs from the conventional steam-
Keywords—Steam Reforming, Direct Reduction, Heat Transfer, reformers in several ways:
Two-Dimensional Model, Simulation • A typical steam-reformer operates at pressures of 20 to
40 bars, but the Midrex® reformer operates at pressures of 2 to
I. INTRODUCTION 3 bars. Shorter tube length and larger tube diameter in a
Midrex® reformer impose less pressure drop compared with

S TEAM reforming is the reaction of a hydrocarbon,

especially methane, with the oxidants water vapor and/or
carbon dioxide to produce hydrogen and carbon
the conventional steam-reformers.
• At high mass velocities in conventional steam reformers
(GM = 10 to 23 kg/m2/sec), the radial diffusivity will be too
monoxide. The chemical reactions taking place in catalytic high to allow the development of significant radial
methane reforming are numerous, among which four reactions concentration. However, in a Midrex® reformer, mass
are more probable under the reforming conditions (Hyman, velocities are considerably lower (GM = 2 to 5 kg/m2/sec), and
1968): significant radial gradients may develop inside the tubes. This,
CH4 + H2O ↔ CO + 3H2 (1) in turn, increases residence time and hence, the risk of carbon
CH4 + CO2 ↔ 2CO + 2H2 (2) formation is amplified (Rostrup-Nielsen, 1984).
CO + H2O ↔ CO2 + H2 (3) • In a Midrex® process, design specifications necessitate
CH4 + 2H2O ↔ CO2 + 4H2 (4) that the reformer feed gas be a mixture of the off gas stream
A typical reformer generally uses nickel catalysts to from the moving bed reactor and fresh natural gas. Since
accelerate reforming reactions. The reforming reactions are hydrogen and carbon-oxide comprise more than half of the off
highly endothermic, meaning that large quantities of heat must gas stream, the reformer feed gas is far above the equilibrium
be added for the reactions to proceed. This is done by reformer condition. As a result, at typical feed temperatures of about 400
o
fire-box. C - 500 oC, set by process economics [11], the reforming
Under industrial reforming conditions, carbon deposition can reactions can not proceed, whereas the carbon formation
occur as in the Boudourd, Beggs, and methane cracking reactions have a high potential of stepping up. Moreover, this
reactions respectively: relatively low temperature of feed gas results in a very large
2CO ↔ CO2 + C (5) gradient between the furnace gas temperature and process gas
CO + H2 ↔ C + H2O (6) temperature at the reactor inlet zone, which consequently
2CH4 ↔ C + 2H2 (7) develops a large gradient inside the tubes.
While conventional steam reformers have been used around • The carbon dioxide content of feed gas in a Midrex®
the world for over 50 years in such varied industries as reformer is significantly higher than most conventional steam-
refineries, ammonia production, and methanol plants, the reformers. This specification, together with a low content of
production of syngas for reducing iron ores in natural gas steam, dictated by design specifications, entails a greater risk
based direct reduction (DR) plants, like Midrex® process, has of carbon deposits, which must be balanced by selecting
been less investigated. Direct reduction of iron is the first step appropriate operating reaction parameters and catalysts.
in converting raw iron ore to steel. The iron ore, which is
III. STEAM REFORMING MODELING
1
K. Vakhshouri is with Chemical and Biological Engineering Department,
University of British Columbia, Vancouver, Canada, V6T 1Z3, Phone: 604- Published reports about the application of two-
827-5197 e-mail: kvakhshouri@chml.ubc.ca dimensional models for simulating industrial steam-reformers
2
M.M.Y.M. Hashemi is with Mork Family Department of Chemical are few in number. Reference [25] applied a pseudo-
Engineering and Material Sciences, University of Southern California,
California, USA
homogeneous two-dimensional model for a steam reformer for

180
 World Academy of Science, Engineering and Technology 34 2007

the first time. It emphasized, however, on the usefulness of ∂xCH 4 Der ρ pg ⎡ 1 ∂xCH 4 ∂ 2 xCH 4 ⎤ ρ B riηi M w,0
such models for the prediction of carbon free operating = ⎢ + ⎥+ (16)
conditions. Reference [5] only compared the results obtained ∂z G ⎢⎣ r ∂r ∂r 2 ⎥⎦ GyCH4 ,0
from a heterogeneous one-dimensional model with a two- ∂xCO2 Der ρ pg ⎡ 1 ∂xCO2 ∂ xCO2 ⎤ ρ B riηi M w,0
2

dimensional one, while in a more recent study Pedernera et al. = ⎢ + ⎥+ (17)

(2003) used their heterogeneous two-dimensional model for ∂z G ⎣⎢ r ∂r ∂r 2 ⎦⎥ GyCH 4 ,0
proposing some theoretical improvements which may The energy balance equation can be written as:
influence the primary reformer performance in a good manner. ∂T pg ⎡ ⎛ 1 ∂T pg ∂ 2T pg ⎞ ⎤
⎟ + ρ B ⎛⎜ ∑ (− ΔH i ) riηi ⎞⎟⎥ (18)
2
[21] However, their model predicted a very large radial Gc p , pg = ⎢λer ⎜ + ⎜ ⎟
gradient, more than 100 oC, between the center of the tubes ∂z ⎢⎣ ⎜⎝ r ∂r ∂r 2 ⎟
⎠ ⎝ i =1 ⎠⎥⎦
and the wall, which is in contrast with the results obtained for In which λer and Der are effective radial conductivity and
conventional steam reformers by others (Rostrup-Nielsen, diffusivity respectively. They are generally calculated by heat
1984, De Deken et al., 1982, and NIPC, 2005). [25, 5, 7] and mass analogy:
In the present study, the steady-state operation of a large- Perm = Perh = Perf (19)
scale Midrex® reformer is investigated by means of a rigorous
mathematical model, accounting for both furnace-side and The methods for the calculation of fluid radial Peclet number,
reactor-side equations. Perf, will be discussed later.
Tallmadge (1970) proposed an extension of Urgan’s (1952)
IV. FURNACE-SIDE MODELING equation apposite for estimation of pressure drop under higher
Reynolds numbers: [28, 10]
Among different types of radiative models appropriate for dp ρ pg u 2pg (20)
− = f
simulation of the fire-box section of an industrial steam dz d cat
reformer, a Roesler flux-type model reevaluated by Filla 1−υ ⎡ n (1 − υ ) ⎤
(1984) is applied for the furnace-side modeling. The f = 3 ⎢m + (21)
υ ⎣ Re ⎥⎦
governing equations in this model are as follow: [12]
m = 1.75 (22)
∂2F
= α ( βF + γ ) (9) n = 4.2 Re 5 / 6 (23)
∂z 2
∂T
c p , fg G fg fg = 4 K a ( F − E fg ) + Q (10)
dz
In which:
α = −(2 K a + At + Ar ) / 2 (11)
β = − (4 K a + 2ε t At ) (12)
γ = (4 K a E fg + 2ε t At Et ) (13)
with boundary conditions:
dF
=0 , at z = 0 & z = L (14)
dz
T fg = 4 F / σ , at z = 0 (15)
In the above equations Q is the heat released along the flame
length due to combustion of fuel and air mixture. It can be
readily found based upon the heat release pattern proposed by
Roesler (1967). [24] Fig. 1 Schematic of a reformer tube

V. REACTOR-SIDE MODELING VI. REACTION KINETICS AND RATE

EXPRESSION
All material and energy balance equations are written for
a differential control volume of a single tube representing any Any pair of reversible reforming reactions (Eq. (1) – (4))
other tube (Fig. 1). It is assumed that axial dispersion effects will account for the stoichiometry of methane reforming.
and external heat and mass transfer resistance are negligible; While the choice of equations is important when dealing with
therefore, the mass balance equations for two major kinetic relationships, any pair of them may be selected when
components, i.e. methane and carbon-dioxide under steady- dealing with equilibrium relationships. In the present study,
state conditions are: reactions (1) and (3) are considered.
The rates of these reactions are evaluated by the following
first-order kinetic expressions:

181
 World Academy of Science, Engineering and Technology 34 2007

Ravi et Al., 1989, Murty and Murthy, 1988, and Rajesh et al.
⎛ − Ea1 ⎞ ⎛⎜ yH3 2 . yCO . p 2 ⎞
⎟
⎜
r1 = a1 p exp⎜ ⎟
⎟ ⎜ yCH4 − y . K (24) 2000) in simulating steam-reformers is the correlation of Beek
⎝ RT ⎠⎝ H2O eq1 ⎠
⎟ (1962).[16, 27, 20, 23, 22, 2] In these studies, however, a
correction factor of about 0.38-0.4 is applied to moderate the
⎛ − E a 3 ⎞ ⎛⎜ y .y ⎞ results predicted by this correlation. The reason may lie in the
r3 = a3 p exp ⎜⎜ ⎟⎟ yCO − H 2 CO2 ⎟ (25) fact that the correlation of Beek (1962) is developed to predict
⎝ RT ⎠ ⎜⎝ y H 2O . K eq 3 ⎟⎠ the heat transfer near the wall of the tubes in which the
The values of activation energy and the pre-exponential temperature of the process gas is considerably higher than the
factor for the methane reforming reaction (Eq.24) are reported mean cross-sectional temperature of fluid. [2]
by Akers and Camp (1955), while the corresponding values A better and more rational solution is to define an overall heat
for the water-gas shift reaction (Eq.25) are reported by Singh transfer based on the two-dimensional heat transfer parameters
and Saraf (1979).[1, 27] The values of equilibrium constant as recommended by Froment (1990) as follows: [13]
(Keq) for all reactions are reported by Elnashaie et al (1990). 1 ⎛⎜ d t ,in d t , out 1 d t ,in ⎞
⎟
For first-order reversible reactions, the effectiveness factor can = ln + + (36)
be calculated by the following expression (Froment and U in ⎜⎝ 2kt d t , in h fw CORR.λer ⎟
⎠
Bischoff, 1990): [9, 13] Beek (1962) stated that the value of CORR is equal to 8.
1
η= [(3Θ) coth(3Θ) − 1] (26) According to Golebiowski (1973), Crider gives the value of
3Θ 2 6.133.[14] Borkink and Westerterp (1992) fitted their
In which the general Thiele modulus, Θ, is defined as: experimental results by a CORR value of 7.4 for various types
of packings.[13] A comprehensive review on the value of
V kρ cat 1 + K eq (27)
Θ = cat . CORR has been carried out by Derkx and Dixon (1996).[6] It
S cat De K eq appears that this remains an issue which has yet to be
completely solved. In the present study, the value of CORR is
assumed to be 6.133. The most commonly used equations for
VII. BOUNDARY CONDITIONS the heat-transfer coefficient at the wall are compiled in Table
I.
At reactor inlet: The correlations found in the literature for the effective
xCO2 = xCH4 = 0 (28) radial thermal conductivity, λer, are semi-empirical expressions
Tpg = Tfeed (29) derived from various experimental results (Kvamsdal et al.,
Ppg = Pfeed (30) 1999). Generally, the value of effective radial thermal
At tube center: conductivity is chosen from an effective radial Peclet number.
∂xCH 4 ∂xCO2 The latter is defined as (Gunn, 1987):[17, 15]
= =0 (31) 1 1 (37)
∂r ∂r = + f ((Re . Pr) −1 )
Perf Perf ( ∞ )
∂T (32)
= 0 TABLE I
∂r
CORRELATIONS FOR HEAT-TRANSFER COEFFICIENT AT WALL (N
At the tube inside walls: = dti /dcat)
∂xCH 4 ∂xCO2 (33)
= =0 Wall to Fluid Heat Transfer Correlations Ref.
∂r ∂r Beek 1/ 3 0.8 0.4 [14]
∂T pg Nufw = 2.58(Re.Pr) + 0.094(Re) (Pr)
λer = U in (Tt , out − T pg ) (34) Yagi & 0 . 054 (Re . Pr) [14]
∂r Kuni Nu fw =
1 + 0 . 0135 (Re) 0 .5 (Pr) 0 . 66
Where the overall heat transfer coefficient is defined as based
Leva [14]
on the inner surface tubes: Nu = 0.813 exp (−6 / N ) (Re) 0.9
Dixon & [30]
1 ⎛⎜ d t ,in d t , out 1 ⎞⎟ (35) Nu fw = (1 − 1 / N ) (Re)0.610 (Pr)1 / 3
= ln + LaBua
⎜
U in ⎝ 2kt d t ,in h fw ⎟⎠ Li & [18]
Finlayson Nu wf = 0.17 (Re) 0.79 (Pr/ 0.72)1 / 3

VIII. HEAT TRANSFER PARAMETERS This equation shows that Peclet number for heat-transfer
becomes independent of the Reynolds number at high values
Special cares should be taken when certain heat transfer
(Re>1000), governing the operation of all conventional steam
correlations are applied. The heat transfer resistance through
reformers. Different values/correlations proposed for
the packing and fluid inside the bed is described by the prediction of limiting value for Peclet number (Perf (∞)) are
effective radial thermal conductivity λer, which lumps together
presented in Table II. Steam reforming standard catalysts are
all heat transfer mechanisms. The extra resistance near the
ring-shaped types; therefore, another important issue that
wall, causing the well known ‘temperature jump’, is described should be considered in applying heat transfer equations is the
by an apparent wall heat- transfer coefficient (Logtenberg and packing shape used in deriving those equations. In the present
Dixon, 1998).[19] The most important wall heat-transfer
study, the values reported by Borkink and Westerterp (1992)
correlation ever used (Hyman, 1968, Singh and Saraf, 1979,

182
 World Academy of Science, Engineering and Technology 34 2007

are applied, while, the equivalent diameters of catalysts are TABLE IV

MOBARAKEH PLANT OPERATING CONDITIONS
calculated as: [3]
Fuel flow rate, (Nm3/hr) 45198
d cat = 3 6 Vcat / π , for heat transfer (38) Combustion air flow rate, (Nm3/hr) 126390
Fuel inlet temperature, (K) 310
d cat = 6 Vcat / S cat , for pressure drop (39) Combustion air Inlet temperature, (K) 873
Fuel & air mixture inlet temperature, (K) 697
3
Feed gas flow rate, (Nm /hr) 107122
TABLE II. CORRELATIONS/VALUES RECOMMENDED LIMITING Feed inlet temperature, (K) 673
PECLET NUMBER FOR HEAT TRANSFER, (N = dti /dcat) Inlet pressure of feed gas, (kPa) 246
Correlations for Prediction of Fluid radial Peclet number for
Ref.
Heat Transfer mole % CH4 C2H6 C3H8 C4H10 H2 H2O CO CO2 N2
2 [3]
Perf ( ∞ ) = A(1 + 19.4 / N ) 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
Fahien and Smith Typical values for A:
A = 10 for spheres
A = 5 for nonspheres XI. RESULTS AND DISCUSSIONS
[3]
Perf ( ∞ ) = A(2 − (1 − 2 / N ) 2 )
Schlunder Typical values for A: A. Output Conditions
A = 7 for spheres The reactor and furnace output conditions predicted by
A = 4.6 for nonspheres different one and two-dimensional models are gathered in
Pe = 10.9 for spheres [3]
Borkink and
rf (∞ ) Tables V-VII.
Westerterp Perf (∞ ) = 7.6 for cylinders
TABLE V
Perf (∞ ) = 4.2 for rings FURNACE AND TUBE OUTPUT RESULTS PREDICTED BY
Perf (∞ ) = 12 for spheres [29] DIFFERENT TWO-DIMENSIONAL MODELS (SPHERE PELLETS)
Heat Transfer Flue Gas Process Gas Effluent
Dixon Perf (∞ ) = 7 for cylinders Correlation Temperature, Temperature, Methane
based on (K) (K) Mole
Perf (∞ ) = 6 for rings
Sphere Pellets Fraction, (%)
Beek 1323 1159 1.82
Yagi & Kuni 1365 1158 2.22
IX. SOLUTION PROCEDURE Leva 1358 1161 2.19
Dixon & LaBua 1323 1189 1.69
The solution starts with an initial guess for the tube-skin Li & Finlayson 1325 1180 1.84
outer surface temperature. By means of this value, the furnace- Plant Data 1393 1198 1.90
side and reactor-side equations are solved separately. The new
value for the tube-skin temperature can be calculated by An overview of the results reveals that in most cases one
making an energy balance on the tubes: dimensional model can predict the reactor output conditions
U out ( Tt , out − T pg ) = ε t ( F − Et ) (40)
very well. However, reasonable predictions by two-
dimensional models can only be achieved when certain
Convergence is assumed to have been achieved whenever the correlations for wall to fluid heat transfer and the radial Peclet
maximum difference between two sequential steps is less than number are applied. As predicted before, the shape of catalyst
the convergence criterion. All hydrocarbons heavier than and the type of heat transfer correlations have considerable
methane in the feed are assumed to instantaneously effects on final results. The best results for both furnace and
hydrocrack into CH4 and CO. The thermodynamic properties reactor outputs are predicted when the correlation of Yagi is
of combustion products and process gas are obtained by using used accompanied by a proper value of the limiting radial
S.R.K equation of state, while their transport properties are Peclet number for ring-shaped catalysts. On the other hand, if
obtained from DIPPR reported data (Daubert and Danner, the catalysts are assumed to be spherical, the best results are
1991). [4] obtained by using the correlation of Beek, followed by the
correlation of Li & Finlayson (1977) (Table 1).[18] In all
X. MODEL VALIDATION cases, one-dimensional models have a better estimation for
flue gas temperature.
In order to validate the model, the simulation results from
both one-dimensional and two-dimensional modeling are TABLE VI
compared against the available data of a Midrex® direct FURNACE AND TUBE OUTPUT RESULTS PREDICTED BY
DIFFERENT TWO-DIMENSIONAL MODELS (RING PELLETS)
reduction plant. The furnace and process conditions of this Heat Transfer Flue Gas Process Gas Effluent
unit are listed in Tables III and IV. Correlation based Temperature, Temperature, Methane
TABLE III PLANT FURNACE DATA on Ring Pellets (K) (K) Mole
Furnace dimension (L × W × H), (m) 41 × 15 × 8 Fraction, (%)
total number of tubes 432 Beek 1312 1218 1.41
reformer tubes inside diameter , (m) 0.200 Yagi & Kuni 1346 1185 1.88
reformer tubes outside diameter , (m) 0.224 Leva 1334 1193 1.78
heated length of reformer tubes , (m) 7.9 Dixon & LaBua 1302 1222 1.32
total number of burners 168 Li & Finlayson 1313 1216 1.42
emissivity of tubes 0.85
Plant Data 1393 1198 1.90
flame length ,(m) 4

183
 World Academy of Science, Engineering and Technology 34 2007

TABLE VII
F FURNACE AND TUBE OUTPUT RESULTS PREDICTED BY
DIFFERENT TWO-DIMENSIONAL MODELS (RING AND SPHERE
PELLETS)
Heat Transfer Flue Gas Process Gas Effluent
Correlation based Temperature, Temperature, Methane
on Ring and Sphere (K) (K) Mole
Pellets Fraction, (%)
Beek 1353 1198 1.81
Yagi & Kuni 1353 1198 1.80
Leva 1346 1203 1.74
Dixon & LaBua 1342 1205 1.67
Li & Finlayson 1354 1097 1.81
Plant Data 1393 1198 1.90

B. Process Gas Temperature Profiles

Axial and radial temperature profiles are shown by a three
dimensional graphs in Fig. 2 for the correlation of Yagi. Since
the best results are obtained by the correlation of Beek for
spherical catalyst and by the correlation of Yagi for ring- Fig. 3 Effect of heat transfer model on fluid temperature difference
shaped catalyst, the temperature gradient profiles between the between tube center and tube wall
wall and center of tubes are also drawn in Figure 3 for these
two models. The reason for this can be clarified by looking at Table 2,
in which it is clear that ring-shaped catalysts have higher
effective radial conductivity resulting in a lower radial
temperature difference inside the tubes. In both cases, it can be
seen that the radial temperature gradients will be considerably
decreased in the active catalyst zone in comparison with the
inert zone. In fact, after contacting with the active catalyst, the
process gas loses its temperature due to the occurrence of
highly endothermic reforming reactions. Since the furnace
heat load reaches the catalyst bed via tube walls, the
temperature of process gas is first increased near the walls and
then is conducted to the tube center in a radial direction.
Consequently, the process gas temperature is constantly higher
close to the wall than close to the center of the tubes;
therefore, the rate of reforming reactions is more rapid near
the wall. As a result, the radial temperature difference between
Fig. 2 Radial and axial temperature profile inside the the wall and the center of the reformer tubes will be lower in
the active catalyst zone than in the inert catalyst zone. Another
As expected, the radial temperature gradients in Midrex® noticeable feature which can be observed from these figures is
reformer tubes are considerably higher than those in the continued existence of the process gas temperature
conventional steam reformers due to the lower process gradient throughout the total length of tubes. The radial
Reynolds numbers, higher ratio of tube diameter to tube temperature differences between the tube centerline and the
length, and lower feed gas temperature. While the maximum wall reach 35 oC and 80 oC for spherical and ring-shaped
temperature gradient between the bed centerline and the wall catalysts, respectively.
is predicted to be 190 oC for sphere catalysts, the
corresponding value is 130 oC for ring-shaped catalysts. To C. The Effect of Catalyst Loading Profile
have an apposite measure, the developed model was applied to The radial and axial methane fractional conversion profile
simulate a conventional steam reformer used in a MeOH plant is shown in Fig. 4. As stated above, the reformer feed gas is
(NIPC, 2005).[7] The average superficial process Reynolds far above the equilibrium condition, resulting in an
number for this plant is about 6000, up to 2.5 times that of a undesirable situation in which the reforming reactions will not
Midrex® reformer. The results show that the maximum proceed unless the feed gas temperature rises before contact
temperature difference in this reformer is 49 oC. This is in with the nickel catalyst zone. In real operational conditions,
agreement with the value of 33 oC reported by De Deken et al catalyst loading involves the utilization of non-active materials
(1982) for another conventional steam reformer used in an at the bottom of the catalyst bed. This will provide an inert
Ammonia plant. It should be reminded that the average zone in which the temperature of process gas will be increased
superficial process Reynolds number for an Ammonia plant above the upper limit of carbon formation reactions, and the
steam reformer is typically higher than that of a MeOH plant lower limits of reforming reactions.
steam reformer. A better view on Fig. 3 shows that when the
catalysts are assumed to have spherical shapes, more
temperature gradients exist between the tube center and the
tube wall.[5]

184
 World Academy of Science, Engineering and Technology 34 2007

If this length is increased, the feed will have a greater

temperature when it comes into contact with the active zone,
and the rate of reforming reactions will be more rapid; as a
result, a higher temperature drop on the active zone will
happen again. This, in turn, will enhance the risk of reversed
reforming reactions as well as carbon formation. Therefore,
the safe design may not be achieved. On the other hand, if the
amount of inert catalyst is lowered, the temperature of the feed
may not reach the proper temperature required for reforming
reactions, and the risk of carbon deposition will be enhanced
again. A conclusive solution needs more comprehensive
investigations including the optimization of process conditions
and the use of dual or triple catalyst loading profiles inside the
reformer tubes (Sadri et al., 2007). [26]
Fig. 4 Methane radial and axial fractional conversion inside the D. Tube wall Temperature
reformer tube Tube wall temperature is an important parameter in
design and operation of steam reformers. The tube material is
The effect of this catalyst loading profile on the thermal exposed to an extreme thermal environment. Higher
and process performance of the reformer can be better temperatures at the tube wall, and thus within the tube, lead to
understood from Figures 3 and 5 in which a conversion drop is increased carbon lay-down on the catalyst and a consequential
apparent at the interface between different catalyst zones, loss of catalytic activity as well as potential catalyst breakage
especially in the center of the tubes, due to the considerable resulting in increased pressure drop and hence, low flow. Both
temperature gradients between the wall and the center of the lead to a decrease in the local reaction rate and associated
tubes. The process gas temperature drops between the catalyst endotherm. A common rule of thumb is that a tube wall
bed centerline and the wall at the interface is about 60 oC for temperature increase of 20 °C will foreshorten a tube life by
both models using the correlations of Yagi, and the over 50 % from its design period of 10 years to less than 5
correlations of Beek. However, no temperature drop is seen years [8]. An apparent mistake seen in published works is the
along the tube length in the axial direction. As is shown in Fig. assumption at which the outside tube skin temperature is
5, the conversion of methane is negative in the first segments considered to be identical for both one and two dimensional
of the active catalyst zone, meaning that the reversed reactions models [5, 21]. This is clearly a false assumption, because
happened. The calculation results show that the temperature there is a substantial temperature difference between the
rise in the inert zone is approximately five times greater than average fluid temperature in one dimensional models and the
in the active zone. As a result, when the hot feed gas reaches fluid temperature flowing near the inner wall of tubes in two
the high active catalyst, the endothermic reforming reactions dimensional models, particularly in the first half length of
will occur very rapidly and will cool the feed gas back down reformer tubes. As a result, the temperature gradient between
into an undesirable zone for forward reforming reactions, the furnace gas and process gas and consequently the tube
especially in the center of tubes where the risk of carbon wall temperature, will differ in these models. The tube
deposition is also high. It comes into question whether the temperature profiles obtained from one and two dimensional
length of the inert catalyst zone should be increased or not? models are shown in Fig. 6.

Fig. 5 Methane conversion in center and next to the wall of a

Fig. 6 The difference between the results obtained by one and two
reformer tube (two-dimensional model)
dimensional models on tube wall temperature

185
 World Academy of Science, Engineering and Technology 34 2007

From this figure, it is seen that the one-dimensional model when the effect of catalyst loading profile and risk of carbon
apparently predicts higher values for tube wall temperatures. formation are the matters of concern in designing steam
The flue gas temperature profiles inside the furnace are also reformers.
shown in this figure showing that this variable has higher
values in the one-dimensional model too. Therefore, it can be NOMENCLATURE
inferred that the one dimensional model performs the total
heat balance total of a steam reformer box (furnace and tubes) ai pre-exponential factor for reaction i, i=1, 3,
at a higher level than does the two-dimensional model. (kmol/kgcat . s)
At half tube surface area per unit free volume, (m2/m3)
XII. CONCLUSION Ar half of refractory surface area per unit free volume,
(m2/m3)
The giant quantity of heat transfer models proposed in cp specific heat at constant pressure, (J/kg. K)
open literatures for studying fixed bed catalytic reactors d diameter, (m)
during the last 40 years, comes most likely from a lack of De effective diffusivity of gas components, (m2.s)
understanding of how exactly heat transfer in a fixed bed Der effective radial diffusivity, (m2.s)
catalytic reactor should be described and modeled. Due to the E black body emissive power, (W. m2)
considerably lower operating Reynolds number, lower tube Ea energy of activation, (J. mol)
length to diameter ratio, together with the non-availability of f parameter defined by Eq.21
extra steam in the feed gas, causing a large temperature F half-sum of forward and backward axial fluxes, (kW.
gradients in radial dimension, the operation of steam reformers m2)
used for production of reducing gas may be a suitable case to G mass velocity, (kg/m2 s)
examine the accuracy of theoretical heat and mass transfer hfw fluid-wall heat-transfer coefficient, (W/m2 .K)
correlations proposed in open literatures for studying the ΔH heat of reaction, (J/mole)
behavior of packed-bed catalytic reactors under real k reaction rate constant, (kmol/ kgcat. S. Pa)
conditions. For that reason, a rigorous two-dimensional model kt thermal conductivity of tube material, (W/m2 .K)
is developed for simulating the operation of this less Ka furnace gas absorption coefficient, (1/m)
investigated-type steam reformer. Both the process side and Keq equilibrium constant
furnace side have been included in this integrated model. The L tube length, (m)
model is capable of not only predicting reactor output m parameter defined by Eq.22
conditions based on longitudinal changes, but also inspecting Mw molecular weight, kg kmol-1
radial heat and mass transfer inside the tubes. A number of n parameter defined by Eq.23
heat transfer models predicting values for effective radial p pressure, (Pa)
conductivity and diffusivity have been examined during the Perf fluid radial Peclet number, dimensionless
simulation of reactor side. Simulation results have been tested Perf (∞) limiting value for Peclet number, dimensionless
against available data from an actual plant. A comparison Perh effective radial Peclet number for heat, dimensionless
between the calculated and available data shows that the two Perm effective radial Peclet number for mass,
dimensional models can represent the reactor and furnace dimensionless
actual data very well but not exceptional. In all cases, strong Pr Prandtl number, dimensionless
radial temperature gradients inside the reformer tubes have Q fuel calorific value, (W/ m2)
been found. In some cases, the results show substantial r radial distance, (m)
discrepancy between these models, and it is revealed that ri rate of reaction i, i=1, 3, (kmol/kgcat. S)
reasonable predictions in all aspects, such as effluent R universal gas constant, 8.314 (J/mol. K)
composition and tube-wall maximum temperature, can only be Re Reynolds number based on superficial mass velocity,
achieved when certain correlations for wall to fluid heat dimensionless
transfer equations are applied. The two dimensional heat S surface, (m2)
transfer model in the catalyst bed may seem to be more T temperature, (K)
accurate than the one-dimensional model at first glance. U overall heat transfer coefficient, (W/m2. K)
However, in equations based on the former model, the linear V volume, (m3)
dimension of the catalyst likewise does not adequately denote x conversion of reactants, dimensionless
the effect of the form and size of the catalyst on heat transfer y mole fraction of gas components, dimensionless
[14]. The changes of the effective thermal conductivity of the z axial distance, (m)
bed along the radius due to non-uniform flow, effects of
chemical reaction over the catalyst, and the system of catalyst Greek Symbols
packing also not considered in deriving most of these α parameter defined by Eq.11
equations. Moreover, it should be noted that the two β parameter defined by Eq.12
dimensional model is not completely adequate for describing a γ parameter defined by Eq.13
packed bed in which the value of N (= dti/dcat) exceeds 10 [14]. ρ density, (kg/m3)
Because of high accuracy, one-dimensional models are yet Θ Thiele modulus defined by Eq.27
comparable with two-dimensional ones in many aspects; σ Stefan-Boltzmann constant, 5.667e-8 (W/m2. K4)
however, developing two-dimensional models are necessary

186
 World Academy of Science, Engineering and Technology 34 2007

υ catalyst bed void fraction, dimensionless [19] Logtenberg, S. A., Dixon, A. G., “Computational Fluid Dynamics
Studies of the Effects of Temperature-Dependent Physical Properties on
ε emissivity Fixed-Bed Heat Transfer”, Ind. Eng. Chem. Res., 37 (1998) 739-747.
λer effective radial thermal conductivity, (kW. m2. K) [20] Murty, V.S., Murthy, M.V.K., “Modeling and simulation of a top-fired
ηi effectiveness factor i, i=1, 3, dimensionless reformer”,Ind. Eng. Chem. Res 27 (1988) 1832-1840.
[21] Pedenera, M. N., Pina, J., Borio, D.O., Bucala, V., “Use of a
Heterogeneous Two-dimensional Model to Improve the Primary Steam
Subscripts Reformer Performance”, Chem. Eng. J. 94 (2003) 29-40.
B catalyst bed [22] Rajesh, J.K., Gupta, S.K., Ray, A.K., “Multiobjective optimization of
cat catalyst steam reformer performance using genetic algorithm”, Ind. Eng. Chem.
fg flue gas Res. 39 (2000) 706-717.
[23] Ravi, K., Joshi, Y. K., Guha, B. K., “Simulation of Primary and
in inner surface of tubes Secondary Reformers for Improved Energy Performance of an
0 initial condition Ammonia Plant”, Chem. Eng. Technol. 12 (1989) 358-364.
out outer surface of tubes [24] Roesler, F.C., “Theory of radiative heat transfer in co-current tube
pg process gas furnaces”, Chem. Eng. Sci. 22 (1967) 1325-1336.
[25] Rostrup-Nielsen, J. R., Catalysis. Sci. Technology, Vol IV, Springer,
r refractory Berlin (1984).
t tube [26] Sadri, M., Vakhshouri, K., Motamed Hashemi, M. M. Y, ”Coke
fw fluid to wall Formation Possibility during the Production of Reducing Gas in Large
ti tube internal diameter Scale Direct Reduction Plant”, Ironmaking and Steelmaking, 34 (2007) .
[27] Singh, C. P. P., Saraf, D. N., “Simulation of side-fired hydrocarbon
reformers”, Ind. Eng. Chem. Process. Des. Dev 18 (1979) 1-7.
REFERENCES [28] Tallmadge, J. A., “Packed Bed Pressure Drop-An Extension to Higher
[1] Akers, W. W., Camp, D.P., “Kinetics of the Methane-Steam Reaction”, Reynolds Numbers”, AIChE. J. 19 (1970) 1092-1093.
AIChE. J. 4 (1955) 471-474. [29] Dixon, A. G., "Wall and Particle –Shape Effects on Heat Transfer In
[2] Beek, J., “Advances in Chemical Engineering”, Vol. 3, Academic Press, Packed Beds Transfer in Fixed Beds at Very Low Tube-to-Particle
New York (1962). Diameter Ratio", Chem. Eng. Communication, 71 (1988) 217-237.
[3] Borkink, J. G. H., Westerterp, K. R., “Influence of Tube and Particle [30] Dixon, A. G., “Heat Transfer in Fixed Beds at Very Low Tube-to-
Diameter on Heat Transfer in Packed Beds”, AIChE. J. 38 (1992) 703- Particle Diameter Ratio”, Ind. Eng. Chem. Res. 36 (1997) 3053-3064.
715.
[4] Daubert, T.E., Danner, R.P., Design Institute for Physical Property Data,
American Institute of Chemical Engineers, Hemispher Publishing
(1991).
[5] De Deken, J. C., Devos, E. F., Froment, G. F., “Steam Reforming of
Natural Gas: Intrinsic Kinetics, Diffusional Influences, and Reactor
Design”, Chemical Reaction Engineering, ACS Symp. Ser., 196, Boston
(1982).
[6] Derkx, O. R., Dixon, A. G.,” Determination of the Fixed Bed Wall Heat
Transfer Coefficient using Computational Fluid Dynamics”, Numer.
Heat Transfer, Part A, 29 (1996) 777-785.
[7] Design and Construction of a Methanol Pilot Plant Based on Reforming
Technology, National Iranian Petrochemical Company (NIPC) Project
NO. 81118018, Iran (2005).
[8] Dixon, A.G., Nijemeisland, M., Stitt., E.H, “CFD Study of Heat Transfer
near and at the Wall of a Fixed Bed Reactor Tube: Effect of Wall
Conduction”, Ind. Eng. Chem. Res. 44 (2005).
[9] Elnashaie, S. S. E. H., Soliman, M.A., Al-Ubaid, A.S., Adris, A., “On
the Non-monotonic Behavior of Methane-Steam Reforming Kinetics”,
Chem. Eng. Sci., 45 (1990) 491-501.
[10] Ergun, S., “Fluid Flow through Packed Beds” Chem. Eng. Prog. 48
(1952) 89-94.
[11] Farhadi, F., Motamed Hashemi, M.M.Y., Bahrami Babaheidari, M.,
“Modeling and Simulation of Syngas Unit in Large Scale Direct
Reduction Plant”, Ironmaking and Steelmaking, 30 (2003) 35-41.
[12] Filla, M., “An Improved Roesler-type Flux Method for Radiative Heat
Transfer in One-dimensional Furnaces”, Chem. Eng. Sci. 39 (1984) 159-
161.
[13] Froment, G. F., Bischoff, K. B., Chemical Reactor Analysis and Design,
Wiley, New York (1990).
[14] Golebiowski, A., Wasala, T., “Thermal Processes in Catalytic
Reforming of Methane with Water Vapor”, International Chem. Eng.,
13 (1973) 133-139.
[15] Gunn, D. J., “Axial and Radial Dispersion in Fixed Beds”, Chem. Eng.
Sci. 42 (1987) 363-373.
[16] Hyman, M.H., “Simulate Methane Reformer Reactions”, Hydrocarbon.
Process. , 49 (1968) 131-137.
[17] Kvamsdal, H. M., Svendsen, H. F., Olsvik, O., “Dynamic simulation and
Optimization of a Catalytic Steam Reformer”, Chem. Eng. Sci. 54
(1999) 2697-2706.
[18] Li, C., Finlayson, A., “Heat Transfer in Packed Beds-A Reevaluation”,
Chem. Eng. Sci. 32 (1977) 1055-1066.

187