Fuel Processing Technology 200 (2020) 106314

Contents lists available at ScienceDirect

Fuel Processing Technology

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

Research article

Multi-scale two-dimensional packed bed reactor model for industrial steam T

methane reforming
⁎
Bhanu Vardhan Reddy Kuncharam1, Anthony G. Dixon
Department of Chemical Engineering, Worcester Polytechnic Institute, 100 Institute Rd, Worcester, MA 01609, United States of America

A R T I C LE I N FO A B S T R A C T

Keywords: A non-isothermal heterogeneous steady-state model was developed for a packed bed reactor for steam methane
Steam reforming reforming employing a multi-scale approach. The model consists of two-dimensional fluid-phase mass and heat
Hydrogen production transport equations accounting for axial and radial dispersion in the reactor tube, as well as accounting for mass
Packed bed and heat transfer resistances at the fluid-solid phase boundary, calculated using empirical equations. Reaction,
Reactor modeling
mass and heat transfer in the catalyst particle are directly coupled with the fluid-phase equations using a 1D
Multi-scale
pellet model, thus avoiding the use of a catalyst effectiveness factor for reaction. The performance of the packed-
bed reactor is compared using three pressure drop equations: the Ergun equation which neglects wall effects and
the Eisfeld-Schnitzlein and Di Felice-Gibilaro correlations which include them. This multi-scale model also ac-
counts for the effects of temperature, pressure and molar change of gas species due to reaction on superficial
velocity using a separate equation. The impact of neglecting these effects through simplified models is evaluated.

1. Introduction flux from burner-side combustion reactions is transferred to the tube

walls.
Synthesis gas (syngas) is used for various applications such as direct Models for the entire reformer have been developed [8–10] but
reduction of raw iron-ore in the steel industry, Fischer-Tropsch pro- these usually require a very simplified representation of the reactor
cesses in the petrochemical industry and also as a source for en- tubes, for example algebraic equations or a one-dimensional model
vironmentally clean fuels and chemicals [1,2]. Syngas is primarily with effectiveness factors. The temperature gradients can be significant
produced from steam reforming of methane, the main component of (larger than 80 °C) in the radial direction due to the high heat flux
natural gas. The steam methane reforming (SMR) reaction proceeds in through the tube wall and the endothermic nature of the main reactions
the following three reaction steps; (1, 2) show methane reforming and [11]. Therefore, two-dimensional modeling of packed-bed tubular re-
(3) shows the water-gas-shift reaction, actors is expected to be necessary for accurate prediction of methane
CH 4 + H2 O ↔ CO + 3H2 ΔH298 = 206.1 kJ /mol conversion, optimizing the yield and production of syngas. Three-di-
(1)
mensional particle resolved CFD has been applied to steam reforming
CH 4 + 2H2 O ↔ CO2 + 4H2 ΔH298 = 165.0 kJ / mol (2) [12], but this approach is at present too computationally expensive for
engineering design and control.
CO + H2 O ↔ CO2 + H2 ΔH298 = −41.15 kJ / mol (3)
Packed-bed reactor models for the SMR reaction of varying com-
The methane reforming reactions are endothermic, and equilibrium plexity have been reported in the literature. They can be classified into
limited. Various methods have been proposed to overcome these lim- three types: pseudo-homogeneous, pseudo-heterogeneous and hetero-
itations, including the use of membrane reactors [3,4], monoliths [5], geneous models. Further classification into either steady state or dy-
sorption-enhanced steam reforming [6] and electro-catalytic steam re- namic, and either one-dimensional or two-dimensional, is possible.
forming [7]. Despite these innovations, most steam reforming opera- Christiansen [13] has given a short review of modeling approaches,
tions are still carried out at high temperature and pressure in tubular while Pashchenko [14] has compared 1D. 2D and 3D types of homo-
reactors packed with catalyst active for reforming reactions. The geneous model.
packed-bed tubular reactors are suspended in furnace units, where heat The earliest work reported pseudo-homogeneous models which

⁎
Corresponding author.
E-mail address: [email protected] (A.G. Dixon).
1
Present address: Department of Chemical Engineering, Birla Institute of Technology and Science, Pilani, Rajasthan, India 333,031.

https://doi.org/10.1016/j.fuproc.2019.106314
Received 20 September 2019; Received in revised form 9 December 2019; Accepted 9 December 2019
0378-3820/ © 2019 Elsevier B.V. All rights reserved.
B.V.R. Kuncharam and A.G. Dixon Fuel Processing Technology 200 (2020) 106314

udp
Nomenclature Peam radial mass Peclet number,
Der
pf μ
C
ap surface area of particles per unit volume, 1/m Pr Prandtl number,
kf
C0 total feed concentration, mol/m3 r radial coordinate, m
Ci,bulk concentration of species i in the bulk fluid, mol/m3 rj rate of jth reaction, mol/kg/s
Cpi concentration of species i in the particle, mol/m3 R radius of the reactor, m
ρf udp
Cpis concentration of species i on the particle surface, mol/m3 Re Reynolds number based on particle diameter,
μ
pf
C heat capacity of fluid, J/kg/K Rg universal gas constant, J/mol/K
dp diameter of catalyst particle, m Rp pellet radius
rp radius of catalyst particle, m s pellet radial coordinate
Dea axial dispersion coefficient, m2/s Sc Schmidt number, μ
ρf DAB
Der radial dispersion coefficient, m2/s k d
Dpi diffusion coefficient of species i in the pellet, m2/s Sh Sherwood number, g p
DAB
ΔHrj heat of reaction of jth reaction, J/mol Tf temperature of fluid in reactor, K
kas effective axial solid thermal conductivity, W/m/K Tfeed feed temperature, K
krs effective radial solid thermal conductivity, W/m/K Tp temperature of catalyst pellet, K
kg particle-fluid mass transfer coefficient, m/s Tps surface temperature of catalyst pellet, K
kf thermal conductivity of fluid, W/m/K ufeed velocity of feed, m/s
kp thermal conductivity of catalyst particle, W/m/K yi mole fraction of species i
L length of the reactor, m yi0 mole fraction of species i in feed
MWave average molecular weight, kg/mol z* axial coordinate, m
MWi molecular weight of species i, kg/mol z dimensionless axial coordinate
MWN molecular weight of species N, kg/mol εb porosity of packed bed (void/total volume)
Nu
h d
Nusselt number, g p εp porosity of catalyst pellet
kf
P pressure, Pa μ viscosity of fluid mixture, kg/m/s
Pfeed feed pressure, Pa ξ dimensionless pellet coordinate
pf dp
ρf uC ρf density of fluid, kg/m3
Peah axial heat Peclet number,
k ea ρp density of catalyst particle, kg/m3
udp
Peam axial mass Peclet number,
Dea
pf dp
ρf uC
Peah radial heat Peclet number,
r

ignored the intraparticle diffusional limitations on the reaction as well Schwaab et al. avoided this problem by using CFD analysis to develop a
as inter-particle mass and heat transfer resistances. Singh and Saraf priori empirical equations for the effectiveness factors in their one-di-
employed a one-dimensional pseudo-homogeneous model to study a mensional unsteady-state simulations of industrial reformers [18].
side fired reformer and validated the model against industrial data [15]. Heterogeneous models explicitly recognize the role of solid catalyst
Kvamsdal et al. developed a dynamic two-dimensional pseudo-homo- particles in a packed-bed, and thus include separate mass and energy
geneous model for a tubular steam methane reformer in a side-fired transport equations for the fluid phase in the tube and for the catalyst
furnace [16]. Pseudo-homogeneous models are not suitable when there particles. These models can avoid the use of the catalyst effectiveness
are significant diffusional limitations on reaction in the catalyst parti- factor since the diffusion limitations in the catalyst are directly eval-
cles and the SMR reaction in industrial packed bed reactors is strongly uated and coupled with the fluid phase. Xu and Froment studied an
diffusion limited. industrial reformer employing a steady-state one-dimensional hetero-
The pseudo-heterogeneous models take into account diffusional geneous model accounting for internal diffusion in the catalyst particle,
limitations on reaction within the catalyst at low computational cost by but ignoring axial and radial dispersion [19,20]. Soliman et al. and
employing a constant effectiveness factor. The value of the effectiveness Elnashaie et al. used a dynamic one-dimensional heterogeneous model
factor is between zero and unity; it is unity when there are no diffu- to study steam reforming in top-fired and side-fired furnaces, the si-
sional limitations on the reaction in the catalyst i.e. ri (pc,i) = ri(psurf,i). mulation results were then compared with industrial reformer perfor-
The catalyst effectiveness factor of a typical reaction i is defined as, mance [21,22]. Quinta Ferreira et al. used a two-dimensional hetero-
r
geneous model for SMR with large-pore catalysts to investigate the
dV
apparent rate ∫ ri (pc, i ) ρcat V effect of strong internal mass-transfer resistances [23], but considered
ηi = = 0
intrinsic rate ri (psurf , i ) ρcat (4) only a single 1st-order reaction. Avci et al. investigated a dual catalyst
autothermal fixed-bed reactor for methane reforming and oxidation for
where ri (pc,i) is the rate of reaction i computed using the partial hydrogen production using a one-dimensional heterogeneous model
pressures (pci) inside the catalyst, ri (psurf,i) is the rate of reaction [24]. Nandasana et al. presented a dynamic one-dimensional hetero-
computed using the partial pressures (psurf,i) on the catalyst surface, ρcat geneous model which incorporated heat transfer and diffusion in the
is the density of catalyst, and V is the volume of the catalyst. catalyst to study an industrial steam reformer [25]. Pedernera et al.
The advantage of using an effectiveness factor approach is realized used a two-dimensional steady-state heterogeneous model with im-
when the factor may be evaluated from analytical or empirical equa- posed wall temperature profiles to represent top-fired and side-fired
tions, independently of the integration of the reactor tube model. A list reformers [26]. However, the authors assumed the catalyst pellets were
of studies using different constant effectiveness factors is provided in isothermal, and ignored particle-fluid heat and mass transfer re-
[17]. The assumption of constant effectiveness factor throughout the sistances. Extensive models were given by Kagyrmanova et al. who
packed-bed might lead to inaccurate estimation of concentration and included two dimensions, volume change of the mixture and complex
temperature gradients in the reactor, however the computational ad- catalyst pellet shapes [27]. However, axial dispersion of heat and mass
vantage is lost if the effectiveness factor has to be evaluated by solving were neglected, and the tube wall temperature was fixed at a constant
the single pellet model to obtain a variable effectiveness factor.

2
B.V.R. Kuncharam and A.G. Dixon Fuel Processing Technology 200 (2020) 106314

value instead of specifying heat flux or temperature profiles. They resistances at the fluid-solid interface. The model also uses more rea-
presented limited temperature profiles and pellet concentration pro- listic heat flux profiles typical for a top-fired furnace reformer.
files. Radiation is neglected here for the high flow rates, long tubes and flux
More recently, dynamic heterogeneous reactor models for SMR wall conditions modeled in this study. The model includes the pressure
were also reported. Pantoleontos et al., extending the work of Xu and drop along the length of the reactor, and compares three empirical
Froment, developed a one-dimensional dynamic heterogeneous model equations, the Ergun equation which neglects wall effects, the Eisfeld-
for optimizing the SMR reactor [17]. The authors in their study ignored Schnitzlein equation [33], and the Di Felice-Gibilaro equation [34]
the radial gradients in the bed and the catalyst particle was assumed to which both account for wall effects to different extents. The superficial
be isothermal. Ghouse and Adams developed a one-dimensional tube velocity is calculated accounting for the effect of molar expansion
multi-scale heterogeneous model for a methane steam reformer to study during reaction along the reactor length as well as the effects of pres-
the dynamic effects of feed disturbances and validated their simulation sure and temperature and is compared with two commonly used sim-
results with industrial data available in the literature [28]. Their model plified versions. The novel contribution in this work is to evaluate some
ignored the radial variation of concentration and temperature gradients common assumptions made in 2D fixed-bed models of SMR. This is
in the reactor. done through the development of a model with minimal simplifications,
The assumptions considered in the above one-dimensional hetero- and using it as a benchmark for comparison with the simplified models.
geneous models may not be valid for an industrial reformer since sig-
nificant radial thermal gradients may exist in the bed and there are
strong gradients in the catalyst particle. A frequent simplification is to 2. Model development
assume constant density of the reacting mixture [29] which implies
constant axial velocity, as in [17] for example. Many models also es- A steady state 2D multi-scale packed bed reformer model is devel-
timated the change in superficial velocity in the packed-bed accounting oped to investigate the catalytic steam methane reforming reaction. The
for temperature and pressure effects but ignoring the effect of molar model takes into account the solid-gas phase mass and energy diffu-
expansion due to reaction [28]. One approach to this difficulty is to sional limitations along the reactor length. In lieu of using the catalyst
formulate the governing equations on a mass basis, rather than a molar effectiveness factor, mass and energy balances in the catalyst particle
one [30,31] which removes molar concentration from the convection are calculated and coupled with the gas phase model. The gas phase
term, but requires some additional effort to convert from the mass- model takes into account the axial and radial dispersion in the N = 5
based variables to the molar-based ones required by the reaction rate tube as well as finite solid-fluid mass and heat transfer coefficients,
terms. Finally, in many cases the pressure drop is either ignored or calculated using empirical correlations available in the literature. The
calculated using the Ergun equation [32] which ignores the wall effects pressure drop in the packed bed is calculated employing the Ergun
that are likely to be present in the narrow, low-N tubes used in steam equation for the base case which does not take into account the wall
reforming. effects. The molar expansion during reaction is also taken into account
This paper presents a two-dimensional multi-scale steady-state non- using an expression for the change in average molecular weight along
isothermal (both gas phase and catalyst phase), heterogeneous model to the reactor length. Fig. 1 shows the schematic of the multi-scale model
study an industrial packed-bed reactor for the SMR reaction. This model with gas-phase and solid-phase variables involved in the packed-bed
couples the fluid phase and catalyst phase mass and energy transport reactor.
equations as well as accounting for the mass and energy transfer The present model simulations employ the following assumptions to
describe the conditions in the reactor and catalyst particle: (a) The SMR

Fig. 1. Schematic of multi-scale model of packed-bed reactor [after [28]].

3
B.V.R. Kuncharam and A.G. Dixon Fuel Processing Technology 200 (2020) 106314

reactions in 1–3 are assumed to be occurring in the reactor; other re- kea ∂Tf ∂Tf ker L ∂Tf qw L
Tf |z = 0 = Tfeed; − = 0; = 0; − =
actions such as methane decomposition, Boudouard reaction, or CO L ∂z ∂r R2 ∂r R
z=1 r=0 r=1
reduction to lead to carbon deposition are ignored. This assumption is
(12)
valid when the steam to methane (S/C) ratio is higher than 1.2 [17]; (b)
Although the model can be extended to higher hydrocarbons, the pre- Fig. 2 shows a typical heat flux (qw) profile used to simulate the case
sent model only considers reforming of methane; (c) All catalyst par- of a top-fired furnace, which is imposed at the reactor tube wall. The
ticles in the packed bed are assumed to be spherical, of the same size profile shows that a high heat flux is imposed nearer to the inlet and
and they are randomly packed in the bed; (d) A 1D single catalyst decreases along the length of the reactor. A high heat flux at the inlet is
particle model is used to calculate the physical conditions inside the necessary because the inlet reforming mixture is not preheated to the
particle at a point (r,z) in the tube, the solid and fluid phase variables desired temperature and the main reactions are endothermic.
are coupled along the axial and radial directions of the reactor. The pure gas specific heat and gas mixture specific heat is calculated
using the following equations from Perry's Handbook, [39].
2.1. Gas phase equations Nc
Cp, mix = ∑i =1 Cpi yi (13)
A steady state two-dimensional equation is formulated to account d
for the mass balance in the gas-phase in the packed-bed reactor tube. Cpi = a + bT + cT 2 +
T2 (14)
The equation incorporates the effect of the change of the average mo-
The pure gas viscosity is calculated using the equation of Chapman-
lecular weight on gas velocity along the tube length [35]. The axial and
Enskog theory and the gas mixture viscosity is calculated using the
radial dispersion in the tube, the mass and heat transfer resistances
Chapman-Enskog-Brokaw-Wilke equation [40,41].
between the gas and solid phases are included and are calculated from
empirical equations available in the literature. The reactor and catalyst
2.2. Pressure drop in the packed bed
parameters used in the model are shown in Table 1 and are typical for
steam methane reforming.
The differential form of the Ergun equation is used to calculate the
The mass balance equation can be written as follows,
pressure drop in the packed bed assuming the radial pressure gradient is
G ∂yi Gyi ∂MWave negligible, the equation is written as
− + + Lap (1 − εb ) k g (Cpis − yi Ctot )
MWave ∂z (MWave )2 ∂z
dP (1 − εb ) u ⎡ (1 − εb ) μ
1 ∂ ⎛ ∂y L 1 ∂y − = 150 + 1.75ρf u⎤
=− Ctot Dea i ⎞ − 2 ⎛Ctot rDer i ⎞
⎜ ⎟ ⎜ ⎟ dz dp εb3 ⎢ ⎣ d p
⎥
⎦ (15)
L ∂z ⎝ ∂z ⎠ R r⎝ ∂r ⎠ (5)
The pressure drop in the reactor is alternatively calculated using the
The above equation is solved with the following boundary condi-
Eisfeld-Schnitzlein [33] model which takes into account the wall effects
tions,
on the pressure drop in packed beds. The equation is written as follows:
Dea ∂y ∂yi ∂yi
yi |z = 0 = yi0 ; − Ctot i = 0; = 0; =0 dP (1 − εb ) u ⎡ K1 Aw (1 − εb ) μ A
L ∂z ∂r ∂r (6) − = + w ρf u⎤
z=1 r=0 r=1
dz dp εb3 ⎢ ⎣ dp Bw ⎥
⎦ (16)
where Ctot = P/RgTf, z = z∗/L and r = r∗/R.
where the wall correction terms are determined using
The average molecular weight in the reactor is defined by the fol-
2 2
lowing equation,
Aw = 1 +
2
and B w = ⎡ ⎛ dp ⎞ ⎤
N 3(D / dp )(1 − εb ) ⎢k1 D + k2 ⎥
⎜ ⎟

⎣ ⎝ ⎠ ⎦ (17)
MWave = ∑ yi (MW )i
i=1 (7) where D is the hydraulic diameter of the reactor tube. The terms k1 and
k2 are the coefficients of the wall effect correction terms, for spherical
The sum of species mole fractions ∑yi = 1 is enforced by solving for
particles their values are equal to 1.15 and 0.87, respectively. K1 is the
only N-1 terms and yN is obtained by subtraction. Therefore, the
coefficient of the pressure drop correlation term, equal to 154 for
average molecular weight can be written as
spherical particles.
N −1
MWave = (MW )N + ∑i =1 yi [(MW )i − (MW )N ] (8) 2.3. Catalyst pellet equations
The effect of the change of average molecular weight due to reaction
is accounted for using the following equation, A spherical catalyst is assumed in the packed bed, the mass and
energy balances in the 1D catalyst particle are calculated and coupled
d (MWave ) N −1 dyi with the above gas phase model. The mass transport in the catalyst
dz
= ∑i =1 [(MW )i − (MW )N ]
dz (9)
Table 1
In Eq. (5) G = ρu and the superficial velocity (u) is computed as-
Model parameters and feed composition (from [36]).
suming an ideal gas,
Parameters Feed composition (mol %)
u (MWave )in Pin T
=
uin MWave P Tin (10) Temperature, T0 (K) 824.15 CH4 23.92
Pressure, P0 (bar) 21.59 CO 0.05
The following equation shows the gas-phase energy balance in the Velocity, u0 (m/s) 1.6237 CO2 7.76
packed bed reactor tube. Reactor tube diameter (m) 0.127 H2 0.5
Reactor tube length (m) 10.6 H2O 67.77
kea ∂2Tf L 1 ∂Tf ⎞ ∂T
pf u f
Catalyst pellet diameter (m) 0.0254
+ 2 ⎛rker ⎜ + Lap (1 − εb ) hg (Tps − Tf ) = ρf C
⎟
Catalyst Pellet density (kg/m3) 1947
L ∂z 2 R r⎝ ∂r ⎠ ∂z
Density of packed bed (kg/m3) 1130
(11) Porosity of packed bed (-) 0.42
Gas thermal conductivity (W/m·K) 0.1635
The above equation is solved with the following boundary condi- Pellet thermal conductivity (W/m·K) 1.0
tions,

4
B.V.R. Kuncharam and A.G. Dixon Fuel Processing Technology 200 (2020) 106314

Table 2
Empirical correlations.
Coefficient Equation Equation Reference
number

Axial mass 1
=
0.73εb
+
0.5 26 [44]
Peam ReSc 9.7εb
dispersion 1+
ReSc
Gas-solid mass Sh = 2 + 1.1Re0.6Sc1/3 27 [45]
transfer
Axial heat 1 kp / kg 0.73εb 28 [44]
= + + 0.5
dispersion Peah RePr RePr
Gas-solid heat Nu = 2 + 1.1Re0.6Pr1/3 29 [45]
transfer
Radial mass 1
=
0.73εb
+
1 30 [46]
dispersion Perm ReSc ⎡ 2 2⎤
7 ⎢2 − ⎛1 − ⎞ ⎥
⎣ ⎝ N⎠ ⎦

Radial heat 1 (krs / kf ) 1 31 [47]

= +
dispersion Perh RePr ⎡ 2 2⎤
7 ⎢2 − ⎛1 − ⎞ ⎥
⎣ ⎝ N⎠ ⎦
Fig. 2. A typical heat flux profile for a top fired furnace reformer imposed at the
reformer wall [37,38].

particle is described using the following equation, Dkl =

1 × 10−7T1.75 ( 1
Mk
+
1
Ml )
P [ (∑ νk )1/3 + (∑ νl )1/3] (25)
NR
∂ ⎛ ∂Cpi ⎞ 2 2 ∂Cpi
− ⎜Dpi
∂ξ ⎝
⎟ = R
∂ξ ⎠
p ∑ αij ρp rj + ξ Dpi ∂ξ where Ml and Mk are the molecular weights of gas species (g/mol), T is
j=1 (18) the temperature (K), P is the total pressure (atm), ∑νk and ∑νl are the
sums of structural atomic volume increments (∑νCH4 = 24.42, ∑νCO =
with boundary conditions
18.9, ∑νCO2 = 26.9, ∑νH2 = 7.07 and ∑νH2O = 12.7).
∂Cpi ∂Cpi
= 0; −Dpi = Rp k g )Cpi |ξ = 1 − yi Cbulk )
∂ξ ξ =0
∂ξ ξ =1 (19) 2.5. Empirical correlations

The energy balance considering heat of reaction is described using Empirical equations are used to calculate the interphase heat and
the following equation mass transfer coefficients and the mass and energy dispersion coeffi-
NR cients in the radial and axial directions as functions of velocity, which is
∂ ⎛ ∂Tp ⎞ 2 2 ∂Tp
− ⎜k p ⎟ = R
∂ξ ⎝ ∂ξ ⎠
p ∑ ρp (−∆Hr,j ) rj + ξ
kp
∂ξ
changing in the reactor. The calculations also consider the changing
j=1 (20) input variables such as heat flux, viscosity and specific heat along the
length of reactor. Table 2 lists the equations used and their sources.
with boundary conditions

∂Tp ∂Tp 2.6. SMR kinetic data

= 0; −kp = Rp hg (Tp |ξ = 1 − Tf )
∂ξ ξ =0
∂ξ ξ =1 (21)
The methane steam reforming kinetics for Ni supported on an alu-
where ξ = s/Rp; s is the pellet radial coordinate and Rp is the radius of mina catalyst are taken from the literature [48]. The rate expressions of
the catalyst pellet. reactions 1–3 are reproduced here.
3 P
PH CO
2.4. Effective diffusivity ⎛1 − 2
/ K p1 ⎞
PCH4 PH2 O 0.5 P CH4 P H2 O
r1 = k1 ⎝ ⎠
The effective diffusivity of species inside the catalyst pellet is cal- PH2.52 DEN 2 (32)
culated using an approach based on Hite and Jackson [42], assuming a
4 P
porosity (εp) of 44% and tortuosity (τ) of 3.54. The ratio of fluxes is ⎛1 −
PH
2 CO2
/ K p2 ⎞
PCH4 PH2 O ⎝ P CH4 P H2 O2
evaluated assuming a dominant reaction and is set to the ratio of the r1 = k2 ⎠
stoichiometric coefficients (Nl/Nk = αl/αk). PH1.75
2
DEN 2 (33)

1
=
N
kl
1
(
∑l =sp1 Δ yl − yk N l
k
N
) PCO PH2 O 0.5 (1 − P H2 PCO2
PCO P H2 O
/ K p3 )
Dke N N
1 − yk ∑l =sp1 N l r1 = k3
k (22) PH0.52 DEN 2 (34)
where where P is the partial pressure of species inside the catalyst particle.
1 1 1 DEN is a dimensionless parameter and defined as
= e + e e ym
Δkl Dkl DKk DKl ∑m e DEN = 1 + K CO pCO + K H2 p H2 + K CH4 pCH4 + K H2 O p H2 O / p H2
DKm (23) (35)

The Knudsen diffusivity DKl is calculated using the kinetic theory of The rate coefficients, equilibrium constants, and adsorption coeffi-
gases [43], cients were all taken from the original reference and are used here
0.5 without further modification. The net rate of production of each species
T
DKl = 97rp ⎛ ⎞ ⎜ ⎟ can be written as follows,
⎝ Mi ⎠ (24)
RCH4 = −(r1 + r2), RCO = r1 − r3, RCO2 = r2 + r3, R H2
The binary molecular diffusivity is calculated using a semi-empirical
method of Fuller, Schettler and Giddings [43]. = 3r1 + 4r2 + r3, R H2 O = −(r1 + 2r2) (36)

5
B.V.R. Kuncharam and A.G. Dixon Fuel Processing Technology 200 (2020) 106314

2.7. Numerical methods separately and in sequential order specified by the user. The simula-
tions were carried out on a computer equipped with Intel Xeon CPU E5-
The model developed above was solved using the finite element 2680 @2.70 GHz dual processor and 120 GB RAM on a 64 bit operating
method available in COMSOL Multiphysics version 4.3a, using separate system.
domains for the gas-phase and the catalyst phase to solve the mass and The tube mesh density was determined from runs with a varying
heat transport as well as pressure drop in the packed bed, similarly to number of elements to ensure mesh-independence, and extensive trial-
[49]. The two models were linked using coupling variables as described and-error was made for the pellet mesh to arrive at a fine enough grid
in the COMSOL user guide. The finite element mesh used in the model near the particle surface, to capture the sharp gradients there. The final
consisted of 400 uniform elements for the 2D domain of the reactor mesh had 20 elements in both axial and radial tube dimensions, and 60
tube and 60 distributed elements for the 1D spherical catalyst pellet, elements along the pellet radius, concentrated at the surface.
refined at the pellet surface, resulting in 341,427 degrees of freedom
(DOF) and 23,442 internal DOFs. The coupled model is solved in
COMSOL using a segregated solver which solves different variables

Fig. 3. Simulation results along the length of packed reactor: (a) CH4 conversion, (b) Mole fraction of gas species, (c) Reforming temperature, (d) Pressure, (e) Linear
velocity, and (f) Average molecular weight.

6
B.V.R. Kuncharam and A.G. Dixon Fuel Processing Technology 200 (2020) 106314

3. Results and discussion of different gas species along the length of the reactor. The CH4 mole
fraction decreases along the length of the reactor, due to its conversion
The resulting model was first solved employing the Ergun equation in reactions 1 and 2. H2 mole fraction increases steadily along the re-
for predicting the pressure drop in the packed bed reactor, the base actor as it is the main product of both reactions. CO2 is initially the most
case. The results were then compared with simulation results from the plentiful side product, as reaction 2 dominates at the lower tempera-
model using the Eisfeld-Schnitzlein equation for pressure drop, which tures at the reactor inlet. Further down the tube, at higher tempera-
takes into account the wall effects. Simulations were also peformed for tures, reaction 1 becomes dominant causing the CO mole fraction to
two different cases with simplification for the superficial velocity (u) in increase. The CO2 mole fraction passes through a maximum due to its
Eq. (5), these cases are defined as follows. In case 1, the reaction initial rapid production in reaction 2, which then decreases relative to
mixture density, and thus the superficial velocity, in the reactor is as- reaction 1 and the CO2 formed is partly consumed by the reverse water-
sumed be constant (u = u0). In case 2, reaction mixture density is gas shift reaction 3. The composition of the outlet gas mixture in terms
evaluated accounting for the change in total molar concentration due to of mole percent on wet basis is CH4 = 1.13, CO = 14.18, CO2 = 6.61,
T, P assuming the ideal gas law, H2 = 47.78. Fig. 3c shows that the temperature increases along the
length of the reactor; the temperature at the reactor outlet is observed
P0 Tf
u = u0 as 1189.6 K. Initially, the endothermic reaction slows the rate of tem-
P T0 (37)
perature increase, then the higher wall heat flux compensates for the
In the base case, the reaction mixture density and superficial velo- endotherm, and a faster increase is observed. The rate of temperature
city is evaluated accounting for all changes in density in the reactor, increase is less at longer tube lengths, because of the lower heating rate
qw.
P0 Tf MWave,0 Fig. 3d presents pressure along the length of the packed bed reactor,
u = u0
P T0 MWave (38) calculated using the Ergun equation (in Eq. (15)). The simulation re-
sults show that pressure decreases nonlinearly along the length of the
where 0 denotes the tube inlet conditions.
reactor, the outlet pressure is predicted as 19.981 × 105 Pa and the
pressure drop is calculated as 1.609 × 105 Pa which is 7.45% of the
3.1. Study using Ergun equation for pressure drop inlet pressure. Fig. 3e shows the linear superficial velocity of the gas
along the length of the packed bed reactor. Results show that the ve-
The 2D non-isothermal model simulations with inlet conditions locity increases strongly along the length of reactor, the outlet velocity
specified in Table 1 were performed using the Ergun equation for the is observed as 3.66 m/s, which is more than twice the inlet velocity,
pressure drop in the packed bed reactor. Fig. 3 presents the model si- caused by the changes in pressure, temperature and the molar expan-
mulation results. Fig. 3a shows that CH4 conversion increases along the sion due to the reactions.
length of the reactor, for the base conditions the conversion at the re- Fig. 3f shows the average molecular weight along the length of the
actor outlet is observed to be 93.2%. Fig. 3b presents the mole fractions

Fig. 4. Tube radial profiles at different axial positions, (a) CH4 conversion, (b) Linear velocity, (c) Reforming temperature, (d) Average molecular weight.

7
B.V.R. Kuncharam and A.G. Dixon Fuel Processing Technology 200 (2020) 106314

packed bed reactor, calculated using Eq. (9). The results show that the which can be attributed to the radial temperature gradients. Note that
average molecular weight of the gas mixture decreases along the length in this model plug flow was assumed, so that there was no decrease in
of the reactor, predicting an outlet value of 0.0135 kg/mol, a decrease velocity at the tube wall due to no-slip.
of over 30% from the initial value. This shows that the change in moles Fig. 4c shows reforming temperature in the radial direction at dif-
and average molecular weight contributes strongly to the increase in ferent axial positions. The results show that reforming temperature
superficial velocity. increases up to 60 K in the radial direction, with maximum temperature
Fig. 4 presents results showing CH4 conversion, reforming tem- at the wall of the reactor tube (r/R = 1) due to the strong heat flux
perature, superficial velocity and average molecular weight in the ra- applied there. Fig. 4d shows the average molecular weight in the radial
dial direction at different lengths of the packed bed reactor. Fig. 4a direction at different axial positions. It can be observed that the average
shows that the CH4 conversion remains constant along the radial po- molecular weight remains constant in the radial direction, showing that
sition at different lengths of packed bed reactor. Fig. 4b shows the radial dispersion has no effect on change in average molecular weight.
variation in superficial velocity in the radial direction at different po- There are also no significant radial concentration gradients in the re-
sitions. The results show that velocity increases slightly with radius, actor.

Fig. 5. Results showing the radial profiles of temperature and mole fraction of gas species inside the catalyst pellet. (a) CH4, (b) CO (c) CO2 (d) H2 (e) H2O (f)
Temperature.

8
B.V.R. Kuncharam and A.G. Dixon Fuel Processing Technology 200 (2020) 106314

Fig. 5 shows the mole fractions of the gas species inside the catalyst positions; the inset figure shows a closer look at the profiles near the
pellet; a representative catalyst pellet was taken at the center of the surface. Results show that for all axial positions the rate of reaction
reactor tube, at different axial locations. Fig. 5a presents the CH4 mole increases from a negative value at the surface of the pellet to a constant
fraction inside the pellet, the results show that CH4 mole fraction de- value and nearly zero inside the pellet. The negative reaction shows
creases from the surface of the pellet to a constant value at lower z/L that the reverse water-gas-shift dominates the net rate of reaction near
and remains constant inside the pellet. This is due to the diffusion and the surface. It also shows that reaction occurs mostly at the surface of
conversion of CH4 in reactions 1 and 2. The gradient near the surface is the pellet using only the outer 2% of the pellet. The reaction rates inside
steeper at lower z/L because the difference between bulk gas and pellet near-wall pellets showed similar trends although at higher rates of re-
is higher at lower z/L. At higher values of z/L the reverse trend is seen, action (therefore, not included here).
there is a slight decrease in CH4 mole fraction from pellet center to In nearly all the plots of species mole fraction in Fig. 5, at z/L > 0.6
surface. The amount of CH4 in the pellet center is much lower but is the species profiles reversed, i.e. the methane and water were slightly
significantly above zero. Fig. 5b presents the CO mole fraction along the
pellet coordinate. It can be observed that at lower z/L the CO mole
fraction increases from the surface of the pellet to a constant value
inside the pellet. The profiles show the opposite trends to those ob-
served in the case of the CH4 mole fraction profile i.e. the CO mole
fraction on the surface of the catalyst pellet is less than inside the
catalyst. At higher z/L the CO mole fraction increases from the surface
value and passes through a small maximum before decreasing to a
lower value in the pellet center. Fig. 5c presents the CO2 mole fraction
along the pellet coordinate. The results show that for z/L = 0 and z/
L = 0.2, the CO2 mole fraction increases from the surface value to a
constant value in the catalyst pellet. The CO2 is produced at a high rate
near the pellet surface at lower z/L, which results in a high CO2 gra-
dient to give a flux sufficient to match the reaction rate. At higher z/L
the CO2 production rate is low leading to lower gradients in the mole
fraction, and at intermediate values the CO2 mole fraction decreases
slightly from the surface value. From z/L = 0.4 to 1.0 it then increases
again to a higher value inside the pellet. At all axial positions CO2 mole
fraction profile is flat inside most of the pellet.
Fig. 5d presents the H2 mole fraction along the pellet coordinate at
different axial positions. The results show that H2 concentration also
increases from the surface value to a constant value inside the catalyst
pellet at lower z/L. It can also be observed that H2 concentration in-
creases from z/L = 0 to z/L = 1, because of its production in the re-
forming reactions. Fig. 5e presents the H2O mole fraction inside the
pellet at the center of the reactor tube. The results show that H2O mole
fraction decreases from the surface of the catalyst pellet and remains
flat inside the catalyst for z/L < 0.4. However, at other axial positions
of the catalyst pellet the mole fraction increases slightly towards the
pellet center. H2O mole fraction decreases at increasing axial position
due to its consumption in the reforming reactions. Fig. 5f presents the
temperature inside the catalyst pellet, the results show that temperature
slightly decreases from surface to inside the pellet due to the en-
dotherm.
Fig. 6 shows the rates of reaction inside the tube-center catalyst
pellet. Fig. 6a presents the rate of reaction 1 inside the pellet at different
axial positions and the inset figure shows detail near the surface. Re-
sults show that for z/L = 0 and z/L = 1, the rate of reaction is nearly
zero for all radial positions, at all other axial positions rate of reaction
decreases from a maximum value at the surface of the pellet to a con-
stant value and is nearly zero inside the majority of the pellet. This
shows that the reaction occurs mostly at the surface of the pellet using
only the outer 2–3% of the pellet radius, inside the pellet equilibrium is
attained. The reaction rate increases for z/L = 0 to 0.4, then decreases.
It is slightly negative at bed exit showing that the reverse reaction is
taking place there.
Fig. 6b with inset presents the rate of reaction 2 inside the pellet at
different axial positions. There is substantial reaction already at z/
L = 0, increasing to z/L = 0.2 then decreasing to near zero further
downstream. The results show that at all other axial positions except at
z/L = 1 where reaction 2 proceeds in the reverse direction, the rate of
reaction decreases from a maximum value at the surface of the pellet to
zero inside the pellet. These results also show that the reaction 2 occurs
mostly at the surface of the pellet using only the outer 5% of the pellet. Fig. 6. The rate of reactions inside the catalyst pellet. (a) Reaction 1, (b)
Fig. 6c presents the rate of reaction 3 inside the pellet at different axial Reaction 2, and (c) Reaction 3.

9
B.V.R. Kuncharam and A.G. Dixon Fuel Processing Technology 200 (2020) 106314

higher in the pellet center than at the surface, while the hydrogen, CO2 the specific volume of the gas flow. The superficial velocity at the re-
and CO were slightly lower. The results in Fig. 6 showed that the re- actor outlet is predicted as 3.66 m/s and 3.61 m/s for the Ergun and
action rates were zero within the pellets for most of the radius, but Eisfeld-Schnitzlein studies respectively. Fig. 8c shows that the CH4
Fig. 5 showed that the reactant mole fractions were not zero, so conversion increases along the length of the reactor in both cases. It is
therefore the pellet interiors must be at equilibrium. Equilibrium cal- also observed that CH4 conversion obtained is the same to within the
culations for the different temperatures and pressures along the tube accuracy of the graph. The CH4 conversion obtained at the reactor
length were run to investigate this idea. A plot of the equilibrium mole outlet is 93.18% and 93.09% in the Ergun and Eisfeld-Schnitzlein stu-
fractions at the temperatures along the tube axis is given in Fig. 7 and dies respectively. Fig. 8d shows that the temperature increases along
compared to the simulated mole fractions from Fig. 3. the length of the reactor in both cases. It is also predicted that the re-
We can see that the mole fraction of CH4 is initially above equili- forming temperature is almost the same over the reactor length, fin-
brium while those for the products are below. By z/L = 0.6, however, ishing higher by 0.4 K in the case of the Eisfeld-Schnitzlein study, which
the curves coincide showing that the fluid and pellet compositions is a negligible difference. The reforming temperatures at the reactor
simply track the changes in equilibrium composition caused by the outlet are predicted to be approximately 1190 K. Fig. 8e shows simu-
increasing temperature. This allows the explanation of the pellet species lation results comparing the average molecular weight predicted using
profiles. At lower z/L as the reaction mixture changes to catch up with the two pressure drop equations. Results predict similar trends in both
equilibrium the profiles show expected behavior and the heat supply cases, the average molecular weight of the gas mixture decreases along
from the tube wall and the heat sink in the pellets result in flat pellet the length of the reactor. However, the average molecular weight for
temperature profiles only slightly lower than the fluid temperature. At the Ergun study is predicted to be slightly lower than for the Eisfeld-
higher z/L where equilibrium prevails, we can notice that due to the Schnitzlein study. This is expected as the higher conversion in the Ergun
reduced endothermic heat sink at the pellet surface and the continued case corresponds to a larger change in moles. Finally, Fig. 8f shows
heating from the tube wall, the pellet surface temperature is raised simulation results for the mole fractions of different gas species along
higher than in the pellet center, but is still lower than the fluid tem- the length of the reactor. Overall, the differences between the two
perature. At equilibrium this results in higher CH4 and H2O, and lower models under the conditions studied were quite small, the largest ef-
product H2 and CO in the pellet center. The equilibrium CO2 mole fects were on pressure and velocity. The difference in pressure drop of
fraction goes through a maximum as it is initially produced at a high about 1.3% had only a very slight effect on reactor performance. This
rate by reaction 2 but is then reduced by the RWGS reaction. So, in the result may give an indication why the Ergun equation continues to be
first part of the tube it behaves like a product, then in the later part it used despite the existence of supposedly more accurate correlations
behaves like a reactant. that account for wall effects.
Effectiveness factors were calculated for all reactions using Eq. (4) To illustrate the effects of a more substantial change in pressure
at the tube centerline and along the axial coordinate. The values started drop, simulations were run with a different correlation, that of Di Felice
out relatively high at the tube inlet but decreased rapidly to small va- and Gibilaro [34]. The stronger effect of the tube wall in their corre-
lues in the range 0.005 to 0.02. These are in reasonable agreement with lation is represented by the following equations
the values found by Xu and Froment [20] which are in the range 0.01 to
0.03, although their values for the two reforming reactions initially dP (1 − εb ) u ⎡ (1 − εb ) μ u ⎤
− = 150 + 1.75ρf
increase before decreasing along the tube similarly to ours. Their values dz dp εb3MDG ⎢ ⎣ d p M ⎥
DG ⎦ (39)
for the water-gas shift reaction change sign at z/L = 0.3 roughly, as the 2
D
reaction proceeds in the reverse direction at the particle surface while ⎛ dp − 1 ⎞
continuing forward in the particle interior. We found that the reverse MDG = 2.06 − 1.06 ⎜ ⎟
⎜ D / dp ⎟
water-gas shift reaction predominated almost the entire length of the ⎝ ⎠ (40)
tube, both on the particle surface and inside it. Only the second re-
The results are shown as the dotted lines in Fig. 8.
forming reaction effectiveness factor showed negative values for our
A lower pressure drop will result in higher partial pressures but
conditions.
lower superficial velocities. The results show that for the lower pressure

3.2. Study including tube wall effects in pressure drop equation

The 2D non-isothermal model was simulated for the same condi-

tions as in the previous study but employing the Eisfeld-Schnitzlein
equation (Eq. (16)) for evaluating pressure drop in the packed bed re-
actor. The Eisfeld-Schnitzlein equation takes into account the wall ef-
fects on the pressure drop in the packed bed using correction factors.
Fig. 8 presents simulation results comparing the performance of the
packed bed reactor using both Ergun and Eisfeld-Schnitzlein equations
for evaluating pressure drop. Fig. 8a presents pressure along the length
of packed bed reactor, calculated using both equations. Results predict
that pressure decreases along the length of the reactor in both cases,
however, the higher pressure drop is observed in the case of the Ergun
study. This is expected as the Eisfeld-Schnitzlein correlation allows for
higher void fraction and thus by-pass flow at the reactor tube wall,
which would decrease pressure drop. The pressure at the reactor outlet
is predicted as 19.981 × 105 Pa and 20.244 × 105 Pa in the case of the
Ergun and Eisfeld-Schnitzlein studies respectively, so that the total
difference is only 1.3%. Fig. 8b shows simulation results of gas super-
ficial velocity using both pressure-drop equations. It is observed that
the superficial velocity increases along the length of the reactor in both Fig. 7. Equilibrium mole fractions of the main species along the tube length at
cases. A slightly higher superficial velocity is predicted in the case of the prevailing temperature at each axial position (dashed curves show equili-
the Ergun study, as the lower pressure throughout the tube increases brium calculations).

10
B.V.R. Kuncharam and A.G. Dixon Fuel Processing Technology 200 (2020) 106314

Fig. 8. Results comparing the performance of packed bed reactor using three pressure drop equations, (a) Pressure, (b) Linear velocity, (c) CH4 conversion, (d)
Reforming temperature, (e) Average molecular weight, and (f) Mole fraction of gas species.

drop, the superficial velocity decreases considerably, the CH4 conver- 3.3. Study with simplified superficial velocity equations
sion decreases by a very small amount, to 92.59%, while the average
molecular weight increases, and the species mole fractions increase or As described peviously, the model was also run for two more case
decrease in line with the decreased conversion, again by small amounts. studies using simplified equations for superficial velocity (u) in the
The temperature decreases slightly, implying that the decreased en- transport equations and using the Ergun equation to calculate the
dotherm, due to lower conversion, is small and has little effect. The lack pressure drop in the packed bed. Case 1 assumes the superficial velocity
of change in conversion and composition when total pressure, and thus remains at the constant inlet value (u = u0) which is equivalent to an
partial pressures of reactants, are reduced is attributed to the reaction assumption of constant reaction mixture density. Case 2 accounts for
rates being already very high. The reactor behavior is controlled by the the change in total molar concentration with temperature (T) and
rate of heat addition and the change in reaction equilibrium with in- pressure (P) assuming an ideal gas, but neglects the effect of the change
creasing temperature. in moles. The results predicted by these two case studies using simpli-
fied models are compared with the rigorous model developed in the

11
B.V.R. Kuncharam and A.G. Dixon Fuel Processing Technology 200 (2020) 106314

earlier section (Base case). shows that the pressure decreases along along the length of the packed
Fig. 9 presents simulation results comparing the performance of the bed reactor in all three cases, however, the higher pressure drop is
packed bed reactor in the three different cases described above. observed in the order Base > Case 2 > Case 1 due to the higher
Fig. 9a shows that CH4 conversion increases along the length of the velocities which are also in that order. The pressure at the reactor outlet
reactor in all three cases. Simulation results predict a higher CH4 con- is predicted as 20.66 × 105 Pa, 20.34 × 105 Pa, and 19.97 × 105 Pa for
version in Case 2 than Case 1, however, Case 2 predicts conversion less Case 1, Case 2 and Base studies, respectively. Fig. 9d shows the su-
than that obtained in the Base model. The conversion obtained at the perficial gas velocity in the packed bed reactor for the three cases. It is
reactor outlet is 75.5%, 82.3% and 93.7% in Case 1, Case 2 and Base predicted that superficial velocity increases along the length of the re-
case studies, respectively. Fig. 9b shows that the reforming temperature actor in Case 2 and Base models. The velocity at the reactor outlet can
increases along the length of the reactor in all cases. The reforming be found as 1.62 m/s, 2.54 m/s and 3.68 m/s for Case 1, Case 2 and
temperatures predicted in Case 2 and Base models initially have neg- Base case, respectively. As can be observed in the results, the effect of
ligible difference, but then diverge slightly with the Base case being molar change on velocity in the reactor is evident and shows substantial
lower due to the higher endotherm from the higher conversion. Fig. 9c increase in velocity along the entire length of the reactor. Fig. 9e shows

Fig. 9. Results comparing the performance of packed bed reactor in three different velocity case studies, (a) CH4 conversion, (b) Reforming temperature, (c) Pressure,
(d) Linear velocity, and, (e) Species mole fractions.

12
B.V.R. Kuncharam and A.G. Dixon Fuel Processing Technology 200 (2020) 106314

simulation results comparing the species mole fractions predicted in the CRediT authorship contribution statement
three different cases. We observe lower H2, CO and CH4, and higher
CO2 for the Base case, while for Case 1 the mole fractions are generally Bhanu Vardhan Reddy Kuncharam: Methodology, Software,
lower than for Case 2. Validation, Investigation, Data curation, Writing - original draft,
The changes in velocity can be straightforwardly attributed to the Visualization. Anthony G. Dixon: Conceptualization, Investigation,
decrease in reactant mixture density as the models progress from con- Writing - review & editing, Visualization, Supervision.
stant density, to including the effect of the increased temperature on
density, and finally to including the effect of the molar expansion to Declaration of competing interest
reduce density through decreased average molecular weight. The total
pressure follows the expected trend with velocity. The conversion of The authors declare that they have no known competing financial
Case 1 and Case 2 is low, which can be traced back to the incorrect interests or personal relationships that could have appeared to influ-
accounting for the total molar flow rate in these two cases. Similar ence the work reported in this paper.
reasoning accounts for the mole fractions of the products for the sim-
plified cases, which are too high, while the mole fractions of all species Acknowledgement
in the Base case are in excellent agreement with equilibrium conver-
sion. This research did not receive any specific grant from funding
From these results we can conclude that it is essential to include the agencies in the public, commercial, or not-for-profit sectors.
molar change on reaction in order to correctly predict methane con-
version and species mole fractions. The axial temperature profile is References
well-represented if only changes in temperature and pressure are in-
cluded. Differences in pressure drop between the cases are approxi- [1] D.J. Wilhelm, D.R. Simbeck, A.D. Karp, R.L. Dickenson, Syngas production for gas-
mately 3%, which is relatively minor and well within typical experi- to-liquids applications: technologies, issues and outlook, Fuel Process. Technol. 71
(2001) 139–148, https://doi.org/10.1016/S0378-3820(01)00140-0.
mental error. [2] J.R. Rostrup-Nielsen, Syngas in perspective, Catal. Today 71 (2002) 243–247,
https://doi.org/10.1016/S0920-5861(01)00454-0.
[3] M.A. Murmura, S. Cerbelli, M.C. Annesini, Modeling fixed bed membrane reactors
for hydrogen production through steam reforming reactions: a critical analysis,
4. Conclusions Membranes 8 (2) (2018) 34, https://doi.org/10.3390/membranes8020034.
[4] R. Currie, M.W. Fowler, D.S.A. Simakov, Catalytic membrane reactor for CO2 hy-
This paper presents a rigorous non-isothermal 2D model of a packed drogenation using renewable streams: model-based feasibility analysis, Chem. Eng.
J. 372 (2019) 1240–1252.
bed reactor for steam methane reforming including axial and radial [5] M.A. Ashraf, O. Sanz, M. Montes, S. Specchia, Insights into the effect of catalyst
dispersion, as well as accounting for mass and heat transfer resistance at loading on methane steam reforming and controlling regime for metallic catalytic
the pellet-fluid boundary. The model also takes into account the heat monoliths, Int. J. Hydrog. Energy 43 (2018) 11778–11792, https://doi.org/10.
1016/j.ijhydene.2018.04.126.
and mass transfer in the catalyst particle and couples with fluid-phase
[6] A. Di Giuliano, K. Gallucci, S.S. Kazi, F. Giancaterino, A. Di Carlo, C. Courson,
equations without using catalyst effectiveness factors. The model also J. Meyer, L. Di Felice, Development of Ni- and CaO-based mono- and bi-functional
accounts for the effect of molar change during reaction. Simulations of catalyst and sorbent materials for sorption enhanced steam methane reforming:
the base case conditions showed a strong increase in the axial super- performance over 200 cycles and attrition tests, Fuel Process. Technol. 195 (2019)
106160, , https://doi.org/10.1016/j.fuproc.106160.
ficial velocity caused by the increase in temperature and the increase in [7] Q. Lu, Y. Hou, S.R. Laraib, O. Khalifa, K. Li, W. Xie, M. Cui, Y. Yang, Electro-cat-
moles for this reaction, for which stoichiometry results in a 44% mole alytic steam reforming of methane over Ni-CeO2/γ-Al2O3-MgO catalyst, Fuel
increase and the temperature increase is substantial, approximately Process. Technol. 192 (2019) 57–64, https://doi.org/10.1016/j.fuproc.2019.04.
021.
330 K or 40% of the inlet temperature. Tube radial profiles of conver- [8] A. Kumar, M. Baldea, T.F. Edgar, A physics-based model for industrial steam-me-
sion were flat, but a significant increase in temperature of up to 60 K thane reformer optimization with non-uniform temperature field, Comp. Chem.
from tube center to wall was observed. The reactions in the pellets took Eng. 105 (2017) 224–236, https://doi.org/10.1016/j.compchemeng.2017.01.002.
[9] A. Tran, M. Pont, A. Aguirre, H. Durand, M. Crose, P.D. Christofides, Real-time
place within a short distance of the pellet surface, as remarked by furnace balancing of steam methane reforming furnaces, Chem. Eng. Res. Des. 134
previous workers, with the pellet interior attaining equilibrium, so that (2018) 238–256, https://doi.org/10.1016/j.cherd.2018.03.032.
pellet profiles were flat over most of the pellet radius with sharp [10] C. Tsay, A. Kumar, T.F. Edgar, M. Baldea, Integrating steady-state and dynamic
models for multi-scale flowsheet optimization: a steam-methane reforming case
changes near the surface, showing strong diffusional limitations. As the
study, Comput. Aided Chem. Eng. 47 (2019) 403–408, https://doi.org/10.1016/
total temperature rise inside the pellet was < 1% of the absolute tem- B978-0-12-818597-1.50064-3.
perature, the usual assumption of an isothermal pellet is confirmed in [11] M.H. Wesenberg, H.F. Svendsen, Mass and heat transfer limitations in a hetero-
geneous model of a gas-heated steam reformer, Ind. Eng. Chem. Res. 46 (2007)
these simulations.
667–676, https://doi.org/10.1021/ie060324h.
Case studies were simulated using the Ergun, Eisfeld-Schnitzlein and [12] G.M. Karthik, V.V. Buwa, Particle-resolved simulations of methane steam reforming
Di Felice-Gibilaro pressure drop equations to evaluate the importance in multilayered packed beds, AICHE J. 64 (2018) 4162–4176, https://doi.org/10.
of accounting for the tube wall effect in this reactor with N = 5. Results 1002/aic.690350109.
[13] L.J. Christiansen, Use of modeling in scale-up of steam reforming technology, Catal.
showed that pressure drop and linear velocity were sllghtly higher Today 272 (2016) 14–18, https://doi.org/10.1016/j.cattod.2016.03.007.
when the Ergun equation was used, but there was no effect on any of [14] D. Pashchenko, Effect of the geometric dimensionality of computational domain on
the other variables that were evaluated. Thus for the reactions and SMR the results of CFD-modeling of steam methane reforming, Int. J. Hydrog. Energy 43
(2018) 8662–8673, https://doi.org/10.1016/j.ijhydene.2018.03.183.
conditions investigated here, the Ergun equation is satisfactory and [15] C.P.P. Singh, D.N. Saraf, Simulation of side fired steam-hydrocarbon reformers, Ind.
there is no gain in using equations that account for wall effects. Eng. Chem. Process. Des. Dev. 18 (1979) 1–7, https://doi.org/10.1021/
Simulations made under the simplifying assumptions of a constant i260069a001.
[16] H.M. Kvamsdal, H.F. Svendsen, T. Hertzberg, O. Olsvik, Dynamic simulation and
density reaction mixture and a density which is dependent on only optimization of a catalytic steam reformer, Chem. Eng. Sci. 54 (1999) 2697–2706,
temperature and pressure, ignoring the change in moles, showed sig- https://doi.org/10.1016/S0009-2509(98)00329-7.
nificant discrepancies in the two simplified cases. It can be concluded [17] G. Pantoleontos, E.S. Kikkinides, M.C. Georgiadis, A heterogeneous dynamic model
for the simulation and optimisation of the steam methane reforming reactor, Int. J.
that inclusion of the molar change is essential in this reaction system.
Hydrog. Energy 37 (2012) 16346–16358, https://doi.org/10.1016/j.ijhydene.
For reactions with weaker changes in mole number, or which are run at 2012.02.125.
low conversions, neglect of the mole change may well yield an accep- [18] M. Schwaab, A.L. Alberton, C.E. Fontes, R.C. Bittencourt, J.C. Pinto, Hybrid mod-
eling of methane reformers. 2. Modeling of the industrial reactors, Ind. Eng. Chem.
table approximation.
Res. 48 (2009) 9376–9382, https://doi.org/10.1021/ie801831m.
[19] J. Xu, G.F. Froment, Methane steam reforming, methanation and water-gas shift: I.
Intrinsic kinetics, AICHE J. 35 (1989) 88–96, https://doi.org/10.1002/aic.

13
B.V.R. Kuncharam and A.G. Dixon Fuel Processing Technology 200 (2020) 106314

690350109. https://doi.org/10.1029/JB088iS01p0B353.
[20] J. Xu, G.F. Froment, Methane steam reforming: II. Diffusional limitations and re- [33] B. Eisfeld, K. Schnitzlein, The influence of confining walls on the pressure drop in
actor simulation, AICHE J. 35 (1989) 97–103, https://doi.org/10.1002/aic. packed beds, Chem. Eng. Sci. 56 (2001) 4321–4329, https://doi.org/10.1016/
690350110. S0009-2509(00)00533-9.
[21] M.A. Soliman, S.S.E.H. El-Nashaie, A.S. Al-Ubaid, A. Adris, Simulation of steam [34] R. Di Felice, L.G. Gibilaro, Wall effects for the pressure drop in fixed beds, Chem.
reformers for methane, Chem. Eng. Sci. 43 (1988) 1801–1806, https://doi.org/10. Eng. Sci. 59 (2004) 3037–3040.
1016/0009-2509(88)87044-1. [35] J. Cropley, L. Burgess, R. Loke, The optimal design of a reactor for the hydro-
[22] S.S.E.H. Elnashaie, B.K. Abdalla, R. Hughes, Simulation of the industrial fixed bed genation of butyraldehyde to butanol, ACS Symp. Ser. 237 (1984) 255–270.
catalytic reactor for the dehydrogenation of ethylbenzene to styrene: heterogeneous [36] A.G. Dixon, CFD study of local transport and reaction rates in a fixed bed reactor
dusty gas model, Ind. Eng. Chem. Res. 32 (1993) 2537–2541, https://doi.org/10. tube: Endothermic steam methane reforming, Chem. Eng. Sci. 168 (2017) 156–177.
1021/ie00023a016. [37] D.A. Latham, K.B. McAuley, B.A. Peppley, T.M. Raybold, Mathematical modeling of
[23] R.M. Quinta Ferreira, M.M. Marques, M.F. Babo, A.E. Rodrigues, Modelling of the an industrial steam-methane reformer for on-line deployment, Fuel Process.
methane steam reforming reactor with large-pore catalysts, Chem. Eng. Sci. 47 Technol. 92 (2011) 1574–1586, https://doi.org/10.1016/j.fuproc.2011.04.001.
(1992) 2909–2914, https://doi.org/10.1016/0009-2509(92)87150-O. [38] L. Lao, A. Aguirre, A. Tran, Z. Wu, H. Durand, P.D. Christofides, CFD modeling and
[24] A.K. Avci, D.L. Trimm, Z. Ilsen Önsan, Heterogeneous reactor modeling for simu- control of a steam methane reforming reactor, Chem. Eng. Sci. 148 (2016) 78–92,
lation of catalytic oxidation and steam reforming of methane, Chem. Eng. Sci. 56 https://doi.org/10.1016/j.ces.2016.03.038.
(2001) 641–649, https://doi.org/10.1016/S0009-2509(00)00271-2. [39] R.H. Perry, D.W. Green, J. Maloney, Perry’s Chemical Engineers’ Handbook, 8th ed.,
[25] A.D. Nandasana, A.K. Ray, S.K. Gupta, Dynamic model of an industrial steam re- McGraw-Hill, New York, 2008.
former and its use for multiobjective optimization, Ind. Eng. Chem. Res. 42 (2003) [40] R.B. Bird, W.E. Stewart, E.N. Lightfoot, Transport Phenomena, 2nd ed., Wiley, New
4028–4042, https://doi.org/10.1021/ie0209576. Jersey, 2007.
[26] M.N. Pedernera, J. Piña, D.O. Borio, V. Bucalá, Use of a heterogeneous two-di- [41] B.E. Poling, J.M. Prausnitz, J.P. O’Connell, The Properties of Gases and Liquids, 5th
mensional model to improve the primary steam reformer performance, Chem. Eng. ed., McGraw Hill Professional, 2000.
J. 94 (2003) 29–40, https://doi.org/10.1016/S1385-8947(03)00004-4. [42] R. Hite, R. Jackson, Pressure gradients in porous catalyst pellets in the intermediate
[27] A.P. Kagyrmanova, I.A. Zolotarskii, N.V. Vernikovskaya, E.I. Smirnov, diffusion regime, Chem. Eng. Sci. 32 (1977) 703–709.
V.A. Kuz’min, N.A. Chumakova, Modeling of steam reforming of natural gas using [43] C. Geankoplis, Transport Processes and Unit Operations, 3rd ed., Prentice-Hall
catalysts with grains of complex shapes, Theor. Found. Chem. Eng. 40 (2006) International, Inc, 1993.
155–167, https://doi.org/10.1134/S0040579506020084. [44] M.F. Edwards, J.F. Richardson, Gas dispersion in packed beds, Chem. Eng. Sci. 23
[28] J.H. Ghouse, T.A. Adams, A multi-scale dynamic two-dimensional heterogeneous (1968) 109–123, https://doi.org/10.1016/0009-2509(68)87056-3.
model for catalytic steam methane reforming reactors, Int. J. Hydrog. Energy 38 [45] N. Wakao, T. Funazkri, Effect of fluid dispersion coefficients on particle-to-fluid
(2013) 9984–9999, https://doi.org/10.1016/j.ijhydene.2013.05.170. mass transfer coefficients in packed beds, Chem. Eng. Sci. 33 (1978) 1375–1384,
[29] B.M. Cruz, J.D. da Silva, A two-dimensional mathematical model for the catalytic https://doi.org/10.1016/0009-2509(78)85120-3.
steam reforming of methane in both conventional fixed-bed and fixed-bed mem- [46] E.U. Schlünder, Wärme- und Stoffübertragung zwischen durchströmten
brane reactors for the production of hydrogen, Int. J. Hydrog. Energy 42 (2017) Schüttungen und darin eingebetteten Einzelkörpern, Chemie Ing. Tech. - CIT. 38
23670–23690, https://doi.org/10.1016/j.ijhydene.2017.03.019. (1966) 967–979, https://doi.org/10.1002/cite.330380909.
[30] J. Solsvik, H.A. Jakobsen, Modeling of multicomponent mass diffusion in porous [47] R. Bauer, E.U. Schlünder, Effective radial thermal conductivity of packings in gas
spherical pellets: Application to steam methane reforming and methanol synthesis, flow. Part I. Convective transport coefficient, Int. Chem. Eng. 18 (1978) 181–188.
Chem. Eng. Sci. 66 (2011) 1986–2000, https://doi.org/10.1016/j.ces.2011.01.060. [48] K. Hou, R. Hughes, The kinetics of methane steam reforming over a Ni/α-Al 2 O
[31] K.R. Rout, H.A. Jakobsen, A numerical study of fixed bed reactor modelling for catalyst, Chem. Eng. J. 82 (2001) 311–328.
steam methane reforming process, Can. J. Chem. Eng. 93 (2015) 1222–1238, [49] J. Petera, L. Nowicki, S. Ledakowicz, New numerical algorithm for solving multi-
https://doi.org/10.1002/cjce.22202. dimensional heterogeneous model of the fixed bed reactor, Chem. Eng. J. 214
[32] S. Ergun, Fluid flow through packed columns, Chem. Eng. Prog. 48 (1952) 89–94, (2013) 237–246, https://doi.org/10.1016/j.cej.2012.10.020.

14