Computers and Chemical Engineering 24 (2000) 2623 – 2632

www.elsevier.com/locate/compchemeng

Mathematical model and numerical simulations of catalytic flow

reversal reactors for industrial applications
F. Aubé, H. Sapoundjiev *
CANMET Energy Di6ersification Research Laboratory (Natural Resources Canada), 1615 Lionel-Boulet Bl6d, P.O. Box 4800, Varennes,
Quebec, Canada, J3X 1S6

Received 28 April 1999; received in revised form 2 August 2000; accepted 3 August 2000

Abstract

A mathematical model and computer program have been developed for the study and the design of Catalytic Flow Reversal
Reactors (CFRR). It combines a transient two-dimensional heterogeneous model with a numerical method allowing the fast
formulation of new reactor configurations during the design phase. The program has been validated with experiments performed
for the oxidation of lean methane emissions. It is shown that the dynamic behaviour of the CFRR can be predicted accurately
for a wide range of conditions including small reactor diameter and low air flow rate if the radial effects related to the thermal
insulation are taken in account. Experiments were also accurately predicted for two different heat removal systems: hot air
withdrawal and an internal heat exchange, both in the mid-section of the reactor. © 2000 Elsevier Science Ltd. All rights reserved.

Keywords: Catalytic Flow Reversal Reactors (CFRR); Forced Unsteady State Conditions (FUSC); Two-dimensional heterogeneous model

1. Introduction pundzhiev, Grosev & Elenkov, 1991), methanol synthe-

In the past two decades, forced unsteady state condi-
tions (FUSC) have been shown to provide a significant
efficiency enhancement to a number of catalytic pro-
cesses. Benefits can include increased process productiv-
ity and selectivity, capital equipment savings and
operating cost reductions.
In fixed bed reactors, one way to create FUSC is a
periodic reversal of the feed stream direction. Recent
studies and applications in this field were initiated by
Matros, Boreskov and their research team in Novosi-
birsk. The applications reported in the literature include
SO2 oxidation (Boreskov, Bunimovich, Matros &
Ivanov, 1982; Sapundzhiev et al., 1988; Sapundzhiev,
Grozev & Elenkov, 1990; Bunimovich, Strots & Gold-
man, 1990; Bunimovich, Vernikovskaya, Strots,
Balzhinimaev & Matros, 1995), treatment of waste air
(Matros, Noskov, Chumachenko & Goldman, 1988;
Matros, Noskov & Chumachenko, 1993; Eigenberger &
Nieken, 1988; Chumachenko & Matros, 1990; Sa-
* Corresponding author. Tel.: +1-450-6525789; fax: + 1-450-
6525994.
E-mail address: hsapound@nrcan.gc.ca (H. Sapoundjiev).
sis (Matros, 1989; Thullie & Burghardt, 1990;
Neophytides & Froment, 1992; Vanden Bussche,
Neophytides, Zolotarski & Froment, 1993), production
of synthesis gas (Blanks, Wittrig & Peterson, 1990),
styrene synthesis (Kolios & Eigenberger, 1999), NOx
reduction (Bobrova, Slavinskaya, Noskov & Matros,
1988; Jirat, Stepanek, Kubicek & Marek, 1999; Matros,
Bunimovich, Strots & Mirosh, 1999), propane combus-
tion (Salinger & Eigenberger, 1996) and methane com-
bustion (Sapundzhiev, Chaouki, Guy & Klvana, 1993).
Recently, Kulkarni and Dudukovic (1996, 1997, 1998)
studied the application of CFRR to the coupling of
exothermic and endothermic reactions (e.g. combustion
and steam reforming of natural gas) in the same reac-
tor. In this case, the heat of exothermic reaction can be
utilized to maintain a favourable regime for the en-
dothermic reaction. In this approach, no external heat-
ing and cooling sources are needed. For the cases where
a part of the heat of exothermic reaction must be
extracted to maintain the catalyst temperature below
allowable limits, the excess heat can be utilized for
many purposes such as heating and power generation.
Related to the latter case, this paper presents the math-
ematical modelling and numerical simulation of the

0098-1354/00/$ - see front matter © 2000 Elsevier Science Ltd. All rights reserved.
PII: S0098-1354(00)00618-9
2624 F. Aubé, H. Sapoundjie6 / Computers and Chemical Engineering 24 (2000) 2623–2632
CFRR technology applied to the elimination of lean
methane emissions from underground coal mine venti-
lation air.
Various models were proposed to deal with the dy-
namic behaviour of the CFRR. For example, Eigen-
berger and Nieken (1988), Gosiewski and Sztaba
(1990) used a one-dimensional pseudo-homogeneous
model. Gawdzik and Rakowski (1988), Silveston,
Hudgins, Bogdashev, Vernijakovskaja and Matros
(1990), Thullie and Burghardt (1990), Bhatia (1991)
proposed a one-dimensional heterogeneous model to
account for the gas – solid temperature difference but
neglected the axial diffusion terms. On the other
hand, Sapundzhiev et al. (1990), Young, Hildebrandt
and Glasser (1992), van de Beld and Westerterp
(1994) and Rehacek, Kubicek and Marek (1998) used
a similar model but including the axial diffusion
terms. At a higher complexity level, Sapundzhiev et
al. (1988) developed a two-dimensional heterogeneous
model including radial diffusions and heat and mass
transfer between the phases. Later on, van de Beld
and Westerterp (1996) concluded that a two-dimen-
sional model must be used to improve the model ac-
curacy at low velocities and for small reactor
diameters.
Each model was developed according to the specific
needs of the respective study. Several recommenda-
tions were made to meet minimal accuracy require-
ments to describe satisfactorily the dynamic behaviour
of this type of reactor.
Concerning the mass balance, despite a few excep-
tions using a homogeneous model (Eigenberger &
Nieken, 1988; Gawdzik & Rakowski, 1989;
Gosiewski, 1993), the heterogeneous model is gener-
ally employed where its mass transfer term is repre-
sented by a linear expression (e.g. Sapundzhiev et al.,
1990; Sapundzhiev et al., 1991; Xiao & Yuan, 1994;
Nieken, Kolios & Eigenberger, 1995; van de Beld &
Westerterp, 1996; Salinger & Eigenberger, 1996). The
latter authors also include convective and diffusion
terms for energy (both fluid and solid phases) and
mass balances. Vanden Bussche et al. (1993) specified
that the introduction of mass and energy capacity
terms for both the bulk gas and the gas phase con-
tained inside the pores of the catalyst is also essential
for an adequate modelling.
Concerning specifically the energy balance, Kiselev,
Matros and Chumakova (1988) found that heat dissi-
pation should be taken into account and can be
defined by several means such as axial heat conduc-
tion, heat transfer between the phases or heat transfer
inside the catalyst pellets. Rehacek, Kubicek and
Marek (1992) and Rehacek et al. (1998) found that
complicated regimes existing inside the reactors can
only be found by introducing both axial dispersion in
the gas phase and axial heat conduction into the
model. Gosiewski (1993) specified that the most im-
portant parameter is the gas–solid heat transfer co-
efficient and observed that an overall radial heat
transfer coefficient has to be introduced to account
for radial heat losses. Furthermore, Sapundzhiev et
al. (1988), van de Beld and Westerterp (1996) added
a heat balance term for the reactor wall since its heat
capacity can be significant. They also suggested that,
due to the presence of a thick layer of insulation, a
considerable heat buffer had been created and the
performance of a model might be improved by con-
sidering its heat capacity. However, Sapundzhiev and
his co-workers observed that heat losses could be ne-
glected for reactors with an inner diameter larger
than 1 m. Finally, Khinast, Jeong and Luss (1999)
concluded that, the safest simulation of cases involv-
ing large temperature difference between the solid and
the gas phase is by using a two-dimensional model.
Concerning numerical procedures found in the liter-
ature to solve mathematical models related to CFRR,
lot of work focused on computational time reduction
by using performing numerical methods, by simplify-
ing mathematical models to one dimension, or by
fixing the boundary values in time at the beginning
and end of the half cycle to avoid the need to solve
over the long transient period before the establish-
ment of the stationary conditions (Gupta & Bhatia,
1991; Sapundzhiev et al., 1993). The last technique
has shown satisfactory results to approximate temper-
ature profiles and is useful for basic parametric stud-
ies. However, in the present case and for the majority
of the CFRR industrial applications where the final
design of large units including testing of the reactor
behavior under critical situations depend on calcula-
tions, the accuracy of simulations is the main con-
cern. Therefore, the computer program must be
written in such a way that it allows the detailed de-
sign of reactors including complex geometries, pieces
of equipment, and other aspects like automatic con-
trol of the process, variations of inlet conditions with
time and different modes of heat removal.
The aim of this paper is to demonstrate the impor-
tance of the detailed modeling of heat processes in
the reactor wall and the thermal insulation for the
prediction of the transient behaviour of laboratory-
scale flow reversal reactors. A two-dimensional het-
erogeneous model has been improved by including
complete energy balance for the reactor wall and the
thermal insulation, and mass transfer between bulk
and pore gas phases. The computer program has been
developed with a unique combination of existing nu-
merical and interfacing techniques characterized by
high flexibility and adaptability. A powerful personal
workstation was used to meet the requirements for
speed and accuracy needed for the simulations.
 F. Aubé, H. Sapoundjie6 / Computers and Chemical Engineering 24 (2000) 2623–2632 2625

2. Description of CFRR 3. Mathematical model

A schematic of a typical CFRR process is shown in The general form of the conservation equations writ-
Fig. 1. It consists of a reactor, two pairs of valves and ten in axisymmetrical cylindrical coordinates for the
a heat removal system in the central region separating proposed heterogeneous model is
the catalyst bed into two parts. At both ends of the ( ( 1 (
reactor, an inert (chemically not active) material bed is (r f)+ (r u f)+ (r r6f)
(t (x r (r
present and acts as a thermal capacity. The flow rever-
sal is controlled by the sets of valves. When valves 1 are =
(  
G
(f
+
1 ( 
rG
(f
+ S,

opened and valves 2 are closed, the reaction mixture (x (x r (r (r
flows from left to right (direct flow operation). After a where f=1 for the continuity equation, f=Cs Ts
certain period of time, the position of the valves is (solid phase) and f= Ca Ta (fluid phase) for the energy
reversed in order to reverse the flow direction from conservation equations, and S is the source term (e.g.
right to left (reverse flow operation). Before initiation thermal power generated by chemical reaction for the
of the process the inert and catalyst beds are preheated fluid energy equation).
by an external heat source to an appropriate tempera- This general form has been used to describe the fluid
ture. During normal operation, the system is energy phase and the solid parts of the reactor. On the other
autonomous and the preheating system is switched off. hand, the following assumptions have been made to
In direct flow operation the reaction mixture enters the describe the porous media
reactor from the left side. It heats up, going through the “ heat transfer occurs between the solids (catalyst and
hot left inert bed and enters the catalyst bed at a inert) and the fluid phase,
sufficient temperature to allow the oxidation of meth- “ Mass transfer only takes place between the fluid
ane to occur and further heat the air. The hot gas heats stream and the porous catalyst. The inert bed is not
the right inert bed before exiting to the atmosphere. porous.
During the direct flow operation, the left inert bed, The energy conservation equation for the solid

 
which was initially hot, cools down, while the right porous media can be written as
inert bed, which was relatively cool, heats up. The (T ( (Ts
process switches to reversed flow when the gas mixture (1−o)Cs rs s = (1− o)ks
(t (x (x
temperature entering the left catalyst bed nears a low-
temperature set point, below which catalytic oxidation
+ (1− o)ks
1 ( (Ts
r
  A
− has (Ts − Ta )
is less efficient. The incoming feed gas now enters from r (r (r V
the right, heats up while rising through the right inert N−1
DHj
bed, reacts in the catalyst section and heats the left inert +(1− o) % hj Wsj (r̄j,Ts )
bed before leaving the reactor. Direct flow is re-estab- j = 1 Mj

lished at the end of the reverse flow. The heat generated where o is the bed voidage and N is the number of
in excess by the exothermic chemical reaction is with- reactives. A/V is the solid–fluid surface area per unit of
drawn by the heat removal system in the mid-section of bed volume. The corresponding energy conservation

   
the reactor. equation for the fluid phase is
((raTa ) ( u o ( 6
o Ca + o Ca raTa + Ca rraTa
(t (x o (r o
   
r
( (Ta 1 ( (T A
=o k +o rka a + has (Ts − Ta )
(x a (x r (r (r V
where the velocity field is determined with the two-di-
mensional form of the momentum balance, using the
Brinkman’s equation (Brinkman, 1947) for porous me-
dia. For each one of the N reactives (indexed j) present
in the fluid stream, the continuity equations can be

   
written as:
(r ( u 1 ( (r A
o j+o rj = o rD j + bj (r̄j − rj )
(t (x o r (r (r V
where bj is the mass transfer coefficient (bj = 0 in the
non-porous inert beds). The radial convection term has
been neglected. Moreover, for each one of theNreac-
tives located inside the volume of the porous catalyst,
Fig. 1. Schematic of a typical CFRR process. the continuity equation is
2626 F. Aubé, H. Sapoundjie6 / Computers and Chemical Engineering 24 (2000) 2623–2632
Fig. 2. Grid representation of a simplified reactor configuration.
Fig. 3. Example of element type allocation in a two-dimensional grid.
(r̄j A
op (1 −o) = − bj (r̄j −rj ) +(1 −o) hj Wsj (r̄j, Ts )
(t V the two heat exchanger nodes are represented by the
where op is the porosity of the catalyst, hj is its effective-
ness and Wsj is the rate of reaction in kinetic regime.
The initial and boundary conditions depend on the
design specifications.
4. Numerical procedure
The two-dimensional heterogeneous model was
solved in axisymmetrical cylindrical coordinates by iter-
ation procedures using an existing finite-volume method
proposed by Baliga (1978). A line-by-line Tri-Diagonal
Matrix Algorithm (Thomas, 1949), with alternating
sweeps in the axial and radial directions was imple-
mented to solve the algebraic discretized equations. A
hybrid scheme (Spalding, 1972) is used to solve the
convection and diffusion terms. In order to solve the
unsteady-state terms, a fully implicit scheme is used.
Moreover, a non-uniform grid spacing is used to take
account accurately of the low thickness of the reactor
wall without imposing a large number of control vol-
umes to the solver. It adds to the numerical procedure
more flexibility and saves calculation time. The result-
ing solver is characterized by fast convergence to the
final solutions and high stability.
The originality of the numerical procedure is the high
adaptability of the control volume definition and the
related interfacing technique, allowing high flexibility
for the design of catalytic fixed bed reactors. Any
characteristic or equation inherent in each of the differ-
ent reactor elements (e.g. catalyst, inert material, fluid
phase, reactor wall, thermal insulation, internal or ex-
ternal heat exchange) was integrated into the computer
program from the outset to avoid the rewriting of the
complete set of equations for each new case studied.
Any internal node of the two-dimensional grid repre-
senting the reactor can be specified automatically by the
designer to be included in one of the different reactor
parts. This is illustrated with a grid representation of
the reactor shown in Fig. 2. In this case, for example,
an internal heat exchanger is represented with two
nodes in the mid-section. A total of 100 nodes in the
axial direction and 41 in the radial direction (20 inside
the reactor, 1 for the reactor wall and 20 for the
thermal insulation) were used in this work to simulate
the reactors.
To transform the reactor configuration specifications
into information readable by the computer program, a
character is assigned to each node type in the two-di-
mensional grid and the resulting array of characters is
written to an input file. For example, the reactor
configuration shown in Fig. 2 has been transformed
into the array illustrated in Fig. 3. For this case the
central axis nodes are represented by the letter (c) and
letter (H). At the beginning of the calculations, just
after the input file is read, the specific physical and
thermodynamic properties of each control volume are
defined automatically and are then updated before each
new iteration if they are temperature-dependent. For
the mass and energy balances between the nodes, an
integrated procedure allows the program to manage
automatically any realistic type of control volume inter-
face (e.g. between the reactor wall and a packed bed,
between the catalyst and inert beds or between any
solid part and external boundary condition as known
ambient temperature, imposed heat flux or external film
resistance).
Additional available features had been added to in-
crease the software capabilities
“ graphical comparisons between experimental and
calculation results at each time step,
“ three different flow reversal modes to control the
cycle duration: (1) time fixed cycle duration; (2) cycle
time determined by the experimental files; (3) flow
reversals based on temperature criteria (automatic
control).
5. Kinetic data and transfer relations used in the model
A non-noble metal-based catalyst was used for meth-
ane combustion. The kinetic characteristics for the
methane oxidation were obtained using a differential
 F. Aubé, H. Sapoundjie6 / Computers and Chemical Engineering 24 (2000) 2623–2632 2627

reactor avoiding concentration and temperature gradi-
ents, and under the assumption that the reaction is of
first order according to methane (oxygen is in excess)
Ws = − rs = − Kr̄ = − Koe − Ea /RTsr̄
where Ko is the reaction rate coefficient and Ea is the
activation energy.
Physical and thermodynamic properties of the gas
mixture flowing through the CFRR, like m, ka, D and
ra vary over a wide range of temperature. This temper-
ature dependence has been taken into account by using
polynomial best curve fits based on handbook data.
For the solid phase, the properties are assumed con-
stant and are shown in Table 1. Moreover, the parame-
ter values needed to solve the model and the transfer
relations, used in the calculations for the packed beds,
are shown in Table 1 for each one of the reactors.
6. Experimental setup
The model and simulation software proposed in this
paper had been validated for the design of CFRR
applied to the oxidation of lean methane emissions. The
simulation results were compared with the experimental
data from a 200 mm I.D. and a 500 mm I.D. laboratory
scale reactors operating with two different configura-
tions for heat removal (mid-section hot air withdrawal
or internal heat exchange).
For each reactor, a set of thermocouples was in-
stalled along the central axis for temperature registra-
tion. Methane conversion was measured by gas
chromatography, analysing the methane concentration
in the feed gas stream, the mid-section gas withdrawal
and the gas at the outlet of the reactor. Methane is
almost totally converted by the chemical reaction. The
outlet gas and the extracted hot air in the mid-section
contain no nitrogen oxides due to the low oxidation
temperature (below 800°C).
7. Model and code validation
The process of code validation for transient simula-
tions implies comparisons with data collected over peri-
ods of time and for a specific range of operations.
Simulations were compared to results obtained from
experiments conducted on the lab-scale reactors, often
lasting for over 8 h and during which operating condi-
tions were varied several times.
During the experiments, the temperature was regis-
tered mainly in the central axis of the reactor. A radial
temperature registration was done at only two other
places in the reactor, both located in the catalyst beds.
No temperature registration was done for the reactor
shell and in the thermal insulation. Moreover, since the
reactor and its thermal insulation both have high heat
capacities, any significant difference between the esti-
mated and real temperature values in these sections can
lead to large discrepancies between the simulation re-
sults and the experimental data. For these reasons, a
calculation procedure was developed to estimate the
two-dimensional ‘initial’ temperature field values in the
reactor beds and shell, as well as in the thermal insula-
tion. This procedure is as follows: temperature data
taken at the central axis of the reactor are selected for
the beginning of the calculations. The temperature data
at that time (set to t= 0) are used to calculate the
‘initial’ temperature profile at any node of the grid
along the central axis of the reactor. For simplicity, the

Table 1
List of input parameters for the model simulations shown in this paper

200 mm I.D. reactor 500 mm I.D. reactor

Inert material bed Catalyst bed Reactor wall and insulation Inert material bed Catalyst bed Reactor wall and insulation

oi = 0.4 oc = 0.51 J oc =0.51 J

Cpw = 460 oi =0.725 Cpw =460
opi =0.0 opc =0.40 kg JK opc =0.40 kg JK
Cpt = 1450 opi =0.0 Cpt =1450
m 2 kg K kg K
(A/V)i = 600 (A/V)c W m2 (A/V)c W
m3 m2 kw = 14.3 (A/V)i =246 m2 kw =14.3
mK m3 mK
=280 3 =280
J m m3
Cpi = 925 J kg J J kg
kg K Cpc =574 rw = 7800 Cpi =1050 Cpc =574 rw =7800
W kg K m3 kg K kg K m3
ki = 1.10 W kg W W kg
mK kc =8.0 rt = 96.3 3 ki =0.35 kc =8.0 rt =96.3 3
kg mK m mkgK mK m
ri = 2400 3 kg dw = 3.2 mm ri =2620 3 kg dw =6.4 mm
m rc =2400 3 m rc =2400 3

   
m m
di = 4.5 mm cd =7.2 mm dt = 0.3 m di =11.7 mm d =7.2 mm
c dt =0.4 m
Re 0.6 Re 0.6
Heat and mass transfer relations: Wakao & Kaguei (1982), Nu =2.0+1.1 Pr1/3 and Sh=2.0+1.1 Sc 1/3

Kinetics data: K = 0.135×105 s−1, E = 54 400 J/mol, DH=−815 000 J/mol o

o a
o
2628 F. Aubé, H. Sapoundjie6 / Computers and Chemical Engineering 24 (2000) 2623–2632
Fig. 4. 200 mm ID reactor: 6 =0.15 m/s, 0.85% v/v of methane;
comparison between the proposed model, adiabatic conditions and
heat buffer term added (t= 2128 s).
Fig. 5. 200 mm I.D. reactor: 6 =0.15 m/s, 0.85% v/v of methane;
beginning to end of the first half-cycle (t= 0, 1268 and 2128 s).
radial temperature gradients are assumed to be negligi-
ble inside the reactor beds for the initial calculations.
Thus, the axial temperature profile is kept unchanged in
the radial direction inside the reactor beds. Based on
this profile, a heat conduction calculation (in two-di-
mensional axisymmetrical cylindrical coordinates) is
performed to estimate the temperature field inside the
reactor shell and the thermal insulation. In order to
take account of the heat accumulated during the pre-
heating phase and the heating of these reactor parts
before the beginning of the calculations, an equivalent
heating time duration is estimated and used in the
procedure.
This calculation procedure therefore enables us to
determine realistic ‘initial’ conditions in order to start
comparisons between the experimental and simulation
results from any point in time during an experiment. It
is useful in saving time and avoiding accumulated dis-
crepancies at the beginning of a comparison.
8. Results and discussion
The first case of interest was to investigate the valid-
ity of the model for small diameter reactors at low gas
velocity. As was mentioned previously by van de Beld
and Westerterp (1996), the one-dimensional model fails
under these conditions and a two-dimensional model
should improve the accuracy of simulations. Fig. 4
presents results obtained with the 200 mm diameter
reactor in the case of an operation without heat re-
moval. The reactor was initially preheated with an
electrical heater located in the right catalytic section of
the reactor. The maximum temperature in this zone
reached around 700°C. The experimental temperature
distribution along the reactor length at the end of the
preheating period was taken as the base in calculating
the ‘initial’ conditions (set to t=0), for the beginning
of the calculations. The resulting calculated profile is
shown with a short dashed line on the figure. At that
moment the electrical heater was switched off, and the
following operating conditions were established: meth-
ane concentration of 0.85% v/v, a feed gas velocity of
0.15 m/s and a temperature of 25°C. The flow direction
was from right to left. The experimental results ob-
tained at the end of the first half-cycle (t= 2128 s) are
shown in hollow circles. The predictions of the pro-
posed model at this moment are compared to those
obtained at adiabatic conditions (no heat exchange with
the reactor shell and the insulation) and also at non-
adiabatic conditions by replacing the two-dimensional
thermal insulation with a heat buffer term taking into
account the heat exchange through the reactor shell and
the insulation as was done by Sapundzhiev et al. (1988)
and van de Beld and Westerterp (1996). The simulation
results obtained by the different models are shown in
Fig. 4. The almost adiabatic behaviour of the curves in
the central zone of the reactor is due to the fact that in
this section the fluid flows at high velocity in a pipe
separating the two reactor sections and that no heat
removal takes place. The adiabatic model shows higher
maximal temperature and a large high temperature
zone in the reactor compared to the experimental re-
sults. When the heat buffer term is added, the predic-
tions are slightly improved compared to the adiabatic
model but, nevertheless, a significant departure from
the experimental results is still observed. The best per-
formance is observed with the proposed model. It can
be explained by the fact that the effect of axial heat
transfer by conduction inside the thermal insulation
and into the reactor shell becomes relatively significant
when the temperature front moves slowly. The axial
heat conduction in the reactor shell and insulation leads
to a more uniform axial distribution of the thermal
energy and, consequently, to an increase of the radial
heat tranfer with the reactor core and a flattening of the
temperature curves in these regions.
Fig. 5 shows comparisons between the experiments
and the predictions of the proposed model for the same
conditions as in Fig. 4 but at different intermediate
times during the first half-cycle. The dynamic behaviour
of the reactor is very well predicted for these
conditions.
 F. Aubé, H. Sapoundjie6 / Computers and Chemical Engineering 24 (2000) 2623–2632
Typical temperature profiles along the reactor length
obtained with the 200 mm reactor are shown in Fig. 6.
The superficial velocity of the gas mixture is 0.65 m/s
and the methane concentration is 0.45% v/v for Fig.
6a – c. The data were recorded at the end of a 10 h
continuous experiment. Fig. 6a shows the results at the
beginning and end of the first simulated half-cycle. The
estimated initial temperature profile along the central
axis is shown as a dashed line. For the subsequent
Fig. 6. 200 mm I.D. reactor. Experimental (circles) and simulation
(lines) temperature profiles along the reactor length. 6= 0.65 m/s,
0.45% v/v of methane. (a) First half-cycle (t= 0–320 s); (b) Second
half-cycle (t= 320 – 764 s); (c) Sixth half-cycle (t= 2034–2454 s); (d)
Prolongation of the sixth half-cycle during which no chemical reac-
tion takes place in the reactor (t= 2454–3384 s).
2629
Fig. 7. 500 mm I.D. reactor. 6 =0.65 m/s, 0.3% v/v of methane, hot
air withdrawal. (a) First half-cycle (t =0 – 200 s); (b) Second half-cycle
(t =200 – 400 s); (c) After 15 cyces (t = 6950 s).
results shown in Fig. 6b–c, the beginning of the half-
cycle is shown as dashed line and with solid circles and
its end is shown with solid lines and hollow circles. It
can be seen that the experimental results are in good
agreement with the predictions of the proposed model.
To evaluate the prediction accuracy of the model
when no chemical reaction takes place in the reactor,
the methane feed stream was closed near the end of the
experiment and only cold air at 25°C fed the reactor.
The results are shown in Fig. 6d for the time interval
covering from 2454 to 3384 s with no flow reversal. The
simulation curves are in good agreement with the ex-
perimental results.
The proposed model was tested with experimental
data obtained from the 500 mm diameter reactor oper-
ating with heat removal in its mid-section. Two heat
removal configurations were utilized: hot air with-
drawal and internal heat removal with an air–oil heat
exchanger.
Time evolution of temperature profiles and distribu-
tion along the reactor length are shown in Fig. 7a–c for
the hot air withdrawal configuration. The experimental
2630 F. Aubé, H. Sapoundjie6 / Computers and Chemical Engineering 24 (2000) 2623–2632
conditions are as follows: feed air superficial velocity of
0.65 m/s, an inlet temperature of 20°C, and a methane
concentration of 0.3% v/v.
Fig. 7a shows the temperature evolution along the
reactor length during the first half-cycle. The flow direc-
tion is from left to right. The initial temperature profile
is shown as a dashed line. The circles and the solid line
represent the temperature distribution and the predicted
profile along the reactor length at the end of the first
half-cycle (200 s). The flow direction is then reversed.
The evolution of the heat front during the second
half-cycle is shown in Fig. 7b (200 – 400 s) as the
difference between the solid and hollow circles. Hot air
extraction in the mid-section implies adjusting opera-
tions conditions in order to achieve complete methane
combustion in the upstream catalyst bed and thus
avoiding the presence of methane in the withdrawn air.
This restriction imposes shorter cycle duration and
leads to the formation of two peaks on the axial
temperature profile, both located at the catalyst beds.
Fig. 7(c) shows temperature profiles after 15 cycles
(6950 s). For this case, the reactor operates in stable
regime and 52% of the heat generated by the chemical
reaction is removed by air withdrawal in the central
region of the reactor. The flow rate of the extracted hot
air was kept constant, with an average temperature
around 700°C.
Good agreement between simulations and experi-
ments is also observed with the internal heat removal
scenario. Typical results are shown in Fig. 8a – d for a
fluid superficial velocity of 0.65 m/s and a methane
Fig. 8. 500 mm I.D. reactor. 6 = 0.65 m/s, 0.35% v/v of methane, internal heat remo6al. (a) First half-cycle (t =0 – 200 s); (b) Second half-cycle
(t=200 – 430 s); (c) After 9 cyces (t = 4100 s); (d) After 24 cycles (t= 10800 s).
concentration of 0.35% v/v. The heat removed through
the air–oil heat exchanger is 50% of the total amount
of heat generated by the chemical reaction. The oil
temperature has been kept to a temperature below
350°C in order to prevent boiling. In comparison with
the hot air withdrawal cases, a significant temperature
decrease in the mid-section and the catalyst beds was
observed. It was caused by the air cooling. The in-
creased temperature gradients greatly influence the rate
constant of the reaction, which is an exponential func-
tion of the temperature and makes reactor behaviour
more sensitive to process variables. It may result in
higher discrepancies between predicted and experimen-
tal results as seen on these figures.
Results obtained at the end of the first and the
second half-cycle are shown in Fig. 8a and Fig. 8b,
respectively. Moreover, results obtained after 9 and 24
cycles (4100 and 10800 s respectively) are shown in Fig.
8c–d. There is a progressive axial temperature drift
between predictions and data with time, particularly at
the mid-section. Despite the fact that higher discrepan-
cies were expected, good agreement for short and long-
term simulations was nevertheless found under these
conditions and for the whole investigated range of
operation.
9. Conclusions
The combination of a two-dimensional heteroge-
neous model and a special numerical procedure allow-
 F. Aubé, H. Sapoundjie6 / Computers and Chemical Engineering 24 (2000) 2623–2632 2631

ing a detailed description of a CFRR was developed c catalyst

and validated on two different laboratory-scale reactors i inert material
for the elimination of lean methane emissions. The j reactive number
experimental results were predicted accurately for all s solid properties
the cases studied, including hot air withdrawal and t thermal insulation
internal heat removal at the mid-section of the w reactor wall
reactors. It was also shown that detailed modelling of Dimensionless numbers
the thermal insulation is particularly necessary Nu Nusselt number
for small diameter reactor and low air velocity condi- Pr Prandtl number
tions. The proposed simulation tool is also appropriate Re Reynolds number
for the design and optimization of industrial CFRR, Sc Schmidt number
including study of reactor behaviour during startup, Sh Sherwood number
analysis and setting of the flow reversal automatic
control and excess heat removal systems for heat uti-
lization.

10. Notation References

A/V fluid–solid surface area per unit of volume Baliga, B. R. (1978). A control-volume based finite element method
(m2/m3) for convective heat and mass transfer. Ph.D. Thesis, University of
C specific heat (J/kg K) Minnesota, Minneapolis, USA.
Bhatia, S. K. (1991). Analysis of catalytic reactor operation with
d particle equivalent diameter (m)
periodic flow reversal. Chem. Eng. Sci., 46 (1), 361 – 367.
D diffusion coefficient (m2/s) Blanks, R. F., Wittrig, T. S., & Peterson, D. A. (1990). Bidirectional
Eaj activation energy (J/mol) adiabatic synthesis gas generator. Chem. Eng. Sci., 45 (8), 2407–
h fluid–solid heat transfer coefficient (W/m2 K) 2413.
DH heat of reaction (J/mol) Bobrova, L. N., Slavinskaya, E. M., Noskov, A. S., & Matros, Yu.
k thermal conductivity (W/m K) Sh. (1988). Unsteady-state performance of NOx catalytic reduc-
tion by NH3. React. Kinet. Catal. Lett., 37, 267 – 272.
Koj reaction rate coefficient (s−1) Boreskov, G. K., Bunimovich, G. A., Matros, Yu. Sh., & Ivanov, A.
M molecular weight (kg/mol) A. (1982). Catalytic processes under non-steady-state conditions;
N number of reactives I. Switching the direction for the feed of the reaction mixture to
P partial pressure (bar) the catalytic bed. Experimental Results. Kinet. Catal., 23, 402–
406.
r radial position (m)
Brinkman, H. C. (1947). On the permeability of media consisting of
rsj reaction rate (kg/m3s) closely packed porous particles. Appl. Sci. Res., A1, 81 –86.
R gas constant (J/mol K) Bunimovich, G. A., Strots, V. O., & Goldman, O. V. (1990). Theory
S source term and industrial application of SO2 oxidation reverse-process for
t time (s) sulphuric acid production. In Yu. Sh. Matros, Unsteady-state
T temperature (K) processes in catalysis (pp. 7 – 24). Utrecht-Tokyo: VNU Science
Press.
u axial superficial velocity (m/s) Bunimovich, G. A., Vernikovskaya, N. V., Strots, V. O., Balzhini-
6 radial superficial velocity (m/s) maev, B. S., & Matros, Yu. Sh. (1995). SO2 oxidation in a
Wsj reaction rate (kg/m3s) reverse-flow reactor: Influence of a vanadium catalyst dynamic
x axial position (m) properties. Chem. Eng. Sci., 50 (4), 565 – 580.
Chumachenko, V. A. & Matros, Yu. Sh., (1990). Catalytic purifica-
Greek letters
tion of gases from organic substances under unsteady-state condi-
b mass transfer coefficient (m/s) tions; Results and predictions, Proceedings of the International
G diffusion parameter Conference, Novosibirsk, U.S.S.R., 1990; VSP BV: Utrecht, The
d reactor wall or insulation thickness (mm) Netherlands, 605 – 612.
o packed bed voidage Eigenberger, G., & Nieken, U. (1988). Catalytic combustion with
periodic flow reversal. Chem. Eng. Sci., 43 (8), 2109 – 2115.
op particle porosity
Gawdzik, A., & Rakowski, L. (1988). Dynamic properties of the
f variable adiabatic tubular reactor with switch flow. Chem. Eng. Sci.,
h effectiveness factor of catalyst 43 (11), 3023 – 3030.
m fluid viscosity (kg/m s) Gawdzik, A., & Rakowski, L. (1989). The methods of analysis of the
r density (kg/m3) dynamic properties of the adiabatic tubular reactor with switch
flow. Comput. Chem. Eng., 13 (10), 1165 – 1173.
r̄ fluid density inside the catalyst particles (kg/
Gosiewski, K., & Sztaba, R. (1990). Simplified design of reverse flow
m3) nonstationary reactor for low reactant concentration. In Yu. Sh.
Subscripts and superscripts Matros, Unsteady state processes in catalysis (pp. 629–635).
a fluid mixture properties Utrecht-Tokyo: VNU Science Press.
2632 F. Aubé, H. Sapoundjie6 / Computers and Chemical Engineering 24 (2000) 2623–2632
Gosiewski, K. (1993). Dynamic modelling of industrial SO2 oxidation
reactors. Part II. Model of a reverse-flow reactor. Chem. Eng.
Process., 32, 233 – 244.
Gupta, V. K., & Bhatia, S. K. (1991). Solution of cyclic profiles in
catalytic reactor operation with periodic flow reversal. Comput.
Chem. Eng., 15 (4), 229–237.
Jirat, J., Stepanek, F., Kubicek, M., & Marek, M. (1999). Nonsta-
tionary operation of a system of catalytic monolithic reactors for
selective NOx reduction. Chem. Eng. Sci., 54, 2609–2618.
Khinast, J., Jeong, Y. O, & Luss, D. (1999). Deoendance of cooled
reverse-flow reactor dynamics on reactor model. American Insti-
tute of Chemical Engineering Journal, 45, 2.
Kiselev, O. V., Matros, Yu. Sh., & Chumakova, N. A. (1988).
Phenomena of heat front propagation in fixed catalyst bed. In Yu.
Sh. Matros, Propagation of heat wa6es in heterogeneous media.
U.S.S.R: Novosibirsk.
Kolios, G., & Eigenberger, G. (1999). Styrene synthesis in a reverse-
flow reactor. Chem. Eng. Sci., 54, 2637–2646.
Kulkarni, M. S., & Dudukovic, M. P. (1996). A bidirectional fixed-
bed reactor for coupling of exothermic and endothermic reac-
tions. American Institute of Chemical Engineering Journal, 42 (10),
2897.
Kulkarni, M. S., & Dudukovic, M. P. (1997). Periodic operation of
asymmetric bidirectional fixed-bed reactors: energy efficiency.
Chem. Eng. Sci., 52 (11), 1777–1788.
Kulkarni, M. S., & Dudukovic, M. P. (1998). Periodic operation of
asymmetric bidirectional fixed-bed reactors with temperature limi-
tations. Ind. Eng. Chem. Res., 37, 770–781.
Matros, Yu. Sh. (1989). Catalytic processes under unsteady state
conditions. Amsterdam: Elsevier.
Matros, Yu. Sh., Noskov, A. S., Chumachenko, V. A., & Goldman,
O. V. (1988). Theory and application of unsteady-state catalytic
detoxication of effluent gases from dioxide, nitrogen oxides and
organic compounds. Chem. Eng. Sci., 43, 2061–2066.
Matros, Yu. Sh., Noskov, A. S., & Chumachenko, V. A. (1993).
Progress in reverse process application to catalytic incineration
problems. Chem. Eng. Prog., 32, 89–98.
Matros, Yu. Sh., Bunimovich, G. A., Strots, V. O., & Mirosh, E. A.
(1999). Reversed flow converter for emission control after auto-
motive engines. Chem. Eng. Sci., 54, 2889–2898.
Neophytides, S. G., & Froment, G. F. (1992). A bench scale study of
reversed flow methanol synthesis. Ind. Eng. Chem. Res., 31,
1583 – 1589.
Nieken, U., Kolios, G., & Eigenberger, G. (1995). Limiting cases and
approximate solutions for fixed-bed reactors with periodic flow
reversal. American Institute of Chemical Engineering Journal, 41,
8.
Rehacek, J., Kubicek, M., & Marek, M. (1998). Periodic, quasiperi-
odic and chaotic spatiotemporal patterns in a tubular catalytic
reactor with periodic flow reversal. Comput. Chem. Eng., 22 (1 – 2),
283 – 297.
Rehacek, J., Kubicek, M., & Marek, M. (1992). Modelling of a
tubular catalytic reactor with flow reversal. Chem. Eng. Sci., 47,
2897 – 2902.
Salinger, A. G., & Eigenberger, G. (1996). The direct calculation of
periodic states of the reverse flow reactor — I. Methodology and
propane combustion results. Chem. Eng. Sci., 51 (21), 4903–
4913.
Sapundzhiev, C., Bunimovich, G. A., Drobishevich, V. I., Grozev,
G., Yausheva, L. V., Matros, Yu. Sh., & Elenkov, D. (1988).
Effect of heat loss on operation of reactors with periodic reverse
gas feed. Theoretical Fundamentals Chemical Engineering, 22 (3),
349 – 355.
Sapundzhiev, C., Grozev, G., & Elenkov, D. (1990). Influence of
geometric and thermophysical properties of reaction layer on
sulphur dioxide oxidation in transient conditions. Chem. Eng.
Technol., 13, 131 – 135.
Sapundzhiev, C., Grosev, G., & Elenkov, D. (1991). Non-steady-state
catalytic decontamination of waste gases. Chem. Eng. Technol.,
14, 209 – 212.
Sapundzhiev, C., Chaouki, J., Guy, C., & Klvana, D. (1993). Cata-
lytic combustion of natural gas in a fixed bed reactor with flow
reversal. Chem. Eng. Commun., 125, 171.
Silveston, P. L., Hudgins, R. R., Bogdashev, S., Vernijakovskaja, S.,
& Matros, Yu. Sh. (1990). In Yu. Sh. Matros, Unsteady State
Processes in Catalysis (pp. 557 – 570). Utrecht-Tokyo: VNU Sci-
ence Press.
Spalding, D. B. (1972). A novel finite-difference formulation for
differential expressions involving both first and second deriva-
tives. Int. J. Num. Methods Eng., 4, 551.
Thomas, L. H. (1949). Elliptic problems in linear difference equations
o6er a network. New-York: Watson Sci. Comp. Lab. Rep., Co-
lumbia University.
Thullie, J., & Burghardt, A. (1990). Application of the flow reversal
reactor to the methanol synthesis. In Yu. Sh. Matros, Unsteady
State Processes in Catalysis (pp. 687 – 692). Utrecht-Tokyo: VNU
Science Press.
van de Beld, B., & Westerterp, K. R. (1994). Air purification by
catalytic oxidation in a reactor with periodic flow reversal. Chem.
Eng. Technol., 17, 217 – 226.
van de Beld, L., & Westerterp, K. R. (1996). Air purification in a
reverse-flow reactor: Model simulations vs. experiments. American
Institute of Chemical Engineering Journal, 42, 4.
Vanden Bussche, K. M., Neophytides, S. N., Zolotarski, I. A., &
Froment, G. F. (1993). Modelling and simulation of the reversed
flow operation of a fixed-bed reactor for methanol synthesis.
Chem. Eng. Sci., 48, 3335 – 3345.
Wakao, N., & Kaguei, S. (1982). Heat and mass transfer in packed
beds. Topics in chemical engineering. New-York: Gordon and
Breach Science.
Xiao, W. D., & Yuan, W. K. (1994). Modelling and simulation for
adiabatic fixed-bed reactor with flow reversal. Chem. Eng. Sci.,
49 (21), 3631 – 3641.
Young, B., Hildebrandt, D., & Glasser, D. (1992). Chem. Eng. Sci.,
47, 1825 – 1837.