Chemical Engineering Science. Vol. 48. No. 19, pp. 3335-3345, 1993. OOO-2509/93 S4.00 + 0.

00
Printed in Grspt Britain. 0 1993 Psrgnmon Press Ltd

MODELLING AND SIMULATION OF THE REVERSED FLOW

OPERATION OF A FIXED-BED REACTOR FOR METHANOL
SYNTHESIS

K. M. VANDEN BUSSCHE, S. N. NEOPHYTIDES, I. A. ZOLOTARSKII and

G. F. FROMENT’
Laboratorium voor Petrochemische Techniek, Universiteit Gent, Krijgslaan 281, B9000 Ghent. Belgium

(Received 22 March 1993; accepted for publication 21 April 1993)

Abstract-The proposed simulation of the reversed flow operation of a fixed-bed reactor for methanol
synthesis is based upon a heterogeneous one-dimensional model. Different levels of transient modelling are.
compared. By using the most sophisticated model, the influence of a number of operating parameters is
assessed. Considerable insight into the reactor behaviour is obtained.

1. INTRODUCTION tions A #B and A + B. Eigenberger and Nieken

Transient operation of a chemical may yield
reactor (1988) also used a pseudo-homogeneous one-dimen-
conversions and selectivities unattainable under sional model, but accounted for axial dispersion, in
steady-state conditions (Renken, 1972; Schldlich et the modelling of a reversed flow reactor for catalytic
al., 1983; Grabmuller et al., 1985; Zolotarski et al., combustion. In their work the continuity equations
1989: Nam et al., 1990: Matros, 1990: Westerterp, were again assumed to be in pseudo-steady-state.
1992; Zagoruiko et al., 1992). Thullie and Burghardt (1990) used a pseudo-homo-
Among the various modes of transient operation, geneous one-dimensional model, in which both the
periodic operation combines the benefits of a perma- energy and the continuity equations contained a tran-
nent unsteady-state regime and a constant time-aver- sient term, for their simulation of a reversed flow
aged production (Silveston, 1987). Examples of reactor for methanol synthesis. Matros (1990) recom-
periodic operation are the periodic regeneration of mended a completely transient one-dimensional
catalysts, pulsing columns and periodic changes in heterogeneous model, accounting for internal and ex-
feed concentration or flow direction of chemical reac- ternal diffusional limitations. Sapundzhiev et al.
tors. In the latter, usually referred to as reversed flow (1992) finally developed a two-dimensional fully tran-
operation (Matros, 1989), the reaction mainly takes sient model, but mainly focused on the heat losses
place in a relatively narrow zone, wandering back and from the bed.
forth across the reactor. The moving heat wave allows In the present paper, a one-dimensional heterogen-
for a combination of regenerative heat exchange and eous model is applied to an investigation of methanol
chemical reaction. The features and possibilities of synthesis. Various levels of transient modelling are
this type of reactor operation have been extensively compared. The most sophisticated model, accounting
discussed in reviews by Froment (1991) and Matros for mass and energy capacities in both solid and gas
(1990) and in an analysis of chemical process dy- phase, is used to assess the influence of a number of
namics by Marquardt (1990). The technique has been operating parameters.
applied for SO, production from metallurgic off gas 2. MODELLING OF THE REVERSED FLOW REACTOR
containing SO, (Boreskov et al., 1983; Matros, 1989),
2.1. Equations
and has shown great potential in catalytic purification
The adequacy of a model in describing the behavi-
of waste gases (Eigenbcrger and Nieken, 1988;
our of the reversed flow reactor is to a large extent
Balmer, 1988; Matros, 1989), in sulphur production
determined by the number of equations containing
by the Claus process (Zagoruiko et al., 1989) and in
a transient term. However, the stiffness of the system
methanol synthesis (Matros, 1989; Thullie and
of differential equations increases considerably with
Burghardt, 1990; Neophytides and Froment, 1992;
the number of capacitive terms, due to the important
Vanden Bussche et al., 1992a, b).
differences in solid- and gas-phase time constants. In
Until now, the modelling of the reversed flow reac-
this work, three levels of modelling originating from
tor behaviour has been based upon simplified models.
a heterogeneous one-dimensional model accounting
Gosiewski and Sztaba (1990) proposed a pseudo-ho-
for both intraparticle and interfacial gradients
mogeneous one-dimensional model, but left out the
(Froment and Bischof. 1990) are compared.
transient term in the continuity equations for the
In model I, only the energy equation of the solid
various species. Gawdzik and Rakowski (1988) used
phase is transient. The gas phase is considered to be in
a similar model for a sensitivity analysis for the reac-
the pseudo-steady-state. This assumption, common in
the modelling literature, is generally speaking justified
‘Author to whom correspondence should be addressed. only for low-pressure processes.

3335
3336 K. M. VANDEN BUSSCHE et al.

In model II, a transient term is added to the gas- Initial conditions:

phase equations, thus assigning both heat and mass
YjCz, 0) = YjO
capacities to the bulk gas phase.
In model III, finally, a time derivative is added to T(z, 0) = To (2) (8)
the solid phase continuity equations, introducing
Ts(z, 0) = L-J(z)
a mass capacity for the gas inside the catalyst pores.
The equations of this most sophisticated model III are N,({, z, 0) = 0.
as follows:
At flow reversal,
Continuity equation for component j in the gas phase:
Yj(Z, t-) = Yj(L - z, t’)

rc ayi = _ F a-rj Pl(Z, t-j = P,(L - 2, t’)

Ix + G7j(Y$s - Yj)
* at
F,(z, t-) = F,(L - z, f’)
j= 1,. . . , N - 1 (1) (9)
T(z, t-) = T(L - 2, t’)
N
- %Yj C kgi(Yb - Yi). T,(z, t-) = T,(L - z, t’)
i=,
Nj(<, z, t-) = Nj(& L - z, t’).
Evolution of the total molar concentration:
All three models are combined with the kinetic equa-
82=- 2 +a, i=5 k,i(y;s I
- yJ. (2) tions proposed by Graaf et al. (1988). Non-ideality
the gas phase is accounted for using the Soave-
of

Energy equation for the gas phase in an adiabatic Redlich-Kwong equation of state (Soave, 1972), as
reactor: recommended by Graaf (1986).

2.2. Numerical solution

c&c,: = - F,c,g + a,h/ (T, - T). (3)
The time integration of the energy equation of the
solid phase was performed by means of the trap-
Prsssure drop equation (Ergun, 1952):

t1
ezoidal rule, while a spatial grid along the reactor
150(1 - E)fi + ~.75M F . length is provided by a global spline collocation
m method (Carey and Finlayson, 1975; Villadsen and
d,
Michelsen, 1978). The spline points were relocated in
(4) every time step, using a redivision routine, described
Continuity equation for component j in an isothermal elsewhere (Vanden Bussche et al., 1993). The solution
spherical particle of the solid phase of the gas-phase equations by means of quasi lin-
earisation and the trapezoidal rule was superimposed
ac, l a(tZNj) on this integration, using the same At and AZ. For
E”X =j5 ay
- rips j = 1, . . , N - 1. (5)
every space step, the continuity equations for the
components inside the pellet were solved using ortho-
Energy equation for the solid phase in an adiabatic
gonal collocation along the c-axis and the trapezoidal
reactor
rule along the time axis.

=x. z
PBC,~ + h,a”(T-- 7’J 3. SIMULATION RESULTS
3.1. Physical interpretation of a typical result
Figure 1 shows the catalyst temperature profiles
(6)
along the reactor at various times during a semi-cycle,
obtained through model III. The profiles in Fig, 1, as
Boundary conditions: all the other profiles shown below, were taken from
a_/iilly converged semi-cycle, i.e. when a permanent
Yj(O, t) = Yy
cyclic regime has been established. The conditions
P,(O, t) = PP used for this simulation are given in Table 1.
For illustrative purposes, a diameter of 0.1 m was
F,(O, t) = Ff
used for a flow of 0.256 mol/s, thus obtaining a super-
T-(0, t) = TO (7) ficial flow rate, representative for an industrial
reversed flow reactor used for the conversion of the
Z(z, t) = 0, z=O,L purge gas of a traditional methanol synthesis loop
(Vanden Bussche et al., 1992a).
Nj(< = 0, Z, t) = 0 For this base case, a reactor length of 0.5 m was
chosen, although further on more important bed
Nj(S = $7 Z, t) = k,j(yj, - yj) depths are also considered. The composition of the
feed, finally, is fairly typical for a purge stream of the
 Simulation of reversed flow operation 3337

Flow- where

E
H, = - ’ 5 <(- AH,i)
T°CPSUi i =,
i-
and
i

%_

It follows from eq. (6’) that the shape and the direction
of the heat wave shown in Fig. 1 are determined by the
interaction of three phenomena: the heat generated by
Fig. 1. Catalyst-bed temperature profiles along the reactor the chemical reactions (H,), the convective heat trans-
at various times during one semi-cycle, as predicted by fer (H,) and the conductive heat transfer (Pe,,,J.
model III. Conditions shown in Table 1. At the upstream end of the wave, the feed gas is
heated and progressively cools the catalyst bed.
Simultaneously, exothermic reactions take off,
Table 1. Conditions used for the reversed generating reaction heat, while the axial conduction
flow reactor simulations term tends to limit the slope of the heat wave by
Reactor transferring heat against the direction of the flow to
the cold section of the bed. The relative magnitude of
Diameter (m,) 0.1 these three phenomena determines the direction of the
Height (m,) 0.5 heat wave. For the conditions used in the present
Bulk density (kg/m:) 1050 paper, the heat wave and the gas flow move concur-
Void fraction (mi/m:) 0.4
rently.
Catalyst In the case of an exothermic equilibrium reaction,
Weight (kg) 4.12 like methanol synthesis, the temperature drop at the
Density (kg/m:) 1750 downstream end of the wave further favours conver-
Porosity (m:/m:) 0.5
Tortuosity 3
sion to methanol, resulting in a further generation of
Pellet diameter (m,) 0.0054 reaction heat. The heat is transferred by conduction
and by convection towards the colder solid down-
Operating conditions stream. Hence, in this region, the phenomena co-
To (K) 373.15
operate in broadening the heat wave as it moves
PP (bar) 50
Fp (mol/mf/s) 32.65 through the reactor (Fig. 1).
Cycle time (s) 300 Upon flow reversal, the shape of the new upstream
end of the wave is quickly adjusted to the new flow
Feed composition direction. Afterwards, the upstream slope of the wave
CO (mol %) 4.5
Hz0 (mol %) 0.0
tends to become constant and is mainly determined
MeOH (mol %) 0.0 by HJ-. From that time onwards, the velocity of the
H, (mol %) 88.5 upstream slope is a function only of the ratio of gas
CO2 (mol %) 2.0 and solid heat capacities, of the heat generated by
Inert (mol %) 5.0
reaction, of the feed temperature and of the flow rate.
It, therefore, becomes constant as soon as the wave is
free from time and spatial boundary effects. In what
methanol synthesis, with an “end of run” catalyst in follows, such a profile is called “well-established”.
the main loop reactor, although one might expect
a higher inert content in that case. The nitrogen and 3.2. Comparison of results obtained with various levels
methane content were kept fairly low, mainly to en- of modelling
able a wide variation of the operating conditions. Mass accumulation. Figure 2 shows the evolution
For the interpretation of this result, the energy of the exit concentration of methanol during one
equation for the solid is rewritten in dimensionless semi-cycle, as predicted by the three models. The
form. With conditions used in the calculations are again those of
Table 1, except for the semi-cycle time, now taken to
be 80 s. The most striking differences in the behaviour
of the three mode1 predictions occur in the first 20 s
equation (6) becomes after reversal. In the section of the reactor upstream of
the hot part of the bed, the synthesis gas is swept out
upon reversal, practically in its original composition.
80,
dr-
-$$+
ha
H/(6' - 0,) + Hr (6’)
This creates a dip in the time-dependent concentra-
CES 4B:19-B
3338 K.M. VANDEN BUSSCHE et al.

Fig. 2. Evolution of the exit concentration of methanol as

calculated by the three models. Condttions shown in Table 1.
Semi-cycle time = 80 s. I, II, III: models 1, II and III, re-
spectively.

tion of the reaction products. This feature, already

reported by Neophytides and Froment (1992) in their
study of a bench-scale reactor for reversed flow meth-
anol synthesis, only appears in the simulated profiles
if a mass capacity is assigned to the gas phase, as is
done in models II and III. If the gas phase inside the
catalyst pores is also accounted for, the gas mass
capacity increases by another 75%, explaining the
difference between profiles I1 and III of Fig. 2, corres-
ponding to models II and III, respectively.
The difference in conversion in the final 60 s is due
to the lower maximum temperature predicted by
model II or III. This is illustrated in Fig. 3(a) and (b),
showing the temperature and methanol concentration Fig. 3. (a) Catalyst temperature profiles along the reactor
profiles along the reactor after 22 s into the semi- after 22 s into the semi-cycle, as predicted by the three
cycle. In model I, which does not contain a gas-phase models. (b) Methanol concentration profiles along the reac-
tor after 22 s into the semi-cycle. Conditions summarised in
mass capacity, the response of the gas-phase composi-
Table 1. Semi-cycle time = 80s. I, II, III: models I, II, III,
tion to variations in the solid is instantaneous. Due to respectively.
the mass capacity assigned to the high-pressure gas
phase in models II and III, however, the fresh feed
entering the creeping reaction front sweeps away the
reacted gas mixture from the reaction zone with methanol, causing the differences in conversion men-
a characteristic time that is not negligible when com- tioned earlier.
pared to that of the heat front propagation. The It is, therefore, clear that mass accumulation plays
presence of methanol and water slows down the an important role in the reactor behaviour, not only
conversion. As a result, the reaction and the heat just after flow reversal, but also during the rest of the
production are spread out over a larger section of the semi-cycle, and has to be included in the model.
low-temperature region of the upstream end of the
wave. This leads to a faster movement of the heat Energy accumulation. The differences in the tem-
front along the reactor bed, in agreement with the perature profiles and therefore in the amount of heat
reaction and heat front theory (Matros, 1989). accumulated in the reactor were explained above us-
According to this theory, a faster movement of the ing heat and reaction front theory.
profile results in a higher maximum temperature, The energy accumulation in the gas phase itself is
which is reached at :he location where the reaction most apparent upon flow reversal. Figure 4(a) and (b)
reaches equilibrium. Since the opposite behaviour is shows the evolution of the temperature differences
observed in the present case, it is concluded that the between gas and solid along the reactor, calculated
set of reactions does not reach equilibrium at the using models I and III, respectively, just before and
observed maximum temperature. Therefore, although after flow reversal. The arrows indicate the direction
the heat front, wandering back and forth along the of flow. In curve a of Fig. 4(a), the cold feed extracts
fixed-bed reactor, exhibits the main characteristics of heat from the solid for [ between 0.1 and 0.4 and
the front of an exothermic reaction, it cannot be eventually reaches the solid temperature. At the
regarded as fully developed. This is ascribed to the downstream end of the wave, the catalyst is now
limited length of the bed compared with the width of heated by the gas; hence, the negative peak at
the reaction zone. The lower maximum temperature c = 0.95. Upon flow reversal (curve b), the gas flow
that results from this is beneficial for the conversion to immediately assumes the new pseudo-steady-state
 Simulation of reversed flow operation

Fig. 5. Influence of inlet temperature on the averaged

methanol concentration at the reactor exit (curve a) and on
the maximum temperature reached in the bed during one
semi-cycle (curve b).

As the inlet temperature is raised, the amount of

heat that needs to be extracted from the bed to take
the feed up to the ignition temperature (180°C) and
beyond, decreases. A smaller fraction of the reactor is
then used for heat regeneration and a larger fraction
for reaction. Since, for the major part of the semi-
Fig. 4. Temperature differencebetween gas and solid phase cycle, the effluent leaves the reactor at a temperature
along the reactor, prior to and just after flow reversal: (a) as below 180°C. at which the reaction no longer pro-
predicted by model I; (b) as predicted by model 111. ceeds, the shift of the equilibrium in the decreasing
part of the temperature profile is optimally used to
produce more methanol. As long as the feed enters the
value, corresponding to the new solid temperature reactor below 180°C. the temperature profiles are
profile and the new flow direction. The large temper- similar to the ones shown in Fig. 1.
ature difference at c = 1 arises as the cold fresh feed at In the cases where the temperature of the feed
373 K reaches the hot catalyst bed. exceeds the ignition temperature, the heat wave no
A more realistic representation is provided by longer moves through the bed with constant velocity
model III in Fig. 4(b). Upon reversal (curve b), a cer- and constant upstream slope. At the upstream end of
tain time is needed for the cold gas at c = 0.3 to be the reactor, the feed is still heated by the solid phase,
replaced by hot gas coming from the ignited section of but it also immediately reacts upon entering, thereby
the reactor. The same applies for the hot gas at the generating heat and creating a heat wave with a posit-
new upstream end of the wave, which is only grad- ive slope already at the reactor entrance. At the down-
ually replaced by the cold feed. After 3 s in the new stream end, the heat generated by reaction and the
semi-cycle (curve c), the transients caused by the re- heat transferred by convection and conduction from
versal have disappeared and a profile analogous to the central part of the bed raise the solid-phase tem-
curve b of Fig. 4(a) is obtained. perature, as shown in Fig. 6, for an inlet temperature
of 250°C. Optimum use is now made of the catalyst in
3.3. Influence of operating conditiovs the first par{ of the bed. The effluent always leaves the
The influence of inlet temperature (TO), semi-cycle reactor at temperatures. above 18O”C, however, so
time (t,), flow velocity (u,) and reactor length(L) on the that part of the beneficial effect of the temperature
reactor hehaviour was investigated using model III, drop at the outlet is lost. As a result, the time-aver-
leaving all other conditions as given in Table 1. aged carbon conversion to methanol drops again at
feed temperatures above 190°C (Fig. 5, curve a).
Influence of inlet temperature. The influence of inlet For inlet temperatures above 28O”C, finally, the
temperature on the reactor behaviour was studied usual monotonously increasing steady-state temper-
using the conditions of the base case, except for the ature profile is reached within the time frame of the
semi-cycle time, which was fixed at 80 s. semi-cycle.
Curve a in Fig. 5 shows the time-averaged exit Curve b in Fig. 5 shows the evolution of the max-
concentration of methanol at various feed temper- imum temperature with increasing feed temperature.
atures. For inlet temperatures below lOO”C, the heat For inlet temperatures below 18O”C, T,,,,, is practic-
generated in the reactor in a semi-cycle is insufficient ally independent of To. This is in agreement with
to compensate for the cooling by the gas phase upon Matros (1989). who derived this behaviour mathemat-
reversal, so that the reactor quickly reaches a non- ically for a single reversible reaction, from heat and
ignited steady state. reaction front theory. For feed temperatures exceed-
3340 K.M. VANDENBUSSCHE et al.

Fig. 7. Effect of semi-cycle length on the catalyst-bed tem-

perature profiles, just before flow reversal: (1) 12.25 s;
Fig. 6. Catalyst-bed temperature profiles along the reactor (2) 80 s; (3) 150 s; (4) 210 s; (5) 240 s.
at various times during one semi-cycle,for an inlet temper-
ature of 250-C, predicted by model III. Conditions as in
Table 1.

ing lSO”C, however, the reaction rate at the inlet is no

longer negligible and the theory can no longer be
applied.

Influence of semi-cycle length. As the heat wave

approaches the exit towards the end of the semi-cycle,
the energy generated by the reaction and accumulated
at the downstream end of the wave in the current
semi-cycle is partially driven out of the reactor
(Fig. I). When the semi-cycle time is increased, the
amount of heat retained inside the reactor decreases. Fig. 8. Effect of semi-cycle length on the evolution of the exit
concentration of methanol.
This causes the heat wave to narrow and eventually
leads to a decrease of the maximum temperature.
These effects are illustrated in Fig. 7, showing the
temperature profiles just before flow reversal for vari-
ous semi-cycle lengths and a feed temperature of about 190°C is reached, implying that the reaction no
IOO”C. longer significantly proceeds at lower temperatures.
If the semi-cycle time is too long, the amount of Indeed, the reaction is known to take off on this
energy retained in the reactor at the end of the semi- catalyst around 180°C. Curve 1 in this figure shows
cycle is too low to allow for a conversion in the next the evolution of the methanol concentration for a case
semi-cycle which is sufficient for compensating for the in which the semi-cycle time is of the order of the
cooling by the gas phase. The reactor then rapidly residence time of the gas in the reactor. In this case,
extinguishes. For the conditions considered here, the the corresponding temperature profile barely moves
reactor reaches a non-ignited state for semi-cycle during a semi-cycle (Fig. 7). Because of this, the size of
times exceeding 240s. An analogous behaviour was the cold area is no longer proportional to the semi-
reported by Thullie and Burghardt (1990) and, using cycle length. Moreover, since the temperature at the
simplified models, by Eigenberger and Nieken (1988), exit remains practically unchanged at 140°C during
Matros (1989) and Gawdzik and Rakowski (1988) for the full semi-cycle, there is no drop in the conversion
reactions A 4 B and A +t B. to methanol at the end of the semi-cycle. This results
Figure 8 compares the evolution of the exit concen- in the completely different concentration history
tration of methanol, plotted vs t/t., for various shown in Fig. 8.
semi-cycle lengths. The coincidence of the dip in the If the semi-cycle time is decreased further, below the
concentration profiles for large t, indicates that the residence time of the gas, the reactor extinguishes, in
corresponding heat waves are well established. In that agreement with Lee and Bailey (1974). In this case,
case, the upstream part of the wave moves along the methanol is produced and accumulates in the central
reactor at a constant velocity, identical for all t,, so part of the reactor. Its presence prevents the further
that, prior to flow reversal, the size of the cold section generation of reaction heat, resulting in a very slow
of the reactor is proportional to the semi-cycle length. extinction. Figure 9(a) and (b) shows the temperature
The nearly, but not exactly, coinciding drop at the end and methanol concentration profiles at various stages
of the curves only occurs when an exit temperature of of the extinction, for a semi-cycle length of 1 s.
 Simulation of reversed flow operation

Fig. 10. Influence of semi-cycle length on the time-averaged

effluent concentration of methanol (curve a) and on the
(a) maximum in time of the space-averaged catalyst temperature
(curve b).

The average catalyst temperature is then given by

Tav = To 0,d< = &To

indicating that T.,. ,,,==is also proportional to t,. This

shows once again that the profiles are well developed
or, in other words, that the individual phenomena
controlling the reactor behaviour are independent
of t,.
W - .*xWd4tMWObWUW_Wl.I-l
Influence of reactor length. The influence of reactor
Fig. 9. (a) Temperature profilesduring the extinction of the length on the reactor behaviour was investigated for
reversed flow reactor. (b) Methanol concentration profiles
during the extinction of the reversed flow reactor. Semi-cycle a semi-cycle time of 80 s. The reactor length was
length 1 s. Profiles taken after 0.5 s into the semi-cycle for varied between 0.35 and 3 m. The other parameters
clock times 56 s, loo0 s, . . , i.e. after 56 reversals, 1000 were kept at their values given in Table 1.
reversals, . Reactors shorter than 0.35 m quickly extinguish.
The heat retained in the vessel at the end of the
As can be seen from Figs 8 and 10 (curve a), the exit semi-cycle is too low to allow for a conversion and
concentration of methanol shows a maximum for heat production in the next semi-cycle compensating
t, = I14 s. For shorter semi-cycles, the central part of for the heat uptake by the feed. For reactors longer
than 3 m, the residence time is again of the order of
the bed, although ignited, does not contribute to
methanol production any further, because of equilib- the semi-cycle time, and the gas phase never becomes
rium limitations, while the steep slope at the down- well established. Even longer reactors are expected to
stream end of the wave prevents any significant fur- extinguish, because of methanol accumulation in the
ther increase in conversion. For t, exceeding 1 I4 s, the central part of the bed, as already shown in Fig. 7.
conversion decreases as the ignited part of the reactor Figure 11 summarises the reactor behaviour. The
becomes smaller. maximum temperature (curve a) goes through a max-
Curve b of Fig. 10 shows that a linear dependence imum at a length of OS m, decreasing slightly with
of the time maximum of the space-averaged catalyst increasing reactor size. The average concentration of
is apparent for all but the ex- methanol (curve b), on the other hand, shows a
temperature, T,,,,,,,
treme cases. Indeed, eq. (6) can be written as monotonous increase with the length, which is caused
by the decrease of the maximum temperature and by
the decreasing slope of the downstream end of the
heat wave. This behaviour is illustrated in Fig. 12(a)
and (b), showing the temperature and concentration
(6”) profiles along the reactor vs dimensionless reactor
length, after 44 s into a semi-cycle.
in which As can be seen from Fig. 12(b), the reaction only
t reaches equilibrium at the maximum temperature for
$ zzz- long reactors. The amount of heat generated at
tc the upstream end of the wave is larger for larger
so that lengths. Moreover, the heat is released in a smaller
30, section of the bed, thus increasing the density of the
7 a t..
a7 heat source distribution in the low-temperature re-
3342 K.M. VANDEN BUSSCHE~~ al.

Fig. Il. Influence of reactor length on (a) the maximum

temperature reached in a semi-cycle; (b) the time-averaged
exit concentration of methanol and (c)the maximum in time
of the space-averaged catalyst temperature.

Fig. 13. Influence of reactor length on (a) the catalyst tem-

perature and (b) methanol concentration profiles along the
reactor for a fixed ratio ui/L. Profiles taken at 44 s into the
semi-cycle: (1) 0.175m; (2) 0.2625 m; (3) 0.7 m; (4)’28 m.
(a)

iz
f
20
__
‘i
The same line of thought can be applied to explain
the behaviour of the downstream slope of the temper-
ature profile, which is responsible for the increase in
conversion (Fig. 11, curve b). Finally, it is worth
noting that the wave velocity is again nearly constant
for all cases. Therefore, the dynamic behaviour of the
B 1.0 reactor is roughly the same for large lengths and short
semi-cycle times or vice versa. This results in a linear
% dependence of T,,. malof l/L, as shown in curve c of
Fig. 11.
0 4
1
Influence of reactor length for constant residence and
(W semi-cycle rime. In an effort to eliminate the influence
of the residence time on the effect of increasing reactor
Fig. 12. Influence of reactor length on (a) the catalyst tem-
perature profiles and (b) methanol concentration profiles length, a series of runs was performed adapting L and
along the reactor. Profiles taken at 44 s into a semi-cycle: (1) ui simultaneously, so that uJL was kept constant.
0.35 m; (2) 0.5 m; (3) 1 m; (4) 3 m. Figure 13 shows the influence of ui (or L), for a fixed
ratio q/L, on the catalyst-bed temperature and gas-
phase methanol concentration profiles after 44 s into
gion of the wave. This results in a smaller creeping the semi-cycle. The main features are the increase in
velocity and hence according to the heat front theory, the maximum temperature of the heat front and, espe-
in a lower maximum temperature. Furthermore, the cially for the shorter reactors, the constant dimension-
increased amount of heat generated in the larger reac- less velocity w ( = v/ui) of the upstream end of the
tor enhances the axial conduction, decreasing the wave. Furthermore, the equilibrium methanol con-
slope of the upstream end of the temperature profile, centration is reached at the maximum temperature for
when plotted vs z. reactors exceeding 0.7 m.
 Simulation of reversed flow operation 3343

Fig. 14. influence of reactor length on the maximum temper-

ature reached in the bed at any time during the semi-cycle.

For shorter reactors. the constant dimensionless

velocity is ascribed to the interaction of two compen-
sating effects: the cooling of the catalyst bed by the gas
phase and the heat release by the reaction at the
upstream end of the wave. In these cases the increase
of the maximum temperature, shown in Fig. 14, is
attributed to the increase in the heat release at higher
flow rates.
For reactors with a length exceeding 0.7 m, the
wave behaves as if the bed were infinitely long and the
reaction and heat front theory can be applied. In this
case, the creeping velocity and the evolution of the
maximum temperature depend on the distribution of W AXWd!hfWX~Ul9rrcarmI-l

the heat generated in the reaction zone. (Matros, Fig. 15. Catalyst temperature (a) and methanol concentra-
1989) More explicitly, this means that the increase in tion (b) profiles along the reactor after 75 s into the semi-
the maximum temperature and in the dimensionless cycle, for various reactor lengths,keeping t,/L and ui con-
creeping velocity can be ascribed to a broader distri- stant:(1) 80 s; (2) I14 s; (3) 228s; (4) 684s.
bution of the heat generation in the low-temperature
region of the upstream end of the heat wave. The
observed asymptotic hehaviour of T,,, (Fig. 14) is in Obviously, the creeping velocity is the same in all
agreement with Rhee et al. (19743 who derived this cases in which the semi-cycle length exceeds 80 s,
behaviour mathematically for an infinitely long bed because the wave is well established and the gas velo-
and for a single irreversible reaction, assuming first- city remains unchanged. This results in an identical
order kinetics. dimensionless size of the cold area downstream of the
heat wave at any time during the semi-cycle, which
InjIuence of reactor lengthfor constant flow velocity explains the coincidence of the dip in the exit concen-
and dimensionless semi-cycle time. In this section, the tration profiles of methanol shown in Fig. 16. The
effect of the semi-cycle duration is studied, under velocity of the downstream end of the wave, on the
simultaneous adjustment of the reactor length L. In other hand, is not constant within a cycle. At this side,
this way, both the ratio t,lL and the gas velocity ui are the heat generated by reaction in the zone where the
kept constant. temperature decreases is added to the heat transferred
Figure 15 shows the catalyst temperature and con- by convection and conduction. This causes an in-
centration profiles for various semi-cycle lengths, crease in the velocity of the downstream end of the
plotted vs 5, at 7’ = t/tc = 0.65. For t, below 80 s, the wave with respect to that of the upstream front, result-
reactor extinguishes because the retained heat does ing in a broadening of the front as it moves through
not allow for a sufficient conversion in the next semi- the reactor. This widening is more explicit for larger
cycle to compensate for the cooling by the gas phase. t,, simply because of the longer duration of the semi-
The main features of the behaviour of the heat cycle, and hence the broadening of the wave for in-
front, due to simultaneous variation oft, and L, are (i) creasing L, featured in Fig. IS(a).
an identical and constant creeping velocity of the While for constant residence time and semi-cycle
upstream end of the wave for all cases with semi-cycle length the maximum temperature can, to a certain
times exceeding 80 s, (ii) a widening of the wave for extent, be controlled by adjusting ui or L, for constant
increasing reactor length and (iii) a shift of the posi- flow velocity the width of the wave is controlled by
tion of the maximum temperature towards the down- simultaneous adjustment of semi-cycle time and
stream end of the wave for longer semi-cycles. length.
3344 K. M. VANDEN BUSSCHE et al.

spectively, ccand l/L. This dependence was also de-

rived from a scrutiny of the mode1 equations.
For a constant ratio of residence time and semi-
cycle length, the well-established downstream front of
the heat wave tends to move with a constant dimen-
sionless velocity. For constant nir the width of this
wave can be controlled by adjusting t, and L, while, for
constant t,, the maximum is determined by ui and L.
The insight into the complex behaviour of the re-
versed flow reactor obtained in this study was used in
the design of a reactor for the conversion of the purge
-warneMo f-1 of a traditional methanol synthesis loop, described by
Vanden Bussche et al. (1992a). It has also contributed
Fig. 16. Evolution of the exit concentration of methanol, for
to the development of an economically interesting
various L, keeping u/L constant: (1) 80 s: (2) 114 s; (3) 228 s;
(4) 456 s; (5) 684 s. reversed flow alternative for the traditional multibed
adiabatic reactor for methanol synthesis (Vanden
Bussche et al., 1992b).

The broadening of the wave is also the reason for Acknowledgement-This work was funded in part by the
the decrease of the methanol concentration towards Commission of the European Communities in the frame-
the end of the semi-cycle, shown in Fig. 16, especially work of the Joule Programme, Subprogramme Rational Use
of Energy.
for long rC(and L), since a significant part of the profile
moves out of the bed, causing a high exit temperature.
As can be seen from Fig. 15(b), the exit concentra- NOTATION
tion of methanol benefits from the decrease in the external particle surface area per unit of
maximum temperature. Due to this, heat is accumu- reactor volume, m:/m,”
lated at the downstream end of the wave as the profile cj concentration of component j, mol/mi
moves along the reactor, resulting in a shift of the specific heat of the gas at constant pressure,
CP
maximum temperature towards this end of the profile. J/mol/K
CPS
specific heat of the solid at constant pres-
4. CONCLUSIONS sure, J/kg/K
It was shown that the introduction of mass and ct total concentration, mol/mj
energy capacity terms for both the bulk gas phase and 4 catalyst pellet diameter, m,
the gas phase contained inside the pores of the cata- F, total molar flux, mol/m~/s
lyst is essential for an adequate modelling of the F total volumetric flow rate, m,3/s
reversed flow operation of a fixed-bed reactor. The s gas solid heat transfer coefficient, W/m:/K
improvement in the accuracy of the model is most 4 heat transfer coefficient, dimensionless
pronounced upon flow reversal, but it is also substan- H, generated reaction heat, dimensionless
tial in the rest of the semi-cycle. The lowest level b mass transfer coefficient for component J,
model I is sufficient for the prediction of certain trends mol/mf/s
in the reactor behaviour, but, if operational aspects L total reactor length, m,
have to be. predicted, a higher level of modelling is m total mass flow rate, kg/s
required. M, average molecular weight, g/mol
For methanol synthesis, comprising a set of revers- N number of components in the mixture, di-
ible reactions, there is no unique correspondence be- mensionless
tween the amount of heat retained in the reactor and Nj molar flux for component j inside the cata-
the conversion, unlike the traditionally studied irre- lyst pellet, mol/m.z/s
versible oxidation reactions. The time-averaged car- PI total pressure, Pa
bon conversion to methanol mainly depends on the Peha Peclet number for energy in the axial direc-
width and height of the high-temperature plateau, and tion, dimensionless
on the upstream and downstream slopes of the heat ‘j reaction rate of component j, mol/kg,/s
wave. reaction rate of component j, averaged over
‘i
For a given reactor and feed flow rate, the variation a catalyst pellet, mol/k&/s
of the semi-cycle length leads to a carbon conversion universal gas constant, J/mol/K
into methanol that exhibits a maximum. From the clock time, s
conversion point of view, the optimum inlet temper- semi-cycle time. s
ature was found to coincide with the ignition temper- gas-phase temperature, K
ature of the reaction. space-averaged catalyst temperature, K
The behaviour for decreasing semi-cycle length and time maximum of the space-averaged cata-
increasing reactor length is more or less analogous, lyst temperature, K
resulting in a linear dependence of T,,,,, on, re- maximum temperature. K
 Simulation of reversed flow operation 3345

T, solid-phase temperature, K Graaf, G. H., Sijtsema P. J. J. M., Stamhuis, E. J. and

W catalyst weight, kg Joosten. G. E. H.. 1986. Chem. Enunrr Sci. 41. 2883-2889.
gas-phase mole fraction of component j, di- Graaf, G. H., Stamhuis, E: j. and B&&akkers,.A. A. C. M.,
Y, 1988, Chem. Engng Sri. 43, 318%3195.
mensionless Grabmullet, H.. Hoffmann. U. and SchPdlich, K.. 1985,
Yi mole fraction of j, at the external catalyst Chem. Engng Sci. 40, 951-960.
surface, dimensionless Lee, C. K. and Bailey, J. E., 1974, Chem. Engng Sci. 29.
2 axial reactor coordinate, m, 1157-I 168.
Marquardt, W., 1990, fnt. them. Engng 30, 585-606.
Matros, Yu. Sh., 1989, Studies SW- Sci. Catal. 43.
Greek letters Matros, Yu. Sh., 1990a. Chem. Engng Sci. 45. 2097-2102.
AH,. j enthalpy of formation of component j, J/mol Matros, Yu. Sh., 199Ob. Angew. Chem. tO2, 1274-1285.
E void fraction of the catalyst bed, m:/m,j Nam, Y. W., Hudgins, R. R. and Silveston, P. L., 1990, Chem.
Engng Sci. 45, 3111-3121.
a* pellet porosity, m,3/m:
Neopbytides, S. N. and Froment, G. F., 1992, Ind. Engng
r axial reactor coordinate, dimensionless Chem. Res. 31, 1583-1589.
8 gas-phase temperature, dimensionless Renken, A., 1972, Chem. Engng Sci. 27, 1925-1932.
& solid-phase temperature, dimensionless Rhee, H. K., Lewis, R. P. and Arnundson, N. R., 1974, Ind.
RCII effective axial conductivity of the solid Engng Chem. Fvndam. 13. 317-323.
Sapundzhiev Kh., Bunimovich, G. A., Drobyshevich, V. I.,
phase, W/m,/K Grozev.G., Yausheva, L.V., Matros, Yu. Sh. and Elenkov,
p gas-phase viscosity, Pa s D.. 1988. Tear. Osnovv Khim. Tekhnol. 22, 349-355.
t coordinate in the catalyst pellet, m, Schiidlich. K.. Hoffmann. U. and Hofmann. H.. 1983. Chem.
bulk density, kg,/m: Engng sci. -38, 137&84.
PS
solid density, kg,/m: Silveston, P. L., 1987, In Reactors and Reaction Engineering
P.
(Edited by R. A. Mashelkar and R. Kumar). Indian
2, 7’ time, dimensionless Academy of Sciences, Bangalore, India.
Thullie, J. and Burghardt, A., 1990, In Unsteady State Pro-
Superscript and subscript cesses in Catalysis (Edited by Yu. Sh. Matros). pp.
0 inlet conditions 687-692. VPS BV. Utrecht.
Vanden Bussche, K. M., Froment, G. F., Glass W., Bosch,
0 initial conditions
H. and Vandierendonck, L. L., 1992a, Proceedings of the
EC Conference Enerav Efficiencv in Process -_
Technoloav.
Vouliagmeni (Athens). - .
Vanden Bussche, K. M., Neophytides, S. N. and Froment,
REFERENCES
G. F. (to be published).
Balmer, J., 1989, Chemie ingr-Tech. 61, 632-635. Vanden Bussche, K. M., Neophytides S. N., Froment, G. F.,
Boreskov, G. K., Matros, Yu. Sh., Ivanov, A. A. and Volkov, Glasz, W., Bosch, H. and van Dierendonck, L. L., 1992b,
V. P., 1983, USSR certificate of authorship, 994400. Proceedings of rhe 8th Znrernational Conference on Large
Carey, G. F. and Finlayson, B. A., 1975, Chem. Engng Sci. 30, Chemical Plants, K.V.I.V., Antwerp.
587-596. Villadsen. J. and Michelsen. M. L.. 1978. Solution ofDifferen-
Eigenberger, G. and Nieken, U., 1988, Chem. Engng Sci. 43, rial Equarion Models ‘by P&no&/ Ap&&arion.
2109-2115. Prentice-Hall. Englewood CM%, NJ.
Ergun, S., 1952, Chem. Engng Prog. 48,89-95. Westertern. K. R.. 1992. Chem. Emma Sci. 47. 2195-2209.
Froment, G. F., 1990, In Unsteady State Processes in Cataly- Zagoruiko,. A. Nl, M&ros, Yu. $hI, Kumar, V. R. and
sis (Edited by Yu. Sh. Matros). VPS BV, Utrecht. Kulkarni. B. D., 1992, Chem. Engng Sci. 47.43154321.
Gawdzik, A. aid Rakowski, L.,.l988, Chem. Engng Sci. 43, Zagoruiko, A. N., Noskov, A. S., Drobyshevich. V. I.,
3023-3030. Yausheva, L. V., Malakhova. I. V. and Matros, Yu. Sh.,
Gosiewski, K. and Sztaba, R., 1990, in Vnsready State Pro- 1989. Tear. Osnovy Khim. Tekhnol. 23, 20--215.
cesses in Catalysis (Edited by Yu. Sh. Matros), pp. Zolotarski, 1. A., Bogdashev, S. M. and Matros, Yu. Sh.,
62-35. VPS BV, Utrecht. 1989, Kiner. Karal. 30, 1310-1317.