Chemical Engineering Science 60 (2005) 56375647

www.elsevier.com/locate/ces
Athree-phase nonequilibriumdynamic model for catalytic distillation
Yongqiang Xu, Yuxiang Zheng, Flora T.T. Ng

, Garry L. Rempel
Department of Chemical Engineering, University of Waterloo, Waterloo, Ont., Canada N2L 3G1
Received 26 February 2004; received in revised form 25 April 2005; accepted 2 May 2005
Available online 24 June 2005
Abstract
A detailed three-phase nonequilibrium (NEQ) dynamic model for simulating batch and continuous catalytic distillation (CD) processes
has been developed. In this model, both molar and energy holdups in liquid and vapour phases are taken into account. Multicomponent
mass transfer and heat transfer between vapour and liquid phases as well as between liquid and solid (catalyst) phases are described by
the MaxwellStefan equations. The resulting differential and algebra equations in this model are implemented in gPROMS and C++.
The simulation results are in good agreement with the experimental data obtained from the batch and continuous CD processes for the
production of diacetone alcohol (DAA) using Amberlite IRA-900 as a catalyst. Sensitivity analysis on the mass transfer and kinetics
using the three-phase NEQ dynamic model indicates that the formation of DAA is controlled by solidliquid mass transfer, whereas the
formation of mesityl oxide is kinetically controlled under the simulation conditions.
2005 Elsevier Ltd. All rights reserved.
Keywords: Catalytic distillation; Dynamic simulation; Nonequilibrium modelling; Mass transfer; Diacetone alcohol
1. Introduction
Catalytic distillation (CD) is a unit operation combining
a heterogeneous catalysed reaction and distillation together
and has many advantages, such as increasing product con-
version and selectivity, saving energy by utilizing the reac-
tion heat and prolonging catalyst lifetime by avoiding the
formation of hot spots because the catalyst is surrounded by
the boiling liquid in the CD reaction zone (Ng and Rempel,
2003; Podrebarac et al., 1997a). Most of the CD processes
at present are continuous processes used for producing large
quantities of chemicals. However, there are also literature
reports on batch CD processes (Podrebarac et al., 1998a;
Schneider et al., 2001).
A dynamic model, which can be used to simulate batch
CD processes and step changes of operating variables as
well as start up and shutdown procedures in the continuous
CD processes, is needed for a better understanding of the

Corresponding author. Tel.: +1 519 888 4567x3979;

fax: +1 519 746 4979.
E-mail address: fttng@cape.uwaterloo.ca (F.T.T. Ng).
0009-2509/$ - see front matter 2005 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ces.2005.05.025
complex behaviour of the CD column. However, most of the
existing dynamic models are based on the equilibrium (EQ)
assumptions between vapour and liquid phases (Abufares
and Douglas, 1995; Sneesby et al., 1998). In the EQ model,
the concept of stage efciency or height equivalent to a
theoretical plate (HETP) is generally introduced. However,
the estimation of stage efciency or HETP is very difcult
for a CD process.
Recently, a number of publications on two-phase nonequi-
librium dynamic models have been presented. A two-phase
nonequilibrium (NEQ) dynamic model was developed by
Kooijman and Taylor (1995) for the simulation of distilla-
tion processes which was based on the MaxwellStefan the-
ory. Kreul et al. (1999) also reported a two-phase dynamic
NEQ model for simulating batch distillation columns. The
dynamic distillation models were further extended to simu-
late the CD columns by including the reaction terms (Kreul
et al., 1998; Schneider et al., 2001). Recently, Baur et al.
(2000, 2001) developed a dynamic NEQ cell model for CD
columns in which the use of multiple well-mixed cells in
the liquid and vapour ow directions accounted for staging
in either uid phase. A comparison of the dynamic EQ and
5638 Y. Xu et al. / Chemical Engineering Science 60 (2005) 56375647
NEQ models was given by Peng et al. (2003). However, all
of these pseudo-homogeneous (two-phase) dynamic models
do not consider the mass and heat transfer between the liquid
phase and external catalyst surface.
In our laboratory, Huang et al. (1998, 2000) reported a
three-phase NEQ model (MECRES equations) for a steady-
state CD process where the empirical overall vapourliquid
and liquidsolid mass transfer coefcients were used. Re-
cently, Zheng et al. (2001, 2003a,b) rened this MECRES
model by taking into account the mass and heat transfer
among phases according to the MaxwellStefan theory. It
was validated by comparison with the experimental data ob-
tained for the CD processes of the aldol condensation of
acetone (Zheng et al., 2001) and the synthesis of ethyl cello-
solve (Zheng et al., 2003a). This model is also effective for
design and can be used to provide the optimal design and
operation parameters of a CD process (Zheng et al., 2003b).
A CD process is more complex than a homogeneous reactive
distillation process due to the existence of the heterogeneous
reactions. In general, a three-phase NEQ model is more ap-
propriate for the simulation of CD because there are con-
centration and temperature gradients between the liquid and
solid phases in most reaction systems. Recently, we reported
that a three-phase NEQ model is more appropriate for a CD
process when the system is solidliquid mass transfer con-
trolled (Zheng et al., 2004).
However, there is no literature report on modelling a
dynamic CD process by using a three-phase NEQ model.
In this paper, a three-phase NEQ dynamic model is devel-
oped which accounts for mass and heat transfer between
liquid and catalyst surface based on the MaxwellStefan
theory. The model reaction in this paper is the aldol con-
densation of acetone to produce diacetone alcohol (DAA)
using the Amberlite IRA-900 anion exchange resin as
a catalyst (Podrebarac et al., 1998a). The catalyst was
held within the breglass bags with approximately 1.5 cm
in diameter inside the CD column, which has been de-
scribed by Smith (1984). The aldol condensation is a
consecutive reaction and DAA is an intermediate prod-
uct which undergoes dehydration to produce mesityl ox-
ide (MO) and water. The reaction scheme is provided
in Eq. (1).
2Ac =DAA MO +H
2
O. (1)
DAA is a very useful solvent for cellulose acetate, ni-
trocellulose, vinyl chloridevinyl acetate, and epoxy resins.
DAA is also an intermediate product for manufacturing
methyl isobutyl ketone (Podrebarac et al., 1998a). The re-
action has an unfavourable equilibrium for the production
of DAA. The equilibrium conversion at the normal boiling
point of acetone, 54

C, is only 4.3 wt%. However, the CD

process was found to increase DAA conversion greatly by
in situ removal of it from the reaction zone (Podrebarac et
al., 1998a).
2. A three-phase NEQ dynamic model
Aphysical model of a typical NEQsection is shown in Fig.
1. The column is divided into a number of NEQsections. The
following assumptions are made in the model development:
(1) The condenser and the reboiler are assumed to be at
equilibrium.
(2) Vapourliquid equilibrium occurs at the vapourliquid
interface.
(3) Perfect mixing for bulk phases.
(4) No mass and energy holdups in the condenser, adjacent
lms, and solid phases.
(5) Constant pressure along the CD column.
(6) The catalyst particles are completely wetted with the
liquid.
The mathematical equations for the three-phase NEQ dy-
namic CD model are described below.
2.1. Total condenser (j =1)
For the total condenser, the mass balance is as follows at
total reux:
l
i1
v
i,2
=0, i =1, 2, . . . , C (2)
and temperature
T
1
=T
bubble
. (3)
2.2. Nonequilibrium sections (j =2, . . . , n 1)
For the vapour phase:
Mass balance
dM
V
ij
dt
=v
ij
v
i,j+1
f
V
ij
+N
V
ij
, i =1, 2, . . . , C. (4)
Heat balance
d

C
i=1
M
V
ij
H
V
j
dt
=
c

i=1
v
ij
H
V
j

c

i=1
v
i,j+1
H
V
j+1

i=1
f
i,j
H
VF
j
+e
V
j
. (5)
Mass transfer rate
N
V
ij
=
c1

k=1
k
V
ikj
a
j
_
y
kj
y
I
kj
_
+y
ij
c

k=1
N
V
kj
,
i =1, 2, . . . , C 1. (6)
Heat transfer rate
e
V
j
=h
V
j
a
j
c
V
j
exp c
V
j
1
_
T
V
j
T
I
j
_
+
c

k=1
N
V
kj
H
V
kj
. (7)
Y. Xu et al. / Chemical Engineering Science 60 (2005) 56375647 5639
l
i,j-1
v
i,j+1
H
j
L
T
j
L
l
ij
H
j
V
T
j
V
v
ij
L
j
H
1
Vapour
Liquid
Catalyst
V
j
T
1 +
L
j
T
1
V
-
L

i
n
t
e
r
f
a
c
e
L
-
S

i
n
t
e
r
f
a
c
e
L
ij
f
LF
j
H
LF
j
T
VF
j
VF
j
V
ij
T H f
V
ij
N
S
ij
N
V
j
e
L
j
e
S
j
e
r
j
Q
V
j
H
1 +
R
ij
l
i,j-1
v
i,j+1
H
j
L
T
j
L
l
ij
H
j
V
T
j
V
v
ij
Section j
L
j
H
1
V
j
T
1 +
L
j
T
1
L
ij
f
LF
j
H
LF
j
T
VF
j
VF
j
V
ij
T H f
L
ij
N
V
ij
N
S
ij
N
V
j
e
L
j
e
S
j
e
r
j
Q
V
j
H
1 +
Fig. 1. Three phase model of a nonequilibrium section.
Summation equation
C

i=1
y
I
ij
1 =0. (8)
For the liquid phase:
Mass balance
dM
L
ij
dt
=l
ij
l
i,j1
f
L
ij
N
L
ij
+N
S
ij
,
i =1, 2, . . . , C. (9)
Heat balance
d

C
i=1
M
L
ij
H
L
j
dt
=
c

i=1
l
ij
H
L
j

c

i=1
l
i,j1
H
L
j1

i=1
f
L
ij
H
LF
j
e
L
j
+e
S
j
. (10)
Mass transfer rate
N
L
ij
=
c1

k=1
k
L
ikj
a
j
_
x
I
kj
x
kj
_
+x
ij
c

k=1
N
L
kj
,
i =1, 2, . . . , C 1. (11)
Heat transfer rate
e
L
j
=h
L
j
a
j
_
T
I
j
T
L
j
_
+
c

k=1
N
L
kj
H
L
kj
. (12)
Summation equation
C

i=1
x
I
ij
1 =0. (13)
For the liquidsolid interface:
Mass balance
N
S
ij
v
i
R
ij
G
j
=0, i =1, 2, . . . , C. (14)
Heat balance
e
S
j
Q
r
j
=0. (15)
Mass transfer rate
N
S
ij
=
c1

k=1
k
S
ikj
a
S
j
_
x
kj
x
S
kj
_
+x
ij
c

k=1
N
S
kj
,
i =1, 2, . . . , C 1. (16)
Heat transfer rate
e
S
j
=h
S
j
a
j
_
T
L
j
T
S
j
_
+
c

k=1
N
S
kj
H
L
kj
. (17)
Summation equation
C

i=1
x
S
ij
1 =0. (18)
For the vapourliquid interface:
Continuity equations
N
V
ij
=N
L
ij
, i =1, 2, . . . , C, (19)
e
V
j
=e
L
j
. (20)
Vapourliquid equilibrium
y
I
ij
=K
I
ij
x
I
ij
, i =1, 2, . . . , C. (21)
2.3. Reboiler (j =n)
Mass balance
dM
L
in
dt
=v
in
+l
in
l
i,n1
, i =1, 2, . . . , C. (22)
Heat balance
d

C
i=1
M
L
in
H
L
n
dt
=
c

i=1
v
in
H
V
n
+
c

i=1
l
in
H
L
n

i=1
l
i,n1
H
L
n1
+Q. (23)
Vapourliquid equilibrium
K
in
x
in
y
in
=0, i =1, 2, . . . , C. (24)
The liquid molar holdup for structured packings in the
reaction zone is computed using the empirical correlation
developed by Xu et al. (1997):
h
d
=0.0336u
0.0109
L
0.429
. (25)
The liquid molar holdup for random packings in the non-
reaction zone is estimated by the empirical correlation de-
veloped by Billet and Schultes (1991):
h
d
=
_
12
p
L
La
2
t
gj
L
m
_
1/3
. (26)
5640 Y. Xu et al. / Chemical Engineering Science 60 (2005) 56375647
The multicomponent mass transfer coefcients k are com-
puted by the MaxwellStefan approach (Taylor and Krishna,
1993). The binary mass transfer coefcients for the non-
reaction zone are estimated using the empirical correlation
developed by Onda et al. (1968):

V
ik
=2
_
W
V
a
t
j
V
m
_
0.7
_
Sc
V
ik
_
1/3
(a
t
d
p
)
2
_
a
t
D
V
ik
P
R
g
T
V
_
, (27)

L
ik
=0.0051
_
W
L
a
w
j
L
m
_
2/3
_
Sc
L
ik
_
0.5
(a
t
d
p
)
0.4

_
gj
L
m
j
L
m
_
1/3
j
L
m
. (28)
Here,
a
w
=a
t
_
_
_
1 exp
_
_
1.45
_
W
L
a
t
j
L
m
_
0.1
_
a
t
(W
L
)
2
gj
L
2
m
_
0.05

_
(W
L
)
2
a
t
o
m
j
L
m
_
0.2_
o
m
o
c
_
0.75
_
_
_
_
_
. (29)
In addition, the effective interfacial area a is calculated using
the correlation developed by Bravo and Fair (1982).
a =9.7912a
t
_
o
0.5
m
Z
0.4
_
_
6W
V
j
L
m
W
L
a
t
j
V
m
j
L
m
o
m
g
_
0.392
. (30)
The binary mass transfer coefcients for the catalyst pack-
ing are evaluated by the empirical correlation developed by
Zheng and Xu (1992).

V
ik
a =1.072 10
3
a
t
D
V
ik
d
P
R
g
T
V
_
4W
V
a
t
j
V
m
_
0.92

_
4W
L
a
t
j
L
m
_
0.24
_
Sc
V
ik
_
0.5
, (31)

L
ik
a =0.149
a
t
D
L
ik
d
P
_
4W
L
a
t
j
L
m
_
0.3
_
Sc
L
ik
_
0.5
, (32)

S
ik
a =0.586
a
t
W
L
j
L
_
4W
V
a
t
j
V
m
_
0.27

_
4W
L
a
t
j
L
m
_
0.28
_
Sc
L
ik
_
2/3
. (33)
Heat transfer coefcients are computed by using the
ChiltonColburn analogy (King, 1980).
h
V
=k
V
av
C
V
pm
_
Le
V
_
2/3
, (34)
h
L
=k
L
av
C
L
pm
_
Le
L
_
1/2
. (35)
The macrokinetic data obtained by Podrebarac et al.
(1997b) are used.
R
DAA
=k
1
C
2
Ac
(k
1
+k
2
)C
DAA
, (36)
R
MO
=R
H
2
O
=k
2
C
DAA
. (37)
The equilibrium constant is computed by
K
I
ij
=
P
o
ij

ij
P
j
, (38)
where the liquid activity coefcient is computed by the
modied UNIQUACmodel (Prausnitz et al., 1980). Thermo-
physical constants such as density, diffusivity, enthalpy, heat
conductivity, viscosity, vapour pressure, and surface tension
were calculated based on the correlations suggested by Reid
et al. (1987) and by Danbert and Danner (1989).
3. Simulation results and discussions
3.1. Simulation results for the batch bench-scale CD
column
Batch CD experiments were carried out by Podrebarac
et al. (1998a) in a 1.2 m tall bench-scale glass column with
internal diameter of 2.54 cm as shown in Fig. 2. A 25.4 cm
layer of 6 mm ceramic Intalox saddles was randomly
dumped above the catalyst zone, and 21.4 ml of catalyst
was put inside one catalyst bag with a height of 9.63 cm
and a diameter 1.5 cm. The remainder of the packed col-
umn below the catalyst zone was lled with 6 mm ceramic
Intalox saddles. The reboiler heater was set to 75V, which
provided a reux ow rate of 12.27 g min
1
of acetone.
The column was operated at the total reux. At the start of
the batch CD experiment, 750 ml of acetone was charged
into the reboiler. The zero time corresponded to the time
when the vapour contacted the bottom of the reaction zone
in the batch CD experiments. The detailed information of
the batch CD experiments is well documented in the paper
by Podrebarac et al. (1998a). However, in this calculation
the zero time was assumed to be when the vapour reached
the condenser due to the limitation of the model calculation
since both vapour and liquid are required in the compu-
tation. Hence, it is assumed that no reaction takes place
before the vapour contacted the condenser. This assumption
is reasonable because the column is very short.
Based on the above operating conditions, the batch CD
column is simulated by setting all feed rates and the reboiler
output to zero in the proposed three-phase dynamic NEQCD
model. The packed column is evenly divided into 50 NEQ
sections along the packed column with 10 sections in the
rectifying zone, four sections in the reaction zones, and 36
sections in the stripping zone. It was found that this number
of sections is sufcient in terms of modelling accuracy. The
condenser and the reboiler are assumed to be at equilibrium.
The predicted mole fractions in the reboiler with the
change of time are shown in Fig. 3. A comparison of the
predicted results with the experimental data reveals that the
newly developed dynamic model predicts the mole fraction
of DAA very well up to 7.5 h. After 7.5 h, the predicted
Y. Xu et al. / Chemical Engineering Science 60 (2005) 56375647 5641
Condenser
Stripping zone
Rectifying zone
Catalyst zone
Reboiler
Heater
1
2
11
12
15
16
51
52
Fig. 2. Schematic diagram of the batch CD column.
0
20
40
60
80
100
0 10 15
Time (h)
L
i
q
u
i
d

M
o
l
e

F
r
a
c
t
i
o
n

(
%
)
0
10
20
30
40
50
L
i
q
u
i
d

M
o
l
e

F
r
a
c
t
i
o
n

(
%
)
Measured Ac
Predicted Ac
Measured DAA
Measured Water
Measured MO
Predicted DAA
Predicted Water
Predicted MO
5
Fig. 3. Measured and predicted liquid mole fractions in the reboiler of
the batch CD column as a function of time.
329.20
329.60
330.00
330.40
Section Number
L
i
q
u
i
d

T
e
m
p
e
r
a
t
u
r
e

(
K
)
at 0 min
at 1 min
at 10 min
at 100 min
1 11 21 31 41 51
Fig. 4. Simulated liquid temperature prole along the batch CD column
as a function of time.
mole fraction of DAA is higher than the measured value
which may be attributed to catalyst deactivation (Podrebarac
et al., 1998a). However, it is somewhat surprising that the
MO production is not affected signicantly if the devia-
tion is due only to catalyst deactivation. Another plausible
explanation for the deviation observed for the DAA liquid
mole fraction may be due to the slight changes in exper-
imental conditions such as the cooling water temperature
and ow rate and uctuations in room temperature which
lowered the reux ow rate and hence the mass transfer
rate. We have shown previously that the DAA productivity
is sensitive to mass transfer, while the production of MO is
kinetically controlled (Zheng et al., 2003b). Therefore low-
ering the reux ow rate will decrease the productivity for
DAA, while the productivity for MO will not be affected.
Fig. 4 shows the liquid temperature proles along the col-
umn as a function of time. It was observed that the tem-
perature increased very quickly in sections 1215 because
the reaction occurred in the reaction zone. However, the ac-
tual temperature change is very small since the reaction to
DAA is mildly exothermic (27 KJ mol
1
) and the reac-
tion to MO is endothermic (25 KJ mol
1
) (Podrebarac et
al., 1998b). These temperatures are even higher than that in
the reboiler at the start of the runs. However, the temper-
ature along the column does not change signicantly after
10 min, while the temperature in the reboiler increases due
to the accumulation of the high boiling point products.
Figs. 58 give the proles of vapourliquid mass transfer
rates for the different components along the column with the
change of time. It was found that the vapourliquid mass
transfer rates for the components could be positive (conden-
sation) or negative (evaporation) with the change of time
5642 Y. Xu et al. / Chemical Engineering Science 60 (2005) 56375647
-6.0E-06
-3.0E-06
0.0E+00
3.0E-06
Section Number
M
a
s
s

T
r
a
n
s
f
e
r

R
a
t
e

(
m
o
l
/
s
)
at 0 min
at 1 min
at 10 min
at 100 min
1 11 21 31 41 51
Fig. 5. Simulated vapourliquid mass transfer rate of Ac along the batch
CD column as a function of time.
-2.0E-07
-1.0E-07
0.0E+00
1.0E-07
2.0E-07
3.0E-07
Section Number
M
a
s
s

T
r
a
n
s
f
e
r

R
a
t
e

(
m
o
l
/
s
)
at 0 min
at 1 min
at 10 min
at 100 min
1 11 21 31 41 51
Fig. 6. Simulated vapourliquid mass transfer rate of DAA along the
batch CD column as a function of time.
which indicates the condensation of the vapour phase or the
vaporization of the liquid phase at the specic conditions.
It is interesting to note that the vapourliquid mass trans-
fer rates of Ac at the sections below the reaction zone are
positive (Fig. 5), while of DAA and MO are negative at the
start of operation (Figs. 6 and 7). This indicates that Ac is
condensed from vapour phase to liquid phase, while DAA
-5.0E-07
-2.5E-07
0.0E+00
2.5E-07
5.0E-07
Section Number
M
a
s
s

T
r
a
n
s
f
e
r

R
a
t
e

(
m
o
l
/
s
)
at 0 min
at 1 min
at 10 min
at 100 min
1 11 21 31 41 51
Fig. 7. Simulated vapourliquid mass transfer rate of MO along the batch
CD column as a function of time.
and MO are vaporized from the liquid phase to the vapour
phase although the boiling point of Ac is much lower than
that of DAA or MO. This is due to the fact that there is no
DAA and MO in the CD column prior to the start of reac-
tions. Furthermore, DAA and MO should appear in the liq-
uid phase rst rather than in the vapour phase immediately
after reaction since only the liquid phase is directly in con-
tact with the solid phase (catalyst). This also shows that the
reactant acetone in the column is being replaced gradually
by the products as the reaction proceeds before the process
reaches the steady state. Fig. 8 shows the variation of the
vapourliquid mass transfer rate of water along the batch
CD column. In the upper part of the column (above the re-
action zone) the water is transferred from the vapour phase
to the liquid phase and the mass transfer rate increases as
the reaction proceeds. However, it is moving into the vapour
phase in the lower part of the column at 10 min after the
start of the reaction and transferred from the vapour phase
to the liquid phase again at 100 min.
Normally the molar and energy holdup of the vapour
phase can be neglected when the operating pressure is low.
We indeed found that the same simulation results were ob-
tained with or without the inclusion of the vapour holdup.
However, it was found that the simplied model without the
inclusion of the vapour holdup became very stiff and was
much more difcult to be solved than the complete model.
Therefore the vapour holdup was included in the three-phase
NEQ dynamic model. Similar observation on the stiffness
for solving a pseudo-homogenous dynamic CD model with-
out the inclusion of the vapour holdup was reported (Peng
et al., 2003).
Y. Xu et al. / Chemical Engineering Science 60 (2005) 56375647 5643
-2.0E-06
-1.0E-06
0.0E+00
1.0E-06
2.0E-06
3.0E-06
4.0E-06
Section Number
M
a
s
s

T
r
a
n
s
f
e
r

R
a
t
e

(
m
o
l
/
s
)
at 0 min
at 1 min
at 10 min
at 100 min
1 11 21 31 41 51
Fig. 8. Simulated vapourliquid mass transfer rate of water along the
batch CD column as a function of time.
3.2. Simulation results for the continuous pilot-plant CD
column
Continuous CD processes were carried out in a pilot-plant
column with a diameter of 2.54 cm and height of 4.88 m.
The column is divided into three parts: the rectifying zone,
reaction zone, and stripping zone. The rectifying zone lo-
cated in the top of the column is 91.44 cm long which was
lled with random packing of 6 mm ceramic Intalox sad-
dles. The Amberlite IRA-900 anion exchange resin catalyst
was put in the reaction zone which was 42.8 ml and enclosed
in two bags with a total height of 19.24 cm and a diame-
ter of 1.5 cm. The remainder of the packed column below
the catalyst zone was also lled with 6 mm ceramic Intalox
saddles. Acetone was fed in the middle of the column at
152 ml h
1
. The column was operated under total reux. The
bottom ow rate was adjusted so that the liquid level in the
reboiler remained constant. The reboiler duty was initially
set at 280W under which the reux ow is at 25.4 g min
1
.
When steady state was reached under these conditions, the
reboiler duty was dropped to 190W under which the reux
ow is 15.6 g min
1
. Detailed information was also docu-
mented by Podrebarac et al. (1998a).
The three-phase dynamic NEQ model was used to sim-
ulate the above continuous CD process. The column was
divided into 50 NEQ sections and two equilibrium stages.
Section numbers from 2 to 11 and from 14 to 51 are non-
reaction zones, and section numbers from 12 to 13 are the
reaction zones. The condenser and the reboiler are consid-
ered as an equilibrium stage.
The predicted and measured mole fractions in the reboiler
as a function of time are shown in Fig. 9. It shows that the
predicted results are in good agreement with the experimen-
0
0.2
0.4
0.6
0.8
1
0 5 10 15 20
Time (h)
L
i
q
u
i
d

M
o
l
e

F
r
a
c
t
i
o
n
0
0.1
0.2
0.3
0.4
0.5
L
i
q
u
i
d

M
o
l
e

F
r
a
c
t
i
o
n
Measured Ac
Predicted Ac
Measured DAA
Measured MO
Predicted DAA
Predicted MO
Fig. 9. Predicted and measured liquid mole fractions in the reboiler of the
pilot-plant CD column as a function of time for a negative step change
in reboiler duty from 280 to 190W.
330
332
334
336
338
340
0 5 10 15 20
Time (h)
T
e
m
p
e
r
a
t
u
r
e

(
K
)
Measured reboiler temperature
Predicted reboiler temperature
Fig. 10. Predicted and measured temperature in the reboiler of the pi-
lot-plant CD column as a function of time for a negative step change in
reboiler duty from 280 to 190W.
tal data although the DAA mole fraction is slightly under-
estimated, while the MO mole fraction is slightly overesti-
mated. Fig. 10 shows the liquid phase temperature in the re-
boiler as a function of time. A good agreement between the
predicted and experimental results is also obtained. It was
found that the temperature along the column decreases with
the decrease of the reboiler duty. This result is expected since
5644 Y. Xu et al. / Chemical Engineering Science 60 (2005) 56375647
-2.5E-04
-1.5E-04
-5.0E-05
5.0E-05
Section Number
M
a
s
s

T
r
a
n
s
f
e
r

R
a
t
e

(
m
o
l
/
s
)
at 0 min
at 10 min
at 100 min
at 1000 min
1 11 21 31 41 51
Fig. 11. Simulated vapourliquid mass transfer rate of Ac along the
pilot-plant CD column as a function of time for a negative step change
in reboiler duty from 280 to 190W.
-2.0E-06
2.0E-06
6.0E-06
1.0E-05
1.4E-05
Section Number
M
a
s
s

T
r
a
n
s
f
e
r

R
a
t
e

(
m
o
l
/
s
)
at 0 min
at 10 min
at 100 min
at 1000 min
1 11 21 31 41 51
Fig. 12. Simulated vapourliquid mass transfer rate of DAA along the
pilot-plant CD column as a function of time for a negative step change
in reboiler duty from 280 to 190W.
the conversion of acetone decreases with the decrease of the
reboiler duty which means less high boiling point products
are produced. Variations of the vapourliquid mass transfer
rates of Ac and DAA along the CD column are shown in
Figs. 11 and 12. It can be seen that the mass transfer rates
decrease with the decrease of the reux ow rates as ex-
pected and it is clearly seen in the lower part of the column
close to the reboiler.
0.08
0.12
0.16
0 5 10 15 20
Time (h)
L
i
q
u
i
d

M
o
l
e

F
r
a
c
t
i
o
n
Normal transient
Increasing LF by 20%
Decreasing LF by 20%
Fig. 13. Effect of the liquidsolid mass transfer coefcient on dynamic
responses of the liquid mole fraction of DAA in the reboiler for a negative
step change in reboiler duty from 280 to 190W (LS: liquidsolid mass
transfer coefcient).
3.3. Sensitivity analysis on mass transfer and kinetics
The mass transfer among phases and the kinetics are the
most important parameters for modelling a CD process us-
ing a three-phase NEQ model. The inuences of vapour lm
and liquid lm mass transfer between the vapour and liquid
phases, solidliquid mass transfer between the liquid phase
and catalyst surface, and the kinetics on the dynamic re-
sponse of the continuous CD column are studied by chang-
ing their corresponding values by 20% of the original ones
during the simulation. There is no signicant effect on the
dynamic response of the liquid mole fraction of DAA and
MO in the reboiler when varying vapour lm or liquid lm
mass transfer rate between the vapour and liquid phases
(Figs. 13 and 14). This result suggests that the vapourliquid
mass transfer is not the controlling step of this reaction sys-
tem.
The inuence of solidliquid mass transfer on the dynamic
responses of the liquid mole fractions of DAA and MO in the
reboiler is also evaluated. Figs. 15 and 16 are the simulation
results to present the differences of the liquid mole fraction
of DAA and MO for variation of solidliquid mass transfer
coefcients within 20% of their original values. It is clear
that the formation of DAA is impacted signicantly by the
solidliquid mass transfer although the formation of MO is
not signicantly affected.
The effect of the kinetics on the dynamic responses of the
CD performance is examined by changing the reaction rate
constant by 20%. Figs. 17 and 18 show the effect of the
kinetics on the liquid mole fractions of DAA and MO in the
reboiler, respectively. It could be observed that the formation
Y. Xu et al. / Chemical Engineering Science 60 (2005) 56375647 5645
0.04
0.06
0.08
0.10
0 5 10 15 20
Time (h)
L
i
q
u
i
d

M
o
l
e

F
r
a
c
t
i
o
n
Normal transient
Increasing VF by 20%
Decreasing VF by 20%
Fig. 14. Effect of the vapour lm mass transfer coefcient on dynamic
responses of the liquid mole fraction of MO in the reboiler for a negative
step change in reboiler duty from 280 to 190W (VF: vapour lm mass
transfer coefcient).
0.08
0.12
0.16
0 5 10 15 20
Time (h)
L
i
q
u
i
d

M
o
l
e

F
r
a
c
t
i
o
n
Normal transient
Increasing LS by 20%
Decreasing LS by 20%
Fig. 15. Effect of the liquidsolid mass transfer coefcient on dynamic
responses of the liquid mole fraction of DAA in the reboiler for a negative
step change in reboiler duty from 280 to 190W (LS: liquidsolid mass
transfer coefcient).
of MO is greatly inuenced by the kinetics, whereas the for-
mation of DAA is not sensitive to the change of the reaction
constant. Therefore, it can be concluded that the formation
0.00
0.02
0.04
0.06
0.08
0.10
0 5 10 15 20
Time (h)
L
i
q
u
i
d

M
o
l
e

F
r
a
c
t
i
o
n
Normal transient
Increasing LS by 20%
Decreasing LS by 20%
Fig. 16. Effect of the liquidsolid mass transfer coefcient on dynamic
responses of the liquid mole fraction of MO in the reboiler for a negative
step change in reboiler duty from 280 to 190W (LS: liquidsolid mass
transfer coefcient).
0.08
0.12
0.16
0 5 10 15 20
Time (h)
L
i
q
u
i
d

M
o
l
e

F
r
a
c
t
i
o
n
Normal transient
Increasing k by 20%
Decreasing k by 20%
Fig. 17. Effect of kinetic rate constants on dynamic responses of the
liquid mole fraction of DAA in the reboiler for a negative step change
in reboiler duty from 280 to 190W.
of DAA is controlled by solidliquid mass transfer where
the formation of MO is kinetically controlled under the sim-
ulation conditions. This conclusion is in agreement with the
analysis of Zheng et al. (2003b) obtained for the steady-state
simulation of the aldol condensation CD process.
5646 Y. Xu et al. / Chemical Engineering Science 60 (2005) 56375647
0.04
0.06
0.08
0.10
0 5 10 15 20
Time (h)
L
i
q
u
i
d

M
o
l
e

F
r
a
c
t
i
o
n
Normal transient
Increasing k by 20%
Decreasing k by 20%
Fig. 18. Effect of kinetic rate constants on dynamic responses of the
liquid mole fraction of MO in the reboiler for a negative step change in
reboiler duty from 280 to 190W.
4. Conclusions
A three-phase NEQ dynamic CD model was developed
and validated using experimental data obtained for the aldol
condensation of acetone in a batch CD column and a con-
tinuous pilot-plant CD column without the use of any ad-
justable parameters. The model takes account of multicom-
ponent mass and heat transfer between the vapour phase and
the liquid phase as well as between the liquid phase and cat-
alyst surface based on the MaxwellStefan equations. The
UNIQUAC model is used to consider the nonideality in the
liquid phase. The molar and energy holdups in the vapour
and liquid phases are also taken into account. Sensitivity
analysis of the mass transfer and kinetic parameters used in
the model indicates the production of DAA is mass trans-
fer controlled and the production of MO is kinetically con-
trolled. The generalized three-phase NEQ dynamic model
could also be used to make a priori predictions for other
CD processes using the mass and heat transfer rates and the
kinetics of a CD process.
Notation
a effective interfacial area, m
2
section
1
a
t
specic surface of packing, m
2
m
3
a
w
wetted area, m
2
m
3
c total number of components
C molar concentration, kmol m
3
C
p
heat capacity, kJ kmol
1
K
1
d
p
equivalent diameter of packing, m
D binary diffusion coefcient, m
2
s
1
e heat transfer rate, kJ s
1
f feed rate of component, kmol s
1
g acceleration of gravity, ms
2
G amount of catalyst, ml section
1
h heat transfer coefcient, kJ m
2
s
1
K
1
h
d
dynamic liquid holdup, m
3
m
3
H mole enthalpy, kJ kmol
1
k
1
DAA formation forward rate constant, kmol
1
m
6
(ml catalyst)
1
s
1
k
1
DAA formation reversible rate constant, m
3
(ml catalyst)
1
s
1
k
2
MO formation rate constant,
m
3
(mlcatalyst)
1
s
1
k multicomponent mass transfer coefcient, kmol
m
2
s
1
K equilibrium ratio
l liquid ow rate for components, kmol s
1
L liquid supercial velocity, m
3
m
2
h
1
Le Lewis number, (zMj
1
C
1
p
D
1
)
M section holdup, kmol
N mass transfer rate, kmol s
1
P Pressure, kPa
Q heat duty, kJ s
1
R macro kinetic rate, kmol (ml catalyst)
1
s
1
R
g
universal gas constant, 8.314 kJ kmol
1
K
1
S
c
Schmidt number, (j j
1
D
1
)
T absolute temperature, K
u vapour supercial velocity, ms
1
v vapour ow rate for components, kmol s
1
W mass ow rate, kg m
2
s
1
x liquid composition, mole fraction
y vapour composition, mole fraction
Greek letters
liquid phase activity coefcient
c thermodynamic factors
p dynamic viscosity, kg ms
1
binary mass transfer coefcient kmol m
2
s
1
j viscosity, Pa s
v stoichiometric coefcient
j density, kg m
3
o surface tension, Nm
1
Subscripts
av average value
i component index
j section index
k alternative component index
m property of mixture
Superscripts
I vapourliquid interface
L liquid phase
Y. Xu et al. / Chemical Engineering Science 60 (2005) 56375647 5647
LF liquid feed
r reaction
S catalyst or solid phase
V vapour phase
VF vapour feed
Acknowledgements
Financial support from Natural Sciences and Engineering
Research Council of Canada is gratefully acknowledged.
References
Abufares, A.A., Douglas, P.L., 1995. Mathematical modeling and
simulation of an MTBE catalytic distillation process using SpeedUp
and AspenPlus. Chemical Engineering Research and Design 73 (A1),
312.
Baur, R., Taylor, R., Krishna, R., 2000. Development of a dynamic
nonequilibrium cell model for reactive distillation tray columns.
Chemical Engineering Science 55 (24), 61396154.
Baur, R., Taylor, R., Krishna, R., 2001. Dynamic behavior of reactive
distillation tray columns described with a nonequilibrium cell model.
Chemical Engineering Science 56 (4), 17211729.
Billet, R., Schultes, M., 1991. Modeling of pressure drop in packed
columns. Chemical Engineering and Technology 14 (2), 8995.
Bravo, J.L., Fair, J.R., 1982. Generalized correlation for mass transfer
in packed distillation columns. Industrial and Engineering Chemistry
Process Design and Development 21 (1), 162170.
Danbert, T.E., Danner, R.P., 1989. Physical and Thermodynamics
Properties of Pure Chemicals: Data Compilation. Hemisphere
Publishing Corp., New York.
Huang, C., Yang, L., Ng, F.T.T., Rempel, G.L., 1998. Application
of catalytic distillation for the aldol condensation of acetone: a
rate-based model in simulating the catalytic distillation performance
under steady-state operations. Chemical Engineering Science 53 (19),
34893499.
Huang, C., Ng, F.T.T., Rempel, G.L., 2000. Application of catalytic
distillation for the aldol condensation of acetone: the effect of the
mass transfer and kinetic rates on the yield and selectivity. Chemical
Engineering Science 55 (23), 59195931.
King, C.J., 1980. Separation Process. McGraw-Hill, New York.
Kooijman, H.A., Taylor, R., 1995. A nonequilibrium model for dynamic
simulation of tray distillation columns. A.I.Ch.E. Journal 41 (8),
18521863.
Kreul, L.U., Gorak, A., Dittrich, C., Barton, P.I., 1998. Dynamic
catalytic distillation: advanced simulation and experimental validation.
Computers and Chemical Engineering 22 (Suppl., European
Symposium on Computer Aided Process Engineering8, 1998),
S371S378.
Kreul, L.U., Gorak, A., Barton, P.I., 1999. Dynamic rate-based model
for multicomponent batch distillation. A.I.Ch.E. Journal 45 (9),
19531962.
Ng, F.T.T., Rampel, G.L., 2003. Catalytic distillation. In: Horvth, J.T.
(Ed.), The Encyclopaedia of Catalysis. vol. 2. Wiley, Princeton, NJ.
pp. 477509.
Onda, K., Takeuchi, H., Okumoto, Y., 1968. Mass transfer coefcients
between gas and liquid phases in packed columns. Journal of Chemical
Engineering of Japan 1 (1), 5662.
Peng, J., Edgar, T.F., Eldridge, R.B., 2003. Dynamic rate-based and
equilibrium models for a packed reactive distillation column. Chemical
Engineering Science 58 (12), 26712680.
Podrebarac, G.G., Ng, F.T.T., Rempel, G.L., 1997a. More uses for catalytic
distillation. CHEMTECH 27 (5), 3745.
Podrebarac, G.G., Ng, F.T.T., Rempel, G.L., 1997b. A kinetic study of the
aldol condensation of acetone using an anion exchange resin catalyst.
Chemical Engineering Science 52 (17), 29913002.
Podrebarac, G.G., Ng, F.T.T., Rempel, G.L., 1998a. The production of
diacetone alcohol with catalytic distillation. Part I. Catalytic distillation
experiments. Chemical Engineering Science 53 (5), 10671075.
Podrebarac, G.G., Ng, F.T.T., Rempel, G.L., 1998b. The production of
diacetone alcohol with catalytic distillation. Part II. A rate-based
catalytic distillation model for the reaction zone. Chemical Engineering
Science 53 (5), 1077.
Prausnitz, J., Anderson, T., Grens, E., Eckert, C., Hsieh, R., Oconnell, J.,
1980. Computer Calculations for Multicomponent VaporLiquid and
LiquidLiquid Equilibria. Prentice-Hall, New Jersey.
Reid, R.C., Prausnitz, J.M., Poling, B.E., 1987. The Properties of Gases
and Liquids. McGraw Hill, New York.
Schneider, R., Noeres, C., Kreul, L.U., Gorak, A., 2001. Dynamic
modeling and simulation of reactive batch distillation. Computers and
Chemical Engineering 25 (1), 169176.
Smith Jr., L.A., 1984. Catalytic distillation structures. U.S. pat.#4443559.
Sneesby, M.G., Tade, M.O., Smith, T.N., 1998. Steady-state transitions
in the reactive distillation of MTBE. Computers and Chemical
Engineering 22 (78), 879892.
Taylor, R., Krishna, R., 1993. Multicomponent Mass Transfer. Wiley,
New York.
Xu, X., Zhao, Z., Tian, S., 1997. Study on catalytic distillation processes.
Part III. Prediction of pressure drop and holdup in catalyst bed.
Chemical Engineering Research and Design 75 (A6), 625629.
Zheng, Y., Xu, X., 1992. Study on catalytic distillation processes. Part
I. Mass transfer characteristics in catalyst bed within the column.
Chemical Engineering Research and Design 70 (A5), 459464.
Zheng, Y., Ng, F.T.T., Rempel, G.L., 2001. Catalytic distillation: a
three-phase nonequilibrium model for the simulation of the aldol
condensation of acetone. Industrial and Engineering Chemistry
Research 40 (23), 53425349.
Zheng, Y., Ng, F.T.T., Rempel, G.L., 2003a. Modeling of the catalytic
distillation process for the synthesis of ethyl cellosolve using a three-
phase nonequilibrium model. International Journal of Chemical Reactor
Engineering V1 (A4),
Zheng, Y., Ng, F.T.T., Rempel, G.L., 2003b. Process analysis for the
production of diacetone alcohol via catalytic distillation. Industrial and
Engineering Chemistry Research 42 (17), 39623972.
Zheng, Y., Ng, F.T.T., Rempel, G.L., 2004. A comparison of a
pseudo-homogeneous non-equilibrium model and a three-phase non-
equilibrium model for catalytic distillation. Chemical Engineering
Journal 100, 119127.