Chemical Engineering and Processing 92 (2015) 45–60

Contents lists available at ScienceDirect

Chemical Engineering and Processing:

Process Intensification
journal homepage: www.elsevier.com/locate/cep

Design and control of reactive dividing-wall column for the production

of methyl acetate
Dengchao An, Wangfeng Cai, Ming Xia, Xubin Zhang * , Fumin Wang
School of Chemical Engineering and Technology, Tianjin University, Tianjin 300072, China

A R T I C L E I N F O A B S T R A C T

Article history: A reactive dividing-wall column (RDWC) is developed for the production of methyl acetate (MeAc) in this
Received 14 April 2014 work. Both design and control of the RDWC are investigated by using commercial chemical simulator
Received in revised form 15 February 2015 Aspen Plus and Aspen Dynamics. The optimum RDWC design in terms of total annual costs (TAC) is
Accepted 25 March 2015
screened based on the proposed optimization procedure. The results show that about 7.7% savings in
Available online 28 March 2015
energy consumption, 8.3% reduction in operating cost and 15.5% decrease in capital investment can be
achieved by the RDWC design compared to the corresponding two-column design. Two control
Keywords:
structures (i.e., basic control structure and improved control structure) for the RDWC are presented. The
Control structures
Methyl acetate
results indicate that the improved control structure can handle disturbances in
20% feed flow rate and
Reactive dividing-wall column 5 wt% feed composition effectively.
Dynamic simulation ã 2015 Elsevier B.V. All rights reserved.

1. Introduction Reactive distillation provides an attractive alternative for

Methyl acetate (MeAc) is a widely-used solvent in the
production of cementing compound, coating materials, nitrocel-
lulose, cellulose acetate, cellulose ethers, and celluloid [1]. In
addition, MeAc as an intermediate is used in the production of a
variety of polyesters [2]. Traditionally, MeAc is synthesized
through esterification of acetic acid (HAc) and methanol (MeOH).
Rönnback et al. developed a kinetic model for the esterification of
acetic acid with methanol in the presence of a homogeneous acid
catalyst [3]. Tesser et al. studied the kinetics of oleic acid
esterification with methanol by using an acid ion-exchange
polymeric resin, and obtained a second-order pseudo-homoge-
neous model [4]. Altıokka and Çıtak studied the kinetics of the
esterification of acetic acid and isobutanol in a stirred batch reactor
with and without heterogeneous catalysts [5]. Peters et al.,
however, made a comparison between homogeneous catalysts
and heterogeneous catalysts, and the results indicated that
Amberlyst 15 is very effective for the esterification [6]. Traditional
strong acids are highly active catalysts in esterification and
transesterification reactions, but they usually suffer from corrosive
nature, uncontrollable side reactions and difficult recovery.
Heterogeneous catalysts can overcome the above-mentioned
defects, and are attractive alternatives to homogeneous catalysts
[7,8].
* Corresponding author. Tel.: +86 22 27890041.
E-mail address: tjzxb@tju.edu.cn (X. Zhang).
process intensification to achieve capital and energy costs
reduction. Traditional reactive distillation columns (RDCs) mainly
involve one-column or two-column process. The former process is
stabilized for neat operation by composition controls, which need
on-line composition analyzers that are usually costly and
introduce large lags. The latter is operated by temperature
inferential control for one reactant excess. For two columns
process, a reactant recovery column is needed, leading to a higher
operation and capital cost [9]. Kiss and Bildea reviewed the
industrial applications of DWCs and related research activities,
including column configuration, design, modeling and control
issues [10–12]. Asprion and Kaibel summarized the fundamentals
and advances of DWC [13]. Ramona and Megan presented a rapid
conceptual design of single-feed hybrid reactive distillation, and
the methodology was illustrated for MTBE production [14,15].
Generally, reactive distillation can achieve high conversion and
selectivity than traditional reaction and separation process [16].
DWCs usually can offer about 30% costs savings in capital and
energy compared to conventional configurations.
Reactive dividing-wall column (RDWC) integrates a reactor and
a separator in one DWC [17]. In such a column, reactive systems
with more than two products, non-reacting components or
excessive reagents can be separated. Kiss et al. studied the RDWC
for the production of FAME with methanol and fatty acids, and the
results indicated that 25% of energy savings could be realized
compared with the base case design [18]. Mueller et al. developed a
special two-step to decompose the RDWC into simple column
sequences [19,20]. Fig. 1(a) shows the integration scheme of an

http://dx.doi.org/10.1016/j.cep.2015.03.026
0255-2701/ ã 2015 Elsevier B.V. All rights reserved.
46 D. An et al. / Chemical Engineering and Processing 92 (2015) 45–60

RDWC consisting of two columns separated by a dividing wall

Nomenclature
(vertical wall). Fresh feed A and B are fed into the left of RDWC,
where A reacts with B generating C and D. The lightest component
C1 Reactive distillation column C is withdrawn from the top of the left part of the column and the
C2 Rectifying column heavier product D is obtained from the bottom of the column. D
CDC Conventional distillation column and the excess B are separated in the right part column, and as a
CRDC Conventional reactive distillation column result, almost pure B escapes from the top of the right column and
DWC Dividing-wall column then recycled back to the left part. It is noticed that the dividing
HAc Acetic acid wall top has no internal liquid split because it joins with the inner
k1 Forward rate constant side wall at a certain location. At the dividing wall bottom, the
k1 Backward rate constant vapor split ratio is determined by the cross-sectional area and
Keq Equilibrium constant for the esterification reaction hydraulics of each side of the wall. The scheme illustrated in
L1 Reflux flow rate in the reactive distillation column Fig. 1(b) is equivalent to Fig. 1(a), which is made up of a reactive
L2 Reflux flow rate in the rectifying column distillation column and a rectifying column. It is also the
mcat Catalyst-weight integration scheme applied in this study.
MeAc Methyl acetate Due to the strong interaction among control loops, control
MeOH Methanol problem for RDWC is a challenge. In 1995, Wolff and Skogestad first
NHAc Acetic acid feed location studied the control of a Petlyuk column for separating an ethanol–
NMeOH Methanol feed location propanol–butanol ternary system [21]. Serra et al. studied the
NR Rectifying stages for the reactive distillation column control of DWCs in a thorough way [22]. Ling and Luyben proposed
NRDC Net reactive distillation column a new control structure through manipulating liquid split bL to
NRX Reactive stages for the reactive distillation column realize minimized energy consumption for the separation of a
NS Stripping stages for the reactive distillation column benzene-toluene-o-xylene ternary mixture in a DWC [23,24]. Lee
NT2 Total stages for the rectifying column et al. studied the control of thermally coupled reactive distillation
NVR Withdrawal location of the vapor sidestream for the production of isopropyl acetate, and 23.14% energy savings
QR Reboiler heat input compared to the conventional process can be realized [25]. Abdul
r Reaction rate Mutalib and Smith investigated the operation and control of the
RCM Residue curve map dividing wall column, and presented two different control
RDWC Reactive dividing-wall column configuration using temperature control [26,27]. Mueller et al.
RGA Relative gain array [19,20] studied the transesterification of acetates process with
RR1 Reflux ratio in the reactive distillation column three types of configurations. They concluded that the RDWC
RR2 Reflux ratio in the rectifying column configuration could realize half of the total costs savings in contrast
SVD Singular value decomposition to others.
TAC Total annual cost So far, the study of reactive distillation in one divided wall
V Flow rate of the total rising vapor from the bottom column is scarce, especially for the control. The purpose of this
VR Vapor sidestream flow rate work is to combine the benefit of reactive distillation with DWC to
XB3 Mass fraction of water in the bottom of reactive produce MeAc, and then investigate the design and control of a
distillation column RDWC with special focus on the establishment of control
Y1 Mass fraction of methyl acetate in the vapor structures. First, qualitative relationship between process flow-
sidestream sheet and phase equilibria is established and then a RDWC
av Vapor split ratio flowsheet is established. Next, a systematic procedure is proposed
bL Liquid split ratio to obtain the optimum design on the basis of total annual cost
(TAC). Finally, two control structures for the RDWC are presented.

Fig. 1. (a) Integration scheme (b) equivalent scheme for a reactive dividing-wall column.
 D. An et al. / Chemical Engineering and Processing 92 (2015) 45–60 47

Table 1
Activity coefficient model parameters for MeAc esterification systems.

Comp. i HAc(1) HAc(1) HAc(1) EtOH(2) MeOH(2) MeAc(3)

Comp. j MeOH(2) MeAc(3) H2O(4) MeAc(3) H2O(4) H2O(4)
aij 0.97039 0.43637 0.051007 0.71011 3.1453 0.010143
aji 2.0346 1.1162 0.29355 0.72476 2.0585 0.96295
bij (K) 390.26 62.186 -422.38 62.972 575.68 593.7
bji (K) 65.245 81.848 98.12 326.2 219.04 265.83
cij (K1) 0.0030613 0.00027235 0.00024019 0.001167 0.0060713 0.0021609
cji (K1) 0.003157 0.0013309 0.000076741 0.0023547 0.0070149 0.00020133

Both design and control are carried out with Aspen Plus and Aspen
Dynamics (version 7.2).
Cond.1

2. Steady state design Cond.2

MeAc
HAC
2.1. Reaction kinetics and phase equilibria
RDC
Both reaction kinetics and phase equilibria are essential for the
design and optimization of a RDWC. MeOH
VR RC
2.1.1. Reaction kinetics
The esterification of acetic acid with MeOH can be expressed in
LR
the following form.
pump
Hac þ MeOH,MeAc þ H2 O
W
reb
In this study, the acidic ion-exchange resin Amberlyst 15 is used
as catalyst, which is simply placed inside the tray with a certain cooler
type of replacement mechanism.
Fig. 2. Equivalent flowsheet for a RDWC.
The reaction kinetics can be expressed according to Eq. (1) [16]:
r ¼ mcat ðk1 aHAc aMeOH  k1 aMeAc aH2 O Þ½kmol=kgcat S (1) Table 3
Utility prices.
where temperature-dependent k1 and k1 are given as

 Utility Price
49190
k1 ¼ 2:961  104exp (2) Low-pressure steam (600 kPa) 13.28 ($/GJ)
RT Cooling water 0.354 ($/GJ)
Catalyst (life of 3 months) 7.7162 ($/kg)


69230
k ¼ 1:348  106 exp
RT
In applying the reaction kinetics to a RDWC, it is assumed that
the solid catalyst occupies 50% of the tray holdup volume and a
catalyst density of 770 kg/m3 is used to conve the volume into
catalyst-weight (mcat) [16].
2.1.3. Phase equilibria
To account for nonideal vapor–liquid equilibrium (VLE) and
vapor–liquid–liquid equilibrium (VLLE) for the quaternary system,
the UNIQUAC model is used and the model parameters are listed in
Table 1 [28]. Since the designed RDWC is operated at atmospheric
pressure, the dimerization of acetic acid is considered for
describing vapor phase nonideality by the Hayden-O’Conell second
virial coefficient model. The built-in parameters in Aspen Plus are
used to compute fugacity coefficients.
(3)
Table 2 gives the azeotrope data for MeAc esterification system.
For azeotropic compositions and temperatures, the maximal
relative errors between the predicted values and experimental
data are less than 1.74% and 0.13%, respectively. Therefore, the
selection of the UNIQUAC model was reliable.
2.2. Process design
In this study, the RDWC concept is applied in the production of
MeAc. Fig. 2 shows the equivalent process flowsheet of the RDWC
consisting of a reactive distillation column and a rectifying column.
The two columns are linked with two interconnected vapor and
liquid streams with each other.
This process flowsheet can be understood from the boiling
point (including azeotropes) temperatures ranking. The light and
heavy products MeAc and H2O are obtained from the top and

Table 2
Azeotrope data for MeAc esterification system [29].

Azeotropes Calculated data Experimental data

Components MeOH/MeAc H2O/MeAc MeOH/MeAc H2O/MeAc

Mass basis 0.1810/0.8190 0.0233/0.9767 0.1950/0.8050 0.0320/0.9680
Temperature 326.75 K 329.79 K 327.15 K 329.55 K
Classification Unstable node Saddle Unstable node Saddle
Type Homogeneous Homogeneous Homogeneous Homogeneous
48 D. An et al. / Chemical Engineering and Processing 92 (2015) 45–60

(a) 1.00 (b) 0.06

VR=28

Y1:mass fraction of methyl acetate

0.98 0.05 VR=29
XB3:mass fraction of water
VR=30
VR=31
0.96 0.04
VR=28
VR=29
0.94 0.03
VR=30
VR=31
0.92 0.02

0.90 0.01

0.88 0.00
15 16 17 18 19 15 16 17 18 19
NVR:wthdraw location of sidestream NVR:withdraw location of sidestream

(c)1840
VR=28
VR=29
1820
QR: reboiler heat input(kw)

VR=30
VR=31
1800

1780

1760

1740
15 16 17 18 19
NVR:withdrawal location of sidestream

Fig. 3. Effects of VR (kmol/h) and NVR on (a) the H2O composition (xB3) in the bottoms of the reactive distillation column, (b) the MeAc composition (y1) in the vapor
sidestream, and (c) the reboiler heat input (QR).

bottom of the reactive distillation column respectively. A vapor noticeable that the HAc concentration of 2 ppm can be achieved
sidestream withdrawn from the reactive distillation column is at the bottoms under these specifications.
rectified in a rectifying column with an excess almost pure MeOH
escaped from the overhead. The overhead excess MeOH is recycled 2.3. Process optimization
back, mixed with the fresh MeOH and then fed to the reactive
distillation column. 2.3.1. Economic analysis
We assume that the reactive distillation column has a total of A number of parameters for a RDWC need to be optimized: the
25 stages, with the condenser being stage 1 and the reboiler being number of rectifying tray (NR), the number of stripping trays (NS),
stage 25. The rectifying column has a total of 14 stages with an the number of reactive trays (NRX) in the reactive distillation
assumed tray pressure drop of 0.7 kPa. The vapor sidestream is column (C1) and the number of stages (NT2) in the rectifying
initially withdrawn from stage 18. The condenser pressures of both column (C2). The reflux ratios (RR1, RR2) in the reactive distillation
columns are set at 100 kPa, which ensure reflux-drum temper- column (C1) and rectifying column (C2), feeding locations (NHAc,
atures (329.5 and 337.5 K, respectively) high enough to use cooling NMeOH) of the fresh feed, and withdrawal location and flow rate of
water as condensing agent. The fresh feed is pure HAc and MeOH the vapor sidestream (NVR,VR). They all should be varied to attain
with a flow rate of 50 kmol/h and 60 kmol/h, respectively. The the most economical design with the specified product yields and
working time is assumed to be 8000 h/y. purities.
The Design Specs/Vary function in Aspen Plus is utilized to TAC is usually used as the objective function to screen process
obtain the desired purities of products. For the reactive distillation alternatives [30,31]. A global economic optimization is carried out
column, the distillate flow rate is fixed at 54.1 kmol/h. The mass on the basis of the TAC that is expressed as [32]:
reflux ratio is assumed to be 1.8 and is varied to achieve the desired !
98 wt% MeAc using a Design Specs/Vary function. The vapor 103 $
TAC ¼C v þ C f þ ðir þ im Þ  FCI (4)
sidestream flow rate is assumed to be 29.2 kmol/h and is varied to y
achieve the desired 99 wt% H2O using another Design Specs/Vary
where Cv is the process variable costs, mostly utilities consumption
function. For the rectifying column, the mass reflux ratio is
(steam, cooling water, and catalyst); Cf is the annual fixed costs,
assumed to be 2.33 and is varied to obtain the desired 99 wt%
such as maintenance expense and wages; FCI is the fixed capital
MeOH using a Design Specs/Vary function. A FORTRAN subroutine
investment; ir is the fixed capital recovery rate applied to FCI; and
is embedded in Aspen Plus to calculate reaction rate. It is
im is the minimum acceptable rate of return on FCI. Cf is assumed to
 D. An et al. / Chemical Engineering and Processing 92 (2015) 45–60 49

Fig. 4. Global economic optimization procedure for MeAc system.

!
be 10% of FCI, and ir + im is assumed to be 20% of the FCI in this 103 $
TAC ¼ C v þ 0:3  FCI0 (6)
optimization; thus, the objective function can be rewritten as y
!
103 $
TAC þ C v þ 0:3  FCI (5)
Cv FCI0 = (installed column shell cost + installed column tray cost)
 1.2 + installed heat exchanger cost
Because the construction and installation of a RDWC is much more
difficult than a conventional distillation column (CDC), an assumed Major equipments for this process are the reactive distillation
20% penalty is added to the column [16]. Moreover, the reactive column vessel (including column internals) and heat exchangers
distillation column diameter is considered as the diameter of (condensers, reboiler, and cooler). Small items such as reflux
RDWC in the calculation because of the integration of the two drums, pumps, valves, and pipes are negligible because their costs
columns into one shell. Thus, the objective function is changed to are much lower than the column vessel and heat exchangers. Since
50 D. An et al. / Chemical Engineering and Processing 92 (2015) 45–60
Fig. 5. The optimization for (a) rectifying stages (NR) and stripping stages (NS), (b) reactive stages (NRX), (c) NHAc and NMeOH of the reactive distillation column (C1); (d) total
stages (NT2) of the rectifying column (C2).
the two columns have small diameters, we specify the tray spacing
as 0.5 m for the reactive distillation column and 0.45 m for the
rectifying column. The heat transfer area for the condensers and
reboilers is calculated using the overall heat transfer coefficient
and a differential temperature driving force. Here, the overall heat
transfer coefficients are assumed to be 0.852 kW/(K m2) and
0.568 kW/(K m2) for the condenser and reboiler respectively,
according to Luyben’s book [33]. All the major equipments costs
are estimated using a cost estimation program CAPCOST of Turton
[34].
It is worth noticing that the decision of the diameter of RDWC is
much more complexity than that of CDC. Tray holdup volume has
significant effect on reaction rate and it must be known before the
column diameter is known. As a result, the decision of diameter of
the RDWC is iterative. A tray holdup of reactive trays is assumed
and the column is designed to achieve the desired conversion. Then
the diameter of the column is calculated from maximum vapor-
loading correlations and the required weir height of reactive trays
is calculated. If the weir height is too large, a new tray holdup is
assumed and the design calculation is repeated. At last, a weir
height of 0.1 m of reactive trays is utilized in this study.
The utility costs mainly involve steam and cooling water. Low
pressure steam with a pressure of 600 kPa (433.1 K) and cooling
water with a temperature of 303.1 K are used in this process. The
utility prices taken from CAPCOST are listed in Table 3.The price
and life time of catalyst used here is proposed by Tang [16].
2.3.2. Sensitivity analysis
Because NVR and VR have significant effects on the composition
profiles and reboiler heat input for a RDWC, sensitivity analysis is
essential to seek the appropriate location of NVR and guess the
initial value of VR.
Fig. 3 illustrates the effects of VR and NVR on the H2O
composition in the bottom of the reactive distillation column,
the MeAc composition in the vapor sidestream, and the reboiler
heat input. Fig. 3(a) shows that, for a given VR, the H2O composition
in the bottom of reactive distillation column is nonmonotonic.
When NVR is on stage 17, the purity of H2O has a minimum value
and increases with decreasing NVR. With the increase of NVR the
purity of H2O reaches a maximum value on stage 18, and then it
begins to decrease. Fig. 3(b) shows that, for a given VR, the MeAc
composition in the vapor sidestream decreases gradually with
increasing NVR and from stage 17 the MeAc composition tend to
zero. Fig. 3(c) reveals that the relationship between QR and VR plus
NVR. It indicates that for a given VR the reboiler duty has a
maximum value on stage 17, and it decreases to a minimum value
with stage increasing to18 then remain stable. Thus, the optimum
withdrawal location and flow rate of vapor sidestream are on stage
18 and 28 kmol/h, respectively.
2.3.3. Economic optimization
Because so many design variables need to be optimized, a
systematic calculation sequence is established to facilitate the
 D. An et al. / Chemical Engineering and Processing 92 (2015) 45–60 51

Fig. 6. Optimum process flow diagram for the reactive dividing-wall column.

optimization. For the reactive distillation column, a sequential
iterative optimization search is used to optimize design, with NRX
as the outer iterative loop, the NR, NS as middle loop, and NHAc,
NMeAc and NVR as the inner iterative loop. As for the rectifying
column, only NT2 needs to optimize as an iterative loop and such a
sequential iterative optimization procedure is clearly demonstrat-
ed in Fig. 4.
The optimization procedure is summarized below.
I Optimization for reactive distillation column (C1):

Fig. 7. (a) Reactive distillation column (C1) composition profile. (b) Rectifying column (C2) composition profile.
52 D. An et al. / Chemical Engineering and Processing 92 (2015) 45–60

(a) 360
(b)
380 Temperature K Temperature K

355
370

Temperature of C2
Temperature of C1

360 350

350
345

340
340
330

4 8 12 16 20 335
2 4 6 8 10 12
Stage of C1 Stage of C2

Fig. 8. (a) Reactive distillation column (C1) temperature profile. (b) Rectifying column (C2) temperature.

(1) Fix two columns top pressure (P1, P2) at100 kPa.
(2) Guess the total stages of rectifying column (NT2).
(3) Guess the reactive stages (NRX) of reactive distillation column.
(4) Guess the rectifying stages (NR) and striping stages (NS) of
reactive distillation column.
(5) Guess the HAc feeding location (NHAc), the MeOH feeding
location (NMeOH), and the sidestream withdrawal location
(NVR).
(6) Change the reactive distillation column reflux ratio (RR1), the
rectifying column reflux ratio (RR2) and the withdrawal rate of
sidestream until the three design specifications can be met.
(7) Go back to step 5 and change NHAc, NMeOH and NVR until QR is
minimized.
(8) Go back to step 4 and change NR, NS until TAC is minimized.
(9) Go back to step 3 and change NRX until TAC is minimized.
II Optimization for rectifying column (C2):
(1) Fix the NRX, NR, NS of reactive distillation column (C1) that has
been optimized.
(2) Change the total stages of rectifying column (NT2).
(3) Vary the reflux ratio (RR2) until the three design specifications
can be met.
(4) Go back to step 2 and change NT2 until the TAC is minimized.
(5) Iteration is over and the optimum design is obtained.
Fig. 5 shows the optimization for reactive stages (NRX),
rectifying stages (NR), striping stages (NS), HAc feeding location
(NHAc), MeOH feeding location of reactive distillation column (C1)
and the total stages of rectifying column (NT2). Fig. 5(a) shows the
variation of TAC with the increase of NS, and there is a vivid
comparison of TAC in three different cases with the NR = 2, 4, and 6,
respectively. Obviously, the optimum values for NS and NR are 4 and
2, respectively. Fig. 5(b) shows that when the NR, NS fixed to 2 and 4
respectively, TAC has a minimum value with NRX = 12. With the
increasing of NRX above 12, the TAC increase due to capital cost
increasing; however with the decreasing of NRX under12, the TAC
also increase due to operation cost increasing mainly the reboiler
heat input augment with the increasing of RR1. Fig. 5(c) shows the
influence of NHAc and NMeOH on the TAC. The best location of NHAc
was decided first, and then the appropriate location of NMeOH was
found. As can be seen the TAC is monotonic decreasing with the
increase of NMeOH. Because the upper of the NMeOH the more
unreactive MeOH is recycled back to react with HAc in the reactive
section of the reactive distillation column, which need more
vapors. Fig. 5(d) shows the effect of NT2 on the TAC. Above 10 the
TAC reach to a plateau, however it increase rapidly below 10. Thus,
the optimum RDWC design is obtained after the optimization is
terminated.
Fig. 6 gives the final optimal flowsheet for this system, with
detailed steam information, heat duties, equipment sizes, and
operating conditions at the steady-state design conditions. Fig. 7

Table 4
Comparison among the optimum RDWC design, one-column design and two-column design.

One-column design RDWC design Two-column design

Total stages
RDC/C1 20 20 20
RC/C2 – 12 13

Reflux ratios
RR1 1.7 1.8 1.8
RR2 – 2.33 3
Column diameters (m) 0.98 1 1/0.43
Tray space (m) 0.5 0.5/0.45 0.5/0.45
QR, total (kw) 1565.6 1754.6 1901.1

Economic data (103$/y)

FCI 610 750.36 888.3
CV 579.06 713.11 777.86
TAC 762.04 938.22 1044.35
 D. An et al. / Chemical Engineering and Processing 92 (2015) 45–60 53

shows the composition profiles in the reactive distillation column
and rectifying column. Fig. 8 displays the temperature profiles in
the two columns of the RDWC.
Detailed economic comparisons among the optimum RDWC
design, one-column design and two-column design are listed in
Table 4. It is calculated that about 7.7% savings in energy
consumption, 8.3% reduction in operating cost and 15.5% decrease
in capital investment can be achieved by the RDWC design
compared to the corresponding two-column design. Obviously, the
one-column design can bring some advantages in energy and cost
savings among the three designs. However, as will be shown later,
the neat operation has a poor dynamic performance, and from the
viewpoints of economic and controllability, the RDWC design will
be more attractive than the two other designs.
3. Control system design
The control objective is to achieve stable on-specification
operation when encountering disturbances in feed flow rate and
feed composition. The product purity are controlled by manipu-
lating the reboiler heat input (QR), vapor split ratio (av) and reflux
rate of rectifying column (R2). Because the control of MeOH purity
is fairly easy, we mainly demonstrate the purity variation of MeAc
and H2O when disturbances occur.
3.1. Implementation in aspen dynamics
The steady-state Radfrac model in Aspen Plus consists of two
columns: one reactive distillation column and one rectifying

Fig. 9. (a) Singular value decomposition analysis for reactive distillation column. (b) Singular value decomposition analysis for rectifying column.
54 D. An et al. / Chemical Engineering and Processing 92 (2015) 45–60
Fig. 10. Basic control structure for the RDWC.
column. In reality, there is only one column, but these two
“fictitious” vessels are used in the simulation to model the real
physical equipment. Before the steady-state file is exported into
Aspen Dynamics, the plumbing system and major equipment sizes
have to be specified to obtain a pressure-driven dynamic
simulation. In addition, three “fictitious” items are inserted
between the two columns. A fictitious value is added to the vapor
sidestream outlet and a fictitious valve and pump are inserted in
the bottom liquid stream to compensate the pressure drop. Reflux
drums and column bases are sized to provide about 5 min of liquid
holdup when half full. Most valves are specified to provide
pressure drops of about 304 kPa with the valve half open at the
design flow rate.
3.2. Interaction between design and control
The control structure for the optimum RDWC design is
established and the control effectiveness is evaluated by feed
flow rate and feed composition disturbances. However, the
optimum design does not generate a good controllability, which is
probably due to the strong interaction between the two columns,
indicating that a tradeoff between design and control should be
made. The vapor sidestream withdrawal rate (VR) has a great
effect on the purities of product. Therefore, an overdesign (VR
changes from 28 kmol/h to 29.2 kmol/h) is taken. At last, a
suboptimum process is developed and the controllability is
studied in later section. This design is called a modified RDWC
design.
3.3. Selection of temperature control trays
As can be seen from Fig. 8(a), the temperature profile in reactive
distillation column is complex, and we use sensitivity analysis and
singular value decomposition (SVD) criteria for selecting the
control trays [33]. For sensitivity analysis, a very small change
(0.1%) is made in one of the two variables (QR and av), while
keeping the other inputs fixed. Then the Aspen Plus simulation is
run and the new temperatures on all the trays are determined. The
change in each tray temperature divided by the change in the
manipulated variable gives the steady-state gain for that tray.
There are two gain matrices, one for the reactive distillation
column and the other for the rectifying column. The reactive
distillation column matrix has 20 rows and two columns. The
rectifying column matrix has 12 rows and one column. Each matrix
is decomposed using the SVD program in Matlab into three
matrices: K = USVT. The three U vectors are plotted against tray
number. Fig. 9(a) shows results of the reactive distillation column.
The steady-state gains for the heat input QR and the vapor split
ratio av are given in the left graph, which indicate peaks at trays 11,
17 and 18. The right graph gives two U vectors that indicate the
sensitive trays in the column are 11, 17 and 18. From the
consideration of the dynamic relationships between the location
of the sensitive trays and the manipulated variables, the temper-
atures of the sensitive trays 11 and 18 are controlled by
manipulating av and QR, respectively. Fig. 9(b) shows the results
of the rectifying column. The steady-state gain for the reflux rate is
given in the left graph and the SVD result is shown in the right
graph, and both of them has the sensitive tray 11.
 D. An et al. / Chemical Engineering and Processing 92 (2015) 45–60 55

0.9820 1.000
+10% feed flow +10% feed flow
- 10% feed flow - 10% feed flow
XD1:mass fraction of MeAc

XB3: mass fraction of H2O

0.9815
0.995

0.9810
0.990

0.9805

0.985
0.9800

0.9795 0.980
0 2 4 6 0 2 4 6
time (h) time (h)

360 380
+10% feed flow +10% feed flow
- 10% feed flow - 10% feed flow

355 375
T1,11 (K)

T1,18 (K)
350 370

345 365

340 360
0 2 4 6 0 2 4 6
time (h) time (h)
Fig. 11. Dynamic responses for basic control structure: 10% feed flow disturbances.

3.4. Control structure (10) The vapor sidestream molar flow is flow controlled (reverse
acting).
3.4.1. Basic control structure (11) The temperature for tray 11 in the rectifying column is
After numerous control structures were tested by trial and controlled by manipulating the reflux rate (direct acting).
error, a basic control structure for the RDWC is proposed in this
section. Fig. 10 illustrates the basic control structure for the RDWC. Conventional PI controllers are used for all controllers except
the four liquid level controllers. Proportional controllers are used
(1) Fresh feed is flow controlled (reverse acting). for all liquid levels with Kc = 2. The proportional and integral (PI)
(2) Reflux drum levels in both columns are held by manipulating settings of the top pressure control loops for both columns are set
the flow of distillates (direct acting). at Kc = 20 and t I = 12 min. A 1 min dead-time element is inserted
(3) Base levels in both columns are held by manipulating the flow into each temperature control loop. Relay-feedback tests are run
of the bottoms (direct acting). on the temperature controllers to determine ultimate gains and
(4) The pressure in the two columns is controlled by manipulating periods, and Tyreus–Luyben tuning is used to attain the gain Kc
the heat removal rate in the condenser of the two columns and integral time constant t I. Table 5 gives the tuning parameters
(reverse acting). for the three temperature controllers.
(5) The total recycle MeOH flow is in proportion to the acetate Dynamic performance of the basic control structure is
acid feed flow. evaluated by feed flow rate and composition disturbances.
(6) The recycle MeOH temperature is controlled by manipulating Fig. 11 shows the dynamic responses of this structure for both
the cooler heat removal (reverse acting). an increase and decrease of 10% in the feed flow rate at t = 0.5 h. As
(7) The temperature for tray 18 in the reactive distillation column can be seen, product purities are held fairly close to their desired
is controlled by manipulating the reboiler heat input (reverse specifications at new steady state. The two controlled tray
acting). temperatures are brought back to their set points in about 2.5 h.
(8) The temperature for tray 11 in the reactive distillation column Fig. 12 shows the dynamic responses for feed composition
is controlled by manipulating the av (VR/QR) (direct acting). disturbances at t = 0.5 h. For a decrease of MeOH concentration
(9) The reflux rate in reactive distillation columns is in proportion from 100% to 95%, stable control can be achieved with controlled
to fresh feed. tray temperature returned to its setpoint, and the product purities
56 D. An et al. / Chemical Engineering and Processing 92 (2015) 45–60
0.984
XD1:mass fraction of MeAc
0.980
0.976
0.972
0.968
0 2 4 6
time (h)
358
357
T1,11 (K)
356
355
354
353
0 2 4 6
time (h)
Fig. 12. Dynamic responses for basic control structure: feed composition disturbances.
are held fairly close to their desired specifications after some
oscillation. However, when subjected to a decrease of 5% in HAc
concentration, a somewhat large decrease (about 1 wt%) from the
specified product purity occurs. As a result, when operating RDWC
with this basic control structure, it is necessary to ensure the water
content in HAc feed below a certain value.
However, for an increase of 20% in the feed flow rate, the Aspen
Dynamics integrator will fail to work and then the simulation will
terminate. The reason for this strange behavior may be due to the
high interaction of the two-column system. The temperatures of
both controllers (TC1 and TC3) decrease, and the variation of
controller (TC3) is more tremendous. Thus the withdrawal rate
of vapor sidestream appears a sharp drop and the increase of heat
input of reboiler is slow, which lead to the integrator failing to
work.
3.4.2. An improved control structure
For an increase of 20% in the feed flow rate, the Aspen Dynamics
integrator will fail to work and then the simulation will terminate.
This is because the reboiler heat input cannot increase immedi-
ately once the feed flow rate increases suddenly. To overcome this
defect, a feed-forward controller QR/F ratio is added to the control
system to offer an immediate change in reboiler heat input when
the feed flow rate increases. This is called an improved control
structure. Fig. 13 illustrates this structure for the RDWC. Note that
the ratio controller is on cascade, so the temperature controller
needs to be returned. The ratio strategy reduces the interaction
between the two columns to some extent.
0.9928
95 wt% HAc 95 wt% HAc
95 wt% MeOH 95 wt% MeOH
XB3:mass fraction of H2O
0.9920
0.9912
0.9904
0.9896
0.9888
0 2 4 6
time (h)
374
95 wt% HAc 95 wt% HAc
95 wt% MeOH 95 wt% MeOH
372
T1,18 (K)
370
368
366
0 2 4 6
time (h)
Fig. 14 gives the dynamic responses for the improved control
structure to 20% feed flow rate disturbance. The results show that
two product purities can maintain very close to their desired
specifications at the new steady state. Also, the two tray
temperature control loops perform well in bringing the temper-
atures back to their setpoint. What is more, the settling time is
shorter than that of the basic control structure.
Fig. 15 gives the dynamic responses of the improved control
structure to feed composition disturbances. As can be seen, the
disturbances can be effectively handled with small offsets except
for a decrease of 5% in HAc concentration resulting in a relatively
large offset (about 1 wt%) in product purity.
3.4.3. Comparison of dynamic performance of RDWC design, one-
column design and two-column design.
The dynamic performances of the one-column design, two-
column design and RDWC design are compared, and the results are
shown in Figs. 16 and 17. As can be seen from Fig. 16, for a + 20% feed
flow rate disturbance, all three designs can maintain the two
product purities very close to their desired specifications at the
new steady state. However, for a 20% decrease in feed flow rate,
the dynamic performance of one-column design is unsatisfactory.
The product purities of MeAc and H2O are, respectively, 95.5 wt%
and 85 wt%, which are far away from their specifications. Fig. 17
gives the dynamics responses for the feed compositions dis-
turbances. For MeOH composition disturbance from 100 wt% to
95 wt%, both RDWC design and two-column design can maintain
the two product purities very close to their specifications at the
new steady state. However, for a decrease of HAc composition from
 D. An et al. / Chemical Engineering and Processing 92 (2015) 45–60 57

Fig. 13. Improved control structure for the RDWC.

0.9812 0.994
XD1:mass fraction of MeAc

+20% feed flow

XB3:mass fracton of H2O

0.992 +20% feed flow

- 20% feed flow
0.9808 - 20% feed flow

0.990

0.9804
0.988

0.9800
0.986

0.9796 0.984
0 2 4 6 0 2 4 6
time (h) time (h)

365 380
+20% feed flow +20% feed flow
- 20% feed flow -20% feed flow
360 376
T1,11 (K)

T1,18 (K)

355 372

350 368

345 364

340 360
0 2 4 6 0 2 4 6
time (h) time (h)

Fig. 14. Dynamic responses for improved control structure: 20% feed flow rate disturbances.
58 D. An et al. / Chemical Engineering and Processing 92 (2015) 45–60

Fig. 15. Dynamic responses for the improved control structure: feed composition disturbances.

0.9824
1.000

RDWC +20% feed flow

XB3:mass fraction of H2O
XD1:mass fraction of MeAc

0.9816 NRDC 0.975

+20% feed flow
TCRD
RDWC
NRDC
0.9808 0.950 TCRD

0.9800 0.925

0.9792 0.900
0 2 4 6 0 2 4 6
time (h) time (h)

0.99 1.00
XD1:mass fraction of MeAc

0.98 0.95 RDWC

XB3:mass fraction of H2O

-20% feed flow NRDC

RDWC TCRD
0.97 -20% feed flow NRDC 0.90
TCRD

0.96 0.85

0.95 0.80

0.94 0.75
0 2 4 6 0 2 4 6
time (h) time (h)

Fig. 16. Comparison of dynamic performance among the RDWC, NRDC and TCRD design for feed flowrate disturbances.
 D. An et al. / Chemical Engineering and Processing 92 (2015) 45–60 59

0.9825 1.00
RDWC
95 wt% HAc NRDC
0.99
XD1:mass fraction of MeAc
TCRD

XB3:mass fraction of H2O

0.9750
0.98
RDWC
NRDC
0.9675 0.97 95 wt% HAc TCRD

0.96
0.9600
0.95

0.9525 0.94
0 2 4 6 0 2 4 6
time (h) time (h)

0.982 0.996
RDWC
95 wt% MeOH NRDC
0.992
XD1:mass fraction of MeAc

XB3:mass fraction of H2O

TCRD
0.981

0.988
0.980
RDWC
0.984 95 wt% MeOH NRDC
TCRD
0.979
0.980

0.978 0.976
0 2 4 6 0 2 4 6
time (h) time (h)
Fig. 17. Comparison of dynamic performance among the RDWC, NRDC and TCRD design for feed composition disturbances.

100 wt% to 95 wt%, all the three design have a relatively poor
response, with the one-column design the worst result.
4. Discussion and conclusions
Design and control of a RDWC for producing MeAc are
investigated in this study and an optimal design in terms of the
TAC is presented. Two control structures with three temperature
controllers and an adjustable vapor split ratio aV are proposed for
the RDWC design. It is found that the basic control structure can
handle all disturbances well, except for a 10% increase in feed flow
rate. Furthermore, an improved control structure with QR/F ratio is
explored to get a more robust control. The dynamics simulation
results demonstrate that the improved control structure can
maintain high purity of the two products for both an increase and a
decrease of 20% in the feed flow rate disturbances.
In further work, we will explore a HAc excess process to attain a
higher purity MeAc, with special focus on the establishment of
control structure for the novel process.
Acknowledgment
The authors are thankful for support by the National Basic
Research Program of China (973 Program) (Grant no:
2012CB720302).
References
[1] W. Yu, K. Hidajat, A.K. Ray, Determination of adsorption and kinetic parameters
for methyl acetate esterification and hydrolysis reaction catalyzed by
Amberlyst 15, Appl. Catal. A 260 (2004) 191–205.
[2] S. Giessler, R.Y. Danilov, R.Y. Pisarenko, L.A. Serafimov, S. Hasebe, I. Hashimoto,
Systematic structure generation for reactive distillation processes, Comput.
Chem. Eng. 25 (2001) 49–60.
[3] R. Rönnback, T. Salmi, A. Vuori, H. Haario, J. Lehtonen, A. Sundqvist, E. Tirronen,
Development of a kinetic model for the esterification of acetic acid with
methanol in the presence of a homogeneous acid catalyst, Chem. Eng. Sci. 52
(1997) 3369–3381.
[4] R. Tesser, M. Di Serio, M. Guida, M. Nastasi, E. Santacesaria, Kinetics of oleic
acid esterification with methanol in the presence of triglycerides, Ind. Eng.
Chem. Res. 44 (2005) 7978–7982.
[5] M.R.Altıokka,A.Çıtak,Kineticsstudyofesterificationofaceticacidwithisobutanol
in the presence of amberlite catalyst, Appl. Catal. A 239 (2003) 141–148.
[6] T.A. Peters, N.E. Benes, A. Holmen, J.T.F. Keurentjes, Comparison of commercial
solid acid catalysts for the esterification of acetic acid with butanol, Appl. Catal.
A 297 (2006) 182–188.
[7] M.A. Harmer, Q. Sun, Solid acid catalysis using ion-exchange resins, Appl. Catal.
A 221 (2001) 45–62.
[8] G.D. Yadav, H.B. Kulkarni, Ion-exchange resin catalysis in the synthesis of
isopropyl lactate, React. Funct. Polym. 44 (2000) 153–165.
[9] W.L. Luyben, Economic and dynamic impact of the use of excess reactant in
reactive distillation systems, Ind. Eng. Chem. Res. 39 (2000) 2935–2946.
[10] A.A. Kiss, C.S. Bildea, A control perspective on process intensification in
dividing-wall columns, Chem. Eng. Process. 50 (2011) 281–292.
[11] Ö. Yildirim, A.A. Kiss, E.Y. Kenig, Dividing wall columns in chemical process
industry: a review on current activities, Sep. Purif. Technol. 80 (2011) 403–417.
[12] I. Dejanovi c, Matijaševic, L.Ž. Oluji
c, Dividing wall column—a breakthrough
towards sustainable distilling, Chem. Eng. Process. 49 (2010) 559–580.
[13] N. Asprion, G. Kaibel, Dividing wall columns: fundamentals and recent
advances, Chem. Eng. Process. 49 (2010) 139–146.
[14] M.D. Ramona, J. Megan, Conceptual design of single-feed hybrid reactive
distillation columns, Chem. Eng. Sci. 60 (2005) 4377–4395.
[15] M.D. Ramona, J. Megan, Conceptual design of single-feed kinetically controlled
reactive distillation columns, Chem. Eng. Sci. 60 (2005) 5049–5068.
[16] Y.-T. Tang, Y.-W. Chen, H.-P. Huang, C.-C. Yu, S.-B. Hung, M.-J. Lee, Design of
reactive distillations for acetic acid esterification, AIChE J. 51 (2005)
1683–1699.
[17] S. Hernández, R. Sandoval-Vergara, F.O. Barroso-Muñoz, R. Murrieta-Dueñas,
H. Hernández-Escoto, J.G. Segovia-Hernández, V. Rico-Ramirez, Reactive
60 D. An et al. / Chemical Engineering and Processing 92 (2015) 45–60
dividing wall distillation columns: simulation and implementation in a pilot
plant, Chem. Eng. Process. 48 (2009) 250–258.
[18] A.A. Kiss, J.G. Segovia-Hernández, C.S. Bildea, E.Y. Miranda-Galindo, S.
Hernández, Reactive DWC leading the way to FAME and fortune, Fuel 95
(2012) 352–359.
[19] I. Mueller, E.Y. Kenig, Reactive distillation in a dividing wall column: rate-based
modeling and simulation, Ind. Eng. Chem. Res. 46 (2007) 3709–3719.
[20] I. Mueller, C. Pech, D. Bhatia, E.Y. Kenig, Rate-based analysis of reactive
distillation sequences with different degrees of integration, Chem. Eng. Sci. 62
(2007) 7327–7335.
[21] E.A. Wolff, S. Skogestad, Operation of integrated three-product (petlyuk)
distillation columns, Ind. Eng. Chem. Res. 34 (1995) 2094–2103.
[22] M. Serra, A. Espuña, L. Puigjaner, Control and optimization of the divided wall
column, Chem. Eng. Process. 38 (1999) 549–562.
[23] H. Ling, W.L. Luyben, New control structure for divided-wall columns, Ind. Eng.
Chem. Res. 48 (2009) 6034–6049.
[24] H. Ling, W.L. Luyben, Temperature control of the BTX divided-wall column, Ind.
Eng. Chem. Res. 49 (2009) 189–203.
[25] H.-Y. Lee, I.K. Lai, H.-P. Huang, I.L. Chien, Design and control of thermally
coupled reactive distillation for the production of isopropyl acetate, Ind. Eng.
Chem. Res. 51 (2012) 11753–11763.
[26] M.I. Abdul Mutalib, R. Smith, Operation and control of dividing wall distillation
columns. Part 1: degrees of freedom anddynamic simulation, Int. Chem. Eng.
76 (1998) 308–318.
[27] M.I. Abdul Mutalib, R. Smith, Operation and control of dividing wall distillation
columns. Part 2: simulation and pilot plant studies using temperature control,
Int. Chem. Eng. 76 (1998) 319–334.
[28] T. Pöpken, L. Götze, J. Gmehling, Reaction kinetics and chemical equilibrium of
homogeneously and heterogeneously catalyzed acetic acid esterification with
methanol and methyl acetate hydrolysis, Ind. Eng. Chem. Res. 39 (2000)
2601–2611.
[29] L.H. Horsley, Azeotropic Data – III. Advances in Chemistry Series, vol. 116,
American Chemical Society, Washington, D. C, 1973.
[30] Q. Wang, B. Yu, C. Xu, Design and control of distillation system for methylal/
methanol separation. Part 1: extractive distillation using DMF as an entrainer,
Ind. Eng. Chem. Res. 51 (2011) 1281–1292.
[31] B. Yu, Q.C. Wang, Xu, Design and control of distillation system for methylal/
methanol separation. Part 2: pressure swing distillation with full heat
integration, Ind. Eng. Chem. Res. 51 (2011) 1293–1310.
[32] Muñoz, J.B. Montón, M.C. Burguet, J. de la Torre, Separation of isobutyl alcohol
and isobutyl acetate by extractive distillation and pressure-swing distillation:
simulation and optimization, Sep. Purif. Technol. 50 (2006) 175–183.
[33] W.L. Luyben, Distillation Design and Control Using Aspen Simulation, John
Wiley & Sons, New York, 2006.
[34] R. Turton, R.C. Bailie, W.B. Whiting, J.A. Shaeiwitz, Analysis, Synthesis and
Design of Chemical Processes, Prentice Hall, New York, 2009.