Ind. Eng. Chem. Res.

2010, 49, 735–749 735

Design and Control of Dimethyl Carbonate-Methanol Separation via Extractive

Distillation in the Dimethyl Carbonate Reactive-Distillation Process
Kai-Yi Hsu, Yuan-Chang Hsiao, and I-Lung Chien*
Department of Chemical Engineering, National Taiwan UniVersity of Science and Technology, Taipei 10607, Taiwan

Dimethyl carbonate is an environmentally benign and biodegradable chemical. It can be produced by a

transesterification reaction of methanol with ethylene carbonate, coproducing another useful product, ethylene
glycol. A reactive distillation column can be utilized for the complete conversion of ethylene carbonate with
methanol in excess. The coproduct, ethylene glycol, is the bottom product while the top product of this reactive
distillation column is a mixture of dimethyl carbonate and methanol close to the azeotropic composition. In
this paper, an economical separation process via extractive distillation is proposed to obtain pure dimethyl
carbonate product and also pure methanol to be recycled to the reactive distillation column. A very simple
procedure is proposed in this paper for the quick comparison of alternative entrainer candidates before rigorous
See https://pubs.acs.org/sharingguidelines for options on how to legitimately share published articles.

process simulation is conducted. Aniline is found to be a very effective entrainer to enhance the relative
volatility between methanol and dimethyl carbonate. A problem with small heavy-boiler impurities in the
feed stream of the extractive distillation process is also pointed out, with a practical solution given for this
Downloaded via UNIV OF LIMERICK on November 12, 2020 at 23:23:27 (UTC).

problem. Simple overall control strategy of this process is also proposed to maintain product purity despite
various feed disturbances. Only one tray temperature control loop is required for each of the columns.

1. Introduction into the system. The entrainer acts to enhance the relative
volatility of the original two components so that one original
Dimethyl carbonate (DMC) is an environmentally benign and component can go overhead in the extractive distillation column
biodegradable chemical. It is a nontoxic substitute for dimethyl and the other component will go with the heavy entrainer to
sulfate or methyl halide as a methylating agent or a substitute the bottom of the column. A second entrainer recovery column
for phosgene as a carbonylating agent.1 This useful chemical is used to purify entrainer, which is recycled back to the
can be produced by a transesterification reaction of methanol extractive distillation column.
with ethylene carbonate (EC), coproducing another useful Entrainer screening is an important step in extractive distil-
product, ethylene glycol (EG).2 In the experimental study by lation before the distillation sequence is designed. Several rules
Fang and Xiao,2 this reaction is reversible with an EC have been proposed to obtain a feasible entrainer candidate for
equilibrium conversion of about 80%. A reactive distillation the azeotropic mixtures.9–12 A comprehensive paper by Gme-
column can be utilized for the complete conversion of EC with hling and Mollmann13 used several examples to show their
methanol in excess. However, the top product of the reactive procedure for the entrainer selection. A recent nice paper by
distillation column is the azeotrope of methanol and DMC Kossack et al.14 used acetone-methanol separation as an
because the azeotropic temperature is the lowest of the system. example to show a systematic synthesis framework for extractive
Further separation of this azeotropic mixture is needed to distillation processes. The effect of solvent on controllability
complete the design flow sheet of the overall process to in acetone-methanol separation was further studied by Luy-
coproduce DMC and EG. ben.15
Conventional methods for the separation of azeotrope without In this paper, design and control of the overall process for
adding a third component are two-fold. One method is for the the coproduction of DMC and EG is studied. Particular focus
cases where the azeotrope is heterogeneous, which means after will be on the separation of methanol-DMC mixture via
cool down it can naturally be separated into two liquid phases. extractive distillation. In Wang et al.,8 no comparisons of
With this unique characteristic, a simple two-strippers/decanter candidate entrainers have been made; only phenol is shown as
configuration can be used to separate the azeotrope.3–5 a feasible entrainer. In this paper, a simple procedure is proposed
Methanol-DMC azeotrope does not exhibit this behavior; thus, for quick comparison of candidate entrainers. Aniline is found
this separation method cannot be utilized for this system. to be a much more effective entrainer for the methanol-DMC
Another method is called pressure-swing distillation. This is separation. A problem with heavy impurities in the feed stream
used for some azeotropic systems that exhibit the desirable of the extractive distillation system will also be pointed out,
property of operating pressure having a strong effect on the with a proposed practical solution of this problem. The design
composition of the azeotrope. When this occurs, a two-column flow sheet with two recycled streams and the overall control
system can be used to achieve the desired separation.6,7 Wang strategy of the process will be developed to meet product
et al.8 studied pressure-swing distillation for the methanol-DMC specifications.
azeotrope and found out that this separation method is not more
economical than extractive distillation. 2. Design of Reactive Distillation Column
Extractive distillation is another way to separate an azeotrope.
In this method, a third component (heavy entrainer) is added The reactive distillation column for this system will be
designed first. The reversible reaction is as follows:
* To whom correspondence should be addressed. Tel: 886-2- The kinetics equation of this reaction is from Fang and Xiao2
27376652. Fax: 886-2-27376644. E-mail: [email protected]. and can be expressed as follows:
10.1021/ie901157g  2010 American Chemical Society
Published on Web 12/04/2009
736 Ind. Eng. Chem. Res., Vol. 49, No. 2, 2010

homogeneous catalyst was assumed to feed into the column with

EC feed, thus the column stages below the EC feed location
are all reactive stages. The design variables of this system
include MeOH/EC feed ratio, total number of stages, number
of reactive stages, and MeOH feed location. Iterative optimiza-
CEGCDMC tion procedure was followed to obtain the best values of the
rEC ) k+CECCMeOH - k- (2-1) above four design variables. For each simulation run, the reactive
CMeOH
tray holdup has to be iteratively obtained to agree with the
k+ ) 1.3246 exp ( -13RT060 ) (2-2)
column sizing calculation. The two degrees of freedom of this
RD column are fixed to obtain EG product at high purity of
99.5 mol % and EC conversion at 99.95%.
-28 600
) 15 022 exp(
RT )
k- (2-3) Because the optimization of this RD column is not the focus
of this paper, we show only a resulting flow sheet of this RD
where rEC is the reaction rate of EC in moles per liter per minute column but not the details of search procedure. From Figure 3,
and Ci is the concentration of the ith component in moles per it is found that the two specifications are met and the distillate
liter. The above kinetic expression was obtained experimentally product is mainly a mixture of MeOH and DMC. The design
with homogeneous catalyst of sodium methylate with its of the downstream flow sheet to separate these two components
concentration between 0.2 and 0.3 wt %. In the following design and recycle MeOH to the RD column will be shown in the next
study, because of very low concentration, this catalyst compo- section.
nent was ignored in the simulation for simplification purpose. A final comment worth mentioning is that experimental
The system to be simulated has four components. The verification of reactive distillation in a pilot plant is important
vapor-liquid equilibrium is described by use of a UNIQUAC to concretely confirm the findings in the simulation study. A
model for the liquid phase and an RK model for the vapor phase. recent paper by Schmitt et al.21 demonstrated that the reaction
The experimental data for the MeOH-EG pair are obtained kinetics measured in the laboratory cannot always directly be
from Gmehling and Onken.16 The VLE data of the four pairs used in the simulation to successfully describe the reactive
MeOH-EC, DMC-EC, DMC-EG, and EG-EC are from a distillation in real plant. A simple “transfer factor” concept is
paper by Fang and Qian.17 The VLE data for the remaining proposed in that paper to close the gap between reactive
pair (MeOH-DMC) are from a paper by Rodriguez et al.18 distillation simulation and real experimental results.
Regression has been performed to obtain suitable UNIQUAC
parameters for the latter five pairs, and the built-in parameters 3. Design of Extractive Distillation System
from Aspen have been used for the MeOH-EG pair. The
resulting parameter set can be found in Table 1. Figure 1a As can be seen in the second row of Figure 1a, a mixture of
displays the prediction of the y-x and T-xy plots of all six MeOH and DMC contains a minimum-boiling azeotrope at
pairs by the thermodynamic model. The figure also contains 86.47 mol % MeOH and azeotropic temperature at 63.73 °C;
the experimental data for comparison. It is found that the thus a single distillation column cannot produce two pure
thermodynamic model is excellent in predicting the VLE products. The azeotropic composition changes with operating
behavior and the two azeotropes19 of the system. pressure, for example, at 10 atm the azeotropic composition is
The reactive distillation design concept is to feed excess predicted to be 92.15 mol % MeOH; however, the shift in the
methanol into the system, thus consuming EC to nearly complete azeotropic composition is not significant. The paper by Wang
conversion. From the residue curve map (RCM) in Figure 2, et al.8 confirmed that pressure-swing distillation is not an
EG product (nbp 197.08 °C) can be withdrawn from the bottom attractive way to separate this azeotrope.
of the column, while the column top composition approaches An alternative method for the separation of this mixture is
the azeotrope of MeOH-DMC (63.73 °C). The design flow to use extractive distillation by adding a heavy-boiling entrainer
sheet of this reactive distillation column is developed by into this system. The purpose of the entrainer in this extractive
minimizing the total annual cost (TAC) of the system. The TAC distillation column is to enhance the relative volatility between
includes the utility cost and the annualized capital cost. The MeOH and DMC, so that one component (e.g., MeOH) goes
utility cost includes steam for the reboiler, cooling water for to the top of the column and the other component (e.g., DMC)
the condenser, and catalyst cost. The capital cost includes the goes to the bottom of the column with the entrainer. The upper
column shell, internal trays, reboiler, and condenser. A capital section of the column (above the entrainer feed location) is
charge factor of 3 years is assumed in the calculation. The cost called the rectifying section, and its purpose is to separate MeOH
calculation can be seen from appendix E of Douglas.20 and the entrainer. The middle section of the column (between
The EC feed flow was assumed to be 10 kmol/h, and the entrainer feed stage and the fresh feed stage) is called the
atmospheric column operation was also assumed. In order to extractive section. The purpose of this section is to suppress
enhance the contact of the two reactants, EC (a heavier boiler) DMC from going up the column. The bottom section of the
and MeOH (a lower boiler) are fed from the top portion and column (below the fresh feed location) is called the stripping
the bottom portion of the RD column, respectively. The section, and its purpose is to keep MeOH from going down the

Table 1. UNIQUAC Model Parameters of EC/MeOH/DMC/EG System for Reactive Distillation Study
i ) DMC, j ) EC i ) MeOH, j ) EC i ) DMC, j ) EG i ) EG, j ) EC i ) MeOH, j ) DMC i ) MeOH, j ) EG
aij 2.5273 -0.540 94 0 0 0 -32.587
aji -6.7598 15.892 0 0 0 2.2712
bij(K) 167.34 4589.3 -236.13 -275.66 -18.065 10 000
bji(K) 442.34 -1278.9 -146.88 -27.961 -263.58 -599.11
cij -0.732 11 -2.2771 0 0 0 0
cji 1.0398 -2.0488 0 0 0 0
 Ind. Eng. Chem. Res., Vol. 49, No. 2, 2010 737

Figure 1a. Continued.

738 Ind. Eng. Chem. Res., Vol. 49, No. 2, 2010

Figure 1. y-x and T-xy plots of the RD system.

column. The bottom product of the column is the mixture of can be recycled back to the extractive column. The conceptual
DMC and the entrainer, and it is fed to another entrainer design of the two-column system via extractive distillation is
recovery column to separate these two components, so entrainer shown in Figure 4.
 Ind. Eng. Chem. Res., Vol. 49, No. 2, 2010 739

Figure 2. Residue curve map for the ternary system of DMC-EG-MeOH Figure 4. Conceptual design flow sheet of the extractive distillation system.
at 1 atm.

Figure 5. Isovolatility curve at 1 atm with phenol as entrainer.

triangle where the concentration of DMC goes to zero. This

observation determines that the overhead product of the extrac-
Figure 3. RD design flow sheet. tive distillation column should be MeOH and not DMC. The
The entrainer selection is extremely important to make this reasoning is that by adding phenol into the system, the relative
design flow sheet more attractive. The concept is to find an volatility between MeOH and DMC will become greater than
entrainer that will enhance most of the relative volatility of the unity and result in MeOH going up the column. Note that the
original two components but is also easy to separate from either relative volatility between MeOH and DMC is greater than unity
of the two components. As mentioned in the Introduction, good in the region below the isovolatility curve. This can be explained
screening procedures through computer-aided molecular design by a binary feed composition of MeOH and DMC mixture in
such as UNIFAC were suggested to find a set of feasible this region. A simple distillation would result in azeotropic
entrainers for a particular separation system.13,14 After some composition going to overhead and DMC going to the bottom;
feasible entrainers have been found, in the following, we will thus the relative volatility between MeOH and DMC is greater
use DMC-MeOH system as an example to illustrate further a than unity in this region. On the contrary, a binary feed
simple procedure to quickly compare the performance of composition in the opposite region would result in azeotropic
candidate entrainers by use of a commercial process simulator. composition going to overhead and MeOH going to the bottom;
3.1. Comparison of Candidate Entrainers. In Wang et al.,8 thus the relative volatility in this region is less than unity.
phenol was used as the entrainer in the extractive distillation It is worth mentioning that, in an extractive distillation system,
system. Figure 5 displays the isovolatility curve for the the lighter of the original two components cannot automatically
MeOH-DMC system at 1 atm with phenol as entrainer. Notice be assumed to be the distillate product of this column. A good
that the volatility between MeOH and DMC is unity in the example is the acetone-methanol separation with chlorobenzene
binary system when it is at the azeotrope (shown on the ordinate as entrainer.14,15 For that extractive distillation system, the
in Figure 5 with a temperature of 63.73 °C). As phenol is added, methanol is the distillate product although its normal boiling
the isovolatility curve moves toward the hypotenuse of the point is higher than that of acetone. The decision of the desirable
740 Ind. Eng. Chem. Res., Vol. 49, No. 2, 2010

Figure 7. Relative volatility at various feed ratios.

Although phenol is a feasible entrainer for this extractive

distillation system, the important question is whether this
entrainer is effective enough in enhancing the relative volatility
upon addition into the system. This question can be answered
by plotting not only the isovolatility curve but also the
equivolatility curves (curves with relative volatilities other than
unity) of the ternary system.10
Figure 6 displays the iso- and equivolatility curves at 1 atm
for three alternative entrainers. The other two entrainers are
ethylene glycol and aniline. These two are picked in a list of
commonly used solvent in Gmehling and Mollmann.13 From
Figure 6, it is easily concluded that aniline is a more effective
entrainer for this system that enhancing more of the relative
volatility between MeOH and DMC.
Another comment for showing the result of using ethylene
glycol as entrainer is elucidated below. Although true in many
cases (e.g., comparing the results for phenol and aniline as
entrainer), we should not solely rely on the location where the
isovolatility curve intercepts the MeOH-entrainer edge of the
triangle to determine which one is the more effective entrainer.
Figure 6 shows that although the xE for using EG as entrainer
(xE ) 0.02) is much smaller than that for using aniline (xE )
0.14), the enhancement of relative volatility between DMC and
methanol is much poorer with EG as entrainer. Thus EG is not
a good entrainer for this separation.
Another way to demonstrate the effectiveness of the entrainer
is using a plot as in Figure 7. In this plot, the starting
composition is right at the azeotrope of MeOH and DMC. By
gradually adding entrainer into the system, we can calculate
the enhancement of relative volatility at various feed ratios. This
figure demonstrates again that aniline is a much more effective
entrainer than phenol for this separation system.
Generating the above analysis plots is very easy by use of a
commercial process simulator such as the one from Aspen
Technologies, Inc. The isovolatility curve is readily available
in Aspen Plus. The way to generate equivolatility curves in
Aspen Plus is to use the Flash2 module in the unit operation
Figure 6. Iso- and equivolatility curves at 1 atm with (a) phenol, (b) EG, library. With any arbitrary feed composition in the ternary
or (c) aniline as entrainer. diagram, under adiabatic flash operating at 1 atm and with
negligible vapor flow, the vapor and liquid composition in
equilibrium can be calculated by Aspen Plus. Information is
distillate product can easily be answered by drawing the excerpted about MeOH and DMC, and the relative volatility
isovolatility curve. between these two components can be calculated for addition
 Ind. Eng. Chem. Res., Vol. 49, No. 2, 2010 741

Figure 8. T-xy plots of (a) MeOH-aniline and (b) DMC-aniline.

Table 2. UNIQUAC Model Parameters of the MeOH/DMC System with either Aniline, Phenol, or Ethylene Glycol as Entrainer
i ) aniline, j ) MeOH i ) aniline, j ) DMC i ) phenol, j ) MeOH i ) phenol, j ) DMC i ) EG, j ) MeOH i ) EG, j ) DMC
aij 0 0 0 0.767 07 2.2712 0
aji 0 0 0 -1.5058 -32.587 0
bij(K) -288.28 -32.733 -420.31 -273.39 -599.11 -146.88
bji(K) 85.941 101.44 350.24 691.93 10 000 -236.13
cij 0 0 0 0 0 0
cji 0 0 0 0 0 0

of any amount of the entrainer component. By collection of the separation of MeOH and aniline. Figure 8b shows the T-xy
feed compositions resulting in the same relative volatility, a plot of the DMC-aniline pair. From this plot, it is also
curve can be drawn with equal relative volatility. concluded that aniline is a good entrainer when we focus on
Other factors that affect the total annual cost of the design the separation in the entrainer recovery column.
flow sheet are the y-x or T-xy plots for the MeOH-entrainer
For completeness of information for generating the above
and DMC-entrainer pairs. For the MeOH-entrainer pair,
plots, the UNIQUAC model parameters for calculating the
these plots can be used to determine the ease of separation
in the rectifying section of the extractive distillation column. vapor-liquid equilibrium are listed in Table 2.
For the DMC-entrainer pair, the plots can be used to From the above analysis, looking into iso- and equivolatility
determine whether the separation in the entrainer recovery curves and binary VLE diagrams for the MeOH-entrainer and
column is easy or not. DMC-entrainer pairs, it is concluded that aniline is a better
Figure 8a shows T-xy plots of the MeOH-aniline pair. From entrainer for this system. In the following, the optimal design
this plot, it is concluded that there is no problem with the flow sheet of this MeOH-DMC separation process with aniline
742 Ind. Eng. Chem. Res., Vol. 49, No. 2, 2010

Table 3. Design Variables and TAC of Extractive Distillation Flow

Sheet
extractive distillation column
FE/FF feed ratio 0.883
total stages (N1) 32
entrainer feed stage (NFE) 5
fresh feed stage (NFF) 27
reboiler cost 59 406
condenser cost 55 998
column cost 158 418
tray cost 14 901
steam cost 168 839
cooling water cost in condenser 2965
entrainer recovery column
total stages (N2) 18
feed stage (NF2) 6
reboiler cost 24 148
condenser cost 15 184
column cost 52 267
tray cost 3301
steam cost 43 752
cooling water cost in condenser 858
Figure 9. Summary total TAC plot at various entrainer feed rates. entrainer cost 17
cooler cost 39 630
cooling water cost in cooler 1608
as entrainer will be established to minimize the total annual total utility cost 218 039
cost of the overall process. annualized total capital cost 423 253
It is worth mentioning that, apart from the important factors total TAC 641 292
above relating to the phase equilibrium behavior, other factors
such as thermal stability, nontoxicity, low price, and other (NT2), and feed tray location of entrainer recovery column (NF2).
physical properties should also be considered in the entrainer Following the suggestion in Knight and Doherty22 and also
selection. summarized in Chapter 5 of the book by Doherty and Malone,3
3.2. Optimal Design Flow Sheet. The proposed design flow entrainer feed temperature is set to be about 10 °C less than
sheet with aniline as entrainer can be seen in Figure 4. The the top temperature of the extractive distillation column, so a
product specifications are set to be 99.99 mol % MeOH in the cooler is included in the conceptual design flow sheet.
distillate of the extractive distillation column and 99.5 mol %
DMC in the distillate of the entrainer recovery column. The There are many design variables to be determined, so to
feed condition is set to be the same as the distillate of the RD simplify the optimization procedure the more important and
column. Reflux drums of both columns are operated at complex extractive distillation column is optimized first. There
atmospheric pressure. Column pressure drops are automatically are three design variables to be determined (NT1, NFE, and NFF)
calculated in the Aspen simulation. at each entrainer feed rate. The two design specifications for
The design variables to be determined in the flow sheet all the Aspen simulations are setting the top composition at 99.99
include the entrainer feed flow rate with fixed feed rate of 67.96 mol % MeOH and the ratio of MeOH to the sum of DMC and
kmol/h from RD distillate, total stages of the extractive MeOH in the bottom stream to be 0.004. The reason for this
distillation column (NT1), entrainer and fresh feed tray locations bottom specification is to set the MeOH loss through the column
(NFE and NFF), total stages of the entrainer recovery column bottom. The above two design specifications can be met by

Figure 10. Design flow sheet of the extractive distillation system.

 Ind. Eng. Chem. Res., Vol. 49, No. 2, 2010 743

Figure 11. Comparison to Wang et al.8 design flow sheet: (a) Wang et al.8 flow sheet; (b) flow sheet with aniline as entrainer.

varying the two degrees of freedom in this column (e.g., reboiler NT1. After the optimum solution of NT1, NFE, and NFF at each
duty and reflux flow rate). entrainer feed rate is obtained, the total TAC of this two-column
Total annual cost (TAC1) of this column is used as the system can be calculated with the entrainer recovery column
objective function to be minimized, which including annualized and the recycle stream included. Additional costs in the total
capital costs and operating costs. A sequential iterative opti- TAC include annualized capital cost for the entrainer recovery
mization search procedure is used to find the optimal design column, costs associated with the cooler for the entrainer recycle,
with the entrainer feed rate as the outer iterative loop, NT1 as operating costs of the steam and cooling water to operate the
the middle loop, and NFE and NFF as the inner iterative loop. entrainer recovery column, and entrainer makeup cost.
The procedure is to fix entrainer feed rate and NT1 first and then Figure 9 shows a summary plot of the total TAC of this two-
to find the right NFE and NFF to minimize TAC1 at this particular column system at various entrainer feed rates. Any point in this
combination of entrainer feed rate and NT1. The procedure is summary plot is a collective result of many simulations in the
repeated for the other combinations of entrainer feed rate and inner iterative loops to determine the total stages and feed
744 Ind. Eng. Chem. Res., Vol. 49, No. 2, 2010

Figure 12. Addition of a heel-purge column in the overall process.

locations of the two-column system. It is observed that the best two components are very high, EC at 248.56 °C and EG at 197.08
entrainer feed rate is 60 kmol/h. Table 3 shows the optimal °C. Because this bottom stream is recycled back to the extractive
design variables and also the minimized TAC results of this distillation column, these two heavy-boiler impurities would not
two-column system. The final design flow sheet of this two- escape from the two distillate streams and therefore will be
column system is shown in Figure 10. It is noticed that the total accumulated in the system. Since this is a very practical situation
stages of the extractive distillation column and also the entrainer in an industrial setting, it is worth discussing here.
recovery column are moderate. The ratio between the entrainer One way to avoid heavy-boiler impurities in the feed stream
feed and MeOH-DMC feed is only 0.883. of the extractive distillation system is to add a heel-purge column
3.3. Comparison to Wang et al.8 Design. In this section, between the RD column and the extractive distillation system.
simulation results with aniline and phenol as entrainer will be The conceptual design flow sheet can be seen in Figure 12.
compared. The simple analysis in section 3.1 demonstrated that Conventional wisdom told us that this will not be the economical
aniline should be a much more effective entrainer for this solution for this problem because a large reboiled ratio resulting
separation system. We will further strengthen this conclusion in extra reboiler duty is required to vaporize most of the
with process simulation. materials to overhead in this new heel-purge column, not to
The feed condition as in Wang et al.8 is adapted in this mention additional column equipment and instrumentation will
comparison, and also all the product purities are set as the same. be needed for this design flow sheet.
We will just use the optimization results obtained in the previous Another easier solution to solve this problem is to increase
section for this separation task with aniline as entrainer; thus the rectifying section of the RD column as in Figure 3 of this
feed ratio is set at 0.883, NT1 ) 32, NFE ) 5, NFF ) 27, NT2 ) paper. Figure 13 shows two design flow sheets of the RD
18, and NF2 ) 6. Figure 11 shows two design flow sheets with column with varying rectifying sections. Figure 13a shows the
either phenol or aniline as entrainer. The simulation result for design flow sheet of the RD column with minimized TAC.
the case with phenol as entrainer is excerpted from Wang et However, because there are only two stages of the rectifying
al.8 It is observed that the sum of reboiler duties for our system section, small impurities of EC and EG (especially EG) are
is reduced by as much as 32.8%. The required feed ratio of the present in the distillate stream. Upon simple addition of several
Wang et al.8 system is much higher (1.965 vs 0.883 in our stages in the rectifying section, as can be seen in Figure 13b,
system). The extractive distillation column in Wang et al.8 is EC and EG become negligible in the distillate stream. Notice
much taller (NT1 ) 48 vs 32 as in our system), as is the entrainer the design specifications (purity of EG at bottom set at 0.995
recovery column (NT2 ) 32 vs 18 as in our system). Wang et and conversion set at 99.95%) are all met for these two design
al.8 did not give information on column diameters. However, flow sheets. The TAC of this feasible flow sheet (Figure 13b)
judging from the much higher reboiler duty, the column diameter is only slightly increased from that of the optimized flow sheet.
should be much larger too. Comparison of TACs for Figure 11
cannot be made because we do not have detailed design 4. Control Strategy for the Overall Process
information on the Wang et al.8 system. However, from the
above comparison of column heights and reboiler duties, their The overall design flow sheet of this process is displayed in
capital and utility costs should be much higher. Figure 14. There are two recycle streams in the overall process.
3.4. Problem with Heavy-Boiler Impurities in the Feed One is to recycle the excess reactant, methanol, back to the
Stream. In Figure 4 of the paper by Wang et al.,8 it is noticed reactive distillation column. The other one is to recycle the
that the bottom composition contains small impurities of EC (2.5 entrainer to the extractive distillation column. In this section,
× 10-3 mol %) and EG (2.8 × 10-3 mol %). These small impurities the overall control strategy of this complete process will be
are carried into this extractive distillation system via the feed stream studied. The purpose is to maintain the purity of the dimethyl
(distillate of the RD column). The normal boiling points of these carbonate product as well as the coproduct, ethylene glycol. In
 Ind. Eng. Chem. Res., Vol. 49, No. 2, 2010 745
the control strategy development, we restrict ourselves to use
only tray temperature control loop(s) in each of the three
columns to indirectly hold the product purity for wider industrial
application purposes.
Pressure-driven simulation in Aspen Dynamics is used in the
control strategy development. Ten minutes holdup time with
50% liquid level is used to calculate the volume of each column
base and reflux drum of each column. Top pressures of the three
columns are all set at atmospheric pressure. The tray rating tool
in Aspen Plus is used to automatically calculate the pressure
drop along all columns.
The inventory and some simple regulatory control loops are
determined first as below:
1. EC feed is flow-controlled. (used as a throughput manipu-
lator).
2. Total MeOH feed is flow-controlled by manipulating a
control valve at fresh feed stream. The total MeOH feed set
point is changed to maintain MeOH/EC feed ratio into the RD
column. This feed ratio can be reset by a tray temperature control
loop.
3. RD column base level is controlled by manipulating the
bottom flow (EG product flow).
4. RD column reflux drum level is controlled by manipulating
the distillate flow.
5. RD column top pressure is controlled by manipulating the
condenser duty.
6. RD column reflux is a ratio to EC feed flow.
7. RD column reboiler duty is a ratio to EC feed flow.
8. Base level of the extractive distillation column is controlled
by manipulating the bottom flow.
9. Reflux drum level of the extractive distillation column is
controlled by manipulating the distillate flow (MeOH recycle
flow).
10. Top pressure of the extractive distillation column is
controlled by manipulating the condenser duty.
11. Reflux flow of the extractive distillation column is ratio
to distillate of RD column.
12. Base level of the entrainer recovery column is controlled
by manipulating the aniline makeup flow.
13. The entrainer feed flow to the extractive distillation
column is flow-controlled and ratio to distillate of RD column.
Figure 13. Effect of increasing rectifying section: (a) smaller rectifying
14. Reflux drum level of the entrainer recovery column is section; (b) larger recitifying section.
controlled by manipulating the distillate flow (DMC product
flow).
column to infer the compositions. The manipulated variable selected
15. Reflux flow of the entrainer recovery column is a ratio
for the RD temperature loop is the MeOH/EC feed ratio; the ones
to distillate of the RD column.
for the extractive distillation and entrainer recovery columns are
16. Top pressure of the entrainer recovery column is
their reboiler duties. The open-loop sensitivity analysis was
controlled by manipulating the condenser duty.
performed to find the temperature control point(s) that is(are) most
17. Entrainer feed temperature is controlled by manipulating
sensitive to the manipulated variable changes while still exhibiting
cooler duty.
near-linear behavior. The resulting temperature control points are
An important inventory control loop in the extractive distillation
system is the bottom level of the entrainer recovery column (item at 27th, 30th, and fourth stage for the RD, extractive distillation,
12). The control of this level was suggested by Grassi23 and and entrainer recovery columns, respectively.
Luyben24 to be held by the entrainer makeup flow. However, Figure 15 displays the overall control strategy of this complete
because this flow is very small, thus bottom level essentially floats process. In the closed-loop simulation runs, P-only controller
indicating whether the entrainer inventory in the system is balanced is used in all level loops. The reason for using P-only controller
or not. With this control pairing, the entrainer feed flow to the is to provide maximum flow smoothing and also because
extractive distillation column is flow-controlled and ratio to distillate maintaining a liquid level at set point value is often not
of RD column. Another control strategy (items 11 and 15) was necessary. Kc ) 2 as suggested in Luyben26 is used in most of
adapted from Arifin and Chien,25 where they found that control the level loops. For the two bottom level loops in the extractive
strategy with fixing R/F ratio is better in rejecting the disturbances distillation system, Kc ) 10 is used so that faster dynamics of
than fixing the reflux ratio. the internal flow of the overall process can be achieved and
As for the product quality control loops, the simplest control also for faster increase or decrease of entrainer makeup into
strategy is proposed to use only one tray temperature in each the system. For the top pressure control loops of the columns,
746 Ind. Eng. Chem. Res., Vol. 49, No. 2, 2010

Figure 14. Design flow sheet of the complete process.

Figure 15. Overall control strategy.

tight PI controller tuning parameters of Kc ) 20 and τI ) 12 distillation column are Kc ) 0.71 and τI ) 10.5 min; the tuning
min are used. constants for the tray 30 temperature loop of the extractive
In each temperature loop, an additional 1-min dead time is distillation column are Kc ) 1.2 and τI ) 3.0 min; and the tuning
included for modeling the other neglected dynamics in the constants for the tray 4 temperature loop of the entrainer
system. The tuning constants are determined via relay feedback recovery column are Kc ) 1.3 and τI ) 5.5 min.
test provided in Aspen Dynamics with Tyreus and Luyben Two feed disturbances are introduced into the system to test
tuning rules.27 The iterative tuning procedure was to tune the the control strategy. The first one is the (20% throughput
temperature loop in the reactive distillation column first, then changes. This can be done by changing the set point of the EC
the extractive distillation column, and then the entrainer recovery feed flow control loop. Figure 16 shows the closed-loop results
column. The procedure is repeated until the tuning parameters with the proposed control strategy. From Figure 16a, it is noticed
from relay feedback test converged. The resulting PI tuning that all three controlled temperatures are returned back to their
constants for the tray 27 temperature loop of the reactive set point values with smooth manipulated variable changes.
 Ind. Eng. Chem. Res., Vol. 49, No. 2, 2010 747

Figure 16. Closed-loop results with (20% throughput changes (solid line, +20%; dashed line, -20%).
748 Ind. Eng. Chem. Res., Vol. 49, No. 2, 2010

Figure 17. Closed-loop results with EC feed composition changes (solid line, EC feed contains 10% EG; dashed line, EC feed contains 20% EG).
 Ind. Eng. Chem. Res., Vol. 49, No. 2, 2010 749
From Figure 16b, the compositions of two products (DMC and (2) Fang, Y. J.; Xiao, W. D. Experimental and modeling studies on a
EG) as well as the compositions of the two recycle streams homogeneous reactive distillation system for dimethyl carbonate synthesis
by transesterification. Sep. Purif. Technol. 2004, 34, 255–263.
(MeOH and aniline) all display only small deviations from their (3) Doherty, M. F.; Malone, M. F. Conceptual Design of Distillations;
specifications. Notice that because the throughput changes are McGraw-Hill: New York, 2001; p 382.
a known disturbance, it is possible to adjust the temperature (4) Luyben, W. L. Control of heterogeneous azeotropic n-butanol/water
set points to make the compositions closer to their specifications. distillation system. Energy Fuels 2008, 22, 4249–4258.
The second feed disturbance is feed composition changes: it (5) Wu, Y. C.; Chien, I. L. ; Luyben, W. L. A two-stripper/decanter
flowsheet for methanol recovery in the TAME reactive-distillation process
is assumed that the EC feed is not pure but contains 10% or Ind. Eng. Chem. Res. 2009 (submitted for publication).
20% EG in this feed stream. The proposed control strategy (6) Luyben, W. L. Comparison of pressure-swing and extractive-
should adjust the MeOH/EC feed ratio to accommodate these distillation methods for methanol-recovery systems in the TAME reactive-
changes. Figure 17 displays the closed-loop results. It is distillation process. Ind. Eng. Chem. Res. 2005, 44, 5715–5725.
(7) Luyben, W. L. Design and control of a fully heat-integrated pressure-
observed that the MeOH/EC feed ratio is reduced, as was swing azeotropic distillation system. Ind. Eng. Chem. Res. 2008, 47, 2681–
desired. The product compositions show only small deviations 2695.
from their base case values. (8) Wang, S. J.; Yu, C. C.; Huang, H. P. Plant-wide design and control
of DMC synthesis process via reactive distillation and thermally coupled
extractive distillation Comput. Chem. Eng. 2009 (in press).
5. Conclusion (9) Foucher, E. R.; Doherty, M. F.; Malone, M. F. Automatic screening
of entrainers for homogeneous azeotropic distillation. Ind. Eng. Chem. Res.
In this paper, design and control of the overall process for
1991, 30, 760–772.
production of dimethyl carbonate is studied. The focus is (10) Laroche, L.; Bekiaris, N.; Andersen, H. W.; Morari, M. Homoge-
particularly on the separation of DMC-MeOH azeotrope via neous azeotropic distillation. Comparing entrainers. Can. J. Chem. Eng.
extractive distillation. Since the ability of the entrainer to 1992, 69, 1302–1319.
enhance relative volatility is crucial to the success of the (11) Rodriguez-Donis, I.; Gerbaud, V.; Joulia, X. Entrainer selection
rules for the separation of azeotropic and close-boiling-temperature mixtures
separation task, a very simple procedure is proposed in this paper by homogeneous batch distillation process. Ind. Eng. Chem. Res. 2001, 40,
to quickly compare alternative entrainer candidates before 2729–2741.
rigorous process simulation is conducted. An isovolatility curve (12) Chen, B.; Lei, Z.; Li, Q.; Li, C. Application of CAMD in separating
is used to determine the desirable distillate product of the hydrocarbons by extractive distillation. AIChE J. 2005, 51, 3114–3121.
(13) Gmehling, J.; Mollmann, C. Synthesis of distillation processes using
extractive distillation column. Equivolatility curves together with
thermodynamic models and the Dortmund data bank. Ind. Eng. Chem. Res.
the relative volatility plot, as in Figure 7, can be used to compare 1998, 37, 3112–3123.
the ability of alternative entrainers to enhance the relative (14) Kossack, S.; Kramer, K.; Gani, R.; Marquardt, W. A systematic
volatility of the original two components. The T-xy plots of synthesis framework for extractive distillation processes. Chem. Eng. Res.
the entrainer versus one of the original two components should Des. 2008, 86, 781–792.
(15) Luyben, W. L. Effect of solvent on controllability in extractive
also be examined to confirm ease of separation in the rectifying distillation. Ind. Eng. Chem. Res. 2008, 47, 4425–4439.
section of the extractive distillation column and in the entrainer (16) Gmehling, J.; Onken U. Vapor-Liquid Equilibrium Data Collection;
recovery column. Dechemav: Frankfurt, Germany, 1977.
For the system of separating methanol and DMC, aniline was (17) Fang, Y. J.; Qian, J. M. Isobaric vapor-liquid equlibria of binary
mixtures containing the carbonate group -OCOO-. J. Chem. Eng. Data 2005,
found to be a much more effective entrainer than the one used
50, 340–343.
in open literature. This paper also pointed out a practical problem (18) Rodriguez, A.; Canosa, J.; Dominguez, A.; Tojo, J. Vapour-liquid
that may occur in industry when the feed stream of an extractive equilibria of dimethyl carbonate with linear alcohols and estimation of
distillation system contains heavy-boiler impurities. If these interaction parameters for the UNIFAC and ASOG method. Fluid Phase
impurities are not eliminated from the feed stream, accumulation Equilib. 2002, 201, 187–201.
(19) Gmehling, J. Azeotropic Data, 2nd ed.; Wiley-VCH:Weinheim,
will occur in the extractive distillation system, causing opera- Germany, 2004.
tional problems. A practical solution for this problem is to (20) Douglas, J. M. Conceptual Design of Chemical Process; McGraw-
increase the rectifying section of the RD column to suppress Hill: New York, 1988.
these heavy-boiler impurities from going up the RD column (21) Schmitt, M.; Blagov, S.; Hasse, H. Mastering the reaction is the
key to successful design of heterogeneously catalyzed reactive distillation:
and showing up in the feed stream of the extractive distillation
A comprehensive case study of hexyl acetate synthesis. Ind. Eng. Chem.
system. Res. 2008, 47, 6014–6024.
The design and control of the overall process have also been (22) Knight, J. R.; Doherty, M. F. Optimal design and synthesis of
developed. The overall process contains two recycle streams, homogeneous azeotropic distillation sequences. Ind. Eng. Chem. Res. 1989,
one to recycle excess methanol reactant and the other to recycle 28, 564–572.
(23) Grassi, V. G. Process design and control of extractive distillation.
entrainer. A simple overall control strategy of this process is In Practical Distillation Control; Van Nostrand Reinhold Press: New York,
also proposed to maintain product purity despite various feed 1992; pp 370-404.
disturbances. Only one tray temperature control loop is required (24) Luyben, W. L. Plantwide control of an isopropyl alcohol dehydra-
for each of the three columns. tion process. AIChE J. 2006, 52, 2290–2296.
(25) Arifin, S.; Chien, I. L. Design and control of an isopropyl alcohol
dehydration process via extractive distillation using dimethyl sulfoxide as
Acknowledgment an entrainer. Ind. Eng. Chem. Res. 2008, 47, 790–803.
(26) Luyben, W. L. Plantwide Dynamic Simulators in Chemical
This work is supported by the National Science Council of Processing and Control; Marcel Dekker; New York, 2002.
ROC under Grant NSC 97-2221-E-011-076-MY3. (27) Tyreus, B. D.; Luyben, W. L. Tuning PI controllers for integrator/
deadtime processes. Ind. Eng. Chem. Res. 1992, 31, 2625.
Literature Cited ReceiVed for reView July 20, 2009
ReVised manuscript receiVed September 27, 2009
(1) Ono, Y. Catalysis in the production and reactions of dimethyl Accepted November 18, 2009
carbonate, an environmentally benign building block. Appl. Catal., A 1997,
155, 133–166. IE901157G