pubs.acs.

org/IECR Article

Extractive Dividing-Wall Column Distillation with a Novel Control

Structure Integrating Pressure Swing and Pressure Compensation
Cong Jing, Jiaxing Zhu, Leping Dang,* and Hongyuan Wei
Cite This: Ind. Eng. Chem. Res. 2021, 60, 1274−1289 Read Online

ACCESS Metrics & More Article Recommendations *

sı Supporting Information
See https://pubs.acs.org/sharingguidelines for options on how to legitimately share published articles.

ABSTRACT: In this work, an energy-efficient extractive dividing-

wall column process with heat integration (E-DWC-HI) is proposed
Downloaded via UNIV OF ENG & TECH LAHORE on March 3, 2021 at 07:42:50 (UTC).

for the first time to separate the dichloromethane (DCM)−methanol

(MeOH) azeotrope, which is a widely existing waste effluent in the
pharmaceutical industry. The result of economic evaluation shows
that the E-DWC-HI process is a preferred choice for the separation of
DCM−MeOH mixture. Designing a robust control strategy is
essential for the E-DWC-HI process. Thus, four control structures are
proposed and tested under the disturbances of ±20% feed flow rate
and ±5% feed composition. The basic control structure (CS1) and
the double temperature difference control structure (CS2) both have
a fixed vapor split. The dynamic response shows that the required
purity cannot be met due to the loss of an important control degree
of freedom. In CS3, variable vapor split is carried out by adjusting the
pressures on the two sides of the dividing wall, and all product
purities are held close to their set points except for MeOH product
purity with an obvious offset. The dynamic performance of CS4 has been significantly improved by integrating pressure swing and
pressure compensation. The offset of MeOH product purity is significantly reduced from 2.05 to 0.03%.

1. INTRODUCTION Extractive dividing-wall column (E-DWC) is a promising process

Dichloromethane (DCM) and methanol (MeOH) are both intensification technology to achieve lower cost by improving
excellent organic solvents and are widely used as solvents in process efficiency, reducing equipment size and energy con-
the pharmaceutical industry, such as in the production of sumption.15,19,20 However, Wu et al.21 examined the economics
prednisone and antibiotics.1,2 During these processes, large of three different chemical separations in a critical assessment of
amounts of waste effluent are generated, which consist of a the applicability of E-DWC. They pointed out that, for new
binary mixture of DCM and MeOH in significant proportions. systems, E-DWC cannot guarantee its superiority in economics
The waste effluents with large amounts of DCM pose a potential over conventional extractive distillation. The reason is the
threat to human health and the environment.3−5 On the other introduction of high boiling point solvents which require steam
hand, as an active step, separating these components and of higher temperature that is more expensive. Therefore, when the
recovering the chemical feedstock would be of significant E-DWC scheme is applied to a new system, the economic benefits
benefits and would be necessary to meet rigorous environmental should be carefully examined. To the best of our knowledge, E-DWC
and resource-saving regulations for modern chemical industry. for separation of DCM−MeOH has not yet been reported.
Moreover, the demands for the treatment of these waste Designing a robust control strategy is essential for the E-DWC
effluents increase sharply with the rapid growth of the antibiotic process to achieve good economic benefits and high product
industry.6 purity. Wu et al.21 investigated the dynamic controllability of
It is difficult to separate and recover DCM and MeOH E-DWC under the assumption that the vapor split could not be
mixtures by a simple distillation due to the presence of a minimum- used as a manipulated variable. Their results show that the
boiling azeotrope. Complex distillation technologies such as
pressure-swing distillation (PSD),7−9 azeotropic distillation
(AD),10,11 and extractive distillation (ED)12−15 are widely Received: August 6, 2020
used to separate the azeotropic mixtures. For the ED process, the Revised: December 28, 2020
azeotrope can be separated by introducing the solvent to change Accepted: December 30, 2020
the relative volatilities of the azeotropic components.15 In recent Published: January 13, 2021
years, the ED process has been widely investigated and shows
excellent economic performance.15−19

© 2021 American Chemical Society https://dx.doi.org/10.1021/acs.iecr.0c03876

1274 Ind. Eng. Chem. Res. 2021, 60, 1274−1289
Industrial & Engineering Chemistry Research pubs.acs.org/IECR Article

Figure 1. Flow sheet of CED process.

Figure 2. Optimized flow sheet of E-DWC process.

dynamic controllability of E-DWC is hampered because both vapor split. Several studies have also highlighted the importance
columns are thermally driven by only one reboiler; an important of a control strategy to adjust the vapor split.22−24
control degree of freedom (duty of one reboiler) is lost. In the Some experimental work on small laboratory columns has
conventional process, two reboiler duties and two reflux flow been reported, which carried out “active vapor split control” by
using valves, dampers, blade-shape structures in the vapor
rates are well controlled. In other literature,22 the dynamic channels, and so on.25−28 However, it is still problematic to
controllability of E-DWC with and without manipulating the implement such moving mechanical parts on a typical large
vapor split was compared. The result shows that product quality industrial scale column. Also, a practical commercial application
could not be guaranteed by the control structure with a fixed has not yet been reported.
1275 https://dx.doi.org/10.1021/acs.iecr.0c03876
Ind. Eng. Chem. Res. 2021, 60, 1274−1289
Industrial & Engineering Chemistry Research pubs.acs.org/IECR Article

Figure 3. Influences of different optimization variables on the TAC of E-DWC.

Figure 4. Flow sheet of extractive divided-wall column with heat integration process.

1276 https://dx.doi.org/10.1021/acs.iecr.0c03876
Ind. Eng. Chem. Res. 2021, 60, 1274−1289
Industrial & Engineering Chemistry Research pubs.acs.org/IECR Article

Figure 5. Temperature and composition profiles for optimized E-DWC process.

Recent papers by Luyben29 and Feng et al.30 demonstrated effective and robust control structure despite various dis-
that it was possible to achieve manipulation of the vapor split turbances. In present work, an energy-efficient E-DWC process
by changing the operating pressures on the two sides of the is developed and optimized to minimize the TAC (total annual
dividing wall in an E-DWC configuration. To the best of our cost). Then, the heat integration between fresh feed and the
knowledge, there have been only a few reports on this method recycle solvent stream is applied to this E-DWC process. The
with pressure swing; its effectiveness and applicability remain controllability of the E-DWC process with heat integration
to be tested. Temperature control loops are widely used (E-DWC-HI) is further investigated. Four control structures
to infer the composition of key components in distillation. including a basic control structure, a double temperature difference
However, a constant temperature of the sensitive tray is not control structure, a control structure with pressure swing, and a
indicative of a constant composition if the pressure varies.31,32 novel control structure integrating pressure swing and pres-
Therefore, the dynamic controllability of this method may be sure compensation are designed and tested under large feed
hampered. For the accuracy of the estimation of composition, disturbances.
pressure compensation becomes very important for the new
method where pressure is significantly changed as an operating 2. PROCESS STUDY
variable. Thus, the new method will be introduced into our In this article, Aspen Plus version 11 is used to rigorously
control structures and be improved by integrating pressure simulate steady state processes. The selection of an appropriate
compensation. thermodynamic model is important for accurate simulation.
The purpose of this paper is to develop a sound design of Different predicted VLE data (at 1 atm) of various built-in
E-DWC for separating DCM−MeOH mixtures and to seek an thermodynamic models in Aspen are compared with two sets of
1277 https://dx.doi.org/10.1021/acs.iecr.0c03876
Ind. Eng. Chem. Res. 2021, 60, 1274−1289
Industrial & Engineering Chemistry Research pubs.acs.org/IECR Article

Figure 6. Profiles of differences between the temperatures on adjacent trays (left side) and the result of open-loop sensitivity test (right side).

different experimental data33,34 on the basis of their calculated The feed flow rate is 100 kmol/h with a composition of 83%
root-mean-square deviations (RMSDs), as shown in Table S1. DCM and 17% MeOH. The solvent is DMF with a solvent-to-
The vapor−liquid equilibrium diagram and T−xy diagram for feed ratio of 0.48. In addition, the tiny loss of solvent is balanced
dichloromethane−methanol is shown in Figure S1. The NRTL by adding a small amount of supplement to this process. The
model has the least RMSD values. Thus, the NRTL physical design specifications of the DCM and MeOH products are all set
property package is used in this simulation, and the behavior of to 99.9 mol %. The detailed flow sheet and stream parameters of
the gas phase is assumed to be ideal gas (IG) behavior. The the conventional extractive distillation are shown in Figure 1.
binary interaction parameters of NRTL are listed in Table S2. 2.2. Extractive Dividing-Wall Column Distillation.
The predicted azeotropic temperature is 37.6 °C with 0.8613 In this section, an E-DWC sequence is designed for the
mole fraction DCM and 0.1387 mole fraction MeOH at 1 atm. separation of the DCM−MeOH azeotrope, shown in Figure 2.
2.1. Conventional Extractive Distillation. The design of The feed flow rate, feed composition, and design specifications
the conventional extractive distillation flow sheet used herein is are consistent with those in the CED process mentioned above.
adapted from that presented by Iqbal et al.18 Process parameters This E-DWC configuration consists of an extractive section
in their work have been optimized based on economy. (upper left side), a recovery section (upper right side), and a
Therefore, the same process parameters and design specifica- stripping section (lower side), which cannot be directly simulated
tions are introduced into the CED process in this paper. Note by a ready-made model of Aspen Plus. Thus, a thermodynamically
that the minimum reflux ratio (RR2) and the duties of reboiler equivalent sequence is used to conduct the simulation of E-DWC,17
and condenser in column 2 can be further optimized. which is shown in Figure S2. All three thermodynamically
1278 https://dx.doi.org/10.1021/acs.iecr.0c03876
Ind. Eng. Chem. Res. 2021, 60, 1274−1289
Industrial & Engineering Chemistry Research pubs.acs.org/IECR Article

Figure 7. Basic control structure.

Table 1. Control Parameters of Temperature Controllers in (Vr) is defined as the ratio of the vapor fed into C101 to the
CS1 vapor from C103.
controlled set point control dead time
controller variable Kc τmin (°C) action (min) 3. OPTIMIZATION AND ECONOMIC EVALUATION
TC1 temp of stage 16 0.81 25.08 70.2 direct 1
in C101 3.1. Economic Basics. TAC as a widely used economic
TC2 temp of stage 5 4.49 14.52 85.0 direct 1 criterion in the chemical industries is adopted to evaluate the
in C102 economic performance of the CED and E-DWC processes. The
TC3 temp of stage 5 1.10 18.48 156.8 reverse 1 TAC includes annualized capital costs and energy costs. There
in C103
are some subitems in capital costs, e.g. column, reboiler, condenser,
equivalent columns C101, C102, and C103 are simulated by the and cooler. The energy costs include the following: steam for the
rigorous distillation model RadFrac. The initial vapor split ratio reboiler; cooling medium (cooling water or chilled water) for the

Figure 8. Dynamic responses for CS1 under ±20% feed flow rate disturbances.

1279 https://dx.doi.org/10.1021/acs.iecr.0c03876
Ind. Eng. Chem. Res. 2021, 60, 1274−1289
Industrial & Engineering Chemistry Research pubs.acs.org/IECR Article

Figure 9. Dynamic responses for CS1 under ±5% feed composition disturbances.

Figure 10. ASVMs in the presence of ±5% variations in feed composition.

condensers and cooler. The parameters of equipment sizing and Figure 3 shows the influence of optimization variables on the
economics13,35−37 are listed in Table S3. TAC of the E-DWC process. The graph of operating pressure
3.2. Optimization Variables and Procedure. A sequen- shows that the TAC reduced sharply when the operating
tial iterative procedure (as shown in Figure S3) is used to obtain pressure increased to 1.3 atm. This is because the pressure of
the optimized parameters by minimizing the TAC. Operating 1.3 atm corresponds to an operating temperature of 47.2 °C,
pressure, vapor split ratio (Vr), solvent flow rate, solvent which is greater than the reflux-drum criterion of 320 K
temperature, NT1, and NT3 are selected as the important (47 °C).19,35 Instead of using the expensive chilled water, the
cheap cooling water can be applied in condenser 1, and it
optimization design variables. NT1 and NT3 are the total
significantly reduces the energy cost. A pressure of 1.4 atm with a
numbers of stages required for the rectifying section and the minimum TAC is selected as the optimal operating pressure.
stripping section, respectively. It can also be found that the TAC has a consistent varying
In addition, fresh feed location (NF1) and solvent feed tendency with the increase of Vr, solvent flow rate, and NT1.
location (NFS) have also been considered as design variables to Note that Vr, solvent flow rate, and NT1 all should have a
minimize the reboiler duty (QR). The minimization is subject to minimum value to ensure the separation efficiency in the C101
the design specifications of product purities which are met by column; in this simulation, smaller Vr, solvent flow rate, and
adjusting RR1 and RR2, respectively. NT1 cannot meet the design specifications of product purities.
1280 https://dx.doi.org/10.1021/acs.iecr.0c03876
Ind. Eng. Chem. Res. 2021, 60, 1274−1289
Industrial & Engineering Chemistry Research pubs.acs.org/IECR Article

Figure 11. Double temperature difference control structure.

Table 2. Control Parameters of Temperature Controllers in 48 kmol/h (20% higher than the minimum value18), which
CS2 is consistent with that used in the CED process. The TAC
controlled set point control dead time
increases slightly with the increase of solvent temperature. The
controller variable Kc τmin (°C) action (min) sweet spot is 50 °C. Besides, with the increase of NT3, the TAC
DTDC1 Δ2T16 0.15 25.10 8.1 reverse 1 decreases sharply first and then increases slightly. The sweet spot
DTDC2 Δ2T5 2.34 10.56 4.0 reverse 1 of NT3 is 10.
TC3 temp of stage 5 1.10 18.48 156.8 reverse 1 3.3. E-DWC Process with Heat Integration. In addition,
in C103 heat integration is feasible since the temperature difference
between the fresh feed (40 °C) and the recycle solvent stream
Thus, the sweet spots of Vr and NT1 are 0.84 and 22, (170.3 °C) is enough. Thus, in order to further optimize the
respectively. For a safe operation margin under the disturbances economy of E-DWC, heat integration is performed on the basis
of throughput and composition, the solvent flow rate is set to of an optimized E-DWC process. The detailed flow sheet of the

Figure 12. Dynamic responses for CS2 under ±20% feed flow rate disturbances.

1281 https://dx.doi.org/10.1021/acs.iecr.0c03876
Ind. Eng. Chem. Res. 2021, 60, 1274−1289
Industrial & Engineering Chemistry Research pubs.acs.org/IECR Article

Figure 13. Dynamic responses for CS2 under ±5% feed composition disturbances.

Figure 14. Control structure with pressure swing.

E-DWC process with heat integration is shown in Figure 4. expensive chilled water. The flow sheet of the CED-IP process
It is worth noting that an important basic principle of energy is shown in Figure S4. Compared with the CED process, the
management is to avoid two-phase heating.38 Thus, in this energy cost savings of the CED-IP and E-DWC processes are
process, the liquid stream is kept at a sufficiently high pres- 25.7 and 42.3%, respectively. This result shows that the energy
sure (12 atm) to prevent the formation of vapor in the heat cost saving and TAC saving of E-DWC are partly due to the use
exchanger. Then, the feed flow will flash when it flows through of cooling water instead of chilled water in E-DWC and partly
the valve into the column which is at a lower pressure. due to the application of the dividing-wall column. In addition,
The economic results are summarized in Table S4. Note that, the E-DWC-HI process further optimizes the economy and has
to make a better economic comparison, a conventional extractive the best economic performance with an energy cost saving of
distillation process with an increased pressure of column 1 50.0% and a TAC saving of 27.7%; it is a preferred choice for the
(named the CED-IP process) is also developed to avoiding using separation of the DCM−MeOH mixture.
1282 https://dx.doi.org/10.1021/acs.iecr.0c03876
Ind. Eng. Chem. Res. 2021, 60, 1274−1289
Industrial & Engineering Chemistry Research pubs.acs.org/IECR Article

Figure 15. Dynamic responses for CS3 under ±20% feed flow rate disturbances.

The temperature and composition profiles of the optimized composition controller will not be applied in the following
E-DWC-HI process are shown in Figure 5. From the temperature control structures.
graph of C101 (upper left), the large reduction in temperature 4.1. Basic Control Structure (CS1). In the basic control
from the 14th tray to the 15th tray (feed stage) is mainly because structure, the commonly used single-ended temperature control
of the large difference in the boiling points between DCM loops are employed. The temperature control trays of the
(39.75 °C, 1 atm) and DMF (153 °C, 1 atm) and the quite large E-DWC-HI process are selected on the basis of the slope criterion
concentration (83%) of DCM in the feed stream. As shown in and the open-loop sensitivity test. The left graphs in Figure 6 give
the composition profile of C101 (upper right), a sudden increase the profiles of the temperature differences between adjacent trays.
in DCM mole fraction (light component) occurs at the same The right graphs in Figure 6 show the results of the open-loop
location. We can also find that the composition is dominated sensitivity test with 0.1% changes in RR1, RR2, and the reboiler
by of MeOH and DMF in the stripping section and recovery duty (QR). The result shows that the 16th tray in C101, the fifth
section, similar to a binary system. Thus, under a constant pressure, tray in C102, and the fifth tray in C103 have the largest slopes and
the composition can be indicated by the temperature. a relatively high sensitivities. Therefore, these trays are selected as
the temperature control trays. Figure 7 shows the overall control
4. CONTROLLABILITY OF THE E-DWC-HI PROCESS scheme of CS1.
As mentioned above, the E-DWC-HI process has great 1. The fresh feed is controlled by the flow controller FC1.
economic advantages and energy efficiency. However, in terms 2. The solvent is proportional to the feed with a fixed ratio
of dynamic controllability, this combination leads to the loss of of 0.48.
an important manipulated variable (duty of one reboiler). 3. Two reflux-drum levels and the base level are adjusted by
Another question is that the vapor split ratio is fixed at design by manipulating distillate flow rates and the solvent makeup flow
the location of the dividing wall.29 Therefore, a mismatch in the rate, respectively.
vapor split ratio is inevitable when the feed composition changes. 4. The pressures of two condensers are controlled by adjusting
Due to these, the dynamic performance of the E-DWC-HI is in condenser heat removal.
question, which will be studied in this section to seek a practical 5. Three temperature controllers (TC1, TC2, and TC3) are
and effective control structure. installed with 1 min dead time, and the corresponding manipulated
The column base and reflux drum are sized to have 10 min of variables are RR1, RR2, and the ratio of reboiler duty to feed flow
liquid holdup. Adequate pressure drops are provided on pumps rate (QR/F), respectively.
and control valves. After dynamic parameters are specified, 6. The vapor flow rate entering C102 is proportional to that
the E-DWC process is exported into Aspen Dynamics based on entering C101 with a fixed ratio (0.19/0.81).
the pressure-driven simulation. Then, four control strategies are The flow control loops have the same controller parameters of
constructed and tested in Aspen Dynamics. Considering that Kc = 0.5 and τmin = 0.3. All level control loops are proportional
the purchase and maintenance of component measurement is control only with Kc = 2. The pressure controllers have Aspen
usually expensive, and it has large measurement lags, an online default controller parameters of Kc = 20 and τmin = 12. As shown
1283 https://dx.doi.org/10.1021/acs.iecr.0c03876
Ind. Eng. Chem. Res. 2021, 60, 1274−1289
Industrial & Engineering Chemistry Research pubs.acs.org/IECR Article

Figure 16. Dynamic responses for CS3 under ±5% feed composition disturbances.

Table 3. Data of Temperature, Pressure, and Composition from the bottom of the dividing wall. However, as mentioned
above, there is no adjustment that can be made for the vapor split.
P2,5 (bar) xMeOH T2,5 (°C)
Thus, a mismatch in the vapor split ratio is unavoidable especially
0.974 0.54 82.75 when the feed composition changes. Therefore, adjusting the
1.453 0.76 82.54 vapor split is very important for the control strategy of E-DWC.
1.453 0.82 79.98 4.2. Double Temperature Difference Control Structure
1.854 0.94 82.53 (CS2). As mentioned above, CS1 has a poor dynamic perfor-
mance in the presence of feed composition disturbances. Recent
in Table 1, control parameters of the temperature controllers are papers30,39,40 show that a double temperature difference control
tuned using relay-feedback tests with the Tyreus−Luyben rule. structure (DTDC) can enhance the ability of resisting feed com-
Disturbances of ±20% feed flow rate and ±5% feed com- position disturbances. Thus, on the basis of CS1, a DTDC
position are introduced to test the dynamic performance of CS1. control structure is constructed in this section. TC1 and TC2 in
The setting value of FC1 is changed from 100 to 120 kmol/h or CS1 are replaced by DTDC1 and DTDC2 in CS2. TC3 is still
from 100 to 80 kmol/h at 0.5 h. The feed composition of DCM installed to manipulate the variable QR/F.30 The reference
is changed from 83 to 87% or from 83 to 79% at 0.5 h. stages for the two temperature control loops are selected by the
Figure 8 gives the dynamic response in the presence of ±20% averaged absolute variation magnitude (ASVM) method.39 For
flow rate disturbance. The results show that all three tray tem- a specified control loop, the distribution of changes in temper-
peratures return to the set values. Product purities of DCM and ature difference between the sensitive stage and the remaining
MeOH are also held close to the specification purity of 99.9%. trays are calculated under the assumption that the disturbances
Figure 9 shows the dynamic responses due to feed composition of the feed composition are completely rejected. The ASVM of
disturbances. The result shows that the product purities of DCM the nth tray is defined as the following equation.
and MeOH have a large offset when facing +5% feed composition
disturbance. Besides, the decrease of RR1 cannot make the tem- 2NC
1
perature return to its set value. Obviously, to hold this temper- ASVM = ∑ αi|Δ(Tn − TSS)i |
ature, more vapor needs to be input to the upper left side of C101 2NC i=1 (1)

Figure 17. Flow sheet equations for pressure compensation.

1284 https://dx.doi.org/10.1021/acs.iecr.0c03876
Ind. Eng. Chem. Res. 2021, 60, 1274−1289
Industrial & Engineering Chemistry Research pubs.acs.org/IECR Article

Figure 18. Control structure integrating pressure swing and pressure compensation.

where NC is the number of feed components; αi (i = 1, ..., NC) in CS1 and CS2, a control structure with pressure swing is
are the weighting coefficients, which represent the relative adapted from a new method that was presented in a recent
importance of rejecting each disturbance of feed composition. paper.29
Tn and TSS are the nth tray temperature and the selected sensitive So far, the two condensers on both sides of the dividing wall
stage temperature, respectively, in the presence of feed com- have operated at the same pressure. Suppose we maintain the
position disturbances. same operating pressure of the upper left condenser and reduce
Figure 10 shows the curves of the ASVMs under the assumption the operating pressure of the upper right condenser. This will
of completely rejecting ±5% feed composition disturbances. As deliver more vapor to the upper right side of the dividing wall
suggested by the published papers of Yuan et al.39 and Feng and less vapor to the upper left side. Therefore, on the basis of
et al.,40 the selection of the reference stages should start from the CS1, instead of manipulating RR1, the 16th tray temperature
sensitive stage. The closest stage with a local ASVM minimum can be controlled by adjusting the set point of the pressure
can capture the coupling from the control loop below or above controller in C102 (PC2). As shown in Figure 14, the reflux ratio
the sensitive stage, so it should be chosen as the reference stage. RR1 is fixed. TC1 has controller parameters of Kc = 0.82 and
As shown in Figure 10a, for the temperature of the sensitive τmin = 33.0 with the reverse control action. PC2 is on cascade
stage in C101 (T16‑C101), ASVM reaches a local minimum at the mode with the default controller parameters of Kc = 20 and τmin
21th stage of C101 which should be selected as the lower = 12. The LMTD (logarithmic mean temperature difference)
reference stage (LRS) of T16‑C101. To determine the upper option for condenser 2 is selected in Aspen, and the manipulated
reference stage (URS) of T16‑C101, the red horizontal dashed variable of PC2 is the cooling water flow rate with an inlet
line is drawn from the LRS to the left, and the intersection with temperature of 30 °C and a heat capacity of 4200 J/(kg·K).
the ASVM curve above the sensitive stage is closer to the 11th In order to increase the adjustment range of heat removal
tray. Thus, the 11th stage is selected as the URS of T16‑C101 for in condenser 2 and thereby ensure the adjustment range of the
the DTDC1. Similarly, as shown in Figure 10b, the fourth and pressure, the UA value (the product of the overall heat transfer
22th stages should be chosen as the URS and LRS of T5‑C102 for coefficient U and heat transfer area A) of condenser 2 is
the DTDC2 in C102. Thus, the double temperature difference increased to 1.5 times its steady state value, which increases the
of DTDC1 and DTDC2 can be expressed as Δ2T16 = (T21‑C101 − TAC slightly by 0.49%. The resulting flow rate of cooling water
T16‑C101) − (T16‑C101 − T11‑C101) and Δ2T5 = (T22‑C102 − is 4617 kg/h with an outlet temperature of 64.4 °C and a
T5‑C102) − (T5‑C102 − T4‑C102). The control structure of CS2 is required UA of 8.34 kW/K. It should be noted that, during the
shown in Figure 11. Table 2 gives the control parameters of the dynamic operation, UA is a fixed value. By manipulating the
temperature control loops. cooling water flow rate, the duty of condenser 2 is adjusted
The same disturbances as that in CS1 are introduced to test directly, while the pressure of C102 is controlled indirectly.
the dynamic performance of CS2. As shown in Figure 12, the Figures 15 and 16 give the dynamic responses in the presence
dynamic response in the presence of feed flow rate disturbances of flow rate and composition disturbances, respectively. All three
is similar to that of CS1: all two product purities of DCM and tray temperatures return close to their set points after the
MeOH can be held close to the specification purity of 99.9%. introduction of feed flow rate disturbances. More importantly,
Figure 13 gives the dynamic response due to feed composition all products are held close to their required purity specifications.
disturbances. The result shows that DCM product purity can be Stable control is achieved.
maintained, which is better than that in CS1. However, the As for feed composition disturbances, all products are close to
purity of MeOH still has a large deviation. their required purity specifications except MeOH product.
4.3. Control Structure with Pressure Swing (CS3). To Compared with CS1, the controllability of CS3 has been greatly
overcome the drawbacks of the fixed vapor split ratio mentioned improved. However, when the composition of DCM decreases,
1285 https://dx.doi.org/10.1021/acs.iecr.0c03876
Ind. Eng. Chem. Res. 2021, 60, 1274−1289
Industrial & Engineering Chemistry Research pubs.acs.org/IECR Article

Figure 19. Dynamic responses for CS4 under ±20% feed flow rate disturbances.

MeOH product purity is dropped from 99.9 to 97.85% with an Table 3 gives the concentration of MeOH under different
apparent static offset of 2.05%, which is unsatisfactory. A similar pressures and temperatures. The data is generated from the
deviation also appears in other published literature.29 It is worth T−xy graph in Aspen Plus. According to the data, the calculated
noting that, when the feed composition of DCM is reduced, a values of c1, c2, c3 and c4 are 0.0088, −0.0361, −0.2745, and
pressure slightly less than the ambient pressure is required (the 3.0912, respectively. As shown in Figure 17, flow sheet equations
bottom picture on the left side of Figure 16). Therefore, vacuum in Aspen Dynamics are used to implement the equations derived
equipment (such as a vacuum pump or steam-jet ejector) is above. The calculated variable (composition of MeOH) is the
required to assist in achieving this vacuum. input signal of DT2 block (a dead time block, 1 min dead time).
4.4. Control Structure Integrating Pressure Swing and Figure 18 shows the flow sheet of CS4. Instead of using TC2 to
Pressure Compensation (CS4). In this section, to overcome control the temperature of the fifth tray in C102 with RR2, a
the drawbacks of static deviation mentioned in CS3, a new pressure-compensated temperature controller (TPC) is employed
control structure with a pressure-compensated temperature to control the calculated composition. The TPC controller has the
control loop is proposed. Note that the temperature of the fifth controller parameters of Kc = 2.36 and τmin = 10.56 with the reverse
tray in C102 (T2,5) is used to infer the composition of the control action. In addition, as shown in Figure 6, the 13th stage may
recovery section. The estimation of composition may be not be chosen as another sensitive tray in C101. Thus, an additional
accurate if the tray pressure varies. temperature control loop (TC4) with the 13th stage temperature as
It can be seen from Figure 5 that the liquid phase composition a controlled variable and RR1 as a manipulated variable is added to
in C102 is mainly MeOH and DMF, so it can be treated as CS4. TC4 has the controller parameters of Kc = 0.79 and τmin =
a binary system. The relationship between temperature, pres- 18.48, direct control action.
sure, and composition can be expressed by the following The same disturbances of feed flow rate and composition are
equation:31,32 also introduced to examine the control performance of CS4.
As shown in Figures 19 and 20, we can find that all product
xMeOH = (c1P2,5 + c 2)T2,5 + c3P2,5 + c4 (2) purities can be driven back to the set values with smaller
transient deviations and shorter settle time, compared with CS3.
where xMeOH is the MeOH concentration of the fifth tray in When subjected to the disturbance of −5% DCM, the dynamic
C102; P2,5 is the pressure of the fifth tray in C102 (bar); T2,5 is performance of CS4 is significantly better than that of CS3.
the temperature of the fifth tray in C102 (°C); c1, c2, c3, and c4 are It can be seen from Figure 20 that the cooling water flow rate
correlation coefficients. increases significantly. This may be due to the relatively low vapor
1286 https://dx.doi.org/10.1021/acs.iecr.0c03876
Ind. Eng. Chem. Res. 2021, 60, 1274−1289
Industrial & Engineering Chemistry Research pubs.acs.org/IECR Article

Figure 20. Dynamic responses for CS4 under ±5% feed composition disturbances.

Table 4. Comparison of Dynamic Performances of Different The process is optimized by a sequential iterative procedure
Control Structures on the basis of the TAC economic index. Then, the heat
disturbances
integration between the fresh feed and the recycle solvent
stream is further performed. Compared with the CED process,
configurations feed flow rate feed composition
the optimized E-DWC-HI process has a better economic
set performance with an energy cost saving of 50.0% and a TAC
point +20% −20% +5% DCM −5% DCM
reduction of 27.7%.
CS1 DCM purity (%) 99.9 99.91 99.88 94.67 99.92
The controllability of this E-DWC-HI process is investigated
MeOH purity (%) 99.9 99.89 99.91 84.10 99.90
CS2 DCM purity (%) 99.9 99.90 99.88 99.97 99.92
by dynamic simulation. Four control structures are proposed
MeOH purity (%) 99.9 99.89 99.91 70.21 99.91 and tested in the presence of ±20% feed flow rate and ±5% feed
CS3 DCM purity (%) 99.9 99.89 99.91 99.91 99.87 composition disturbances. The results show that product quality
MeOH purity (%) 99.9 99.89 99.91 99.94 97.85 cannot be guaranteed by the basic control structure (CS1) and
CS4 DCM purity (%) 99.9 99.90 99.89 99.89 99.91 DTDC control structure (CS2), which have a fixed vapor split.
MeOH purity (%) 99.9 99.89 99.91 99.89 99.87 A mismatch in the vapor split ratio is inevitable when the feed
composition changes. The control structure with pressure
flow rate in C102 compared to that in C101. When the disturbance swing (CS3) is proposed to address this issue by adjusting the
of −5% DCM occurs, the corresponding increase of MeOH pressures on the two sides of the dividing wall. The result shows
composition is quite large, changing from 17 to 21 mol % (∼23.5%
that CS3 has a better dynamic performance than CS1 and CS2.
on a relative basis of 4/17). Thus, a more, large amount of vapor
needs to be input to C102. Correspondingly, the pressure on the However, there is still an obvious offset in MeOH product purity
top of C102 needs to be greatly reduced to provide sufficient in the presence of −5% feed composition disturbance. Then, an
pressure differential as a driving force; that is, a more, large amount improved control structure with a pressure-compensated feature
of cooling water is required. The comparison of dynamic (CS4) is developed. All product purities meet the required
performances of the four control structures is shown in Table 4. specifications with small transient deviations and a short settle
time despite throughput and feed composition disturbances.
5. CONCLUSION Therefore, the control structure integrating pressure swing
In this article, an energy-efficient E-DWC process is proposed and pressure compensation can provide an effective and robust
for the first time to separate DCM−MeOH azeotropic mixture. control for the E-DWC process.
1287 https://dx.doi.org/10.1021/acs.iecr.0c03876
Ind. Eng. Chem. Res. 2021, 60, 1274−1289
Industrial & Engineering Chemistry Research

■
pubs.acs.org/IECR Article

ASSOCIATED CONTENT NFS = solvent feed location

*
sı Supporting Information RRi = reflux ratio of column i
The Supporting Information is available free of charge at A = heat exchanger area
https://pubs.acs.org/doi/10.1021/acs.iecr.0c03876. QR = reboiler duty
ID = column diameter
RMSD comparison of different thermodynamic models LMTD = logarithmic mean temperature difference
using VLE data; binary parameters of NRTL model at UA = product of overall heat transfer coefficient U and heat
1 atm; basis of economics and equipment sizing; com- transfer area A
parison of economic results of CED, CED-IP, E-DWC, Ti,j = temperature of tray j in column i
and E-DWC-HI processes; vapor−liquid equilibrium and Pi,j = pressure of tray j in column i
T−xy diagrams for DCM−MeOH; thermodynamically TPC = pressure-compensated temperature controller
equivalent process; optimization procedure for E-DWC; TE = temperature measurement
flow sheet of CED-IP process; screenshots of control PT = pressure measurement

■
structures and control panel plots used in Aspen for
CS1−CS4 (PDF)
REFERENCES

■ AUTHOR INFORMATION
Corresponding Author
(1) Shao, S.; Gu, Z. Measurement and correlation of vapor-liquid
equilibria for dichloromethane-methanol-acetone-water system. Chem.
Eng. 2012, 40, 40−43.
Leping Dang − School of Chemical Engineering and (2) Han, S.; Yang, J.; Wang, A. Study on the separation of
Technology, Tianjin University, Tianjin 300072, People’s dichloromethane and methanol by extractive distillation. Appl. Chem.
Republic of China; orcid.org/0000-0003-1713-5422; Ind. 2018, 47, 1190−1194.
Email: dangleping@tju.edu.cn; Fax: +86-022-27400287 (3) Macmillan, D. S.; Murray, J.; Sneddon, H. F.; Jamieson, C.;
Watson, A. J. B. Replacement of dichloromethane within chromato-
Authors graphic purification: a guide to alternative solvents. Green Chem. 2012,
Cong Jing − School of Chemical Engineering and Technology, 14, 3016−3019.
Tianjin University, Tianjin 300072, People’s Republic of (4) Abtahi, M.; Mesdaghinia, A.; Yaghmaeian, K.; Nabizadeh, R.;
China Jaafarzadeh, N.; Rastkari, N.; Saeedi, R.; Nazmara, S.; Naddafi, K.
Jiaxing Zhu − School of Chemical Engineering and Technology, Dichloromethane emissions from automotive manufacturing industry
Tianjin University, Tianjin 300072, People’s Republic of in Iran: case study of the SAIPA automotive manufacturing company.
China Toxicol. Environ. Chem. 2013, 95, 757−764.
(5) Nassiri Koopaei, N.; Abdollahi, M. Health risks associated with the
Hongyuan Wei − School of Chemical Engineering and
pharmaceuticals in wastewater. Daru, J. Pharm. Sci. 2017, 25, 9.
Technology, Tianjin University, Tianjin 300072, People’s (6) Antibiotics Market-Growth, Trends and Forecast (2020−2025).
Republic of China Mordor Intelligence. https://www.mordorintelligence.com/industry-
Complete contact information is available at: reports/antibiotics-market (accessed 2020-02-01).
https://pubs.acs.org/10.1021/acs.iecr.0c03876 (7) Hosgor, E.; Kucuk, T.; Oksal, I. N.; Kaymak, D. B. Design and
control of distillation processes for methanol−chloroform separation.
Author Contributions Comput. Chem. Eng. 2014, 67, 166−177.
C.J. and J.Z. contributed equally to this work and should be (8) Li, X.; Zhao, Y.; Qin, B.; Zhang, X.; Wang, Y.; Zhu, Z.
Optimization of pressure-swing batch distillation with and without heat
considered co-first authors.The manuscript was written through
integration for separating dichloromethane/methanol azeotrope based
contributions of all authors. All authors have given approval to on minimum total annual cost. Ind. Eng. Chem. Res. 2017, 56, 4104−
the final version of the manuscript. 4112.
Notes (9) Yang, A.; Shen, W.; Wei, S.; Dong, L.; Li, J.; Gerbaud, V. Design
The authors declare no competing financial interest. and control of pressure-swing distillation for separating ternary systems

■ ACKNOWLEDGMENTS
This research did not receive any specific grant from funding
with three binary minimum azeotropes. AIChE J. 2019, 65, 1281−1293.
(10) Yu, H.; Ye, Q.; Xu, H.; Zhang, H.; Dai, X. Design and control of
dividing-Wall column for tert-butanol dehydration system via
heterogeneous azeotropic distillation. Ind. Eng. Chem. Res. 2015, 54,
agencies in the public, commercial, or not-for-profit sectors.

■
3384−3397.
(11) Wu, Y. C.; Lee, H.-Y.; Huang, H.-P.; Chien, I. L. Energy-saving
ABBREVIATIONS dividing-wall column design and control for heterogeneous azeotropic
DCM = dichloromethane distillation systems. Ind. Eng. Chem. Res. 2014, 53 (4), 1537−1552.
MeOH = methanol (12) Yang, A.; Sun, S.; Shi, T.; Xu, D.; Ren, J.; Shen, W. Energy-
DMF = dimethylformamide efficient extractive pressure-swing distillation for separating binary
PSD = pressure-swing distillation minimum azeotropic mixture dimethyl carbonate and ethanol. Sep.
TAC = total annual cost Purif. Technol. 2019, 229, 115817.
(13) Yi, C. C.; Shen, W.; Chien, I. L. Design and control of an energy-
AD = azeotropic distillation
efficient alternative process for the separation of methanol/toluene/
ED = extractive distillation water ternary azeotropic mixture. Sep. Purif. Technol. 2018, 207, 489−
E-DWC = extractive dividing-wall column 497.
CED = conventional extractive distillation (14) Feng, Z.; Shen, W.; Rangaiah, G. P.; Dong, L. Design and control
RMSD = root-mean-square deviation of vapor recompression assisted extractive distillation for separating n-
Vr = vapor split ratio hexane and ethyl acetate. Sep. Purif. Technol. 2020, 240, 116655.
NTi = total number of stages in column i (15) Li, L. M.; Guo, L. J.; Tu, Y. Q.; Yu, N.; Sun, L. Y.; Tian, Y. Y.; Li,
NFi = fresh feed location of column i W. S. Comparison of different extractive distillation processes for 2-

1288 https://dx.doi.org/10.1021/acs.iecr.0c03876
Ind. Eng. Chem. Res. 2021, 60, 1274−1289
Industrial & Engineering Chemistry Research pubs.acs.org/IECR Article

methoxyethanol/toluene separation: Design and control. Comput. (39) Yuan, Y.; Huang, K. Q.; Chen, H. S.; Zhang, L.; Wang, S. F.
Chem. Eng. 2017, 99, 117−134. Configuring Effectively Double Temperature Difference Control
(16) Luyben, W. L. Comparison of extractive distillation and pressure Schemes for Distillation Columns. Ind. Eng. Chem. Res. 2017, 56,
swing distillation for acetone/chloroform separation. Comput. Chem. 9143−9155.
Eng. 2013, 50, 1−7. (40) Feng, Z.; Shen, W.; Rangaiah, G. P.; Dong, L. Proportional-
(17) Chen, Y. C.; Yu, B. Y.; Hsu, C. C.; Chien, I. L. Comparison of integral control and model predictive control of extractive dividing-wall
heteroazeotropic and extractive distillation for the dehydration of column based on temperature differences. Ind. Eng. Chem. Res. 2018,
propylene glycol methyl ether. Chem. Eng. Res. Des. 2016, 111, 184− 57, 10572−10590.
195.
(18) Iqbal, A.; Ahmad, S. A.; Ojasvi. Design and control of an energy-
efficient alternative process for separation of Dichloromethane-
Methanol binary azeotropic mixture. Sep. Purif. Technol. 2019, 219,
137−149.
(19) Sun, L.; Wang, Q.; Li, L.; Zhai, J.; Liu, Y. Design and control of
extractive dividing wall column for separating benzene/cyclohexane
mixtures. Ind. Eng. Chem. Res. 2014, 53, 8120−8131.
(20) Li, Q.; Feng, Z.; Rangaiah, G. P.; Dong, L. Process optimization
of heat-integrated extractive dividing-wall columns for energy-saving
separation of CO2 and hydrocarbons. Ind. Eng. Chem. Res. 2020, 59
(23), 11000−11011.
(21) Wu, Y. C.; Hsu, P. H.; Chien, I. L. Critical assessment of the
energy-saving potential of an extractive dividing-wall column. Ind. Eng.
Chem. Res. 2013, 52, 5384−5399.
(22) Tututi-Avila, S.; Jimenez-Gutierrez, A.; Hahn, J. Control analysis
of an extractive dividing-wall column used for ethanol dehydration.
Chem. Eng. Process. 2014, 82, 88−100.
(23) Xia, M.; Xin, Y.; Luo, J.; Li, W.; Shi, L.; Min, Y.; Xu, C.
Temperature Control for Extractive Dividing-Wall Column with an
Adjustable Vapor Split: Methylal/Methanol Azeotrope Separation. Ind.
Eng. Chem. Res. 2013, 52, 17996−18013.
(24) Qian, X.; Jia, S.; Skogestad, S.; Yuan, X. Control structure
selection for four-product Kaibel column. Comput. Chem. Eng. 2016, 93,
372−381.
(25) Dwivedi, D.; Strandberg, J. P.; Halvorsen, I. J.; Preisig, H. A.;
Skogestad, S. Active Vapor Split Control for Dividing-Wall Columns.
Ind. Eng. Chem. Res. 2012, 51 (46), 15176−15183.
(26) Huaqiang, G.; Xiangwu, C.; Nan, C.; Wenyi, C. Experimental
study on vapour splitter in packed divided wall column. J. Chem.
Technol. Biotechnol. 2016, 91, 449−455.
(27) King, S. T.; Haas, J. Vapor and liquid flow control in a dividing
wall fractional distillation column. US 8562792 B2, 2013.
(28) Kang, K. J.; Harvianto, G. R.; Lee, M. Hydraulic Driven Active
Vapor Distributor for Enhancing Operability of a Dividing Wall
Column. Ind. Eng. Chem. Res. 2017, 56 (22), 6493−6498.
(29) Luyben, W. L. Vapor split manipulation in extractive divided-wall
distillation columns. Chem. Eng. Process. 2018, 126, 132−140.
(30) Feng, Z.; Shen, W.; Rangaiah, G. P.; Dong, L. Closed-loop
identification and model predictive control of extractive dividing-wall
column. Chem. Eng. Process. 2019, 142, 107552.
(31) Luyben, W. L. Distillation Design and Control Using Aspen
Simulation, 2nd ed.; John Wiley & Sons: 2013.
(32) Qiu, J.; Hua, T.; He, G.; Wu, X.; Shen, H.; Ling, H. Pressure-
compensated temperature control of Kaibel divided-wall column.
CIESC J. 2018, 69 (11), 4798−4813.
(33) Luo, C.; Yue, K.; Qiao, R.; Li, C.; Zhou, J.; Li, Q. Effect of ionic
liquids on the isobaric vapor-liquid equilibrium behavior of dichloro-
methane + methanol at 101.3 kPa. J. Mol. Liq. 2017, 243, 699−705.
(34) Tenn, F. G.; Missen, R. W. A Study of the Condensation of
Binary Vapors of Miscible Liquids: Part I: The Equilibrium Relations.
Can. J. Chem. Eng. 1963, 41, 12−14.
(35) Luyben, W. L. Principles and Case Studies of Simultaneous Design;
John Wiley & Sons: 2011.
(36) Douglas, J. M. The Conceptual Design of Chemical Processes;
McGraw-Hill: New York, 1988.
(37) Turton, R.; Bailie, R. C.; Whiting, W. B.; Shaelwitz, J. A.;
Bhattacharyya, D. Analysis, Synthesis and Design of Chemical Processes,
4th ed.; Prentice Hall: 2012.
(38) Luyben, W. L. Energy management in distillation preheat
systems. Chem. Eng. Process. 2020, 156, 108074.

1289 https://dx.doi.org/10.1021/acs.iecr.0c03876
Ind. Eng. Chem. Res. 2021, 60, 1274−1289