Article

pubs.acs.org/IECR

Design and Control of Extractive Dividing Wall Column for

Separating Benzene/Cyclohexane Mixtures
Lanyi Sun,* Qiuyuan Wang, Lumin Li, Jian Zhai, and Yuliang Liu
State Key Laboratory of Heavy Oil Processing, China University of Petroleum, Qingdao 266580, China
*
S Supporting Information

ABSTRACT: This paper considers the design and control of the separation of benzene and cyclohexane process via extractive
distillation in a dividing wall column. To aid the separation, furfural is used as the heavy boiling entrainer. The optimal ow sheet
with minimum energy requirements has been established using the multiobjective genetic algorithm with constrains. Three
control strategies are proposed: the basic control strategy uses four composition controllers, and two improved control strategies
with and without vapor split ratio use temperature controllers that are more practical in application than the basic control
strategy. The dynamic simulations reveal that the three control strategies can object to the feed disturbances well. The dynamic
responses of two improved control structures show that vapor split ratio as a manipulated variable can maintain the product
purities at their set points when a feed disturbance is introduced.

1. INTRODUCTION distillation is more available for the isolation of aromatics from

Petrochemical industries often require high-purity cyclohexane, pyrolysis gasoline (approximately 6590% aromatics); thus, it is
which is widely used not only in the production of nylon (adipic the main separation technique in industry to separate benzene
acid, caprolactam), paints, and varnishes but also in the synthesis and cyclohexane.35
of pharmaceutical intermediates. Moreover, cyclohexane is an High energy requirements, accompanied by high capital and
excellent solvent for resins, fats, paran oils, butyl rubber, etc. operating costs and process complexity, will be needed to meet such
As an important commodity chemical, cyclohexane is mainly a demand for high purity cyclohexane (99.0% minimum purity).
produced by the addition of hydrogen to benzene over a Ni or Pt Process intensication has received considerable attention all over
catalyst. The reaction is typically carried out at 573 K and in the the world in recent decades, due to the increasing awareness of the
pressure range 2025 atm with a large amount of heat released. limited energy resources.68 In order to obtain further energy saving,
The unreacted benzene, which is present in the reactors euent a novel conguration, which integrated two columns into one shell,
stream, must be removed to obtain high purity cyclohexane. is identied as a dividing wall column (DWC).9 Compared with
Separation process of benzene and cyclohexane is expensive conventional congurations, DWCs often provide more than 30%
and cumbersome by a conventional distillation process, reductions in capital costs and roughly 30% savings in energy costs.
considering the only 0.6 K dierence in the boiling points of Since two columns are integrated into one shell, DWCs can save
benzene and cyclohexane and the small interaction parameter more space required for installation as well as a higher purity of side
between them.1 The most commonly used methods for separa- product compared with conventional sequences.1013 Two-column
tion of aromatic/aliphatic mixtures are given in Table 1,2 which extractive distillation sequence integrated into one shell forms an
extractive dividing wall column (EDWC).
Table 1. Methods for Aromatic Recovery The rest of the paper is organized as follows. First, process ow
sheet and steady state are investigated by Aspen Plus. Next, the
requirements for basic or economical optimal ow sheet is determined by minimizing heat duty of the
process separation problem operation
overall system with the multiobjective genetic algorithm with
liquidliquid BTX separation from lower aromatic content (2065%) constrains. Then, three eective control strategies for this system
extraction reformate gasoline
are provided. Finally, the disturbance rejection capability of this
extractive BTX separation from medium aromatic content (6590%)
distillation pyrolysis gasoline process is evaluated for ow and composition variations by Aspen
azeotropic BTX separation from high aromatic content (>90%) Dynamics, and Conclusions are discussed in the last section.
distillation pyrolysis gasoline
crystallization isolation of p-xylene from distillative preseparation of o-xylene
m/p-mixtures and ethylbenzene from C8 aromatic
2. STEADY-STATE DESIGN AND OPTIMIZATION OF
fractions EXTRACTIVE DISTILLATION PROCESS
adsorption on isolation of p-xylene from continuous, reversible, and selective
solids C8 aromatic fractions adsorption
2.1. Entrainer Screening and Selection. Choosing an
eective solvent, which is a challenging issue to a successful
shows that the major technologies presently available to separate
aromatic/aliphatic mixtures are azeotropic distillation and Received: October 1, 2013
extractive distillation. If the aromatic content is greater than Revised: March 25, 2014
90%, such as with the pyrolysis of gasoline or crude benzene Accepted: April 12, 2014
from coal coking, azeotropic distillation is economical. Extractive Published: April 13, 2014

2014 American Chemical Society 8120 dx.doi.org/10.1021/ie500291a | Ind. Eng. Chem. Res. 2014, 53, 81208131
Industrial & Engineering Chemistry Research Article

separation, has a profound eect on the design of an extractive

distillation process. Several demanding rules should be followed
to get a suitable solvent for any close boiling mixture,14,15 e.g., the
selectivity, the volatility of solvent, its environmental impact, or
even the toxicity of solvent.16
Four entrainers for the benzenecyclohexane system have
been studied in the same conditions, and furfural is found to
be the best candidate. The main physical properties and the
simulation results with four dierent entrainers are given in
Figure 1 and Table 2, respectively. The results of calculating the

Figure 2. Residual curve map (RCM) of the benzenecyclohexane

furfural system at 101.3 kPa.

the problem of high energy requirements, the EDWC technique

is applied to the benzenecyclohexane system. Since there is no
o-shelf DWC unit in Aspen Plus, PRO/II, and other currently
available process simulators, the steady design of an EDWC always
uses a thermodynamically equivalent process for simulation, which
Figure 1. Eects of solvent on VLE of cyclohexane(1)benzene(2) is shown in Figure 3b. The main column (column I) is equivalent
system n(cyclohexane):n(benzene):n(entrainer) = a:(0.5 a):0.5.

Table 2. Comparison of the Main Physical Properties for

Dierent Entrainers
molecular molecular boiling density at 293 K,
solvent formula weight, g/mol point, K kg/m3
furfural C5H4O2 96.08 435 1160
DMSO C2H6OS 78.13 462 1100.4
DMF C3H7NO 73.09 425 948
sulfolane C4H8O2S 120.17 558 1261

relative volatility are comparable based on the same thermo-

dynamics model.17 Figure 1 shows that the dimethyl sulfoxide
(DMSO), N,N-dimethylformamide (DMF), and sulfolane
provide high selectivity, while DMSO and DMF are not taken
into consideration for their high prices and great toxicities.
Although a novel extractive distillation operating second recovery
column under vacuum pressure18 has been proposed to avoid the
high bottom temperature caused by sulfolane, the low temper-
ature in the condenser would result in a consequent increase in
the operation cost. In this paper, furfural will be used in the
separation of benzene and cyclohexane azeotropic system.
2.2. Residue Curve Map (RCM) for BenzeneCyclohexane
System. For distinguishing between feasible and infeasible
sequences, a residue curve map (RCM) technique is introduced Figure 3. (a) Optimum process ow diagram for the conventional
as a simple method in the benzenecyclohexanefurfural congurations. (b) Process ow diagram for the EDWC.
system. Figure 2 shows the RCM curves for this ternary system.
As can be seen in Figure 2, the benzenecyclohexane azeotrope to the rst column in Figure 3a. The rectifying column (column II)
is the unstable node, furfural is the stable node, and both benzene and the stripping column (column III) are equivalent to the
and cyclohexane are the saddles. Therefore, the products cannot rectifying section and stripping section of the entrainer recovery
be obtained simultaneously. There are no distillation boundaries column, respectively, as shown in Figure 3. Highly pure cyclo-
in the RCM, which indicates that the benzene/cyclohexane hexane is distilled at the top of the main column, the benzene is
mixtures can be separated easily in the presence of furfural. collected from the top of the side column, and the furfural of
2.3. Process Design and Sensitivity Analysis. 2.3.1. the stripping column is recycled back to the main column at an
Process Flow Sheet and Steady-State Design. In order to solve appropriate feeding location. Figure 3b (right) shows the process
8121 dx.doi.org/10.1021/ie500291a | Ind. Eng. Chem. Res. 2014, 53, 81208131
Industrial & Engineering Chemistry Research Article

ow diagram implemented in Aspen Plus, which also presents the

design parameters of this EDWC conguration.
Assume that the main column has a total of 30 stages, and
the rst stage is condenser. The stripping column consists of
10 stages, and the side column comprises 7 stages. The condenser
pressure of the main column is set at 101.325 kPa and the side
column is 106.325 kPa, which give reuxed drum temperatures
353.8 and 355.2 K, respectively. Thus, it is high enough to permit
the use of cooling water in the overhead condensers. At stage 19
(NF = 19) of the main column, a feed rate of 1000 kmol/h at
353.15 K is fed in. The recycle solvent subcooled to a temperature
of 343.15 K in the following simulations returns back to the stage
of 8, which is also a process design variable. The initial value of
the reux ratio for the main column is 4.2. The split fraction is
assumed to be 0.68 to achieve the desired products. 0.77 is chosen
as the reux ratio of the side column. Under these specications,
the purities are 99.3 and 99.8 wt % for cyclohexane and benzene,
respectively, and 99.9 wt % furfural can be obtained in the EDWC Figure 5. Eect of the reux ratio of the main column RR1 and vapor
(stripping section) bottom. NRTL is used in the simulations since split ratio V on the reboiler heat duty (QR).
it is capable of describing the physical properties of this system.5
2.3.2. Sensitivity Analysis. Because the reux ratio of the main 2.4. Optimization of Extractive Distillation Process.
column (RR1) and vapor split ratio (V) have signicant eects Multiobjective genetic algorithms applied to optimize the
on the composition proles and reboiler duty for an EDWC, dividing wall distillation columns have been published in recent
sensitivity analysis has to be carried out to seek the appropriate papers.19,20 However, there are only a few papers that applied
range of vapor split ratio and the optimum RR1. multiobjective genetic algorithm to optimize an EDWC process.
Figure 4 shows the inuence of the reux ratio of the main Evaluating of the objective function using a multiobjective genetic
column and vapor split ratio on the cyclohexane composition on algorithm with constrains, coupled with Aspen Plus21 as a design
tool, can obtain the rigorous results. Instead of obtaining one
optimal design, a set of optimal designs is obtained throughout
this procedure, which integrates the Pareto front. In this way, the
engineer can choose a trade-o by picking some points along
the Pareto front. In the optimization of multivariable functions,
the stochastic optimization methods present a reasonable com-
putational eort, and they just need calculations of the objective
function without problem reformulation.
As the thermodynamically equivalent of the EDWC, three
coupled RADFRAC units are used in Aspen Plus. The
manipulated variables include heat duty, reux ratio, total number
of stages, locations of the feed, extracting agent ow, and vapor
ow to the side column.
In terms of multiobjective optimization, there are three
objectives to minimize: the number of stages of the total
columns, the heat duty of the sequence, and the extracting agent
ow, which are in competition and constrained by the desired
purities and recoveries in each product stream, and therefore the
Figure 4. Eect of the reux ratio of the main column RR1 and vapor
split ratio V on the cyclohexane composition XD1 in the distillate of the
objectives must be optimized simultaneously. This problem can
EDWC. be expressed as follows:
min(Ni , Q R , FEA ) = f (Q R , RR i , Ni , NF, i , FEA , V2)
the top of the main column (XD1). It can be seen that the s.t.
cyclohexane composition is higher with the increase of vapor yk xk
split ratio at a higher reux ratio. Nevertheless, the value of
the vapor split ratio must be between 0.55 to 0.65 in case the where Ni is the number of stages of the column i, QR is the
excessive vapor ow back into the main column with benzene reboiler heat duty, FEA is the extracting agent ow, RRi is the
may cause a lower cyclohexane composition and energy waste. reux ratio, NF,i is the feed stage number of column i, and V2 is the
In addition, the vapor split ratio also has a great eect on reboiler value of the vapor stream owed into the side column. yk and xk
duty. A drop of the vapor split ratio decreases reboiler heat duty are the vectors of obtained and required purities and recoveries
as presented in Figure 5. This could be owed to the lower vapor for the k components, respectively. For the EDWC, 2000
ow returning into the main column that causes less liquid in the individuals and 40 generations are chosen as parameters of the
bottom to be vaporized. genetic algorithm, with 0.80 and 0.05 of crossover and mutation
Therefore, the EDWC may be operated at a reux ratio above fraction.
3.5 with vapor split ratio of 0.6 or higher to satisfy the 99.3% mass Here are the processes. First, a feasible initial design of the
specication for cyclohexane. EDWC is given as the original solution to the algorithm of
8122 dx.doi.org/10.1021/ie500291a | Ind. Eng. Chem. Res. 2014, 53, 81208131
Industrial & Engineering Chemistry Research Article

each run. The algorithm generates N individuals (i.e., new designs) algorithm, and the objectives of conventional conguration to
based on the initial solution to make up the rst population. The minimize are heat duties of the sequence, extracting agent ow,
manipulated variables of each of the N individuals are sent to and the total number of stages.
Aspen Plus to perform the simulation, and then Aspen Plus gives According to the recent paper by Wu et al.,23 since a heavy
the values of objective functions and constraints for each individual entrainer is introduced in the extractive distillation system, often
to the algorithm. The population is divided into subpopulations cases show that the reboiler duty is reduced but with adversely
in terms of the number of satised constraints with the increasing total steam cost that has a strong eect on the total
retrieved information; at this time, the best individuals satisfy annual cost (TAC). Hence, the TAC of the conventional
the c constraints, followed by those individuals that reach c 1 two-column system and the EDWC are compared in Table 3.
constraints, etc. Inside each subpopulation, the individuals are
ranked according to the value of the tness function. The original Table 3. Comparison between the Conventional Two-
objective functions can be optimized through the classication of Column Design and the EDWC Design
the population, which can also minimize the dierence between
the required and obtained constraints (recoveries and purities). conventional two-column
Finally, a set of optimal designs of the EDWC is obtained. More congurations C1 C2 EDWC
detailed information about this algorithm and its link to Aspen reboiler duty (Gcal/h) 16.8 17.5 26.9
Plus can be found in the original work.21 condenser duty (Gcal/h) 8.5 15.1 18.2
Figure 6 shows the Pareto front for the benzene/cyclohexane required steam type low-pressure medium- medium-pressure
mixtures, which includes the objectives to minimize: heat duty pressure
(50 psig, (150 psig, (150 psig, 458 K)
421 K) 458 K)
total reboiler duty (Gcal/h) 34.3 (0%) 26.9 (21.6%)
(% dierence)
total steam cost (1000 $/y) 4839.3 (0%) 4750.9 (1.8%)
(% dierence)
total operating cost (1000 $/y) 4914.0 (0%) 4797.1 (2.4%)
(% dierence)
total capital cost (1000 $/y) 1607.2 (0%) 1411.0 (12.2%)
(% dierence)
total annual cost (1000 $/y) 6521.1 (0%) 6208.1 (4.8%)
(% dierence)

The results show that the TAC of EDWC is 4.8% less than that
of the conventional design, and the steam cost is also reduced
by 1.8%. Needing only one column is the additional advantage
of this EDWC design conguration, which reduces the space
requirement. To a certain extent, the separation of benzene/
cyclohexane mixtures can benet from the EDWC conguration.
Figure 6. Pareto front of the EDWC for benzene/cyclohexane mixtures. Figures 79 show the temperature and composition proles of
the three columns for the ow sheet, respectively. There is a rapid
rise in the temperature for stage 8 and a rapid fall for stage 19
of the sequence, extracting agent ow, and the total number of
stages. At the end, 20 optimal designs are observed that make up
the Pareto front, which indicated that a dividing wall distillation
column can perform an extractive separation. These optimal
designs satisfy the specied purities and recoveries with the lowest
energy possible. In our work, the energy requirements is the
criterion for choosing our particular needs; thus the design
with the minimum reboiler duty is chosen as the nal design.
The number of stages for the main column, the side column, and
the stripping column are 32, 10, and 15, respectively. Bravo-Bravo
et al.22 reported that the classical dividing wall column with Figure 7. Main column for (a) temperature prole and (b) composition
symmetry on both sides of the wall may not correspond to the prole.
best scheme with lower energy requirements; thus, the results
indicate that the use of sizing on both sides of the wall can be an
optimization variable for the optimal design of EDWC. The
recycle solvent returns back to stage 9 (NFS = 9), and the feed
ows to stage 19 (NF = 19). 3.80 is chosen as the reux ratio
of the main column. For the side column, the value of the reux
ratio is set to 0.74. In addition, the V2 is 1046.9 kmol/h, and
the extracting agent ow is 2610 kmol/h. Under the above
conditions, the reboiler heat duty is 26.9 Gcal/h, which presents
approximately 22% savings of the total reboiler duty compared
with the heat duty of the conventional conguration shown in Figure 8. Side column for (a) temperature prole and (b) composition
Figure 3a, which is also optimized using the multiobjective genetic prole.

8123 dx.doi.org/10.1021/ie500291a | Ind. Eng. Chem. Res. 2014, 53, 81208131

Industrial & Engineering Chemistry Research Article

Table 4. Controller Tuning Parameters

controlled manipulated controller gain controller integral
variable variable KC (%) time I (min)
CC1 y9(Ben) S 0.68 27.0
CC2 x15(Fur) QR 48 22.0
CC3 y32(Cyc) v 0.47 42.5
CC4 x1(Ben) L 28 49.6

Figure 9. Stripping column for (a) temperature prole and (b)

composition prole. 3. CONTROL OF EXTRACTIVE DIVIDING WALL
COLUMN
3.1. Exporting to Aspen Dynamics. To test the control
strategies of this system, pressure-driven simulation in Aspen
Dynamics is applied. Before exporting the Aspen Plus steady-
state simulation to Aspen Dynamics, the tray-sizing option in
Aspen Plus is used to determine the needed size of equipment,
such as diameters of the three columns, weir height, base, and
reux drum of each column.
The base and reux drum volumes of each column are all sized
to give 10 min holdup with 50% liquid level. Pumps and valves
are almost specied to have pressure drops of about 3 atm with
the valve half open to handle changes in ow rates.
At this point, all equipment has been sized. There are two
items remaining to consider. First, the pressure of the feed
stream leaving valve must be exactly equal to the pressure in the
stage where it is fed in. Second, there is not any real valve or
compressor in an EDWC inside, and the ctitious valves and
pumps are to produce the pressure drop for the requirement of
Aspen Dynamics. Then it is ready to export the developed
steady-state simulation to Aspen Dynamics,
3.2. Basic Control Strategy (CS1). 3.2.1. State for Basic
Control Structure (CS1). The basic control structure for this
Figure 10. (a) Basic control structure CS1 for the EDWC. (b)
system is proposed in a series of works.2426 The basic control
Controller faceplates. structure CS1 and the corresponding controller faceplates are
shown in Figure 10.
in Figure 7a, at which the entrainer and fresh feed are fed in, The basic control schemes are given as below.
respectively. It is obvious that stage 16 displays a fairly steep slope (1) The fresh feed to the main column is controlled by
for the temperature. In Figures 8a and 9a, two steep slopes are manipulating the ow rate (reverse acting).
found near the 9th and 3rd stages, respectively, and the prole (2) The base level in the stripping column is held through
distinguishing features indicate that these stages are proper controlling the makeup ow rate of furfural (reverse
temperature control points for the corresponding columns. acting).

Figure 11. Dynamic responses to feed disturbances for the CS1: 20% in feed ow rate.

8124 dx.doi.org/10.1021/ie500291a | Ind. Eng. Chem. Res. 2014, 53, 81208131

Industrial & Engineering Chemistry Research Article

Figure 12. Dynamic responses to feed disturbances for the CS1: 10% in benzene composition.

(3) The base levels in the main column and side column are
held by manipulating the respective bottom ow rate
(direct acting).
(4) The reux drum levels in the main column and side
column are held by manipulating the corresponding
distillates ow rate (direct acting).
(5) The total recycle solvent ow rate to the fresh feed ow
rate ratio is added.
(6) The top pressures of the main column and side column are
controlled by adjusting the corresponding condenser heat
removal rate (reverse acting).
(7) The reux ratio in the main column is xed.
(8) The feed temperature of recycle entrainer is controlled by
manipulating the cooler heat removal (reverse acting).
(9) The solvent composition in the bottom of the stripping
column is controlled by the reboiler duty, which is ratioed
to feed ow rate forming a feed-forward controller.
(10) The composition of benzene in the distillate of the side
column is controlled by manipulating the corresponding
reux ow rate (reverse acting).
(11) The vapor composition of benzene on the feed stage is
controlled using the CC1 controller by adjusting the S/F
ratio. The S/F ratio is on cascade since it obtains the set
point signal from the CC1 controller. Figure 13. (a) Improved control structure CS2 for the EDWC; (b)
(12) The vapor sidestream (V2) molar ow rate is controlled controller faceplates.
through adjusting the valve V7 (reverse acting).
(13) The vapor sidestream molar ow rate is ratioed to the
during the operation of the EDWC. It should be noticed that
reboiler heat input. The FC3 controller is on cascade for its adjusting the vapor split ratio is impractical in the chemical
set point signal received from the VR/QR ratio. industry now for the EDWCs. The normal settings of the
(14) The cyclohexane composition in the vapor sidestream proportionalintegral (PI) type power-ow controller are KC =
is controlled using the CC3 controller by adjusting the 0.5 and I = 0.3 min. All level loops are P-only and the KC is 2, and
VR/QR ratio (reverse acting). Since the set point signal the pressure controller is PI with its default values. Considering
of the VR/QR ratio comes from the CC3 controller, the measurement and actuator lags in any real physical system,
VR/QR ratio is on cascade. a 3 min dead time is inserted into the corresponding composition
(15) Reboiler heat duty to feed ow (QR/F) ratio is added to control loops. To determine the ultimate gains and integral
this system to change reboiler heat input immediately time constant for the four composition controllers, the Tyreus
when feed uctuates. Luyben tuning is used, and the tuning results are listed in Table 4.
Many control loops described above are typical distillation For the detailed process of tuning and setting the reader is
control structures. The bottom level of the stripping column referred to Luybens book.29
is held by the entrainer makeup ow suggested by Grassi27 and 3.2.2. Performance of Basic Control Structure CS1. Two
Luyben.28 types of disturbances are used to test the proposed control
The CC3 controller is set to adjust the vapor split ratio when structure performance, namely, 20% changes in fresh feed ow
the cyclohexane composition in the vapor sidestream changes rate and 10% changes in fresh feed benzene composition.
8125 dx.doi.org/10.1021/ie500291a | Ind. Eng. Chem. Res. 2014, 53, 81208131
Industrial & Engineering Chemistry Research Article

Figure 14. Dynamic responses to feed disturbances for the CS2: 20% in feed ow rate.

The simple PI composition control actually works quite well relatively long time to settle is not good for the controllability and
for the benzene and cyclohexane system. As the feed-forward will lead to economic losses.
controller QR/F ratio is added for the control system, it can oer 3.3. Improved Control Strategy (CS2). Taking into
an immediate adjustment in reboiler heat input when subjected account the applicability and cost, the usage of basic control
to 20% changes in feed ow rate. structure CS1 with only composition controllers will be limited.
Results of the basic control structure for disturbances in both Comparing with composition control, temperature control brings
feed ow rate and feed composition are illustrated in Figures 11 in shorter drag time and costs less in both equipment and
and 12. Although the system reaches a new steady state at last, maintenance, and few articles on temperature control of EDWC
the basic control structure presents relatively large overshooting could be found in the open literature. Thus, the use of temperature
and settles after a long time for the two kinds of disturbances, control instead of direct composition will be explored in this
especially for benzene composition decreased by 10%. The section.
8126 dx.doi.org/10.1021/ie500291a | Ind. Eng. Chem. Res. 2014, 53, 81208131
Industrial & Engineering Chemistry Research Article

Figure 15. Dynamic responses to feed disturbances for the CS2: 10% in benzene composition.

After an amount of structures are evaluated by trial and error,

an improved control structure is proposed. Figure 13 shows the
improved control structure CS2 for the EDWC and its controller
faceplates.
The dierence from the CS1 is shown as follows:
(1) The stage 16 temperature of the main column and the
stage 9 temperature of the side column are controlled
using the TC16 and TC9 controllers by adjusting the RR1
and RR2, respectively.
(2) The temperature of stage 3 in the stripping column is
controlled using the TC3 controller by adjusting the
VR/QR ratio (reverse acting).
(3) The vapor sidestream molar ow rate V2 is ratioed to the
reboiler heat input. Because the set-point signal of the Figure 16. Improved control structure CS3 for the EDWC.

8127 dx.doi.org/10.1021/ie500291a | Ind. Eng. Chem. Res. 2014, 53, 81208131

Industrial & Engineering Chemistry Research Article

Figure 17. Dynamic responses to feed disturbances for the CS3: 20% in feed ow rate.

FC3 controller is received from the VR/QR ratio, the FC3 3.4. Improved Control Strategy without Vapor Split
controller is on cascade. Ratio (CS3). The control structure with or without vapor split
(4) The total recycle solvent ow rate to the fresh feed ow has become the hot topic.3033 On the one hand, the vapor split
rate ratio is added to control the total recycle solvent ow ratio can keep the products on high purities when feed
rate, and the FC2 is on cascade. disturbances are introduced. On the other hand, the control of
vapor split ratio would be dicult to manipulate in a real process.
Figure 14 shows the dynamic responses for 20% changes in
fresh feed rate at the time of 1 h. The product composition settles Thus, we study the improved control structure for the case when
in less than 5 h for this system, and steady-state osets are quite the vapor split ratio is not available as a degree of freedom. Then a
small for the products. new control structure without vapor split ratio is put forward
For 10 mol % benzene feed composition changes, product based on the CS2, and the corresponding ow sheet is present in
compositions settle in about 5 h as shown in Figure 15. Figure 16. The main dierences are that the QR/F ratio controls
8128 dx.doi.org/10.1021/ie500291a | Ind. Eng. Chem. Res. 2014, 53, 81208131
Industrial & Engineering Chemistry Research Article

Figure 18. Dynamic responses to feed disturbances for the CS3: 10% in benzene composition.

the stage 3 temperature instead of the furfural composition of the in products purity than control structure CS3, which indicate
stripping column and the vapor split ratio is not the manipulate that the control of vapor split ratio is good for maintaining the
variable compared with the CS2. Figures 17 and 18 give the product purities at their set points when a feed disturbance
responses to feed disturbances for the CS3, which are similar to occurs. However, considering the impossibility of adjusting
those of the CS2, and the purities of both products are brought vapor split ratio in the chemical industry at present, the control
close to the set points. structure CS3 may be the preferred choice to separate benzene
It is interesting to compare the responses of the three and cyclohexane via EDWC.
alternative control structures for the EDWC to feed ow rate and
feed composition disturbances. The improved control structures 4. CONCLUSIONS
CS2 and CS3 have lower peak transients and shorter settle The design and control of the benzene and cyclohexane
times than the basic control structure CS1. The short settle time separation process implemented in the EDWC are thoroughly
is due to temperature controllers, which are characterized by investigated in this paper. To optimize the ow sheet, the
low price, good reliability, and rapid response compared with multiobjective genetic algorithm with constrains is used, which
the composition controllers. Moreover, the dynamic response of minimizes the overall heat duty with desired product composi-
control structures with vapor split ratio show better performance tions. From the optimal results, it is found that the optimum ow
8129 dx.doi.org/10.1021/ie500291a | Ind. Eng. Chem. Res. 2014, 53, 81208131
Industrial & Engineering Chemistry Research Article

sheet of EDWC can achieve up to 22% savings of the total NF,i = feed stage number of column i
reboiler duty; meanwhile, the steam cost and the TAC are Ni = the number of stages of column i
reduced by 1.8% and 4.8%, respectively. After the investigation of PI = proportionalintegral
the benzenecyclohexane system thoroughly, we highly approve QR = reboiler heat duty
the view reported by Wu et al.,23 namely, that the dividing-wall RRi = the reux ratio of column i
column is a promising technology to save energy and the space S = recycle solvent ow rate
required for separating ternary azeotrope, while the energy- VR = vapor sidestream ow rate
saving potential is limited. Thus, the total reboiler duty, the steam V2 = vapor stream owed into the side column
cost, and the TAC should be checked carefully in the design of an XD1 = cyclohexane composition on the top of the main
EDWC conguration. column
As for the overall control strategy of the process, it is found x1 = mole fraction of cyclohexane in liquid phase
that the basic control strategy can work well during 20% in the x1 = benzene composition in the side column of stage 1
feed ow rate and 10% in the benzene composition changes. x15 = furfural composition in the stripping column of stage 15
However, the relatively large overshooting and long time to y1 = mole fraction of cyclohexane in vapor phase
settle are not good for the controllability. In addition, there is a y9 = benzene composition in the main column of stage 9
limit in the application of using four composition controllers in y32 = cyclohexane composition in the main column of stage 32
industry. Thus, the improved control structures CS2 and CS3 are V = vapor split ratio
recommended, which can eectively handle the disturbances V = latent heat of the steam
and keep the two products on high purities. And the CS3 using I = controller integral time constant

temperature control without vapor split ratio is more practical in
industry than the CS2 and CS1. All the performances of the REFERENCES
control structures reveal that robust control can be achieved for
(1) Garcia-Villaluenga, J. P.; Tabe-Mohammadi, A. A Review on the
the extractive distillation in a dividing wall column, and the CS3
Separation of Benzene/Cyclohexane Mixtures by Pervaporation
is worth considering in the separating of benzene/cyclohexane Processes. J. Membr. Sci. 2000, 169, 159.
mixtures.

(2) Arpe, H.-J.; Weissermel, K. Industrial Organic Chemistry; Wiley-

VCH: New York, 2010.
ASSOCIATED CONTENT (3) Berg, L. Separation of Benzene from Close Boiling Hydrocarbons
*
S Supporting Information by Extractive Distillation. U.S. Patent 5,458,741, 1995.
Detailed calculation procedures of TAC. This material is (4) Yin, W.; Ding, S.; Xia, S.; Ma, P.; Huang, X.; Zhu, Z. Cosolvent
available free of charge via the Internet at http://pubs.acs.org. Selection for BenzeneCyclohexane Separation in Extractive Distil-

lation. J. Chem. Eng. Data 2010, 55, 3274.

AUTHOR INFORMATION (5) Kumar, U. A.; Mohan, R. Quinary and Eight-Component Liquid
Liquid Equilibria of Mixtures of Alkanes, Aromatics, and Solvent
Corresponding Author (Furfural). J. Chem. Eng. Data 2013, 58, 2194.
*L.S. Tel.: +86 13854208340. Fax: +86 0532 86981787. E-mail: (6) An, W. Z.; Yuan, X. G. A Simulated Annealing-Based Approach to
sunlanyi@upc.edu.cn. the Optimal Synthesis of Heat-Integrated Distillation Sequences.
Notes Comput. Chem. Eng. 2009, 33, 199.
The authors declare no competing nancial interest. (7) Errico, M.; Rong, B. G.; Tola, G.; Spano, M. Optimal Synthesis of

Distillation Systems for Bioethanol Separation. Part 2. Extractive

Distillation with Complex Columns. Ind. Eng. Chem. Res. 2013, 52, 1620.
ACKNOWLEDGMENTS (8) Kiss, A. A.; Suszwalak, D. J. P. C. Enhanced Bioethanol
This work was supported by the National Natural Science Dehydration by Extractive and Azeotropic Distillation in Dividing-
Foundation of China (Grant 21276279) and the Development of Wall Columns. Sep. Purif. Technol. 2012, 86, 70.
key technologies project of Qingdao Economic and Techno- (9) Kiss, A. A. Advanced Distillation Technologies: Design, Control and
logical Development Zone under Grant 2013-1-57. Further the Applications; John Wiley & Sons, Ltd.: Chichester, 2013.
authors are grateful to the editor and the anonymous reviewers (10) Kiss, A. A.; Bildea, C. S. A Control Perspective on Process
for their helpful comments and constructive suggestions with Intensification in Dividing-Wall Columns. Chem. Eng. Process. 2011, 50,
regard to the revision of the paper. 281.

(11) Dejanovic, I.; Matijasevic, Lj.; Olujic, Z . Dividing Wall Column-A

Breakthrough Towards Sustainable Distilling. Chem. Eng. Process. 2010,
NOMENCLATURE 49, 559.
Ben = benzene (12) Asprion, N.; Kaibel, G. Dividing Wall Columns: Fundamentals
Cyc = cyclohexane and Recent Advances. Chem. Eng. Process. 2010, 49, 139.
Cs = the saturated steam price (13) Kiss, A. A.; Ignrat, R. M. Innovative Single Step Bioethanol
DMF = N,N-dimethylformamide Dehydration in an Extractive Dividing-Wall Column. Sep. Purif. Technol.
DMSO = dimethyl sulfoxide 2012, 98, 290.
DWC = dividing wall column (14) Rodriguez-Donis, I.; Gerbaud, V.; Joulia, X. Entrainer Selection
Rules for the Separation of Azeotropic and Close-Boiling-Temperature
EDWC = extractive dividing wall column
Mixtures by Homogeneous Batch Distillation Process. Ind. Eng. Chem.
F = fresh feed ow rate
Res. 2001, 40, 2729.
FEA = the extracting agent ow (15) Chen, B.; Lei, Z.; Li, Q.; Li, C. Application of CAMD in Separating
Fur = furfural Hydrocarbons by Extractive Distillation. AIChE J. 2005, 51, 3114.
KC = controller gain (16) Hsu, K. Y.; Hsiao, Y. C.; Chien, I. L. Design and Control of
L = reux ow rate in the rectifying column Dimethyl Carbonate Methanol Separation via Extractive Distillation in
NF = feeding location of the fresh feed the Dimethyl Carbonate Reactive-Distillation Process. Ind. Eng. Chem.
NFS = feeding location of the recycle solvent stream Res. 2009, 49, 735.

8130 dx.doi.org/10.1021/ie500291a | Ind. Eng. Chem. Res. 2014, 53, 81208131

Industrial & Engineering Chemistry Research Article

(17) Tang, C. M.; Li, X. L. Computer-Assisted Choice of Extractive

Solvent in the Distillation of Acetic-H2O. Chin. J. Appl. Chem. 2006, 23,
1066.
(18) Qin, J.; Ye, Q.; Xiong, X.; Li, N. Control of Benzene-Cyclohexane
Separation System via Extractive Distillation Using Sulfolane as
Entrainer. Ind. Eng. Chem. Res. 2013, 52, 10754.
(19) Gomez-Castro, F. I.; Segovia-Hernandez, J. G.; Hernandez, S.;
Gutierrez-Antonio, C.; Briones-Ramrez, A. Dividing Wall Distillation
Columns: Optimization and Control Properties. Chem. Eng. Technol.
2008, 31, 1246.
(20) Gomez-Castro, F. I.; Rodrguez-A ngeles, M. A.; Segovia-
Hernandez, J. G.; Gutierrez-Antonio, C.; Briones-Ramrez, A. Optimal
Designs of Multiple Dividing Wall Columns. Chem. Eng. Technol. 2011,
34, 2051.
(21) Gutierrez-Antonio, C.; Briones-Ramrez, A. Pareto Front of Ideal
Petlyuk Sequences Using a Multiobjective Genetic Algorithm with
Constraints. Comput. Chem. Eng. 2009, 33, 454.
(22) Bravo-Bravo, C.; Segovia-Hernandez, J. G.; Gutierrez-Antonio,
C.; Duran, A. L.; Bonilla-Petriciolet, A.; Briones-Ramrez, A. Extractive
Dividing Wall Column: Design and Optimization. Ind. Eng. Chem. Res.
2010, 49, 3672.
(23) Wu, Y. C.; Hsu, P. H. C.; Chien, I. L. Critical Assessment of the
Energy-Saving Potential of an Extractive Dividing-Wall Column. Ind.
Eng. Chem. Res. 2013, 52, 5384.
(24) Luyben, W. L.; Chien, I. L. Design and Control of Distillation
Systems for Separating Azeotopes; John Wiley & Sons: New York, 2010.
(25) Xia, M.; Yu, B.; Wang, Q.; Jiao, H.; Xu, C. Design and Control of
Extractive Dividing-Wall Column for Separating Methylal-Methanol
Mixture. Ind. Eng. Chem. Res. 2012, 51, 16016.
(26) Wang, Q.; Yu, B.; Xu, C. Design and Control of Distillation
System for Methylal/Methanol Separation. Part 1: Extractive
Distillation Using DMF as an Entrainer. Ind. Eng. Chem. Res. 2012,
51, 1281.
(27) Grassi, V. G. Practical Distillation Control; Van Nostand Reinhold
Press: New York, 1992.
(28) Luyben, W. L. Plantwide Control of an Isopropyl Alcohol
Dehydration Process. AIChE J. 2006, 52, 2290.
(29) Luyben, W. L. Plantwide Dynamic Simulators in Chemical
Processing and Control; Marcel Dekker: New York, 2002.
(30) Zhang, H.; Ye, Q.; Qin, J.; Xu, H.; Li, N. Design and Control of
Extractive Dividing-Wall Column for Separating Ethyl Acetate
Isopropyl Alcohol Mixture. Ind. Eng. Chem. Res. 2013, 53, 1189.
(31) 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.
(32) 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, 15176.
(33) Dwivedi, D.; Halvorsen, I. J.; Skogestad, S. Control Structure
Selection for Three-Product Petlyuk (Dividing-Wall) Column. Chem.
Eng. Process. 2013, 64, 57.

8131 dx.doi.org/10.1021/ie500291a | Ind. Eng. Chem. Res. 2014, 53, 81208131