2696 Ind. Eng. Chem. Res.

2008, 47, 2696-2707

Comparison of Extractive Distillation and Pressure-Swing Distillation for

Acetone-Methanol Separation
William L. Luyben*
Department of Chemical Engineering, Lehigh UniVersity, Bethlehem, PennsylVania 18015

Two of the most common methods for separating a binary homogeneous azeotrope are pressure-swing
distillation and extractive distillation. The former is effective if the composition of the azeotrope changes
significantly with pressure. The latter method is effective if a suitable solvent can be found. This paper compares
the steady-state design and the dynamic control of these two methods when applied to the acetone-methanol
binary system. The minimum-boiling azeotrope at 1 atm contains 77.6 mol % acetone at 328 K. At 10 atm
the azeotropic composition is 37.5 mol % acetone at 408 K, so pressure-swing separation is feasible. Extractive
distillation is also feasible using water as the solvent. Both systems require two distillation columns. Purities
of the two products are set at 99.5 mol %. Results show that the extractive distillation system has a 15%
See https://pubs.acs.org/sharingguidelines for options on how to legitimately share published articles.

lower total annual cost. However, a third component (water) is introduced that appears as trace impurities in
both the acetone and methanol products. It is also much more difficult to attain higher purities in the extractive
distillation system than in the pressure-swing system because of ternary vapor-liquid equilibrium constraints.
The dynamic controllabilities of the two alternative processes are quite similar. Steady-state designs and
Downloaded via FH FRANKFURT on July 30, 2025 at 13:21:42 (UTC).

control structures are also developed for the two methods when the columns are heat integrated. Heat integration
is straightforward in the pressure-swing system because the condenser temperature in the high-pressure column
is 60 K higher than the base temperature in the low-pressure column. In the extractive distillation system, the
pressure in the second solvent recovery column must be increased from 1 to 5 atm to provide the necessary
temperature differential driving force.

1. Introduction dynamics and control. Kossack et al.10 presented a systematic

synthesis framework for extractive distillation systems.
Azeotropic distillation is widely applied in industry and has The purpose of the present paper is to compare two alternative
been widely studied in the literature. Most papers consider only distillation separation methods for the acetone-methanol sys-
steady-state design, which is a challenging problem because of tem, which forms a homogeneous minimum-boiling azeotrope.
the high degree of nonlinearity, distillation boundaries, and The comparison will be in terms of both steady-state design
heterogeneous liquid-liquid equilibrium. One of the earliest and dynamic control. Systems with and without heat integration
studies of extractive distillation was the work of Grassi.1 Chien are considered.
and co-workers2-4 have studied both extractive and azeotropic
distillation systems for the separation of isopropyl alcohol and
water. Both design and control issues are presented, as well as 2. Acetone-Methanol Separation using Pressure-Swing
experimental results. The control structure they developed for Distillation
the heterogeneous azeotropic system uses an “inverse double- Figure 1A gives Txy diagrams for the acetone-methanol
loop” strategy in the azeotropic column in which multiple system at pressures of 1 and 10 atm. The large shift in the
temperatures are used. Luyben5 developed a control structure composition from 77.6 to 37.5 mol % acetone indicates that a
for the ethanol-water-benzene heterogeneous system in which pressure-swing system should be feasible. The Uniquac
only one temperature in each column is controlled. An interest- physical property package is used in the Aspen Plus simulation.
ing counterintuitive aqueous level control loop on the decanter Knapp and Doherty9 used the van Laar method, but the
is described. Plantwide control of an isopropyl alcohol extractive azeotropic search in Aspen Plus using van Laar predicts that
distillation system has also been studied.6 there are no azeotropes. Therefore, the Uniquac method is
One of the earliest papers7 that considers both design and used in this work. Notice that Figure 1A shows an unusual
control studied the use of pressure-swing azeotropic distillation behavior in the pure component boiling points. At low pressure,
to separate a binary homogeneous azeotropic mixture. The acetone is the lower-boiling component. At high pressure,
numerical example was the tetrahydofuran (THF)-water system. methanol becomes the lower-boiling component. Figure 1B
This work was extended in a recent paper8 in which the tradeoff gives a residue curve map for the acetone-methanol-water
between steady-state economics and dynamic controllability was system, which is useful in the design of the extractive distillation
demonstrated. system.
Knapp and Doherty9 studied heat-integration of binary 2.1. Steady-State Design. Figure 2 shows the flowsheet
homogeneous azeotropic systems using extractive distillation with stream information, heat duties, equipment sizes, and
methods. One of their examples considered the acetone- operating conditions. The design is heuristically optimized
methanol system with water as the solvent. They did not by running the rigorous simulation with an increasing number
consider pressure-swing distillation, nor did they consider of stages until the reflux ratio stopped decreasing, which gives
an approximation of the minimum reflux ratio. These runs are
* E-mail: [email protected]. Phone: (610) 758-4256. Fax: (610) made with the distillate and bottoms compositions held at their
758-5057. specified values. In the low-pressure column, the bottoms
10.1021/ie701695u CCC: $40.75 © 2008 American Chemical Society
Published on Web 03/20/2008
 Ind. Eng. Chem. Res., Vol. 47, No. 8, 2008 2697

Figure 1. (A) Txy diagrams for acetone-methanol at 1 and 10 atm. (B) Residue curve map for acetone-methanol-water.

Figure 2. Flowsheet for pressure-swing distillation; acetone-methanol.

composition specification is 99.5 mol % methanol, and the high-pressure column. The optimum feed stage locations are
distillate composition specification is 75 mol % acetone (2.5 determined by finding the feed trays that minimized reboiler
mol % lower than the azeotrope at 1 atm). In the high-pressure heat input. Figures 3 and 4 give temperature and composition
column, the bottoms composition specification is 99.4 mol % profiles in the two columns.
actone and the distillate composition specification is 40 mol % The first column in Table 1 gives design parameters and
acetone (2.5 mol % higher than the azeotrope at 10 atm). economic data for this pressure-swing distillation system.
The minimum reflux ratios are found with 103 stages to be Column diameters are found using Aspen Tray Sizing. The
2.3 and 3.08 in the low- and high-pressure columns, respectively. relationships given in Luyben11 are used for sizing condensers
The number of stages required for reflux ratios about 1.2 times and reboilers and for calculating capital cost of equipment.
these minima are 52 in the low-pressure column and 62 in the Temperature differentials are 34.8 K in reboilers. In condensers,
2698 Ind. Eng. Chem. Res., Vol. 47, No. 8, 2008

Table 1. Design Results

pressure swing extractive
C1 Nstage 52 57
ID (m) 3.30 2.79
QR (MW) 15.8 11.4
QC (MW) 15.1 9.87
QHX (MW) - 1.35
Treflux (K) 328 329
AC (m2) 996 609
AR (m2) 758 576
AHX (m2) - 52.7
shell (106 $) 1.13 1.03
HX (106 $) 1.19 1.02
total capital (106 $) 2.32 2.06
energy ($106 /yr) 2.22 1.69
TAC ($106 /yr) 3.00 2.38
C2 Nstage 62 26
ID (m) 1.89 1.89
QR (MW) 7.42 7.10
QC (MW) 5.86 6.89
Treflux (K) 407 338
AC (m2) 72.2 288
AR (m2) 383 350
shell (106 $) 0.722 0.346
HX (106 $) 0.466 0.624
total capital (106 $) 1.19 0.970
energy ($106 /yr) 1.13 1.03
TAC ($106 /yr) 1.52 1.38
total system total capital (106 $) 3.51 3.03
total energy ($106 /yr) 3.35 2.72
TAC ($106/yr) 4.52 3.75
Figure 3. (A) Low-pressure column temperature profile. (B) Low-pressure
column composition profiles.
would be attractive in this pressure-swing system, and this
design is explored later in the paper.
Energy cost is assumed to be $4.7 per million kilojoules, and
a payback period of 3 years is used to calculate annual capital
cost. The total annual cost (TAC) of this system is $4,520,000
per year.
2.1. Control System Design. The Aspen Plus file is exported
to Aspen Dynamics as a pressure-driven simulation after reflux
drum and base volumes are specified to provide 5 min of holdup
when at the 50% liquid level. Pumps and valves are sized to
give adequate pressure drops to handle changes in flow rates.
Figure 5 shows the control structure developed for this
system. Conventional PI controllers are used for all flows,
pressures, and temperatures. Proportional controllers are used
for all liquid levels. Relay-feedback tests are run on the two
temperature controllers to determine ultimate gains and periods,
and Tyreus-Luyben12 tuning is used. Each temperature control-
ler has a 1-min deadtime in the loop. Reflux ratios are held
constant in each column (2.84 in the low-pressure column and
3.11 in the high-pressure column).
Temperature control tray location is selected by choosing the
tray where there is a large temperature change from stage to
stage. In the low-pressure column, Stage 48 is selected with a
setpoint of 342.8 K (see Figure 3A). In the high-pressure
column, Stage 53 is selected with a setpoint of 413 K (see Figure
4A). Controller faceplates are shown at the bottom of Figure 5.
Table 2 gives controller parameters for the two temperature
controllers.
Figure 6 demonstrates the effectiveness of this control
Figure 4. (A) High-pressure column temperature profile. (B) High-pressure structure. Stable regulatory control is achieved for several large
column composition profiles. disturbances. The responses of the system to large 20% increases
and decreases in feed flowrate are shown in Figure 6A.
the temperature differential is the reflux drum temperature minus Responses for feed composition disturbances from 50 to 60 mol
310 K cooling water. Notice that reflux drum temperature in % acetone and from 50 to 40 mol % acetone are shown in Figure
the high-pressure column (407 K) is much higher than that in 6B. Product purities (xB1(M) and xB2(A)) are held quite close to
the low-pressure column (328 K). This results in a very small their specifications. The system takes about 2 h to come to a
area condenser. These temperatures suggest that heat integration new steady state.
 Ind. Eng. Chem. Res., Vol. 47, No. 8, 2008 2699

Figure 5. Control structure for pressure-swing distillation system.

Table 2. Controller Tuning Parameters

pressure swing with extractive with
pressure swing extractive heat integration heat integration
TC1
controlled variable T1,48 ) 342.8 K T1,50 ) 347.6 K T1,48 )342.8 K T1,50 ) 347.6 K
manipulated variable QR1 QR1 QAUX QAUX
transmitter range 300-400 K 300-400 K 300-400 K 300-400 K
controller output range 0 to 6.6 × 106 cal/s 0 to 6 × 106 cal/s 0 to 40 GJ/ha -30 to 30 GJ/hb
ultimate gain 26.1 7.15 44 10.3
ultimate period 2.4 min 4.2 min 3.6 4.8
Kc 8.16 2.23 13.7 3.2
τI 5.28 min 9.24 min 7.9 10.6
TC2
controlled variable T2,53 ) 413 K T2,22 ) 371.7 K T2,53 ) 140 °C T2,21 ) 136.3 oC
manipulated variable QR2 QR2 QR2 QR2
transmitter range 350-450 K 325-425 K 350-450 K 300-400K
controller output range 0 to 3.6 × 106 cal/s 0 to 3.6 × 106 cal/s 0 to 3.6 × 106 cal/s 0 to 6 × 106 cal/s
ultimate gain 13.1 4.23 24.9 1.20
ultimate period 3.6 min 4.8 min 4.8 6.0
Kc 4.11 1.32 7.8 0.375
τI 7.92 min 10.6 min 10.6 13.2
a Faceplate shows heat input in metric (GJ/h). b Auxiliary condenser needed to handle increase in methanol in feed.

3. Acetone-Methanol Separation Using Extractive 3.1 Steady-State Design. The Knapp-Doherty optimum
Distillation design had a 57-stage extractive column and a 26-stage methanol
column, both operating at 1 atm. They do not provide informa-
Knapp and Doherty9 present the economic optimum design
tion about feed tray locations, so these were found empirically
of an acetone-methanol separation using water as the extractive
solvent. The design used in this paper is based on their work. by finding what locations give minimum energy consumption.
Using extractive distillation has the inherent disadvantage of Note that the solvent is fed to the extractive column on a tray
introducing a third component into the system that will appear above the feed tray and below the top of the column so that
as impurity in the product streams. So the extractive system high-purity acetone can be produced in the distillate.
must show significant economic and/or control advantages over Figure 7 shows the flowsheet used in this paper with stream
the pressure-swing system to make it the process of choice. information, heat duties, equipment sizes, and operating condi-
2700 Ind. Eng. Chem. Res., Vol. 47, No. 8, 2008

of the methanol column from 378 to 320 K before it is fed into

the extractive column.
This flowsheet differs from that given in the Knapp-Doherty9
design in one important feature: the solvent-to-feed ratio.
Knapp and Doherty use a ratio unity. For a feed flowrate of
540 kmol/h, this corresponds to a solvent flowrate to the
extractive column of 540 kmol/h. However, as Figure 7 shows,
a solvent flowrate of 1100 kmol/h is used in this paper. We are
unable to get the desired purity of the acetone product at lower
solvent flowrates.
The extractive column has three design degrees of freedom
once the total stages and feed locations are fixed: reflux ratio,
solvent flowrate, and reboiler heat input. Figure 8 shows the
effects of changing the solvent flowrate and the reflux ratio on
the compositions of the distillate stream from the extractive
column, which is the acetone product. The composition of
acetone in the bottoms is held at 0.05 mol % acetone by
manipulating the reboiler heat input for all the cases shown in
Figure 8. Figure 8A shows that for a given solvent flowrate S,
there is an optimum reflux ratio, RR, that gives a maximum
acetone purity. The higher the solvent flowrate, the higher the
acetone purity. To achieve the desired 99.4 mol % acetone
purity, the solvent flowrate must be about 1100 kmol/h at a
reflux ratio of about 4. Figure 8B shows how the impurities of
methanol and water are affected by solvent and reflux ratio.
Increasing solvent flow decreases methanol impurity but
increases water impurity.
These results illustrate that it can be difficult to achieve the
desired acetone product purity in this extractive distillation
because of the competing effects of solvent flowrate and reflux
ratio. This problem becomes less severe as the volatility
Figure 6. (A) Pressure-swing system; 20% feed flow disturbances. (B) difference between the light key component and the solvent
Pressure-swing system; feed composition disturbances. increases. The acetone-water system does not have a large
volatility, which leads to difficulty in attaining the desired
tions. Notice that a small makeup water stream is required to acetone purity. Other solvents with higher boiling points should
account for the small losses of water in both product streams. permit higher acetone purities. However, the base temperature
Notice that there is more water impurity in the acetone product in the solvent recovery column would be higher, which would
leaving in the distillate from the extractive column than methanol require a higher-cost energy source. The non-monotonic rela-
(0.39 mol % water versus 0.11 mol % methanol), despite the tionship between acetone purity and reflux ratio can also lead
fact that water is a higher boiler than methanol. A cooler is to difficult operating problems, as pointed out by Knapp and
also required to cool down the water recycle from the bottom Doherty.9

Figure 7. Flowsheet for extractive distillation; acetone-methanol.

 Ind. Eng. Chem. Res., Vol. 47, No. 8, 2008 2701

Figure 10. (A) Extractive column temperature profile. (B) Extractive

column composition profiles.

Figure 8. (A) Effect of RR and solvent in extractive column on acetone

purity. (B) Effect of RR and solvent in extractive column on impurities.

Figure 9. Energy demand in pressure-swing distillation for high purities.

Figure 11. (A) Methanol column temperature profile. (B) Methanol column
composition profiles.
This inherent problem of an extractive distillation should be
compared to the ease of attaining higher purity levels in a in each column. The reboiler heat duties are QR1 and QR2 in
pressure-swing distillation system. As Figure 9 illustrates, very the low- and high-pressure columns, respectively. This feature
high product purities can easily be obtained with only modest is a clear advantage of the pressure-swing process.
increases in energy costs in the pressure-swing distillation The acetone-methanol system using water as the solvent is
system. These results are generated with a fixed number of trays explored by Kossack et al.,10 and they report (their Table 5)
2702 Ind. Eng. Chem. Res., Vol. 47, No. 8, 2008

Figure 12. Control structure for extractive distillation system.

that the economic optimum design had a solvent-to-feed ratio

of 1.886. This is much closer to our results.
The other difference between this design and the Knapp and
Doherty design is the reflux ratios in the two columns. They
report 2.76 and 1.06, while the present design values are 3.44
and 1.61. This difference is probably due to our higher solvent
flowrate.
Temperature and composition profiles in the two columns
are given in Figures 10 and 11. The extraction column composi-
tion profiles given in Figure 10B are quite unusual. The solvent
is fed on Stage 25 and the feed is introduced on Stage 40.
The economics of the extractive distillation process are
presented in the second column of Table 1. Notice that the
column diameters are smaller, heat exchanger areas are smaller,
and energy consumption is smaller in the extractive system
compared to the pressure-swing system. The TAC is about 17%
lower for the extractive system.
The TAC given by Knapp and Doherty for their extractive
system is $2,790,000 per year, compared to the $3,750,000 per
year calculated in this work. The difference is probably due to
our much higher solvent circulation rate.
2.2. Control System Design. The Aspen Plus file of this
extractive distillation system is exported to Aspen Dynamics
after dynamic parameters are specified using the same assump-
tions made in the pressure-swing simulation. Figure 12 shows
the control structure developed for this system. Relay-feedback
testing and Tyreus-Luyben tuning of the temperature loops give
the controller parameters given in Table 2. The temperature
controllers have 1-min deadtimes in the loops. Reflux ratios
are held constant in each column (3.44 in the extractive column
and 1.61 in the methanol column).
Temperature control tray location is selected by choosing the
tray where there is a large temperature change from stage to Figure 13. (A) Extractive distillation system; 20% feed flow disturbances.
stage (Stage 50 at 347.6 K in the extractive column and Stage (B) Extractive distillation system; feed composition disturbances.
22 at 371.7 K in the methanol column). See Figures 10A and
11A. the methanol column (the “LC21” controller) is controlled by
The two essential features of this control structure are (see manipulating the makeup water flowrate.
the block arrows in Figure 12) (1) the solvent flow is ratioed to Although the water makeup stream is much smaller than the
the feed flow (the “ratio” multiplier) and (2) the base level in total solvent flow, the 5-min holdup time in the base of the
 Ind. Eng. Chem. Res., Vol. 47, No. 8, 2008 2703

Controller faceplates are shown at the bottom of Figure 12.

Notice that the controller “FCtot” is on “cascade” because it
receives its remote setpoint from the multilier “ratio” so that
the solvent flowrate is ratioed to the feed flowrate.
Figure 13 demonstrates the effectiveness of this control
structure. Stable regulatory control is achieved for several large
disturbances. The responses of the system to large 20% increases
and decreases in feed flowrate are shown in Figure 13A.
Responses for feed composition disturbances from 50 to 60 mol
% acetone and from 50 to 40 mol % acetone are shown in Figure
13B. Product purities (xD1(A) and xD2(M)) are held quite close to
their specifications. The system takes about 1.5 h to come to a
new steady state.
At this point it is useful to give a direct comparison of the
dynamic performances of the two alternative systems. Figure
14 provides this comparison. The solid lines are for the
extractive distillation system and the dashed lines are for the
pressure-swing distillation system. Overall, these responses are
quite similar. The deviations in acetone purity are larger in the
pressure-swing system (bottoms from the high-pressure column)
than those seen in the extractive system (distillate from the
extractive column). This difference could be explained by the
response of a column to a change in feed flowrate. If the feed
is liquid, the bottoms product is more disturbed than the
distillate. Since the acetone product comes out the bottom of
the high-pressure column, it is disturbed more than in the
extractive distillation system in which the acetone product comes
off the top of the extractive column.
If this explanation is valid, one would expect to see larger
disturbances in the methanol product coming out the bottom of
Figure 14. (A) Comparison of extractive distillation and pressure-swing the low-pressure column in the pressure-swing system than those
systems; feed flowrate disturbances. (B) Comparison of extractive distillation
and pressure-swing systems; feed composition disturbances.
seen in the methanol product coming out the top of the methanol
column in the extractive system. However, this is not observed
methanol column is able to handle dynamic changes in the water in Figure 14A.
solvent circulation rate. When the feed is increased, the ratio From a purely conceptual dynamic point of view, one would
immediately increases the solvent fed to the extractive column. expect the extractive distillation process to exhibit less product
The level in the base of the methanol column immediately quality variability because both products are removed as
begins to drop. However, the increased flows to the extractive distillate streams from columns that have liquid feed streams.
column eventually produce an increase in the feed to the The results shown in Figure 14 do not give a clear demonstration
methanol column, which brings the base level back up. of this principle.

Figure 15. Flowsheet for extractive distillation with heat integration.

2704 Ind. Eng. Chem. Res., Vol. 47, No. 8, 2008

Table 3. Design Results for Heat-Integrated Systems In this section we compare the steady-state design and the
pressure swing extractive dynamic control of heat-integrated extractive and pressure-swing
with heat with heat processes. The same numbers of trays used in the base-case
integration integration designs are used in both systems. The systems have not been
C1 Nstage 52 57 re-optimized for heat integration.
ID (m) 3.30 2.79 3.1. Heat-Integrated Extractive Distillation Process. 3.1.1.
QCR (MW) 5.86 9.31
QAUX (MW) 8.93 2.10
Design. Figure 15 gives the flowsheet for the revised extractive
QC (MW) 15.1 9.87 distillation process in which the pressure in the methanol column
QHX (MW) - 1.35 is increased from 1 to 5 atm. This raises the reflux-drum
Treflux (K) 328 329 temperature to 385 K with the 99.5 mol % methanol distillate.
ARC (m2) 110 514 The base of the extractive column is 364 K, so there is a
AC (m2) 982 609
AHX (m2) - 52.7 reasonable 21 K differential temperature driving force to size
AAUX (m2) 450 106 the reboiler/condenser heat exchanger.
shell (106 $) 1.13 1.02 The higher pressure in the methanol column requires a higher
condenser (106 $) 0.643 0.567 reflux ratio and more heat input than the low-pressure operation
condenser/reboiler (106 $) 0.155 0.421
aux reboiler (106 $) 0.388 0.151 because the higher pressure makes the separation more difficult.
cooler HX (106 $) - 0.096 The reflux ratio increases from 1.61 to 2.91, and the reboiler
aux condenser (106 $) - ? heat input increases from 7.1 to 11.4 MW.
total capital (106 $) 1.58 2.26 The condenser duty in the methanol column is now 9.31 MW,
energy ($106/yr) 1.32 0.311
C2 Nstage 62 26 which can be used to produce vapor in the base of the extractive
ID (m) 1.89 1.61 column. The total energy needed in the extractive column is
QR (MW) 7.59 11.3 11.4 MW, so an auxiliary steam-driven reboiler is required with
AC (m2) - - a duty of 2.1 MW. The extractive column itself is unchanged
AR (m2) 383 570
shell (106 $) 0.722 0.292
except it has two reboilers.
condenser (106 $) - - The reboiler/condenser is sized by using an overall heat
reboiler (106 $) 0.349 0.451 transfer coefficient of 0.00306 GJ h-1 m-2 K-1. The heat transfer
total capital (106 $) 1.07 0.743 rate is 33.53 GJ/h (9.31 MW) and the temperature difference is
energy ($106/yr) 1.13 1.67
total total capital (106 $) 2.65 3.00
21.2 K, giving an area of 514 m2.
system The second column in Table 3 gives sizing and cost
total energy ($106/yr) 2.45 1.98 information. Notice that the diameter of the methanol column
TAC ($106/yr) 3.33 2.98 (1.61 m), which is operating at 5 atm, is smaller than the
diameter in the non-heat-integrated case at 1 atm (1.89 m),
3. Heat-Integrated Systems despite having a larger reboiler heat input (11.3 versus 7.59
The studies reported above have considered systems with MW). The larger vapor density at higher pressure results in a
independent reboilers and condensers on both columns. The smaller diameter column.
large differences in the temperature between the high- and low- The TAC of the heat-integrated extractive system is $2,980,-
pressure columns in the pressure-swing process make heat 000 per year. This should be compared with the base-case TAC
integration a natural. In the extractive process, Knapp and of $3,750,000 per year. Total capital investment is essentially
Doherty9 suggest raising the pressure in the methanol column the same in the base and heat-integrated designs ($3,000,000),
so that its overhead vapor can be condensed in the reboiler of while energy cost is reduced from $2,720,000 to $1,980,000.
the extractive column. They point of that the alternative of 3.1.2. Control. Implementing heat integration in Aspen
raising the pressure in the extractive column is infeasible because Dynamics requires the used of Flowsheet Equations. The method
of distillation boundaries that occur in the ternary diagram at for simulating heat integration in distillation systems is discussed
higher pressures. in detail in a recent paper.8

Figure 16. Aspen Dynamics flowsheet equations for heat integration and pressure-compensated temperature in high-pressure column.
 Ind. Eng. Chem. Res., Vol. 47, No. 8, 2008 2705

Figure 17. Control structure for extractive distillation system with heat integration.

Figure 18. Control structure for pressure-swing distillation system with heat integration.

There are two conditions that must be satisfied at each point in the reboiler/condenser QRC must be equal to the product of
in time during the dynamic simulation: (1) The heat transfer the fixed area ARC, the fixed heat-transfer coefficient U and the
2706 Ind. Eng. Chem. Res., Vol. 47, No. 8, 2008

Figure 19. (A) Comparison of extractive distillation and pressure-swing systems with heat integration; feed flowrate disturbances. (B) Comparison of
extractive distillation and pressure-swing systems with heat integration; feed composition disturbances.

temperature difference between the condenser temperature in that have been made from the base-case control structure: (1)
the methanol column and the base temperature in the extractive The output signal from the TC1 controller in the extractive
column. This heat transfer rate is the heat removed in the column is the auxiliary heat input QAUX, which is 7.61 GJ/h
condenser of the methanol column. (2) The total heat transfer (note Aspen uses metric units). (2) The PC2 pressure controller
in the base of the extractive column is the sum of the heat in the methanol column has no input signal and is on manual.
removed in the condenser of the methanol column QRC and the (3) The deadtime element “dead2” has no input shown in Figure
heat added in the auxiliary reboiler QAUX in the extractive 17, but its input signal is the pressure-compensated temperature
column. signal as defined in the flowsheet equation given in Figure 16A.
These conditions correspond to the first and second laws of The temperature and pressure on Stage 21 in the methanol
thermodynamics. Figure 16A shows the flowsheet equations for column are measured, and a “pressure-compensated” temper-
the extractive heat-integrated system. The two conditions listed ature13 TPC is fed into the deadtime element as its input signal.
above are specified in the top two equations in the Text Editor The design pressure on Stage 21 is 5.136 atm with a composition
window. of 45 mol % methanol and a temperature of 396.19 K.
The third equation is used to provide a “pressure-compen- Generating a Txy diagram at 6 atm gives a temperature of
sated” temperature13 measurement in the methanol column. This 401.66 K at the same composition. The following equation is
is needed because in the heat-integrated system the pressure in used to calculate the signal fed to the deadtime element
the methanol column is not controlled. It floats with operating and subsequently into the temperature controller as the PV
conditions. If more heat transfer is required in the reboiler/ signal.
condenser, a larger temperature difference is required, and this
- 396.19
is achieved by the pressure in the methanol column increasing,
which raises the bubblepoint temperature in the reflux drum.
TPC ) T21 - (P21 - 5.136) (401.66
6 - 5.136 )
Figure 17 shows the control structure for the heat-integrated The two temperature controllers are tuned using the methods
extractive system. The three block arrows point to the changes discussed above. Tuning parameters are given in Table 2.
 Ind. Eng. Chem. Res., Vol. 47, No. 8, 2008 2707

Remember that the temperature controller in the extractive handle this disturbance. In the Aspen Dynamics simulation this
column is manipulating the energy input QAUX from the auxiliary is easily achieved by specifying the minimum controller output
reboiler (2.10 MW or 7.6 GJ‚h as given in the TC1 faceplate signal of TC1 to have a negative value (-30 GJ/h as shown in
in Figure 17). Table 2). The additional capital cost of an auxiliary condenser
The effectiveness of this control structure on the heat- is not included in the economic data shown in Table 3.
integrated extractive distillation system is compared with that
of the heat-integrated pressure-swing distillation system in the 4. Conclusion
following section.
Steady-state and dynamic comparisons have been presented
3.2. Heat-Integrated Pressure Swing Process. The heat-
of an extractive distillation process and a pressure-swing
integrated flowsheet in the pressure-swing system is essentially
distillation process. The extractive system has lower costs and
the same as without heat integration. The only change is that
equivalent dynamics compared to the pressure-swing process.
there is a reboiler/condenser and an auxiliary steam-heated
However, it is more difficult to achieve high product purities.
reboiler on the low-pressure column. The temperature difference
In addition, the introduction of a third component can lead to
between the condenser of the high-pressure column and the base
impurity issues in the products.
of the low-pressure column is quite large (407.5 - 345.3 )
Using heat integration reduces energy costs. Capital costs in
62.2 K). The heat duty in the condenser of the high-pressure
the extractive system are essentially the same with and without
column is 5.86 MW, which gives a heat-transfer area of 110
heat integration. In the pressure-swing system, the capital cost
m2.
of the heat-integrated process is actually lower than the base-
The first column in Table 3 gives process and economic data
case process. This is the result of lower reboiler capital cost
for this process. The TAC of the heat-integrated pressure-swing
because of the smaller areas that result from the large temper-
system is $3,330,000 per year. This should be compared with
ature difference between the top of the high-pressure column
the base-case TAC of $4,520,000 per year and the heat-
and the base of the low-pressure column.
integrated extractive process of $2,980,000 per year. Total
In the heat-integrated system, the pressure in the high-pressure
capital investment is unexpectedly smaller than for the base
column changes with operating conditions, so a pressure-
design ($3,330,000 instead of $3,510,000). This is the result of
compensated temperature control structure is required to main-
the small reboiler/condenser due to the large temperature driving
tain product specification in these columns.
force. The energy cost is reduced from $3,350,000 to $2,450,000
per year.
The energy requirement in the low-pressure column is 14.79 Literature Cited
MW, so the auxiliary reboiler must provide 14.79 - 5.86 ) (1) Grassi, V. G. Process design and control of extractive distillation,
8.93 MW (32 GJ/h, as shown in the TC1 faceplate in Figure Practical Distillation Control; Van Nostrand Reinhold: New York, 1992;
18). Similar flowsheet equations are needed in this system and p 370.
are given in Figure 16B. The pressure-compensated temperature (2) Chien, I. L.; Wang, C. J.; Wong, D. S. H.; Lee, C. H.; Cheng, S. H.;
Shih, R. F.; Liu, W. T.; Tsai, C. S. Experimental investigation of
measurement uses the temperature (412.9 K) and pressure conventional control strategies for a heterogeneous azeotropic distillation
(10.354 atm) on Stage 53. column. J. Process Control 2000, 10, 333.
(3) Chien, I. L.; Zeng, K. L.; Choa, H. Y. Design and control of a
- 411.51
TPC ) T53 - (P53 - 10.354) (414.26
11 - 10.354 )
complete heterogeneous azeotropic distillation column system. Ind. Eng.
Chem. Res. 2004, 43, 2160-2174.
(4) Arifin, S.; Chien, I. L. Design and control of an isopropyl alcohol
dehydration procee via extractive distillation using dimethyl sulfoxide as
Figure 18 gives the control structure developed for the heat- an entrainer. Ind. Eng. Chem. Res. 2008, 47, 790-803.
integrated pressure-swing process. (5) Luyben, W. L. Control of a multi-unit heterogeneous azeotropic
Figure 19 gives a direct comparison of the two alternative distillation process. AIChE J. 2006, 52, 623-637.
heat-integrated processes for feed flowrate and feed composition (6) Luyben, W. L. Plantwide control of an isopropyl alcohol dehydration
process. AIChE J. 2006, 52, 2290-2296.
disturbances. The solid lines are for the pressure-swing system. (7) Abu-Eishah, S. I.; Luyben, W. L. Design and control of a two-column
The dashed lines are for the extractive system. Both systems azeotropic system. Ind. Eng. Chem. Process Des. DeV. 1985, 24, 132-
provide stable base-level regulatory control. The results are quite 140.
similar to those found in the non-heat-integrated systems. (8) Luyben, W. L. Design and control of a fully heat-integrated pressure-
swing azeotropic distillation system, Tuning of PI controllers for integrator/
These results show that there is no control penalty for going deadtime processes. Ind. Eng. Chem. Res. 2008, 41, 2681-2695.
to the heat-integrated systems. This result is to be expected (9) Knapp, J. P., Doherty, M. F. Thermal integration of homogeneous
because only partial heat integration is used so there is no loss azeotropic distillation sequences. AIChE J. 1990, 36, 969-984.
of control degrees of freedom. Similar results were found in a (10) Kossack, S.; Kraemer, K.; Gani, R.; Marquardt, W. A systematic
previous paper.8 synthesis framework for extractive distillation processes. Chem. Eng. Res.
Des. 2008, submitted.
One comment should be made about the extractive heat- (11) Luyben, W. L. Distillation Design and Control using Aspen
integrated system. A close look at the flowsheet in Figure 15 Simulation; Wiley: New York, 2006; p 88.
will disclose an auxiliary condenser on the extractive column (12) Tyreus, B. D.; Luyben, W. L. Tuning of PI controllers for integrator/
in addition to an auxiliary reboiler. When the feed composition deadtime processes. Ind. Eng. Chem. Res. 1992, 31, 2625-2628.
(13) Buckley, P. S.; Luyben, W. L.; Shunta, J. P. Design of Distillation
disturbance from 50 to 40 mol % acetone is introduced into the Column Control Systems; Instrument Society of America: Research Triangle
system, the large production rate of methanol increases the Park, NC, 1985; p 234.
reboiler heat input in the methanol column, causing the heat
input in the auxiliary reboiler in the extractive column to be ReceiVed for reView December 11, 2007
driven to zero. The system is unable to maintain acetone product ReVised manuscript receiVed February 8, 2008
Accepted February 13, 2008
purity because the temperature in the extractive column becomes
too high. Thus, an auxiliary condenser would be needed to IE701695U