Accepted Manuscript

Title: Design and Control of a Pressure-Swing Distillation

Process with Vapor Recompression

Author: William L. Luyben

PII: S0255-2701(17)30869-3
DOI: https://doi.org/10.1016/j.cep.2017.09.020
Reference: CEP 7084

To appear in: Chemical Engineering and Processing

Received date: 28-8-2017

Revised date: 28-9-2017
Accepted date: 29-9-2017

Please cite this article as: William L.Luyben, Design and Control of a Pressure-
Swing Distillation Process with Vapor Recompression, Chemical Engineering and
Processing https://doi.org/10.1016/j.cep.2017.09.020

This is a PDF file of an unedited manuscript that has been accepted for publication.
As a service to our customers we are providing this early version of the manuscript.
The manuscript will undergo copyediting, typesetting, and review of the resulting proof
before it is published in its final form. Please note that during the production process
errors may be discovered which could affect the content, and all legal disclaimers that
apply to the journal pertain.
Submitted to Chemical Engineering and Processing: Process Intensification

Design and Control of a Pressure-

Swing Distillation Process with Vapor
Recompression

William L. Luyben

Department of Chemical Engineering

Lehigh University
Bethlehem, PA 18015
USA

September 1, 2017
Revised September 28, 2017

[email protected]; 610-758-4256; FAX 610-758-5057

High Lights

 Pressure-swing distillation using vapor recompression in both columns is

studied.
 An effective control structure is developed for this unique process.
 Energy costs are 16% lower than those given in a published paper.

Abstract
Pressure-swing distillation is widely used to separate minimum-boiling azeotropes
when the azeotropic composition has significant pressure dependence. The two columns
operate at different pressures with distillate streams having compositions close to their
respective azeotropes. Heat integration is typically used since column temperatures are
sufficiently different to permit heat transfer.
A novel modification to achieve process intensification has recently been
proposed that uses vapor recompression on both columns. The study developed a
complex heat exchanger network along with the design of the two columns. Dynamic
controllability was not studied.
The purpose of this paper is to develop a robust control structure for this process.
A novel control scheme features the flowrate of the distillate recycle from the high-
pressure column as the throughput manipulator with the fresh feed brought in on level
control.

Key Words
Pressure-swing distillation; azeotropic distillation; vapor recompression

1. Introduction
Vapor recompression (heat pumping) has many applications in a wide variety of
domestic and industrial uses. The basic thermodynamic principle of operation is to use
work to produce temperature differences that permit heat transfer. The home refrigerator

2
is the most widely applied and most simple example in which high-value electrical
energy is used to achieve transfer of heat from the low temperature inside the refrigerator
to the warmer air in the kitchen. The closed cycle of the circulating refrigerant can be
visually conveyed on an enthalpy-pressure diagram with energy flowing in from the low-
temperature source (air in refrigerator) and energy flowing out to the high-temperature
sink (room air). Work is required to raise the pressure of the vapor refrigerant so it can be
condensed by the high-temperature sink. Then expanding the high-pressure liquid
refrigerant to a low pressure produces a partially vaporized low-temperature stream that
can be completely vaporized by heat from the source. The final step is compressing the
refrigerant vapor from the low-pressure “evaporator” to the high pressure in the
“condenser.”
Early industrial applications that involve vapor recompression of process streams
include distillation and evaporation. The separation of components with similar boiling
points (low relative volatility) such as propylene/propane1 can be achieved in a
distillation column by compressing the overhead vapor and condensing it in the reboiler
to provide vapor boilup in the column.
Several recent papers have studied vapor recompression in more complex
distillation systems. Li et al2 proposed using vapor recompression to provide the heat
required in one of the reboilers of a divided-wall column separating a heterogeneous
azeotrope. Effective dynamic control of this system has been demonstrated3. Luo et al4
proposed the use of vapor recompression in a single divided-wall column for bioethanol
dehydration, and Patrascu et al5 developed a control system for this complex interacting
system. An improved process design and a more robust control structure for this process
has been presented6.
The use of vapor recompression in pressure-swing distillation systems has
recently been proposed by Xia et al7. This is a novel switch from conventional pressure-
swing distillation in which heat integration is typically used between the two columns
operating at two pressure and temperature levels. The economics of a conventional
pressure-swing distillation process with and without heat integration were compared with
the vapor-recompression design. The process with vapor recompression had a much
higher capital investment (73 %) compared to the heat integrated case because of the two

3
expensive compressors. However the energy cost is much smaller (48 %), producing a
modest reduction in total annual cost of 8.6 %.
A complex heat-exchanger network was proposed in the Xia et al7 paper that used
both the sensible heat and the latent heat of the compressed vapor streams to achieve
“self heating” with no external source of heat required. The only energy cost was the
compressor power of the two compressors. Complex process flow patterns were proposed
with two process streams split for preheating in parallel heat exchangers with different
heat source streams. A total of nine separate heat exchangers were used. A much more
simple heat-exchanger system is used in this paper.

2. Process Studied
The numerical example studied by Xia et al7 is the separation of di-isopropyl
ether (DIPE) and isopropyl alcohol (IPA) whose normal boiling points are 63.3 oC and
82.26 oC, respectively. These components form minimum-boiling azeotropes as shown in
Figure 1. At 1.01 bar, the composition of the azeotrope is 77.43 mol% DIPE at 65.97 oC.
At 4.05 bar, the composition of the azeotrope is 61.36 mol% DIPE at 110.8 oC. The
reasonable pressure dependence of the azeotropic composition permits the economic use
of pressure-swing distillation to achieve high-purity products. Aspen steady-state and
dynamic simulations are used in this study with NRTL physical properties.
Figure 2 gives the xy-diagram for the two columns with the feed and products
streams of the two columns. With a feed of 25 mol% DIPE, the bottoms from the low-
pressure column C1 is high-purity IPA with composition xB1. The distillate D1 has a
composition xD1 slightly smaller than the azeotropic composition at 1.01 bar. The
distillate is fed to the high-pressure column C2 in which the bottoms is high-purity DIPE
with composition xB2. The distillate has a composition xD2 slightly larger than the
azeotropic composition at 4.04 bar. The approach composition difference (yaz – xD) is a
key design optimization variable that is discussed in the next section. The distillate D2
from the high-pressure column is recycled back to the low-pressure column.
Figure 3 gives the flowsheet of the pressure-swing distillation process with vapor
recompression studied in this paper. The column configurations are those used in the Xia

4
et al7 paper (number of trays, feed locations and column pressures), but a more simple
heat exchanger network is used to avoid dynamic controllability problems.
The overhead vapor from each column is compressed to a pressure that is high
enough to provide a differential temperature driving force ∆T in the two reboilers of 5 oC.
This is the design ∆T used by Xia et al7 and is reasonable because expensive compression
is involved, which pushes the design trade-off between heat exchanger area and
compressor power to large area and small ∆T to reduce compression costs.
The low-pressure column operates at 1.01 bar and has 20 stages with 0.01 bar
pressure drop per stage. The bottoms product is high-purity IPA, giving a base
temperature of 86 oC. Therefore the condensing temperature on the hot side of the LPC
reboiler must be 91 oC. The compressor discharge pressure that gives this temperature of
the LPC overhead vapor is 2.43 bar.
The high-pressure column operates at 4.05 bar and has 28 stages with 0.01 bar
pressure drop per stage. The bottoms product is high-purity DIPE, giving a base
temperature of 121 oC. Therefore the condensing temperature on the hot side of the HPC
reboiler must be 126 oC. The compressor discharge pressure that gives this temperature of
the HPC overhead vapor is 6.08 bar.
The energy requirement to achieve the specified separation in the LPC is 1101
kW, but the heat required to desuperheat and condense the stream coming from the
compressor K1 exceeds this amount by 57.0 kW. Therefore, an auxiliary condenser is
installed using cooling water. The flowrate of the cooling water will be used to control a
temperature in the LPC.
The energy requirement to achieve the specified separation in the HPC is 767 kW,
but the heat released in desuperheating and condensing the stream coming from the
compressor K2 is only 677 kW. A steam-heated auxiliary reboiler is installed to provide
the required additional vapor boilup (90 kW). The flowrate of the steam will be used to
control a temperature in the HPC.
The total energy requirement of the process shown in Figure 3 is 171 kW of
expensive electric power to drive the compressors and 90 kW of inexpensive low-
pressure steam. It is important to note that the compressor power requirements of the Xia
et al7 paper were 248.1 kW. So the proposed process shown in Figure 3 achieves the

5
separation with 31% less high-level compressor power requirements. However, the
proposed process requires 90 kW of energy in the auxiliary reboiler. Assuming energy
costs of $16.8 per GJ for compressor power and $7.78 per GJ for low-pressure steam, the
Xia et al process energy cost is 16% higher than the energy cost of the proposed process.

3. Optimum Distillate Composition

There are two important design optimization variables in heat-integrated pressure-
swing distillation systems. The first is the selection of the two operating pressures, which
affects the differences in the two azeotropic compositions and the temperature
differentials for heat transfer. The second critical variable is the closeness of the distillate
composition to the azeotropic composition. This ∆x = yaz – xD specification has a very
significant effect on the flowrates of the distillate streams and on the difficulty of the
separation (the required reflux ratio). The design trade-off must balance these two
competing effects. A small ∆x means the separation requirement are larger, which results
in higher reflux ratios. Vapor boilup and compressor power requirements tend to
increase. On the other hand, a small ∆x results in smaller distillate flowrates, as discussed
below. A logical criterion for selecting the optimum ∆x is to determine the value that
minimizes compressor power requirements.
First let us analyze the overall system, which is binary so only two material
balances can be applied. The overall balances given below apply for any selection of ∆x.
F  B1  B2
Fz  B1 xB1  B2 xB 2
If the feed flowrate (F), feed composition (z) and the two bottoms product compositions
(xB1 and xB2) are given, the two product flowrates (B1 and B2) can be calculated. These
values apply for any pressure selection, any distillation configuration and any choice of
distillate compositions.

6
  x z 
B1  F  B 2 
 xB 2  xB1 
B2  F  B1
Now the total and component balances for just the HPC column C2 can be combined to
find the important relationship that shows how the recycle flowrate D2 is affected by the
selection of the two distillate compositions.
D1  D2  B2
D1 xD1  D2 xD 2  B2 xB 2
Combining these two equations gives:
D2 xD 2  B2 xB 2
D1 
xD1
D2 xD 2  B2 xB 2
D2  B2 
xD1
Solving for the recycle flowrate D2 gives:
 x x 
D2  B2  B 2 D1 
 xD1  xD 2 
The recycle flowrate is strongly affected by the difference between the selected distillate
compositions. The smaller this difference is, the larger the recycle flowrate.
It should be emphasized that this material balance relationship clearly shows that
small changes in the distillate compositions will strongly affected the recycle flowrate.
Now the distillate compositions must be related to the azeotropic compositions at
the two pressures. The LPC distillate composition must be somewhat smaller than the
azeotropic composition at 1 atm. The HPC distillate composition must be somewhat
larger than the azeotropic composition at 4 atm. We define this difference as ∆x, which
for simplicity is assumed to be the same for both columns.
xD1  y1az  x  0.7743  x
xD 2  y2az  x  0.6136  x
Figure 4 shows how a number of important variables are affected by the selection
of ∆x. Product purities (xB1 and xB2) are kept constant for all cases, so the product
flowrates B1 and B2 are also constant. The upper left graph demonstrates that the distillate
flowrates increase as ∆x is increased. However, since the separation is easier, the reflux

7
flowrates R1 and R2 decrease. The net effect is non-monotonic changes in the overhead
vapor flowrates through the compressors. A ∆x of 1.5 mol% gives the minimum total
compressor power.

4. Simulation Issues
Any process with recycle streams can present simulation convergence problems.
The procedure used in this work is to “tear” three streams in the steady-state simulation
(D2, R1 and R2) as shown in Figure 5. For a given ∆x case, the flowrate and composition
of D2 are known, so these are fixed. Two Aspen flowsheet design specs are used to
achieve the two specified distillate compositions by varying the two reflux flowrates R1
and R2. These tear streams are closed in Aspen Dynamics.
The heat-transfer relationships are handled in Aspen Dynamics by using
Flowsheet Equations, as given in Figure 6. In the LPC, the net reboiler duty is calculated
by subtracting from the heat transferred in the “REB1” block the heat removed in the
auxiliary condenser (the output signal from the temperature controller in C1). In the HPC,
the total reboiler duty is calculated by adding to the heat transferred in the “REB2” block
the heat added in the auxiliary reboiler (the output signal from the temperature controller
in C2).

5. Plantwide Control
Many process systems with recycles can be sensitive to small changes in
throughput that can produce large changes recycle flowrates. This “snowballing”
phenomenon8 has been reported in a variety of systems and led to the suggestion made
two decades ago of using a control structure that places a flow controller somewhere in
the liquid recycle loop.
A number of papers have discussed the control of various types of pressure-swing
distillation systems with and without heat integration. Minimum-boiling azeotropes have
distillate recycles. Maximum-boiling azeotropes have bottoms recycles. A review of
many aspects of pressure-swing distillation, including control structures, has been

8
recently presented9. A pioneering paper10 proposed a control structure for the minimum-
boiling THF/water azeotropic separation with partial heat integration. The snowball
effect was not observed despite having no flow controller in the distillate recycle loop.
Successful control of the dimethyl carbonate/methanol separation was reported11 using a
control structure without a flow controller in the distillate recycle loop, but heat
integration was not considered. Ye et al12 studied a partially heat-integrated pressure-
swing distillation system separating the maximum-boiling ethylene diamine/water
azeotropic mixture. The proposed control structure fixed the ratio of the bottoms recycle
from the LPC to the fresh feed, so a flow controller was in the recycle loop. However,
their control scheme appears to not control the level in the base of the LPC, so the
effectiveness of this plantwide control structure is questionable.
The process studied in this paper does not use heat integration between the two
columns. Vapor recompression of the overhead vapor from each column is used to
provide heat to each reboiler. It is not unexpected to find that a different control structure
will be required with this new process configuration. A number of alternative control
structures were studied. The only structure that provided robust and effective control is
described below and shown in Figure 7. All of the other schemes resulted in snowballing
with the recycle D2 stream undergoing very large changes when small disturbances were
introduced into the system, which resulted in valve and reboiler duty saturation.
The key feature of the proposed control structure is to flow control the recycle
flowrate D2 and use it as the throughput manipulator. Then the fresh feed is brought into
the system to control the level in the reflux drum of the first column. The location of the
temperature control trays are selected by looking at the temperature profiles given in
Figure 8. Level controllers are proportional with KC = 2. Pressure controller tuning used
the Aspen default values. Temperature controllers had 1-minute deadtimes and were
tuned using relay-feedback tests and Tyreus-Luyben tuning rules. The loops are
enumerated below.
1. Recycle D2 is flow controlled.
2. Reflux-drum level in C2 is controlled by manipulating the flowrate D1, which is
the feed to C2.
3. Reflux-drum level in C1 is controlled by manipulating the flowrate F.

9
 4. Base levels are controlled by manipulating bottoms flowrates.
5. Column pressures are controlled by manipulating compressor power.
6. Reflux flowrates are ratioed to distillate flowrates.
7. The temperature on Stage 16 in C1 is controlled by manipulating heat removal in
the auxiliary condenser.
8. The temperature on Stage 25 in C2 is controlled by manipulating heat input in the
auxiliary reboiler.

Figures 9 and 10 give the responses of the system to large disturbances in

throughput and feed composition. In Figure 9 the setpoint of the recycle D2 flow
controller is changed by 20% at 0.5 hours. Solid lines are 20% increases, and dashed lines
are 20% decreases. Both product compositions are held close to their specifications. Note
that a 20% change in recycle D2 produces about a 20% change in fresh feed F. Both
temperatures are well controlled. An increase in throughput results in more compressor
power and larger duties in the two reboilers, the auxiliary reboiler and the auxiliary
condenser.
The response of the fresh feed to changes in the throughput manipulator recycle
D2 is quite interesting. When the recycle D2 is increased, the initial response of the fresh
feed is to decrease before eventually increasing to the new steady-state flowrate. This
inverse response behavior can be explained by remembering that the D2 distillate from
the HPC is at a higher temperature and pressure than the conditions in the LPC. So an
increase in D2 increases the vapor going to the LPC condenser and results in a temporary
increase in the reflux-drum level. The level controller reduces the fresh feed (top right
graph in Figure 9). However, the increase in D2 (and R2 since it is ratioed to D2) causes
the reflux-drum level in C2 to drop, which eventually increases D1 (and R1). This starts to
drop the level in the C1 reflux drum and pulls in more fresh feed.
Figure 10 gives responses to changes in feed composition. Solid lines are when
DIPE composition is increased from 25 to 30 mol% with a corresponding reduction in
IPA. Dashed lines are when DIPE composition is decreased from 25 to 20 mol% with a
corresponding increase in IPA. Keep in mind that the throughput manipulator D2 is
constant for these feed composition disturbances. The control structure adjusts the

10
flowrate of fresh feed to handle the changes in feed composition. An increase in DIPE
feed composition results in less fresh feed (top right graph in Figure 10) fed to the LPC .
Putting more DIPE into the system increases the separation load in the HPC where this
component is removed as a bottoms product. With fixed distillate and reflux flowrates in
this column, its feed must be reduced to achieve the required bottoms purity. Therefore
the distillate and reflux streams in the LPC are reduced, which results in lower fresh feed
flowrates. Note that there is more heat removed in the auxiliary cooler in the LPC (Qcool
in the 4th right graph in Figure 10).
The heat duties in the two auxiliary heat exchangers are relatively small
percentages of the total duties (4.9% in the condenser and 12% in the reboiler) and could
saturate during severe transients. However, process is demonstrated to handle quite large
disturbances in both throughput and feed composition.

Conclusion
A pressure-swing distillation process that uses vapor recompression on both
columns has been studied in terms of both steady-state design and dynamic
controllability. The proposed process has lower energy cost than the process in the
published paper. A robust plantwide control structure is demonstrated to effectively
handle large disturbances in throughput and feed composition. The control scheme
features the distillate recycle flowrate as the throughput variable.

11
References
1. Muhrer, C. A., Collura, M. A., Luyben, W. L. “Control of Vapor-Recompression
Distillation Columns'', Ind. Eng. Chem. Research 29 (1990) 59.
2. Li, R., Ye, Q., Suo, X., Dai, X., Yu, H., Feng, S., Xia, H. “Improving the
performance of heat pump-assisted azeotropic dividing wall distillation” Ind. Eng.
Chem. Res. 55 (2016) 6454-6464.
3. Luyben, W. L. “Control of Azeotropic DWC with Vapor Recompression”
Chemical Engineering and Processing: Process Intensification 109 (2016) 114-
124.
4. Luo, H., Bildea, C. S., Kiss, A. A. “Novel heat-pump-assisted extractive
distillation for bioethanol purification” Ind. Eng. Chem. Res. 54 (2015) 2208-
2213.
5. Patrascu, I., Bildea, C. S., Kiss, A. A. “Dynamics and control of a heat-pump-
assisted extractive dividing-wall column for bioethanol dehydration” Chem. Eng.
Res. Des. 119 (2017) 66-74.
6. Luyben, W. L. “Improved plantwide control structure for extractive divided-wall
columns with vapor recompression” Chem. Eng. Res. Des. 123 (2017) 152-164.
7. Xia,H., Ye, Q., Feng, S., L, R., Suo, X. “A novel energy-saving pressure seing
distillaion process based on self-heat recuperation technology” Energy (2017),
submitted
8. Luyben, W. L. “Snowball Effect in Reactor/Separator Processes with Recycle”
Ind. Eng. Chem. Res. 33 (1994) 299-305.
9. Liang, S., Cao, Y., Liu, X., Li, X, Zhoa, Y., Wang, Y., Wang, Y. “Insight into
pressure-swing distillation from azeotropic phenomenon to dynamic control”
Wang Chem. Eng. Res. Design 117 (2017) 318-335.
10. Abu-Eishah, S., Luyben, W. L. “Design and control of two-coulmn azeotropic
distillation system” IEC Proc. Des. Dev. 24 (1985) 132-140.
11. Wei, H., Wang, F, Zhang, J., Liao, N., Xiao, F., Wei, W., Sun, Y. “Design and
control of dimethyl carbonate-methnaol separation via pressure-sweing
distillation” Ind. Eng. Chem. Res. 52 (2013) 11463-11478.

12
 12. Li, R., Ye, Q., Suo, X., Dai, X., Yu, H. “Heat-integrated pressure-swing
distillation process for separation of a maximum-boiling azeotrope ethylene
diamine/water” Ye Chem. Eng. Res. Des. 105 (2016) 1- 15.
13.

Figure Captions
Figure 1 – Txy diagrams at 1.01 and 4.05 bar
Figure 2 – xy diagram
Figure 3 – Flowsheet
Figure 4 – Effect of ∆x
Figure 5 – Aspen Plus PFD with torn streams
Figure 6 – Aspen Dynamics flowsheet equations
Figure 7 – Control structure
Figure 8 – Temperature profiles
Figure 9 – 20% changes in D2
Figure 10 - Feed composition disturbances

13
14
15
16
17