Chemical Engineering and Processing 50 (2011) 1266–1274

Contents lists available at ScienceDirect

Chemical Engineering and Processing:

Process Intensification
journal homepage: www.elsevier.com/locate/cep

Separation of di-n-propyl ether and n-propyl alcohol by extractive distillation

and pressure-swing distillation: Computer simulation and economic
optimization
Estela Lladosa ∗ , Juan B. Montón 1 , Ma Cruz Burguet 2
Departamento de Ingeniería Química, Escuela Técnica Superior de Ingeniería, Universitat de València, 46100 Burjassot, Valencia, Spain

a r t i c l e i n f o a b s t r a c t

Article history: Azeotropic mixtures are impossible to separate by ordinary distillation. Two of the most common meth-
Received 18 January 2011 ods for separating a binary homogeneous azeotrope are pressure-swing distillation (PSD) and extractive
Received in revised form 29 June 2011 distillation (ED). The PSD process is effective if the azeotropic composition changes significantly with
Accepted 21 July 2011
pressure. The ED process is effective if a suitable solvent can be found.
Available online 30 July 2011
This paper compares these two alternatives to separate a mixture made up of 50 mol% of di-n-propyl
ether and 50 mol% of n-propyl alcohol by means of a practical case of a plant to treat 12,000 Tm/year of
Keywords:
this mixture.
Pressure-swing distillation
Extractive distillation
The simulation has been carried out satisfactorily by mean of a package of commercial software (Aspen
Computer simulation Hysys® ) using the thermodynamic model UNIQUAC with binary parameters obtained experimentally by
Di-n-propyl ether us in previous papers.
n-Propyl alcohol The two processes evaluated have been optimized independently from each other and the best config-
urations have been evaluated economically. Results show that, for this particular case, the PSD is more
attractive than the extractive distillation.
© 2011 Elsevier B.V. All rights reserved.

1. Introduction a homogeneous minimum boiling point azeotrope at atmospheric

Distillation is the most widely used separation process in the
chemical industries. In a typical chemical plant, distillation columns
and their support facilities can account for about one-third of the
total capital cost and more than half of the total energy consump-
tion. Consequently, the design and optimization of the distillation
train have a critical impact on the economics of the entire process.
Moreover, when the problem is separate an azeotropic mixture,
rigorous, robust and reliable thermodynamic models are crucial for
the synthesis and design of these separation systems.
In this work, it has been studied the separation of binary
azeotrope di-n-propyl ether (DPE) + n-propyl alcohol (PA). The
aliphatic ethers are obtained normally by dehydration of the cor-
responding alcohol in the presence of an adequate catalyst. In this
case, di-n-propyl ether can be prepared from n-propanol by dehy-
dration with sulphuric acid. Final purification of ether in traditional
technologies is a relative complex procedure due to the existence of
∗ Corresponding author. Tel.: +34 96 35 44325; fax: +34 96 3544898.
E-mail addresses: Estela.Lladosa@uv.es (E. Lladosa),
jbmonton@uv.es (J.B. Montón), cruz.burguet@uv.es (M. Burguet).
1
Tel.: +34 96 35 44317.
2
Tel.: +34 96 35 44318.
pressure.
There are many techniques suitable for separation of azeotropic
mixtures, such as pressure swing distillation [1], extractive and
azeotopic distillation [2], liquid–liquid extraction [3], adsorption
[4], pervaporation using membrane [5], salt addition [6], and some
new coupling separation techniques and so on. Among the vari-
ous techniques available for breaking azeotropes, how to select a
suitable one, design and desired separation process, and optimize
the separation sequences are a complex task. In this work, only
pressure-swing distillation and extractive distillation have been
considered.
Although these two alternatives are quite similar in process
flow-sheet, the pressure-swing distillation is preferred because of
the need of adding a solvent in the extractive distillation may rise to
serious environmental concerns. Studies on the PSD processes have
not received much attention so far, although it has been known
since the 1920s.
Laboratory experiments in either pressure-swing distillation or
extractive distillation are time-consuming and expensive because
of the large number of parameters involved. It would be desirable to
predict the experimental data with the help of available simulation
programs.
The use of computer simulations for process modelling
and design is an established best practice for rapid process

0255-2701/$ – see front matter © 2011 Elsevier B.V. All rights reserved.
doi:10.1016/j.cep.2011.07.010
 E. Lladosa et al. / Chemical Engineering and Processing 50 (2011) 1266–1274
development and optimization in the chemical and petrochemical
industry. Such technology is also contributing to the development
and optimization of process technologies and production plants.
However, one of the challenges limiting the use of process mod-
elling and design is the lack of proven models and databanks for
estimating thermophysical properties. A major problem appears to
be obtaining a reliable, consistent set of plant data. Nevertheless, in
this case, it has been used consistent thermophysical experimental
data determined by us in previous papers [7].
Aspen Hysys® v.2006.5 was selected as a process simulator for
both its simulation capabilities and its ability to incorporate calcu-
lation using the spreadsheet tool.
In a previous paper, our research group [8] studied also these
separation processes (extractive distillation and pressure-swing
distillation) applied to the separation of isobutyl alcohol + isobutyl
acetate azeotropic mixture.
1.1. Pressure-swing distillation
On the one hand, homogeneous azeotropic compositions that
are pressure-sensitive can be separated using pressure-swing dis-
tillation (PSD), which utilizes two or more distillation columns
operating at different pressures together with appropriate recycle
strategies to achieve the desired separation.
Lewis [9] was the first, who proposed distilling the azeotropic
mixtures by PSD. This process has been suggested by other authors
to separate azeotropic mixture; e.g. Black [10], Abu-Eihah and Luy-
ben [11], Chang and Shih [12]. Phimister and Seider [13] were the
first who studied the batch application of binary PSD by simula-
tion. They investigated the separation of a minimum azeotrope
(THF-water) by semicontinuous PSD.
The sensitivity of azeotropes to changes in pressure has been
known and studied for years. The magnitude of pressure effects
depends on the mixture. Sometimes, composition of azeotropes
change very little (e.g. the ethanol–water azeotrope). However,
there are mixtures where compositions of some azeotropes change
rapidly with pressure and even azeotropes that appear and dis-
appear as pressure varies. In our case, the DPE + PA azeotropic
composition is pressure-sensitive.
From the viewpoint of green chemical principles, additional
solvents should be avoided as much as possible in chemical
process. Following these principles, PSD seems to be more attrac-
tive and should be preferentially selected, compared to the
azeotopic/extractive distillation. However, it is not often exploited
commercially, because in many case the relative volatility remains
close to 1.0 at the top of the column (for minimum-boiling
azeotrope) or at the bottom (for maximum-boiling azeotrope). In
such cases, a high reflux ratio and a large number of equilibrium
stages are required to achieve complete separation, so the inten-
sity of energy consumption may lead to be a process economically
noncompetitive.
To investigate how the PSD works with the DPE + PA azeotropic
system, we have done a computer simulation of the vapour–liquid
equilibrium using ComThermo® Aspen v.2006.5 at different pres-
sures with the interaction parameters obtained from experimental
VLE data obtained by us [7]. Based on these results we have decided
to carry out the design and optimization of the pressure-swing
distillation process.
1.2. Extractive distillation
On the other hand, extractive distillation (ED) can be used to
separate the components of an azeotropic mixture adding a sol-
vent (entrainer) that is capable of strongly modifying the relative
volatility of the mixture. The synthesis and design of extractive dis-
tillation processes take place in two steps [14]. The first one involves
1267
Table 1
UNIQUAC binary interaction parameters.
Component i Component j Aij (J/mol) Aji (J/mol)
Di-n-propyl ether n-Propyl alcohol 3426.70–2.18Ta 2209.81–3.48Ta
Di-n-propyl ether 2-Ethoxyethanol 1984.08 −692.81
n-Propyl alcohol 2-Etoxyethanol 123.54 17.48
a
T is the temperature in Kelvin.
the selection of one or more candidate solvents (which facilitate
the separation by changing the relative volatilities in the mixture
through physical or chemical interactions with the original compo-
nents), and the choice of one or more column configurations. The
second step, process design, involves the search for optimal process
parameter values. The success of the second step depends on the
solutions obtained for the first one because efficiency in extrac-
tive distillation is largely determined by the choice of a suitable
entrainer.
In this work, based on the guidelines for the solvent screen-
ing, initially, it had been chosen five solvents: 1-pentanol [15],
n-butyl propionate [16], N,N-dimethylformamide [17] and 2-
ethoxyethanol [18]. Therefore, in order to be able to select the best
solvent among them, we have carried out simulations with Aspen
Hysys® v.2006.5 of Aspen Technology Inc., using the binary inter-
action parameters correlated from experimental data obtained for
all binaries involved and determined by our research group.
After all, the aim of this work is to study the influence of the
operation variable values a column configuration on the perfor-
mance of the DPE + PA separation by extractive distillation using an
entrainer and by swing-pressure distillation with the help of a com-
mercial simulator (Aspen Hysys® v.2006.5 of Aspen Technology
Inc.). Finally, we have chosen the best alternative for the sepa-
ration of the azeotropic mixture under study from the economic
point of view. The results from the study will provide basic design
information in applications associated with extractive distillation.
2. Simulation
2.1. Problem definition
The two alternatives considered in this study (PSD and ED) were
simulated starting from the same initial data. The feed is a mixture
made up of 50 mol% of di-n-propyl ether and 50 mol% of n-propyl
alcohol, with a flow rate of 12,000 Tm/year; we took 8000 working
hours per year, which is a mass flow of 1500 kg/h (∼ =18.49 kmol/h).
2.2. Property package
Computer simulation using commercial process simulators is
a useful tool to predict qualitatively the influence of the operating
variables on the column performance, provided that the interaction
binary parameters for the studied mixture are available in their
own databank. The accuracy of the simulated results is strongly
dependent on the quality of the binary parameters from the liquid-
phase activity coefficient models.
In this case, UNIQUAC activity model [19] was chosen and we
have used the binary interaction parameters published by us in
previous papers [15–18]. The parameters used are listed in Table 1.
2.3. Pressure-swing distillation
It is well known that, in some cases, changing the system pres-
sure can affect the vapour–liquid equilibrium (VLE) of a mixture.
This effect can be used to separate a binary mixture containing
a minimum boiling azeotrope, provided that this mixture signif-
icantly changes composition over a moderate pressure range.
1268 E. Lladosa et al. / Chemical Engineering and Processing 50 (2011) 1266–1274

1.0 Table 2
Specification of the design variables in the pressure-swing distillation process.

Variable Specification

0.8 Binary feed Temperature (◦ C) 50.0

Molar flow (kmol/h) 18.49
Molar composition 50% DPE
a Low pressure Pressure (kPa) 30.0
kP
20 column (LPC) Bottom purity of PA (molar%) 99.0
0.6
ydi-n-propyl ether

a High pressure Pressure (kPa) 101.3

kP
1 .3 column (HPC) Bottom purity of DPE (molar%) 99.0
10

0.4
According to that, the low pressure column (LPC) will work at
30 kPa and the high pressure column (HPC) at atmospheric pressure
(101.3 kPa).
0.2

2.3.2. Sequencing of the pressure-swing distillation process

0.665

0.0
0.0 0.2
Fig. 1. Estimated composition diagram for di-n-propryl ether + n-propyl alcohol
system at different pressures using the UNIQUAC model with the parameters given
in Table 1.
Fig. 1 shows the xy diagram for the di-n-propyl ether–n-propyl
alcohol system at pressures of 20 and 101.3 kPa, experimentally
determined by us in a previous paper [7]. As can be seen, the shift in
the composition from 82.3 to 66.5 mol% di-n-propyl ether indicates
that a pressure-swing system should be feasible.
2.3.1. Operating pressures selection
To investigate the pressure sensitivity of di-n-propyl ether + n-
propyl alcohol azeotropic mixture, we begin with a simulation of
the VLE at different pressures, using the UNIQUAC parameters listed
in Table 1. In Fig. 2, the di-n-propyl ether mole fraction and tem-
perature of the azeotrope are plotted as a function of pressure.
We can observe a significant pressure influence on the azeotropic
composition.
The actual operating pressure should be chosen within such
a range that water can be used as coolant for the overhead con-
denser and steam can be used as a heating medium for the reboiler.
1.0
0.9
0.8
x azeotrope di-n-propyl ether
0.823 For a binary mixture presenting a pressure-sensitive minimum
0.4 0.6 0.8 1.0 boiling point azeotrope, the separation sequence is formed by two
x di-n-propyl ether columns operating at different pressures [20]. For a binary mixture
forming a pressure-sensitive minimum boiling azeotrope, our case,
the separation sequence works as shown in Fig. 3, where it is repre-
sented the Txy diagram for this di-n-propyl ether + n-propyl alcohol
azeotropic mixture for both pressures (30 and 101.3 kPa). The fresh
feed, F0 , is mixed with the recycled stream from the second column
to form the feed stream, F1 , to the first column, which operates at
30 kPa. Because F1 lies to the left of the azeotrope at 30 kPa, pure
n-propyl alcohol can be obtained as a bottom product, B1 and a
mixture near the azeotropic composition at 30 kPa is the distillate,
D1 . Stream D1 is the feed stream to the next column with different
pressure, in this case 101.3 kPa. Because F2 (∼ =D1 ) now lied to the
right of the azeotrope at 101.3 kPa, the other pure component, di-n-
propyl ether, can recovered in the bottom product stream, B1 , and
a near azeotropic mixture becomes the distillate, D2 , for recycling
to the first column.
Fig. 4 shows the pressure-swing sequence for the separation of
DPE from PA. The feed enters the low pressure distillation column
at 30 kPa (LPC) and the distillate of this column has a composi-
tion closed to the low pressure azeotrope. This distillate is the feed
stream to the high pressure column at 101.3 kPa (HPC), which distil-
late has a composition that approaches the high pressure azeotrope.
This distillate has a composition that is similar to the feed compo-
sition and it is recycled to mix with the feed to LPC. High purity of
180 PA (99 molar%) is produced as a bottom stream from the LPC and
DPE (99 molar%) is produced as a bottom stream from the HPC.
160
2.3.3. PSD optimization
140 2.3.3.1. Partial optimization based on the total reboiler heat duty as

a reference variable. In order to select the best conditions to carry

0.7 120 out the global economical optimization, we start with a partial opti-
mization specifying some variables and using the total reboiler duty
T(ºC)

0.6 100 as a reference variable.

The variables chosen to be specified can be characterized as
0.5 80
either design variables or optimization variables. Design variables
are those whose values are set by market demands or physical con-
ditions. The design variables selected in this case are the flow rate,
0.4 60
composition and temperature of binary feed, the operating column
pressures and the purities of the bottom streams in both columns.
0.3 40
Table 2 shows the specifications chosen for all the previously men-
tioned variables. Once design variables have been specified, their
0.2 20 values remain constant throughout the optimization procedure. On
0 25 50 75 100 125 150 175 200
the contrary, optimization variables are those that must arbitrarily
P (kPa) be assigned a value. For the PSD process, the optimization variables
were the number of trays in each column, the recycle flow rate and
Fig. 2. Di-n-propyl ether mole fraction and temperature of the azeotrope as a
function of pressure. Solid line: azeotrope temperature vs. pressure. Dashed line: the LPC distillate composition. Their values are subject to change
azeotrope composition vs. pressure. as we proceed from the base case to the optimal design.
 E. Lladosa et al. / Chemical Engineering and Processing 50 (2011) 1266–1274 1269

Fig. 3. Temperature-composition diagram for di-n-propyl ether + n-propyl alcohol azeotropic mixture and distillation sequence.

The optimization procedure using Aspen Hysys® software
requires setting the number of trays in both columns. With the
short-cut design tool of Aspen Hysys® we did a preliminary opti-
mization (case studies) analyzing the variation of stage number and
reboiler heat duty (RHD) as a function of the reflux ratio, in order
to set the number of ideal trays and feed position for both columns.
From the results of these case studies (Fig. 5) we selected 12 ideal
trays (feed entry at stage 7) for the LPC and 12 ideal trays (feed
entry at stage 6) for the HPC.
With the design variable specifications showed in Table 2 and
once the number of trays was fixed, the system then converged suc-
cessfully in the rigorous facility of Aspen Hysys® . To carry out the
optimization, we studied six cases varying the LPC distillate com-
position from 0.665 to 0.779 molar fraction DPE, selecting in each
case the best recycle flow rate that minimizes the total reboiler heat
duty needed for both columns of PSD process. The variation range
of this optimization variable was established to produce reasonable
results. When the value approaches either of the azeotropic com-
positions at both pressures, the design is critical, or the system does
not converge or it requires very high reflux ratios. Fig. 6 shows the
variation of the total reboiler heat duty vs. the LPC distillate com-
position where it can be seen a clear minimum for the PSD process
corresponds to 0.765 mole fraction of DPE of the LPC column.
2.3.3.2. Economic evaluation. As it can be seen in Fig. 6 the
global optimum for the PSD process corresponds to a 0.765 mole

Fig. 4. Hysys process flow diagram (PFD) of pressure-swing distillation.

 1270 E. Lladosa et al. / Chemical Engineering and Processing 50 (2011) 1266–1274

(a) 17 7.0e+6
Table 3
Utility prices (Chemical Engineering Plant Cost Index 2009 = 521.9 and project life 10
years).
16
6.0e+6 Capital cost
15 Utility Price
5.0e+6 Low-pressure steam (D /t) 10

RHD (kJ/h)
14
Cooling water (D /m3 ) 0.04
stages

Electricity (D /kWh) 0.07

13 4.0e+6
Solvent (2-ethoxyethanol)a (D /kg) 1.3
a
12 Advanced Petrochemicals Ltd.
3.0e+6
11
2.0e+6 In order to compare the PSD process with the ED process, it
10 will be necessary to evaluate economically the PSD optimum, that
is, calculate the total annual costs (TAC) [21], using the following
9 1.0e+6
1 2 3 4 5 objective function:
Reflux ratio
TAC (D /year) = CV + Cf + (ir + im ) · FCI (1)
(b) 16 5.0e+6
where CV is the process variable cost, mostly annual utility con-
15 sumption (steam, cooling water and electrical power); Cf the annual
4.0e+6
fixed costs, i.e. maintenance and wages; FCI is the fixed capital
14 investment; ir is the fixed capital recovery rate applied to FCI; im is
the minimum acceptable rate of return on FCI.
RHD (kJ/h)

13 In this economic evaluation, Cf was assumed to be 10% of FCI; ir

stages

3.0e+6 was 8.3%, corresponding to a lineal recovery of 12 years and im was

12 supposed to be 7% of FCI; hence Eq. (1) could be rewritten as:

11 TAC (D /year) = CV + 0.30 · FCI (2)

2.0e+6

10 Fixed capital investment (FCI) was estimated using the costs

estimation program CAPCOST® of Turton et al. [22] with an updated
9 1.0e+6 Chemical Engineering Plant Cost Index (CEPCI 2009 = 521.9). For eco-
1 2 3 4 5 nomic evaluation, a 10-year project life is considered. Variable costs
Reflux ratio were obtained applying the utility costs from Table 3.
Details of the columns for PSD process together with the total
Fig. 5. Variation of stage number (—) and reboiler heat duty (– –) as a function of
annual costs (TAC) for this case are listed in Table 4. In Appendix A
the reflux ratio for the LPC (a) and HPC (b).
are listed the estimated capital investment of each individual unit.

fraction of DPE. This optimum is based on the minimum reboiler

heat duty but it should correspond with the global economic opti-
mum, since, having previously fixed the number of trays in each
column, the operation costs (mainly the total reboiler heat duty)
clearly dominate the total costs.
6.0e+6
2.4. Extractive distillation
Extractive distillation has the inherent disadvantage of intro-
ducing a third component into the system (the solvent) that will
appear in the product streams, which necessarily have to be recov-
ered. So, the extractive system must show significant economic
advantages over the pressure-swing system to make it the process
of choice.

5.6e+6
2.4.1. Solvent selection
Since the solvent is the core of extractive distillation, more
Total RHD(kJ/h)

5.2e+6
attention should be paid on the selection of potential solvents.

4.8e+6
Table 4
Global economic optimum for PSD process.
4.4e+6
Design parameters Low pressure High pressure
column (LPC) column (HPC)
4.0e+6
Number of stages 12 12
Feed (top down stage number) 7 7
3.6e+6 Assumed tray efficiency (%) 70 70
0.71 0.72 0.73 0.74 0.75 0.76 0.77 0.78 0.79 Number of trays 18 18
xDPE in LPC column Reflux ratio 0.72 1.07
Reflux rate (kg/h) 3622.19 4413.57
RHD (kJ/h) 1.89 × 106 1.92 × 106
Fig. 6. Total reboiler heat duty vs. azeotropic composition of low pressure column
Total annual costs (TAC) = 454 × 103 D /year
in the pressure-swing distillation process.
 E. Lladosa et al. / Chemical Engineering and Processing 50 (2011) 1266–1274
Fig. 7. Hysys process flow diagram (PFD) of extractive distillation.
The influence of the solvent is usually quantified in terms of the
so-called selectivity, Sij , which is defined as:
(˛Tij )
Sij = (3)
(˛Bij )
where (˛Tij ) is the relative volatility of the keys components (DPE
and PA) in presence of a solvent (ternary mixtures i + j + solvent);
(˛Bij ) is the relative volatility of the keys components (binary mix-
ture i + j).
At low pressures, due to the vapour pressure relation with the
components keys is practically constant with and without solvent,
Eq. (3) can be rewritten as:
T
(i /j )
Sij = B
(4)
(i /j )
where  i and  j are the activity coefficients of the components i
and j.
As the activity coefficients depend on the phase composition and
the solvent effect tends to increase as its concentration increases,
it is common practice to consider, at least in a preliminary solvent
selection, the situation at infinite dilution. The selectivity at infinite
dilution is defined as the ratio of the activity coefficients of both key
components at infinite dilution in the solvent:
T reference variable. As in the PSD process, in order to select the best
(i∞ /j∞ )
Sij∞ = B
(5)
(i∞ /j∞ )
The extractive agent should have the following characteristics:
lower volatility than the volatilities of the mixture components,
increases the relative volatility of the mixture, preferably does not
form an azeotrope with any of the mixture components, and is read-
ily available and cheap. In previous papers [15–18] it has been made
an exhaustive study to separate the DPE + PA azeotropic mixture
1271
with different solvents. Table 5 shows some characteristics of the
different solvents. It can be observed that N,N-dimethylformamide
is, beyond all doubt, the best solvent studied in order to achieve
the separation, although 2-ethoxyethanol could be a good solvent.
Nevertheless, according environmental aspects (2-ethoxyethanol
is much less aggressive than N,N-dimethylformamide) it was
decided to made the study of the extractive distillation with 2-
ethoxyethanol as entrainer.
2.4.2. Sequencing of extractive distillation process
After the entrainer has been selected, attention is directed to the
sequence of the distillation columns. The process configuration is
shown in Fig. 7, in which the solvent is added at the top trays of the
extractive column (EC). Since 2-ethoxyethanol is much less volatile
than either DPE or PA, it flows down the column, dragging to PA, to
leave with the bottom product.
The solvent recovery column (SRC) removes PA from 2-
ethoxyethanol. This is an easy separation because the solvent is
much less volatile than PA. The lean solvent is then cooled and recy-
cled back to the extractive column. If the recovery of the solvent, in
this column, is high a very small amount of pure solvent make-up
is required to maintain the solvent-to-feed ratio constant.
2.4.3. Optimization
2.4.3.1. Partial optimization based on the total reboiler heat duty as a
Table 5
Normal boiling point and selectivity of the different solvents for DPE + PA separation.
Solvent Teb (K) Sij∞
1-Pentanol 410.62 1.924
n-Butyl Propionate 418.69 0.787
N,N-dimethylformamide 425.63 5.092
2-Ethoxyethanol (Cellosolve) 407.95 2.756
1272 E. Lladosa et al. / Chemical Engineering and Processing 50 (2011) 1266–1274

Table 6 Table 7
Specification of the design variables in the extractive distillation process. Specifications in the extractive column for each case.

Variable Specification Case

Feed streams EC-1 EC-2 EC-3 EC-4 EC-5 EC-6

Binary feed Temperature (◦ C) 85.0
Molar flow (kmol/h) 18.49 Ideal trays number 35 37 40 45 50 55
Molar composition 50% DPE Solvent entry stage (optimum) 12 13 14 15 16 17
Solvent make-up Temperature (◦ C) 25 Feed entry stage (optimum) 26 28 31 35 40 44
Molar composition 100% 2-ethoxyethanol
Extractive column
Distillate Purity of DPE (molar%) 99.0
Recovery of DPE 99.9% the point that the curvature is most sharply changing, in this col-
Solvent recovery column umn between 18 and 22 ideal trays. Therefore, we decided to set
Distillate Purity of PA (molar%) 99.0
the number of ideal trays for this column at 22 (feed entry at stage
Recovery of PA 99.9%
number 10).
In the extractive column, due to it is a ternary mixture, we can-
not use the short-cut column design facility so, we have studied
six different cases varying the number of ideal trays from 35 to 55
conditions to carry out the ED separation process, we start with a
(fewer than 35 do no converge whatever the solvent-to-feed ratio
partial optimization specifying some variables and using the total
is; more than 55 is not reasonable), selecting in each case the best
reboiler duty as a reference variable. As previously, we specified the
solvent and feed entry stage. In Table 7 the six cases are specified.
temperature, pressure, flow rate and composition of binary feed. In
With the specification showed in Table 6 and for the cases spec-
addition, we also specified distillate purity and recovery of DPE in
ified in Table 7, the columns were then evaluated with the rigorous
the extractive column and distillate purity and recovery of PA in
steady-state column facility of Aspen Hysys® and converged suc-
the solvent recovery column. Finally, solvent make-up was cho-
cessfully for all six cases. From previous simulation runs, it was
sen as a pure component at ambient conditions. All the previously
found that the solvent-to-feed ratio is very critical, having a signif-
mentioned variables are taken as design variables in this problem
icant effect on the reflux ratio in both columns, and therefore on
and the selected values are shown in Table 6. Once design variables
the total reboiler heat duty (RHD), so the partial optimization was
have been specified, their values remain constant throughout the
made over this variable. In each case, the solvent-to-feed ratio was
optimization procedure. In this case, extractive distillation process,
then adjusted to minimize the total RHD needed for the process.
the solvent-to-feed ratio in the extractive column and the number
Table 8 shows a summary of the partial optimization base on total
of trays in each column are optimization variables.
reboiler heat duty as a reference variable.
As in the previous case, using the short-cut design facility of
Aspen Hysys® we did a case study analyzing the variation of stage
number and reboiler heat duty as a function of the reflux ratio, in 2.4.3.2. Global economic optimization. As it can be seen in Table 8
order to set the number of ideal trays and solvent position of the the minor reboiler heat duty obviously corresponds to EC-6 case.
solvent recovery column. Fig. 8 shows a plot of theoretical num- However, this case requires a big number of stages in the extrac-
ber of stages and reboiler heat duty vs. reflux ratio. In general, for tive columns; so, it would be an expensive column. Therefore, to
the higher reflux region, utility consumptions are very large even determine the best of these six alternatives it is necessary to carry
though initial equipment costs are relatively low, while for the low out an economic evaluation based on the minimum total annual
reflux region, initial investment costs are low. According to the costs (TAC), using the objective function described by Eqs. (1) and
results obtained, using the short-cut column utility, the theoret- (2) and the utility costs of Table 3.
ical stage region that minimizes the total costs will be located at All variables have the same meaning as for the PSD process
although in this case, the process variable costs (CV ) should include
the 2-ethoxyethanol make-up cost.
Fig. 9 shows the calculated TAC data for the six cases. As it can
32 2.4e+6 be observed, there is a minimum for EC-4 case (solvent to feed
ratio = 1.9). Table 9 lists details of extractive and solvent recovery
30 columns needed to meet the design objectives for this optimum.
2.2e+6 Moreover, in Appendix B are listed the estimated capital invest-
28
ment for each individual unit.
26 2.0e+6
2.5. Alternatives comparison
RHD (kJ/h)

24
stages

1.8e+6 As can be seen in Tables 4 and 9, the annual costs estimated

22 for the PSD process are quite lower than the ones corresponding
to the ED process (454 × 103 D against 644 × 103 D ) both obtained
20 1.6e+6

18
Table 8
1.4e+6
Optimum solvent-to-feed ratio in the extractive column for each case.
16
Case Solvent-to-feed ratio Total reboiler duty (kJ/h)
14 1.2e+6 EC-1 3.8 4.58 × 106
2.0 2.5 3.0 3.5 4.0 4.5 5.0
EC-2 3.0 3.93 × 106
Reflux ratio EC-3 2.4 3.40 × 106
EC-4 1.9 2.93 × 106
Fig. 8. Solvent recovery column case study in the extractive distillation process. EC-5 1.6 2.69 × 106
Variation of stage number (—) and reboiler heat duty (– –) as a function of the reflux EC-6 1.5 2.55 × 106
ratio.
 E. Lladosa et al. / Chemical Engineering and Processing 50 (2011) 1266–1274 1273

alternatives would be less although the PSD process would always

7.2e+5 be favourable. On the other hand, the cost of the steam for the
reboilers is, by a wide margin, the most important of the util-
ity costs, which justifies the decisions previous to carrying out
7.0e+5 the partial optimization (in those cases in which there is no FCI
variation) only based on these costs. As it can be seen, the cost
TAC (€/year)

of steam is clearly higher in the PSD process than in the ED

6.8e+5 process, because the reflux rates are higher in the first process
(Tables 4 and 9).
All these results have been obtained for a practical case base on
6.6e+5 a plant for the treatment of 12,000 Tm/year of a DPE + PA mixture.
If we consider larger plants, the difference between both alter-
natives becomes less and less, precisely because the steam costs
6.4e+5 grow almost proportionally to the flow rate, while the costs asso-
ciated with the investment grow much more slowly, an therefore
EC-1 EC-2 EC-3 EC-4 EC-5 EC-6 the total costs of both alternatives are much closer. We have to
Case consider, however, that when we increase the size of the plant
significantly, the costs of the make-up solvent grow in the same
Fig. 9. Total annual cost (TAC) for the different cases in the extractive distillation. proportion, if we maintain the specification of purity and recovery
of the products.

Table 9
Global economic optimum for extractive distillation process (EC-4). 3. Conclusions
Design parameters Extractive Solvent recovery
column (EC) column (SRC)
The simulation of processes with a commercial software pro-
gram (Aspen Hysys® ) used appropriately is a very powerful tool to
Number of stages 45 22
analyze the separation of complex mixtures.
Feed (top down stage number) 15 (solvent) 10
35 (binary In this case, steady-state comparisons have been presented of a
feed) pressure-swing distillation process and an extractive distillation
Assumed tray efficiency (%) 70 70 process to separate the di-n-propyl ether and n-propyl alcohol
Number of trays 65 32
azeotropic mixture.
Reflux ratio 3.59 1.91
Reflux rate (kg/h) 3414.64 1064.43
The computer simulation and economic evaluation of the two
RHD (kJ/h) 1.72 × 106 1.22 × 106 separation alternatives allow us to conclude that, to process
Total annual costs (TAC) = 644 × 10 D /year
3
12,000 Tm/year (approximately, 1500 kg/h), the process that uses
PSD is much more attractive in terms of steady-state economics.
Capital investment is very significant lower.

using the same criteria and evaluation procedures. This result

may seem surprising, since usually the ED is a process more Acknowledgements
attractive than PSD process; provided that an appropriate sol-
vent is used in the extractive column and that it becomes easy to Financial support from the Ministerio de Ciencia y Tecnología of
recover. Spain, through project No. CTQ2010-18848 is gratefully acknowl-
We will analyze more slowly the structure of the total annual edged.
costs. In Table 10 we can see more precisely the details of the
annual costs for each separation alternative. The first thing that Appendix A.
calls our attention in Table 10 is the difference of the capital
investment value between both alternatives, because of the big
size of the columns in the ED process (see Table 9) needed to
obtain the desired purity of the products and recover the solvent Estimated capital investment of each individual unit in the pressure-swing distilla-
efficiently. This big difference of FCI leads to an important differ- tion (PSD) process.
ence of the costs associated with the capital investment (mainly 103 D
fixed capital recovery and minimum acceptable rate of return).
Low pressure column (LPC)
At the end, these costs determine the difference between both Tower + trays 349.46
alternatives. By lowering these costs the difference between both Reboiler 38.11
Condenser 25.09
Reflux pump 50.19
Table 10 Reflux vessel 40.89
Summary of economic results. Vacuum system 97.59
Total 601.33
Pressure-swing Extractive
distillation distillation High pressure column (LPC)
Fixed capital investment (103 D ) 1062.32 1698.03 Tower + trays 304.85
Cost proportional to FCI (103 D /year)a 286.82 431.10 Reboiler 19.52
Steam (103 D /year) 146.48 112.86 Condenser 18.59
Cooling water (103 D /year) 18.85 14.80 Reflux pump 50.19
Electric power (103 D /year) 2.05 0.37 Reflux vessel 41.82
Solvent make-up (103 D /year) – 84.81 Total 434.97
Total annual costs (103 D /year) 454.20 643.94 Recycled pump 26.02
a
Cost proportional to FCI = Cf + (ir + im )·FCI. Fixed capital investment for PSD process 1062.32
1274 E. Lladosa et al. / Chemical Engineering and Processing 50 (2011) 1266–1274

Appendix B. [7] E. Lladosa, J.B. Montón, M.C. Burguet, R. Muñoz, Isobaric vapor–liquid equi-
libria for the binary systems 1-propyl alcohol + dipropyl ether and 1-butyl
alcohol + dibutyl ether at 20 and 101.3 kPa, Fluid Phase Equilib. 247 (2006)
47–53.
Estimated capital investment of each individual unit in the extractive distillation [8] R. Muñoz, J.B. Montón, M.C. Burguet, J. de la Torre, Separation of isobutyl
(ED) process. alcohol and isobutyl acetate by extractive distillation and pressure-swing
distillation: Simulation and optimization, Sep. Purif. Technol. 50 (2006)
103 D
175–183.
Extractive column (EC) [9] W.K. Lewis, Dehydrating alcohol and the like, U.S. Patent, 1,676,700, July 10,
Tower + trays 918.25 1928.
Reboiler 21.38 [10] C. Black, Distillation modelling of ethanol recovery and dehydration processes
for ethanol and gasohol, Chem. Eng. Prog. 76 (1980) 78–85.
Condenser 17.66
[11] S.I. Abu-Eishah, W.L. Luyben, Design and control of two-column azeotropic col-
Reflux pump 57.62
umn azeotropic distillation system, Ind. Eng. Chem. Proc. Des. Dev. 24 (1985)
Reflux vessel 33.45
132–140.
Total 1048.36 [12] T. Chang, T.T. Shih, Development of an azeotropic distillation scheme for purifi-
Solvent recovery column (SRC) cation of tetrahydrofuran, Fluid Phase Equilib. 52 (1989) 161–168.
[13] J.R. Phimister, W.D. Seider, Semicontinuous, pressure swing distillation, Ind.
Tower + trays 486.11
Eng. Chem. Res. 39 (2000) 122–130.
Reboiler 20.44
[14] I. Rodriguez-Donis, V. Gerbaud, X. Joulia, Entrainer selection rules for the sepa-
Condenser 17.66
ration of azeotropic and close-boiling-temperature mixtures by homogeneous
Reflux pump 36.25 batch distillation process, Ind. Eng. Chem. Res. 40 (2001) 2729–2741.
Reflux vessel 28.81 [15] E. Lladosa, J.B. Montón, M.C. Burguet, R. Muñoz, Vapor–liquid equilibria in the
Total 589.27 ternary system dipropyl ether + 1-propanol + 1-pentanol and the binary sys-
tems dipropyl ether + 1-pentanol, 1-propanol + 1-pentanol at 101.3 kPa, Fluid
Cooler 33.45
Phase Equilib. 247 (2006) 175–181.
Recycled pump 26.95
[16] E. Lladosa, J.B. Montón, M.C. Burguet, R. Muñoz, Isobaric vapor–liquid equilibria
Fixed capital investment for ED process 1698.03 for binary and ternary mixtures of dipropyl ether, 1-propyl alcohol, and butyl
propionate, J. Chem. Eng. Data 51 (2006) 2233–2238.
[17] E. Lladosa, J.B. Montón, M.C. Burguet, R. Muñoz, Phase equilibria involved
in extractive distillation of dipropyl ether + 1-propyl alcohol using N,N-
References dimethylformamide as entrainer, J. Chem. Eng. Data 52 (2007) 532–537.
[18] E. Lladosa, J.B. Montón, M.C. Burguet, R. Muñoz, Phase equilibria involved
in extractive distillation of dipropyl ether + 1-propyl alcohol using 2-
[1] T.C. Frank, Break azeotropes with pressure-swing sensitive distillation, Chem.
ethoxyethanol as entrainer, Fluid Phase Equilib. 255 (2007) 62–69.
Eng. Prog. 12 (1997) 52–63.
[19] D.S. Abrams, J.M. Prausnitz, Statistical thermodynamics of liquid mixtures: a
[2] M.F. Doherty, M.F. Malone, Conceptual Design of Distillation Systems, McGraw-
new expression for the excess Gibbs energy of partly or completely miscible
Hill, New York, 2001.
systems, AIChE J. 1 (1975) 116–128.
[3] J.W. Drew, Design for solvent recovery, Chem. Eng. Prog. 71 (1975) 92–99.
[20] J.D. Seader, E.J. Henley, Separation Process Principles, Wiley, 2006.
[4] D.R. Garg, J.P. Ausikaitis, Molecular sieve dehydration cycle for high water con-
[21] P. Langston, N. Hilal, S. Shingfield, S. Webb, Simulation and optimization
tent streams, Chem. Eng. Prog. 79 (1983) 60–64.
of extractive distillation with water as solvent, Chem. Eng. Proc. 44 (2005)
[5] H.L. Fleming, Consider membrane pervaporation, Chem. Eng. Prog. 88 (1992)
345–351.
45–52.
[22] R. Turton, R.C. Bailie, W.B. Whiting, J.A. Shaeiwitz, Synthesis and Design of
[6] W.F. Furter, Extractive distillation by salt effect, Can. J. Chem. Eng. 55 (1977)
Chemical Processes, Prentice Hall, New Jersey, 1998.
229–239.