Journal of the Taiwan Institute of Chemical Engineers 91 (2018) 251–265

Contents lists available at ScienceDirect

Journal of the Taiwan Institute of Chemical Engineers

journal homepage: www.elsevier.com/locate/jtice

Energy-efficient and ecologically friendly hybrid extractive distillation

using a pervaporation system for azeotropic feed compositions in

alcohol dehydration process

Felicia Januarlia Novita

a , Hao-Yeh Lee

b , ∗, Moonyong Lee

a , ∗

a School of Chemical Engineering, Yeungnam University, Dae-dong 712-749, Republic of Korea b Department of Chemical Engineering, National Taiwan University of Science and Technology, Taipei 10607, Taiwan

a r t i c l e i n f o

Article history:

Received 20 January 2018

Revised 25 April 2018

Accepted 16 May 2018

Available online 10 July 2018

Keywords:

Alcohol dehydration process

Extractive distillation

Pervaporation-distillation hybrid system

Azeotropic feed compositions

Economic savings

a b s t r a c t

A hybrid extractive distillation column (EDC) with high selectivity pervaporation (PV) is implemented in

three alcohol dehydration processes. For these dehydration processes, glycerol is chosen as an entrainer.

Initially, a conventional configuration using two columns in sequence, one of which is an EDC and the

other is a recovery column (RC) for separating the glycerol/water mixture, is introduced. However, such

a conventional configuration is expensive and requires high energy owing to the low operating pressure

conditions in the RC to avoid glycerol degradation. To overcome these limitations, a high selectivity PV

from a cellophane membrane is proposed to replace the RC used in the conventional configuration for

separating the glycerol/water mixture. In the hybrid configuration, the membrane cost increases owing to

the presence of some glycerol at the bottom of the EDC. However, the operating cost reduces due to less

energy requirement. Consequently, the simulation results show that the hybrid configuration can save up

to 25% and 41% of the total annual cost and energy, respectively, compared to those of the conventional

configuration. The proposed hybrid configuration for the alcohol dehydration process using glycerol as

an entrainer is a promising technology for real-life industrial applications owing to the availability of a

commercial cellophane membrane.

© 2018 Taiwan Institute of Chemical Engineers. Published by Elsevier B.V. All rights reserved.

1

p

t

t

t

l

m

fl

s

l

t

u

w

t

e

n

r

t

n

w

u

i

i

n

c

c

l

a

d

i

c

p

n

h

1

. Introduction

In the chemical industry, distillation is one of the most im-

ortant separation processes, mainly because it allows the separa-

ion of ideal and non-ideal mixtures in large-scale units. Moreover,

he separation of homogeneous and heterogeneous azeotropic mix-

ures has received significant attention from the industry, particu-

arly from an energy perspective [1,2] . Several types of distillation

ethods are used, including simple distillation, partial distillation,

ash distillation (equilibrium distillation), rectification, pressure-

wing distillation [3–5] , azeotropic distillation [6,7] , reactive distil-

ation [8] , and the use of an extractive distillation column (EDC). Of

hese distillation methods, EDC is one of the well-known methods

sually implemented in the separation process [9,10] . An entrainer,

hich is the heaviest component in a mixture, is added to the sys-

em to separate the minimum-boiling binary azeotropes [11] . The

ntrainer does not form any azeotropes with the original compo-

ents and is completely miscible with them at all amounts. The

∗ Corresponding authors.

E-mail addresses: haoyehlee@mail.ntust.edu.tw (H.-Y. Lee), mynlee@yu.ac.kr

(M. Lee).

I

o

m

a

ttps://doi.org/10.1016/j.jtice.2018.05.023

876-1070/© 2018 Taiwan Institute of Chemical Engineers. Published by Elsevier B.V. All r

elative volatility of the original two components increases when

his heavy entrainer is added into the system. One original compo-

ent can go directly to the top part of the column, and the other

ill go along with the heavy entrainer to the bottom of the col-

mn.

Many aqueous alcohol solutions such as ethanol (EtOH)/water,

sopropanol (IPA)/water, and tert–butyl alcohol (TBA)/water used

n various chemical and biotechnologies are usually diluted and

eed to be separated. An EtOH/water mixture is of great industrial

oncern, owing to its potential as a renewable energy source that

an be used as a supplement or complete replacement to gaso-

ine, as well as a raw material for use in alcohol industry. EtOH is

verified clean-burning fuel. Therefore, the use of EtOH can also

ecrease the amount of pollution released into the air [12] . IPA

s a basic chemical and solvent used in the production of many

hemicals, intermediates, and washing cleaners for semiconductor

rocesses. However, IPA is commonly mixed with water and can-

ot be used or reused directly, and hence, studies on separating

PA/water mixtures are worth conducting [13] . Anhydrous TBA is

ften used in the pharmaceutical industry for producing artificial

usk, essence, and raw medicines [14] . It is well known that these

lcohols form azeotropes with water, and cannot be extracted from

ights reserved.

252 F.J. Novita et al. / Journal of the Taiwan Institute of Chemical Engineers 91 (2018) 251–265

Fig. 1. (a) Isovolatility curve of EtOH-water-GLY and (b) EtOH-water-EG systems.

a

i

t

a

d

i

n

[

t

o

c

t

t

t

v

t

t

a

p

d

r

i

o

f

m

C

f

b

i

c

u

e

i

l

f

t

a

a

w

p

E

t

aqueous solutions at a high concentration through conventional

distillation methods [15] , and thus, a typical dehydration process

system is applied to obtain these alcohols in high purity from an

azeotropic feed composition using EDC distillation with a heavy

entrainer.

Before designing a distillation sequence in an EDC system, en-

trainer screening is an important step [16–18] . The presence of the

entrainer is the key difference between an EDC and simple distil-

lation, offering a greater degree of freedom in the overall system

design. An isovolatility curve is a useful tool for the selection of a

suitable entrainer and for determining the desirable product in the

distillate stream of an EDC [19] . Fig. 1 (a) and 1(b) show isovolatility

curves of an EtOH-water-glycerol (GLY) system and an EtOH-water-

ethylene glycol (EG) system, respectively.

From Fig. 1 , it can be seen that glycerol is an effective entrainer

for an EtOH-water system. As glycerol is added, the triangle area

above the isovolatility line in Fig. 1 (a) (the dashed line near the

EtOH apex of the triangle) becomes smaller than that in Fig. 1 (b).

The isovolatility curve also shows that EtOH is a distillate product

in the system, because when glycerol is added to the system, the

relative volatility becomes greater than 1, causing EtOH to move

up the column [19] .

Furthermore, in previous studies [20–22] , the researchers fo-

cused on the dehydration processes of EtOH, IPA, and TBA using

glycerol as an entrainer. Glycerol is a byproduct in the production

of biodiesel, and is nontoxic, environmentally friendly, highly avail-

able, and inexpensive [23,24] . These features have stimulated the

exploration of new uses, e.g. , as an entrainer for EDC distillation.

The isovolatility curve and the previous research results show that

glycerol can be used for bioethanol dehydration [25–27] , replac-

ing ethylene glycol, which is toxic and environmentally unsafe, and

may become prohibited in the future [27] .

The present study describes the dehydration processes of EtOH,

IPA, and TBA. A process flowsheet of the dehydrate EtOH process

developed by Gil et al. [20] was taken as a reference for the three

dehydration processes, as well as for glycerol, which is as an en-

trainer. However, the product specifications for each dehydration

process are set differently depending on the use of each type of

alcohol. There are two distillation columns, one EDC and one re-

covery column (RC). The RC is used to separate glycerol from wa-

ter and later recycle the glycerol back to the EDC. The separation

of a glycerol/water mixture requires a process with a low operating

pressure condition because glycerol decomposes at high tempera-

tures (290 °C at normal pressure). Separation processes that use a

very low operating pressure, in which water cannot be utilized as

coolant, are usually very expensive, not only monetarily but also

n terms of the energy required for separation. Therefore, reducing

he total annual cost (TAC) and energy requirement has become

n important issue for the separation of a glycerol/water mixture

uring the dehydration process.

Nowadays, the need to reduce the energy consumption in the

ndustry is increasingly attracting attention, indicating that it is

ecessary to further develop and improve the existing processes

28] . Membrane technology can be an alternative to the separa-

ion technology because the former technology does not depend

n the volatility of the components. A membrane separates the

omponents based on the difference in the partial pressures of

he permeant across the membrane [29] . To create a difference in

he partial pressure in a membrane, a vacuum pump is applied

o the permeating part. This type of separation offers several ad-

antages, such as low energy consumption and high selectivity. Of

he different types of membrane technology methods, pervapora-

ion (PV) is a promising method. In PV, the separation between

binary or multicomponent liquid mixture is introduced through

artial vaporization using a lipophilic membrane. The feed mixture

irectly interacts with one side of the membrane. The permeate is

emoved as a vapor phase through the opposite side and moves

nto a vacuum pump before being condensed, and separation can

ccur based on the liquid-vapor phase change in the PV [30,31] .

Many studies have focused on the membrane materials used

or each type of system. One of the most common commercial

embranes used in separation processes is a cellulosic membrane.

ellulose has been well known over the last 30 years as a pre-

erred membrane material because it is an ecologically friendly

iopolymer, is compatible with biological compounds, is relatively

nexpensive, and has hydrophilic properties [32] . Cellophane is a

ommercial cellulosic membrane. Biswas et al. [33] studied a PV

nit using a cellophane membrane for the dehydration of a glyc-

rol/water mixture.

In this study, a hybrid configuration using a membrane system

s proposed. Note that the membrane system is not directly uti-

ized to separate the alcohol and water, which makes it different

rom the work by Malekpour et al. [34] that utilized a membrane

o separate IPA from water directly. The aim of the membrane

pplication in this study is to develop a cheaper method to sep-

rate glycerol/water mixtures. Consequently, an innovative EDC

ith a hybrid PV configuration is suggested for the dehydration

rocess of EtOH, IPA, and TBA. Based on the design procedure,

DC is applied to obtain pure EtOH/IPA/TBA at the top stream of

he EDC. The membrane unit, PV, is used for further purification

F.J. Novita et al. / Journal of the Taiwan Institute of Chemical Engineers 91 (2018) 251–265 253

Table 1

NRTL binary parameters for the EtOH-water-glycerol system.

Component i Glycerol Glycerol Water

Component j Water EtOH EtOH

aij −0.7318 0 3.4578

aji −1.2515 0 −0.8009

bij (K) 170.917 36.139 −586.081

bji (K) 272.608 −586.081 246.18

cij 0.3 0.3 0.3

o

p

p

a

c

P

u

P

f

E

2

2

2

a

p

i

a

e

t

s

a

a

N

2

8

b

o

t

I

r

[

o

i

i

t

p

c

m

m

[

2

g

b

s

t

t

Table 2

NRTL binary parameters for the IPA-water-glycerol system.

Component i Water IPA Water

Component j Glycerol Glycerol IPA

aij 0 0 5.3852

aji 0 0 −2.5041

bij (K) 617.62 259.42 −1005.06

bji (K) −499.09 402.3 850.87

cij 0.3 0.3 0.3

Table 3

NRTL binary parameters for the TBA-water-glycerol system.

Component i Water TBA TBA

Component j Glycerol Glycerol Water

aij −1.2515 9.46553 −0.6 86 8

aji −0.7318 −1.53891 7.0893

bij (K) 272.608 −1659.26 203.419

bji (K) 170.917 58.5991 −1372.38

cij 0.3 0.3 0.3

c

a

o

o

p

2

e

p

e

t

t

u

t

m

r

m

m

t

b

n

w

c

κ

p

n

p

(

i

i

(

p

w

d

t

f the glycerol/water mixture. A sequential iterative optimization

rocedure is applied to optimize several design variables in each

roposed configuration for minimizing the TAC. Six configurations

re proposed for EtOH, IPA, and TBA: three conventional EDC

onfigurations and three hybrid EDC and PV configurations. Aspen

lus V10.0 is used as a commercial simulator for running the sim-

lations, and Aspen Custom Modeler V10.0 is used to develop the

V unit. The PV unit is integrated with Aspen Plus V10.0. The TAC

or the distillation column is calculated using an Aspen Process

conomic Analyzer (APEA) [35] provided by Aspen Plus V10.0.

. Model development

.1. Thermodynamic models

.1.1. EtOH-water-glycerol system

An EtOH/water mixture has a minimum-boiling homogeneous

zeotrope at 78.1 °C (at atmospheric pressure) with an EtOH com-

osition of 89 mol%. Accordingly, this mixture cannot be separated

n a single distillation column, and if fed to a column that operates

t atmospheric pressure, the purity of EtOH in the distillate cannot

xceed 89 mol%, whereas high purity water can be obtained from

he bottom. The NRTL physical property model [36] is used to de-

cribe the non-ideality of the liquid phase, and the vapor phase is

ssumed to be ideal. The detailed NRTL model binary parameters

re taken from the Aspen Plus database [37] . Table 1 shows the

RTL binary parameters of the EtOH-water-glycerol system.

.1.2. IPA-water-glycerol system

Along with Gmehling [38] , IPA (normal boiling point of

2.35 °C) has a minimum-boiling azeotrope with water (normal

oiling point of 100 °C) at 1 atm with an azeotropic temperature

f 80.1 °C and azeotropic composition of 68.70 mol% IPA. Therefore,

hese two components cannot be separated into a single column.

n this study, glycerol is chosen as an entrainer to aid in the sepa-

ation. The reason for choosing glycerol, as reported by Zhang et al.

21] , is that, at the EDC operating temperature, the vapor pressure

f glycerol is very low and thus the properties of glycerol are sim-

lar to those of ionic liquids [39,40] , which have been widely stud-

ed for their potential use in EDC distillation [41–46] .

For this three-component system, the NRTL model will be used

o describe the non-ideality of the liquid phase, whereas the vapor

hase is assumed to be ideal. The NRTL model parameters of the

omponents are taken from Zhang et al. [21] . The complete NRTL

odel parameters are shown in Table 2 . All other physical property

odel parameters are taken from the built-in values of Aspen Plus

37] .

.1.3. TBA-water-glycerol system

Liquid-vapor equilibrium data for a binary mixture of TBA-

lycerol were reported by Aniya et al. [22] . TBA has a minimum-

oiling azeotrope with water at 1 atm with an azeotropic compo-

ition, namely, a TBA mole fraction of 0.6259 and an azeotropic

emperature of 79.9 °C. The NRTL model is used to predict the ac-

ivity coefficients, as well as the ideal gas model for the fugacity

oefficient. The binary interaction parameters for the NRTL model

re taken from the Aspen Plus database [37] , with the exception

f the TBA-glycerol pairing, which is obtained through a regression

f the experimental data from Aniya et al. [22] . The NRTL binary

arameters of the TBA-water-glycerol system are shown in Table 3 .

.2. Membrane model

Biswas et al. [33] reported the influence of the process param-

ters for the dehydration of glycerol-water mixtures with perva-

oration both experimentally and through a simulation. From their

xperiment results, pure water was found as the permeant because

he membrane is selective to water. The selectivity coefficient of

he membrane is infinite [33,47,48] . In this study, the membrane

nit used is a commercially available cellophane film, the same as

hat used by Biswas et al. Based on this model, the rate of per-

eation is mainly controlled by the mass transfer boundary layer

esistance. The effects of the process parameters on the rate of per-

eation are estimated and verified using a simple unsteady-state

odel for the transport of solute under a concentration polariza-

ion regime [33] . The flux through a membrane for pure water can

e written as

w

=

D wg

δm

(C

u w

− C

d w

)(1)

here C

u w

and C

d w

are correlated with the bulk concentrations.

The permeability of pure water is temperature dependent and

an be expressed through the Arrhenius relation:

= κ0 exp ( −E / RT ) . (2)

Therefore, the flux equation becomes as shown in Eq. (3) when

ure water is obtained as the permeant:

w

= κ(P

sat w

− P

d )/ δm

V w

(3)

A detailed derivation of Eq. (3) can be found in a published pa-

er by Biswas et al. [33] .

In these equations, variable C is the total concentration

kmol/m

3 ); D wg is the diffusivity of water in glycerol (m

2 /s); E

s the activation energy (kJ/kmol); n is the flux (kmol/m

2 s); P d

s the downstream pressure (N/m

2 ); P sat is the vapor pressure

N/m

2 ); R is the universal gas constant (kJ/kmol K); T is the tem-

erature (K); V is the molar volume (m

3 /kmol); subscripts i and

are the component and water; superscripts d and u are the

ownstream interface and upstream interface, respectively; δm

is

he membrane thickness (m); κ is the permeability of pure water

254 F.J. Novita et al. / Journal of the Taiwan Institute of Chemical Engineers 91 (2018) 251–265

Fig. 2. Flux of water (kg m

−2 s −1 ) vs. temperature ( °C).

w

T

w

b

s

g

t

fi

a

[

c

t

p

c

i

s

m

d

s

3

a

m

t

a

t

v

t

b

s

c

r

(

p

e

w

o

g

(m

3 (stp)m m

−2 s −1 (N m

−2 ) −1 ); and κ0 is the Arrhenius constant

(m

3 (stp)m m

−2 s −1 (N m

−2 ) −1 ).

From Biswas et al.’s study [33] , data on the permeation

rate of pure water as a function of temperature were obtained.

Thus, the calculated values of κ0 and E/R are 0.65227 cm

3 (stp)

cm cm

−2 s −1 (cmHg) −1 and 5322 K, respectively. From the perme-

ability data, the flux of water can be obtained. As the flux of wa-

ter ( n w

= kmol/m

2 .s) is known, the flux of glycerol is calculated by

multiplying the flux of water by the ratio of the molar composition

of glycerol to water in the permeate. As noted in this work, the ra-

tio of the molar composition of glycerol to water in the permeate

was assumed as 0.001/0.999. Based on those information, a per-

vaporation model was developed in Aspen Custom Modeler V10.0.

Later on, the pervaporation model was integrated with Aspen Plus

V10.0. Fig. 2 shows the results of the model validation between the

reference [33] and the present study for the relation between the

flux of water (kg m

−2 s −1 ) and the temperature ( °C).

3. Dehydration processes using conventional EDC configuration

The process flowsheet applied for the dehydration process using

glycerol as an entrainer is referenced from Gil et al. [20] . In their

Fig. 3. Optimal process flowsheet of

ork, they utilized two distillation columns: an EDC and an RC.

he azeotropic feed compositions and glycerol are fed to the EDC,

here the dehydration of the desired component takes place. The

ottom product of the EDC is fed to the RC, where the glycerol is

eparated from water and is recycled to the EDC. Note that a small

lycerol stream is added to balance the small glycerol losses in the

op stream of both columns. The operating pressure in the EDC is

xed at 1 atm, whereas that in the RC is fixed at 0.02 atm, to avoid

thermal degradation of glycerol owing to the high temperature

20] . This process flowsheet will be implemented for the three al-

ohol dehydration processes described in this study. However, the

arget purities are set differently based on the use of each alcohol

roduct. As a similarity among the three alcohol dehydration pro-

esses considered in this study, the feed composition fed to the RC

s set for no alcohol to remain in the stream, and thus the further

eparation task focuses only on both components (glycerol/water

ixture). Details of this process, including the optimization proce-

ure for each dehydration process, are given in the following sub-

ections.

.1. Dehydration EtOH process

The optimal flowsheet of EtOH/water separation with glycerol

s the entrainer can be seen in Fig. 3 . The EtOH/water azeotrope

ixture and the entrainer are fed into the EDC using stages be-

ween the feed stages of the entrainer and azeotropic (AZ) feed

s the extractive section, and the stages below the AZ feed as

he stripping section. The presence of glycerol adjusts the relative

olatility between the EtOH and water, causing the EtOH to move

o the top part of the column, and the water to move down to the

ottom part. EtOH with the desired purity is obtained in the top

tream of the EDC. In the stripping section, EtOH as the lightest

omponent is stripped toward the extractive section of the column

esulting in only negligible EtOH in the bottom stream of the EDC

gly/water feed). This gly/water feed stream is fed into the RC to

roduce nearly pure water in the distillate and nearly pure glyc-

rol at the bottom of the column. Glycerol as a heavy entrainer

ill be recycled back to the EDC. To balance the very small losses

f entrainer in the top stream of both columns, a small stream of

lycerol should be added.

the EtOH dehydration process.

F.J. Novita et al. / Journal of the Taiwan Institute of Chemical Engineers 91 (2018) 251–265 255

i

c

(

o

l

t

t

A

t

(

K

t

v

e

t

t

p

i

t

a

e

t

T

f

s

1

r

d

b

v

R

v

b

i

c

f

f

f

A

t

t

F

2

s

d

o

t

w

T

c

i

N

t

o

p

o

s

a

T

2

Fig. 4. Flowchart of sequential iterative optimization procedure.

3

s

s

e

p

fi

(

b

e

o

0

The product specifications are set as follows: 99.7 mol% of EtOH

n the EtOH stream, and 99.8 mol% of glycerol in the glycerol re-

ycle stream. The AZ feed flow rate is assumed to be 100 kmol/h

89 mol% of EtOH and 11 mol% of water). The operating pressure

f the EDC is 1 atm, and that of the RC is 0.02 atm, owing to the

imitation of glycerol, which can be decomposed at high tempera-

ures.

The design variables to be determined in the flowsheet include

he following: the total number of stages of the EDC (NT1), the

Z feed stage (NF1), the glycerol mixed feed stage (NFE), the to-

al number of stages of the RC (NT2), the gly/water feed stage

NF2), and the glycerol recycle flow rate (E). Based on studies by

night and Doherty [49] and Doherty and Malone [11] , the en-

rainer feed temperature can also be considered as another design

ariable, and thus a cooler is built into the flowsheet. Moreover, Gil

t al. [20] mentioned that low reflux operations need an entrainer

o be fed at temperatures of between 70 °C and 80 °C to maintain

he purity of the distillate, and they chose an entrainer feed tem-

erature of 80 °C. Therefore, the same entrainer feed temperature

s applied in the present study.

A sequential iterative optimization search is applied to find

he optimal value for each of the six design variables mentioned

bove. Note that an entrainer recycle flow rate is taken as an

xternal cycle. Thus, after the glycerol recycle flow rate is changed,

he total stages and feed stage should be adjusted to minimize

AC [50–53] . The two design specifications in the EDC are the

ollowing: setting the top composition at 99.7 mol% EtOH, and

etting the bottom composition at an EtOH mole fraction of

e −5 . These two design specifications can be met by varying the

emaining two degrees of freedom in this column ( e.g. , reboiler

uty and reflux ratio). The design specification of the RC is a

ottom composition of 99.8 mol% glycerol, which can be met by

arying the reboiler duty of the RC. One degree of freedom in the

C is chosen as the bottom flow rate, which is one of the design

ariables that will be optimized.

The total annual cost (TAC) is used as the objective function to

e minimized, which includes the annualized capital and operat-

ng costs. The capital costs include the column shell, trays, reboiler,

ooler, and condenser. The operating costs include the electricity

or operation of the pump, steam, refrigerant, and cooling water

or operation of the reboiler, condenser and cooler. The calculations

or both columns are calculated using an Aspen Process Economic

nalyzer (APEA) [35] provided by Aspen Plus V10.0. Fig. 4 shows

he optimization procedure used to minimize the TAC by varying

he six design variables.

Fig. 5 shows the sensitivity results for the six design variables.

rom this plot, the optimal number of stages of the EDC (NT1) is

0, with the best NF1 at the 12th stage and the best NFE at the

econd stage (stages are counted from top to bottom, with the con-

enser as the first stage and the reboiler as the last stage). The

ptimal number of stages of the RC is 6 with the best NF2 at the

hird stage. The separation capability of the columns will improve

hen the total number of stages of both the EDC and RC increases.

his means that the energy requirement will decrease while the

apital cost increases. The opposite also occurs, which means there

s a trade-off for all stages of both columns. The optimal NT1 and

T2 are the best number of stages for the given specifications in

his EtOH dehydration process. As a result, the total duty of this

ptimal conventional EDC configuration for the EtOH dehydration

rocess is 1668.4 kW with an optimal TAC of $6.56 × 10 6 , and the

ptimal glycerol flow rate is 43 kmol/h. Compared to Gil et al.’s

tudy, the duty of the optimal conventional case of this study is

bit higher owing to the target purities, which are set differently.

he total duty of Gil et al.’s study [20] was 1627.8 kW, which is

.4% lower than the result in this study.

.2. Dehydration IPA process

Based on the flowsheet of the EtOH dehydration process, the

ame flowsheet is applied for the IPA dehydration process. Fig. 6

hows the optimal flowsheet of the IPA/water separation with glyc-

rol as the entrainer. The difference between the IPA dehydration

rocess and the EtOH dehydration process is in the product speci-

cation, which is set as follows: 95 mol% of IPA in the IPA stream

over 96 wt% of IPA [13] ). The AZ feed flow rate is assumed to

e 100 kmol/h (67 mol% of IPA and 33 mol% of water). The glyc-

rol composition in the glycerol recycle stream is set as 99.8 mol%

f glycerol. In addition, the operating pressure is set as 1 and

.02 atm for the EDC and RC, respectively.

256 F.J. Novita et al. / Journal of the Taiwan Institute of Chemical Engineers 91 (2018) 251–265

Fig. 5. Optimization results of the EtOH dehydration process.

Fig. 6. Optimal process flowsheet of the IPA dehydration process.

i

w

c

t

E

s

o

N

b

E

o

4

3

r

There are also six design variables that will be optimized using

a sequential iterative optimization search. These six design vari-

ables are the total number of stages of the EDC (NT1), the AZ feed

stage (NF1), the glycerol mixed feed stage (NFE), the total number

of stages of the RC (NT2), the gly/water feed stage (NF2), and the

glycerol recycle flow rate (E). The minimum TAC is still utilized as

the objective function, which includes the annualized capital and

operating costs. The calculations for both columns are calculated

in a similar manner as the dehydration EtOH process using an As-

pen Process Economic Analyzer (APEA) [35] provided by the Aspen

Plus V10.0. The optimization procedure used to minimize the TAC

by varying the six design variables follows the steps of the EtOH

dehydration process, as described in the previous section.

Fig. 7 shows the sensitivity results for these six design vari-

ables. From this plot, the optimal number of stages of the EDC

(NT1) is 16, with the best NF1 at the tenth stage and the best NFE

at the third stage. The separation capability of the columns will

mprove when the total number of stages of the EDC increases,

hich means that the energy requirement will decrease while the

apital cost will increase. The opposite also occurs, which means

hat there is a trade-off regarding the total number of stages of the

DC. The optimal NT1 is the best number of stages for the given

pecifications in this IPA dehydration process. The optimal number

f stages of the RC is 4, with the best NF2 at the third stage. For

T2 and NF2, at lower than this stage, the desired purity cannot

e achieved. As a result, the total duty of this optimal conventional

DC configuration for the IPA dehydration is 2076.0 kW with an

ptimal TAC of $6.67 × 10 6 , and the optimal glycerol flow rate is

3 kmol/h.

.3. Dehydration TBA process

For the TBA dehydration process, the process flowsheet also

efers to the dehydration EtOH process. The optimal flowsheet for

F.J. Novita et al. / Journal of the Taiwan Institute of Chemical Engineers 91 (2018) 251–265 257

Fig. 7. Optimization results of the IPA dehydration process.

Fig. 8. Optimal process flowsheet for the TBA dehydration process.

t

i

c

9

fl

3

a

s

t

I

t

o

a

I

[

u

f

i

F

5

b

s

t

m

c

o

m

s

d

T

a

he TBA/water separation with glycerol as the entrainer is shown

n Fig. 8 . The product specifications set for this dehydration pro-

ess are as follows: 99.999 mol% of TBA in the TBA stream, and

9.8 mol% of glycerol in the glycerol recycle stream. The AZ feed

ow rate is assumed to be 100 kmol/h (62.59 mol% of TBA and

7.41 mol% of water). The operating pressure is maintained at 1

nd 0.02 atm for the EDC and RC, respectively.

For this dehydration process for a TBA/water system, there are

ix design variables that will be optimized using a sequential itera-

ive optimization search, similar as done for dehydration EtOH and

PA processes. The minimum TAC (including the annualized capi-

al and operating costs) is still utilized as the objective function

f the optimization procedure. The calculations for both columns

re calculated in a similar manner as in the dehydration EtOH and

PA processes using an Aspen Process Economic Analyzer (APEA)

35] provided by Aspen Plus V10.0. The optimization procedure

sed to minimize the TAC when varying the six design variables

ollows the steps used in the dehydration EtOH process described

n the previous section.

Fig. 9 shows the sensitivity results for the six design variables.

rom this plot, the optimal number of stages of the EDC (NT1) is

1, with the best NF1 at the 19th stage, whereas the optimal num-

er of stages of the RC (NT2) is 7, with the best NF2 at the third

tage. The separation capability of the columns will improve when

he total number of stages of both the EDC and RC increases. This

eans that the energy requirement will decrease while the capital

ost increases. The opposite also occurs, which means that a trade-

ff exists for the total number of stages of both columns. The opti-

al NT1 and NT2 are the best total number of stages for the given

pecifications in this TBA dehydration process. As a result, the total

uty of this optimal conventional EDC distillation configuration for

BA dehydration is 1432.4 kW with an optimal TAC of $7.32 × 10 6 ,

nd the optimal glycerol flow rate is 80 kmol/h.

258 F.J. Novita et al. / Journal of the Taiwan Institute of Chemical Engineers 91 (2018) 251–265

Fig. 9. Optimization results of the TBA dehydration process.

Table 4

Detailed information for each dehydration process using conventional EDC.

EtOH case IPA case TBA case

GLY/Water feed (mol%)

Glycerol 80 59 68

Water 20 41 32

Cost ($)

Capital cost $4,582,230 $4,614,030 $5,401,560

Operating cost $1,978,940 $2,057,720 $1,921,830

Total Cost $6,561,170 $6,671,750 $7,323,390

Duty (kW)

EDC 1542.7 1230.9 948.9

RC 125.7 845.1 483.5

Total Duty 1668.4 2076.0 1432.4

e

o

4

c

g

t

e

s

t

m

i

o

t

a

B

w

r

i

t

a

p

b

w

a

a

r

n

p

b

p

g

v

P

o

f

3.4. Comparison of three conventional dehydration processes using

EDC

Table 4 shows the TAC and energy requirements for each dehy-

dration process, as well as information on the glycerol/water com-

position fed to the RC. As Table 4 indicates, the most expensive of

the three processes is the TBA dehydration process, which is pos-

sibly due to the total number of stages of the EDC (NT1) in the

TBA case, which requires many stages to obtain the desired purity.

However, the IPA dehydration process has the highest energy re-

quirement, which can be explained based on two considerations.

First, if the EDC is observed, the highest reboiler duty is found

for the EtOH case. This occurs owing to the smaller AZ water feed

composition in the EtOH case compared to the two other cases. In

the EtOH case, the AZ water feed composition is 11 mol%, whereas

the AZ water feed composition in the IPA case is 33 mol%, and is

37.41 mol% in the TBA case. A smaller AZ water feed composition

entering the EDC means a higher reboiler duty is required to move

the desired alcohol toward the top section of the column. Sec-

ond, if the RC is observed, the highest reboiler duty is obtained in

the IPA case. This occurs owing to the glycerol composition in the

gly/water feed affecting the duty of the RC. The simulation results

show that a smaller glycerol composition in the gly/water feed re-

sults in a higher duty of the RC, the reason for which is more water

ntering the RC, which means a higher duty is required to move all

f the water to the top stream of the RC.

. Dehydration processes using hybrid EDC with PV

onfiguration

As mentioned in the Introduction, a separation process for a

lycerol/water mixture normally requires a low operating pressure

o separate glycerol from water owing to the limitation of glyc-

rol decomposing at high temperatures. In a conventional EDC, a

ingle recovery column is utilized to separate glycerol from wa-

er. The recovery column is operated at 0.02 atm to avoid a ther-

al degradation of glycerol at high temperature. The low operat-

ng pressure increases the TAC of this process compared to normal

perating pressure conditions. For this reason, a novel configura-

ion is proposed in this study to overcome the high cost of sep-

rating a glycerol/water mixture during the dehydration process.

y referring to the conventional EDC configuration, a hybrid EDC

ith a membrane configuration is proposed in this study. Pervapo-

ation (PV) with high selectivity is applied as the membrane unit

n this hybrid configuration. A cellophane membrane is chosen for

his study owing to its high selectivity toward water on the perme-

te side. The PV unit acts as a recovery column in a conventional

rocess for separating glycerol from water.

For the design of the PV unit, it is assumed as a simple mem-

rane model that considers the mass balance in a single block

ithout any discretization, as shown in Fig. 10 . Several additional

ssumptions are made when building this membrane unit, such as

constant total pressure in the three parts of the membrane (feed,

etentate, and permeate), a solution-diffusion transport mecha-

ism, the permeability being unrelated to pressure, the partial

ressure of the component in the feed being the partial pressure

efore the membrane, the partial pressure of the component in the

ermeate being the partial pressure after the membrane, a negli-

ible pressure drop from the feed to the retentate, and a constant

olume. The following sub-sections discuss the hybrid EDC with a

V configuration for the three alcohol dehydration processes. The

ptimal PV unit is explored, and the optimal flowsheet is obtained

or each dehydration process.

F.J. Novita et al. / Journal of the Taiwan Institute of Chemical Engineers 91 (2018) 251–265 259

Fig. 10. Simple membrane model assumed as a single block unit ( n is the flux

(kmol/m

2 s), and x and y are the mole fractions).

4

o

(

t

T

m

s

s

c

t

c

i

a

b

a

w

p

d

i

p

a

t

p

s

p

p

i

c

t

a

o

r

a

A

m

c

t

p

s

p

p

t

m

m

u

T

[

t

c

m

p

t

w

t

.1. Hybrid dehydration EtOH process

Fig. 3 in the previous section shows that the separation process

f the glycerol/water mixture occurs in a single recovery column

RC) operated at 0.02 atm. In this study, a PV unit is proposed

o handle the role of the RC in the conventional configuration.

he proposed hybrid configuration is expected to make the process

ore profitable compared to the conventional configuration, as ob-

erved from the TAC perspective or energy requirement. Fig. 11

hows the optimal flowsheet of the hybrid dehydration EtOH pro-

ess configuration.

As shown in Fig. 11 , a membrane unit, which is a high selec-

ivity PV unit, is implemented during the process. A commercial

ellophane membrane is applied as a membrane type for a PV unit

n this study, as mentioned earlier. The membrane feed temper-

ture is one of the important variables that can allow the mem-

rane to work optimally. Dogan and Hilmioglu [47] reported that

significant weight loss of the cellophane membrane takes place

ithin a temperature range of 30 0 °C to 60 0 °C. This high tem-

erature is also a limitation for glycerol because glycerol can be

Fig. 11. Flowsheet of the hybrid EDC with PV confi

ecomposed at 290 °C. Therefore, the membrane feed temperature

s set at 185.6 °C, allowing the thermal degradation of the cello-

hane membrane and glycerol to be avoided. Note that the perme-

te pressure of the PV unit is set at 0.0067 bar. Thus, the area of

he PV unit required is 45 m

2 to obtain 99.9 mol% of water on the

ermeate side and a glycerol purity of 99.8 mol% on the retentate

ide.

In this hybrid configuration, the optimization process is im-

lemented as well and the objective function of the optimization

rocess is the minimum TAC. The TAC of the hybrid configuration

ncludes the operating cost and the annualized capital cost of the

olumn and membrane. The capital costs of the column include

he column shell, trays, reboiler, cooler, and condenser. The oper-

ting costs of the column include the electricity for the operation

f the pump, the steam and cooling water for operation of the

eboiler, condenser and cooler. The column cost is calculated using

n Aspen Process Economic Analyzer (APEA) [35] provided by

spen Plus V10.0. However, for the membrane unit, the cost is

anually calculated because Aspen Plus does not provide such

alculations for this membrane model. Since the information of

he cost and lifetime of the cellophane membrane have not been

ublished yet in open literature, in this study those are assumed

imilar to the work of Szitkai et al. [54] , wherein their membrane

rice was based on industrial practice for a dehydration ethanol

rocess. In this study, the membrane cost consists of two parts:

he membrane material which has cost per m

2 and the membrane

odules which have cost per module [28] . The membrane replace-

ent cost was taken from Szitkai et al. [54] as 1111,37 US$/m

2 ,

pdated with the Chemical Engineering Plant Cost Index (CEPCI).

he lifetime of the membrane was assumed to be three years

54] . For the vacuum pump, the capital and operating costs refer

o Woods [55] and Oliveira et al. [56] , respectively. Thus, the

apital costs of the membrane include the membrane material,

embrane module, membrane replacement, vacuum pump, feed

ump, and cooler. The operating costs of the membrane include

he electricity for operation of the vacuum pump and the chilled

ater for operation of the cooler.

Fig. 12 shows the flowchart of the optimization procedure for

he hybrid configuration. There are five variables require to be

guration for the dehydration EtOH process.

260 F.J. Novita et al. / Journal of the Taiwan Institute of Chemical Engineers 91 (2018) 251–265

Fig. 12. Flowchart of optimization procedure for hybrid configuration.

o

s

m

r

o

o

f

m

e

t

4

c

F

s

h

o

o

t

E

t

t

s

i

m

$

c

g

4

E

Fig. 13. Optimization results of the dehydration

ptimized. These five design variables are the total number of

tages of the EDC (NT1), the AZ feed stage (NF1), the glycerol

ixed feed stage (NFE), the membrane area (A), and the glycerol

ecycle flow rate (E). Fig. 13 shows the sensitivity result of this

ptimization procedure.

From Fig. 13 , it is found that the optimal recycling flow rate

f glycerol is at 48 kmol/h. As a result, this hybrid configuration

or the dehydration EtOH process has an optimal energy require-

ent and TAC of 1529.2 kW and $5.20 × 10 6 , respectively, which is

quivalent to 8% and 21%, respectively, compared to the conven-

ional configuration of the dehydration EtOH process.

.2. Hybrid dehydration IPA process

Based on the hybrid configuration of the dehydration EtOH pro-

ess, the same flowsheet is applied for the dehydration IPA process.

ig. 14 shows the optimal hybrid configuration for the IPA/water

eparation with glycerol as the entrainer.

The product specifications of the PV unit are set similar to the

ybrid configuration of the dehydration EtOH process: 99.9 mol%

f water on the permeate side, and a glycerol purity of 99.8 mol%

n the retentate side. However, the membrane area required by

he IPA dehydration process is larger than that required by the

tOH dehydration process. The membrane area required to obtain

he desired product specifications is 248.5 m

2 . The membrane feed

emperature in this process is set to 145.4 °C.

The minimum TAC is also analyzed for this process. Fig. 15

hows the optimal results for the hybrid dehydration IPA process.

The calculation of the membrane is the same as that applied

n the EtOH dehydration process. As a result, the energy require-

ent and TAC of this hybrid configuration are 1220.2 kW and

4.99 × 10 6 , which are equivalent to 41% and 25% compared to the

onventional configuration, respectively, where the optimal rate of

lycerol recycling in this hybrid configuration is 78 kmol/h.

.3. Hybrid dehydration TBA process

Based on both hybrid configurations of the dehydration

tOH and IPA processes, the same flowsheet is applied for the

EtOH process in the hybrid configuration.

F.J. Novita et al. / Journal of the Taiwan Institute of Chemical Engineers 91 (2018) 251–265 261

Fig. 14. Flowsheet of the hybrid EDC distillation with PV configuration for the IPA dehydration process.

Fig. 15. Optimization results of the dehydration IPA process in the hybrid configuration.

d

c

e

o

o

h

t

g

T

p

a

w

c

4

w

c

f

a

m

r

b

ehydration TBA process. Fig. 16 represents the optimal hybrid

onfiguration for the TBA/water separation with glycerol as the

ntrainer.

The product specifications of the PV unit are set as 99.9 mol%

f water on the permeate side, and a glycerol purity of 99.8 mol%

n the retentate side. The membrane area required in the TBA de-

ydration process is 221.2 m

2 . The membrane feed temperature in

his process is 152.2 °C.

The minimum TAC is also analyzed for this process. The optimal

lycerol recycle flow rate is 80 kmol/h, as can be seen in Fig. 17 .

he calculation of the membrane is the same as that applied in the

revious hybrid configurations. As a result, the energy requirement

nd TAC of this hybrid configuration are 947.7 kW and $5.75 × 10 6 ,

hich are equivalent to 34% and 21%, respectively, compared to the

onventional configuration.

.4. Comparison of three hybrid dehydration processes using EDC

ith PV

Table 5 shows the TAC and energy requirements for each hybrid

onfiguration of the three dehydration processes, as well as in-

ormation on the membrane feed temperature and the membrane

rea for each dehydration process. As indicated in Table 5 , a higher

embrane feed temperature means a smaller membrane area is

equired. The membrane feed temperature is set similar to the

ottom temperature of the EDC in the conventional configuration,

262 F.J. Novita et al. / Journal of the Taiwan Institute of Chemical Engineers 91 (2018) 251–265

Fig. 16. Flowsheet of the hybrid EDC with PV configuration for the TBA dehydration process.

Fig. 17. Optimization results of the dehydration TBA process in the hybrid configuration.

w

3

p

o

o

t

t

I

c

l

p

which can save the heater/cooler costs required for setting the

membrane feed temperature. The bottom temperature in the

EDC has a correlation with the amount of water in the bottom

stream (gly/water feed stream). When the amount of water in the

gly/water feed stream is low, the bottom temperature is high ow-

ing to the higher energy required to move the alcohol component

to the top part of the EDC, which is the reason why the bottom

temperature of the EDC in the EtOH dehydration process is the

highest compared to the two other dehydration processes. The

composition of water in the gly/water feed stream in the conven-

tional configuration of the EtOH dehydration process is 20 mol%,

hereas in the IPA and TBA dehydration processes it is 41 and

2 mol%, respectively. Owing to this condition, the IPA dehydration

rocess has the highest membrane capital cost but the lowest

perating cost (for column and membrane) compared to the two

ther dehydration processes. The lowest total operating cost in

he IPA dehydration process occurs as a consequence of replacing

he RC with the PV unit. In the conventional configuration of the

PA dehydration process, the RC has the highest reboiler duty

ompared to that of the two other dehydration processes. The

owest operating cost verifies that a higher economic saving can be

rovided when using a hybrid EDC with a PV configuration for a

F.J. Novita et al. / Journal of the Taiwan Institute of Chemical Engineers 91 (2018) 251–265 263

Table 5

Detailed information on each dehydration process using hybrid configuration com-

pared to conventional configuration (shown in Table 4 ).

EtOH case IPA case TBA case

Membrane feed temperature ( °C) 185.6 145.4 152.2

Membrane area (m ²) 45.0 248.5 221.2

Cost ($)

Hybrid configuration

Column

Capital cost $3,280,050 $3,178,430 $3,915,320

Operating cost $1,754,910 $1,359,210 $1,370,060

Membrane

Capital cost $129,304 $345,301 $326,529

Operating cost $40,227 $108,938 $138,213

Total Cost $5,204,491 $4,991,879 $5,750,122

Conventional configuration

Total Cost $6,561,170 $6,671,750 $7,323,390

Saving (%) 21% 25% 21%

Duty (kW)

Hybrid configuration

EDC 1529.2 1220.2 947.7

Total Duty 1529.2 1220.2 947.7

Conventional configuration

Total Duty 1668.4 2076.0 1432.4

Saving (%) 8% 41% 34%

d

a

i

c

c

p

a

a

f

t

5

m

c

d

t

u

d

t

fi

p

d

g

i

p

u

g

m

a

a

o

c

i

w

t

d

h

i

t

a

t

A

a

o

A

g

f

t

s

u

s

d

A

A

m

c

[

t

p

a

f

A

T

A

T

w

A

[

m

A

T

w

[

A

T

w

ehydration process using glycerol as an entrainer. As an additional

dvantage, the thermal degradation of glycerol can also be avoided

n the hybrid configuration. The simulation results prove that de-

reases of up to 25% and 41% in the TAC and energy requirements

an be achieved, respectively, when the hybrid configuration is im-

lemented in a dehydration process using glycerol as an entrainer,

s compared to a conventional configuration. The high potential of

hybrid configuration can be implemented in the actual industry

or alcohol dehydration using glycerol as an entrainer owing to

he availability of a commercial cellophane membrane.

. Conclusions

In this study, a hybrid configuration combining an EDC and a

embrane system was proposed for three alcohol dehydration pro-

esses (EtOH, IPA, and TBA). In the conventional configuration, the

ehydration process utilizes two columns in series for separating

he glycerol/water mixture: an EDC and an RC. The RC is operated

nder low operating pressure conditions owing to the thermal

egradation of glycerol, which must be avoided at high tempera-

ures. A process flowsheet of the EtOH dehydration process was

rst built up, and later implemented in the two other dehydration

rocesses. Glycerol is chosen as an entrainer for the three alcohol

ehydration processes as a substitute for ethylene glycol, because

lycerol is nontoxic, environmentally friendly, highly available, and

nexpensive. In the conventional configuration, the glycerol com-

osition in the gly/water feed affected the duty of the RC. The sim-

lation results showed that a smaller glycerol composition in the

ly/water feed resulted in a higher duty of the RC, because when

ore water entered the RC, additional duty was required to move

ll water to the top stream of the RC. The RC was operated under

low pressure of 0.02 atm owing to the thermal degradation

f glycerol at high temperatures. Consequently, the conventional

onfiguration would have a high TAC owing to this limitation. An

ntensified hybrid configuration utilizing a high selectivity PV unit

as proposed to enhance the TAC and the energy efficiencies of

he conventional EDC distillation configuration during the alcohol

ehydration process. By replacing the RC with a PV unit, the

ybrid EDC with the PV configuration provided a further reduction

n the TAC and energy required for the dehydration process. Thus,

he proposed hybrid EDC with a PV configuration afforded TAC

nd energy savings of up to 25% and 41%, respectively, compared

o the conventional EDC configuration for the dehydration process.

hybrid configuration for alcohol dehydration using glycerol as

n entrainer can be a promising technology in the actual industry

wing to the availability of a commercial cellophane membrane.

cknowledgments

This study was supported by the Basic Science Research Pro-

ram through the National Research Foundation of Korea (NRF)

unded by the Ministry of Education ( 2018R1A2B6001566 ), and by

he Priority Research Centers Program through the National Re-

earch Foundation of Korea (NRF) funded by the Ministry of Ed-

cation ( 2014R1A6A1031189 ). The authors are also grateful for the

upport from the Ministry of Science and Technology of Taiwan un-

er a MOST 104-2221-E-011-149-MY2 grant.

ppendix

. Membrane cost calculation

The data used to calculate the membrane costs (the membrane

aterial and module costs) were taken from Lee et al. [28] . The

ost of the membrane replacement was taken from Szitkai et al.

54] . The membrane lifetime was assumed as three years during

he calculation. The capital and operating costs of the vacuum

ump referenced Woods [55] and Oliveira et al. [56] . In this study,

Chemical Engineering Plant Cost Index (CEPCI) of 567.3 was used

or cost updating.

.1. Cost of membrane material

he membrane material cost [ $ ] is 387 . 5

[m

2 ]

(A1)

.2. Cost of membrane module

he membrane module cost [$] is 125550 ×(

A

324

)0 . 3

(A2)

here A [m

2 ] is membrane area for each module.

.3. Cost of membrane replacement

The membrane replacement cost is taken from Szitkai et al.

54] as 1111.37 US$/m

2 , updated with the CEPCI. In this study, the

embrane life time is assumed as 3 years.

.4. Cost of vacuum pump

he vacuum pump cost [$] is 4200 × CEPCI 13

CEPCI 83

×(

60 × F p × 8 . 314 × 273 . 15

3600 × 101 . 325

)0 . 55

(A3)

here Fp is the total permeation rate through the membrane

kmol/h].

.5. Cost of feed pump

he feed pump cost [$] is 26700 × CEPCI 13

CEPCI 83

×(

24 × F F × 3600

50 0 0 0

)0 . 53

(A4)

here F is the total feed rate to the pump [m

3 /s].

F

264 F.J. Novita et al. / Journal of the Taiwan Institute of Chemical Engineers 91 (2018) 251–265

[

A.6. Power of vacuum pump

The annual cost of power [$] is 8150 × 0 . 04

×{ (

F P 8 . 314 T

3600

)(k r

k r −1

)[ (1 . 013

P op

)( k r / k r −1 )

−1

] }

(A5)

where k r is the heat capacity ratio (1.33), and P op is the pressure

on the permeate side.

A.7. Chilled water cost

In this work, the permeate side is 0.0067 bar. The use of chilled

water is necessary to cool the permeate in the liquid. The price of

chilled water is $4.43/GJ.

References

[1] Figueirêdo MFD , Guedes BP , Araújo JMD , Vasconcelos LGS , Brito RP . Optimaldesign of extractive distillation columns – a systematic procedure using a pro-

cess simulator. Chem Eng Res Des 2011;89:341–6 .

[2] Jaint S , Smith R , Kim JK . Synthesis of heat-integrated distillation sequence sys-tems. J Taiwan Inst Chem Eng 2012;43:525–34 .

[3] Liang S , Cao Y , Liu X , Li X , Zhao Y , Wang Y , Wang Y . Insight into pres-sure-swing distillation from azeotropic phenomenon to dynamic control. Chem

Eng Res Des 2017;117:318–35 . [4] Li R , Ye Q , Suo X , Dai X , Yu H . Heat-integrated pressure-swing distillation pro-

cess for separation of a maximum-boiling azeotrope ethylenediamine/water.Chem Eng Res Des 2016;105:1–15 .

[5] Zhu Z , Xu D , Liu X , Zhang Z , Wang Y . Separation of acetoni-

trile/methanol/benzene ternary azeotrope via triple column pressure-swingdistillation. Sep Purif Technol 2016;169:66–77 .

[6] Wang SJ , Huang K . Design and control of acetic acid dehydration system viaheterogeneous azeotropic distillation using p-xylene as an entrainer. Chem Eng

Process Process Intensif 2012;60:65–76 . [7] Luyben WL . Economic optimum design of the heterogeneous azeotropic dehy-

dration of ethanol. Ind Eng Chem Res 2012;51:427–32 .

[8] Li X , Wang R , Na J , Li H , Gao X . Reversible reaction-assisted intensificationprocess for separating the azeotropic mixture of ethanediol and 1,2-butanediol.

Ind Eng Chem Res 2018;57:710–17 . [9] Kiss AA , Suszwalak DJPC . Enhanced bioethanol dehydration by extractive

and azeotropic distillation in dividing-wall columns. Sep Purif Technol2012;86:70–8 .

[10] Chen YC , Yu BY , Hsu CC , Chien IL . Comparison of heteroazeotropic and extrac-

tive distillation for the dehydration of propylene glycol methyl ether. ChemEng Res Des 2016;111:184–95 .

[11] M.F. Doherty, M.F. Malone, Conceptual design of distillation systems, McGraw-Hill, 2001.

[12] Vorayos N , Kiatsiriroat T , Vorayos N . Performance analysis of solar ethanol dis-tillation. Renew Energy 2006;31:2543–54 .

[13] Zhigang L , Jinchang Z , Biaohua C . Separation of aqueous isopropanol by reac-

tive extractive distillation. J Chem Technol Biotechnol 2002;77:1251–4 . [14] Xu S , Wand H . Separation of tert-butyl alcohol–water mixtures by a

heterogeneous azeotropic batch distillation process. Chem Eng Technol2006;29:113–18 .

[15] Lei Z , Zhang J , Chen B . Separation of aqueous isopropanol by reactive extractivedistillation. Chem Technol Biotechnol 2002;77:1251–4 .

[16] Saiful A , Chien IL . Design and control of an isopropyl alcohol dehydration pro-

cess via extractive distillation using dimethyl sulfoxide as an entrainer. Ind EngChem Res 2008;47:790–803 .

[17] Li H , Zhou P , Zhang J , Li D , Li X , Gao X . A theoretical guide for screening ionicliquid extractants applied in the separation of a binary alcohol-ester azeotrope

through a DFT method. J Mol Liq 2018;251:51–60 . [18] Simasatitkul L , Kaewwisetkul P , Arpornwichanop A . Techno-economic assess-

ment of extractive distillation for tert-butyl alcohol recovery in fuel additive

production,. Chem Eng Process Process Intensif 2017;122:161–71 . [19] Luyben WL , Chien IL . Design and control of distillation systems for separating

azeotropes. Hoboken, New Jersey: John Wiley & Sons, Inc; 2010 . [20] Gil ID , Gómez JM , Rodríguez G . Control of an extractive distillation pro-

cess to dehydrate ethanol using glycerol as entrainer. Comput Chem Eng2012;39:129–42 .

[21] Zhang L , Zhang W , Yang B . Experimental measurement and modeling ofternary vapor −liquid equilibrium for water + 2-propanol + glycerol. J Chem Eng

Data 2014;59:3825–30 .

[22] Aniya V , De D , Singh A , Satyavathi B . Isobaric phase equilibrium of tert-butylalcohol + glycerol at local and subatmospheric pressures, volumetric prop-

erties, and molar refractivity from 303.15 to 333.15 k of tert-butyl alco-hol + glycerol, tert-butyl alcohol + water, and water + glycerol binary systems.

J Chem Eng Data 2016;61:1850–63 .

[23] Gu Y , Je ́ro ̂me F . Bio-based solvents: an emerging generation of fluids for thedesign of eco-efficient processes in catalysis and organic chemistry. Chem Soc

Rev 2013;42:9550–70 . [24] Yang D , Hou M , Ning H , Zhang J , Ma J , Yang G , Han B . Efficient SO 2 absorption

by renewable choline chloride −glycerol deep eutectic solvents. Green Chem2013;15:2261–5 .

[25] Lee F-M , Pahl RH . Solvent screening study and conceptual extractive distilla-tion process to produce anhydrous ethanol from fermentation broth. Ind Eng

Chem Process Des Dev 1985;24:168–72 .

[26] García-Herreros P , Go ́mez JM , Gil ID , Rodríguez G . Optimization of the designand operation of an extractive distillation system for the production of fuel

grade ethanol using glycerol as entrainer. Ind Eng Chem Res 2011;50:3977–85 .[27] Navarrete-Contreras S , Sa ́nchez-Ibarra M , Barroso-Mun ̃oz FO , Herna ́ndez S ,

Castro-Montoya AJ . Use of glycerol as entrainer in the dehydration ofbioethanol using extractive batch distillation: simulation and experimental

studies,. Chem Eng Process Process Intensif 2014;77:38–41 .

[28] Lee HY , Li SY , Chen CL . Evolutional design and control of the equilibrium-lim-ited ethyl acetate process via reactive distillation-pervaporation hybrid config-

uration. Ind Eng Chem Res 2016;55:8802–17 . [29] Aptel P , Cuny J , Jozefowicz J , Morel G , Neel J . Liquid transport through mem-

branes prepared by grafting of polar monomers onto poly(tetrafluoroethylene)films. I. Some fractionations of liquid mixtures by pervaporation. J Appl Polym

Sci 1972;16:1061–76 .

[30] Neel J . Introduction to pervaporation. Pervaporation membrane separation pro-cesses. Huang RYM, editor; 1991 .

[31] Meares P . The sorption and diffusion of vapours in polymers. In: Bakish RA,editor. Proceedings of the Third International Conference on Pervaporation Pro-

cessesin the Chemical Industry, 12-20. Englewood, NJ: Bakish Materials Corpo-ration; 1988 .

[32] Nunes SP , Peinemann K-V . Membrane technology in the Chem industry.

Nunes SP, Peinemann K-V, editors, Weinheim, Germany: Wiley-VCH VerlagGmbH&Co; 2001. Membrane preparation .

[33] Biswas K , Datta S , Chaudhuri S , Kargupta K , Datta S , Sanyal SK . Dehydration ofglycerol-water mixtures using pervaporation: Influence of process parameters.

Sep Sci Technol 20 0 0;35:1391–408 . [34] Malekpour A , Mostajeran B , Kookmareh GA . Pervaporation dehydration of bi-

nary and ternary mixtures of acetone, isopropanol, and water using polyvinyl

alcohol/zeolite membranes. Chem Eng Process Process Intensif 2017;118:47–53 .[35] Aspen Process Economic Analyzer (APEA) from Aspen Plus V10.0 software

(2017). [36] Renon H , Prausnitz J . Journal of Local compositions in thermodynamic excess

functions for liquid mixtures. AIChE J 1968;14:135–44 . [37] Aspen Technology, Aspen physical property system, Version Number V10.0

(2017).

[38] Gmehling J . Azeotropic data. Second ed. Weinheim: Wiley-VCH; 2004 . [39] Brennecke JF , Maginn EJ . Ionic liquids: innovative fluids for chemical process-

ing. AIChE J 2001;47:2384–9 . [40] Marsh KN , Boxall JA , Lichtenthaler R . Room temperature ionic liquids and their

mixtures: a review. Fluid Phase Equilib. 2004;219:93–8 .

[41] Pereiro AB , Arau ́jo JMM , Esperanc ̧ a JMMS , Marrucho IM , Rebelo LPN . Ionic

liquids in separations of azeotropic systems: a review. J Chem Thermodyn2012;46:2–28 .

[42] Li Q , Xing F , Lei Z , Wang B , Chang Q . Isobaric vapor −liquid equilibrium for iso-propanol + water + 1-ethyl-3-methylimidazolium tetrafluoroborate. J Chem Eng

Data 2008;53:275–9 . [43] Li Q , Zhang J , Lei Z , Zhu J , Wang B , Huang X . Isobaric vapor-liquid equi-

librium for (propan-2-ol + water + 1-butyl-3-methylimidazolium tetrafluorobo-

rate). J Chem Eng Data 2009;54:2785–8 . 44] Kim H-D , Hwang I-C , Park S-J . Isothermal vapor-liquid equilibrium data at

T = 333.15 K and excess molar volumes and refractive indices at T = 298.15 Kfor the dimethyl carbonate + methanol and isopropanol + water with ionic liq-

uids. J Chem Eng Data 2010;55:2474–81 . [45] Navarro P , Larriba M , García S , García J , Rodríguez F . Physical prop-

erties of binary and ternary mixtures of 2-propanol, water, and

1-butyl-3-methylimidazolium tetrafluoroborate ionic liquid. J Chem EngData 2012;57:1165–73 .

[46] Orchille ́s AV , Miguel PJ , Gonzalez-Alfaro V , Vercher E ,Martínez-Andreu A . Isobaric vapor-liquid equilibria of

1-propanol + water + trifluoromethanesulfonate-based ionic liquid ternarysystems at 100 kPa. J Chem Eng Data 2011;56:4454–60 .

[47] Dogan H , Hilmioglu ND . Zeolite-filled regenerated cellulose membranes for

pervaporative dehydration of glycerol. Vacuum 2010;84:1123–32 . [48] Eichinger, D., Jurkovic, R., Astegger, H.F., Hinterholzer, P., Weinzierl, K., Zikeli, S.

(1993). Method of producing shaped cellulosic articles. US Patent Office, Pat.No. 5 216 144.

[49] Knight JR , Doherty MF . Optimal design and synthesis of homogeneousazeotropic distillation sequences. Ind Eng Chem Res 1989;28:564–72 .

[50] Zhao Y , Ma K , Bai W , Du D , Zhu Z , Wang Y , Gao J . Energy-saving ther-

mally coupled ternary extractive distillation process by combining withmixed entrainer for separating ternary mixture containing bioethanol. Energy

2018;148:296–308 . [51] Bao Z , Zhang W , Cui X , Xu J . Design, optimization and control of extractive

distillation for the separation of trimethyl borate −methanol. Ind Eng Chem Res2014;53:14802–14 .

F.J. Novita et al. / Journal of the Taiwan Institute of Chemical Engineers 91 (2018) 251–265 265

[

[

[

[

[

52] Wang Y , Liang S , Bu G , Liu W , Zhang Z , Zhu Z . Effect of solvent flow rates oncontrollability of extractive distillation for separating binary azeotropic mix-

ture. Ind Eng Chem Res 2015;54:12908–19 . 53] Zhao Y , Zhao T , Jia H , Li X , Zhu Z , Wang Y . Optimization of the composition

of mixed entrainer for economic extractive distillation process in view of theseparation of tetrahydrofuran/ethanol/water ternary azeotrope. J Chem Technol

Biotechnol 2017;92:2433–44 .

54] Szitkai Z , Lelkes Z , Rev E , Fonyo Z . Optimization of hybrid ethanol dehydrationsystems. Chem Eng Process 2002;41:631–46 .

55] Woods DR . Cost estimation for the process industries. Hamilton, Ontario,Canada: McMaster University; 1983 .

56] Oliveira TA , Cocchini U , Scarpello JT , Livingston AG . Pervaporation mass trans-fer with liquid flow in the transition regime. J Membr Sci 2001;183:119–33 .