Ewa

YAa

b

a

A

R

R

2

A

A

K

W

r

P

P

H

D

P

1

Tibpdtrnt

h0

Process Safety and Environmental Protection 1 0 3 ( 2 0 1 6 ) 413–423

Contents lists available at ScienceDirect

Process Safety and Environmental Protection

journa l h om ep age: www.elsev ier .com/ locate /ps ep

nhanced recovery of PGME and PGMEA fromaste photoresistor thinners by heterogeneous

zeotropic dividing-wall column

us Donald Chaniagoa, Gregorius Rionugroho Harviantoa,lireza Bahadorib, Moonyong Leea,∗

School of Chemical Engineering, College of Engineering, Yeungnam University, Gyeongsan, Republic of KoreaSchool of Environment, Science and Engineering, Southern Cross University, Lismore, Australia

r t i c l e i n f o

rticle history:

eceived 30 November 2015

eceived in revised form 10 May

016

ccepted 16 May 2016

vailable online 21 May 2016

eywords:

aste photoresistor thinner

ecovery

GME

GMEA

eterogeneous azeotrope

ividing-wall column

rocess integration

a b s t r a c t

Propylene glycol monomethyl ether (PGME) and propylene glycol monomethyl ether acetate

(PGMEA) are representative photoresistor thinners used extensively and generated as waste

during the display and semiconductor material manufacturing processes. Although the

waste thinner is normally retrieved by distillation, the azeotropes of these two thinner com-

ponents with water limit the distillation performance. In this paper, an extensive design

study of enhanced distillation processes was carried out to determine a favorable path for

waste thinner recovery. Appropriate thermodynamic models for the design of a waste thin-

ner recovery process were obtained through the regression and validation of experimental

vapor-liquid-liquid equilibrium data. An optimal direct sequence using three conventional

distillation columns with a decanter was introduced as a base design to overcome the distil-

lation boundary by azeotropes. Several advanced distillation configurations were examined

to further improve the energy efficiency of the conventional recovery process. A novel het-

erogeneous azeotropic dividing-wall column was developed based on process intensification

and integration. The proposed enhanced recovery process reduced the energy requirement

for waste thinner recovery significantly by 33.1%. The advanced distillation configuration

can be an attractive option for improving the economic and environmental efficiency of the

commercial waste thinner recovery and recycling processes.

© 2016 Institution of Chemical Engineers. Published by Elsevier B.V. All rights reserved.

quers and paints, chemical intermediates for manufacturing

. Introduction

he display and semiconductor material manufacturingndustry is one of the largest industries in the world and it haseen declared to be one of the most important and growthower of nations. Flat panel displays, such as liquid crystalisplays (LCD), plasma display panels (PDP), and light emit-ing diodes (LED) are widely used in everyday life and haveeplaced cathode ray tubes because they are lighter, thin-er, have low-power consumption, and are less harmful to

he environment (Chaniago et al., 2015). During display and

∗ Corresponding author. Tel.: +82 538102512.E-mail address: mynlee@yu.ac.kr (M. Lee).

ttp://dx.doi.org/10.1016/j.psep.2016.05.012957-5820/© 2016 Institution of Chemical Engineers. Published by Elsev

semiconductor material manufacturing processes, thinnersare used as basic materials for removing the photoresistorfrom the substrate edges or dispensing nozzles. Propyl-ene glycol monomethyl ether (PGME) and propylene glycolmonomethyl ether acetate (PGMEA) are representative pho-toresistor thinner components. A large amount of wastethinners containing PGME and PGMEA is generated whenthe unreacted photoresistor is removed using a photoresistorthinner. Besides, PGME is used in many other applications: lac-

PGMEA, a coalescing agent for water-based paints, coatings

ier B.V. All rights reserved.

414 Process Safety and Environmental Protection 1 0 3 ( 2 0 1 6 ) 413–423

VLE/VLL E data selecti on. Thermod ynamic mod el selection and

VLE/VLLE data regr ession

Design inquiry to find acc eptable conv entional column sequence

Direct Sequ ence

Optimization of convention al column sequ ence by bru te fo rce approac h

Enhance d optimal conv entional column by advance d design (d ividing-wall column and

therma l integrat ion)

Indirect Sequence

Possible? STOPNo

Yes

STOP

Economic analysis fo r all pro cess schema

for automotive architectural and metal finishing, chemicalsfor manufacturing copper-clad lamination products (Chinn,2004a). PGMEA is also used in automotive paints and coatings,architectural coatings, metal-coil coatings, industrial mainte-nance coatings, electronic manufacturing, silk-screen printinginks, and metal finisher (Chinn, 2004b). Nevertheless of theirexpensive cost, the use of PGME and PGMEA has rapidlyincreased in a wide range of products because of their impor-tant advantages such as low systemic toxicity and minorparticle formation. This great value of PGME and PGMEA fromeconomic and environmental considerations highlights theneed of an efficient recovery of waste thinner.

The waste photoresistor thinner is generally reclaimedby distillation. Yoon et al. (2010) investigated feasible recov-ery of PGMEA from acetone + PGMEA and toluene + PGMEAbinary mixtures using batch distillation apparatus. Chang(2012) experimented batch distillation to identify the co-existing compounds in waste thinners and their hinderingmechanism in order for possible improvement of distil-lation application. Han et al. (2014) evaluated scale-upseparation characteristics for PGMEA recovery from awater + PGMEA binary mixture. PGMEA was recovered with90 wt.% purity and 74.6% recovery by using 10-stage batchdistillation apparatus. Wang and Li (2015) investigated amethod for purifying PGMEA from the waste PGMEA liq-uid mixtures. The process consists of two conventionaldistillation columns for recovering PGMEA with a purity of95 wt.% and a recovery of 95%. A coupling tower design wasapplied to reduce steam consumption.

Distillation is one of the most widely-used separation pro-cesses but is also one of the most energy-consuming processesin the chemical industry. The energy-intensive character ofdistillation curtails the economics of the waste thinner recov-ery process. Further, distillation has a limited use when themixtures involved form azeotropes. The co-existing targetconstituents (PGME and PGMEA) in the waste thinners hasazeotropes with water, which also hinders the efficient recov-ery of the target photoresist thinner compounds because aconventional distillation process cannot achieve higher purityand recovery beyond the azeotropic points.

Several alternative methods have been used to satisfy theseparation specifications in azeotropic systems. In the hetero-geneous azeotropic system, the formation of a liquid-liquid(decanting) phase using a proper entrainer is a way to accom-plish more efficient distillation (Wang and Huang, 2012). Anazeotrope can be also eliminated by extractive distillation withan appropriate solvent, such as for the separation of ethanoland water (Ravagnani et al., 2010). Pressure adjustment to shiftthe composition profile of a mixture in pressure swing dis-tillation is another way to recover the target fraction for ahomogeneous azeotropic system in the absence of a solvent(Mahdi et al., 2014). However, only a few mixtures can be influ-enced by pressure modification, limiting this application tospecific mixtures.

Increasing requirement for reducing the energy require-ment and meeting the stringent environmental regulationshas also highlighted the need of efficient separation methodsto handle azeotropic mixtures. Therefore, the application ofadvanced distillation columns, such as a thermally-coupleddistillation (TCD) column, dividing-wall column (DWC), andheat integrated distillation, is progressing. The DWC is a sin-gle shell, fully intensified TCD column that is introduced

to overcome a large number of sequence possibilities andreduce the operation cost. The DWC provides considerable

Fig. 1 – Design approach used in this study.

potential energy savings of up to 30% compared to the con-ventional distillation sequence (Schultz et al., 2002; Parkinson,2007). The DWC can be also applied to azeotropic separa-tion to reduce the energy consumption. Despite the manypublications addressing the separation of azeotropic systemsvia conventional distillation, there are only a few reportson advanced design approaches for azeotropic systems byexploiting the DWC: Azeotropic-DWCs yield good perfor-mance for pyridine-water-toluene separation (Wu et al., 2014),acetic acid dehydration (Le et al., 2015), bio-ethanol dehy-dration (Kiss and Suszwalak, 2012), and ethanol dehydrationprocesses (Sun et al., 2011).

Energy savings in distillation process can also be achievedthrough thermal integration (TI) between two columns ina sequence, called a multi-effect configuration or pressure-staged, where one column is operated at a higher pressurethan the other, and the condenser duty of higher pressurecolumn can provide the required reboiler duty in the lowerpressure column (Seider et al., 2009). However, the TI canbe also implemented for non-pressure-staged in a varietyof systems provided that there are sufficient energy sourcesfor another column and feasible heat transfer from the heatsource to a cold source.

The aim of this study was to develop an advanced dis-tillation configuration for the enhanced recovery of PGMEand PGMEA from waste photoresistor thinners in the dis-play and semiconductor material manufacturing industries.Fig. 1 provides an outline of the design approach used in thisstudy to achieve this task. Binary parameters of thermody-

namic model candidates were first obtained for the wastethinner mixture through stringent regression and validation

Process Safety and Environmental Protection 1 0 3 ( 2 0 1 6 ) 413–423 415

scsth

2

CkTed(aotfmNVfUr

2

AfdacBdttNirSUPete

l

G

�

˛

�

G

m

teps. Based on the thermodynamic model, several processonfigurations were then examined using a rigorous processimulator. A novel heterogeneous azeotropic DWC configura-ion was finally proposed to recover PGME and PGMEA in aighly efficient way.

. Thermodynamic model

hoosing the appropriate physical property parameters is theey to a successful and reliable simulation (Carlson, 1996).his depends on the correct estimation data for vapor-liquidquilibrium (VLE) and vapor-liquid-liquid equilibrium (VLLE)ata for water + PGME + PGMEA system. Chiavone-Filho et al.

1993) examined the VLE behavior of water + PGME mixturet 353.15 K and 363.15 K. Hsieh et al. (2006) studied the VLLEf water + PGMEA binary mixture and water + PGME + PGMEAernary mixture. Tochigi et al. (2007) examined the isobaric VLEor water + PGME, water + PGMEA, and PGME + PGMEA binary

ixtures, and correlated the experimental data using theRTL equation. In this study, those experimental data for theLE and VLLE of the water + PGME + PGMEA system obtained

rom the open literature were correlated using the NRTL andNIQUAC models for process design purpose. The regression

esults were validated qualitatively and quantitatively.

.1. Regression and validation of VLE/VLLE data

commercial process simulator (Aspen Plus V8.6) was usedor rigorous simulations of every distillation process candi-ate. Both NRTL and UNIQUAC activity coefficient modelsnd HOC (Hayden and O’Connell, 1975) second virial coeffi-ient model with the association parameters were evaluated.ecause the binary interaction set was unavailable in theefault Aspen Plus databank, a set of parameters for theernary system was obtained by interpolating the experimen-al data. In this study, the Aspen Plus regression result ofRTL-HOC and UNIQUAC-HOC were tested by binary exper-

mental data to verify that the regressed parameter had beeneproduced properly for binary VLE. All experimental data inection 2 were considered and regressed by NRTL-HOC andNIQUAC-HOC in the Aspen Plus interpolation. The Aspenlus physical parameter regression system provides a pow-rful tool for regressing and correlating the experimental datao obtain the binary coefficient values. The adjustable param-ters of those two models are defined as follows:

NRTL equation:

n �i =∑

jxj�jiGji∑kxkGki

+∑

j

xjGij∑kxkGkj

(�ij −

∑m

xm�mjGmj∑kxkGkj

)(1)

ij = exp (−˛ij�ij) (2)

ij = aij + bij

T(3)

ij = cij + dij(T − 273.15 K) (4)

ii = 0 (5)

ii = 1 (6)

where � i is the activity coefficient of component i; x is theole fraction; �ij( /= �ji) is the interaction parameter; ˛ij is the

non-randomness constant for binary interactions; aij, bij, cij,and dij are the binary interactions between a pair of compo-nent i and j, respectively.

UNIQUAC equation:

ln �i = ln˚i

xi+ z

2qi ln

�i

˚i− qi ln ti − qi

∑j

�j�ij

tj+ li + qi

− ˚i

xi

∑j

xjlj (7)

�i = qixi

qr; qr =

∑k

qkxk (8)

˚i = rixi

rr; rr =

∑k

rkxk (9)

li = z

2(ri − qi) + 1 − ri (10)

ti =∑

k

�k�ki (11)

�ij = exp

(aij + bij

T+ cij ln T + dijT

)(12)

where Ф is the segment fraction; � is the area fraction; r andq are the pure component relative volume and surface areaparameters, respectively; �ij( /= �ji) is the interaction parame-ter; z is the coordination number and equal to 10 for the liquidphase; aij, bij, cij, and dij are the binary interactions between apair of component i and j, respectively.

In the Aspen Plus simulator, the binary parameters, aij, bij,cij, and dij in the NRTL and UNIQUAC models are obtained bythe generalized least-squares method to minimize the follow-ing maximum likelihood objective function (Q):

Q =NDG∑n=1

wn

NP∑i=1

[(Te,i − Tm,i

�T,i

)2

+(

Pe,i − Pm,i

�P,i

)2

+NC−1∑j=1

(xe,i,j − xm,i,j

�x,i,j

)2

+NC−1∑j=1

(ye,i,j − ym,i,j

�y,i,j

)2⎤⎦ (13)

where NDG, NP, and NC are the number of data groups in theregression case, the number of points in data group n, and thenumber of components present in the data group, respectively.wn is the weight of data group n; e is the estimated data and m isthe measured data; i is the data for data point i; j is the fractiondata for component j; and � is the standard deviation of theindicated data. T, P, x, and y are the temperature, pressure,liquid, and vapor mole fractions, respectively.

Britt–Luecke method (Britt and Luecke, 1973) was chosenfor the optimization in Aspen Plus. In this study, owing to thelimited experiment data, only bij was regressed in the UNI-QUAC model, and bij and cij in the NRTL model while the otherbinary parameters were simply set to zero, as proposed by

many researchers (Harvianto et al., 2016; Hsieh et al., 2006;Resk et al., 2014).

416 Process Safety and Environmental Protection 1 0 3 ( 2 0 1 6 ) 413–423

Table 1 – Binary interaction parameters obtained from regression.

System NRTL-HOC UNIQUAC-HOC

bij bji cij AAD bij bji AAD

PGMEA + Water 452.17 1069.18 0.43 0.0391 −494.3 59.1223 0.0185Water + PGME 584.86 9.73 0.47 0.0037 241.841 −492.419 0.0027PGME + PGMEA 30.59 72.34 0.3 0.0051 66.5676 −103.641 0.0053

Fig. 2 – Water (1) + PGME (2) VLE phase diagram comparisonwith literature (Exp.) values: (a) at 363 K, (b) at 53.3 kPa. Fig. 3 – PGME (1) + PGMEA (2) VLE phase diagram

comparison with literature (Exp.) values: (a) at 393.15 K, (b)at 53.3 kPa.

diagram.

2.2. Validation of regression result

The NRTL-HOC and UNIQUAC-HOC regression results werevalidated using experimental data to ensure that the regressedbinary parameters were reliable for the design purpose. Table 1lists all the regressed outcomes of the thermodynamic param-eter of NRTL-HOC and UNIQUAC-HOC. The average absolutedeviation (AAD) values defined as follows were calculated forevery binary pair to validate the regression result quantita-tively, as listed in Table 1.

AAD =n∑

i=1

((xcal

i− x

expi

)

n

)(14)

where n is the number of data; xcali

and xexpi

refer to thecalculated and experimental mole fraction for component i,respectively.

2.2.1. Water and PGMEThe regression result was validated from the literature data(Chiavone-Filho et al., 1993; Tochigi et al., 2007) at variouspressures and temperatures. Fig. 2 shows the VLE phase dia-grams of Water + PGME at 363.15 K and at 53.3 kPa.

As shown in the figures, both UNIQUAC-HOC andNRTL-HOC matched the experimental data closely. Further,UNIQUAC-HOC had a slightly better prediction than NRTL-HOC overall.

2.2.2. PGME and PGMEAThis binary VLE was validated using literature data (Hsiehet al., 2006; Tochigi et al., 2007) for the Txy and Pxy equilibriumphase diagrams, as shown in Fig. 3.

No significant difference was observed between NRTL-HOCand UNIQUAC-HOC for the Pxy and Txy equilibrium phase

Process Safety and Environmental Protection 1 0 3 ( 2 0 1 6 ) 413–423 417

Fig. 4 – Water (1) + PGMEA (2) VLE phase diagramcomparison with literature (Exp.) values at 93.3 kPa.

VLLE pha se diagram at 358 .11 K

Water

0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0

PGME

0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1.0

PGM

EA

0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1.0

Organic pha se (UNIQUAC-HOC)Aqu eou s phase (UN IQUAC-HOC)Vapor phase (UNIQUAC-HOC)Organic pha se (NRTL- HOC)Aqu eou s phase (NR TL- HOC)Vap or pha se (NR TL- HOC)Organic phase (Exp. Hsieh et al.)Aqu eou s phase (Exp. Hsieh et al.)Vap or pha se (Exp. Hsieh et al.)

Fig. 5 – Water (1) + PGMEA (2) + PGME (3) VLLE phasediagram comparison with literature (Exp.) values in molefraction at 358.11 K.

2Vtsta

2Tidpvtpdpa

position point until point 3. The top separation is limited bythe distillation boundary or line B at point 2 which indicates

.2.3. Water and PGMEAalidation was only done by an experimental liquid phase dueo the data availability in the literature (Tochigi et al., 2007), ashown in Fig. 4 at pressure 93.3 kPa. From the Tx phase diagramrend, UNIQUAC-HOC showed a closer result than NRTL-HOCt 1 atm or 101.325 kPa.

.2.4. Ternary VLLE phase diagramernary mixtures of water + PGME + PGMEA can have an

mmiscible liquid-liquid phase. In this case, a good VLE pre-iction for each binary case does not guarantee a good VLLErediction of the ternary case. A vapor-liquid-liquid phase isalid with this ternary mixture. An inaccurate VLLE predic-ion will result in an unreliable design. The predicted VLLEhase behavior was also validated with the ternary VLLE phaseiagram from the experimental data. Water and PGMEA areartially miscible in ternary mixtures containing water, PGME,nd PGMEA, as shown in Fig. 5.

As can be seen from the AAD values in Table 1, UNIQUAC-HOC more closely matches the experimental result comparedto the NRTL-HOC result and the deviation is very small.

2.2.5. Azeotropic PointWater + PGME and water + PGMEA have two azeotropic points.Table 2 compares the simulator prediction result of theazeotropic points with the literature data. Both two modelsmatched the experimental data well. In particular, NRTL-HOC is more accurate in predicting the water compositionat the azeotropic point with PGMEA at 53.3 kPa, whereasUNIQUAC-HOC is more accurate in predicting the azeotropictemperature, as can be seen from Table 2.

As observed from the figures and Tables 1 and 2, UNIQUAC-HOC provided a more accurate prediction than NRTL-HOCfor most regression results. Therefore, UNIQUAC-HOC wasfinally chosen for the thermodynamic model for simulationand design study.

3. Conventional distillation sequence

Direct and indirect conventional distillation sequences werefirst investigated for a base design to identify a favorable sepa-ration path. A waste thinner feed of 1000 kg/hr was consideredfor process design. Based on the composition of the wastethinner samples from a real industry, the chosen feed compo-sition was 31 wt.% water, 23 wt.% PGME, and 46 wt.% PGMEA.Water, PGME, and PGMEA are the most volatile, middle, andleast volatile components, respectively. The target purity ofeach PGMEA and PGME for sale purposes were set more than99.95 wt.%. The required recovery of PGMEA and PGME wereassumed to be more than 99.95 wt.% for avoiding the loss ofvaluable thinner components. For distillation column simu-lation in Aspen Plus, convergence tolerance error of 1 × 10−6

was adopted to secure sufficient preciseness.

3.1. Direct sequence

In a direct sequence, the most volatile component is removedas the top product from the first column and the remainingbinary mixture is then separated in the next second col-umn. Therefore, for the water + PGME + PGMEA mixture, wateris removed as the top product in the first column and thePGME + PGMEA mixture as the bottom product is separatedin the next column. Fig. 6 shows the feasible separation inthe first column of the direct sequence. The ternary mix-ture exhibits a distillation boundary, or separatrix, by theazeotropes (line B). This distillation boundary divides theternary diagram into two distillation regions, i.e., Regions I andII, and the separation is limited by this line. To overcome thisbarrier, the composition in Region I (R-I) should be moved tothe distillation boundary line within the liquid-liquid enve-lope so that the additional liquid-liquid separation moves thecomposition to Region II (R-II) beyond the distillation bound-ary.

In the direct sequence, the first column recovers most of thePGME and PGMEA from the bottom and removes water fromthe top completely. PGME and PGMEA from the bottom arethen separated to the required purity in the next column. Thewaste thinner feed is introduced at point 1 and goes throughby a mass balance line (Line A) from the maximum water com-

418 Process Safety and Environmental Protection 1 0 3 ( 2 0 1 6 ) 413–423

Table 2 – Azeotropic data of water + PGME and water + PGMEA at 53.3 kPa.

Component (1) Component (2) Method T (K) xaz1 (mole)

Water PGME LiteratureTochigi et al. 354.03 0.8290NRTL-HOC 353.98 0.8143UNIQUAC-HOC 353.79 0.8369

Water PGMEA LiteratureTochigi et al. 352.47 0.8880NRTL-HOC 353.74 0.8911UNIQUAC-HOC 352.91 0.8632

Fig. 6 – Ternary map for water + PGME + PGMEA in mass fraction of direct sequence at 1 atm.

Fig. 7 – Separation using a single heterogeneous azeotropiccolumn.

the maximum feasible water composition on the top product.Because point 2 is in the liquid-liquid region, the top productcan be separated further to be points 4 and 5 by a liquid-liquidseparator. The main target of the first column is to removemost of water in the feed onto the top product. The top vaporproduct mixture containing most of water discharges into thecondenser and then separated in the decanter to be water-rich point 4 and organic-rich point 5. The bottom product, inwhich most of the water is stripped, is rich in the thinner com-ponents. Point 2 indicates that some PGME and PGMEA stillremain in the top product, which suggests that an additionalcolumn is required to recover the remaining PGME and PGMEAfrom water.

Fig. 7 shows the maximum separation when a single het-erogeneous azeotropic column is used to remove water. Asshown in the figure, the organic solvent and water are not sep-arated well due to the distillation boundary. Approximately101.95 kg/hr of PGMEA and 33.42 kg/hr of PGME are lost inthe lean thinner stream. An additional column is essentialfor recovering the valuable organic components from the topproduct.

Fig. 8 presents the proposed direct sequence using threeconventional distillation columns. Note that the top productof the additional column (C-1b) is recycled to the azeotropiccolumn (C-1) for the maximum recovery of PGME and PGMEA.C-1b removes water on the bottom product in a high purityof 99.99 wt.%. As shown in Fig. 6, trajectory arrow toward to

water, which means water is the heaviest component in R-II. C-1b strips PGME and PGMEA to the top product. The top

stream is then recycled to C-1, which makes each PGME andPGMEA recovery in the C-1 bottom increase up to 99.99%. The

last column (C-2) is to separate the PGME and PGMEA with therequired purity.

Process Safety and Environmental Protection 1 0 3 ( 2 0 1 6 ) 413–423 419

Fig. 8 – Proposed direct sequence using three conventional columns (without and with thermal integration): red-dotted linei

susnfdeft

3

Irtcttptpcct

4

Cerr

s energy transfer from a heat source.

For optimization of the column structure, the number oftages was first varied for a wide range while keeping the prod-ct purity and recovery higher than 99.95%. The number oftages was chosen and fixed as the minimum one that giveso further apparent decrease in reboiler duty. The optimal

eed location was then found to give the minimum reboileruty. This sequential optimization procedure was applied toach column. The proposed direct sequence was close to per-ect separation for PGME and PGMEA recovery, and showed theotal reboiler duty of 1539 kW.

.2. Indirect sequence

n an indirect sequence, the least volatile component isemoved first as the bottom product from the first column andhe remaining binary mixture is then separated in the nextolumn. In contrast to the direct sequence, which still methe feasible separation to overcome the distillation boundary,he indirect sequence does not reach the feasible separationoints in Region II, as shown conceptually in Fig. 9. The dis-illation boundary prevents the composition shift to R-II atoint 2; hence, the entire water removal is not feasible. As aonsequence, the indirect sequence was excluded for furtheronsideration and the direct sequence was finally chosen ashe base case for another improvement.

. Thermally coupled distillation column

onventional distillation sequences can have a remixingffect, which introduces inefficiency in the separation. This

emixing problem can be avoided or reduced by utilizing mate-ial flows to provide some of the necessary heat transfer by

direct contact, which is known as a TCD configuration. Fig. 10shows a conventional direct sequence and several equivalentTCD configurations originating from the direct sequence. In athermally coupled direct sequence (Fig. 10b), the reboiler of thefirst column in the conventional direct sequence is replacedby a thermal coupling. A side rectifier arrangement (Fig. 10c)is one of the most popular TCD configurations and topologi-cally equivalent to the thermally coupled direct sequence. Thetwo TCD configurations were further intensified to be a topdividing–wall column (TDWC) (Fig. 10d) by integrating the twoconventional columns into a single shell with a vertical wallon the top section.

4.1. Enhanced recovery by HA-DWC configuration

C-1 and C-2 in the direct sequence (Fig. 8) can be interlinkedor coupled thermally by carrying two interconnecting streams(one in the liquid phase and the other in the vapor phase)between the two columns. Fig. 11 shows a thermally-coupleddirect sequence of C-1 and C-2. In this configuration, the liquid(L-1) from C-1 bottom is fed to C-2, but the reboiler of C-1 isremoved and the vapor (V-1) from C-2 is utilized as a sourceof the vapor stream of C-1. The reboiler of C-2 takes all theresponsibility for supplying energy to C-1 and C-2. The leanorganic of the top C-1 product is proceeded to C-1b to recoverthe lost organic in the lean organic stream.

In this study, the thermally coupled direct sequence inFig. 11 was intensified further into the equivalent TDWCconfiguration. Fig. 12 presents the resulting overall processflow sheet using the heterogeneous azeotropic dividing-wall

column (HA-DWC). The total number of stages in the HA-DWCwas simply set to be the same as C-1 in the direct sequence.

420 Process Safety and Environmental Protection 1 0 3 ( 2 0 1 6 ) 413–423

Fig. 9 – Ternary map for water + PGME + PGMEA in

(a) (b )

(c) (d )

A B C

A

B C

B

C

A B C

BA

C

A B C

A

B

C

A B C

B

C

A

Fig. 10 – Conventional direct sequence and its equivalentthermally coupled configurations: (a) conventional directsequence; (b) thermally coupled direct sequence; (c) siderectifier arrangement; (d) TDWC configuration.

equipment (capital cost), was used to evaluate the economic

The number of stages in the dividing-wall section was alsodetermined to be the same as that of the rectifying sectionin C-2 for topological consistency. Note that each side ofthe dividing-wall section usually has the same number ofstages to accommodate the practical implementation of thetwo columns into a single shell. The feed location and vaporsplit in the HA-DWC were then adjusted to have the mini-mum reboiler duty. The top and bottom product contained99.98 wt.% of PGME and 99.99 wt.% of PGMEA, respectively.The reboiler duty of HA-DWC was 800 kW, which is much

lower than the sum of reboiler duty C-1 and C-2 (1332 kW)

mass fraction of indirect sequence at 1 atm.

in the direct sequence. Overall, the total reboiler duty of theentire recovery process was reduced from 1539 kW to 1169 kW,which is equivalent to a 24.0% energy saving compared tothe optimal direct sequence. Note that the energy saving wasbased on the conservative conditions. Additional energy sav-ings will be possible if rigorous optimization of the HA-DWCstructure is carried out.

5. Combined configuration: HA-DWC withthermal integration

In the proposed direct sequence, a huge amount of energyis removed in the condenser of C-1. However, heat transferfrom the C-1 top is not feasible because the top temperatureof C-1 is lower than the bottom C-1b or C-2. A higher temper-ature heat source from the top vapor of C-2 can be used asan energy source to reduce the energy required in reboilingthe bottom stream of C-1b. The energy consumption of theC-1b reboiler can be reduced to 41 kW through this thermalintegration, as depicted in Fig. 8. Therefore, the total reboilerduty was reduced to 1373 kW, which is equivalent to a 10.8%energy saving compared to the base design without thermalintegration. The minimum approach temperature of the heatexchanger was assumed to be 10 ◦C.

This thermal integration could also be implemented forthe HA-DWC configuration. In the HA-DWC configuration, thelatent heat that is rejected in the condenser of HA-DWC istransferred to cover approximately 140 kW in the reboiler ofC-1b, as shown in Fig. 12. Finally, the total reboiler duty wasreduced significantly to 1029 kW, which is equivalent to 33.1%energy savings compared to optimal base design.

6. Economic evaluation

The total annual cost (TAC), which involves the cost of thetotal energy requirements (operating cost) and purchased

performance of the processes considered. The capital cost

Process Safety and Environmental Protection 1 0 3 ( 2 0 1 6 ) 413–423 421

sequ

wae1td

Fht

Fig. 11 – Thermally coupled direct

as estimated mainly considering the distillation columnsnd heat exchangers using Guthrie’s modular method (Bieglert al., 1997) and annualized based on a plant lifetime of0 years. The column diameter was designed at 80% ofhe load at the flooding point. The capital cost for the

istillation column consists of the cost of the column,

ig. 12 – Heterogeneous azeotropic DWC (without and with thermeat source. (For interpretation of the references to color in this fi

his article.)

ence configuration of C-1 and C-2.

internals, reboiler, condenser, and decanter. The data for thecapital and utility cost calculation were taken from the lit-erature (Turton et al., 2012). The Chemical Engineering PlantCost Index of 576.1 in 2014 was used for cost updating. Lowpressure steam (160 ◦C, 5.2 barg) is commonly used for all

reboilers. The appendix outlines the detailed method used

al integration). Red-dotted line is energy transfer from agure legend, the reader is referred to the web version of

422 Process Safety and Environmental Protection 1 0 3 ( 2 0 1 6 ) 413–423

Table 3 – Energy and economic performances of process configurations studied.

Conventional (base case) Conventional (with TI) HA-DWC HA-DWC (with TI)

PGME (kg/hr) 229.96 229.96 229.98 229.98PGME (wt%) 99.98 99.98 99.98 99.98PGMEA (kg/hr) 459.96 459.96 459.97 459.97PGMEA (wt%) 99.99 99.99 99.99 99.99Total reb. duty (kW) 1539 1373 1169 1029Energy saving (%)* – 10.8 24.0 33.1Capital cost (million US$) 1.204 1.204 0.972 0.972Op. cost (million US$/yr) 1.399 1.253 1.128 1.005TAC (million US$/yr) 1.579 1.433 1.273 1.149TAC saving (%)a – 9.3 19.4 27.2

a Compared to the conventional direct sequence (base case).

Table A.1 – Utility cost data (Turton et al., 2012).

Utility Price ($/GJ)

Cooling water 0.35Steam (low pressure) 13.28

in the cost calculation. Table 3 lists the results of the energyand economic performance. As seen in the table, the proposedHA-DWC configurations without/with thermal integration canreduce the TAC significantly by up to 19.4% and 27.2%, respec-tively.

7. Conclusions

To construct the appropriate thermodynamic models for thedesign of the waste thinner recovery process, the binaryparameters of the UNIQUAC-HOC and NRTL-HOC models forthe water + PGME + PGMEA mixture were obtained after strin-gent regression and validation. The UNIQUAC-HOC modelmatched the experimental VLLE data, which involves water,PGME, and PGMEA, and was finally chosen. A direct sequenceusing three conventional distillation columns without anentrainer was proposed as a favorable base design. The pro-posed conventional direct sequence eliminated successfullythe azeotrope difficulty and recovered the PGME and PGMEA inhigh purity. An indirect sequence was also examined but failedto recover PGME and PGMEA in high purity and secure highwater removal. A novel HA-DWC configuration was proposedto enhance the energy efficiency of the recovery process. Thisconfiguration resulted in large savings of 24.0% in the reboilerduty and of 19.4% in the TAC compared to the conventionaldirect sequence. Furthermore, applying thermal integration tothe HA-DWC reduced remarkably the total reboiler duty andTAC by 33.1% and 27.2%, respectively. The proposed recoveryprocess using the thermally integrated HA-DWC is expected toprovide an attractive option for enhancing the waste thinnerrecovery efficiency in the display and semi-conductor materialmanufacturing industries.

Acknowledgments

This study was supported by Basic Science Research Programthrough the National Research Foundation of Korea (NRF)funded by the Ministry of Education (2015R1D1A3A01015621).This work was also supported by Priority Research Cen-ters Program through the National Research Foundationof Korea (NRF) funded by the Ministry of Education(2014R1A6A1031189).

Appendix. Column cost correlations

A.1. Capital cost (CC)

Guthrie’s modular method was applied (Biegler et al., 1997).Chemical Engineering Plant Cost Index of 576.1 (2014) was

used for cost updating.

CC = BMC column + BMC wall + BMC tray stack

+ BMC condenser + BMC reboiler (A.1)

Updated bare module cost (BMC) = UF × BC × (MPF + MF − 1)

(A.2)

where UF is the update factor : UF = present cost indexbare cost index

(A.3)

BC is the bare cost of the heat exchanger : BC = BC0 x

(A

A0

)˛

(A.4)

where MPF is the material and pressure factor; MF is the mod-ule factor (a typical value), which is affected by the base cost.

Area of the heat exchanger, A = Q

U�T(A.5)

Material and pressure factor : MPF = Fm + Fp + Fd (A.6)

A.2. Operating cost (OC)

Table A.1

OC = Csteam + CCW (A.7)

where Csteam is the cost of the steam; CCW is the cost of coolingwater.

A.3. Total annual cost (TAC)

The total annual cost includes the annual capital cost (ACC)and the annual operating cost (AOC). The annual investment

Process Safety and Environmental Protection 1 0 3 ( 2 0 1 6 ) 413–423 423

cp

A

wy

T

R

B

B

C

C

C

C

C

C

H

H

H

H

H

ost was obtained from the literature and refers to the annualayments over the life of the project (Hicks and Chopey, 2012).

CC = CC

[i(1 + i)n

(1 + i)n − 1

](A.8)

here, i is the interest rate per year (8%), and n is the projectear (10 years).

AC = ACC + AOC (A. 9)

eferences

iegler, L.T., Grossmann, I.E., Westerberg, A.W., 1997. SystematicMethods of Chemical Process Design. Prentice Hall Inc., NewJersey.

ritt, H.I., Luecke, R.H., 1973. The estimation of parameters innonlinear, implicit models. Technometrics 15, 233–247.

arlson, E.C., 1996. Don’t gamble with physical properties forsimulations. Chem. Eng. Prog. 92, 35–46.

hang, J.F., Master Thesis, Environmental Engineering 2012.Affect the Recovery of Waste Photoresist Thinner Factor ofDistillation. National Central University, Taiwan.

haniago, Y.D., Minh, L.Q., Khan, M.S., Koo, K.K., Bahadori, A.,Lee, M.Y., 2015. Optimal design of advanced distillationconfiguration for enhanced energy efficiency of waste solventrecovery process in semiconductor industry. Energy Convers.Manag. 102, 92–103.

hiavone-Filho, O., Proust, P., Rasmussen, P., 1993. Vapor-liquidequilibria for glycol ether + water systems. J. Chem. Eng. Data38, 128–131.

hinn, H., 2004a. Glycol Ethers, Marketing Research Report:Chemical Economics Handbook. SRI Consulting, pp. 50.

hinn, H., 2004b. Glycol Ethers, Marketing Research Report:Chemical Economics Handbook. SRI Consulting, pp. 52–53.

an, S.K., Jung, H.S., Ahn, D.H., Kim, J.Y., 2014. A study on theseparation characteristics of multistage distillation tower forpurification of waste propylene glycol monomethyl etheracetate (PGMEA). J. Korea Soc. Waste Manag. 31, 406–412.

arvianto, G.R., Kim, S.E., Kang, K.J., Bahadori, A., Lee, M.Y., 2016.Liquid–liquid equilibrium data and correlation for quaternarysystems of acetic acid + water + methyl acetate + p-xylene at313.2 K. J. Ind. Eng. Chem. 35, 369–375.

ayden, J.G., O’Connell, J.P., 1975. A generalized method forpredicting second virial coefficients. Ind. Eng. Chem. ProcessDes. Dev. 14, 209–216.

icks, T., Chopey, N., 2012. Handbook of Chemical EngineeringCalculations, 4th ed. McGraw Hill Professional.

sieh, C.T., Lee, M.J., Lin, H.M., 2006. Multiphase equilibria formixtures containing acetic acid, water, propylene glycol

monomethyl ether, and propylene glycol methyl etheracetate. Ind. Eng. Chem. Res. 45, 2123–2130.

Kiss, A.A., Suszwalak, D.J.P.C., 2012. Enhanced bioethanoldehydration by extractive and azeotropic distillation individing-wall columns. Sep. Purif. Technol. 86, 70–78.

Le, Q.K., Halvorsen, I.J., Pajalic, O., Skogestad, S., 2015. Dividingwall columns for heterogeneous azeotropic distillation.Chem. Eng. Res. Des. 99, 111–119.

Mahdi, T., Ahmad, A., Nasef, M.M., Ripin, A., 2014.State-of-the-art technologies for separation of azeotropicmixtures. Sep. Purif. Rev. 44, 308–330.

Parkinson, G., 2007. Dividing-wall columns find greater appeal.Chem. Eng. Prog. 103, 8–11.

Ravagnani, M.A.S.S., Reis, M.H.M., Filho, R.M., Wolf-Maciel, M.R.,2010. Anhydrous ethanol production by extractive distillation:a solvent case study. Process Saf. Environ. Prot. 88, 67–73.

Resk, A.J., Peereboom, L., Kolah, A.K., Miller, D.J., Lira, C.T., 2014.Phase Equilibria in systems with levulinic acid and ethyllevulinate. J. Chem. Eng. Data 59, 1062–1068.

Schultz, M.A., Stewart, D.G., Harris, J.M., Rosenblum, S.P., Shakur,M.S., O’Brien, D.E., 2002. Reduce costs with dividing-wallcolumns. Chem. Eng. Prog. 98, 64–71.

Seider, W.D., Seader, J.D., Lewin, D.R., 2009. Product & ProcessDesign Principles: Synthesis, Analysis and Evaluation, (withCD). John Wiley & Sons.

Sun, L.Y., Chang, X.W., Qi, C.X., Li, Q.S., 2011. Implementation ofethanol dehydration using dividing-wall heterogeneousazeotropic distillation column. Sep. Sci. Technol. 46,1365–1375.

Tochigi, K., Takahara, H., Shiga, Y., Kawase, Y., 2007. Isobaricvapor–liquid equilibria for water + propylene glycolmonomethyl ether (PGME), water + propyleneglycolmonomethyl ether acetate (PGMEA), and PGME + PGMEA atreduced pressures. Fluid Phase Equilib. 260, 65–69.

Turton, R., Baile, C.R., Whiting, W.B., Shaeiwitz, J.A.,Bhattacharyya, D., 2012. Analysis, Synthesis and Design ofChemical Processes, 4th ed. Prentice Hall, Upper Saddle River,NJ.

Wang, S.J., Huang, K., 2012. Design and control of acetic aciddehydration system via heterogeneous azeotropic distillationusing p-xylene as an entrainer. Chem. Eng. Process.: ProcessIntensif. 60, 65–76.

Wang, Z.J., Li, H.L. Method for purifying PGMEA from PGMEAwaste liquid. State Intellectual Property Office of The P.R.C.Patent Number CN 104370742A; 2015.

Wu, Y.C., Lee, H.Y., Huang, H.P., Chien, I.L., 2014. Energy-savingdividing-wall column design and control for heterogeneousazeotropic distillation systems. Ind. Eng. Chem. Res. 53,1537–1552.

Yoon, K.J., Lee, W.M., Kwon, O.H., Kim, J.Y., 2010. Purification ofwaste organic solvent containing propylene glycol

monomethyl ether acetate (PGMEA). Appl. Chem. Eng. 6,616–620.