Chemical Engineering Science 190 (2018) 297–311

Contents lists available at ScienceDirect

Chemical Engineering Science

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

Reactive distillation with pervaporation hybrid configuration forenhanced ethyl levulinate production

https://doi.org/10.1016/j.ces.2018.06.0240009-2509/� 2018 Published by Elsevier Ltd.

Abbreviations: LAEE, ethyl levulinate; LA, levulinic acid; PV, pervaporation; RD, reactive distillation; RDLA, reactive distillation using an excess LA reactant;reactive distillation using an excess EtOH reactant; TAC, total annual cost; TCD, thermally-coupled distillation; TCRD, thermally-coupled reactive distillation;thermally-coupled reactive distillation with a side-stripper; TCRDSS-PV, thermally-coupled reactive distillation�pervaporation.⇑ Corresponding authors.

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

Felicia Januarlia Novita a, Hao-Yeh Lee b,⇑, Moonyong Lee a,⇑a School of Chemical Engineering, Yeungnam University, Dae-dong 712-749, Republic of KoreabDepartment of Chemical Engineering, National Taiwan University of Science and Technology, Taipei 10607, Taiwan

h i g h l i g h t s

� A novel RD combined with PV processwas proposed for enhanced LAEEproduction.

� The excess EtOH reactant showed asignificant advantage over the excessLA case.

� The proposed hybrid configurationreduced the energy and TACremarkably.

� Thermal coupling enhanced theefficiency of the hybrid processsuccessfully.

� Series PV arrangement was suitablefor LAEE production process.

g r a p h i c a l a b s t r a c t

a r t i c l e i n f o

Article history:Received 10 May 2017Received in revised form 7 June 2018Accepted 9 June 2018Available online 18 June 2018

Keywords:Ethyl levulinate productionReactive distillation-membrane hybridprocessThermally coupled reactive distillationPervaporation membrane arrangementExcess reactantEnergy efficiency

a b s t r a c t

A hybrid process design based on reactive distillation (RD) combined with pervaporation membrane sys-tems was proposed for the production of ethyl levulinate (LAEE) via esterification. In the proposed hybridconfiguration, LAEE is produced from the bottom stream of the RD column while the ethanol/water mix-ture from the overhead is fed to an additional column. The azeotropic mixture in the overhead of theadditional column is separated in the pervaporation unit, and the ethanol-rich stream is recycled tothe RD column. An RD-only configuration with excess ethanol reactant was also examined to investigatethe benefit of excess ethanol reactant and the azeotrope mixture separation. The proposed RD with per-vaporation hybrid configuration provided 61.0% and 71.0% reductions of the total annual cost and energy,respectively, compared to the conventional RD configuration with excess levulinic acid (LA) reactant and6.0% and 13.0%, respectively, as compared to the conventional RD configuration with excess ethanol reac-tant. The results demonstrated the advantage of the proposed hybrid RD process using pervaporation forlowering the cost and improving the energy efficiency of LAEE production. Different pervaporationarrangements (such as series and parallel arrangements) were discussed in connection with the two mainvariables (operating temperature and concentration gradient) affecting the membrane flux. The seriesarrangement was more suitable for the LAEE hybrid process because the influence of the operating tem-perature is dominant in this process. Thermal intensification in the RD-pervaporation was also examinedfor exploring further energy reduction. This thermally coupled RD-pervaporation hybrid configuration

RDEtOH,TCRDSS,

298 F.J. Novita et al. / Chemical Engineering Science 190 (2018) 297–311

provided further reductions in total annual cost and energy by 63.0% and 73.0%, respectively, compared tothe conventional RD configuration with excess LA reactant.

� 2018 Published by Elsevier Ltd.

1. Introduction offer lower capital and utility costs and can be used for a wide

Ethyl levulinate (LAEE), which is one of main derivatives of levu-linic acid (LA), has attracted considerable interest as a next-generation fuel additive with its many advantages such as higherengine efficiency and lower CO and NOx emissions (Fagan andManzer, 2006; Joshi et al., 2011). The global LAEEmarket is expectedto reach USD 11.8 million by 2022 (Grand View Research, 2016).Because the global LAEE industry is still in its nascent stage, researchhas focused primarily on R&D for a novel commercial LAEE produc-tion technology. Unlu and Hilmioglu (2016) mentioned that thesynthesis of LAEE can be generally carried out using a homogeneouscatalyst in a conventional batch reactor. However, Huber et al.(2006) reported that ester levulinates, including LAEE, are normallyproduced in two reaction steps: LA is dehydrated to angelica lactoneand then mixed with an alcohol to form levulinate. However, thetwo reaction steps make the process complicated and expensive.To address the problem associated with the two reaction steps,Nandiwale et al. (2013) carried out an esterification experiment,in which LA reacted directly with alcohol to form levulinates.Because the esterification process is a reversible reaction wherehigh conversion can only be achieved if the backward reaction isminimized, they employed excess ethanol (EtOH) to reach the highconversion of LA in the esterification reaction. Recently, to over-come the equilibrium limitation in the esterification of LAmore effi-ciently, Novita et al. (2017) applied a reactive distillation (RD)technique to the LAEE production process and proposed severalreactive distillation configurations with excess LA reactant.

Reactive distillation is an important example of combined reac-tion and separation in a single column, and is especially useful forequilibrium-limited reaction systems (Buchaly et al., 2007; Taylorand Krishna, 2000). Reactive distillation offers numerous advan-tages, such as improved yield and selectivity, decreased energyrequirements, and avoidance of hot spots (Novita et al., 2015).These advantages are promising for the application of reactive dis-tillation columns in esterification, where the conventional processhas an equilibrium limitation and high operating costs.

Thermally-coupled distillation (TCD), where either the con-denser or the reboiler, or both, in one column are replaced by addi-tional thermal links (an interconnecting vapor liquid stream) to theother columns, is an attractive alternative that offers lower energyconsumption in distillation processes (Triantafyllou and Smith,1992). Both reactive distillation and thermally-coupled distillationare further developments of the conventional distillation unit(Wang et al., 2010) but represent two different methods of integra-tion. However, reactive distillation and thermally-coupled distilla-tion can be combined to create a new process integration concepttermed ‘‘thermally-coupled reactive distillation” (TCRD). By imple-menting thermally-coupled reactive distillation, the advantages ofboth processes (reactive distillation and thermally-coupled distil-lation) can be harnessed and enhanced, and further cost reductionis possible compared to the conventional distillation unit. In partic-ular, thermally-coupled reactive distillation is suitable to particularreaction-separation systems in which the unreacted componentsand products are integrated in at least a ternary mixture(Galindo et al., 2011).

Along with the distillation system, membrane systems are nowrecognized as a possible alternative to traditional energy intensiveseparation methods such as distillation. Membrane systems often

range of separations (Verhoef et al., 2008). Pervaporation (PV)membrane is most suitable if the liquid mixtures form an azeo-trope or contain thermally-unstable components and/or have com-ponents with close boiling points (Nagai, 2010). In pervaporation,the feed and retentate are in the liquid-state on the high-pressure side of the membrane, while the permeate is removedas a vapor on the low-pressure side of the membrane. Pervapora-tion has been the subject of many papers for several decades asmentioned by Luyben (2009). As a stand-alone process, pervapora-tion is probably not economically feasible. However, in a hybrid orcombined process, such as when coupled with a distillation unit ora reactor, the overall efficiency of pervaporation can be improved(Lipnitzki et al., 1999; Verhoef et al., 2008; Singh and Rangaiah,2017). Based on the advantages of membrane and distillation sys-tems, their combination with pervaporation is anticipated to findincreasing application in the future (Aiouache and Goto, 2003).Recently, the research effort for distillation-membrane hybrid sys-tems was expanded to combine it with more complicated distilla-tion systems such as thermally coupled and reactive distillation(Lee et al., 2016; Harvianto et al., 2017a, 2017b).

In this paper, a novel hybrid process combining reactive distil-lation and pervaporation is proposed for the enhanced productionof LAEE via esterification. To force the forward reaction in theesterification of LA to LAEE, an LA or ethanol reactant needs to beprovided in excess to the reactive distillation column while theproduct H2O recycled to the reactive distillation column, if any, isminimized. In the proposed hybrid process, an excess EtOH reac-tant configuration was chosen to utilize a highly selective mem-brane unit for EtOH/H2O separation. A reactive distillation-onlyconfiguration with the excess EtOH reactant, where the EtOH/H2O azeotrope mixture is recycled to the reactive distillation col-umn without further separation, was also examined to investigatethe effect of different excess reactants and the azeotrope mixtureseparation on overall process performance. Thermal intensificationwas also investigated for exploring further energy and cost reduc-tions in the RD–pervaporation hybrid process. The arrangements ofthe pervaporation unit are also discussed to highlight its signifi-cance in real applications of membrane hybrid configuration.

2. Model development

2.1. Thermodynamic and kinetic models

There are four components in the studied system, i.e., two reac-tants (EtOH and LA) and two products (LAEE and H2O). The non-random two liquid (NRTL) model was used to calculate the liquidactivity coefficient, and the Hayden and O’Connell (HOC) modelwas used to account for association in the vapor phase (Novitaet al., 2017). The EtOH and H2O pairing parameters were obtainedfrom the Aspen Plus built-in database, and the remaining parame-ters refer to the regressed results from Resk’s paper (Resk et al.,2014).

The reversible esterification reaction for the studied system isrepresented as:

C5H8O3 þ C2H5OH ()k1k2

C7H12O3 þH2O ð1Þ

ðLAÞ ðEtOHÞ ðLAEEÞ ðH2OÞ

F.J. Novita et al. / Chemical Engineering Science 190 (2018) 297–311 299

The kinetic model for the above reaction is obtained from theresearch studies by Tsai (2014) and Novita et al. (2017). A hetero-geneous AmberlystTM 39 catalyst was used in the study. Thenonideal-quasi-homogeneous (NIQH) kinetic model was used tofit the experimental data. The data of this kinetic model, basedon the activity, are summarized in Table 1. In the table,R = 8.314 kJ/kmol�K is the gas constant and T (K) is the absolutetemperature. Note that the unit of the reaction rate constants iskmol/kgcat/s, which is converted to kmol/m3/s in the simulationstudy by assuming a catalyst density of 770 kg/m3.

2.2. Pervaporation unit

In this study, a commercial simulator, Aspen Plus V8.6, wasused to obtain a rigorous simulation. The RadFrac module basedon the equilibrium stage model was used to calculate the totalmaterial, components, and energy balance in each column. A sim-ple engineering model from Luyben (2009) was adopted to per-form a pervaporation unit. A pervaporation model was built upusing Aspen Custom Modeler V8.4 software and then integratedinto the Aspen Plus V8.6 software.

The dynamics of the membrane model from Luyben (2009) wasbased on the pervaporation performance curves reported bySander and Soukup (1988). In Sander and Soukup’s report, theyclearly discussed about utilizing a highly selective polyvinyl alco-hol (PVA) composite membrane as a pervaporation unit for theethanol dehydration process. In this context a PVA membrane,which has the hydrophilic characteristic that allows the water topass selectively across the membrane, was chosen in this study.The amount of mass transfer in the membrane depends on the dif-fusivity of each species. Water has a larger diffusion coefficientthan other components because the membrane is hydrophilic.

The curve for the EtOH/H2O system indicates relationshipsbetween three variables, i.e., the flux of H2O through the pervapo-ration membrane, the composition of the permeate vapor, and theconcentration of EtOH in the liquid feed. The pervaporation modelassumes that diffusion through the membrane depends on the con-centration difference between the retentate and permeate sides ofthe membrane for each component:

fluxj ¼ DjðCRj � CPjÞ ð2Þwhere fluxj denotes the molar flux of component j (kmol/h�m�2), Dj

is the diffusion coefficient (m/h), CRj is the concentration of compo-nent j in the retentate liquid (kmol/m3), and CPj is the concentrationof component j in the permeate vapor (kmol/m3).

Normally, the diffusion coefficient, Dj, is temperature depen-dent, and can be expressed as the Arrhenius form:

Di ¼ Di;0 exp�Ei

RT

� �ð3Þ

Based on the data from the curve for the EtOH/H2O system(Sander and Soukup, 1988) and using Eq. (2), Luybencalculated the diffusion coefficient of EtOH and H2O (Luyben,2009). In his work, the average values of DH2O ¼ 0:0137 m/h and

Table 1Kinetic model and parameters for the studied system.

Kinetic model (catalyst) k1 Ka

Quasi-homogeneous model(AmberlystTM 39)

�rA ¼ k1 aLAaEtOH � aLAEEaH2OKa

� �

k1 ¼ 133:33 exp�37790

RT

� �

k2 ¼ 3:077 exp�36915:37

RT

� �

at T = 353.5 K,

k1 ¼ 3:47� 10�4

[kmol/kgcat/s]

at T = 353.5 K,Ka ¼ K1

K2 ¼ 32:18

DEtOH = 2.4 � 10�5 m/h at 373 K were used for calculation of themembrane parameters. The diffusion coefficient used in this study,referenced from Luyben’s work, are presented below:

DH2O ¼ 1:158� 109e�9385=T

DEtOH ¼ 2:028� 106e�9385=T ð4ÞThe flux of each component in each cell was calculated by using

Eqs. (2) and (4) given that all compositions and the temperatureare known. The NRTL-HOC relationship was used in the PV unit,where the feed in the pervaporation unit consists of EtOH andH2O only.

3. Discussion of process flowsheet

The esterification of LA and EtOH can be classified as Type IRaccording to the classification system based on the relative volatil-ities of the reactants and products presented by Tung and Yu(2007). Moreover, the excess design is preferable for a system withthe lightest reactant, type IR (Lin, 2007). Based on these premises,Novita et al. (2017) proposed a reactive distillation process, whichconsists of one reactive distillation and two simple columns in ser-ies that employs excess LA for enhanced LAEE production. Fig. 1shows the flowsheet of the reactive distillation process with theexcess LA design for the LAEE production process (Novita et al.,2017). This conventional RD-only configuration using LA as theexcess reactant (RDLA-only configuration) required a total reboilerduty of 7.77 Gcal/h with a total annual cost (TAC) of $5.27 � 106 toproduce 99.5% mole fraction of LAEE.

In the RDLA-only configuration, because the reaction zone islocated in the top section of the reactive distillation column andLA was chosen as the excess reactant, no EtOH remained in the pro-duct stream. Thus, the azeotrope formedbetweenEtOHandH2Oandthe separation associated with the azeotrope mixture was avoided.Chua et al. (2017) studied the effect of different excess reactantsfor the RD process to produce isopropyl alcohol and showed thatusing excess H2O is more economical than using excess propylene,even though the excess H2O involves an azeotrope in the process.

In this study, a new RD configuration using EtOH as the excessreactant was investigated for enhanced LAEE production. In thecase of excess EtOH reactant, the azeotropic EtOH/H2O mixture,which is difficult to separate, is essentially involved in the process.Two different approaches can be considered to treat this azeotropicEtOH/H2O mixture: recycling the excess EtOH either without sepa-ration or after separation of the azeotropic EtOH/H2O mixture. Therecycling of excess EtOH after removing H2O is desirable for theesterification reaction because the reverse reaction by H2O can beminimized. However, the separation of the azeotrope mixturerequires additional capital and energy costs. Therefore, there mightbe a trade-off, and thus the overall process economics of using anexcess EtOH reactant will be mainly determined by the availabilityof an effective way to separate the azeotrope mixture. It is easilyexpected that use of a conventional distillation system for the sep-aration of the azeotropic EtOH/H2Omixture through extractive andazeotropic distillation is undesirable because of its high energy andcost-intensive features. Many practical advantages of pervapora-tion such as less equipment, lower cost, and lower energy require-ments, as well as the commercial availability of a high-performancemodule for alcohol dehydration, prompted investigation into theRD–pervaporation hybrid process using excess EtOH reactant pro-posed in this study. To investigate the effect of excess EtOH reactantand azeotrope separation on the RD process performance and eco-nomics, an RDEtOH-only configuration, where excess EtOH is directlyrecycled to the RD column without azeotrope separation, was alsostudied, as discussed in the following sub-section.

Fig. 1. Flowsheet of the excess LA design for the LAEE production process; X in XEtOH, XH2O, XLAEE, and XLA denotes the mole fraction (Novita et al., 2017).

300 F.J. Novita et al. / Chemical Engineering Science 190 (2018) 297–311

3.1. Design and optimization of the RDEtOH-only configuration

Fig. 2 shows the flowsheet of the RDEtOH-only configuration forthe LAEE production process. In this RD configuration, because

Fig. 2. Flowsheet of the RDEtOH-only configuration; X in X

EtOH was selected as the excess reactant, the reaction zone wasimplemented in the lower section of the reactive distillation col-umn. LA is fed from the third stage and the EtOH mixture is fedfrom the 32nd stage. In particular, 99.5 mol% of LAEE is produced

EtOH, XH2O, XLAEE, and XLA denotes the mole fraction.

F.J. Novita et al. / Chemical Engineering Science 190 (2018) 297–311 301

from the bottom stream of the RD column with no remaining LA.Moreover, unreacted EtOH and H2O are obtained as the distillate.The EtOH/H2O distillate mixture with a mole fraction ratio of0.07/0.93 is then fed into a subsequent distillation column (C1)for further separation, where pure H2O is obtained in the bottomstream and a near-azeotropic EtOH/H2O mixture is obtained inthe overhead stream. Note that less H2O is desirable in the recycledfeed to mitigate the reverse reaction. Thus, the near-azeotropicEtOH/H2O mixture with a mole fraction ratio of 0.88/0.12 is thendirectly recycled into the reactive distillation column.

The RDEtOH-only configuration was optimized to minimize thetotal annual cost (TAC) by using an iterative optimization proce-dure. Note that this sequential optimization procedure does notresult in a global optimal solution, but it does give useful insight

8 9 10

2.20

2.21

2.22

2.23

TAC

(106 $)

EtOH recycle (kmol/hr)

32 34 36

2.590

2.597

2.604

TAC

(106 $)

EtOH feed location

49 50 51

2.59

2.60

2.61

TAC

(106 $)

Reactive stages (NRXN

)

Ns = 2, N

R = 2

Fig. 3. Sensitivity plots for the main des

into the sensitivity of the design variables on the process perfor-mance and economics (Novita et al., 2017). Eq. (5) indicates thatthe payback period (n) for the TAC, which comprises the annualoperating cost (AOC) and the annualized total capital cost (TCC),is three years. In this study, the capital cost covers the costs ofthe column, trays, and heat exchangers (condenser and reboiler),and the operating cost includes the costs of the heating media(the steam or the fuel oil depending on the bottom temperatureof the columns), cooling water, refrigerant, and catalyst. The calcu-lation was based on the study by Douglas (1998), ignoring the costof piping and pumps.

TAC ¼ AOC þ TCCn

ð5Þ

36 37 382.198

2.200

2.202

2.204

2.206

TAC

(106 $)

C1 total stages (NT,C1

)

NF,C1

= 34 N

F,C1 = 35

NF,C1

= 36 N

F,C1 = 37

322.5898

2.5900

2.5902

2.5904

2.5906

2.5908

TAC

(106 $)

Stripping stages (Ns)

NR = 2, N

RXN = 49

ign variables in the column section.

6 12 18 24 30 36 42 48 5480

100

120

140

160

180

200

Tem

pera

ture

(C)

Stage

Temperature (C)

NRXNNR NS

Fig. 4. Temperature profile of the RD column in the RDEtOH-only configuration.

6 12 18 24 30 36 42 48 540.0

0.2

0.4

0.6

0.8

1.0

Mol

e fr

actio

n

Stage

EtOH H

2O

LAEE LA

Fig. 5. Liquid composition profile of the RD column in the RDEtOH-onlyconfiguration.

302 F.J. Novita et al. / Chemical Engineering Science 190 (2018) 297–311

There are six main design variables to be optimized in the col-umn section: (1) the number of reactive stages in the reactive dis-tillation column (Nrxn,RD); (2) the number of stripping stages in thereactive distillation column (Ns,RD); (3) the EtOH feed location; (4)the number of stages in C1 (NC1); (5) the C1 feed stage (NF,C1), and(6) the EtOH recycling. Fig. 3 presents the sensitivity plots for thesevariables. The steps used in the iterative optimization procedurefor the column section are as follows: (1) Place the heavy reactantfeed (NF,LA) on the upper part of the reaction section and locate thelight reactant feed (NF,EtOH) several trays away from the lower partof the reaction section. (2) Fix the number of stripping sections (Ns,

RD). (3) Guess the number of stages in the reactive trays (Nrxn,RD).(4) Go back to 3 and vary Nrxn,RD until the TAC is minimized. (5)Go back to 2 and vary Ns,RD until the TAC is minimized. (6) Go backto 1 and vary the NF,EtOH until the TAC is minimized. (7) Fix thenumber of C1 columns (NT,C1). (8) Guess the C1 feed location (NF,

C1). (9) Go back to 8 and change NF,C1 until the TAC is minimized.(10) Go back to 7 and vary NT,C1 until the TAC is minimized. (11)Vary the flowrate of recycled EtOH until the TAC is minimized.

Three assumptions were made to design the reactive distillationcolumn in the design procedure (Novita et al., 2017): (1) the down-comer occupies a 10% tray area; (2) the liquid holdup is set as half-full with the AmberlystTM 39 catalyst in each reactive stage; (3) theweir height is 0.1 m, giving a reactive liquid holdup of 0.069 m3. Inthis study, the catalyst cost was also included in the AOC, wherethe catalyst life was assumed to be three months (Tang et al.,2005; Lee et al., 2016). The column pressures were fixed as 1atm, the same as in the excess LA reactant process, for the use ofcooling water and for fair comparison by excluding the pressureeffect on efficiency improvement. The maximum operating tem-perature of AmberlystTM 39 is 130 �C, which was considered themain constraint in the design of the reactive distillation column.The reactive distillation column has 53 stages where the reactivestage (i.e., the 3rd to 51th stages) consists of 49 stages. Increasingthe number of reactive stages would increase the reactive holdup,which would further reduce the required reboiler duty. However,more stages increase the capital and catalyst cost. As seen fromFig. 3, using a total of 49 reactive stages produced a trade-offbetween these two competing effects. It should be noted that thenumber of reactive stages could not be less than 49 because ofthe maximum allowable temperature limit of the catalyst. Thesame consideration was applied for the stripping and rectifyingstages. These stages could not be less than two for each stage(including condenser and reboiler). The EtOH feed location is alsoan important variable that can reduce the TAC. Because the EtOHfeed exceeds the temperature limit below the 32nd stage, the32nd stage was chosen as the EtOH feed location. As the numberof C1 stages increases, the capital cost increases, but the energyrequirement decreases due to the improved separation ability ofthe C1 column. The optimal number of stages for the C1 columnwas found to be 37 stages. The optimal feed stage of the C1 columnwas the 36th stage. A smaller recycle flowrate corresponds to asmaller column diameter and a lower recycling cost. However, ata recycled EtOH flowrate of less than 8 kmol/h, the temperatureof the reactive stages exceeds the limit. Figs. 4 and 5 show the tem-perature and liquid composition profiles, respectively, of the reac-tive distillation column for the RDEtOH-only configuration. It can beseen from Fig. 4 that the temperature in the reactive stages (the3rd to 51st stages) is below 130 �C. As a result, the total duty ofthe RDEtOH-only configuration was 2.61 Gcal/h, with an optimalTAC of $2.2 � 106.

An interesting phenomenon (i.e., a gradual temperatureincrease while going to the top section) was found from the tem-perature profile in the reactive section (from the 5th to 50thstages), as seen in Fig. 4, which is impossible in a normaldistillation column. In the reactive distillation column, EtOH as

the lightest component of the system enters into the lower partof the reactive section (i.e., the 32th stage). As seen in Fig. 5,because of the esterification reaction, its concentration decreasesas it travels to the upper section. However, LA, which is the heav-iest component of the system, enters into the top part of the reac-tive section (i.e., the third stage). Its concentration decreasesbecause of the esterification reaction as it travels to the lower sec-tion. Both LAEE and H2O as products are in the middle ranking ofthose two reactant components and fractionated in the sameway as in normal distillation, i.e., the concentration of LAEE, whichis heavier than H2O, increases, whereas the H2O concentrationdecreases as it travels to the lower section. The stage temperatureof the reactive section is finally determined from this competingeffect between feed and product components in the reactive sec-tion. As a result, in the RD column for the esterification of LA, agradual decrease in stage temperature occurred between the 3rdto 32th stages in the reactive distillation section, as seen inFig. 4. Because very little LA remains by the esterification reaction,whereas only the LAEE and the excess reactant EtOH remain,almost no reaction occurs and a component pinch zone is formedin the reactive distillation zone from the 33th to the 51th stage,

F.J. Novita et al. / Chemical Engineering Science 190 (2018) 297–311 303

which results in minimal temperature changes in the correspond-ing stages, as seen in Fig. 4.

3.2. Design and optimization of RD–PV hybrid configuration

In the hybrid configuration combining reactive distillation andpervaporation (the RD–PV hybrid configuration), the near-azeotropic EtOH/H2O mixture from the overhead of the C1 columnis transferred to the pervaporation unit for separation of EtOH fromH2O; the purified excess EtOH is then recycled to the reactive dis-tillation column. Fig. 6 shows a flowsheet of the proposed RD–PVhybrid configuration. The separation task in the columns and thedesign procedure assumptions were the same as those describedin Section 3.1.

An iterative optimization procedure was also applied to thishybrid configuration. Fig. 7 presents the sensitivity plot of the maindesign variables in the column section for the hybrid configuration.As shown in Fig. 7, the reactive distillation column has the samestructure as the RDEtOH-only configuration, i.e., a total of 53 stageswith 49 reactive stages (from the 3rd to the 51th stages), whereasthe optimal number of stages in the C1 column was found to be 22stages due to reduced C1 column feed flow. Note that the optimalrecycle flowrate in the hybrid configuration could be significantlyreduced by up to 6 kmol/h as compared to the RDEtOH-only config-uration, which leads to lower capital and operating costs for boththe RD and C1 columns. At less than 6 kmol/h, the temperatureof the reactive stages exceeds the limit. Fig. 8 shows the tempera-ture profile of the reactive distillation column in the RD–PV hybridconfiguration. It can be seen from Fig. 8 that the temperature in thereactive stages remains below 130 �C as the maximum allowabletemperature of the catalyst.

As a result, the TAC of the RD-PV hybrid configuration was$2.06 � 106 (comprising the column section: $1.97 � 106 and themembrane section (in series): $0.95 � 105) and the total reboilerduty was 2.28 Gcal/h. The design of the membrane section is dis-cussed in the following subsection.

Fig. 6. Flowsheet of the proposed RD-PV hybrid configuration;

3.3. Discussion of membrane arrangement

In the membrane separation section, the arrangement of thepervaporation unit is an important design issue. Thus far, to thebest of our knowledge, the relationship between the arrangementof the pervaporation unit and the main process variables affectingthe membrane flux has not been documented. In this study, thearrangement of the pervaporation unit will be studied as well todetermine the best arrangement of the pervaporation unit for thisLAEE production process. There are two different basic arrange-ments of the membrane: parallel and series arrangements (Naiduand Malik, 2011). Fig. 9 shows a schematic of these twoarrangements.

In the parallel arrangement, the feed stream is first preheatedand split into several streams. Each stream is then sent to the per-vaporation unit to meet the product specification in the retentateside via the single module of the pervaporation unit. On the otherhand, in the series arrangement, the single feed stream is pre-heated and sent to the pervaporation unit directly without anysplitting. The required product specification is achieved in theretentate side via several pervaporation modules arranged inseries.

The membrane flux is an important variable in designing thepervaporation unit because it is closely related to the requiredmembrane area. When the membrane flux increases, the requiredmembrane area decreases. Eqs. (2) and (3) indicate that the con-centration gradient and operating temperature are the two mainvariables affecting the membrane flux.

In this study, both membrane arrangements were investigatedunder the same inlet conditions. The membrane and the feed con-ditions in this study have taken from to the work by Luyben (2009),where the mole fraction of the EtOH/H2O feed mixture was0.85/0.15 at 101.8 �C. For the maximum temperature tolerated byPVA membrane, Sander and Soukup (1988) stated that it is ulti-mately determined by the thermal stability of the PVA membrane.In the work by Tsuchiya and Sumi (1969), they reported that thethermal decomposition temperature (Td) of PVA begins at about

X in XEtOH, XH2O, XLAEE, and XLA denotes the mole fraction.

48 49 50 512.45

2.46

2.47

2.48

2.49

2.50

TAC

(106 $)

Reactive stages (NRXN

)

Ns = 2, N

R = 2

322.460

2.463

2.466

2.469

TAC

(106 $)

Stripping stages (Ns)

NR = 2, N

RXN = 49

32 36 41 462.20

2.25

2.30

2.35

2.40

2.45

2.50

TAC

(106 $)

EtOH feed location

18 19 20 21 222.24

2.25

2.26

2.27

TAC

(106 $)

C1 feed location

NT,C1

= 21 N

T,C1 = 22

NT,C1

= 23

6 8 10 112.05

2.10

2.15

2.20

2.25

TAC

(106 $)

EtOH recycle (kmol/hr)

Fig. 7. Sensitivity plot of the main design variables in the column section for the hybrid configuration.

304 F.J. Novita et al. / Chemical Engineering Science 190 (2018) 297–311

200 �C. Thus, the inlet operating temperature of 101.8 �C is stillallowable for PVA membrane. The azeotropic EtOH/H2O mixturefrom the C1 column was compressed to a pressure of 7 atm toensure that the retentate remained in the liquid-phase, and fedinto the pervaporation unit. The permeate pressure was 0.155atm, and the vapor permeate streams from each module weretransferred to a refrigerated condenser (with an area of 0.203 m2

and an overall heat transfer coefficient of 0.0203 cal/s cm2 K),where the flow rate for condensing the permeate vapor using 0�C refrigerant was 5.66 kmol/h. The heat duty of the condenserwas 0.01 Gcal/h. The retentate flow (6 kmol/h) from the pervapora-tion unit was fed back into the RD column.

The result showed that the series arrangement with two mod-ules requires a total membrane area of 88 (44 � 2) m2 to separate

the azeotropic EtOH/H2O mixture. On the other hand, the parallelarrangement requires more modules (three modules) of the perva-poration unit and a larger total membrane area of 348 (116 � 3)m2. This result illustrates that the arrangement of the pervapora-tion unit is an important factor that must be considered in themembrane separation section. Note that each module has threecells (or lumps) for both membrane arrangements.

Table 2 presents the details of each membrane arrangement.Notably, the data in Table 2 show that when the parallel arrange-ment was applied, the H2O concentration in the feed stream wasmaintained at the highest value (0.15 mol fraction of H2O) for allof the pervaporation modules, despite the large temperature drop(DT = 36.7 �C). On the other hand, the H2O concentration in thefeed stream decreased gradually in the series arrangement, moving

Table 2Detailed stream information for different membrane arrangements.

Parallel arrangement

Module 1 Module 2

Flux (g/m2 h) 11.72 11.72Stream type F R F(Feed: F)(Retentate: R)Pressure (atm) 7 6.99 7Temp. (�C) 101.8 65.1 101.8

Composition (mole fraction)EtOH 0.85 0.955 0.85H2O 0.15 0.045 0.15

Series arrangement

Module 1

Flux (kmol/m2 h) 59.42Stream type F R(Feed: F)(Retentate: R)Pressure (atm) 7 6.97Temp. (�C) 101.8 81.4

Composition (mole fraction)EtOH 0.85 0.91H2O 0.15 0.09

Feed

Feed

tnemegnarrAlellaraP

1

2

n

1

Permeate

Retentate

Permeate

Fig. 9. Schematic of the parallel and

6 12 18 24 30 36 42 48 548090

100110120130140150160170180190200210

Tem

pera

ture

(C)

Stage

Temperature (C)

NR NRXN NS

Fig. 8. Temperature profile of the RD column in the hybrid RD-PV configuration.

F.J. Novita et al. / Chemical Engineering Science 190 (2018) 297–311 305

from a mole fraction of 0.15–0.09. However, the temperature drop(DT = 15–20 �C) was smaller than that in the parallel arrangement.

In addition, the simulation results showed that the seriesarrangement requires a smaller total membrane area than the par-allel arrangement because of the larger membrane flux in the for-mer (the membrane flux in the series arrangement was 3.5–5-foldhigher than that in the parallel arrangement). These results indi-cate that in the series arrangement, the influence of the operatingtemperature is much more important than that of the concentra-tion gradient.

A lower total membrane area also has an economic impact.Fig. 10 shows a cost comparison of both membrane arrangements.The membrane price of 775 US$/m2 was taken from Szitkai et al.(2002). It was also assumed that one-third of the membrane areaneed to be replaced each year, on average based on the membranelifetime of three years (Szitkai et al., 2002). In this study, the mem-brane price was divided into two parts: (1) the membrane materialwhich has cost per m2 and (2) the membrane modules which havecost per module (Santoso, 2010; Lee et al., 2016). The capital andoperating cost of the vacuum pump were taken from Woods

Module 3

11.72R F R

6.99 7 6.9965.1 101.8 65.1

Out from PV0.955 0.85 0.955 0.9550.045 0.15 0.045 0.045

Module 2

41.58F R

6.97 6.94101.8 87.1

Out from PV0.91 0.955 0.9550.09 0.045 0.045

tnemegnarrAseireS

2

Retentate

n

Permeate

series membrane arrangements.

Pararel Series0

50

100

150

PV arrangement

Cos

t (10

3 $/ye

ar)

Membrane cost (103$/year) Modules cost (103$/year) Total cost (103$/year)

Fig. 10. Comparison of cost effectiveness of both membrane arrangements.

306 F.J. Novita et al. / Chemical Engineering Science 190 (2018) 297–311

(1983) and Oliveira et al. (2001), respectively. As shown in Fig. 10,for the LAEE hybrid process it turned out that the series arrange-ment is more economical and suitable than the parallel arrange-ment. However, in the case where the influence of theconcentration gradient is greater than that of the operating tem-perature, the parallel membrane arrangement can be moreadequate.

3.4. Comparison of RD-PV hybrid configuration with the RDLA-only andRDEtOH-only configurations

Table 3 compares the details of the cost distribution of the RDLA-only configuration (Novita et al., 2017), the RDEtOH-only configura-

Table 3Detailed cost distribution of the RDLA-only configuration (Novita et al., 2017), the RDEtOH-

Conventional RDLA-only configuration Conventional RDEtOH-only c

RD column capital cost ($ 1000/year)Column 254.88 ColumnTrays 49.19 TraysHeat exchangers 204.53 Heat exchangers

RD column operating cost ($ 1000/year)Catalyst 34.44 CatalystLP 1791.55 HPCooling water 66.95 Cooling water

C1 column capital cost ($ 1000/year)Column 62.22 ColumnTrays 8.19 TraysHeat exchangers 96.28 Heat exchangers

C1 column operating cost ($ 1000/year)HP 1536.05 LPCooling water 19.48 Cooling waterC2 column capital cost ($ 1000/year)Column 100.58Trays 15.18Heat exchangers 113.00

C2 column operating cost ($ 1000/year)Fuel oil 897.15Cooling water 22.12

Total annual cost (TAC) ($ 1000/year)5271.79 2197.96

tion, and the RD–PV hybrid configuration. Although a pervapora-tion unit is additionally included, the proposed hybridconfiguration offers an important advantage derived from the com-bination of distillation and membrane systems as well as the cleverchoice of the excess reactant in the reactive distillation design. Thetotal duty (2.28 Gcal/h) and TAC ($2.06 � 106) of the RD–PV hybridconfiguration were significantly lower than those of the RDLA-onlyconfiguration (total duty 7.77 Gcal/h, TAC = $5.27 � 106) and lowerthan those of the RDEtOH-only configuration (total duty 2.61 Gcal/h,TAC = $2.20 � 106). Moreover, the RDLA-only configuration had ahigher catalyst cost than the RDEtOH-only configuration becauseof a reactive condenser, for which the holdup was set as the max-imum kinetic holdup, which was taken to be 20 times that of thetray holdup (Tung and Yu, 2007; Novita et al., 2017).

In this proposed process configuration, the LAEE (as the heavi-est product) exits the reactive distillation column first, and thus,the heaviest reactant, LA, is not present in the product stream,which makes separation in the C1 column much easier. Thiscontribution is clear from the low reboiler duty of the C1 column(0.31 Gcal/h in the proposed hybrid configuration and 0.57 Gcal/hin the RDEtOH-only configuration compared with 2.37 Gcal/h inthe RDLA-only configuration). Thus, the separation problem dueto the azeotropic mixture could be solved in the most efficientway by embedding the pervaporation unit instead of usingadditional columns, which also reduces the total duty and totalcost of the process: the TAC of the C2 column was $11.48 � 105

in the RDLA-only configuration, whereas the pervaporation unit inseries only cost $0.95 � 105.

3.5. TCRDSS–PV hybrid configuration

Although the RD-PV hybrid configuration afforded a largeimprovement in term of TAC and energy saving compared to theRD-only configurations, the process can still be further optimized,especially in terms of energy saving. Fig. 11 shows the liquid

only configuration, and the RD–PV hybrid configuration.

onfiguration RD-PV hybrid configuration

258.81 Column 258.8150.13 Trays 50.13121.58 Heat exchangers 119.92

1.64 Catalyst 1.641325.15 HP 1280.3130.66 Cooling water 30.45

61.98 Column 30.066.66 Trays 2.5454.99 Heat exchangers 36.68

279.18 LP 151.827.18 Cooling water 3.82

Pervaporation capital cost ($ 1000/year)Membrane 11.37Modules 45.98Vacuum pump 3.64Heat exchangers 3.38Feed pump 0.48Pervaporation operating cost ($ 1000/year)Membrane replacement 22.73LP 4.49Electricity 1.12Refrigerant 1.80

2061.17

6 12 18 24 30 36 42 48 540.0

0.2

0.4

0.6

0.8

1.0

Mol

e fr

actio

n

Stage

EtOH H

2O

LAEE LA

1 2 3 40.0

0.2

0.4

0.6

0.8

1.0

Mol

e fr

actio

n

Stage

EtOH H

2O

LAEE LA

Fig. 11. Liquid composition of the RD column in the RD-PV hybrid configuration.

F.J. Novita et al. / Chemical Engineering Science 190 (2018) 297–311 307

composition for the reactive distillation column in the RD-PVhybrid configuration.

It can be seen from Fig. 11 that there is a slight remixing effectin the top part of the reactive distillation column. Thus, modifyingthe column section to the thermally-coupled structure, where onereboiler or condenser, or both, is eliminated and replaced with theinterconnecting vapor liquid stream, may possibly offer savings interms of both energy and capital costs by reducing the remixingeffect.

In the proposed hybrid configuration, the apparent remixingeffect mostly occurs in the rectifying section of the reactive distil-lation column. Therefore, thermal links must be placed in theupper section of the column by eliminating one condenser; a ther-mally coupled reactive distillation with a side-stripper (TCRDSS)configuration. Fig. 12 shows the proposed TCRDSS-PV hybridconfiguration.

Fig. 12. Flowsheet of the TCRDSS -PV hybrid configuration; X

Note that the TCRD–PV hybrid configuration was also imple-mented in the study of Harvianto et al. (2017a) for n-butyl acetateproduction process. Besides its different application, the structureof the TCRD–PV hybrid configuration used in this study was differ-ent from that of Harvianto et al. (2017a,b). Because the pervapora-tion membrane for EtOH–H2O separation has a sufficiently highseparation performance, no further separation and recycling ofthe permeate stream from the membrane was required, whichmakes the hybrid configuration simple and practical.

To focus on the energy saving derived from the column config-uration only, the main structures (the total number of stages, etc.,)of the columns and the membrane section in the TCRDSS-PV hybridconfiguration were kept the same as those in the RD-PV hybridconfiguration. In this way, all of the energy saving is derived fromthe liquid side stream flow rate and the locations of the thermallink feed and EtOH feed. Further improvements in energy efficiency

in XEtOH, XH2O, XLAEE, and XLA denotes the mole fraction.

140 143 146

0.9940

0.9945

0.9950LA

EE m

ole

frac

tion

Liquid side stream flow rate (kmol/hr)14 16 18

0.9940

0.9945

0.9950

LAEE

mol

e fr

actio

n

Thermal links feed location

50 51 52

0.9947

0.9948

0.9949

0.9950

LAEE

mol

e fr

actio

n

EtOH stage location

Fig. 13. Effect of three variables in the TCRDSS-PV configuration on the production of LAEE (mole fraction).

308 F.J. Novita et al. / Chemical Engineering Science 190 (2018) 297–311

and cost saving will be possible if the structures of the TCRD andmembrane section are fully optimized. Fig. 13 shows the optimiza-tion results for these design variables. The optimal liquid sidestream flowrate was 143 kmol/h, the location of the thermal linkfeed was in the 16th stage, and the EtOH feed location was in the52nd stage. By eliminating one condenser in the reactivedistillation column and replacing it with the thermal links, the

6 12 18 24 30 36 42 48 540.0

0.2

0.4

0.6

0.8

1.0

Mol

e fr

actio

n

Stage

EtOH H

2O

LAEE LA

Fig. 14. Liquid composition profile for reactive distillation column in hybridTCRDSS-PV configuration.

total duties of the reactive distillation and C1 columns could bereduced by approximately 0.18 Gcal/h.

Fig. 14 shows the liquid composition profile for the reactive dis-tillation column in the TCRDSS-PV hybrid configuration. The liquidcomposition profile at the bottom section of the reactivedistillation column was similar to that in the RD-PV hybrid config-uration. On the other hand, the remixing phenomenon in the top

6 12 18 24 30 36 42 48 54100

120

140

160

180

200

Tem

pera

ture

(C)

Stage

Temperature (C)

NRXN

Fig. 15. Temperature profile of the reactive distillation column in the TCRDSS-PVhybrid configuration.

Table 4Comparison of the RD-only, RD-PV hybrid, and TCRDSS–PV hybrid configurations.

Conventional Hybrid

RDLA-only RDEtOH-only RD–PV TCRDSS–PV

Total duty (Gcal/h) 7.77 2.61 2.28 2.10TAC ($) 5.27 � 106 2.20 � 106 2.06 � 106 1.97 � 106

Energy saving (%) 66.0 71.0 (13.0a) 73.0 (8.0b)TAC (%) 58.0 61.0 (6.0a) 63.0 (4.0b)

a Compared to the RDEtOH-only configuration.b Compared to the RD–PV hybrid configuration.

F.J. Novita et al. / Chemical Engineering Science 190 (2018) 297–311 309

section of the reactive distillation column was removed. Conse-quently, the total energy requirement of the TCRDSS-PV systemwas 2.10 Gcal/h, which is equivalent to energy savings of around8.0% compared to the RD-PV hybrid configuration. Note that thetemperature in the reactive section is far under the limit of the cat-alyst (Fig. 15). The TCRDSS-PV hybrid configuration had a TAC of$1.97 � 106 (column section: $1.88 � 106 and membrane section(in series): $0.95 � 105), which is equivalent to a TAC savings of4.0% compared to the RD-PV hybrid configurations. Table 4 sum-marizes the design results for the two RD-only and the two hybridconfigurations.

4. Conclusions

In this study, hybrid configurations combining reactive distilla-tion and membrane systems were proposed for enhanced LAEEproduction. The RD-PV hybrid configuration utilizes two columnsin sequence combined with the series-pervaporation arrangement.The selection of EtOH as an excess reactant for reactive distillationresulted in a remarkable reduction in the required energy of thesubsequent column. Consequently, the proposed RDEtOH-only con-figuration resulted in significant savings in the TAC and total dutyof as much as 58.0% and 66.0%, respectively, as compared to theRDLA-only configuration. In the proposed RD–PV hybrid configura-tion, the pervaporation unit was successfully utilized to separatethe azeotropic EtOH/H2Omixture in the most efficient way, achiev-ing further savings in the TAC and energy of as much as 6.0% and13.0%, respectively, as compared to the RDEtOH-only configuration.

Membrane arrangement for the pervaporation units was also animportant design factor in the membrane hybrid application. Theoperating temperature and concentration gradients are the twomain design variables affecting the diffusion rate of the membrane.In the proposed hybrid configuration, because the operating tem-perature was more influential than the concentration gradient,the series arrangement was found to be more advantageous thanthe parallel arrangement. The higher membrane flux in the seriesarrangement led to a smaller total membrane area. An intensifiedhybrid configuration utilizing the TCD, i.e., the TCRDSS-PV hybridconfiguration, was found to be an effective way to enhance theenergy efficiency of the RD-PV hybrid configuration. By removingthe remixing effect in the top section of the reactive distillationcolumn, the TCRDSS-PV hybrid configuration provided a furtherreduction in the TAC and energy for LAEE production. Thus, theproposed TCRDSS-PV hybrid configuration afforded TAC and energysavings of up to 63.0% and 73.0%, respectively, compared to theRDLA-only configuration, and 4.0% and 8.0%, respectively, comparedto the RD-PV hybrid configuration. These hybrid configurations canbe extended to other similar chemical processes involving esterifi-cation reactions and azeotropic mixtures.

Notes

The authors declare no competing financial interest.

Acknowledgements

This study was supported by the Basic Science Research Pro-gram through the National Research Foundation of Korea (NRF)funded by the Ministry of Education (2018R1A2B6001566) andby the Priority Research Centers Program through the NationalResearch Foundation of Korea (NRF) funded by the Ministry of Edu-cation (2014R1A6A1031189). The authors are also grateful for thesupport from the Ministry of Science and Technology of Taiwanunder grant MOST 104-2221-E-011-149-MY2.

Appendix A. Total annual and membrane costs calculation

The equipment cost estimation follows the procedure ofDouglas (1998). A payback period of three years for the exchangerand column is assumed, and a M&S index of 1473.3 (2010, 3rdquarter) is used in the calculation. The utility costs were takenfrom Turton et al. (2009). The membrane price was taken fromSzitkai et al. (2002). The membrane lifetime was assumed as threeyears during the calculation. In this study, the membrane price wasdivided into two parts: (1) the membrane material which has costper m2 and (2) the membrane modules which have cost per mod-ule (Santoso, 2010; Lee et al., 2016). The capital cost and the oper-ating cost of the vacuum pump refer to Woods (1983) and Oliveiraet al. (2001), respectively.

A.1. Reboiler heat exchange area (AR)

The reboiler heat exchange area (AR) is determined using thefollowing relation:

AR½ft2� ¼ QR

URDTRðA1Þ

where QR [BTU/h] is the reboiler duty, the overall heat-transfercoefficient UR is assumed 250 BTU/(h ft2 �F), and the temperaturedriving force DTR [�F] in the reboiler depends on the steam (for ahigh bottom temperature that cannot use a high pressure steam,it is used a fuel oil and DTR is assumed 45 �F).

A.2. Condenser heat exchange area (AC)

The condenser heat exchange area (AC) can be calculated usingthe expression,

AC½ft2� ¼ QC

UCDTCðA2Þ

where QC [BTU/h] is the condenser duty, the overall heat-transfercoefficient UC is assumed 150 BTU/(h ft2 �F), and DTc is given by

DTC ¼ 120� 90

In Tb�90Tb�120

� �with Tb being the condenser operating temperature (in �F).

310 F.J. Novita et al. / Chemical Engineering Science 190 (2018) 297–311

A.3. Height of column (LC)

The column height (LC) is calculated assuming 2-ft tray spacingand allowing 20% more height for base level volume:

LC ½ft� ¼ 2:4NT ðA3Þwhere NT is the total number of trays in the column.

A.4. Cost of column

The column cost is given by the expression

column cost½$� ¼ M&S280

� 101:9� D1:066C � L0:802C � ð2:18þ 3:67Þ

ðA4Þ

A.5. Cost of tray

The tray cost in the distillation columns can be defined as

tray cost½$� ¼ M&S280

� 4:7� D1:55C � LC � ð1þ 1:8þ 1:7Þ ðA5Þ

A.6. Cost of heat exchanger

The following expression is used to determine the heat exchan-ger cost,

heat exchanger cost ½$� ¼ M&S280

� 101:3� A0:65ð2:29þ FcÞ ðA6Þ

where Fc is 5.0625 for the reboiler and 3.75 for the condenser.

A.7. Utility costs

Utility type

Cost ($/GJ) (by Turtonet al., 2009)

Cooling water

0.354 Low pressure steam (LP) (5 barg,

160 �C)

13.28

Medium pressure steam (MP)(10 barg, 184 �C)

14.19

High pressure steam (HP) (41 barg,254 �C)

17.70

Fuel oil

14.2 0 �C refrigerant 4.92

A.8. Cost of membrane material

Membrane material cost ½$� ¼ 387:5=m2 ðA7Þ

A.9. Cost of membrane module

Membrane module cost ½$� ¼ 125550� A324

� �0:3

ðA8Þ

where A [m2] is membrane area for each module.

A.10 Cost of membrane replacement

The membrane replacement cost is taken from Szitkai et al.(2002) as 775 US$/m2. In this study, the membrane life time isassumed as 3 years. Therefore, one-third of the total membranearea has to be replaced each year, on average.

A.11. Cost of vacuum pump

vacuum pump cost ½$� ¼ 4200� CEPCI10CEPCI83

� 60� Fp � 8:314� 273:153600� 101:325

� �0:55

ðA9Þ

where Fp is the total permeate rate through themembrane [kmol/h].

A.12. Cost of feed pump

feed pump cost ½$� ¼ 26700� CEPCI10CEPCI83

� 24� FF � 360050000

� �0:53

ðA10Þwhere FF is the total feed rate to the pump [m3/s].

A.13. Power of vacuum pump

annual cost of power ½$� ¼ 8150� 0:04

� FP8:314:273:153600

� �kr

kr � 1

� ��

� 1:013Pop

� �ðkr�1=krÞ� 1

" #)ðA11Þ

where kr is the heat capacity ratio (1.33). Pop is the pressure on thepermeate side.

A.14. Catalyst cost (assuming a catalyst life of 3 months)

catalyst cost ½$� ¼ catalyst loading ½kg� � 7:7162$

kgðA12Þ

References

Aiouache, F., Goto, S., 2003. Reactive distillation-pervaporation hybrid column fortert-amyl alcohol etherification with ethanol. Chem. Eng. Sci. 58, 2465–2477.https://doi.org/10.1016/S0009-2509(03)00116-7.

Buchaly, C., Kreis, P., Gorak, A., 2007. Hybrid separation process-combination ofreactive distillation with membrane separation. Chem. Eng. Process.: ProcessIntensification 46, 790–799. https://doi.org/10.1016/j.cep.2007.05.023.

Chua, W.J., Rangaiah, G.P., Hidajat, K., 2017. Design and optimization of isopropanolprocess based on two alternatives for reactive distillation. Chem. Eng. Process.:Process Intensification 118, 108–116. https://doi.org/10.1016/j.cep.2017.04.021.

Douglas, J.M., 1998. Conceptual Design of Chemical Processes. McGraw-Hill.Fagan, P.J., Manzer, L.E., 2006. Preparation of levulinic acid esters and formic acid

esters from biomass and olefins. U.S. Patent 7,153,996.Galindo, E.Y.M., Herna’ndez, J.G.S., Herna’ndez, S., Antonio, C.G., Ramı’rez, A.B., 2011.

Reactive thermally coupled distillation sequences: Pareto front. Ind. Eng. Chem.Res. 50, 926–938. https://doi.org/10.1021/ie101290t.

Grand View Research. Ethyl Levulinate Market Worth $11.8 Million by 2022,March 2016. http://www.grandviewresearch.com/press-release/global-ethyl-levulinate-market (Accessed 13.09.2017).

Harvianto, G.R., Ahmad, F., Lee, M., 2017a. A thermally coupled reactive distillationand pervaporation hybrid process for n-butyl acetate production with enhancedenergy efficiency. Chem. Eng. Res. Des. 124, 98–113. https://doi.org/10.1016/j.cherd.2017.05.007.

Harvianto, G.R., Ahmad, F., Lee, M., 2017b. A hybrid reactive distillation processwith high selectivity pervaporation for butyl acetate production viatransesterification. J. Mem. Sci. 543, 49–57. https://doi.org/10.1016/j.memsci.2017.08.041.

Huber, G.W., Iborra, S., Corma, A., 2006. Synthesis of transportation fuels frombiomass: chemistry, catalysts, and engineering. Chem. Rev. 106, 4044–4098.https://doi.org/10.1021/cr068360d.

Joshi, H., Moser, B.R., Toler, J., Smith, W.F., Walker, T., 2011. Ethyl levulinate: apotential bio-based diluent for biodiesel which improves cold flow properties.Biomass Bioenergy 35, 3262–3266. https://doi.org/10.1016/j.biombioe.2011.04.020.

Lee, H.Y., Li, S.Y., Chen, C.L., 2016. Evolutional design and control of the equilibrium-limited ethyl acetate process via reactive distillation-pervaporation hybridconfiguration. Ind. Eng. Chem. Res. 55, 8802–8817. https://doi.org/10.1021/acs.iecr.6b01358.

F.J. Novita et al. / Chemical Engineering Science 190 (2018) 297–311 311

Lin, L.C., 2007. Effects of relative volatility ranking to the design of reactivedistillation: excess-reactant design. M.S. Thesis, National Taiwan University,Taipei.

Lipnitzki, F., Field, R.W., Ten, P., 1999. Pervaporation-based hybrid processes: areview of process design, applications and economics. J. Membr. Sci. 153, 183–210. https://doi.org/10.1016/S0376-7388(98)00253-1.

Luyben, W.L., 2009. Control of a column/pervaporation process for separating theethanol/water azeotrope. Ind. Eng. Chem. Res. 48, 3484–3495. https://doi.org/10.1021/ie801428s.

Nagai, K., 2010. Fundamentals and perspectives for pervaporation. In: Drioli, E.,Giorno, L. (Eds.), Comprehensive Membrane Science and Engineering. Elsevier,Oxford, pp. 243–271.

Naidu, Y., Malik, R.K., 2011. A generalized methodology for optimal configurationsof hybrid distillation-pervaporation processes. Chem. Eng. Res. Des. 89, 1348–1361. https://doi.org/10.1016/j.cherd.2011.02.025.

Nandiwale, K.Y., Sonar, S.K., Niphadkar, P.S., Joshi, P.N., Deshpande, S.S., Patil, V.S.,Bokade, V.V., 2013. Catalytic upgrading of renewable levulinic acid to ethyllevulinate biodiesel using dodecatungstophosphoric acid supported ondesilicated H-ZSM-5 as catalyst. Appl. Catal. A Gen. 460, 90–98. https://doi.org/10.1016/j.apcata.2013.04.024.

Novita, F.J., Lee, H.Y., Lee, M., 2015. Self-heat recuperative dividing wall column forenhancing the energy efficiency of the reactive distillation process in the formicacid production process. Chem. Eng. Process.: Process Intensification 97, 144–152. https://doi.org/10.1016/j.cep.2015.09.007.

Novita, F.J., Lee, H.Y., Lee, M., 2017. Energy efficient design of ethyl levulinatereactive distillation process via thermally coupled with external heat-integrated arrangement. Ind. Eng. Chem. Res. 56 (24), 7037–7048. https://doi.org/10.1021/acs.iecr.7b00667.

Oliveira, T.A., Cocchini, U., Scarpello, J.T., Livingston, A.G., 2001. Pervaporation masstransfer with liquid flow in the transition regime. J. Membr. Sci. 183, 119–133.https://doi.org/10.1016/S0376-7388(00)00576-7.

Resk, A.J., Peereboom, L., Kolah, A.K., Miller, D.J., Lira, C.T., 2014. Phase equilibria insystems with levulinic acid and ethyl levulinate. J. Chem. Eng. Data 59 (4),1062–1068. https://doi.org/10.1021/je400814n.

Sander, U., Soukup, P., 1988. Design and operation of a pervaporation plant forethanol dehydration. J. Membr. Sci. 36, 463–475. https://doi.org/10.1016/0376-7388(88)80036-X.

Santoso, A., 2010. Design and control of hybrid distillation-membrane systems forseparating azeotropicmixtures. M.S. Thesis, National TaiwanUniversity, Taiwan.

Singh, A., Rangaiah, G.P., 2017. Review of technological advances in bioethanolrecovery and dehydration. Ind. Eng. Chem. Res. 56, 5147–5163. https://doi.org/10.1021/acs.iecr.7b00273.

Szitkai, Z., Lelkes, Z., Rev, E., Fonyo, Z., 2002. Optimization of hybrid ethanoldehydration systems. Chem. Eng. Process.: Process Intensification 41, 631–646.https://doi.org/10.1016/S0255-2701(01)00192-1.

Tang, Y.R., Chen, Y.W., Huang, H.P., Yu, C.C., Hung, S.B., Lee, M.J., 2005. Design ofreactive distillations for acetic acid esterification. AIChE J. 51, 1683–1699.https://doi.org/10.1002/aic.10519.

Taylor, R., Krishna, R., 2000. Modelling reactive distillation. Chem. Eng. Sci. 55,5183–5229. https://doi.org/10.1016/S0009-2509(00)00120-2.

Triantafyllou, C., Smith, R., 1992. The design and operation of fully thermallycoupled distillation columns. Trans. Inst. Chem. Eng. 70A, 118.

Tsai, C.Y., 2014. Kinetic behavior study on the synthesis of ethyl levulinate overheterogeneous catalyst. M.S. Thesis, NTUST, Taiwan.

Tsuchiya, Y., Sumi, K., 1969. Thermal decomposition products of poly (vinylalcohol). J. Polym. Sci. A-1 Polym. Chem. 7 (11), 3151–3158. https://doi.org/10.1002/pol.1969.150071111.

Tung, S.T., Yu, C.C., 2007. Effects of relative volatility ranking to the design ofreactive distillation. Aiche J. 53 (5), 1278–1297. https://doi.org/10.1002/aic.11168.

Turton, R., Bailie, R.C., Whiting, W.B., Shaeiwitz, J.A., 2009. Analysis, Synthesis, andDesign of Chemical Processes. Prentice Hall.

Unlu, D., Hilmioglu, N.D., 2016. Synthesis of ethyl levulinate as a fuel bioadditive bya novel catalytically active pervaporation membrane. Energy Fuel 30, 2997–3003. doi: 0.1021/acs.energyfuels.5b02911.

Verhoef, A., Degreve, J., Huybrechs, B., Veen, H.V., Pex, P., Bruggen, B.V.D., 2008.Simulation of a hybrid pervaporation-distillation process. Comput. Chem. Eng.32, 1135–1146. https://doi.org/10.1016/j.compchemeng.2007.04.014.

Wang, S.J., Lee, H.Y., Ho, J.H., Yu, C.C., Huang, H.P., Lee, M.J., 2010. Plantwide designof ideal reactive distillation processes with thermal coupling. Ind. Eng. Chem.Res. 49, 3262–3274. https://doi.org/10.1016/j.compchemeng.2007.04.01410.1021/ie900786u.

Woods, D.R., 1983. Cost Estimation for the Process Industries. McMaster University,Hamilton, Ontario, Canada.