O
pen
A
rchive
T
OULOUSE
A
rchive
O
uverte (
OATAO
)
OATAO is an open access repository that collects the work of Toulouse researchers and
makes it freely available over the web where possible.
This is an author-deposited version published in :
http://oatao.univ-toulouse.fr/
Eprints ID : 15864
To link to this article : DOI : 10.1016/j.cherd.2016.01.026
URL :
http://dx.doi.org/10.1016/j.cherd.2016.01.026
To cite this version : You, Xinqiang and Rodriguez-Donis, Ivonne
and Gerbaud, Vincent Low pressure design for reducing energy cost
of extractive distillation for separating Diisopropyl ether and
Isopropyl alcohol. (2016) Chemical Engineering Research and
Design, vol. 109. pp. 540-552. ISSN 0263-8762
Any correspondence concerning this service should be sent to the repository
administrator:
staff-oatao@listes-diff.inp-toulouse.fr
Low
pressure
design
for
reducing
energy
cost
of
extractive
distillation
for
separating
diisopropyl
ether
and
isopropyl
alcohol
Xinqiang
You
a,b,c,
Ivonne
Rodriguez-Donis
b,c,
Vincent
Gerbaud
b,c,∗aDepartmentofChemistryandLaboratoryforAdvancedMaterialandGemstoneTestingCenterofECUST,East
ChinaUniversityofScienceandTechnology,Shanghai200237,China
bUniversitédeToulouse,INP,UPS,LGC(LaboratoiredeGénieChimique),4alléeEmileMonso,F-31432Toulouse
Cedex04,France
cCNRS,LGC(LaboratoiredeGénieChimique),F-31432ToulouseCedex04,France
Keywords:
Extractivedistillation Energysaving
Lowpressuredistillation Extractiveefficiencyindicator Optimization
a
b
s
t
r
a
c
t
Weshowhowreducingpressurecanimprovethedesignofa1.0-1amixturehomogeneous extractivedistillationprocessandweuseextractiveefficiencyindicatorstocomparethe optimalityofdifferentdesigns.Thecasestudyconcernstheseparationofthediisopropyl ether(DIPE)–isopropylalcohol(IPA)minimumboilingazeotropewithheavyentrainer 2-methoxyethanol.WefirstexplainthattheunexpectedenergycostOFdecreasefollowingan increaseofthedistillateoutputsisduetotheinterrelationofthetwodistillateflowrates andpuritiesandtheentrainerrecyclingthroughmassbalancewhenconsideringboththe extractivedistillationcolumnandtheentrainerregenerationcolumn.Then,wefindthatfor thestudiedcasealowerpressurereducestheusageofentrainerandincreasestherelative volatilityofDIPE–IPAforthesameentrainercontentintheextractivecolumn.A0.4atm operatingpressureisselectedtoenabletheuseofcheapcoolingwaterinthecondenser.We runanoptimizationoftheentrainerflowrate,bothcolumnsrefluxratios,distillatesand feedlocationsbyminimizingthetotalenergyconsumptionperproductunit.Doubledigit savingsinenergyconsumptionareachievedwhileTACisreducedsignificantly.Anextractive efficiencyindicatorthatdescribestheabilityoftheextractivesectiontodiscriminatethe desiredproductbetweenthetopandthebottomoftheextractivesectionoftheextractive sectioniscalculatedforcomparingandexplainingthebenefitofloweringpressureonthe basisofthermodynamicinsight.
1.
Introduction
Diisopropylether(DIPE) hasbecomeanimportantgasoline additive over the past decade and it is also widely used inmany other fields,suchastobacco productionand syn-theticchemistry(Lladosaetal.,2008).Isopropylalcohol(IPA) isextensivelyusedinpharmaceuticalindustryasachemical
∗ Correspondingauthorat.UniversitédeToulouse,INP,UPS,LGC(LaboratoiredeGénieChimique),4alléeEmileMonso,F-31432Toulouse
Cedex04,France.Tel.:+33534323651.
E-mailaddress:Vincent.Gerbaud@ensiacet.fr(V.Gerbaud).
intermediateandsolvent(Wangetal.,2008).IPAcanbe pro-ducedbyusingsolidacidorliquidacidascatalyticagent,with DIPEasacoproduct(LogsdonandLoke,2000).Theseparation ofDIPE/IPAisthekeydownstreamprocessthatdeterminesthe entireprocesseconomicbenefits.However,IPAandDIPE can-notbeseparatedbyconventionaldistillationprocessbecause theyformabinaryminimumboilinghomogeneousazeotrope.
(a)
F
EF
(A+B)
A(B)
Extractive section Rectification section Stripping section VL
TS+B
(A)
B(A)
V
L
T(b)
F
EF
(A+B)
A(B)
Extractive section Rectification section Stripping sectionV
L
TS+B
(A)
B(A)
V
L
TFig.1–Flowsheetoftypicalextractivedistillationwith(a)heavyentrainerand(b)lightentrainer. Extractive distillation is the common method for the
separationofazeotropicmixtureinlargescale productions (DohertyandMalone,2001;Petlyuck,2004;LuybenandChien, 2010). The relative volatility ofthe original components is alteredbyaddinganentrainerasitinteractsdifferentlywith the light and heavy components. For the separation of a minimum-(resp.maximum-)boilingazeotropicmixtureA−B byusingadirect(resp.indirect)splitconfigurationasshown inFig.1,thecommonbutnotexclusivethoughtisthatone shouldaddaheavy(resp.light)entrainerEthatformsnonew azeotrope.ThecorrespondingternarymixtureA−B−Ebelongs toSerafimov’sclass1.0-1a(Kivaetal.,2003),whichaccounts for21.6% ofall azeotropic ternarymixtures (Hilmen et al., 2002).
Thermodynamicinsightofextractivedistillationwas stud-iedby Laroche et al. (1991)through using the univolatility lineandvolatilityorder.Theyshowedthatoncethevolatility orderdiagramoftheternarymixtureazeotropeisobtained, the feasible separation sequence flowsheet is determined. A general feasibility criterion was then derived for batch extractivedistillation(Rodríguez-Donis etal.,2009a,b,2012; Rodríguez-Donis et al., 2012a,b). Later, it was extended to continuousmode byShen etal. who discussed the conse-quence on the process feasible ranges of reflux ratio and entrainer-to-feedflowrateratio, andtheinterconnectionof thecompositionprofileinthethreecolumnsection,rectifying, extractiveandstripping(Shenetal.,2013;ShenandGerbaud, 2013).Basedonthethermodynamicinsightanalysisofresidue curvemap, univolatilitylineand univolatilityorderregions oralternativelyofunidistributioncurvesandunidistribution order(Petlyuketal.,2015),thefeasibilityofextractive distilla-tionprocessescanbepredictedbyusingthegeneralfeasibility criterionwithoutanysystematiccalculationsofcomposition profiles.Inthispaper,thethermodynamicinsightofthe pro-cessfeasibilityfortheextractivedistillationincludesknowing whichcomponentwillbewithdrawninthefirstdistillatecut, whattheadequatecolumnconfigurationis,andwhetherthere issomelimitingoperatingparametervalueornot( Rodríguez-Doniset al.,2009a).Subsequentcalculationofcomposition profilescanthenhelprefinetherefluxratioandentrainerfeed flowraterange(Petlyuketal.,2015).
Accordingto Lladosaet al.(2007) the azeotropeof DIPE and IPA is very sensitive to pressure following the equi-librium diagrams of them based on experimental data. Therefore,theazeotropecanbeseparatedbypressureswing distillation(Luyben,2012;Luo etal., 2014).Theauthor also reported that 2-methoxyethanol is an excellent solvent to breakthe azeotropebasedon the vapor–liquid equilibrium experimentaldata. In2014, Luo et al.(2014) comparedthe
pressure-swing distillation and extractive distillation with 2-methoxyethanolfortheseparationofDIPEandIPAand per-formedafullheatintegrationoftheseprocesses.Theresults show thatthe fullyheat-integrated pressureswing distilla-tionsystemoffersa5.75%reductioninthetotalannualcost and7.97%savingsinenergyconsumptionascomparedtothe extractivedistillationprocess.
Weshowedinourpreviouswork(Youetal.,2015a,b)thata suitabledecreaseofthepressureinextractivedistillation col-umnforacetone–methanolminimumboilingazeotropewith waterallowedfordoubledigitsavingsinTACandenergycosts. However,inmanycases,thecostofpullingvacuumtoincrease therelativevolatilityofasystemovercomeswhatissavedby loweringthethermaldutythankstolowerboiling tempera-turesatlowerpressures.
In most extractive distillationprocess designstudies, a high product purity and recovery are sought (Luyben and Chien,2010;Luoetal.,2014;Youetal.,2015a,b)andasuitable distillateoutputflowrateisusuallyfoundbyrunninga sensi-tivityanalysisbasedonsimulationsinclosedloopflowsheet wheretheentrainerleavingtheregenerationcolumngoesto theextractivecolumnentrainerfeed(Youetal.,2015a,b).But severalliteratureworkshaveshownresultswhereonecannot getthetargethighpurityandrecoveryforAwhatevertheheat duty(Luoetal.,2014)orahighpurityforbothAandB prod-ucts(Giletal.,2009).Evidentlywiththeentrainerrecycling, distillate purities andflow ratesare coupledthroughmass balances.
Inthiswork,wederiveinSection2forthefirsttimesofar, therelationofthetwodistillatesintheextractivedistillation process to rationallyselect adequate purities and distillate flowratesandtoanticipatethemaximumimpuritycontent intheentrainerrecyclestream.Itallowsustolaterexplain thenon-intuitivebehaviorthattheenergycostdecreases fol-lowing the increaseofthedistillate output flows.Thenwe investigate in Section 3 whether reducing the pressure is worthfortheextractivedistillationoftheDIPE–IPAminimum boilingazeotropewith2-methoxyethanolheavyentrainer sys-temwiththesakeofreducingtheenergycostandTACforthe separationprocessitselfandwhetheritwouldbecompetitive withthebestliteraturedesignproposedbyLuoetal.(2014). OurmethodologyisshowninFig.2.Itappliestoall extrac-tivedistillationprocessfortheoftenencounteredseparation ofminimumboilingazeotropesABwithaheavyentrainerE wheretheunivolatilitycurve˛AB=1goestotheAEedge
(1.0-1a-m1extractiveseparationclassseparation).Itreliesupon usingthethermodynamic insightanalysisofresiduecurve, isovolatilitylineandternarymaptoselectpressure,andthen runningtwostepoptimizationprocedure(Youetal.,2015a,b)
Fig.2–Blockflowdiagramoftheproposedmethodology for1.0-1a-m1extractiveseparationclass.
toobtaintheoptimaldesignandevaluatingitscostbenefit. InSection4,wecomparethedesignresultswiththebest lit-eraturedesignsandanalyzethemwiththehelpofextractive efficiencyindicatorsofextractivesectionEextandperstagein
extractivesectioneextthatweproposedrecently(Youetal.,
2015a,b).FinallyweinvestigateinSection5thesensitivityof thedesignovertheextractivecolumnpressureandoverthe totalnumberoftrays.
2.
Steady
state
design
2.1. Interrelationofdistillatesintheextractive distillationprocess
Product purities and recovery are key specifications for at the designstepforany separationprocess. For the extrac-tivedistillationofanazeotropic mixtureAB, high purityA andhigh purityBcolumnsdistillates(resp.bottomliquids) areobtainedforadirect(resp.indirect)splitprocess configu-ration.Wederivenowtheinterrelationbetweenthem.Fora binaryazeotropicmixtureABseparatedinadirectsplit pro-cessconfigurationwithflowrateFandcontentxF,A,xF,B,the
Table1–Relationshipbetweendistillates,purityand recoveryforbinarymixturefor100kmol/hbinary mixture.
Feedcomposition Purityspecification
xF,A 0.75 xD1,A 0.98 xF,B 0.25 xD2,B 0.98 Rangeofdistillate (Eqs.(1)and(2)) D1(kmol/h)(Eq.(1)) 75.00( A=98%) 76.53( A=100%) D2(kmol/h)(Eq.(2)) 25.00( B=98%) 25.51( B=100%) Rangeofdistillate (Eqs.(3)and(4)) D1(kmol/h)(Eq.(4)) >76.53( B=93.877%) >75.00( B=94%) D2(kmol/h)(Eq.(3)) >25.51( A=99.3197%) >25.00( A=99.333%)
productpuritiesandprocessrecoveryarexD1,A,xD2,Band A, B,andobeytherelations
A=D1xD1,A FxF,A →D1= FxF,A A xD1,A (1) B=D2xD2,B FxF,B →D2= FxF,B B xD2,B (2)
But,wecanalsowritethemassbalanceforA(Eq.(3))andB (Eq.(4))anduseEqs.(1)and(2)toobtaintherelationinfluence ofonecomponentrecoveryontheotherdistillateflowrate:
D2= FxF,A−D1xD1,A 1−xD2,B−xD2,E = FxF,A−
Fx F,A A xD1,A xD1,A 1−xD2,B−xD2,E = FxF,A(1− A) 1−xD2,B−xD2,E> FxF,A(1− A) 1−xD2,B (3) D1= FxF,B−D2xD2,B 1−xD1,A−xD1,E = FxF,B− Fx F,B B xD2,B xD2,B 1−xD1,A−xD1,E = FxF,B(1− B) 1−xD1,A−xD1,E> FxF,B(1− B) 1−xD1,A (4)wherexD1,EandxD2,Earetheentrainercontentinthetwo distillates.
From theseequationsweobservethatthedistillateflow rate D2 (resp.D1)iscontrolledbytherecoveryandproduct
purityofB(resp.A)andbytherecoveryofA(resp.B).Soan unreasonablechoiceofdistillateflowratemayleadtopoor productqualityand/ortoenhancethedifficultyofthe sepa-ration,requiringmoreenergyandcosttoachievethesame productrecovery.Basedontheequationabove,wetestthe DIPE–IPAsystematlowpurityandrecovery.Theresultsare showninTable1.
FromTable1,weknowthat(1)followingEqs.(1)and(2), thevaluerangeofA-richdistillateD1andB-richdistillateD2
iseasytocalculatefrom Aand Brespectively,butitisnot
strictenough.Forexample,componentA’srecovery A(seeEq.
(3))hastobewithinaverynarrowvaluerangearound99.32%. OtherwiseD2cannotreachthereasonablerangevalue.
Like-wisefor B(seeEq.(4)).Therefore,forabinarymixtureofAB
with100kmol/handxF,A=0.75,ifthetwoproducts
specifica-tionsaresetat98%,weshouldchoose Abetween[99.3197%,
99.333%],whichwillensurethat Biswithin[98%,100%].
Theserelationsalsoindicate thatthechoiceofdistillate impactsthenecessaryentrainerfeedflowrateandits impu-ritycontent.Forexample,ifweassumeD1=75,thenequation
1gives A=0.98forapurityofxD1,A=0.98.Thustherewill be75×(1−0.98)=1.5kmol/hofAenteringthesecond regen-eration column. As D2=25kmol/h, the purity of B will be
nothigherthanxD2,B=(1−1.5/25)=0.94ifneglectthelossof Ainthe recycledentrainer streamW2. Onthe otherhand,
if wewanttoachieve xD2,B=0.98,themaximum valueofA that should reach in D2 is 0.5kmol/h (25×(1−0.98)). Sub-tractedfromthe1.5kmol/henteringtheregenerationcolumn, 1kmol/hAhastoberecycledwiththeentrainerrecyclestream
W2.Thatimpurity inthe entrainerrecycleimplies thatthe
entrainer flow ratehastobe over1000(1/(1−0.999))kmol/h in order to guarantee our specification of0.999mol% pure entrainerrecycledtotheextractivecolumn.Thiscorresponds to a very large entrainer-to-feed flow rate ratio (FE/F=10),
leadingtolargeenergycost.Amorereasonableexample is toincreaseD1 to 76.1kmol/h.ThenwithxD1,A=0.98 Eq.(1) gives A=0.99437and76.1×(1−0.0.99437)=0.428kmol/hof Awillenterthesecondcolumn.Beingbelowthemaximum 0.5kmol/hofAtoentertheregenerationcolumn,thepurities ofBandrecyclingentrainercanbemetandthe entrainer-to-feedflowrateratiodoesnothavetobeveryhigh,whichcould resultsinenergycostsavings.
2.2. Objectivefunctionandoptimizationapproach Ourobjectivefunctionisdefinedasfollows(Youetal.,2014)
minOF=Qr1+m·Qc1+Qr2+m·Qc2 k·D1+D2 subjectto: xDIPE,D1≥0.995 xDIPE,W1≤0.001 xIPA,D2≥0.995 x2-methoxyethanol,W2≥0.999 (5)
whereQr1:extractivecolumnreboilerduty,Qc1:extractive
columncondenserduty,Qr2:entrainerregenerationcolumn
reboilerduty,Qc2:entrainerregenerationcolumncondenser
duty,D1:distillateflowrateoftheextractivecolumn,D2:
dis-tillateflowrateoftheentrainerregenerationcolumn.Factors
kandmrespectivelydescribethepricedifferencesbetween productsDIPEand IPAand betweenthe condenser cooling andreboilerheatduties:k=3.9(productpriceindex),m=0.036 (energypriceindex).ThemeaningofOFisthe energy con-sumptionusedperproductunitflowrate(kJ/kmol).TheOFwe usedinthisworkaccountsfornotonlybothcolumns’energy demand,but alsoreflectstheweight coefficientofthetwo productpriceskandreboiler–condensercostpricem.Another advantageofOFisthatitcanreflectstheeffectsofthe vari-ablesinentrainerregenerationcolumnonthetotalprocess, asevidencedlatter.
Thisobjective function is used withthe two step opti-mizationmethodologypresentedinpreviouswork(Youetal., 2015a,b).First,inopenloopflowsheetwithnorecycleofthe entrainerandfreshentrainerfeedtoeaseconvergence,Aspen plus’sequentialquadraticprogramming(SQP)methodisused fortheoptimizationoftheprocessoveranenergy consump-tionobjectivefunction,underpurityandrecoveryconstraints andagivencolumnstructure,bymanipulatingthecontinuous variables:columnrefluxesR1,R2andtheentrainerflowrate
FEforthechoiceoftwodistillatesthroughcomparingOFwith
Luo’sdesignasinitialvaluesforothervariables.Secondly,a sensitivityanalysistoolisperformedtofindoptimalvalues ofthefeedtraylocationsNFE,NFAB,NFReg,whileSQPisrun
foreach set ofdiscrete variablevalues.Thefinal optimiza-tionisfoundthroughminimizingOFvalueanditisvalidated byrerunningthesimulationincloseloopflowsheetwhere theentrainerisrecycledfromtheregenerationcolumntothe extractivecolumnandanentrainermake-upfeedisadded. Theclose-loopsimulationrequiresadjustingtherefluxratio ifnecessary inorderto overcomethe effectofimpurity in recyclingentraineronthedistillatepurity.Finally,theTACis calculatedtocomparetheseparationsequences,similarlyto whatwasdoneinYouetal.(2015)byconsideringa3years paybackperiodisandbyusingDouglas’cost formulaswith MarshallandSwift(M&S)inflation2011inde xequalsto1518.1
(Youetal.,2015a,b).Theenergycostofthereboileris3.8$per GJ,afterconsultingachemicalcompanyinChongqingChina. An alternative optimization approach less sensitive to convergence issues hasbeen usedby usin another paper (Youetal.,2015a,b).Couplinganexternalstochasticgenetic algorithm with the process simulator, it proved suitable to optimize directly the closed loop configuration with all discreteandcontinuousvariablesanddoingsowitha multi-objective criterion, provided that the initial population of process designsets of parameters had enough individuals withoutconvergenceproblems.
3.
Process
analysis
3.1. Pressuresensitivityoftheazeotropicmixture With the purpose of changing the operating pressure to improve the extractive distillation sequence, we report in
Table2theDIPE–IPAazeotropiccompositionchangewith pres-sure.WeusedthesameVLEmodelastheoneusedtocompute theresiduecurvemapinFig.3.
Table 2 shows that the DIPE/IPA azeotrope is sensitive enoughtopressurechange.ThemixtureexhibitsaBancroft point(ElliottandRainwater,2000)near5.8atm witha boil-ingtemperaturecloseto134.8◦C,wheretheirvolatilityorder
isreversed.Consideringthelargepressuresensitivityofthe azeotrope,thePSDprocessislikelyfeasibleforthemixture. Since this was investigatedby Luo et al. (2014) we donot considerthisprocesshere.However,Table2alsoshowsthat thecontentofDIPEintheazeotropicmixtureincreaseswhen the pressure decreases.AsseeninFig. 3forP1=1atm and
P1=0.4atmthisalsopromptstheunivolatilitycurveto
inter-sect the A–E edge nearer the DIPE vertex. As cited in the literaturesurveyanddiscussedinthenextsection,itmeans thatless entrainerisneededtobreaktheazeotrope,which couldreducethecapitalcost.Besidesaloweroperating pres-sureimplieslowerboilingtemperaturesandpossibleenergy costsavings.Inparticular,ifweassumethattheextractive columndistillateisalmostpureDIPEandconsidera conser-vativevalueof40◦Castheminimumallowedtemperaturein
thecondensertousecheapcoolingwater,0.4atmisthe low-estpressurefromTable2sincetheDIPEboilingpointisthen computedat42.5◦C.
3.2. Thermodynamicinsightanalysisofextractive processfeasibility
The separation of the minimum boiling azeotropes DIPE A(68.5◦C)–IPAB(82.1◦C)(x
azeo,A=0.78 @66.9◦C)withheavy
entrainer2-methoxyethanolE(124.5◦C)belongstothe
1.0-1a-m1extractiveseparationclass(GerbaudandRodriguez-Donis, 2014). The univolatility curve ˛AB=1 intersects the binary
sideA–EasshowninFig.3displayingalsotheresiduecurve map and otherisovolatility lines at1atm and 0.4atm.The vapor–liquidequilibriumofthesystemisdescribedwiththe nonrandom two-liquid (NRTL) (Renon and Prausnitz, 1968) thermodynamicmodel,whilethevaporphaseisassumedto beideal.Thebinary parametersofthemodelarethesame asLuoetal.(2014)whogotthevaluesbyregressingfromthe experimentaldata(Lladosaetal.,2007).
FromFig.3,componentA(DIPE)isaresiduecurvemap sad-dle[Srcm]andcannotbedistilledbyazeotropicdistillation.But
544
chemicalengineeringresearchanddesign 109 (2016) 540–552Table2–DIPE–IPAazeotropictemperatureandcompositionatdifferentpressureswithNRTLmodel.
P(atm) TbDIPE(◦C) TbIPA(◦C) Tbazeo(◦C) AzeotropeDIPEmolfraction
10.0 161.8 155.72 154.6 0.2957 5.0 128.1 129.5 123.7 0.5551 4.0 118.4 121.9 114.6 0.6004 2.5 99.7 107.1 96.9 0.6732 1.0 68.5 82.1 66.9 0.7745 0.8 61.7 76.6 60.4 0.7950 0.6 53.4 69.8 52.4 0.8200 0.5 48.4 65.7 47.5 0.8351 0.4 42.5 60.9 41.8 0.8531
Fig.3–ExtractivedistillationcolumnconfigurationandDIPE–IPA–2-methoxyethanol.Class1.0-1aresiduecurvemapat 1atmwithisovolatilitycurvesat0.4and1atm.
productthankstotheoccurrenceofanextractivesectionin thecolumnbyfeedingtheentrainerFEatadifferentlocation
thanthemainfeedFAB.
Fromthermodynamicinsight,theregionABEinFig.3 sat-isfiesthegeneralfeasibilitycriterion(Rodríguez-Donisetal., 2009a),sowecanexpectthat(1)DIPEwillbethedistillatein theextractivecolumnasAisthemostvolatilecomponentin theregionandthereisaresiduecurvelinkedAEfollowingthe increaseoftemperature.Therefore,thecolumnconfiguration isadirectsplit(Fig.1a).(2)Thereisminimumentrainer-to-feed flowrateratio(FE/F)minthatguaranteestheprocessfeasibility.
WhenFE/Fislowerthan(FE/F)min,theprocessforachievinga
highpuritydistillateisimpossiblebecausethestablenodeof extractivesectionSNextislocatedonthe˛AB=1curve.Above
thatamount,therelativevolatility˛ABisalwaysgreaterthan
one.Indeed,theintersectionpointxPoftheunivolatilitycurve
andthetriangleedgeAEgivesustheinformationtocalculate the(FE/V)minbythemethodshowninLelkesetal.(1998)and
thentranslate(FE/V)mininto(FE/F)minbyusingthetransferring
equation(Youetal.,2015a,b).
Fig.3alsodisplaystheisovolatilitylines˛AB=1,˛AB=2and
˛AB=4at0.4atm. Theydemonstrate clearly that the
sepa-rationiseasierinlowpressuresincemuchlessentraineris neededtoachievethesamevolatilityofAB.Therefore, lower-ingpressureislikelyapotentialwaytoreduceenergycostand TACforthestudiedsystem,butadetaileddesignisnecessary toassessthebenefitofloweringpressure.
Finally,wehavedrawninFig.4theliquid–vapor equilib-riumcurveforthebinarymixtures2-methoxyethanol–DIPE. Itexhibits apinch pointnear pureDIPE whichremains at lowpressure.Thishints that asignificant number oftrays arenecessary intherectifyingsection toreachhigh purity
DIPE.Thispointisfurtherprovedbyourdesigninthenext section.
4.
Results
and
discussions
Aiming atsaving energycost forthe extractivedistillation sequence of DIPE–IPA with 2-methoxyethanol, the operat-ingpressureofextractivecolumnisadjustedtoP1=0.4atm,
which givesatop temperaturenear42.5◦Cenabling touse
cheapcoolingwateratcondenser.IntendingtorevisitLuo’s
chemicalengineeringresearchanddesign 109 (2016) 540–552
545
Table3–Relationshipbetweendistillates,purityandrecoveryforDIPE–IPA.
Feedcomposition Purityspecification
x1 0.75 xD1,A 0.995
x2 0.25 xD2,B 0.995
Rangeofdistillate(Eqs.(1)and(2))
D1(kmol/h)(Eq.(1)) 75.00( A=99.5%) 75.37( A=100%)
D2(kmol/h)(Eq.(2)) 25.00( B=99.5%) 25.12( B=100%)
Rangeofdistillate(Eqs.(3)and(4))
D1(kmol/h)(Eq.(4)) >75.37( B=98.493%) >75.00( B=98.50%)
D2(kmol/h)(Eq.(3)) >25.12( A=99.8325%) >25.00( A=99.833%)
design,wekeep hisproposed totalnumber oftrays inthe extractivecolumn(Next=66) andinthe entrainer
regenera-tion column(NReg=40). Wealsouse the same feed as Luo
etal.(FAB=100kmol/h,0.75(DIPE):0.25(IPA))and preheatthe
entrainerto328.15Kastheydid.Butweselecthigherproduct purityspecificationsequalto0.995molarfractionsforboth DIPEandIPAwhereastheyobtained0.993forDIPEand0.996 forIPA.Otherdesignparametersarefoundbyusingtwosteps optimizationprocedure.Luo’sdesignat1atmisusedasthe initialpointoftheoptimizationthatwerunat0.4atm.
InthecalculationofTAC,theheatexchangerannualcost forcoolingrecyclingentraineristakenintoaccountinorderto emphasizetheeffectoftherecyclingentrainerflowrateonthe process.Thepressuredropofpertrayisassumed0.0068atm, thesame asLuo etal.(2014) andthetrayefficiencyis85% (Figueiredoetal.,2015)forcalculatingcapitalcostofthe col-umn.
Itshouldbenoticedthatavacuumpumpisneededto pro-ducelowpressureinextractivecolumnwhiletheprocessis started,afterthattheoperatingpressureofthecolumnis con-trolledbythe condenserheatduty.Theelectricitycostand capitalcostofthevacuumpumpismuchlowerthanthatof thenormalliquid-conveyingpumpbecauseitonlyworks a littletimeduringeachstart-upanditsvapor-conveyingflow (thevolumeoftheextractivecolumn)ismuchsmallerthan thatofthenormalliquid-conveyingpump(totalliquidflowof recyclingentrainer,feedandproductforoneyear).Sincethe costofliquid-conveyingpumpcouldbeneglectedinthe con-ceptualdesignstage,thecostofvacuumpumpshouldnotbe takenintoaccount.Theothercostssuchaspipesandvalves arealsoneglected.
4.1. Choiceofdistillateflowrate
Havingsetnewproductpurityspecificationsat0.995,weuse Eqs.(1)–(4)toselectsuitabledistillateflowrate,asshownin
Table3.
ThroughtheuseofEqs.(3)and(2),Table3showsthatthe recoveryofA (DIPE)shouldbeatleastgreater than 99.83% andthattherecoveryofB(IPA)shouldbeabove99.5%.Our finaldesignresultsshownlaterovercomethesevalues.This tablealsoexplainwhyLuo’designthatfixedD1=75.44kmol/h
preventedhimtoreachapurityof0.9950forDIPEsinceitis abovethe75.37kmol/hlimitvalueforthatpurity.Heobtained 0.9930whichcorrespondstoa A=99.88%fromEq.(1).
AsideeffectofthestronglynonlineardependencyofD1
andD2ontheproductpurityisthatthesimulation cannot
beconvergedsteadilywhenwedirectlytreatthedistillates asanoptimizedvariableintheSQPmethod.SoD1isvaried
withadiscretestepof0.05kmol/hfrom75kmol/h( A=99.5%)
to75.35kmol/h( A=99.96%)andD2isvariedwithadiscrete
stepof0.03kmol/hfrom25kmol/h( B=99.5%)to25.12kmol/h
Fig.5–EffectsofD1andD2onOFwithD1,D2,FE,R1andR2
asvariables.
( B=99.98%),andtheSQPoptimizationisruntoobtainFE,R1
andR2.TheresultsshowninFig.5displaythevalueofthe
objectivefunctionOFvs.D1andD2.
FromFig.5,weobservethatOFdescribingthetotalenergy demand per product flow rate decreases quicklywith the increaseofD1.Thesamestatementwasmadeinourprevious
studyfortheextractivedistillationoftheacetone–methanol azeotropicmixturewithwaterasaheavyentrainer.Alsothis phenomenoniscounterintuitive:normally,themoreproduct atspecifiedpurityisobtained,themoreenergy(reboilerduty) shouldbeused.ButasexplainedinSection2.4andevidenced inFig.5,lessproduct(e.g.adistillateD1of75kmol/h)induced
alargeentrainerfeed–mainfeedratioincolumn1tomeet thepurityspecifications.Besides,seekingalargerD1DIPE-rich
distillateincreasestheseparationdifficultyintheextractive columnbuteasesthe separationintheentrainer regenera-tion column asless DIPE impurity entersit, hereinducing global energysavings. Obviously,our objective functionOF accountingforbothcolumnscandescribequantitativelythe trade-offbetweenthetwocolumns.Meanwhile,Fig.5shows thatanincreaseinD2increasestheOFvalue.Noticethatfor
D1lowerthan75.1kmol/h,theOFvalueisnotdisplayedasitis
wellabove45,000kJ/kmolbecauseoftoomuchDIPEentering regenerationcolumn,leadingtohighenergycost.Thispoint agreeswiththerelationoftwodistillate mentionedbefore. ThusD1lowerthan75.1kmol/hcannotbeconsideredfurther
asasuitablevalue.
As we will optimize other variables such as NFE, NFF,
NFReg, Next and NReg in the subsequent steps, we select
D1=75.35kmol/handD2=25kmol/h;thosevaluescorrespond
toa productrecoveryof99.96%forDIPE-rich distillate and 99.5%forIPA-richdistillate,respectively.Thecorresponding
546
chemicalengineeringresearchanddesign 109 (2016) 540–552Table4–OptimizedvaluesofFE,R1,andR2forthe
extractivedistillationofDIPE–IPAwith
2-methoxyethanol,caseAat1atmandcaseBat0.4atm.
CaseA CaseB P1(atm) 1 0.4 P2(atm) 1 1 N1 66 66 N2 40 40 D1(kmol/h) 75.35 75.35 D2(kmol/h) 25 25 NFE 30 30 NFAB 56 56 NFReg 10 10 FE(kmol/h) 96.9 84.6 R1 1.85 1.70 R2 1.80 1.33 OF(kJ/kmol) 37174.9 30639.9
OFvalueis37174.9kJ/kmol,withFE=96.9kmol/h,R1=1.85and
R2=1.80atP1=0.4atm.
4.2. ContinuousvariablesFE,R1andR2
TheoptimizedFE,R1,R2valuesareshowninTable4whilethe
othervariablesarekeptconstant.Noticethattheinitialvalues ofthethreefeedlocationsincaseAandcaseBaretakenfrom Luo’sdesign.
Weobserve(1)underlower pressure(caseB),thevalues ofFE,R1,R2incaseBarelowerandOFdecreasesby17.6%.It
materializesthe benefitofthe pressurereductionthatwas anticipated throughtheanalysisofresidue curvemap and univolatility line for a1.0-1a class mixture. (2) The reduc-tioninR2ismorepronouncedthatinR1.Iflessentraineris
fedtothe extractivecolumn,a greaterR1 isneededtoget
thesameseparationeffect.Meanwhiletheconcentrationof entrainer fedtothe regenerationcolumndecreases due to massbalance,andlessenergy(R2decrease)isusedto
recy-cletheentrainer.Overall,theseobservationsagreewiththe studyfortheseparationofacetone-methanol-watersystem (Youetal., 2015a,b).These resultsalsoshow thebenefitof optimizingbothcolumnstogethertoimprovefurthertheOF. 4.3. Selectingsuitablefeedlocations
Thesensitivity analysisover the three feed positions with ranges [20;40] forNFE, [>NFE; 66]for NFAB, [5;40] forNFReg
wasmade byusingexperimentalplanning proceduresoas toavoidlocalminimum.Foreachsetofvalues,FE,R1,R2are
optimizedwhileD1andD2arefixed.Table5showstheresults
consideringnowP1=0.4atmandP2=1atmasincaseBabove.
Asawhole,theshiftingofthethreefeedlocationsimproves furthertheOFvalueby1.9%butalsoimpactstheprocess effi-ciencythatwillbediscussedinSection4.5.Afewstatements canbemadefromTable5:(1)OFismoderatelysensitiveto smallchangesofthethreefeedlocationswhenFE,R1,andR2
areoptimized.(2)Inconventionaldistillationanincreaseof thenumberoftraysintherectifyingsectionallowstousea lowerrefluxratioR1,but hereforanextractivecolumn,the
oppositeisfoundasfeedlocationofentrainermovesupthe columnfrom30to28forthelowestOFdesign.Thereasonwas statedinLelkesetal.(1998):toomuchtraysintherectifying sectionisnotrecommendedinextractivedistillationforthe 1.0-1aclassbecausethepureproductDIPEisasaddleinthe RCMandlengtheningtherectifyingcompositionprofilewill
Table5–OpenloopoptimalresultsofFE,R1,R2,NFE,
NFAB,NFRegunderfixedD1andD2fortheextractive
distillationofDIPE–IPAwith2-methoxyethanol, P1=0.4atmandP2=1atm.
NFE NFAB NFReg FE R1 R2 OF(kJ/kmol)
26 58 13 75.1 1.76 1.20 30292.8 27 59 14 75.0 1.75 1.17 30081.9 28 58 13 75.1 1.76 1.17 30144.8 28 59 15 74.1 1.77 1.16 30110.5 28 60 15 75.0 1.74 1.17 30042.9 28 60 16 74.0 1.77 1.16 30122.2 28 61 15 74.0 1.78 1.15 30213.2 29 55 9 80.0 1.73 1.37 30827.4 30 56 10 84.6 1.70 1.33 30639.9 30 57 11 84.4 1.66 1.34 30410.7 31 55 10 86.5 1.67 1.39 30740.8 31 56 11 86.6 1.65 1.38 30619.0
approachinsteadtheunstablenodethatshouldlienearthe unwantedazeotrope.(3)TheminimumvalueofOFisfound withsixextranumberoftraysintheextractivesectionthan Luo’sdesign.Asdiscussedearlier,akeyfactortoachievethe extractiveseparationistoreachtheextractivesectionstable nodeSNextneartheDIPE–2-methoxyethanoledge.Thiscanbe
achievedbymoretraysintheextractivesection.Another con-sequenceisthattheefficiencyoftheextractivesectionthat willbediscussedinSection4.5isimproved.(4)Thelowest energycostforper unitproductOFis30042.9kJ/kmolwith
NFE=28,NFAB=60andNFReg=15.Itrepresentsafurther1.9%
reductioninenergycostcomparedthedesignwithNFE=30,
NFAB=56andNFReg=10.
4.4. Closedloopdesignandoptimaldesignparameters Theoptimaldesignisre-simulatedinclosedloopflowsheet inordertomakesuretheproductpurityisachievedandthat theeffectofimpurityinrecyclingentrainerontheprocessis accountedfor.Formoredetailedinformation,seeourearlier study (You et al., 2015a,b). Case2under P1=0.4atm
corre-spondstotheopenloopdesignofcaseB.Itiscomparedtocase 1thatcorrespondstoourbestclosedloopdesignunderP1=1
atm.IndeedcaseA(P1=1atm)inTable4thatcorrespondedto
adesignwithLuo’sfeedlocationsdidnotallowedunderclosed loopsimulationtoreachthetargetedpurityforthedistillates andcannotbeusedforcomparison.
Thedesignandoperatingvariablesandthecostdataare shown inTable 6, referring to the flow sheet notations in
Fig.3.Table7displaystheproductpurityandrecovery val-ues. Figs. 6 and 7 show the temperature and composition profilesoftheextractiveandentrainerregenerationcolumns forthebestcase,namelycase2,whereMEnotationholdsfor 2-methoxyethanol.
Table6confirmsthataprocessdesignoperatingat1atm likecase1canbeimprovedfurther,whilekeepingthesame numberoftraysintheextractiveandregenerationcolumns.In summary,(1)theentrainerflowratedecreaseddrasticallyby 25%from100kmol/hincase1to75kmol/hincase2,showing thatimprovementwaspossibledueprincipallytoa combi-nationofashiftoffeedlocationsandpressurereduction,as deducedfromthepressuredependencyoftheazeotrope com-positionandisovolatilitycurves.(2)Theenergyconsumption underlyingtheOFvalueincase2isreducedby11.2%and13.4% comparedtocase1and caseLuo,respectively. Itismostly attributedtoareductioninthe entrainerflowrateallowed
chemicalengineeringresearchanddesign 109 (2016) 540–552
547
Table6–OptimaldesignparametersandcostdatafromclosedloopsimulationfortheextractivedistillationofDIPE–IPA with2-methoxyethanol.
Column CaseLuoa Case1 Case2
C1 C2 C1 C2 C1 C2 Next 66 66 66 P1(atm) 1 1 0.4 FAB(kmol/h) 100 100 100 FE(kmol/h) 100 100 75 D1(kmol/h) 75.44 75.35 75.35 NFE 30 29 28 NFAB 56 59 60 R1 1.54 1.86 1.83 QC(MW) 1.553 1.746 1.838 QR(MW) 2.279 2.205 1.948 NRreg 40 40 40 P2(atm) 1 1 1 D2(kmol)/h 25.03 25 25 NFReg 10 12 15 R2 1.93 1.88 1.18 QC(MW) 0.823 0.811 0.614 QR(MW) 0.748 0.737 0.658 Column C1 C2 C1 C2 C1 C2 Diameter(m) 1.44 0.78 1.37 0.69 1.66 0.65 Height(m) 46.33 28.05 46.33 28.05 46.33 28.05 ICS(106$) 0.720 0.251 0.683 0.220 0.838 0.207 AC(m2) 66 23 74 22 164 18 AR(m2) 116 38 112 38 99 34 IHE(106$) 0.347 0.172 0.354 0.170 0.443 0.151 Costcap(106$) 1.184 0.449 1.145 0.412 1.426 0.378 Costope(106$) 0.938 0.313 0.911 0.308 0.809 0.274 CostCA(106$) 1.333 0.462 1.293 0.445 1.284 0.399 QHA(MW) 0.407 0.407 0.305 CostHA(106$) 0.016 0.016 0.013 TAC(106$) 1.810 1.754 1.696 OF(kJ/kmol) 35098.7 34245.2 30398.9
a Luo’sprocessdistillatemolarpurityforDIPE-richproductisbelowourspecifications.
Table7–ProductpuritiesandrecoveriesforcaseLuo,cases1and2designs.
Molefraction D1 D2 W2 W1 Recovery
CaseLuo@1atm DIPE 0.99350 0.00202 4.67E−27 0.00041 99.93%
IPA 0.00044 0.99637 0.00028 0.20044 99.64%
ME 0.00606 0.00161 0.99972 0.79915
Case1 @1atm DIPE 0.99504 0.00096 5.67E−26 0.00019 99.97%
IPA 0.00076 0.99772 0.00034 0.20033 99.78%
ME 0.00420 0.00132 0.99966 0.79948
Case2@0.4atm DIPE 0.99501 0.00105 1.39E−23 0.00026 99.96%
IPA 0.00090 0.99730 0.00042 0.25044 99.73%
ME 0.00409 0.00165 0.99958 0.74930
bythereduced pressure.(3)Meanwhile, TAC savingsreach 3.4%mainlyduetothe decrease oftheentrainer flowrate andrefluxratiointhesecondregenerationcolumn.(4)There arealsodrawbacksofloweringpressureexceptthenegligible costofthevacuumpumpmentionedabove.Ifwecompare theextractivecolumnincase1andcase2,thedecreaseof thepressureleadstoa21.2%increaseofthecolumn diame-ter,anda2.2timesincreaseofthecondenserheatexchanger areaduetothedecreaseofcondensertemperature.However, thedecreaseofthepressureresultsinthedecreaseofreboiler dutyby11.6%.Asaresult,thebenefitofdecreasingpressure overcomesthepunishmentbecausetheannualcostof extrac-tivecolumnisslightlyreducedfrom1.293to1.284(106$per
year).
RecallinTable7, Luo’sdesigndoesnotmeetour molar purity specification on the DIPE distillate. To meet that
specification,case1resultshowsthatR1 mustbeincreased
andtheheatdutyaswell.Despiteallthis,Table6showsthat case1 givesa3.1%and 2.4%reductionin TACand energy cost OFover Luo’sdesignatthesame pressure dueto the decrease ofreboiler dutythankstomoresuitable distillate andfeedlocations.Thisprovestheeffectivenessofthetwo stepoptimizationprocedureweused.
Table6alsoshowsthatshiftingthefeedtraylocationsand runninganewoptimizationimprovestheOFandthusreduces theprocessenergyconsumption.Thispointisevidencedby comparingthetotalreboilerheatdutyofcase1(2.942MW)and case2(2.606MW):a11.4%totalheatdutyissaved.Regarding productpurityandrecovery,Table7showsthatpuritytargets aremetandthatrecoveriesarehighandinagreementwiththe relationshipbetweenthetwodistillateflowratesasdiscussed inSection2.1.
Fig.6–Temperatureandcompositionprofilesofcase2extractivecolumnfortheextractivedistillationofDIPE–IPAwith 2-methoxyethanol.
4.5. Analysisfromextractiveefficiencyindicatorsand ternarycompositionprofilemap
Shiftingfeedtraylocationsinduceschangesinthenumber oftrays in each section. We already discussed the impact ofthenumberoftraysintherectifyingsection.Butforthe extractivedistillationprocess,thegeneralfeasibilitycriteria (Rodríguez-Donisetal.,2009a;Youetal.,2015a,b)andthe fea-sibilityanalysiswehaveconductedinSection2.1showthat thenumberoftraysintheextractivesection,Next,shouldbe
highenoughtoallowthecolumncompositionattheentrainer feedtraytolieclosetothestablenodeoftheextractive pro-filemap.Inourprevious workonanother 1.0-1aextractive separationclass mixture,weproposed extractiveefficiency indicatorsdescribingtheabilityoftheextractivesectionto dis-criminatethedesiredproductbetweenthetopandthebottom oftheextractivesection.
Eext=xP,H−xP,L (6)
eext= Eext
Next (7)
where Eext: the total extractive efficiency indicator of
extractivesection,xP,H:productmolefractionatoneendof
extractivesection,xP,L:productmolefractionatanotherend
ofextractivesection,eext:extractiveefficiencyindicatorper
tray,Next:traynumberofextractivesection.Here,weusethe
entrainerfeedandthemainfeedtrayslocationsasendsofthe extractivesection.
TheextractiveefficiencyindicatorsEextandeextareshown
inTable8forthecases1and2andLuo’sdesign,alongwiththe DIPEcompositionatthefeedtrays.Fig.8displaysthe extrac-tivecolumncompositionprofilesinaternarymap.
From Table8andFig.8,weremarkthat(1)Luo’sdesign extractiveefficiencyindicatorsarenegative,statingthatthe extractivesectionisnotabletoimprovetheDIPEpurity.(2)A lowerentrainerflowrateasincase2operatingatlow pres-sure allows the top end ofthe extractive section SNext to
Fig.7–Temperatureandcompositionprofilesofcase2entrainerregenerationcolumnfortheextractivedistillationof DIPE–IPAwith2-methoxyethanol.
indicatorsareslightlyhigherforcase2thanforcase1.Indeed the lower operating pressure leads to more favorable iso-volatilitycurvesandSNextlocationasshowninFigs.3and8,
whichdirectlyaffecttheextractiveefficiencyindicatorsEext
andeext.
ComparedwithLuo’sdesign,bothcase1andcase2show positiveandevidentlyhigherextractiveefficiencyindicators. Case2withalowerTAChasahigherefficiencythancase1. Butwecannotconcludethatthedesignwithamoresuitable extractiveefficiencyindicatorismoreeconomicalandwehave
shownthatthereexistsanoptimalextractiveefficiency indi-catorvaluecorrespondingtothelowestTACdesigninanother work(Youetal.,2015a,b).
5.
Design
sensitivity
to
the
extractive
column
pressure
and
the
total
number
of
trays
Inthispart,wedoasensitivityanalysisoftheprocessdesign byovertheextractivecolumnpressureandoverthetotal num-beroftrays.
Table8–EfficienciesofpertrayandtotalextractivesectionfortheextractivedistillationofDIPE–IPAwith 2-methoxyethanol.
DIPEcomposition Next Eext/10−3 eext/10−3
Entrainerfeedtray(SNext) Feedtray
CaseLuo 0.463 0.492 27 <0 <0
Case1 0.501 0.393 31 108 3.48
Fig.8–LiquidcompositionprofilesforcaseLuo,cases1 and2extractivedistillationcolumndesignsforthe extractivedistillationofDIPE–IPAwith2-methoxyethanol.
Pressureisonlyvariedforthefirstextractivedistillation columnwithvaluesbetween1atmandalowerlimitequalto 0.4atmthatweselectedforusingcoolingwaterforthe con-denserkeepingthesametotaltraynumber.Thetotalnumber oftraysisincreasedonebyonefrom66to70afterselecting themostsuitablepressure,namely0.4atm.
The two step nonlinear programming optimization methodologywithSQPmethodisrunforeachcase.Carrying afulloptimizationofthepressure and traynumberatthe sametimethantheothervariableswouldrequiresolvinga mixed-integernonlinearprogramming(MINLP)problemwith adequateoptimizationschemesthatwehavenotconsidered here.ComparisonisdoneontheOFvalueonly,namelythe energydemandperproductflowrateandTAC.Forreasons givenearlier,wedonotaccountforadditionalcostsatlow pressureliketechnologicalpricefactorincreaseandvacuum pumpandsub-cooledfluidcosts.
5.1. Selectionofthepressurefortheextractivecolumn Resultsofthesensitivitytopressureoftheextractivecolumn ontheprocessOFandTACvaluesaredisplayedinFig.9.Lines areshownforeyeguidance.
FromFig.9,thelowerboundofthepressurerangestudied, 0.4atmgivesthedesignwiththelowestOFandTACvalues whilekeepingthesametotaltraynumber.First,thisagrees wellwiththermodynamicsinsightanalysisdisplayedinFig.3: therelativevolatilityincreasesquicklyatlowerpressureand theboilingtemperatureislower,whichleadtogethertouse lessenergytoachievethesameseparationintermsofpurity andrecovery.Second,theprocessTACdecreaseatlower pres-sureismainlyduetothedecreaseoftheprocessenergycost. Third,TACdecreaseslittlefrom0.5atmto0.4atmisbecause thebenefitofloweringpressureiscrippledbytheincreaseof capitalcostforcondenserduetothedecreaseoftemperature drivingforce.
5.2. Selectionofthetotaltraysnumberforthe extractivecolumn
Fig.10displaystheOFandTACvaluesforthetotaltraynumber oftheextractivecolumn,andTable9showsthedesignatthe
Fig.9–EffectofpressureonenergycostOFandTAC.
TAC minimumvalue.Thelowerbound,66trays,istheone usedintheprevioussections,andwastakenfromLuo’sdesign (Luoetal.,2014).
For aconstant pressure,wefindthattheprocessdesign wouldbenefitfromanincreaseoftheextractivecolumn num-beroftrayupto68asbothOFandTAC decrease.TheTAC decreaseisduetocapitalcostdecreasefromthereductionof columndiameterand reboiler-condenserheattransferarea thankstoalowervaporflowrate,andalsoduetothe operat-ing costdecrease asthelower refluxallowsasmallerheat duty tobeused. For these valuesit compensatesfully the capitalcostincreaseduetothehighercolumnheight.Then fortraynumbersequalto69and70,therefluxremains sim-ilartothe68traycasevalue.Theoperatingcostsaresimilar but the extra capitalcost ofthe columnheight dominates again and theTAC increases. Inthe meanwhile, correlated totheenergydemand,theOFvalueremainssimilartothat forNext=68.Hence,oneshouldprefertodesignwith68trays
inthe extractivedistillationcolumnwhenpressure iskept at0.4atm.
Fig.10–Effectoftotaltraynumberofextractivecolumnon energycostOFandTAC.
chemicalengineeringresearchanddesign 109 (2016) 540–552
551
Table9–EffectofNextonthedesignparametersandcostdatafromclosedloopsimulationfortheextractivedistillation
ofDIPE–IPAwith2-methoxyethanol.
Next 67 68 69 Column C1 C2 C1 C2 C1 C2 P1(atm) 0.4 0.4 0.4 FAB(kmol/h) 100 100 100 FE(kmol/h) 76 77 75 D1(kmol/h) 75.35 75.35 75.35 NFE 29 31 31 NFAB 61 62 63 R1 1.75 1.70 1.71 QC(MW) 1.786 1.754 1.760 QR(MW) 1.901 1.873 1.874 NRreg 40 40 40 P2(atm) 1 1 1 D2(kmol/h) 25 25 25 NFReg 14 13 13 R2 1.25 1.27 1.27 QC(MW) 0.634 0.639 0.639 QR(MW) 0.677 0.682 0.679 Column C1 C2 C1 C2 C1 C2 Diameter(m) 1.63 0.66 1.62 0.66 1.62 0.66 Height(m) 46.94 28.05 47.55 28.05 48.77 28.05 ICS(106$) 0.831 0.210 0.834 0.210 0.851 0.210 AC(m2) 159 18 156 18 157 18 AR(m2) 97 35 95 35 95 35 IHE(106$) 0.435 0.154 0.430 0.155 0.431 0.155 Costcap(106$) 1.408 0.385 1.407 0.385 1.429 0.385 Costope(106$) 0.790 0.282 0.778 0.284 0.778 0.282 CostCA(106$) 1.259 0.410 1.247 0.412 1.255 0.411 QHA(MW) 0.309 0.313 0.304 CostHA(106$) 0.013 0.014 0.013 TAC(106$) 1.681 1.672 1.677 OF(kJ/kmol) 30069.1 29801.1 29785.1
6.
Conclusions
Aimingatevaluatingthebenefitofloweringpressurefor sav-ingenergycostandtotalannualcost,wehaveoptimizedthe designofahomogeneousextractivedistillationprocess for theseparation ofthe DIPE–IPAminimumboilingazeotrope with heavy entrainer 2-methoxyethanol. The process flow sheetincludesboththeextractivedistillationcolumnandthe entrainerregenerationcolumn.Thecuriousbehaviorthatthe energycostOFdecreasesfollowingtheincreaseofdistillate areclarifiedbytheinterrelationofthetwodistillateflowrates andpurities andthe entrainer recyclingthroughmass bal-ance.
OurmethodologystatedinSection1isillustratedwiththe studiedcase.Byusinginsightfromtheanalysisoftheternary residuecurvemapandisovolatilitycurves,wehavenoticed thebeneficial effectofloweringthe pressureinthe extrac-tivedistillationcolumn.Alowerpressurereducestheusageof entrainerasitincreasestherelativevolatilityofDIPE–IPAfor thesameentrainercontentinthedistillationregionwherethe extractivecolumnoperates.A0.4atmpressurewasselected toenabletheuseofcheapcoolingwaterinthecondenser.
Thenwehaverunanoptimizationaimingatminimizing thetotalenergyconsumptionper productunit,thus defin-ingtheobjectivefunctionOF.OFincludesbothproductsand bothcolumnsenergydemandsatboilerandcondenserand accountsforthepricedifferenceinheatingandcoolingenergy andforthepricedifferenceinproductsales.Rigorous sim-ulations inclosed loop flow sheet were done in all cases. Thetotal number ofcolumntraysis identicalto literature
worksofLuoetal.forthesakeofcomparison.Othervariables havebeenoptimized;twodistillates,entrainerflowrate,reflux ratios,entrainer feed location and mainfeed location.The totalannualizedcost(TAC)wascalculatedforallprocesses.
Acompetitive designisfound: energycostand TACare reduced by13.4%and 6.3%,respectivelycomparedtoLuo’s design.Thecausesare examinedinthe lightofthedesign parameter interrelations and through explanations of the extractiveefficiencyindicatorsandprofilemapbasedonthe analysisofthermodynamicinsight. Threeimportantissues haveemerged.Firstthereductionofthepressureisbeneficial totheseparatingcase.Second,thedecreaseoftheenergycost functionOFwhenthedistillateflowrateincreasesatconstant purityhasbeenexplainedbytheinterrelationoftwodistillates throughmassbalance.Third,extractiveefficiencyindicators oftheextractivesectionthatdescribestheabilityofthe extrac-tivesectiontodiscriminatethedesiredproductbetweenthe topandthebottomoftheextractivesectionarerelevant cri-terionstoassesstheperformanceofanextractivedistillation processdesign.
Finally,thesensitivityanalysisoverthepressureandthe totalnumberoftraysintheextractivecolumnwasdone.The resultsshowedthatenergycostperunitproductisreducedby 15.1%meanwhileTACissavedby7.6%.
A further investigation isconducted toassess the rela-tionoftheextractiveefficiencyindicatorsontheprocessTAC andtotalenergyconsumptionbyusingmultiobjectivegenetic algorithm.Thenumberoftotaltraysinthetwocolumnsand thevalueoftheoperatingpressurewillberegardedas opti-mizationvariables.
Acknowledgment
XinqiangYouthankstheChineseScientificCouncilforits sup-port.
Appendix
A.
Sizing
and
economic
cost
calculation
Thediameterofadistillationcolumniscalculatedusingthe
traysizingtoolinAspenPlussoftware.
Theheightofadistillationcolumniscalculatedfromthe equation:
H= N
eT×0.6096
Ntray stageexceptcondenser and reboiler,eT tray
effi-ciencyistakenas85%inthiswork.
Theheattransferareasofthecondenserandreboilerare calculatedusingfollowingequations:
A= Q
u×T
u:overallheattransfercoefficient(kWK−1m−2),u=0.852
forcondenser,0.568forreboiler.
Thecapitalcostsofadistillationcolumnareestimatedby thefollowingequations:
Shellcost=22522.8D1.066H0.802 Traycost=1423.7D1.55H
Heatexchangercost=9367.8A0.65
References
Doherty,M.F.,Malone,M.F.,2001.ConceptualDesignof DistillationSystems.McGraw-Hill,NewYork.
Elliott,J.R.,Rainwater,J.C.,2000.TheBancroftpointand vapor–liquidequilibriainthesystembenzene+isopropanol. FluidPhaseEquilibria175,229–236.
Figueiredo,M.F.,Brito,K.D.,Ramos,W.B.,Vasconcelos,L.G.S., Brito,R.P.,2015.Effectofsolventcontentontheseparation andtheenergyconsumptionofextractivedistillation columns.Chem.Eng.Commun.202,1191–1199.
Gerbaud,V.,Rodriguez-Donis,I.,2014.Extractivedistillation.In: Gorak,A.,Olujic,Z.(Eds.),DistillationBook,II.Elsevier, Amsterdam,p.201,ISBN978-0-12-386878-7(Chapter6).
Gil,I.D.,Botía,D.C.,Ortiz,P.,Sánchez,O.F.,2009.Extractive distillationofacetone/methanolmixtureusingwateras entrainer.Ind.Eng.Chem.Res.48,4858–4865.
Hilmen,E.K.,Kiva,V.N.,Skogestad,S.,2002.Topologyofternary VLEdiagrams:elementarycells.AIChEJ.48,752–759.
Kiva,V.N.,Hilmen,E.K.,Skogestad,S.,2003.Azeotropicphase equilibriumdiagrams:asurvey.Chem.Eng.Sci.58,1903–1953.
Laroche,L.,Andersen,H.W.,Morari,M.,Bekiaris,N.,1991.
Homogeneousazeotropicdistillation:comparingentrainers. Can.J.Chem.Eng.69,1302–1319.
Lelkes,Z.,Lang,P.,Benadda,B.,Moszkowicz,P.,1998.Feasibility ofextractivedistillationinabatchrectifier.AIChEJ.44,810– 822.
Lladosa,E.,Montón,J.B.,Burguet,M.C.,Munoz,R.,2007.Effectof pressureandthecapabilityof2-methoxyethanolasasolvent inthebehaviourofadiisopropylether–isopropylalcohol azeotropicmixture.FluidPhaseEquilibria262,271–279.
Lladosa,E.,Montón,J.B.,Burguet,M.C.,DelaTorre,J.,2008.
Isobaric(vapour+liquid+liquid)equilibriumdatafor (di-n-propylether+n-propylalcohol+water)and(diisopropyl ether+isopropylalcohol+water)systemsat100kPa.J.Chem. Thermodyn.40,867–873.
Logsdon,J.E.,Loke,R.A.,2000.IsopropylAlcohol.Kirk-Othmer EncyclopediaofChemicalTechnology.Wiley,NewYork.
Luo,H.,Liang,K.,Li,W.,Li,Y.,Xia,M.,Xu,C.,2014.Comparisonof pressure-swingdistillationandextractivedistillation methodsforisopropylalcohol/diisopropyletherseparation. Ind.Eng.Chem.Res.53,15167–15182.
Luyben,W.L.,2012.Pressure-swingdistillationforminimum-and maximum-boilinghomogeneousazeotropes.Ind.Eng.Chem. Res.51,10881–10886.
Luyben,W.L.,Chien,I.L.,2010.DesignandControlofDistillation SystemsforSeparatingAzeotropes.Wiley,NewYork.
Petlyuck,F.B.,2004.DistillationTheoryanditsApplicationto OptimalDesignofSeparationUnits.CambridgeUniversity Press,Cambridge,UK.
Petlyuk,F.,Danilov,R.,Burger,J.,2015.Anovelmethodforthe searchandidentificationoffeasiblesplitsofextractive distillationsinternarymixtures.Chem.Eng.Res.Des.99, 132–148.
Renon,H.,Prausnitz,J.M.,1968.Localcompositionsin
thermodynamicexcessfunctionsforliquidmixtures.AIChEJ. 14,135–144.
Rodríguez-Donis,I.,Gerbaud,V.,Joulia,X.,2009a.
Thermodynamicinsightsonthefeasibilityofhomogeneous batchextractivedistillation.1.Azeotropicmixtureswith heavyentrainer.Ind.Eng.Chem.Res.48,3544–3559.
Rodríguez-Donis,I.,Gerbaud,V.,Joulia,X.,2009b.
Thermodynamicinsightsonthefeasibilityofhomogeneous batchextractivedistillation.2.Low-relative-volatilitybinary mixtureswithaheavyentrainer.Ind.Eng.Chem.Res.48, 3560–3572.
Rodríguez-Donis,I.,Gerbaud,V.,Joulia,X.,2012a.
Thermodynamicinsightsonthefeasibilityofhomogeneous batchextractivedistillation.3.Azeotropicmixtureswithlight boilingentrainer.Ind.Eng.Chem.Res.51,4643–4660.
Rodríguez-Donis,I.,Gerbaud,V.,Joulia,X.,2012b.
Thermodynamicinsightsonthefeasibilityofhomogeneous batchextractivedistillation,4.azeotropicmixtureswith intermediateboilingentrainer.Ind.Eng.Chem.Res51, 6489–6501.
Shen,W.,Gerbaud,V.,2013.Extensionofthermodynamic insightsonbatchextractivedistillationtocontinuous operation.2.Azeotropicmixtureswithalightentrainer.Ind. Eng.Chem.Res.52,4623–4637.
Shen,W.,Benyounes,H.,Gerbaud,V.,2013.Extensionof thermodynamicinsightsonbatchextractivedistillationto continuousoperation.1.Azeotropicmixtureswithaheavy entrainer.Ind.Eng.Chem.Res.52,4606–4622.
Wang,S.J.,Wong,D.S.H.,Yu,S.W.,2008.Effectofentrainerlosson plant-widedesignandcontrolofanisopropanoldehydration process.Ind.Eng.Chem.Res.47,6672–6684.
You,X.,Rodriguez-Donis,I.,Gerbaud,V.,2014.Extractive distillationprocessoptimisationofthe1.0-1aclasssystem, acetone-methanolwithwater.In:Klemes,J.J.,Varbanov,S.V., Liew,P.Y.(Eds.),24thEuropeanSymposiumonComputer AidedProcessEngineering.Elsevier,Amsterdam,ISBN 978-0-444-63456-6.
You,X.,RodriguezDonis,I.,Gerbaud,V.,2015a.Improveddesign andefficiencyoftheextractivedistillationprocessfor acetone-methanolwithwater.Ind.Eng.Chem.Res.54, 491–501.
You,X.,RodriguezDonis,I.,Gerbaud,V.,2015b.Investigationof separationefficiencyindicatorfortheoptimizationofthe acetone-methanolextractivedistillationwithwater.Ind.Eng. Chem.Res.,http://dx.doi.org/10.1021/acs.iecr.5b02015.