• Aucun résultat trouvé

Experiments and dynamic modeling of a reactive distillationcolumn for the production of ethyl acetate by consideringthe heterogeneous catalyst pilot complexities

N/A
N/A
Protected

Academic year: 2021

Partager "Experiments and dynamic modeling of a reactive distillationcolumn for the production of ethyl acetate by consideringthe heterogeneous catalyst pilot complexities"

Copied!
15
0
0

Texte intégral

(1)

To cite this version :

Fernandez, Mayra Figueiredo

and Barroso, Benoît and Meyer, Xuân-Mi

and Meyer, Michel

and Le Lann, Marie-Véronique

and Le Roux, Galo

Carrillo and Brehelin, Mathias Experiments and dynamic modeling of a reactive

distillationcolumn for the production of ethyl acetate by consideringthe

heterogeneous catalyst pilot complexities. (2013) Chemical Engineering

Research and Design, vol. 91 (n° 12). pp. 2309-2322. ISSN 0263-8762

O

pen

A

rchive

T

OULOUSE

A

rchive

O

uverte (

OATAO

)

OATAO is an open access repository that collects the work of some Toulouse researchers

and makes it freely available over the web where possible.

This is an author’s version published in :

http://oatao.univ-toulouse.fr/9924

Official URL :

https://doi.org/10.1016/j.cherd.2013.05.013

Any correspondence concerning this service should be sent to the repository administrator :

(2)

Experiments

and

dynamic

modeling

of

a

reactive

distillation

column

for

the

production

of

ethyl

acetate

by

considering

the

heterogeneous

catalyst

pilot

complexities

Mayra

F.

Fernandez

a,b,c,d,e,f

,

Benoît

Barroso

f

,

Xuân-Mi

Meyer

a,b,∗

,

Michel

Meyer

a,b

,

Marie-Véronique

Le

Lann

c,d

,

Galo

C.

Le

Roux

e

,

Mathias

Brehelin

f

aUniversitédeToulouse,INPT,UPS,LaboratoiredeGénieChimique,4AlléeEmileMonso,F-31030Toulouse,France bCNRS,LaboratoiredeGénieChimique,F-31030Toulouse,France

cCNRS,LAAS,7avenueduColonelRoche,F-31077Toulouse,France dUniversitédeToulouse,INSA,LAAS,F-31077Toulouse,France

eDepartmentofChemicalEngineering,PolytechnicSchooloftheUniversityofSãoPaulo,AvenidaProfessorLineuPrestes,05088-900São

Paulo,Brazil

fSolvayResearchandInnovationCenter,85avenuedesFrèresPerret,BP62,F-69192StFons,France

a

b

s

t

r

a

c

t

Greatefforthasbeenappliedtomodelandsimulatethedynamicbehaviorofthereactivedistillationasasuccessful processintensificationexample.However,verylittleexperimentalworkhasbeencarriedoutintransientconditions. Theworkpresentsaseriesofexperimentsfortheproductionofethylacetatefromesterificationofaceticacidand ethanolinareactivedistillationpilotcolumn.Thesteady-stateapproachperformedexperimentswithbothexcess ofalcoholandstoichiometricfeedconfiguration.Predictedandmeasuredresultsshowgoodagreementandreveal astrongdependencyofthestructuredpackingcatalystactivityonthepilotgeometryanditsoperatingconditions. Thetransientprocessbehavioroftheheterogeneouslycatalyzedsystemwasdeeplyinvestigatedandcontinuous anddynamicdatawerecollectedforanequilibriummodelvalidation,afterdifferentperturbationsonparameters. Theexperimentalvalidationisshowntobeessentialtoproviderealistichydrodynamicparameters,tounderstand thesensitiveparameterssuchasheatlossesandtoadaptvaluesforthecatalystholdupasafunctionofthesystem.

Keywords:Reactivedistillation;Modeling;Heterogeneouscatalyst;Ethylacetate;Experimentalvalidation

1.

Introduction

In recent years, increasing attention has been directed toward reactive distillation processes as a successful pro-cessintensificationexample.Reactivedistillationmeansthe simultaneousimplementationofreactionanddistillationin acounter currentlyoperatedcolumn,wherechemical reac-tions(mainlyequilibriumlimitedreactions)aresuperimposed onvaporliquidequilibrium.Conversioncanbeincreasedfar beyondwhat is expectedby the chemical equilibriumdue to the continuous removal of reaction products from the

Correspondingauthorat:UniversitédeToulouse,INPT,UPS,LaboratoiredeGénieChimique,4AlléeEmileMonso,F-31030Toulouse, France.Tel.:+330534323652.

E-mailaddress:xuan.meyer@ensiacet.fr(X.-M.Meyer).

reactivezone,reducing costsand contributingtoa sustain-ableproduction.However,thereactivedistillationprocessisa complexsysteminwhichthecombinationofseparationand reactionoperationsleadstonon-linearinteractionsbetween phaseequilibrium,masstransferrates,diffusionand chemi-calkinetics.Asaconsequence,theanalysisoftransientregime operationismadenecessarytobetterunderstandtheprocess behaviorandthepresentnonlinearities.

Thedynamicbehaviorofreactiveseparationsystemshas attractedattention inrecentacademicand industrial stud-ies. Although great effort has been applied to model and

(3)

tosimulatethe dynamicbehavioroftheprocess,verylittle experimental work has been carried out in transient con-ditions.Someauthors developedrigorousdynamicmodels, but only experimental data from steady state operations were considered for model validation. Kenig et al. (1999)

developedarigorousrate-baseddynamicmodelforreactive absorptionprocessesthatwasvalidatedbythecomparisonof thesourgasesreactiveabsorptioninairpurificationpacked columnssimulationagainstpilot-plant steady-state experi-ments.Mihaletal.(2009)studiedahybridreactiveseparation systemconsistingofaheterogeneouslycatalyzedreactive dis-tillationcolumnandapervaporationmembranelocatedinthe distillatestream.Thesteadystatebehaviormodelwas vali-datedbycomparisonwiththeexperimentdatafromthework

donebyKotoraetal.(2008)andthesystemdynamic

behav-iorwas investigatedbysimulation.Other authors reported dynamicexperimentaldataonbatch-operated columnsfor the production of methyl acetate: Schneider et al. (1999)

includedthe explicit calculationofheatand masstransfer rates in a rigorous dynamic rate-based approach and the experiments on a batch distillation column showed good agreementwithsimulationresults.Noeresetal.(2004) con-sideredarigorous rate-baseddynamicmodel fordesigning abatch heterogeneously catalyzedreactive distillationand goodagreementwasverifiedforcompositionsand temper-atures through the column after forced perturbations on refluxratio.Singhetal.(2005)studiedesterificationreaction of acetic acid with n-butanol in a packed distillation col-umnwiththecommercialcatalyticpackingKATAPAK-Sand non-catalyticwiregauze.Adynamicequilibriumstagemodel wasdevelopedtoanalyzetheinfluenceofvariousoperating parameters and several trials were carried out; reasonably good agreementbetween the experimentaland simulation resultswassaidtobeverified,but resultswereonlyshown for one representative attempt. Xu et al. (2005) developed adetailed three-phasenon-equilibriumdynamicmodel for simulating batch and continuous catalytic distillation pro-cesses.Thesimulationresultswereingoodagreementwith theexperimentaldataobtainedfromtheproductionof diace-tonealcohol.Experimentswereperformedwiththecolumn under total reflux and the transient behavior was studied after a decrease ofthe reboiler duty. Finally, Völker et al.

(2007)conductedclosed-loopexperimentstovalidateacontrol

structure.Theauthorsdesignedamultivariablecontrollerfor amedium-scalereactivedistillationcolumnandsemi-batch experimentsonclosed-loopconfigurationwereconductedso astodemonstratecontrolperformancefortheproductionof methylacetatebyesterification.Sequentialperturbationson refluxratioandonacidfeedwereintroducedtoabatch oper-ationreactivecolumn.

Toourknowledge,there isalackofexperimental stud-iesconcerningtheethylacetatereactivedistillationsystem incontinuousdynamicconditionsintheliterature.Adetailed experimentalanalysiswouldbeofgreatimportanceinorder toprovideagoodrepresentativesimulationmodelforthe het-erogeneouslycatalyzedsystem.Theparametersconcerning columngeometry(reboilerdesign,column diameter), tech-nology(catalyst,packingcharacteristics)andhydrodynamics (liquid retentions,flooding considerations) require realistic valuesthatcanonlybewellidentifiedbasedon experimen-talvalidation.Theobjectiveofourstudyisthusthedefinition ofbothsteady-stateanddynamicmodelsandthegeneration oftherequiredexperimentaldataonacontinuous heteroge-neouslycatalyzedreactivepilotcolumn.Severalexperimental

Fig.1–Columnsimplifiedscheme.

trials were conducted to investigate the transient process behavior and to collect continuous and dynamic data for modelvalidation.Discussionsweredevelopedconcerningthe complexitiesofthereactivedistillationprocess,thepossible steady statemultiplicitiesand thesensitivities duetoheat losses,specificoperatingconditionsandtheheterogeneous catalysis.Theimportanceofconsideringallthese peculiari-tiesintheinterpretationofadynamicmodelishighlighted andhencethedevelopedmodelcombininginformationfrom thesteadystateandfromthedynamicregimeisacceptedfor therepresentationoftheethylacetatesystem.

2.

Materials

and

methods

2.1. Pilotcharacteristics

Thereactivedistillationfortheheterogeneouscatalyzed ester-ification of ethanol (EtOH) and acetic acid (HOAc) toethyl acetate(EtOAc)andwater(H2O)isstudied.Experimentswere

carriedoutinalab-scalepilotcolumn.Theconsidered reac-tioniswrittenas:

HOAc+EtOH↔ EtOAc+H2O

Thepilotplantconsistsofaglasscolumnwithaninner diameter of75mmand aheightof7m.Itisdivided into7 modularsectionsof1m,withaliquiddistributoratthetopof eachsection,numberedbottom-up(D1–D7).Thedistributors allowauniformdistributionoftheliquidfeedinthepacking, avoidinganyliquidflowthroughoutthewall.Thedistillation pilotcolumnisschematizedinFig.1.Thepackingstructure hasthefollowingcharacteristics:

- Themodularsectionatthetopofthecolumnisfilledwith the structured packing SulzerDX (numberof theoretical stages∼8).

- The 5central modular sections are filled with the reac-tivestructuredpackingKATAPAK SP-Labo, withanacidic ion-exchange resin as the heterogeneous catalyst. The structuredpackingenlargestheinternalsurfaceand pro-motesturbulencessothatthemasstransferbetweenthe liquidandthevaporphase,andtheinteractionoftheliquid phaseandthereactivecatalystporesareincreased(number oftheoreticalstages∼11).

(4)

- Themodularsectionatthebottomofthecolumnisfilled withthestructuredpackingSulzerCY(numberof theoreti-calstages∼8).

Thecolumnworks atatmosphericpressureand is ther-mally insulated. Thecolumn operates with threedifferent feedflows:theaceticacidfeedisinjectedatdistributorD6, theethanolfeedisinjectedatdistributorD1andathirdfeed flow,whichiscalledReflux,isassimilatedtoanexternalreflux andisintroducedintothecolumnatdistributorD7.This exter-nalrefluxisrepresentativeoftheorganicphasecomingfroma decanterinwhichdifferentstreamsoftheprocessaremixed. Aclassicalglasscondenserisverticallypositionedabove thecolumn;thevaporstreamwouldbefullycondensedand withdrawnasthe distillate flow.After passingthroughthe condenser,thedistillategoestoaheatexchangertobecooled to ambient temperature by glycolic water at 5◦C so as to

avoidtheevaporationoftheethylacetateduringsample with-drawal.Thedistillategoesfurthertoa5-ldecanterwhereit splitsinto twodifferentphases.Theinterfacelevelis regu-latedmanuallybecausetheproductionoftheaqueousphase isrelativelysmall.

Thecolumnreboilerisequippedwithasensorofcoaxial wavestomeasuretheliquidlevelandafacilityenablesthe liquidlevel regulationinthereboiler byactingonthe out-letresidueflowrate.Theheatofthecolumnisregulatedby controllingthetemperatureofthe circulatingoilinsidethe reboilerheatelement.Inordertomaintainaconstantheat duringpilotexperiments(soastoensureconstantdistillate flowrate), the difference between the oiltemperature and theliquidtemperaturewasmaintainedconstant.Theexternal perturbationswereneglected.

Theaceticacidfeedlineiselectricallyheatedupto30◦C,so

astoavoiditssolidification(itsmeltingpointisabout16.6◦C).

The flow meter measurements are numerically smoothed beforeitsconsiderationinthe regulators,whichactonthe frequencyofthepumpvariations.Thus,regularandconstant flowsareobtainedovertime.

Itisworthmentioningthatthe pilotplantconfiguration andoperating conditionswere notdesignedtoprovidethe best productivity or productpurity, but toacquire dataon column characteristics and system behavior, following its dynamictendenciesandresponsestoperturbations.

2.2. Dataacquisition

Theplantinstrumentationprovidesseventemperature mea-surementsinthevaporphaseofeachliquiddistributor,and twotemperature measurements inthe liquidphase ofthe reboileranddistillateline.Inaddition,temperature measure-mentsofthecoolingliquidenteringandexitingthecondenser andtheoilthatheatsthereboilerareprovided.Thefiveflow rates–threefeedflows,theproduceddistillateandresidue– andthepressuredroponthecolumnarealsoregistered. Mea-surementsofboththeflowrateandthetemperatureofthe coolingfluidareplacedinthecondenserinputand output. Allthe process variables,suchasflowrates, temperatures, reboilerliquid level andsystem pressure are collected and monitoredbyastandarddigitalprocesscontrolsystem.

During experimental tests, liquid phase samples were withdrawnfrom fourliquid distributorsD2, D3,D4and D5 (becauseD1,D6andD7receivedthefeedflowsandtheirvalves werenotavailable)aswellasfromthedistillate.Atthe bot-tomofthecolumn,thegeometryofthereboilerresultsinan

importantresidencetime.Inordertowithdrawa representa-tivesampleofthecompositionatthebottomline,aderivation of the down-coming liquid was introduced just above the reboiler.

Theesterificationreactionstudiedincorporatesfour com-ponents:ethanol,aceticacid,ethylacetateandwater.With the purpose of quantifying the composition ofall quater-narysystemcomponents,threedifferentanalyticalmethods wereapplied:gaschromatography,KarlFischermethodand acid–basetitration.

Then,thedatareconciliationprocedurewasconductedby acomputationaltool.Thesetofdataconsistsof5flow-rate and20masscompositionsmeasurements.Randomerrorsare assumedtofollowanormaldistributionandthe reconcilia-tionprocedureminimizestheweightedleastsquaresofthe errors betweenthe reconciledand themeasured variables. Theweightistheinverseofthemeasurescovariancematrix, which isaclassicapproachcalledGauss–Markovestimator

(WalterandPronzato,2010).Bybalances,reactionequations

andphysicalconstraintsconsiderations,thecalculationswere made and they resulted in values of flow rates and com-positions thatrespect the columnmassbalance with high accuracy(error<10−6).

2.3. Experimentalprocedure

Beforestartingtheexperimentaltestswiththeesterification components,someexperimentswerecarriedoutwithwater inordertocalibratethepumps,toestimateheatlosses,to verifyheatequipments,todeterminestart-upandshut-down procedures.

Allthetestswereperformedasfollows:thedaybeforethe experiment,thecolumnwasheatedupundertotalreflux con-ditionswithoutanyfeedsandremainedintheseconditions foratleast12h.Atthebeginningofthetest,thenext morn-ing,thethreefeedstreams:acid,alcoholandexternalreflux wereswitchedonandthesystemwasobserveduntilsteady stateconditionswerereached.Stableoperationalconditions werenormallyreachedafterapproximately7hofexperiment. Duetothefactthatonlytemperaturesweremeasuredonline, theidentificationofsteadystateconditionswasassumedto happenwhentemperatureswerestableatthecolumn distrib-utors.Theknowledgeofthecompositionswasonlypossible aftertheexperimentalruns,becauseofflinelaboratory anal-yseswereadopted,whichrequiredmoretimeforpreparation andcalculations.

Through12experiments,differentperturbationswere per-formedandasubstantialnumberofdatawerecollected,such as flow rate measures, temperature and composition pro-files. The tests were chosen towork under alcoholexcess feed configuration inorder toconsumeall the acidand to meetthestringentacidspecificationforacetates.Forthe pur-posesofcomparison,anadditionaltestatsteadystatewith

Table1–Operatingparametersoftests.

Test Steady-stateconfiguration Dynamicperturbation

1 Ethanolexcess –

2 Stoichiometricfeed –

3 Ethanolexcess +10%refluxmassflow 4 Ethanolexcess −10%refluxmassflow 5 Ethanolexcess +10%acidmassflow 6 Ethanolexcess +10%ethanolmassflow 7 Ethanolexcess Heatperturbation

(5)

Table2–Productstreamscompositions.

Test Feedratio(molar) Distillate(%mass) Bottom(%mass)

EtOH/HOAc EtOAc EtOH H2O HOAc H2O EtOH EtOAc

1 1.13 89.2 3.7 7.1 11.9 88.1 0.0 0.0 2 1.04 91.5 2.6 5.9 21.2 78.9 0.0 0.0 3 1.12 88.7 3.7 7.6 17.4 82.6 0.0 0.0 4 1.12 88.8 3.5 7.7 11.2 88.8 0.0 0.0 5 1.13 89.9 3.6 6.5 8.6 91.4 0.0 0.0 6 1.12 89.1 3.5 7.4 14.4 85.6 0.0 0.0 7 1.13 88.8 3.5 7.7 7.8 92.2 0.0 0.0

Table3–Reactantconversionrates.

Test XHOAc XEtOH

1 97.2% 86.2% 2 93.6% 90.0% 3 96.1% 86.1% 4 97.7% 86.1% 5 97.8% 86.5% 6 96.7% 86.7% 7 98.4% 86.9%

stoichiometric feed configuration was conducted. Table 1

showstheoperatingfeedconditionsandtheperturbation con-ductedateachtest.

Intestsn◦3,4,5,6and7,aftersteady-stateconditionswere

obtained,aperturbationofoneparameterwascausedinthe column,withtheattempttokeepallthe otherparameters constant.Theseperturbationsstronglydisturbedthesystem –temperatureschangedrapidly–anditsbehaviorwas moni-toredforthenextapproximately4or5h.Asaconsequence ofthe laboratoryopeningtimes constraints, therewas not alwayssufficienttimetowaitforthesystemtoreachthenew operatingpoint.

3.

Experimental

results

3.1. Steadystateanalysis

Exceptfortestn◦2(differentfeedratio),thetargetsteadystate,

withethanolexcessfeedconfiguration,wasthesameforall thetests.Thefeedratioandtheresultsoftheproductstream compositionsareshowninTable2.

Fig.2–Temperaturemeasurementsateachdistributor,for alltests.

Whiledistillatecompositionswerenearlythesameforall tests,thecompositionsatthebottomwerelessreproducible. Conversionrateswerecalculated(Table3):

XEtOH=

NfeedEtOH−NbottomEtOH Nfeed

EtOH

XHOAc=

NfeedHOAc−NbottomHOAc Nfeed

HOAc

(6)

Foranapproximately12%ethanolmolarexcessfeed(tests n◦1,3,4,5, 6and 7),theconversion rateswere almostthe

samewithameanvalueof96.9%foraceticacidand86.9%for ethanol.

AsshowninFig.2,temperaturesfromthecolumnlower sections had a marked increase after test n◦4. Thesteady

statesobtainedatthe beginning oftheexperimental cam-paign(testn◦1,3,4)weredifferentfromthelaterones(tests

n◦5,6,7),althoughtheyremainsimilaramongthem.

InFig.3,itcanbeobservedthatthecompositionsinD4 andD5keptalmostconstant,whilecompositionsinD2and D3stronglychangedinthelast testsofthe campaign;this factconfirmstheobservationoftheincrementinthecolumn lower sectionstemperatures. Actually,with theincrease in watercontentand the decrease inethanolcontent,higher valuesoftemperatureareexpected.Theobtainingoftwo dif-ferentsteadystatesthroughtheexperimentsisthusaccepted andthisisfurtherunderstoodwiththesimulationresults.It isworthnotingthat, despitethepresenceofdifferent pro-filesinsidethecolumn,thecompositionsofproductstreams remainsimilar.Thiscanbetheconsequenceofthe separation-onlysectionsaboveandbelowthereactivesection.

3.2. Studyofthetransientregimes

Theexperimentalcampaignresultedinfivetestswith rep-resentativetransientdata.Theperturbationsoccurredunder open-loop conditions, by changing only one variable. The momentoftheperturbationisrepresentedinthegraphsby averticalstraightline.

Theperturbation ofrefluxrateand feedflowrateswere imposedasstepchangesandtheycorrespondtothe experi-mentalperturbationbecausetheactionwasconductedatthe valuesgiventothepumps,whichrespondalmost instanta-neously.Theperturbationontheheatdutywasconductedby droppingthesetpointoftheheatdutycontroller,i.e.a nega-tivestepchangeof3◦Cofthetemperaturedifferencebetween

oilandreboilerliquid.

3.2.1. Testn3:+10%ofexternalrefluxflowrate

Beforetheperturbation,thetemperatureinthereboilerwas approximately100◦Candalltheothertemperatures

through-outthecolumnwerebetween70and85◦C.Thetemperatures

evolutionalongtestn◦3isrepresentedinFig.4.The

temper-aturechangesoccurfirstatthestagesinwhichthereactants feedstreamsareinjected:liquiddistributorsD1andD6.Their responsesare fasterand haveahighergainthan theother ones.Then,the temperaturesatD4andD5alsodecreased, andnewsteadystateconditionsseemtobeobtained2hafter theperturbation.

Regardingcompositionsbeforetheperturbation(Fig.5),the compositionatthe distillate wasobserved tobemore sta-blethanthecompositionatthebottomofthecolumn.The contentsofwaterandethanolatthebottomwerenotconstant even when the constant temperatures allow the assump-tionthatthesteadystateconditionswerereached.Thisfact highlightsthedisadvantageofnothavingonlinecomposition measuresduringtheoperationoftheseintensifiedsystems. Theincrementoftherefluxratio,atconstantheatconditions, resultedinanincreaseindistillateflowrateandindistillate estercontentandadecreaseinthewatercontent.Atthe bot-tom,bothwaterandacidcontentsdecreasedresultingina higherethanolfraction(Fig.6).

Fig.4–Temperaturesbeforeandaftera10%externalreflux feedflowrateincrease.

Fig.5–Masscompositionsatthedistillate(a)andatthe bottom(b)beforeandaftera10%increaseoftheexternal refluxfeedflowrate.

3.2.2. Testn4:

10%ofexternalrefluxflowrate

Afterapproximately2hofassumedsteadystateconditions, the externalreflux massflowratewas reducedby10%.An unexpectedbehaviorwasverifiedinthetemperature atD1 beforetherefluxperturbation:apositivestepofapproximately 2◦Csuddenlyoccurred.Thistemperatureseemstobehighly

sensitive to the operation conditions and this fact can be verified also after perturbation, because the temperatures from D1and D6werethe firstonestoreact.Both ofthem aremeasuredwherethereactantfeedsarelocated,andthis behaviorwasalsoverifiedfortestn◦3.Theothertemperatures

alsoroseovertime,andmarkedgradientswereobservedatD2 andD3.

Regardingcomposition(Fig.7),asexpected,thebehavior wasintheoppositedirectionoftheoneobservedfortestn◦3:

therewasadecrease indistillate flowrateandindistillate ester content,beingreplacedbywaterand ethanol. Atthe columnbottom,the watercontentincreased,replacing the

(7)

Fig.6–Temperatureinliquiddistributorsbeforeandaftera 10%decreaseoftheexternalrefluxfeedflowrate.

Fig.7–Masscompositionsatthedistillate(a)andatthe bottom(b)beforeandaftera10%decreaseoftheexternal refluxfeedflowrate.

3.2.3. Testn5:+10%ofaceticacidfeedflowrate

Duringtestn◦5,theaceticacidfeedflowratewasincreased

by10% andconsequentlyalmostallthe temperaturesrose, buttheyshoweddifferentresponses.TemperaturesatD4and D5remainwithconstant positivegradientuntilthe endof the experiment. Atdistributor D6,a stepofapproximately 2◦Cwasverifiedandthetemperaturecontinuedtoincrease

afterit.ThetemperatureatD3decreasedrightafterthe per-turbationbutitsbehaviorchangedlateranditstartedtorise. ThetemperaturesfrombothdistributorsD1andD2showed oscillationsbutremainwithmeanvaluesclosertotheones observedatnominalregime.Actually,thetopsectionsofthe columnweremoreaffectedbytheperturbationonacidfeed thanthebottomsections,duetotheproximitytoacidfeed location.Theexperimentwas stoppedbeforeanewsteady statewasreached(Figs.8–12).

Thecompositionsanalysisexhibitsthatthewatercontent inthedistillateroseanditbecamelesspureinester.The bot-tomcompositionbehaviortestifiesthatthesystemwasnot exactlyinsteadystateconditionsbeforetheperturbationof acidfeed.

Fig.8–Temperatureinliquiddistributorsbeforeandaftera 10%increaseoftheaceticacidfeedflowrate.

Fig.9–Masscompositionsatthedistillate(a)andatthe bottom(b)beforeandaftera10%increaseoftheaceticacid feedflowrate.

3.2.4. Testn6:+10%ofethanolfeedflowrate

Aftertheperturbationbydecreasingtheethanolfeedflowrate, thetemperaturesdecreasedthroughthecolumnduetothe strongerpresenceofalightcomponent.Theirresponseswere lessstrongthaninthecaseoftheincreaseinacidflowrate. ThetemperaturesatD1,D2,D3andD4changedfasterthan

Fig.10–Temperatureinliquiddistributorsbeforeandafter a10%increaseofethanolfeedflowrate.

(8)

Fig.11–Compositionsatthedistillate(a)andatthebottom (b)beforeandaftera10%increaseofethanolfeedflowrate.

thetemperaturesatD5andD6.TemperatureatD1driftedat approximately1h30aftertheperturbation.

Asaresultoftheperturbation,astrongerinfluencewas verifiedinthebottomcomposition;theincrementofethanol resultedinahigherconversion,droppingtheamountofacetic acid.

3.2.5. Testn7:reductionofheatduty

After the decrease of 3◦C in the temperature difference

betweenheatoilandreboilerliquid,thetemperaturesatthe bottomsections,D1toD3,wereobservedtodropwithsimilar velocityamongthem,butthe measuresinthetopsections remainedconstant.Actually,with lessheat tothecolumn, lessvapor isproduced and theamount oftheinner liquid increases.

Thefirstconsequenceinproductionwasasharpdecrease indistillateflowrate,followedbyadeclineinheavy compo-nentcontentthroughthecolumn.Bothcontentsofesteron distillateandethanolatthebottomthusincreased(Fig.13).

Fig.12–Temperatureinliquiddistributorsbeforeandafter areductionofheatduty.

Fig.13–Masscompositionsatthedistillate(a)andatthe bottom(b)beforeandafterareductionofheatduty.

4.

Steady

state

model

To represent the continuous reactive distillation system, a modelwasdevelopedwiththeAspenPlus®software.An

equi-libriumstagemodelwasconsideredanditshouldbeadapted fortheprocesssimulationandbehaviorprediction.Itisworth mentioning thatnon-equilibrium models normally provide more details and more precise informationto the simula-tionthantheequilibriummodelsinthecaseofconventional packeddistillationcolumns.However,theavailabilityof reli-ablemasstransfercorrelationsforthecatalyticpackingwould beaprerequisitefortheuseofanon-equilibriumstagemodel. EventhoughBehrensetal.(2006)proposessuchcorrelationfor KATAPAK®-SP,itcannotbeconsideredreliable,since

HETP-valuesresultingfromthiscorrelationarealwaysindependent ofthetypeofpacking,thetestsystem,aswellasthegasload andtheliquidmisdistributioneffects.Consequently,theuse ofsuchcorrelationswouldnotimprovetheaccuracyofthe simulationresults,butcouldevenlowerstheirqualitysince thevariationinseparationperformanceisnotconsideredfor theirdefinition.

Table 4 presents the different parameters to be

deter-minedforthe equilibriummodel. Theintrinsic parameters werechosenfrompreviousstudiesonthermodynamicsand kinetics and from pilot analysis. The NRTL activity coeffi-cient modelwas consideredforthe phaseequilibrium and theHayden–O’Conellequationofstatewasusedtoaccount fortheaceticaciddimerisationinthevaporphase.Toaccount fortheequilibriumchemicalreaction,twodifferentkinetically

Table4–ParametersdefinedintotheAspenPlus®

steadystatemodel.

Intrinsicparameters Operating parameters

Adjustable parameters

-thermodynamics -flowrates -reactionefficiency -kinetics -heatduty -heatloss -pilotgeometry -pressure

(9)

Table5–Adjustablecoefficientforreactionefficiency.

Ethanolexcess Stoichiometricfeed(testn◦3)

C=0.5 C=1

controlledreactionsweredefined:onetorepresentthedirect reaction and another to represent the inverse one. The operatingparameterssuchasflowrates,pressureandheat duty were adapted from the conditions of each test. Con-cerningtheadjustableparameters,itwasnecessarytoadapt valuesoftheglobalreactionefficiencytobetterfitthe simu-latedconversiontoexperimentalresults.Differentfeedratios wereverifiedtoinfluencethecompositionsthroughthe col-umnand,asaconsequence,thecatalystresinactivity.Todeal withthisfact,anadjustablecoefficientCwasconsidered;this factisfurtherclarifiedinTable5.

Theheatlosswasinitiallycalculatedfromtemperatures andmaterialspresentintothecolumnandtheresultingvalue is200W,inwhichapproximately25%isthelossatthereboiler andthe rest islinearlydistributed throughoutthecolumn. However,itwasalsoobservedthattheenvironmental condi-tionsofeachdaystronglyaffectthepilotoperationconditions andtheadjustmentofthisparametershouldbeconsideredin themodel.

Fig.14comparesthesimulatedmasscompositionofthe distillate and ofthebottomwith theexperimentalresults. Theyshowgoodagreementbetweenthem.

Fig.15isasuperpositionofsimulationsand experimen-talresultsoftests n◦1,3and 4.Thesimulatedprofiles are

drawnbycontinuelines(−)andtheirexperimentalvaluesare representedwithdiamond-shapes().Thestraighthorizontal continuous lines represent the range ofmeasured compo-sitions and it can be verified that simulation curves show agreementwiththe straightcontinuelines, concludingthe reliabilityofthemodel.

Forthesamesteadystateoutputs,thecompositionprofiles variedinsidethereactivezone(heightbetween1mand6m), but theirvalueswerealmost similarinsidethe separation-onlyzones. Incontrast withthereactive sectionthathasa

Fig.14–Massfractionsatthedistillate(a)andatthe bottom(b).

flattemperatureprofile,theseparativesectionsshowmarked temperature gradients. Atemperature measurementinside theseparativesectionwouldbehighlysensitivetoasystem dysfunctionorachangeinthesteadystateconditions.Itis thuspossibletoinfertheregulationofhybridreactivecolumns bymeasuresplacedintheseparativesections.Thisfactis veri-fiedintheliteraturebydifferentauthors(Laietal.,2007;Kumar

andKaistha,2009).

4.1. Understandingtheadjustablecoefficientfor reactionefficiency

Asmentionedbefore,toadapttheglobalreactionefficiencyin AspenPlus®,anadjustablecoefficientwasconsideredinthe

(10)

reactionkineticsequationstoaccountfortheheterogeneous catalystactivitysensitivities.Thekineticlawbecame: directreaction: HOAc+EtOH →EtOAc+H2O

inversereaction:EtOAc+H2O→ HOAc+EtOH

rdir=C·ko,dir·[HOAc]·[EtOH]·exp



−Ea,dir RT



rinve=C·ko,inv·[EtOAc]·[H2O] ·exp



− Ea,inv

RT



Someassumptionscanbedrawntojustifythiscoefficient:

• Water inhibits the activity of the ions exchanger resin, whensimultaneouslypresentwiththeorganiccompounds thatshouldreact(Brehelin,2006;DargeandThyrion,1993;

Grob and Hasse, 2006). When aqueous components are

present,disadvantageoustransfercharacteristicsoccurfor theorganiccomponentsonthecatalyticpackingdueto dif-ferenttransferratesbetweenwaterandorganicmolecules totheporesofthecatalyst;whenthefeedisat stoichio-metric proportion,the compositionofwaterthroughthe columnisobservedtobesignificantlylowerthanwhenthe feedisatethanolexcess(Fig.16).

• Themodelsupposesthattheliquidreactionoccursat con-tinuous stirred tank conditions. However, the supposed liquidflowconditionsarenotverifiedinourtests,where thePecletnumberisapproximately30.

• Liquidflowthroughthecatalystbagscanbeinfluencedby somephenomenathatdependonthesolutioncomposition: the existence of preferential paths caused by the non-homogeneous swelling of the resin or the variable wettability ofthe catalyststructure infunction ofwater solutioncontent.

The need for this adjustable coefficient in the catalyst activityhasalreadybeendiscussedintheliterature(Harbou

etal.,2011;Beckmannetal.,2002).Theauthorsbelievethat

thespecificcharacteristics ofthecatalyticpackingand the

Fig.16–Comparisonofcompositionprofilesfortestn◦2

(stoichiometricfeed)andn◦6(ethanolexcess).

disadvantageousflowcharacteristics,inadditiontothe dif-ferentphysicalpropertiesofthesolutions,suchastherelative volatilities,explainthedifferentbehavioroftheprocess.Their considerations arecoherentwiththeassumptionstakenin thiswork.

Thevariationincompositions profiles,regardingresults fromonetestwithstoichiometricfeed(testn◦2)andanother

one withethanol excess feed(test n◦6) representative of

tests n◦1, 3,4, 5and 7 iscomparedin Fig. 16. Itcan be

observedthatthecompositionindistillateisnearlythesame, buttheincreasedamountofaceticacidunder stoichiomet-ricfeedconditionsexitsthecolumnbychangingthebottom composition(Fig.17).

4.2. Understandingtheadjustablecoefficientforheat losses

Despitethefactthatthetargetsteady-statewasthesamefor thetwelvetests,twodifferentsteady-stateconditionswere obtained.Differentweatherconditionsandthusdifferentheat losseshappenedduringthetestsanditcanbeconcludedthat theheatlosshasanimportantinfluenceonthepilot operat-ingconditions.Inordertoimprovethesystemrepresentation, theinitialheatlosscalculatedforthecolumnwaschanged soastodecreasethedistillateflowrateandtobetterfitthe

(11)

compositionprofiles.Morerepresentativevalueswerefound whentheheatlosswasincreasedby11%intestsn◦1,3and4.

Whensimulatingtheprocesswiththisnewvalue,the distil-lateflowrateisreducedby1.2%,whichisalmostinvisiblein historicaldata,butsufficienttoimprovethepredictedprofiles atthecolumnbottomsections.Thus,thedifferenceamong thesteadystatesobtainedthroughoutthecampaigncouldbe thecloserattentiongrantedtothepilotmanipulationasfrom testn◦4.

Thephenomenonofsteady-statemultiplicities,commonly studiedinreactivedistillationcolumns,couldalsoinfluence theattainmentofdifferentsteadystatesthroughthe exper-imentalcampaign.Adeeperanalysisofthesepossibilitiesis developedinthesectiondealingwiththedynamicsimulation oftheprocess.

5.

Dynamic

model

Oncethecolumnconfigurationandtheoperatingparameters werevalidated,thesystembehaviorintransientregimewas analyzeddevelopingadynamicmodel.Here,theneedofthe experimentalcampaignishighlightedfortheacquisitionof realisticvaluesforcolumngeometry,technologyand hydrody-namics.Theseparametersareveryimportanttoinitializethe dynamicsimulationandasmalldeviationcaninduceerrors inthesensitivities,theinstabilitiesandtheresponsesofthe process.

Thevalues for reboiler design, diameter ofthe column and height of theoretical stages were directly considered frompilotobservation.Theliquidholdupinthereboilerwas assumed constant during the experimental tests and the dynamicsimulation,due tothepresenceofa level regula-tion.Specificliquidvolumefractionswereinitializedforstages with structured reactive packing and flooding calculations werepermitted.Alltherequiredinformationswerefedinto the AspenPlus® modeland the steady stateobtainedwas

automaticallyexportedastheinitializationforthedynamic simulation inAspen Plus Dynamics®

. Thevalues concern-ingcolumntechnology,geometry,heatloss,pressure,system thermodynamicsandreactionkineticsremainconstant dur-ingdynamiccalculations.

Thedynamicmodeloperatesunderopen-loopcontrol con-ditions,i.e.theregulationsaresetinmodemanualanddirectly deliverthefixedmanipulatedvariablesand noinformation fromtheoutputsisconsidered.Forthepurposeofbetter rep-resentingtheexperiments,theheatdutyandtherefluxratio arethespecificationsforthesimulationdegreesoffreedom andtheproductsflowratesandthroughputsarethesystem responses.

First,thedynamicmodelrepresentsthesteadystate evolu-tionovertime.Theresultisastablesteadystatethatremains inthevaluesobtainedwiththeAspenPlus®simulation.Due

tothedifference thatsomesimulatedsteadystateshowed ascomparedtotheexperimentalresults,atemperaturebias wasconsideredineachmeasuresoastocomparethedynamic responsesgainsanddelaysinthenextdiscussions.Inorder torepresent all the transient responses, each perturbation wasintroducedintothemodel.AspenPlusDynamics®

pro-videsthevaluesofawiderangeofprocessvariablesthrough thetransientregimes;theevolutionofthetemperature val-uesandthecompositionsinthedistillateandinthebottom productcanbethusanalyzed.Forclaritypurposesand due tothefactthat themostimportanttemperature responses

Fig.18–Experimentalversussimulatedtemperature(a) anddistillatemasscomposition(b)evolutionintestn◦3.

areverifiedinsidethecolumn,thegraphsarepresentedwith thetemperaturesfromdistributorsD2toD7,fortheperiodof approximately2hbeforeuntil2haftertheperturbation; prod-uctoutputtemperaturesdonotexertsuchstronginfluences.

5.1. Testn3:+10%ofexternalrefluxflowrate

AscanbeseeninFig.18a,modelpredictionsandexperimental resultsareingoodagreementfortemperature.Nonetheless, thefinalvaluesforthenewsteadystatedonotexactlymatch theexperimentaldataforD5andD6,thedistributorscloser tothetopofthe column.Theunexpectedbehaviorverified inD1,whichincreasedbeforetheperturbationanddecreased later,wasnotpredictedbythemodel.InFig.18b,thesimulated behaviorisingoodagreementwiththemeasuredethylacetate andethanolcontents,buttheexperimentalvaluesforwater contentdecreasefasterthanthemodel.

5.2. Testn4:

−10%ofexternalrefluxflowrate

The responses from the model and from the experimen-tal data agree in directions, gain magnitudes and time constants for D3 to D6. Fig. 19a allows observing that the temperatures at D1 and D2 drifted and the cause of this phenomenon is not considered in the model. It can be concluded that this behavior is not a direct consequence of the perturbation. The distillate compo-sition evolution is well represented by the model in

Fig.19b.

5.3. Testn5:+10%ofacidfeedflowrate

Inthecaseoftestn◦5,themodelpredictionsshowedsimilar

responsesdirectionstotheexperiments, buttheirbehavior werenotthesame:theexperimentaldatahadmore instabil-ityaftertheperturbationandalthoughthetemperaturesat D1,D2andD3returnedtotheirpreviousvalues,the obtain-mentofanewsteadystatecannotbeassuredinthenext2h.

(12)

Fig.19–Experimentalversussimulatedtemperature(a) anddistillatemasscomposition(b)evolutionintestn◦4.

Similarlytotheunexpectedbehaviorobservedduringtestn◦

4,sometemperatures(D5andD6)driftedandthecauseofthis phenomenonisnotconsideredinthemodel.Whencomparing thepredictedandthemeasuredvaluesfordistillate composi-tion(Fig.20b),theirmagnitudesaftertheperturbationarenot thesame.Incoherencewiththetemperaturesevolution,the pilotwasobservedtoexertstrongerinfluencesthantheones predictedbythemodel.

Fig.20–Experimentalversussimulatedtemperature(a) anddistillatemasscomposition(b)evolutionintestn◦5.

Fig.21–Experimentalversussimulatedtemperature(a) anddistillatemasscomposition(b)evolutionintestn◦6.

5.4. Testn6:+10%ofethanolfeedflowrate

Aftertheincreaseoftheethanolamountinthecolumn,both measured and simulated temperature values followed the sametendencies,atapproximatelythesamevelocity(Fig.21). Yet,anewsteadystatewasnotreachedineitherthecasesand thevaluesattheendofthetestwerenotthesamebetween themodelandtheexperiment.Thefinaldistillate composi-tionhassimilarmeasuredandpredictedvalues.Therewere noimportantconsequencesduetothisperturbation.

5.5. Testn7:reductionofheatduty

Theperturbationonheatdutyduringtestn◦7wascarriedout

bydroppingby3◦Cthedifferencebetweenheatoil

tempera-tureandreboilerliquid.Therealresponseoftheheatdevice couldbeanalyzedbythedynamicevolutionoftheoil temper-atureandforthepurposesofrepresentationonthemodel,a 20-minrampthatdecreasedtheheatdutyby5%wasassumed. Thebeginningandtheendofthisramparerepresentedbytwo differentverticallinesinthegraphs(Fig.22).Thecomposition analysisexhibitsthatsteadystateconditionswerenotreally verified inthe experimentalresults and thus,stabilitiesof thesetemperaturesaftertheperturbationcouldbeexpected. Again, themodel respondsslowerthan themeasured data aftertheperturbation.Aninterestingobservationisthatthe temperatureatD1showsanoscillation,whichisfollowedby themodel.Later,bothresultsstabilize,butatdifferentvalues. Theresponsesofthedistillatecompositionhavethesame directioninthemodelandintheexperiment.Incoherence withthetemperatureresponses,themodelrespondsslower thanthemeasureddatatotheperturbation.

Finally,aftertheanalysisofeachtestandthemodel repre-sentation,itcanbeassumedthatthecolumntemperaturesare stronglysensitivetoexternalconditionsandthattheobserved drifts intemperaturesare theconsequenceofanoperation conditionthat isnotrepeatableforall thetestsandit was notidentifiedduringpilotmanipulations.Externalconditions

(13)

Fig.22–Experimentalversussimulatedtemperature(a) anddistillatemasscomposition(b)evolutionintestn◦7.

were observedtohavestronginfluenceon thesystem due tothepilotgeometryand thishypothesismay beaccepted becausethepilotdimensionsprovidelargesuperficialcontact withtheenvironment.Forexample,itispossiblethat,during sometests(approximately14hfrommorningtoevening),the evolutionoftheambienttemperatureinsidethelaboratory resultedindifferentheatlossvalues,butthemodelconsiders itconstant.Thisgeometricissueisexpectednottooccurin industrial-scaledevices.

Theanalysisofthecolumntransientregimeshowsgreater discrepancy of the predicted and measured temperatures at D1 and D6; the fact that the feeds are positioned at thesestagesmayinduceadditionalperturbations.Moreover, several studies have shown that, as a consequence ofthe nonlinearinteractions,complexopen-loopbehaviorssuchas steadystatemultiplicities,trajectorieswithcomplex attract-ingsetsanddynamicbifurcationscanoccurquitefrequently inreactive distillation, depending on the characteristics of thereactionsystemandontheoperationconditions(

Rosales-QuinteroandVargas-Villamil,2009;Ramzanetal.,2010;Chen

etal.,2002).Theauthorsdetectthedifficultoperatingregions inparameterspacefocusingtheuseofcommerciallyavailable processsimulators.GehrkeandMarquardt(1997)andReder

et al.(1999) deeplyanalyzedthe multiplicityphenomenon:

theyemployedcontinuationalgorithmsinasimulation soft-wareandfoundaninfinitenumberofsteadystatesolutionsin thecolumnwithaninfinitenumberoftraysatinfinitereflux ratio.Itis,however,understoodthatanextendednumberof steadystateswillmostlikelynotoccurinarealcolumn.The authors performedsomeexperimentaltests,inwhich sus-tainedoscillationscouldbefoundandthreemultiplesteady stateswereattainedintherealcolumnforroughlythesame bottomsflowrate.Thesecomplexesevidencesarein coher-encewiththeresultsfoundinthiswork.

Moreprecisely, KumarandKaistha (2008) and Leeet al.

(2006)foundthroughsimulationthatatfixedrefluxrate,

out-put multiplicity, with multipleoutput valuesfor the same reboilerduty, causes thecolumnto driftto anundesirable

Fig.23–Comparisonofmeasuredandsamplingsbubble temperaturesfortestn◦3(a)andtestn4(b).

steady-stateunderopenloopoperation.Bothworksagreethat it can beavoidedforafixedreflux ratiopolicy. Due tothe factthatourpilotisunderfixedrefluxrateconfiguration,the resultsmaybeofimportanceforfurtherstudiesaimingatan experimentalconfirmationofthesteadystatemultiplicities.

5.6. Verificationofthetemperaturesensorsreliability

Duringtheexperimentalcampaign,somedriftsin tempera-tureswereobserved–mainlyduringtestsn◦4and5andthis

phenomenoncouldnotbepreciselyexplained.Anyspecific actionoranychangeinoperationalconditionswasidentified asthereasonforthisbehavior.

Samplings of the solution inside the column were withdrawn during the experiment and compositions were measured by analytical methods. Their theoretical bub-ble temperature were calculated and compared to the experimentallymeasuredtemperaturestoverifythe reliabil-ityofthetemperaturemeasures.

It is worth noting that each liquid distributor has two accesses;oneistheentryforthethermocouple–presentin all distributors–and theotheroneallowseitherthe place-mentofavalvetowithdrawliquidsamplesortheintroduction ofa feedstream. Thus, itwas notpossibletoobtain sam-plingsfromD1andD6,becausetheyreceivefeedstreams.This factisaninconvenientbecausesomeunexpectedbehaviors wereobservedexactlyattheselocationsandtheycannotbe verified.Weaccepttheresultsfromthecomparisonatother distributors.

It can beconcluded from Fig. 23 that the experimental measuresarecoherentwiththesamplingsinthemajorityof cases.Thisfactvalidatesthereliabilityofthetemperature sen-sorsandthustheexistenceofunexplainableperturbationsin thecolumn,whichwerenotpredictedbythemodel.

(14)

Fig.24–Experimentalandsimulatedtemperature evolutionintestn◦5,where“Simul”representsthe

adaptedvaluesforthehydrodynamicsand“BadSimul” considersthedefaultvalues.

5.7. Understandingtheadjustableinitialvaluesfor liquidholdup

Inorder toobtaina reliabledynamicmodel ofthe process it isknownthat its geometric,technological and hydraulic parameters need to be detailed. When these parameters concernexternalmeasures,suchascolumnheightand diam-eter or reboiler and condenser dimensions, for example, thevaluescanbeobtainedfrompilotobservation.However, whentheparametersconcerninternalmeasuressuchasthe flow hydraulics due topacking characteristics, the evalua-tionbecomesmoredifficult.Itmaybepossibletoacceptthe manufacturerspecificationsforsomepacking types,but in the case ofstructured reactive packing, the simple accep-tanceofthe manufacturerspecificationswouldnotbevery reliable.

Concerning the structured packing provided by Sulzer Chemtech® and usedin our separative sections,extensive

dataobtainedfromexperimentalstudiescanbefoundinthe literature.Dimaetal.(2006)and Olujicetal.(2007) investi-gatethehydrodynamicsofacounter-currentgas–liquidflow laboratory-scalecolumnstructuredwithSulzerBXand Mel-lapakPlus,respectively.Thedynamicholdupwascalculated infunctionofthe liquid loadand valuesfrom 0.02 to0.10 werefoundfortheinitialliquidholdupateachstage.The pro-cesssimulatorAspenPlus®proposesadefaultfractionvalue

of0.05.Thevaluesareinagreement.

KATAPAK-SPLabowasusedinthereactivesection,which isa structured catalystsupport foruse in gas–liquid reac-tionsystemsinwhichcatalystpelletssuchasion-exchange resinscan beembedded. Bycombining catalystcontaining wiregauzelayers(catalyticlayers)withlayersofwiregauze packing (separation layers), it can achieve separation effi-cienciesequivalent toup to4 theoreticalstages permeter and catalystvolumefractionsup to50% (Gotze and Bailer, 2001).TheperformanceoftheKATAPAK-SPdependshowever onmany parameters;the mostimportantaredynamic liq-uidhold-up,pressure drop, residence timebehavior,liquid physicalproperties and catalyticload point. Behrenset al.

(2008)experimentallydeterminedthestaticanddynamic

liq-uid holdup characteristics of the catalyst-filled pockets as

Table6–Initialstageliquidfractionforeachpacking.

SulzerDX KatapakSP-Labo SulzerCY

0.02 0.45 0.05

encounteredinKATAPAK-SP.Theauthorsexplainedthatthe value for dynamic liquid holdup was between those for the static liquidholdupand forthe catalyticload point. A methanol–water mixturewasusedandstaticliquidholdup fractions higher than 0.3 were verified. Kramer (1998) also stated that under gas–liquid trickling flow conditions, the staticholdupatapackedbedofsphericalparticlesmay rep-resentupto25%or33%ofthetotalliquidholdup.Itcanbe thusconcludedthattheinitialliquidfractioninthereactive sectionismuchhigherthantheholdupsintheseparative sec-tions.Itwasthennecessarytodefinedifferentvaluestomodel theinitialliquidfractionateachsectionofthecolumn.The valuesthatbetterrepresentthesystembehavioraregivenin

Table6.

Alltherequiredspecificationsforthestructuredcatalytic packinghighlighttheneedofspecialattentionwhen model-ing aheterogeneouscatalyzedcolumn,wherethepresence ofsolidparticles stronglyinfluencesthesystem.Thevalue adoptedforthereactivesection(themostdifferentfromthe defaultvalue proposedbyAspenPlus®)is deeplyrelatedto

the specific operational conditions ofthe process and this is far from the idea ofproposinga generic approach. The difficulties observed with the heterogeneous catalyzed columnsexplainwhythegreatmajorityofindustrialcolumns areunderhomogeneouscatalysisconfiguration.

For thepurposeofcomparison,Fig.24 showsthe exper-imentaland differentpredicted valuesforthe temperature evolution inthe column duringtest n◦5, forexample.The

continuouslinesrepresentthemodelwiththeadaptedand coherent valuesfor thehydrodynamic parameters and the dottedlinesaccountforthesimulationwiththedefault val-ues.Itcanbeverifiedthattherightdefinitionofthehydraulic parametersisofgreatimportanceofthemodelreliability.

6.

Conclusions

and

perspectives

An experimentalcampaignwas conductedforthe produc-tion ofethyl acetatefrom esterificationof acetic acid and ethanolinaheterogeneouslycatalyzedreactivepilotcolumn. Severaltestswere performedtodeterminethesteadystate conditions for afeed configuration with excess ethanol. A thorough analysison steady state characteristics was per-formedandeachtestwassimulatedusingtheAspenPlus®

software.Goodagreementwasobtainedbetween experimen-talandsimulationresults.Oneadditionaltestwasconducted under stoichiometricfeedconfiguration and it wasverified that the feed composition strongly influences the catalyst activitysothatthereliabilityofthemodelrequiresan adap-tationofthereactionkineticsforeach operatingcondition. Important sensitivities of the pilot to heat duty and heat losseswerealsoobserved.Inordertostudythereactive sys-tem dynamics,fiveexperimentaltestswere performedand theyprovidedrepresentativeresults.Perturbationswere car-riedoutinalcoholandacidfeedstreams,refluxrateandheat dutyand sufficientdata wasavailablefordefiningrealistic geometry, technologyand hydrodynamics ofthe pilot.The modelwasdevelopedinAspenPlusDynamics®considering

all the parametersand conditions present atthepilot and a specific discussion on the best representation of the heterogeneouslycatalystandtherelatedholdupwas devel-oped. Thesystem hydraulics isalso shown to be strongly dependentonthepresentsolutionandits operating condi-tions.Thevaluesforliquidholdupmustbeadaptedfromthose

(15)

providedbythemanufacturersorthesimulationsoftwarein ordertoberepresentative.Animportanteffortwasnecessary todevelopa uniquemodelthat qualitativelyand quantita-tivelyrepresentsthesystemtendenciesandresponses.The dynamicmodelobtainedisareliable representationofthe proposedreactivedistillationprocessand itcanbeusedto predict other possible perturbations that anindustrial site mayfacesuchasanimpurityofwateronthefeedstreams, forexample.Itisconcludedthatthereliabilityofacomplex system modellies on the deep knowledgeofits operating conditions and sensitive parameters, specially in the case ofheterogeneous catalyst. Therequirement of experimen-talmanipulationstoobtaincoherentmodelconsiderationsis highlighted.

Theimportantinterests,incomparisontopreviousworks, isthattheoperatingconditionswereanalyzedfora continu-ousprocessunderdifferentperturbationsonfeedflowrates, refluxflowrateandheatdutyandthesamederivedmodel isinagreementwithall the conditions. Theapplicationof thisdynamicapproachtotheheterogeneouslycatalyzedethyl acetateesterificationisasignificantnewcontributiontothe actualresearchconcerningreactivedistillation.

References

Beckmann,A.,Nierlich,F.,Popken,T.,Reusch,D.,Scala,C., Tuchlenski,A.,2002.Industrialexperienceinthescale-upof

reactivedistillationwithexamplesfromC4-chemistry.Chem.

Eng.Sci.57,1525.

Behrens,M.,Olujic,Z.,Jansens,P.J.,2006.Hydrodynamicsand

masstransferperformanceofmodularcatalyticstructured

packings.Chem.Eng.Res.Des.84(A5),381.

Behrens,M.,Olujic,Z.,Jansens,P.J.,2008.Liquidholdupin

catalyst-containingpocketsofamodularcatalyticstructured

packing.Chem.Eng.Technol.31(11),1630.

Brehelin,M.,2006December.Analysedefaisabilité,conceptionet simulationdeladistillationréactiveliquide–liquide–vapeur. Applicationetvalidationexperimentalesurlaproductionde l’acétateden-propyle.Ph.D.Thesis,InstitutNational PolytechniquedeToulouse,France.

Chen,F.,Huss,R.S.,Doherty,M.F.,Malone,M.F.,2002.Multiple

steadystatesinreactivedistillation:kineticeffects.Comput.

Chem.Eng.26,81.

Darge,O.,Thyrion,F.C.,1993.Kineticsoftheliquidphase

esterificationofacrylicacidwithbutanolcatalysedbycation

exchangeresin.J.Chem.Technol.Biotechnol.58,351.

Dima,R.,Soare,G.,Bozga,G.,Plesu,V.,2006.Gas–liquid

hydrodynamicsincountercurrentcolumnswithKatapak-S

andBXstructuredpacking.Rev.Roum.Chim.51(3),219.

Gehrke,V.,Marquardt,W.,1997.Asingularitytheoryapproachto

thestudyofreactivedistillation.Comput.Chem.Eng.(Suppl.),

S1001.

Gotze,L.,Bailer,O.,2001.SulzerChemtechreactivedistillation

withKATAPAK®

.Catal.Today69,201.

Grob,S.,Hasse,H.,2006.Reactionkineticsofthehomogeneously

catalyzedesterificationof1-butanolwithaceticacidinawide

rangeofinitialcomposition.Ind.Eng.Chem.Res.45,1869.

Harbou,E.,Schmitt,M.,Parada,S.,Grossmann,C.,Hasse,H., 2011.Studyofheterogeneouslycatalysedreactivedistillation

usingtheD+Rtray—anoveltypeoflaboratoryequipment.

Chem.Eng.Res.Des.89,1271.

Kenig,E.,Schneider,R.,Gorak,A.,1999.Rigorousdynamic

modellingofcomplexreactiveabsorptionprocesses.Chem.

Eng.Sci.54(21),5195.

Kotora,M.,Buchaly,C.,Kreis,P.,Gorak,A.,Markos,J.,2008.

Reactivedistillation–experimentaldataforpropylpropionate

synthesis.Chem.Papers62(1),65.

Kramer,G.J.,1998.Staticliquidhold-upandcapillaryrisein

packedbeds.Chem.Eng.Sci.53(16),2985.

Kumar,M.V.P.,Kaistha,N.,2008.Roleofmultiplicityinreactive

distillationcontrolsystemdesign.J.ProcessControl18,692.

Kumar,M.V.P.,Kaistha,N.,2009.Evaluationofratiocontrol

schemesinatwo-temperaturecontrolstructureforamethyl

acetatereactivedistillationcolumn.Chem.Eng.Res.Des.87,

216.

Lai,I.,Hung,S.,Hung,W.,Yu,C.,Lee,M.,Huang,H.,2007.Design

andcontrolofreactivedistillationforethylandisopropyl

acetatesproductionwithazeotropicfeeds.Chem.Eng.Sci.62,

878.

Lee,H.,Huang,H.,Chien,I.,2006.Bifurcationinthereactive

distillationforethylacetateatlowerMurphreeplate

efficiency.J.Chem.Eng.Jpn.39(6),642.

Noeres,C.,Dadhe,K.,Gesthuisen,R.,Engell,S.,Gorak,A.,2004.

Model-baseddesign,controlandoptimisationofcatalytic

distillationprocesses.Chem.Eng.Process.43(3),421.

Mihal,M., ˇSvandová,Z.,Markoˇs,J.,2009.Steadystateand

dynamicsimulationofahybridreactiveseparationprocess.

Chem.Papers64(2),193.

Olujic,Z.,Behrens,M.,Spiegel,L.,2007.Experimental

characterizationandmodelingoftheperformanceofa

large-specific-areahigh-capacitystructuredpacking.Ind.Eng.

Chem.Res.46,883.

Ramzan,N.,Faheem,M.,Gani,R.,Witt,W.,2010.Multiplesteady

statesdetectioninapacked-bedreactivedistillationcolumn

usingbifurcationanalysis.Comput.Chem.Eng.34(4),460.

Reder,C.,Gehrke,V.,Marquardt,W.,1999.Steadystate

multiplicityinesterificationdistillationcolumns.Comput.

Chem.Eng.(Suppl.),S407.

Rosales-Quintero,A.,Vargas-Villamil,F.D.,2009.Onthe

multiplicitiesofacatalyticdistillationcolumnforthedeep

hydrodesulfurizationoflightgasoil.Ind.Eng.Chem.Res.48,

1259.

Schneider,R.,Noeres,C.,Kreul,L.U.,Gorak,A.,1999.Dynamic

modellingandsimulationofreactivebatchdistillation.

Comput.Chem.Eng.(Suppl.),S423.

Singh,A.,Hiwale,R.,Mahajani,S.M.,Gudi,R.D.,Gangadwala,J., Kienle,A.,2005.Productionofbutylacetatebycatalytic

distillation.Theoreticalandexperimentalstudies.Ind.Eng.

Chem.Res.44(9),3042.

Völker,M.,Sonntag,C.,Engell,S.,2007.Controlofintegrated

processes:acasestudyonreactivedistillationina

medium-scalepilotplant.ControlEng.Pract.15(7),

863.

Walter,E.,Pronzato,L.,2010.IdentificationofParametricModels:

FromExperimentalData.Springer,LondonLtd.

Xu,Y.,Zheng,Y.,Ng,F.T.T.,Rempel,G.L.,2005.Athree-phase

nonequilibriumdynamicmodelforcatalyticdistillation.

Figure

Fig. 1 – Column simplified scheme.
Table 1 – Operating parameters of tests.
Table 3 – Reactant conversion rates.
Fig. 4 – Temperatures before and after a 10% external reflux feed flow rate increase.
+7

Références

Documents relatifs

L’archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des

In particular, the objectives of this study included the characterization of the spring hydrograph considering the effect of (1) the amount of exchange flow by assigning

A generic plate model has been formulated using warping functions in order to describe the transverse shear stresses.. This model can handle classical plate theories as well as

25 and 75 m depth, and the Strengbach catchment in the Vosges Massif (France), where several shallow boreholes, 50 to 120 m deep were drilled in the upper part of

An advantage (which turns out to be also a disadvantage) of the approach lies in its lovingly intensive description of training at the primary school Ievel, There is probably no

For cellulose acetate grafted with ammonium ionic liquids containing different anions, the anion influence on the sorption data generally followed the same trends as for the first

In this study, an experimental work on the rheology of the reactive PA6 resin has been presented, which showed the influence of polymerization and crystallization on the dynamic

Table 2: Molecular constants of ethyl acetate (trans conformer) obtained with program XIAM and comparison with results of BELGI-C s and ab initio results. 14