You are on page 1of 26

JouleThomsonColdFingerDesignandAnalysis

AJouleThomson(JT)deviceachievescoolingbythrottlingarealgasfromahighpressuretoa lowpressure.AperfectgasexhibitsnoJTeffect:theisenthalpicexpansionofaperfectgasis alsoanisothermalprocess.Therefore,notallexpansionsresultinadropintemperature:in somecases,thetemperaturecanrise.Gasessuchasnitrogen,argon,krypton,andhelium exhibitatemperaturedropwithexpansion,andcanthereforebeusedforcoolingapplications. Furthermore,eachofthesegaseshasadifferentstartingpressureforachievingthemaximum coolingeffect.Fornitrogen,forexample,theJTcoolingeffectpeakswhenstartingnear5000 psia. JTdevicescanbeusedinstages,andincombinationwithothercryocoolingtechniques.They arenotveryefficientthermodynamically,sotheytendtobeusedinapplicationswheresmall, compact(andmotionless)coldfingersarerequired,orinshortlivedapplicationsinwhicha transient(blowdown)sourceissufficient,suchascoolingIRsensorsinamissileorprobe.A closedloopcycleinvolvingaJTthrottleiscalledaLindeHampsoncooler,whichisbasicallyaJT coldfingerwithcontinuoushighpressuregasprovidedbyacompressorinsteadofabottle.The hotgasattheoutletofthecompressoriscooledusingtheenvironmentasasink. Asimpleventureororificecanbeusedasathrottlingdevice,butoftenamorecomplexoption isemployedtoavoidtheneedforverysmallopenings(whichcanclogduetothefreezingof contaminantgases).Oneoptionisanozzletubeororificetube.Anotherisalabyrinthseal1 effectivelyaseriesoflargerapertureorificesinseries. TheperformanceofaJTdevicecanbesignificantlyenhancedbyprecoolingthehighpressure (HP)fluidusingthecolderexhaustfluidfromthelowpressure(LP)sideofthethrottle.A GiauqueHampsoncounterflowfinnedpipeheatexchanger(alsoknownasaprecooleror recuperator)isacommonoption. Thepurposeofthisdocument,andtheassociatedthermal/fluidmodels,istoillustratethe applicationofSINDA/FLUINT,Sinaps,andThermalDesktop/FloCADto
1

JTcryostatanalysis Modelingofgaslabyrinthseals(eventhoughtheypurposelyleakinthisapplication) Modelingofheatexchangersingeneral,atboththesystemanddetailedlevels

www.gpcvalves.com/downloads/jt.pdf LabyrinthflowspoilersisthetermusedbyGPC,butsincemanyapplicationsofsteppedlabyrinthsaretoreduce flowinrotatingturbomachinery,thetermsealorthrottlewillbeusedhere.

Apreliminarydesignwillbedevelopedfirst.Atthistoplevel,thethrottlewillbeconsideredasa simple(single)orifice,andtheheatexchangerwillbetreatedatasystemlevelusing effectiveness/NTUmethods.ThispreliminarymodelwillbepreparedusingSinaps. Subsequently,amoredetailedmodelwillbedevelopedinwhichthelabyrinthsealwillbe designedandmodeledinmoredetail.Also,theheatexchangerwillbemodeledinmoredetail inthissecondround,demonstratingapplicationofcompactheatexchangertechniques.This detailedmodelwillalsobepreparedusingSinaps,althoughaThermalDesktop/FloCAD representationofthefinaldesignwillalsobepresented.

DesignRequirementsandConstraints
Theunitwillberequiredtocoola1Wloadat80K.Theloadhasathermalmassof30gm(0.066 lbm)andaspecificheat(Cp)of300J/kgK,andisconnectedtothecoldfingerviaalinear conductanceof1W/K. Nitrogenat300Kand5000psia(35MPa)isavailable,andmaybeexhaustedtoambient pressure.Minimalnitrogenconsumptionisdesired.Acompactsinglestagecryostatisdesired, withanoveralllengthofnomorethanabout4cm(approx1.5)anddiameternogreaterthan about1.5cm(approx.). Thedetailsofthesourceofnitrogenhavebeenpurposelyomitted:itmightbeaverylarge bottle,anditmightbethecompressorandgascoolerinaLindeHampsoncycle.Eitherway,the implicationofassuminganinfinitesourceisthatonlyasteadystateanalysisneedbeused.This hastremendousimpactontheresultingdesignandanalysis.Therefore,itshouldbementioned thatblowdownJTsystemscanandhavebeenanalyzedusingSINDA/FLUINT,includingcool downtransientsandevenautomatedoptimizationofcomponentstoachievebothrapidcool downaswellassustainedoperation.Insuchablowdownsystem,areservoirisusually availabledownstreamofthethrottletostoreanyliquidthatisgenerated,andsomemeansis providedforpreferentiallyexhaustingonlyvaporfromthatreservoiroratleastdiscouraging lossoflowenthalpyliquid.Forasteadysystem,asmallamountofliquid(e.g.,highqualitytwo phasefluid)mayenterthelowpressuresideoftheprecooler. Manydetailsareobviatedbyassumingthatthisdeviceoperatesatsteadystate,sincethe purposeofthismodelistodemonstrateothermodelingtechniques(e.g.,labysealsandheat exchangers).However,sincemanyJTapplicationsareindeedblowdownsystemsinwhichcool downcharacteristicsandliquidproductionandconsumptionareimportant,theFloCADbased modelincludesademonstrationofsuchtransientcapabilities. Thehighpressurelineis0.013(0.33mm)ID,andismadeof7030coppernickelalloy.Soldered copperfinsareavailable,butwiththesmallestfinthicknessavailablebeing0.0025(0.064mm).

PreliminaryDesign
Forpreliminarydesignpurposes,thethrottlewillbetreatedasaneffectiveorifice,anda toothedlabyrinthinepassageorlabysealwillbesizedlatertoprovidethesamethrottling actiononcetheJTflowrateandthrottleinletconditionsareknown. Similarly,theheatexchangerwilllargelybesizedbasedonNTU(systemlevel,inlet/outlet) methods,andamoredetailedmodelwillbeusedoncethebasicdesignisknown. However,theoverallUAoftheheatexchangercanonlybeestimatedatthispreliminarystage ofdesign.TheoverallUAwillbeastrongfunctionnotonlyofthegeometryoftheheat exchanger(e.g.,axialfinspacingandODoftheannularfins),andofthethrottle(whichlargely determinestheflowrateandwhichcannotbesizedindependentlyoftheheatexchangers performance).ButtheoverallUAwillalsobeafunctionofthelocaltemperaturesand pressures,whichvarygreatlyonbothsidesoftheheatexchanger.Furthermore,unlikemany heatexchangersizingexercises,thetemperatureonthecold(lowpressure,orLP)sideisnot knownaheadoftime:itisafunctionofthethrottleandofthedesignoftheheatexchanger itself.Inotherwords,thisisastronglycoupledproblem.Alsothepossibilityoftwophasefluid enteringtheLPsideoftheheatexchangerfurthercomplicatesthedesign.Adetailedmodel couldeasilyencompassallofthesedetails,butitwouldbeslowertosolve,andclearlymany iterationswillberequiredtoarriveatasatisfactorydesign.Therefore,ahighlyapproximate modelwillfirstbeusedtonarrowtherangeofpossibilities,andtoarriveatastartingpointfor themoredetaileddesign.Aswillbeseen,thistoplevelmodelisacrossbetweenaspreadsheet analysisandafourpointcycle(thermodynamicstate)analysis. ThetermoverallUAisnottobeconfusedwiththeFLUINTconvectiontieUA,whichisthe localconductanceofthattie.Unlessitisusedasasegmenttie,theUAofatieisactuallythe hAinheatexchangerterminology:thelocalfilmcoefficienttimesthelocalheattransferarea thatisthebasisofthatcoefficient.Inotherwords,theoverallUAoftheheatexchangercan becalculatedifallofthelocaltieUAsareknownalongwithotherfactorssuchasfin efficiency,constrictionresistances,foulingresistances,etc. TheSinapsmodelusedforthepreliminarydesignisavailableforinspection(orforuseasa startingpoint):sizing.smdl.Therefore,noattemptwillbemadetodocumentthemodelin detail.Rather,thiswriteupseekstoprovideanexplanationoftheapproachused,andshould beviewedasacompaniontothedataavailableinthemodel.BasicfamiliaritywithbothSinaps andSINDA/FLUINTisassumed.

ModelingtheLoad
Actually,thespecificationof1Wat80Kwouldoverspecifyanysinglesteadystateanalysis: eithertheloadcanbemodeledasaboundarynodeat80Kandthedesignadjustedsuchthat 1Wisextractedfromit,ortheloadcanbemodeledasadiffusionorarithmeticnodewitha sourceof1W,andthenthedesignisadjustedsuchthatthenodeismaintainedat80K.The lattermode(givenQ,findT)wasusedinthisanalysis,butprovisionsfortheoppositemode (givenT,findQ)areavailableaswell.2 Theloadwasmodeledasasinglediffusionnode,load.1,anditwastiedtothedownstreamof thethrottleusinganHTUtiewithUA=1. Someoftheregistersusedtodefinetheloadandboundaryconditionsareshownbelow:

ModelingandSizingtheThrottle
Thethrottleisinitiallymodeledasasimpleorifice,sincetheadiabaticexpansioninsucha devicewillbesimilartomostmechanismswhichcauseapressuredrop.Anorificeessentially hasonlyonedegreeoffreedom:thethroatsizeoraperture.Thissimplifiedtreatment facilitatesansweringthecentralpreliminarydesignquestion:Whatminimumflowrateachieves 1Wofcoolingat80K?Theorificemodelwillbereplacedbyalabyrinthsealmodelwithmore specificdimensionsatalaterstage,buttheprimarygoalofthatmorecomplexdesignwillbeto achievethesamemassflowrateasdidtheorifice(albeitwithalargerminimumflowareafor reducedriskofclogging). Unfortunately,theflowratecannotbedeterminedseparatelyfromtheheatexchangerdesign: theymustbesizedsimultaneously.Thestrongestcrossinfluenceisthetemperatureofthe precooledfluidattheoutletofthethrottlingdevice,butasecondaryinfluenceisthepressure dropoftheheatexchanger.Ifalargeheatexchangerisneededtoaccomplishthedesign,then
ThediffusionnodecanbeheldasatemporaryboundarynodeviaacalltoHTRNODinOPERATIONS.Thiscallis optionallyinvokedviauseoftheLoadModeregister.Ofcourse,thenodetypecouldalsobesimplychangedin Sinapspermanently.
2

anonnegligiblepressuredropintheheatexchangerwillmeanalowerpressuregradient availabletothethrottle:bothadecreasedupstreampressure(duetoflowlossesintheHPside oftheheatexchanger),andanincreaseddownstreampressure(duetoflowlossesintheLP side). Someamountofiterationwillthereforebeneeded.Infact,thiswouldbeagoodproblemfor theSINDA/FLUINTSolver(designoptimizationmodule),whichcouldautomaticallyachievea honedpreliminarydesign,presumingthattheusersufficientlyexplainstheprobleminterms ofobjectivesandconstraints.However,becausethisexampleproblemisalreadylengthy,and becausethepreliminarymodelishighlysimplified,asimplerbutcrudermethodwillbeused:a coupleofparametricsweepswhichareexercisedafewtimeseachinordertoclosethedesign iteration. Assumingthatanapproximateheatexchangerdesignhasbeenestablished(asexplainedinthe nextsection),asimpleparametricsweepoforificeholesizesismade,plottingboththe temperatureoftheload(at1Wdissipation),andtheestimatedNTUoftheheatexchanger (registerNTUeff).Therequireddesignwillbethatwhichachievesthe80Krequirement,and whichdoesntrequiremorethantheestimatedheatexchangerperformance.Asseeninthe plotbelow,theminimumorificesizeis0.14timestheflowareaoftheHPline(correspondingto amassflowrateofabout0.86gm/sor6.8lbm/hr)sincethatisthesizethatyieldsan80Kload. ThisthrottleflowareafractionisexpressedastheregisterOratio,notingthattheIDoftheHP lineiscontainedintheregisterIDhp:

Theplotbelowwasmadeusinganinput(assumed)NTUof5(effectivenessof0.8333for counterflow),andatthedesignpoint(Oratio=0.14)theestimatedperformanceoftheheat exchangerisaboutNTUeff=4.9.Thisagreementverifiesthatthepredictedperformanceofthe heatexchanger(NTUeff)isaboutthesameasthatrequired(NTU):theorificesizingcan thereforebetrusted.Iftheheatexchangerdesignwasenlargedorreduced,thentheabove sizingwouldneedtoberepeatedusingthenewbaselineheatexchangerdesign.Fortunately, fewsuchiterationsarerequiredtocloseonadesign.

ThesuddenjumpintheNTUeffatanOratioofabout0.08correspondstothetransitionfrom laminartoturbulentontheLPside.Inotherwords,thedesignchosencorrespondstoa turbulentflowsolutionontheLPside,butnotbymuch.Also,thedesignisverycloseto producingliquidintheoutletoftheorifice:slightvariationscancausetwophasefluidtoenter theLPsideoftheheatexchanger.ThiscanbeseenpastanOratioof0.15,orpastanassumed NTUof5.05(asshownbelow).

ModelingandSizingtheHeatExchanger(precooler)
SINDA/FLUINTregistersplayanimportantroleindefiningtheheatexchanger.Thekeyvariables aredefinedbelowintermsofa(nottoscale)sketch):

ODhp

Lfins Fthk

1/Fpitch

ODhx IDhx

Fromafewkeyvariables,(theHPlineIDIDhp,theODofeachfindiskFODhp,thefindensity FpitchandthicknessFthk),theconfigurationcanbedefined.Theblockageratio(sigratio),heat transferarea(Aheat)andareapervolume(AtoVrat),core(minimum)flowarea(Acore),and hydraulicdiameter(DHhx)canallbecalculated.Withaguessataheattransfercoefficienton theLPside,thefineffectiveness(Feff)canalsobeestimated.

NTUisthenumberofheattransferunits,definedastheoverallUAoftheheatexchanger(UAo, whichcorrespondstotheregisterUAeff)dividedbytheminimumm*Cpproduct(C)forthe HPorLPsides: NTU=UAo/Cmin whereCmin=min(Chot,Ccold). NTUisthenondimensionalsizeofanyheatexchanger,anditmayberelatedtothe effectivenessoftheheatexchanger,whereisdefinedastheratioofactualheatexchangeto

thetheoreticalmaximum.ForsinglephaseconstantCpflows,thetheoreticalmaximumheat transportis: Qmax=Cmin*(Thot,inletTcold,inlet) whichisbasedontheinflowingtemperaturesonthehot(HP)andcold(LP)sidesoftheheat exchanger. Foracounterflowheatexchanger: =NTU/(1+NTU) NTUisausefulsizingparameter,butitisdifficulttogeneralizetotwophaseflows,orevento singlephaseflowswithstronglyvaryingproperties.Thisdesignwilldefinitelyexperiencestrong propertyvariationsbecauseofthewiderangeoftemperaturesandpressures.Itmayalso experiencetwophaseflowattheinlettotheLPsideoftheprecooler.3 Fortunately,effectivenesscanbeeasilygeneralizedtocomplexsinglephaseandeventwo phaseflows,usingenthalpyasafunctionofpressureandtemperatureandquality:h(P,T,X). Neglectingquality(i.e.,assumingsinglephase)forclarity: Qmax,hot=m*[h(Phot,outlet,Thot,inlet)h(Phot,outlet,Tcold,inlet)] Qmax,cold=m*[h(Pcold,outlet,Tcold,outlet)h(Pcold,outlet,Thot,inlet)] Qmax=min(Qmax,hot,Qmax,cold) Heatexchangersizingismoredifficultinthismodelthaninmostbecausethecoldinlet(LP side)conditionsarenotspecified:theymustbesimultaneouslysolvedalongwiththerestofthe thermal/fluidnetwork,aspartofanydesigniteration. Usingonenodeorlumptorepresenttheinlet,andanothernode/lumptorepresenttheexit, therearetwoways4tomodelatoplevelheatexchanger:(1)calculatethecurrentexit

Forablowdownsystem,thisliquidwouldbetrappedinareservoirsuchthat(ideally)onlysaturatedvapor enteredtheLPsideoftheprecooler.Inthissteadyversion,thereisnosuchtransientaccumulationofliquid. Instead,ifliquidformsandisnotvaporizedbytheload,thenitwillentertheprecoolerwhereitwillface supercriticalwalltemperaturesandwillvaporizequickly.Nonethelesstheeffectsofthishighqualityfluidcanbe significant,andthemodelmustbegeneralizedtosurvivetheseconditions.Evenifthefinalanswerisallvapor enteringtheLPsideoftheprecooler,duringiterationssometwophaseflowmaybeexperienced. Actually,ifCmax=(constantwalltemperature,perhapsbecausetheoppositesideisboilingorcondensing),then athirdwayispossibleusingeitherHTUSorHTNSsegmentties.Separatesupportnotesontoplevelheat exchangermodelingareavailableatwww.crtech.com


4

temperaturesofthenodes/lumpsrepresentingtheexit,or(2)calculatethetotalheatexchange andsetthecurrentheatratesintoandoutoftheexitnodes/lumpsaccordingly. ForaSINDA(thermalonly)network,inwhichfluidflowisapproximateusingoneway conductors,ifthetemperatureisspecified(method#1above)thentheoutletshouldbea boundaryorheaternodewhosetemperatureisadjustedinVARIABLES1.Ifinsteadtheheat rateattheoutletisspecified(method#2),adiffusionnodeorarithmeticnodeshouldbeused torepresenttheoutletstate. Inthismodel,asingleFLUINTfluidnetworkisusedtorepresentbothsidesoftheheat exchanger.Fortoplevelheatexchangeranalysis,theFLUINTequivalentofaboundarynodeat theexit(method#1above)isnotaplenum,butratheraheaterjunction:ajunctionforwhich HTRLMPhasbeencalled.Aheaterjunctionwillholdconstantthelumpenthalpy(andtherefore temperature,approximately),butnotthepressureaswell(aswouldaplenum).Thealternative (method#2)istousetankorjunction,andapplytheheatrateQLtotheexitsbasedon effectivenesscalculationsmadeinFLOGIC0.Thislatermethodischosenforthismodel becauseitiseasiertocalculateQLratherthanHLorTL(aswouldbeneededfortheheater junctionmethod),5eventhoughthisapproachcanbealittlelessstablewhenconverging.6

TheQLmethodisalsomoreapplicabletotopleveltransients,thoughthosearenotplannedandwouldbeof questionablevalueinthisparticularcase,unlikemanyliquidliquidheatexchangers. Asimplerunningaveragemethodfordampingisapplicableifconvergenceproblemsoccur,suchasQLnew= QLnew*(1F)+QLold*F,whereF=0fornodampingandFapproachesunityforheavydamping.Usually0.1<F<0.5is adequate.Nodampingwasrequiredinthisparticularmodel.


6 5

Theheatexchangereffectiveness(registerEff)wasestimatedusinganestimateoftherequired NTU(registerNTU).BecauseNTUisnotentirelyappropriate,specifyingNTUisreallyjusta meansofspecifyingeffectivenessindirectly. ThecalculationoftheoutletQLisshowbelow,asextractedfromFLOGIC0:

Notethatintheabovelogic,thepossibilityforthepresenceofliquidatjunction#2(theLPor coldsideinlet)istakenintoaccountbythedirectuseoftheenthalpyofthelumpitself:HL2.No presumptionofvaporstateismadeattheinlet,unliketheoutlets.Attheoutlets,thevirtual statesareassumedtobevaporpertheuseoftheVHvaporenthalpyroutine.(Ifliquidhad beenassumed,theVHLIQliquidenthalpyroutinecouldhavebeenusedinstead.TheVQUALH routinecouldhavebeenusedtodeterminethestateifitwereuncertain,orifitweretwo phase.)

AnotherreasonforspecifyingNTUandthencalculatingeffectiveness(Eff)isthattheestimated actualNTUforthecurrentheatexchanger(NTUeff)canbecalculated,withtheexpectationthat whenNTUisaboutthesameasNTUeff,asuccessfulheatexchangersizinghasbeenachieved. NTUeffisthereforeestimatedinlogic(specifically,OUTPUTCALLSforthefluidsubmodel):

TheabovecalculatesUAeffusetheUBandAHT(heattransfercoefficientandheattransfer area,respectively)offourties:twoforbothinlets,andtwomoreforbothoutlets,eachwith halfoftheheattransferareafortheirrespectivesides.Thisapproachrepresentsacrude attempttoaveragepropertiesovertheheatexchanger,whichhasnotbeensubdividedaxially (aswillbedoneinsubsequentmodels).Similarly,fakeboundarynodesareusedtorepresent thewall,andtheDUPL(lumpsideDUPfactor)hasbeensettozerotokeepthetiesfrom actuallytransferringenergyto/fromtheendpointjunctions.Inotherwords,theentirepurpose ofthesefourties(#2,3,4,and5)istoexploitFLUINTsbuiltinheattransfercalculationsto estimatethefilmcoefficientsonbothsides,leadinguptothecalculationoftheoverallUAo. Directcallstotheunderlyingcorrelationscouldhavebeenusedinstead. Thesebuiltin(default)pipeflowcorrelationsareadequateontheHPside,whichisplain(albeit smalldiameter)tubing.However,thedefaultcorrelationsdonottakeintoaccountthe restartingoftheturbulentboundarylayerwitheachfinontheLPside,andsotheyareknown tounderestimatetheoverallheattransferforthisregime.Anaugmentationfactorof TurbMagH=3isthereforeappliedtothetiesXNTM(turbulentNusseltnumbermultiplier).This factorischosensomewhatarbitrarilybasedroughlyonexperiencewithsimilarfinnedtube designs.ApplicationofmorerealisticColburnJhfactor(=StPr1/3)isdescribedinthedetailed model.

Similarly,thefrictionontheLPsidewillbehigherthanthatpredictedbytheplaintube correlations,soanaugmentationfactorofTurbMagF=4isappliedtothepathsFCTM (turbulentfrictionmultiplier,againbasedonexperiencewithsimilardesigns).Applicationofa morerealisticFanningfrictionfactorisalsodescribedinthedetailedmodel. Althoughslight,theeffectsoffinefficiency,wallandfinareacorrections,andtemperaturedrop acrosstheHPtubeareallaccountedfor.Thefinefficiency(Feff)isestimatedbasedona rectangularfinandestimatedfilmcoefficientfrompriorruns,buttheresultissohighastobe essentiallyunityforthisroughsizingexercise.Infact,anadequateapproximationwouldbeto assumeUAoUAcoldsincethecold(LP)sideisthelimiting(bottleneck)conductance. NotealsothattheNTUeffcalculationnecessarilyrequirestheassumptionofsinglephaseflow (becauseoftheneedtocalculateCpusingtheVCPVvaporspecificheatroutine),thoughthe resultingmodelcantoleratedeparturesintotwophaseflowduringiterations. Inanutshell,thestrategyofthisapproachisasfollows: 1. ChooseanNTUvalueasrequired,andcalculatethecorrespondingeffectiveness(in registers) 2. Imposetheresultingeffectivenessonthemodel,assumingsinglephasebutallowingfor strongpropertyvariationsusingenthalpyfunctions(VH,VHLIQ).UsetheQLmethod (versustheheaterjunctionmethod),withlogicplacedinFLOGIC0tosettheoutletheat rates. 3. UsingfaketiestoestimatetheoverallUA,andusethisvaluetocalculatetheactual NTU(asNTUeff)inOUTPUTCALLS. 4. ThedesigniterationcanbeconsideredconvergedifNTU(required)isapproximatelythe sameastheactualNTUeffforthecurrent(baseline)orificesize.Ifthesevaluesaretoo farapart,thenthecurrentheatexchangeriseithertoosmall(NTUeff<NTU)ortoolarge (NTUeff>NTU).

Theaboveparametricsweepismadeatthedesignthrottlesetting,whichwasdescribedabove. Notethattheabove(throttlesizingsweep)wasmadeatthecurrentheatexchangersize.Both sweepsmustbeiterated(sincetheSolverwasnotemployed)ifthereissubstantial disagreementinthepredictedNTUeff,butsmalldisagreementswillberesolvedbymore detailedmodelsthatfollow.Intheabovediagram,therequiredNTUis5(toachieve80K),and theestimatedperformanceoftheheatexchangeisinapproximateagreement:NTUeff=4.9.

DetailedAnalysis(inSinaps)
Inthepreliminarydesign,roughsizingoftheheatexchangerwasperformedusingcrudely estimatedfrictionandheattransferaugmentationfactors(TurbMagF,TurbMagH)appliedto thedefaultplaintubecorrelations.Thethrottlewasmodeledasasimpleorificewhose aperturewassizedtoprovidetheminimumflowraterequiredtoachievealoadtemperatureof 80K. Theoriginalmodelwillbeexpandedinstagestoreplacethetoplevelheatexchangermodel usedinthepriormodelwithamoredetailed(spatiallyresolved)heatexchangerusingtypical dataforsuchaheatexchanger(detailed1.smdl).Theorificewillthenberesizedusingthis refinedheatexchanger,andwillthenbereplacedbyalabyrinththrottlethatprovidesthesame flowrateunderequivalentconditions(detailed2.smdl).

HigherFidelityHeatExchangerModeling
Fromheatexchangersofsimilardesign,theperformanceofthefinnedpipeGiauqueHampson (sizedatthesystemlevel)isestimated7attwoReynoldsnumbers,basedontheFanningffactor andtheColburnjhfactor.ThismethodologyisbasedonthatdescribedinCompactHeat ExchangersbyKaysandLondon. Reynolds(Re) 2200 4400 Fanning(f) 0.040 0.033 Colburn(jh) 0.010 0.008

Fromthesevalues,thefollowingrelationshipscanbefit: f=0.476Re0.322 jh=0.177Re0.373

Theseformulaearecontainedwithinthemodelviatheuseofregisters,asfollows: f=cofRepowf jh=cojRepowj

Forthelaminarregime,8aNusseltnumber(XNUL)of12.0isexpected. DetailedmethodsforincorporatingtheaboveparametersintoaFLUINTmodelare documentedatwww.crtech.com,andsotheywillonlybebrieflydescribedbelow. Inthepriormodel,theheatexchangerwasmodeledastwoinletsandtwooutlets,withheat transferestimatedinlogicbasedoneffectiveness,andtheresultingheatratesappliedtothe outlets,withheattransferperformanceestimatedusingfakeFLUINTties.Inthismodel,the heattransferpassageswillbesubdividedinto20segmentsaxially,bothinthehighpressure (HP)sideandinthelowpressure(LP)side.Furthermore,themetallicwallandfinswillbe

Notreally.Likemuchofthisexample,dataisguessedifunavailable.Thesenumbersaretypical,butarefictitious andshouldnotbeusedforanyrealdesignwork.Correctionsforviscosity,whichvariesgreatlywithinthisheat exchanger,shouldalsobeappliedinarealisticcase. SINDA/FLUINTattemptstocalculateatransitionalReynoldsnumberasneededtoachievenumericalcontinuity, butdoesthisnumberisconstrainedtotherangebetween1000and10,000.Forthisproblem,thetransitionoccurs atthelowerend(aboutRe=1000).


8

representedbySINDAthermalnodes,includingaxialandradialconduction,fineffectiveness, etc. Althoughthemodelwillonlyberuninsteadystatemode,tubesareusedforthehighdensity HPside,anddiffusionnodesareusedforoneside(theHPside,somewhatarbitrarily)ofthe heatexchanger.9 Atatoplevel(withtheheatexchangerscollapsedforclarity),thenetworkappearsasfollows:

Anextramass(metalvolume)factor,ExtraVol,isaddedtoaccountforthemassoftheheatexchangerinner andoutercanister,whichisdistributedevenlytoallnodes.Usingdiffusionnodesforbothsidesoftheheat exchangerwouldunnecessarilyslowanytransientsolutionbecauseofthesmallresistancebetweeneachsideof theheatexchanger.Inotherwords,thetemperaturedropwithinthemetalitselfisalmostnegligible:athin walledheatexchangercouldhavebeenmodeledusingasinglerowofnodes.


9

Thedepictionoftheexpandedheatexchanger(postprocessedtoshowtemperaturesatthe designpoint)isshowbelow:

TheconductancebetweentheIDandODoftheHPtubeisusedtocalculatetheconductance betweenthetworowsofnodes.Then,thefinefficiency(f)andthecorrectedefficiency(o) o=1(Afin/Aheat)(1f) issettobetheregisterFeff0,whichisappliedtothetiescalingfactorUAMontheLPside. NocorrectionsarenecessaryontheHPside,sinceitisacirculartube.Byhavingsubdividedit, however,theSINDA/FLUINTcodewillcalculateheattransferandpressuredropalongits length,includingslightpressurerecoveryasthefluidcoolstoalargedensityalongthelength. FortheLPside,thecompactheatexchanger(CHX)ffandjhfactorsneedtobeappliedtothe STUBEconnectorsandHTNCtiesrepresentingtheLPflowpassages. Theheattransferisrepresentedusingtheregisterscojandpowjasdescribedabove.These registersareappliedtothetiesasCDB=coj,andUER=1+powj,withUEC=UEH=1/3.TheCDBis theDittusBoeltercoefficient,UERistheReynoldsnumberexponent,andUECandUEHare theheatingandcoolingexponentsonthePrandtlnumber,respectively.Settingthese HTN/HTNCtieconstantsoverridestheinternalturbulentcorrelationstocorrespondtothe equationpresentedabove:jh=cojRepowj Forlaminarflow,theNusseltnumber(XNUL)issetto12. TheabovecorrectionsdonotoverwritethetwophasecalculationswithinSINDA/FLUINT.When thethrottlegeneratestwophasefluidatsteadystate,thensomeliquidenterstheLPsideof

theheatexchangerbecausethereisnoaccumulationofliquidinthecoldreservoirdownstream ofthethrottle.However,thisliquidencountersahotwall:thepipeandfinswillbeabovethe criticalpointofnitrogen(approx.126K).Therefore,filmboilingwilloccurbasedoninternal correlations,ascanbeseenbytheFinthe2PcolumnoftheTIETABtextoutput.Theresulting predictionsarenotfarfromthatofthesubsequentsinglephasezoneintheLP,sonoeffort wasmadetocorrectthem.10 NotealsothattheReynoldsnumberintheLPsidewerealwaysgreaterthan1000,sotheXNUL (laminarNusseltnumber)inputwasirrelevant,thoughusefulduringsolutioniterations. ThefrictionrelationshipisappliedbyestimatingaKfactorcorrection.RecallthatFK= fd*TLEN/DHwherefdistheDarcyfrictionfactor,or4ff*TLEN/DHwhereffistheFanningfriction factor.Theextralosses(aboveandbeyondthosecalculatedbySINDA/FLUINTusingitsdefault tubecorrelations)canthenbeestimatedasFK=(4fffd)*TLEN/DH,whereff=cofRepowfandfdis calculatedinternallybyFLUINTusingtheFRICTfunction(seeSection7intheUsersManual).In ordertobeabletoavoidtheneedtocallafunctionwithintheFKexpression,FRICTwillbe replacedbytheapproximationfd=0.316Re1/4.Theinternalwallroughnessfraction(WRF)will beleftaszerotomakethisapproximationmoreaccurate. Inotherwords,theFLUINTisallowedtocalculatefrictioninternally,includingforanytwo phaseorlaminarregimesencountered,butthenthatcirculartubecorrelationiscorrected usinganexpressionappliedtoallLPpathsasfollows:11 FK=(4*cof*Max(1,REY#this)^powf0.316/max(1,REY#this)^0.25)*TLEN#this/DH#this NotethatREY#thismeansthecurrentReynoldsnumberofthispath,andthatextraeffort mustbeundertakentoavoidraisinganinitialzerovaluetoapower.Thisisaccomplishedby onlyoperatingonthemaximumof1andthecurrentReynoldsnumber,whichconsiderably complicatestheaboveexpression.Excludingalsothe#thisdereferencingyieldsthefollowing invalidbuteasiertocomprehendexpression: FK=(4*cof*REY^powf0.316/REY^0.25)*TLEN/DH Anevenmorecomplexandcompleteexpressionmayberequiredtoavoidunderestimatingfd forlaminarflow(approximatelyREY#this<2300),inwhichregimefd=64/REY#this.
Actually,nomethodexiststocorrectthem.Theincomingdropletswouldstrikethefinsandpipewall,and thereforetheactualheattransferwillbehigherthanthedefaultfilmboilingcorrelationswillpredict.However, thisonlycausesthefluidontheLPsidetotransitiontosinglephasevaporinanevenshorterlength:totalenergy transferwouldnotbegreatlyaffectedbyanimprovedestimateofthefilmboilingregime.
11 10

SINDA/FLUINTiscaseinsensitive.Anycapitalizationsaremadeforclarity,butarenotrequired.

Whenthedetailedmodeloftheheatexchangerwascreated,itoutperformedtheestimates madeinthepreliminarymodel.Asaresult,theorificeaperturewasreducedfromOratio=0.14 to0.095toyield80Kfora1Wload.

DesignofanEquivalentLabyrinthThrottle
GeneralPneumaticsCorp.12usesalabyrinthsealasetofflattoppedteetharrangedin seriesasathrottlingdevice.Normally,labyrinthsealsareusedtodiscourageleakagein rotatingmachinery,suchasinthegapsbetweenimpellorsandpumphousings.Inthiscase however,bothwallsarestationary,andtheleakageisintentional.Thehousingisahollow conical(approx.15degreehalfangle)nozzle.Teetharecutintoaseparatepart:theODofa matchingconicplug.Bymovingtheplugbackandforthaxially,thegapbetweenthetop (land)oftheteethandthehousingcanbevaried,providinganadjustmentofflowrate. Inthepreliminarydesign,aneffectivesingleorificewasusedtosizetheflowrate.Atthedesign point,anydevicewhichachievesthesamepressuredropandflowrateunderthesame conditionswillsuffice.13 Thissingleorificewillbereplacedbyaseriesoflabyrinthteeththatachievethesamethrottling effect.TheconeuponwhichtheteetharecuthasadiameterofIDhx(0.25or6.35mm)atthe entrance,and0.032(0.81mm)attheexit,ascanbeseenintheabovediagram.Seven(= Nteeth)teethofspacing0.06(1.52mm=1/PitchLab)fitonthiscone,usingadepthofcutof 0.008(=DeepLaby).Theclearancebetweenthelandareaoftheteethandthehousing (GapLaby)willbeadjusteduntilitachievesthesamethrottlingeffectasthesimpleorificeat thedesignpoint.

12

www.gpcvalves.com/downloads/jt.pdf Differenceswillappearatotheroperatingpoints,andduringtransients.

13


WideLaby+LandLaby =1/PitchLab LandLaby

DeepLaby

OutDLaby

InDLaby =IDhx

GapLaby

Formodelinghightemperature(nearperfectgas)labyrinthseals,amethodoutlinedinthe SINDA/FLUINTUsersManualusingtheMartinVermesoptionforTABULARconnectorsmaybe used.Onedrawbackofthismethodisthatifchokingoccursatanunknownpoint(suchthatit cannotbebuiltintothetabularinputarray),thenitisbettertomodelthefirst(upstream)N1 teethusingtheTABULARconnector,andthenuseaseparateORIFICEconnectortomodelthe finaltooth.ThistreatmentisespeciallyimportantinaJTmodel,incasethefluidchangesphase beforeflowingpastthefinaltooth. TheexampleSinapsmodel(detailed2.smdl)showssuchaTABULARORIFICEmodeloflabyrinth seals.Theinlet(ith)endoftheTABULARhasbeendupedout(DUPI=0)ashastheoutlet(jth) endoftheORIFICE(DUPJ=0),leavingtheoriginalsingleorificestillincontroloftheflowrate. Usingthismethod,theclearance(GapLaby)canbeadjusteduntiltheflowthroughthe TABULARORIFICEpathisthesameasthatthroughtheoriginalorifice.Aftertheclearancehas beensized,theoriginalORIFICEcouldbedeleted(ordupedoutusingDUPI=DUPJ=0)andthe TABULARORIFICEpathopened(settingitsDUPfactorsbacktounity). UsinganORIFICEtorepresentlossesthroughanannulargapisclearlyanapproximation.The longorificeoptionisinvoked,usingthehydraulicdiameterofDh=2*GapLaby,resultingin ELLD=LandLaby/(2*GapLaby)whereLandLabyisthelengthofthetoothlandarea. UsingtheMartinVermesmethodforrepresentingthelabyrinthflowisclearlyastrained approachforthissupercritical,dense,cold,andevenflashingsituation.Itisreallyonlyprovided fordemonstrationpurposes,andisnotarecommendedapproachforcryostats.Infact,the normaluseoftheMREF=4optionforTABULARconnectorsisimpossiblesincetheinletcan containliquidduringsolutioniterations,andthatsituationisillegalforMREF=4.Therefore,the

MREF=1optionisusedinstead,withGCF=(R/gc)(seetheUsersmanualfordetails).However, thepredictionsarenotcompletelyoutoflinewithothermethods,andthismethodhasthe benefitoflettingthenumberofteethbevariableforsizingsealsorthrottles. Analternativemethodisthereforealsopresented:usingaseparatelong(ELLD>0)ORIFICEfor eachtoothandnotjustthelasttooth.SevenORIFICEconnectorsinseries,withflowareas reducedineachtoothalongtheconicalthrottle.Thisseriesofpathsisalsodupedoutsuchthat ithasnoeffectontheprimaryflow,againallowingitsclearancetobeadjustedwithout disruptingtheprimarythermohydraulicsolutionbasedonthesingleorificethrottle.

Theendresult(notedabote)isapredictionofaclearanceof0.11thousandsofaninch(2.8 microns).Almostallofthepressuredropoccursinthelasttooth,whichischoked.Therefore,a verysmallflowareaisrequiredtomatchthe91%reductioninareathatwasrequiredintheHP lineintheoriginalsingleorificeapproach,notingthattheHPlinehasanIDofonly0.013 (0.33mm). Theabovemethodscanonlybedescribedasbeingusefulforpreliminarydesignwhentestdata islacking.Iftestdataforsimilarunitswereavailable,modelscouldbecalibratedasrequiredto match,includingestimatinganappropriateKvalueintheMartinVermesmodel(register KayLaby)ortheCdoftheorificesinseries.

EquivalentGeometric(FloCAD)Model
AFloCADbasedmodel(JTtransient.dwg)wasconstructedthatwasbasedonthedetailed1.smdl Sinapsmodel,usingasimple(single)orificeasathrottle.Theprimarypurposeofthismodelis

tomakethemodelavailabletoFloCADusers,butasecondarypurposeisthesimulationofa blowdown(chilldown)transient.(Thereisnoreasonwhysuchatransientmodelcouldnot havebeendoneusingtheSinapsbasedmodelaswell.) FloCADenables,butdoesnotrequire,theuseofgeometryformodelconstruction.Insome cases,theuseofgeometryassistsboththegenerationofthemodelandthepresentationofthe results.Inothercases,detailedgeometrycanbeadistraction. TheplaintubeHPsideoftheprecooler,forexample,couldberepresentedasasinglestraight line(FloCADPipe),withthedepictedlengthequalto1.1(LenHX)whiletheactuallengthwas muchlongerduetocoiling.However,itisnotdifficulttodepictahelicalcenterlineforthe FloCADPiperepresentingtheHPline.14 TheHPpipe(Pipe.1)isasimple,plaincirculartube.DiffusionnodesareusedfortheIDandOD oftheHPline.(Thecapacitanceofthewallmaterialwillbemagnifiedtotakeintoaccountthe thermalmassofthefinsandsolder,aswillbedescribedlater.) Thecenterlinehasabout29turns(Nturns).Withaveryfineresolution(say,300axial segments),thelumpswouldappeartofollowthecenterlinenicely.Butonlycoarseresolution (about2030)isrequired.Forsuchlowresolution,thelumpsappeartobefollowingalooser helix,andthepathsappeartoshortcircuitthefinecoil,jumpingfromlumptolumpdirectly. Thisisanartificeofthegraphicalmethods,anddoesnotaffectanswers:thePipeknows aboutthefinelycoiledhelixcenterline.

However,helicalPipescanbeslowtodrawinCADinthecurrentversionofThermalDesktop(V5.1).Under Preferences,underGraphicsVisibility,leavethePipeoptionuncheckedasmuchaspossible.UnderGraphicsSize, settheCurveMeshingResolutiontobeashighaspossiblewhileworkingwiththemodel,settingitbacktoa smallervalue(forbetteraccuracy)beforerunningthemodel.


14

TheLPsideisadifferentsituationentirely.Inthatcase,depictingeachofthethousandsoffins isimpractical,andexplicitlymodelingthegradientswithinthosefins(usingtensorhundredsof nodesperfin)iscomputationallyintractable.Inotherwords,fortheLPside,detailedgeometry isadistractionratherthanafeature.Therefore,theLPsideismodeledasasingle,straightline (visibleatthetopoftheabovediagram)thatconnectsthermallytotheODoftheHPtube, usingoverallcorrections(oorFeff0,asexplainedinthedetailedmodel)andapplyingtheheat transfercorrectionsforthejhrelationship(again,asdocumentedearlier).Inotherwords,the effectofthefinsonheattransferwillbemodeledusingcoefficientsandcorrectionfactorsto avoidtheirexplicitdepictionandmeshing. TheLPpipe(Pipe.2)istiedtothesurfacerepresentedbytheHPhelix(Pipe.1),withtheUAM, CDB,UER,etc.factorsspecifiedtorepresenttheoandjhcalculations.Aswiththedetailed Sinapsmodel,theffcalculationhasbeenperformedusinganestimatedKfactorcorrection basedonFK=(4fffd)*TLEN/DH.

TransientChilldownExample
Thesupplyisnolongerinfinite:ithasbeenassignedavolume(Vbottle),equalto100liters. Heattransferfromthewallofthebottleisneglected. Similarly,thereservoirsvolume(Vreser)isspecifiedas10cc.Ifliquidisformedinthisreservoir, itistrapped.15Themechanismfortrappingliquidisnotspecified,soitismodeledasvaporonly egress(STAT=VSintheexhaustpath). Asteadyrun(caseSteady)verifiessubstantiallythesameanswersasthedetailedSinapsmodel. Becausethesourceisnowatank(representingthefinitevolumebottle)insteadofaplenum,in ordertorunasteadystatethissupplymustbeheld(madetoactlikeatemporaryplenum usingtheutilityHLDLMP)inOPERATIONSbeforethecalltoSTEADY(a.k.a.FASTIC). Aseparatecase(Chilldown)hasbeenaddedtomodelthetransientresponse,startingfroman initialtemperatureof300K. Sincethepriormodelsconcernedthemselvesonlywithsteadystateoperation,thedetailsof thehousing,mandrel,labyrinthplug,etc.werenotnecessary.Theywillbeimportantfora transientchilldownmodel.However,thosedetailsexceedthepurposeofthissampleproblem. Tokeepthisdemonstrationsimple,themissingmasses(includingthoseofthefins)willbe modeledasasimpleaugmentationfactorfortheHPlinespecificheatproperties:MoreMass=3. Becauseofthesmalldiameter,longlengths,andhighdensityintheHPline,thatlinewillbe modeledwithtubesinsteadofSTUBEs.(Thisdistinctionwasirrelevantforsteadystate analyses.)However,becauseofthesmallvolumes,themasswithintheHPlinewillbe neglected:junctionswillbeusedinsteadoftanks. Ifthetransientwerestartedwithallthelumpsatthelowexhaustpressure,thenthe considerableCPUwouldbewastedinthefirstfewmicrosecondsofthetransientinorderto pressurizethesystem(atleastthereservoir)andestablishinitialflowrates.Recognizingthat thiseventisalmostexclusivelyathermaloneandnotahydrodynamicone(otherthanthe fillingandemptyingofliquidinthereservoir)enablesconsiderablesavingstobemade.This computationalaccelerationisachievedbystartingthetransientwiththethermalboundary conditionsheldattheinitial300K,butallowingthefluidsubmodeltostartfromaflowingstate, comingtoequilibriumwiththehotwallsaroundit.

Liquidistrappedduringatransient,sinceatankhasactualvolumeinthatcase.However,duringasteadystate (whenthereservoiristreatedasajunction),thereisnoplacetostoreliquid,soitcontinuestoflowoutofthe junction,butleavesthejunctioninaverywet(fullofliquid)statethatistheonlyeffectofSTAT=VSduringa steadystate.

15

InsideofOPERATIONS,thethermalsubmodelsMAINandHPSIDEareheld(makingalldiffusion nodesactastemporaryboundarynodes)usingtheBDYMODutility.Asteadystateiscalledto initializethefluidsubmodel(withagaintheinletbottletemporarilyheld),thentheholdsare releasedbeforethetransientisinitiated. Finally,tomaketheproblemmorerealistic,the1Wloadisconvertedintoanequivalent conductorfromaboundary(theenvironment)at300K.Inotherwords,thesourceismodeled asaparasiticheatleak,notasadissipativesource.Anewnode(main.2)andconductorare addedtorepresentthissituation.

TransientResultsandDiscussion
Theplotbelowdepictsthetemperatureprofilesforthefirsthourofthetransientcooldown. Afteranhour,theloadhasbeencooledto140Kand12%oftheoriginalnitrogenpressurehas beenexpended.

Evenwithagenerouslysizedpressurantbottleandanunderestimatedthermalmass,thecold fingerisstillcoolingdownafter1hourandisalongwayfromthetargetof80K.Removingthe heatloadentirely(pertheChilldownNLcase,achievedbyzeroingtheconductancemain.g1 fromtheenvironmentandtheload)makeslittledifferenceintheresponse.

Thistransientwasperformedfordemonstrationpurposesonly,butitrevealsafundamental flawinthepreminarydesignprocedure:itwasperformedforsteadyconditionsonly,withno marginoroversizingneededtoachievechilldowninareasonabletime.

You might also like