運(yùn)用現(xiàn)代CFD方法設(shè)計(jì)高效率的汽輪機(jī)_英文_

1、第32卷第2期2003年6月熱力透平THERMALTURBINEVo.3No.2June2003ApplicationofAdvancedCFD2MethodstotheDesignofHighlyEfficientSteamTurbinesT.Thiemann,A.deLazzer,M.Deckers(SteamTurbineEngineering,SiemensAGPowerGeneration,Mülheim,Germany)Abstract:Thepresentpaperinvolvestheapplicationofamoderncomputationalfluiddyna

2、mics(CFD)methodinthedesignprocessofhighlyefficientsteamturbines.Themethodisappliedtotypicalsteamturbinedesigncases,namelythecalculationofthefullythree2dimensionalflowthroughsteamturbinestages,bladesealsandturbineex2hausthoods.Remarkablygoodagreementwasachieved.Furthermore,theresultsofstructuredandun

3、structuredcodeswerecomparedagainsteachother.ExcellentagreementwasfoundandtheuniqueabilitiesofanunstructuredCFDcodetomodelcomplexgeometriesaredemonstrated.Keywords:steamturbinedesign;blade2pathdesign;labyrinthseal;exhausthoodCLCnumber:TK212Document:AArticle:-)運(yùn)用現(xiàn)代CFD摘要:(Dynamics,CFD)技術(shù)設(shè)計(jì)高效率汽輪機(jī)的方法,。計(jì)算

4、結(jié)果和實(shí)際是相符的。進(jìn)一步對(duì),已證明,非結(jié)構(gòu)單元CFD程序用于復(fù)雜形狀的流場(chǎng)計(jì)算時(shí),并具有獨(dú)特的功能。關(guān)鍵詞:汽輪機(jī)設(shè)計(jì);葉片通流部分設(shè)計(jì);迷宮式密封;排汽缸;計(jì)算流體動(dòng)力學(xué)1IntroductionTheoverallefficiencyofadvancedsteamturbinepowerplantsisstronglyrelatedtotheefficiencyatwhichthepotentialenergyofthesteamisconvert2edintorotationalkineticenergywithinthesteamturbine.Itisthereforethetas

5、kofasteamturbinedesignengineertokeeptheenergylossesaslowaspossibleinallcomponentsofthesteamflowpath,suchastheturbineinletduct,thebladepassage,andtheexhaustcasing.Thisisadifficulttask,sincetheflowalongthispathisextremelycomplexduetoitsinherentthree2dimensionalandunsteadynature.Computationalfluiddynam

6、ics(CFD)methodscanprovidearemedytothissituation.Thesemethodsmustnecessarilyemployrathercrudeturbulencemodelsandtheymaynotresolveallflowfeaturessuchasseparationbubbles,transi2tionregionsandunsteadyvorticesinallcasesaccu2rately.Also,theyareneitherfullygridindependentnorcannumericalerrorsbecompletelyav

7、oided.Nevertheless,therapiddevelopmentsinCFDandcomputerpowerinrecentyearshaveprovidedverysophisticatedandreliabletoolsthatareabletopre2dictoveralltrendsandperformance.Therefore,three2dimensionalNavier2Stokesmethodsarecon2sideredtobebasicdesigntoolsthesedaysandtheyarenowbeingroutinelyusedataveryearly

8、stageinthedesignprocess.CFDtechniquesprovideaplatformforenhancedunderstandingofthecomplexfluiddynamicmecha2nismsinturbomachines.Thiswillultimatelyleadtonewandinnovativedesignfeaturessuchasbowedturbinebladesasshown,forinstance,byJansenandUlm1.Alargenumberofdifferentde2signoptionscanbestudiedingreatde

9、tail,consum2ingonlyafractionofthetimeneededtocarryoutasingleexperiment.ThedesigncycletimeandhenceReceivedDay:2003-03-05Biography:Dr.ThomasThiemann(1966-),male,studiedmechanicalengineeringattheRuhr-UniversityofBochum,Germany.Afterre2ceivinghisdoctoraldegree(Dr.-Ing.)inthefieldofcompressoraerodynamics

10、heworkedseveralyearsasaprojectleaderonindus2trialthermo-andfluiddynamicprojects.SincejoiningSiemensPowerGeneration,heisresponsibleforthefluiddynamicdesignofsteamturbinecomponents.In2002hewasnominatedasprojectleaderfortheproductdevelopmentoftheLPsteamturbines.第2期熱力透平87designcyclecostsarethereforedras

11、ticallyreduced.Someoftheprogressthathasbeenachievedbyus2ingmodernCFDmethodshasbeenreportedbyOeynhausenetal.2,Cofer3,andScarlin4.SinceaconsiderableportionofthetimerequiredforacompleteCFDanalysismustbeaccountedforgridgeneration,ahighlyefficientdesignenviron2mentisofconsiderableimportanceforindustriala

12、pplications.Inthiscontext,theapplicationofun2structuredmeshesespeciallyformodellingcomplexgeometriesisofincreasingrelevance.However,inmostcasesthisleadstoincreaseddemandsonhard2wareresourcesandcomputationtimes.Forthisreason,thedecisiononwhethertouseunstructuredorstructuredmeshescannotbeansweredingen

13、eralbutdependsonthecurrentproblem.Inthesubsequentsections,adescriptionofthecomputationalmethodisgivenandgeneralrequire2mentsforgridgenerationtoolsforengineeringposearedescribed.Themethodisto2calsteamturbinedesign,tionofthefullysteamturbinetheofleakageflowsandtheexhausthoods.Acomparisonisbetweentheca

14、lculatedresultsandexperimentaldatainordertoassessthepredic2tivecapabilityofthepresentmethod.Further2more,thedifferentmeshingschemesarecomparedagainsteachother.manufacturingcostsleadtoasteadyincreaseoftheutilisationfactorofthebladematerialsunderoper2atingconditions.Underthesepremises,calculationmetho

15、dshavetoberefinedtoassuresufficientme2chanicalstrengthoftheaerofoil.Herebydetailedknowledgeoftheloaddistributiononthebladesisrequiredandpossiblyeventhreedimensionalfluid2structurecouplinghastobeconsidered.2.2PhysicalModelIngeneral,theissueofCFDinclassicalturboma2chineryapplications,istoprovideasolut

16、iontothefollowingequationsforagivenpuresinglecompo2nentmediuminsomekindofflowpath(Byrd5,6White):1.conservationofmassdt0(1)thevectorofveloci2ofmomentum+S=-󰂈p-󰂈Dt(2)whereisthestrainontheflowandSisdenotesanoptionalsourcetermofvolumeforces(e.g.gravity,electricalormagneticforces)3.conserv

17、ationofenergyandwiththeheatfluxq.(+v󰂈h)=-󰂈q-:󰂈v+dt(2TurbineDesignUsingCFD2.1GeneralAspectsStateoftheartturbineandblade2pathdesignin2evitablyincorporates3DCFDanalysisforseveralreasons.Firstly,althoughnotalwaysmentionedasamajorreason,experimentalinvestigationshavesimplybecomeve

18、ryexpensive.Experimentalstudiesoverawideparameterrangecannotbeaffordedanymoreasameanstodevelopaspecificbladeoranexhausthooddesign.Nowadays,parameterstudiescanonlybeaffordednumerically,whereasexperimentalvalidationusu2allywillberestrictedtoacertainfinal(orclosetofinal)design.Secondly,currentdemandson

19、tur2bineefficiencyrequirebladeaerofoildesignphiloso2phiestakingintoaccountatleastthree2dimensionaleffectsinformofan”averagedsteadystate”3Dflowfieldbutpossiblyalsocallforconsiderationoftransientandunsteadyeffects.Thisisofspecialimportancefortheturbineexhaustwherethepres2surerecoverydependssignificant

20、lyontheflowfielddistributionleavingthelaststage.Furthermore,thedemandsonefficiencyaswellasthoseonplantproductivityaimedatreducingthematerialanddt+v󰂈p)(3)Forcompressiblefluids,onehasalsotoconsidertheappropriateequationofstate.Equations(1)2(3)giveacompletedescriptionofthemacroscopicflowofasing

21、lecomponentflowfield,iftheflowfieldcouldbecompletelymod2elled.However,thiswouldrequireextremelyfinegridstoresolvethetiniestturbulencestructureswhichishardlyeverpossiblefortheflowchannelsizesrelevantforturbomachinery.Accordingly,theequationsaretimeaveragedandlinearisedwithrespecttomeanlocalflowandflu

22、idproperties,herebyintroducingadditionalvariablesfortheun2known(turbulent)velocityfluctuationsandtheirderivatives.Bymeansofturbulencemodels,theseso2calledReynolds2AveragedNavier2Stokesequa2tionsareclosed.Generally,therearemanydifferentapproachesinturbulencemodels.Awellknownanalyticalmixinglengthmode

23、lsistheBaldwin2Lomaxmod2el,thentherearetheso2calledoneandtwoequa2tionmodels2namedafterthenumberofadditionaltransportequationstobesolvedtogetherwiththe,k2areexamplesoftwo2e2flowfield2,wherek288ApplicationofAdvancedCFD2MethodstotheDesignofHighlyEfficientSteamTurbinesquationmodelsandmostofthemarethemse

24、lvesdi2videdupintoseveraldifferentapproaches.Asum2modelsisgivenin7.Final2maryofdifferentk2ly,thereareReynolds2StressandAlgebraicReynolds2Stressmodelswhichtrytosolveforalltheunknownfluctuationcomponentsinthestresstensor,withtheresult,that6orevenmoreaddi2tionaltransportequationshavetobesolved.Un2fortu

25、nately,everysingleturbulencemodelhasitsfavouritefieldofapplicationanditsspecificneedsongridresolution.Forindustrialapplications,ak2typeturbulencemodelisoftenused.Thechoiceherebyisbasedonitsrelativerobustness,moderateneedsregardinggridrefinementnearwalls(e.g.modelsusuallyrequireratherfineresolutionof

26、k2theboundarylayer)andthefact,thatturboma2chineryflowsingeneralarehighlyturbulent.Herebyonehastotakeintoaccountasystematicerrorregardingwallfrictionandboundarymodelconsideration:Ak2layersasturbulent,thus2upnewboundary2ingedgeandtoaturbu2lentstateisFurthermore,thestandardformofthesemodelsincorporates

27、somemodellinglimitsregardingtheturbulentkineticenergypro2ductionrate,leadingtoanover2estimationoftheturbulentkineticenergygenerationinstagnationpointflow;asflowsaroundturbinebladesalwayshaveanatleasttwodimensionalstagnationpoint,thiseffecthastobecapturedbyappropriatemodifi2cationstothemodel(oneofthe

28、seistheso2calledmodelwhichKato2Launderextensiontothek2wasusedinthenumericalcodeusedforthepre2sentedresults).Forthepresentstudies,thecommercialCFX2TASCflow󰃑andCFX25.5󰃑codeswereused.ThesecodesbothsolveafinitevolumeformulationoftheReynolds2AveragedNavier2Stokesequationsusingacoupledimpl

29、icitsolver.CFXTASCflow󰃑herebyisrestrictedtostructuredhexahedralmeshes,where2asCFX25.5󰃑canalsohandleunstructuredtetrahe2dralandtriangularprismmeshes.Turbulencewasmodel,especiallyfortur2consideredusingthek2binestageswhilstusingtheKato2Launderexten2sion.Calculationsweredoneforsteadystat

30、eusual2lywithmixingplanesusedforcouplingreferencesystemsofdifferentstateofrotation.3ApplicationofCFDtoSteamTurbineBladePathandTurbineDesign3.13DBladeDesignUsingCFD3.1.1GridGenerationProvidingasuitablegridfortheproblemtobestudiedwithCFDisofessentialimportanceinor2dertoobtainreliableresults.Unsuitable

31、gridsorcellswithunfavourableaspectratiosdonotonlyworsentheconvergencebehaviourofthecode,buttheymightalsostronglyinfluencethenumericalre2sults,whichmayevenleadtounrealisticflowsepa2ration.Forturbomachineryapplications,agridprovidinggoodorthogonalitywithrespecttothebladeaerofoilandfollowingtheactualfl

32、owexitan2gleshouldbepreferredinordertoprovidegoodnu2mericaltreatmentofthenearwallregionandtoob2tainawellresolvedbladewake.Gradientswithintheflowfieldwillnecessarilybesmearedinpropor2tiontotheextensionofthealongthegra2dients.,forwellalignedwithnormaltothe,beobtainedthanaspectratiocrossingthewakeat,ev

33、enwhenthegridresolutiona2thewakeismuchcoarser.Anotheraspectofgridgenerationisthecouplingofperiodicboundariesordifferentgridblocks.Re2gardingorthogonalityofameshsurroundingatur2bineblade,node2to2nodecorrespondenceonthepe2riodicinterfacemightnotbedesirableandanarbi2trary(orinterpolated)couplingmightbe

34、pre2ferred.However,interpolationalgorithmsforgridcouplingusuallylooseoneorderofmathematicalac2curacy.Thussomedistortionofthegridmightbeanecessarycompromisetoobtainawellresolvedflowfieldwithinthebladepassage.Basedontheseconsiderations,thegridgenera2tionstrategyfollowedforthecurrentstudyontur2bineblad

35、epathandaerofoildesignisasfollows:1.providegoodefficientandtrust2worthynearwalltreatmentofthebladeprofilesbyapplyinganO2typegridwithbestpossibleorthogonalitycloselyaroundthebladeprofile.2.Fillthespacetotheperiodicboundaries(whicharechosentobeapproximatelyinthemid2dleoftheflowchannel)withaC2grid,star

36、tingsomewhereclosetothesuction2sidetrailingedgeoftheprofile,extendingforwardaroundthenoseofthebladeandbacktosomewherenearthepressuresidetrailingedge.Herebyprovidenode2to2nodecorrespondenceontheperiodicboundarytoavoidadecreaseinnumericalaccuracy.3.FilltheremainingspacearoundthetrailingedgewithtwoH2gr

37、idsalignedinawaysuchthatgoodresolutionofthebladewakeisensured.Onceagain,providenode2to2nodecorrespondenceontheperiodicboundaries.4.Ifadditionalgridsareadded,e.g.forsimu2第2期熱力透平89latingshroudleakageflows,attachthosegridsus2inginterpolatednon2node2to2nodematchinginter2faces,sincematchinggridswouldrequ

38、iretremen2dousgriddingefforts.Thegridresolutionitselfnecessarilydependsonboththetypeofproblemtobestudiedandtheavailablehardwareresources.Additionally,forin2dustrialapplicationstheabilitytoreproducespecificgridpropertiesisofvitalimportance.Inmanycas2es,itisnotpracticabletorefinegridsuntilgrid2in2depe

39、ndentresultsareobtained,sincea)numericalresourcescanbelimitedandb)duetothefactthatmostCFDcodesusewallfunctionsfornear2walltreatment.Thesewallfunctionsrequire,thatthedistancebetweenthewallandthefirstgridnodewithintheflowfieldlieswithinacertainfiniterangeofdistances.Furtherrefinementofthegridwillviola

40、tetheunderlyingphysicalassumptionsforthesewall2functions.Thus,theresultsstillchangingdonotimprovebutHowever,sufficientlyisonofdifferenttheinfluenceofbeassumedtobesimilarforallconsidered.Inordertoen2suresimilargridproperties,gridgenerationshouldbebasedonsomekindofstandardisedtemplates,functionsorgrid

41、dingcoefficients,whichallowtosaveacharacteristicgridlayoutandtoapplyittodifferentgeometricparameters.Bytheway,suchtemplatescanalsobesurroundedbyahandywork2ingenvironmentforgriddefinition,parameterad2justment,gridgeneration,controlofgridproper2ties(e.g.extremalgridvolumes,extremalaspectratiosorextrem

42、algridangles)andparametersetmanagement,herebysignificantlyspeedingupthegenerallyiterativegridgenerationprocess.Espe2ciallyforusershavingnodistinct3D2CFDback2ground,thesetoolslikewiseserveassomeindirectmeanstoassureatleastacertainminimumofgridquality,especiallywhennumericalstudiesarehand2edovertosub2

43、contractorsortemporalemployees.AnexampleofthegridforasteamturbinebladesimilartothoseusedinthepresentstudyisgiveninFigure1,showinggridintheblade2to2bladeplane.Figure1:Blade2to2Thedifferentgridsmen2blademulti2blockgridfortionedaboveareclearlyvis2aturbinebladerowible,soisthenode2to2nodegridcorresponden

44、cealongtheperiodicboundary.Inthepresentcase,thegridresolutionwaslimitedforthesakeofamultistageanalysisincludingbladeshroudgeometries.Inthefollowing,someresultsofthenumericalanalysisofafourstagehighpressuremodelturbinewillbediscussed.Themodelturbinewasexperi2mentallystudiedintheSiemensPGMülheimh

45、ighpressureturbinerig.Experimentswereperformedunderpressurisedconditions(inletpressurearound40bar).Thusthemodelturbinewasoperatingun2derReynoldsnumberssimilartothoseofrealhighpressuresteamturbines.Theturbinewasequippedwithbladeshroudsandsealsprovidingabladepathgeometrysimilartotherealsteamturbinebla

46、de2pathinordertoachieveasarealisticoverallflowfieldturbineaspossi2entireconfigurationbladeseals,isdepictedinNotethesealingarrangement(inthepresentcase,steppedshroudswith3sealingfins)andtheresultingcavitiesupstreamanddownstreamofeachindividualbladerow.Asmentionedabove,thesealinggridsarecoupledtothefl

47、owpathusinginterpolatedinterfaces,Figure3.Figure2:Configurationoftheentiremodelturbinewithseals3.1.2ResultsforthefourstageturbinemodelTheresultsoftheanalysisofthemultistagetesttur2binerigisfocusedonthefollowingissues:1.Theinfluenceofleakageflowsontheover2allflowfieldandturbineefficiency.Figure3:Deta

48、ilsofthegrid2.Theimportanceofattachmentofthesealregionmultistageanalysisre2(redgrid)tothemainflowgardingtheaccuracyofPathgrid(black)showingthepredictedflowfieldtheinterpolated(nonnode2)andofthepredictedeffi2to2nodematchingattach2mentciency.Forthesakeofcompar2ison,themachinehasalsobeenmodelledwithre2

49、ducedsize(twoandthreestages)andalsowithouttheentiresealingarrangements(leadingtosmoothinnerandouterannuluscontours).Furthermore,a90ApplicationofAdvancedCFD2MethodstotheDesignofHighlyEfficientSteamTurbinesconfigurationwithblockedstatorbladesealswassimulatedwheretheshroudofthebladeswerein2creasedinthe

50、numericalmodelsuchastocontacttherotorcontour.Thisledtoaconfigurationwerethecavitiesup2anddownstreamofthebladerowswerepresentbutnoleakagemassfractioncouldoc2cur.Theimportanceoftheleakagemassflowontheoverallperformanceoftheturbinecanbeestimatedqualitatively,havingalookatstreaklinesstartingnearthehuban

51、dcasingwallupstreamofthefirststatorasdepictedinFigure4.field2andmuchlikelyontheoverallperformance2ex2ceedstheoneoftherotorbladeleakagefraction.ThisassumptionissupportedbyFigure4whencomparedtothenumericalsimulationofaconfigurationwithblockedstatorseals,Figure5.Thedifferenceofthestreamlinepatterninthe

52、hub2regionisstriking,whereasthepatternintheouterflowpathregimere2mainsunchanged.Figure5:SimilartoFigure4,butstatorbladesealsFigure4:Streaklinesplot,showingtheand2tributionofleakageflowpath;Obviously,whenenteringtherotorbladerowisimmediatelydrivenfarupintoflowregionandremainsthere.Sincetheleakagefrac

53、tionofthesubsequentsta2torbladeswillbehavesimilarly,alargeregionofdis2turbedmainflowwillresult.Theleakagefractionleav2ingtherotorbladesealisdrivenbackradiallyoutwardswithinthenextstatorbladerowandthusitsinfluenceismuchmorelocallyrestricted.Onecanconclude,thattheinfluenceofstatorbladeleakageontheover

54、allflowTheoftheflowalsobecomesandrotoraxialexitinFigure6.Ob2fractiondoesstronglyen2ofthechannelvortexandleadsshiftoftheflowfielddisturbancetowithinthemainflowregion.Accordingtothestreaklineplots(Figures4and5)theinfluenceatthehubisstronger(velocitypeakisshiftedfrom10%to25%spanfromhub)thanatthecasing(

55、peakve2locityshiftedtolessthan20%fromcasing).Asonewouldexpectfromthestreaklineplots,thenewvelocitypeakundertheinfluenceofrotorbladeleakageflowatabout80%radialspanismuchmoreconstricted(orlesssmeared)inthespanwisedirectionthantheonearound25%radialspangen2eratedbythestatorbladeleakage.Figure6:Radialdis

56、tributionofstatorandrotorbladeexitaxialvelocity.Variationsforstages1to4withnormalseals(opensymbols)andblockedstatorseals(filledsymbols)JudgingfromFigures4to6,gradualflowfielddegradationalongthemachineisobvious.Aboveall,theperformanceofthefirststagewillbebetterascomparedtoallsubsequentstages,sincethe

57、flowfielddisturbanceduetoleakagemassflowhasnotbeendevelopedyet.Asaruleofthumb,onecould第2期熱力透平91concludethatdownstreamofthesecondstage(sta2tor/rotor)repeatingflowconditionshavebeenachieved.Thisshouldbecomeobviouswhencomparingtheefficiencypredictionsforconfigurationswithdiffer2entstagecounts.InFigure7

58、,asummaryoftheseeffectsisgiven.Theefficiencyofa22stageconfigu2ration(thefirsttwostages)withoutseals(alsonosealcavitiesweremodelled)showsnotmuchcorre2spondencewiththemeasureddata.Eventheover2allshapeoftheefficiencychartdeviatesfromthemeasureddata2theefficiencydecreaseforbothsmallandhighloadisunderest

59、imated.Byconsider2ationofstatorandrotorbladesealsonly,thecalcu2latedefficiencyconsiderablyapproachesthemea2sureddataandtheagreementintheoveralltrendscanbeseentoimprove.Whenconsideringfurthermore,thattheefficiencyisdistributedunevenlya2longthemachine(thefirststageperformsthanallsubsequentstages)leadingaa