AComparisonofRANS,URANS,andDDESforHigh-LiftSystemsfrom
HiLiftPW-3
RiccardoBalinandKennethE.JansenAnnandH.J.Smead DepartmentofAerospaceEngineeringSciences
UniversityofColorado- Boulder
AIAASciTechForumJanuary10th,2018
Outline
• Overviewofcasesstudiedandnumericalcomputations
• Numericalresults• GridconvergencestudyonHL-CRMmodel• EffectsofinitialconditionsJSM• RANS,URANS,andDDESonJSM
• Conclusions
2
WorkshopCasesStudiedHL-CRM
JSM
3
Cases AnglesofAttack(AoA) Notes
1a 8°,16° • gridrefinementstudy• full-gapgeometry• B1committeegrids,Coarse-Medium-Fine
1b 16° • gridadaptationstudy• full-gapgeometry• in-house,Simmetrix grids
Cases AnglesofAttack(AoA) Notes
2a 4.36°,10.47°,14.54°,18.58°,20.59°,21.57°
• nonacelle• C1committeegrid,M
2b 21.57° • nonacelle• DDES• adaptationstudy,in-houseSimmetrix grids
2c 4.36°,10.47°,14.54°,18.58°,20.59°,21.57°
• withnacelle• C1committeegrid,M
WorkshopCasesStudiedHL-CRM
JSM
4
Cases AnglesofAttack(AoA) Notes
1a 8°,16° • gridrefinementstudy• full-gapgeometry• B1committeegrids,Coarse-Medium-Fine
1b 16° • gridadaptationstudy• full-gapgeometry• in-house,Simmetrix grids
Cases AnglesofAttack(AoA) Notes
2a 4.36°,10.47°,14.54°,18.58°,20.59°,21.57°
• nonacelle• C1committeegrid,M
2b 21.57° • nonacelle• DDES• adaptationstudy,in-houseSimmetrix grids
2c 4.36°,10.47°,14.54°,18.58°,20.59°,21.57°
• withnacelle• C1committeegrid,M
inprogress
inprogress
WorkshopCasesStudiedHL-CRM
JSM
5
Cases AnglesofAttack(AoA) Notes
1a 8°,16° • gridrefinementstudy• full-gapgeometry• B1committeegrids,Coarse-Medium-Fine
1b 16° • gridadaptationstudy• full-gapgeometry• in-house,Simmetrix grids
Cases AnglesofAttack(AoA) Notes
2a 4.36°,10.47°,14.54°,18.58°,20.59°,21.57°
• nonacelle• C1committeegrid,M
2b 21.57° • nonacelle• DDES• adaptationstudy,in-houseSimmetrix grids
2c 4.36°,10.47°,14.54°,18.58°,20.59°,21.57°
• withnacelle• C1committeegrid,M
inprogress
inprogress
NumericalSet-Up
6
• ComputationscarriedoutwithPHASTAstabilized,finiteelementflowsolver.• Spalart-Allmaras (SA)one-equationmodel(QCRresultsrun,notfocushere).
• Allcomputationsrunfullyturbulent,nospecifiedtransition.
• IncompressibleNavier-Stokesequationssolved.
• Allresultsarewithglobaltimestepping:willcitetimestepinchordflights.
SliceacrosswingsectionoftheJSMgridused
HL-CRM– GridConvergenceStudy
7
Lift:• About5%under-predictionwithCoarse• Mediumwithin1%ofFineforbothAoA• Mediumconverged to“true”solution
Liftanddragcoefficientsvs.numberofgridpointsto-2/3power
Drag:• Slowerconvergence,Med.gridnot
within1%ofFine
LiftCoefficient DragCoefficient
Coarse
MediumFine
HL-CRM– GridConvergenceStudy
8
Pressurecoefficientprofilesat24%and68%ofthehalf-spanfor16° AoA
• ExcessiveflowseparationoverbothflapswithCoarsegrid• MediumandFinegridsalmostidentical.
PS2
PS6
HL-CRM– GridConvergenceStudy
9
Pressurecoefficientprofilesatotherpressurestationsfor16° AoA
HL-CRM– GridConvergenceStudy
10
Pressurecoefficientprofilesatotherpressurestationsfor16° AoA
HL-CRM– GridConvergenceStudy
11
SurfaceLineIntegralConvolutionofWallShearStressat16° AoA
HL-CRM– GridConvergenceStudy
12
SurfaceLineIntegralConvolutionofWallShearStressat16° AoA
Separationlineoninboardflapatmid-chord
HL-CRM– GridConvergenceStudy
13
SurfaceLineIntegralConvolutionofWallShearStressat16° AoA
Separationlineonoutboardflapfurtherdownstream,flowstaysattachedforlonger
HL-CRM– GridConvergenceStudy
14
SurfaceLineIntegralConvolutionofWallShearStressat16° AoA –Zoomonflapgap
HL-CRM– GridConvergenceStudy
15
SurfaceLineIntegralConvolutionofWallShearStressat16° AoA –Zoomonflapgap
Largerregionofseparatedflowattheflapgap
HL-CRM– GridConvergenceStudy
16
Sliceat24%ofhalf-spancoloredbyspan-wisevorticity
Negativevorticity(outofscreen)Positivevorticity(intoscreen)
Flowdirection
HL-CRM– GridConvergenceStudy
17
Sliceat24%ofhalf-spancoloredbyspan-wisevorticity
Distortedshearlayerduetolackofresolution
Shearlayersaccuratelycomputed
HL-CRM– GridConvergenceStudy
18
Sliceat24%ofhalf-spancoloredbyspan-wisevorticity
Morenarrowjetofirrotationalflowthoughgap,slowermovingfluidovertheflapleadingedge
HL-CRM– GridConvergenceStudy
19
Sliceat24%ofhalf-spancoloredbyspan-wisevorticity
Boundarylayerseparation
HL-CRM– GridConvergenceStudy
20
Interimsummary:• Mediumgridsufficientforconvergencetowithin1%forlift,slightlymore
than1%fordrag.• Coarsegridhasexcessiveseparationovertheflaps.• Causeofexcessiveseparationisthepoorresolutionoftheflapcoveshear
layerseparation,themainelementwake,andtheflapgap.
Adaptivity:• Thiscaseposesadifficultchallengeforadaptivity:Mediumgridonly3xlargergives
closetofinesolutionleavesnarrowmarginforadaptive“win”.Finegridonly9xlarger.
• Inourexperience,fullyautomaticanisotropicadaptivity canrequire4ormorecyclesofadaptationandresultingridsaslargeasmedium.Worthwhile?
• Weexploredasimplerapproach:• StartadaptationfromagridthatusesCoarse“surface”gridwithselected
improvementingapsandMediumnormalspacing,growth,andtrailingedgethickness(newmeshis14.5Mnodesvs{8,26.5,70}Mfor{C,M,F}),
• Attempt,inoneadaptation,toimprovelocationsofsurfacegridinadequacytothesamelevelasfine.
• Goal:yieldsamequalityasfineforlesscomputationaleffortthanmedium.
HL-CRM– CustomGridforAdaptivity
21
Pressurecoefficientprofilesatinboardpressurestationsfor16° AoA
HL-CRM– CustomGridforAdaptivity
22
Pressurecoefficientprofilesatoutboardpressurestationsfor16° AoA
HL-CRM– CustomGridforAdaptivity
23
Pressurecoefficientprofilesatoutboardpressurestationsfor16° AoA
SelectimprovementsofB2CommitteeCoarsegrid(normalspacing,trailingedges,andmodestgapresolution)eliminatestheextraseparationandbringtheotherwiseB2CommitteeCoarsegridresolutionintosameflowregimeasMediumandFinegrids(e.g,.1%CL difference).
PreliminaryAdaptivity
Skinner,Doostan,Peters,Evans,andJansen 24
PreliminaryAdaptivity:PreserveSurfaceAnisotropy
Skinner,Doostan,Peters,Evans,andJansen 25
Adaptivity:FineGridResolutionOnlyWhereRequired
Skinner,Doostan,Peters,Evans,andJansen 26
JSM– EffectsofInitialConditions
• MostgroupsusedsteadyRANS,butobservedtwomainstrategiesforinitialconditions• Startingeveryangleofattackfromfree
streamconditions• Usingconvergedsolutionatsmallerangleof
attack– alphacontinuation
27
RANScomputationsontheJSMno-nacellemodelfrom3rd AIAAHigh-LiftWorkshop1
• Significantvariationinparticipantpredictionsdueto:• Flowsolver(numerics)• Turbulencemodel• Modelingstrategy(initialconditions(IC),timestepsize,etc.)• Grids
numericalexperiment
JSMLiftCurve
JSM– EffectsofInitialConditions
28
LinearSectionoftheLiftCurve• MultiplesolutionsforthesameAoA• FreestreamICleadstounder-predictionoflift• Alphacontinuationresultsmatchexperimentalliftwell
Liftcoefficientvs.angleofattack(AoA)
JSM– EffectsofInitialConditions
29
LinearSectionoftheLiftCurve– 14.54° AoA
• FreestreamICshowsmassiveseparationdownstreamoftracks7and8
• Alphacontinuationsolutiononlyseparateddownstreamoftrack8,agreeingwithexperimentaldata
Time-averagedwallshearstressalongthestream-wisedirection(Wss_X)
FreestreamIC Alphacontinuation
Tr8Tr7
Tr8Tr7
JSM– EffectsofInitialConditions
30
LinearSectionoftheLiftCurve– 14.54° AoA Inwakeoftrack7
Inwakeoftrack8
JSM– EffectsofInitialConditions
31
Maximumliftandstall• MultiplesolutionsforthesameAoA• Bothapproachesover-predictmaximumliftsignificantly• StallonlypredictedwithfreestreamIC
Liftcoefficientvs.angleofattack(AoA)
JSM– EffectsofInitialConditions
32
• Bothsolutionsmissrootseparationseeninexperiment,over-predictinglift
• UsingfreestreamICleadstoseparationattrack7,betteragreementinliftforwrongreason,wrongstallmechanism
Time-averagedwallshearstressalongthestream-wisedirection(Wss_X)
FreestreamIC Alphacontinuation
Experimentaloilflowimageat21° ofJSM
Tr8Tr7
Tr8Tr7
JSM– EffectsofInitialConditions
33
Interimsummary:• MultiplesolutionsexistforthesameAoA dependingonICs• Alphacontinuationapproachprovidesimprovedflowfieldsolutions• Alphacontinuationisparticularlyeffectiveinlinearpartoftheliftcurve• Freestreaminitialconditionscanleadtooverlyseparatedflow
• Alphacontinuationcanbecomputationallyexpensive,requiresmultiplecomputations
• CanweconvergetothehighliftsolutionifthetransientphaseisnotneglectedwithsteadyRANS,andinsteadweperformaURANSfromfreestreamIC?
JSM– UnsteadyRANS
34
• URANSfromfreestreamICat21.57° AoA
• ∆𝑡# = ∆𝑡𝑐&'#/𝑈* = 0.05 and∆𝑡# = 0.01
• timeindependentsolutionachievedwith∆𝑡# = 0.05
• URANSachievessamesolutionassteadyRANSwithalphacontinuation,forfractionofcost
HiLiftPW-3,DenverCO,June2017 35
JSM– DDESPressureProfiles:Post-Stall
HiLiftPW-3,DenverCO,June2017 36
StartingfromURANSmayormaynotbeOKbecauseseparated.ResolutioninadequateforDDES.
JSM– DDESPressureProfiles:Post-Stall
Conclusion
37
• JSM:MultiplesolutionsexistforthesameAoA dependingonICs• Alphacontinuationapproachagreeswellwithexperimentsperformed
similarly• RANSfromfreestreaminitialconditionscanleadtooverlyseparatedflow• URANSachievessameresultsasalpha-continuationwithsubstantiallyless
cost(forasingleangleofinterest).• PreliminaryDDESperformslightlybetterthanURANSbutmorerefinement
needed• SofarDDESisnotshowingrootstallasseeninexperiments• Willitneedbettertransitionmodeltocapturethiseffect?• HLCRM:coarsegridshowsexcessiveseparationbutmediumandfinein
goodagreement.• Adaptivegridsbeingpursuedtounderstandifcoarsegrid+adaptivity can
getfinegridqualityatlessthanmediumgridcost.• Semi-automaticadaptivity thatpreservessurfacegridanisotropyisshowing
promisetoreducenumberofadaptationcycles.
Acknowledgements
AnawardofcomputertimewasprovidedbytheInnovativeandNovelComputationalImpactonTheoryandExperiment(INCITE)program.ThisresearchusedresourcesoftheArgonneLeadershipComputingFacility,whichisaDOEOfficeofScienceUserFacilitysupportedunderContractDE-AC02-06CH11357.Specifically,theproductionrunsweredoneonMiraandCetuswhilethepost-processingwasdoneonCooley.ThisworkalsoutilizedtheJanus supercomputer,whichissupportedbytheNationalScienceFoundation (awardnumberCNS-0821794)andtheUniversityofColoradoBoulder.TheJanussupercomputerisajointeffortoftheUniversityofColoradoBoulder,theUniversityofColoradoDenverandtheNationalCenterforAtmosphericResearch.Specifically,theseresourceswereusedinmeshgenerationandpre-processing.Finally,wearegratefultoacknowledgeSimmetrix Inc.fortheirmeshingandgeometricmodelinglibraries,Acusim SoftwareInc.(acquiredbyAltairEngineering)fortheirlinearalgebrasolverlibrary,andKitware (ParaView)fortheirvisualizationtools.TheSCOREC-coremeshpartitioningandadaptationtoolsusedinthisresearchweresupportedbytheU.S.DepartmentofEnergy,OfficeofScience,OfficeofAdvancedScientificComputingResearch,underawardDE-SC00066117(FASTMath SciDAC Institute).
38
Questions
39
References
1J.Slotnick,T.Wayman,D.Simpson,andS.Fowler,“HiLiftPW-3:Case2Results.”https://hiliftpw.larc.nasa.gov/Workshop3/HiLiftPW3-Presentations/Summary_Case2.pdf.
40