LAVA Contribution to HiLiftPW-3 · • Trailing edge had 25 points instead of the 9 specified in the document • Spanwisespacing at the root and tip of elements was based on local

LAVAContributiontoHiLiftPW-3

JamesJensen,MarieDenison,Gerrit Stich,JeffreyHousman,andCetinKiris

ComputationalAerosciences BranchNASAAmesResearchCenter

PID033

3rd HighLiftPredictionWorkshopDenver,COJune3-4,2017

Summaryofcasescompleted:- LAVA,OversetStructured,SA-QCR2000

Case Alpha=8,Fully turb,grid

study

Alpha=16,Fully turb,grid

study

1a(fullgap) yes yes

1c(partialseal) yes yes

HiLiftPW-3,DenverCO,June2017 2

Case Polar, Fullyturb Polar,specifiedtransition

Polar,wtransitionprediction

2a (nonacelle) yes no no

2c (withnacelle) yes no no

- LAVA,UnstructuredPolyhedral,SACase Polar, Fullyturb Polar,specified

transitionPolar,wtransitionprediction

2c (withnacelle) yes no no

Case 2DVerificationstudy

3 yes

Launch Ascent and Vehicle Aerodynamics (LAVA) Framework

Far FieldAcoustic Solver

Structural Dynamics

ObjectOrientedFramework

DomainConnectivity/SharedDataC++/FortranwithMPIParallelism

LAVA

Multi-Physics:Multi-PhaseCombustionChemistryElectro-Magnetics……

6 DOF Body Motion

Post-ProcessingTools

Conjugate Heat Transfer

Other Solvers& Frameworks

Not Yet Connected

Connected Existing

Future3

Framework

Developing

Other Development Efforts• Higher order methods• Curvilinear grid generation• Wall modeling• LES/DES/ILES Turbulence• HEC (optimizations, accelerators,

etc) Kiris at al. AST-2016 and AIAA-2014-0070

Prismatic Layers

Structured Curvilinear

Navier-Stokes

Unstructured Arbitrary Polyhedral

Navier-Stokes

Structured Cartesian AMR

Navier-Stokes

LatticeBoltzmann

Actuator DiskModels

• High quality body fitted grids • Low computational cost• Reliable higher order methods

• Grid generation largely manual and time consuming

• Essentially no manual grid generation

• Highly efficient Structured • Adaptive Mesh Refinement (AMR)

• Low computational cost• Reliable higher order methods• Non-body fitted -> Resolution of boundary layers inefficient

• Partially automated grid generation

• Body fitted grids • Grid quality can be challenging• High computational cost• Higher order methods yet to fully mature

LAVA Computational Grid Paradigms

4

Structured Cartesian AMR

Unstructured Arbitrary Polyhedral

Structured Curvilinear

• 2nd order cell centered• MUSCL Scheme• AUSMPW+ convective flux function• BDF2 with GMRES linear solver (PETSc)• SA turbulence model

LAVA Solver Details

5

Unstructured Arbitrary Polyhedral Structured Curvilinear

HiLiftPW-3,DenverCO,June2017

AllsimulationsarerunwithRANSusingcoldstartswithoutCFLramping

• 2nd order accurate Modified Roe Flux Difference Splitting for the convective terms

• 2nd order central differencing for the viscous terms

• SA turbulence model with QCR-2000 for viscous flux tensor

Briefoverviewofgridsystem(s)

OversetGridforCase2c 6

GridSystem Case(s) If committeegrid,reportanyproblems/issuesIfusergrid,reasonforgeneratinggridsystem

Committee(GridA,StructuredOverset) 1a and1c Observedsolutiondecouplinginunexpectedareasofthegrid

Committee(GridA,StructuredOverset) 2aand2c Usedcustom domainconnectivitywithcommitteegrids

User(UnstructuredPolyhedral) 2c Customgeneratedwithdifferentwakerefinements

UnstructuredGridforCase2c

ProvidedtheCommitteeOversetGridsforCase2aandCase2c• GeneratedwiththeChimeraGridTools(CGT)• Thegridwasgeneratedasamediumgridusingtheparametersfromthegriddingguidelinesdocumentprovidedbythecommittee

• Trailingedgehad25pointsinsteadofthe9specifiedinthedocument• Spanwise spacingattherootandtipofelementswasbasedonlocalcellsizes• Usedseparatewakemeshesratherthanaregionofuniformspacinginthevolumes

HiLiftPW-3,DenverCO,June2017 7

Case2a

Case2c

ReasoningBehindWakeGrids

HiLiftPW-3,DenverCO,June2017 8

WakeStreamlinesFinal“Geometric”WakeGrids

• Decidedtousea“geometric”basedwakegrid

• Originalplanwastogenerateanwakegridbasedonthestreamlines• Thestreamlinesvarygreatlywithangleofattacksothestreamlinebasedwakeswerenotgeneral

OversetStructuredandUnstructuredPolyhedralMeshParadigms• Closelookattheslat-wingjunction

HiLiftPW-3,DenverCO,June2017 9

LAVAModifiedImplicitHoleCutting(MIHC)Procedure

HiLiftPW-3,DenverCO,June2017 10

• FirsttheDCFroutinesinOVERFLOWareusedtoprovideaminimumholecutandensurethattherequirednumberoffringelayersareavailable(withminimumhole)fortheflowsolver(doublefringeinthiscase)

• Theglobalwall-distanceisalsocomputedandstoredtofile• NexttheMIHCcodeisappliedtotheoversetgridsystemwiththeminimumholescut

MinimumHole FinalHole

WakeGrids• Thedifferentconnectivitycodesincorporatedthewakegridsdifferently

HiLiftPW-3,DenverCO,June2017 11

MIHCWakeGrids Peg5WakeGrids

WhatisSolutionDecoupling?

HiLiftPW-3,DenverCO,June2017 12

• Somesolutiondecouplingisexpectedwithintheareasofoverlapbetweengridsbecausethesurfacepointsareslightlydifferent(notpointmatched)

• Thisdecouplingissomethingthatreducesasthemeshisrefined

Coarse

Medium Fine

HL-CRM(Case1)TopologyDifference

HiLiftPW-3,DenverCO,June2017 13

• SawlargerthanexpectedsolutiondecouplingintheslatCp plots

• Firstsuspicionwasthatthesurfaceswerenotidenticalbuttheyweremadebysplittingonesurfacemesh

• Differentvolumegridgenerationmethods

• Leadstoadifferentamountofsolutiondecouplingatthegridinterfaces

HL-CRM(Case1a)GridConvergenceEffect

HiLiftPW-3,DenverCO,June2017 14

• Liftisconvergingwellatbothanglesandhasconvergedtoanalmostidenticalvaluefrombothsolversatbothangles

• Dragisconvergedat⍺=8andisconvergingat⍺=16• Thesolversarewithin10dragcountsofeachotheratbothanglesofattack

JSM(Case2a)ForceandMomentComparisons

15

JSM(Case2a)Cp ComparisonsD-D

HiLiftPW-3,DenverCO,June2017 16

• Atbothanglesofattacktheflapisfairlydifferentbetweenthetwosolvers

• At⍺=18.58° thedifferenceinthesolutionsattheflapbecomesevenmorepronounced

• OverflowappearstomatchthepeakbetterandLAVAappearstomatchbetterintherecoveryregion

JSM(Case2a)Cp ComparisonsG-G• Thedifferencesbetweenthesolversismorepronouncedatthisstation

• Atthehigherangleofattackthedifferencesbecomemorepronounced

• OntheflapLAVAisclosertotheexperimentatthepeakandOverflowiscloserattheTE

HiLiftPW-3,DenverCO,June2017 17

JSM(Case2c)ForceandMomentComparisons

18

JSM(Case2c)Cp ComparisonsD-D

HiLiftPW-3,DenverCO,June2017 19

• AtthelowerangleofattackOverflowismorecloselymatchingtheexperimentaldatabutatthehigherangleofattackLAVAmorecloselymatchestheexperiment

• OntheslatOverflowandtheLAVAUnstructuredresultareclosertotheexperimentatbothanglesofattack

JSM(Case2c)Cp ComparisonsG-G• Atthelowangleontheflapweseeasimilartrendasbefore;OverflowisclosertotheexperimentnearthepeakbutLAVAiscloserintherecoveryregion

• Atthehighangleofattackallofthesolvershavesimilarpredictions

HiLiftPW-3,DenverCO,June2017 20

Summary• GeneratedthecommitteeoversetgridsystemfortheJSMmodel(used“geometric”wakegrids)

• ObservedsomeunexpectedsolutiondecouplinginCase1duetothewaythegridshadbeengenerated

• LAVAforceresultsagreedwellwiththeOverflowresultsforCase1

• TheCp andForceresultsaresimilarbetweenallofthesolversforbothCase2aandCase2c(flapisareaoflargestdifferences)

• NeedtoexploretheinfluenceoftheRCaswellastransitionmodels

HiLiftPW-3,DenverCO,June2017 21

Acknowledgements• TransformationalToolsandTechnologies(T3)projectunderAeronauticsResearchMissionDirectorate(ARMD)

• NASfacilityforcomputertimeonPleiades

HiLiftPW-3,DenverCO,June2017 22