Computational Analysis of Dual Radius Circulation Control Airfoils

18 Pages·2006·5.87 MB·English

36thAIAAFluidDynamicsConference,June5-8,2006,SanFrancisco,California
Computational Analysis of Dual Radius Circulation Control
Airfoils
E.M.Lee-Rausch∗ V.N.Vatsa† C.L.Rumsey‡
I. Abstract
Thegoaloftheworkistousemultiplecodesandmultipleconfigurationstoprovideanassessmentofthecapability
ofRANSsolverstopredictcirculationcontroldualradiusairfoilperformanceandalsotoidentifykeyissuesassociated
withthecomputationalpredictionsoftheseconfigurationsthatcanresultindiscrepanciesinthepredictedsolutions.
Solutions were obtained for the Georgia Tech Research Institute (GTRI) dual radius circulation control airfoil and
the General Aviation Circulation Control (GACC) dual radius airfoil. For the GTRI-DR airfoil, two-dimensional
structuredandunstructuredgridcomputationspredictedtheexperimentaltrendinsectionalliftvariationwithblowing
coefficient very well. Good code–to–code comparisons between the chordwise surface pressure coefficients and the
solution streamtraces also indicated that the detailed flow characteristics were matched between the computations.
For the GACC-DR airfoil, two-dimensional structured and unstructured grid computations predicted the sectional
lift and chordwise pressure distributions accurately at the no blowing condition. However at a moderate blowing
coefficient,althoughthecode–to–codevariationwassmall,thedifferencesbetweenthecomputationsandexperiment
weresignificant. Computationsweremadetoinvestigatethesensitivityofthesectionalliftandpressuredistributions
tosomeoftheexperimentalandcomputationalparameters,butnoneofthesecouldentirelyaccountforthedifferences
intheexperimentalandcomputationalresults. Thus,CFDmayindeedbeadequateasapredictiontoolfordualradius
CCflows,butlimitedanddifficulttoobtaintwo-dimensionalexperimentaldatapreventsaconfidentassessmentatthis
time.
II. Introduction
Circulation control (CC) technology has great potential for use in the next generation of aeronautical vehicles.
There has also been interest in CC technologies for naval applications. CC not only has the capability to delay or
eliminate separation through boundary layer control, but it also can augment an airfoil’s circulation considerably,
thus enhancing maximum lift. In order to evaluate and design individual CC technologies as well as integrate these
technologiesontoavehicle,accurateandrobustanalysistoolsareneeded. Recentstudieshavefocusedonassessing
and validating the capabilities of computational fluid dynamics (CFD) to predict the performance of different flow
controltechnologiesincludingsyntheticjets1 andCCairfoils.2 TheCFDcodeswerevalidatedthroughcode–to–code
comparisonsaswellasthroughcomparisonwithexperimentaldata. AlthoughthecirculartrailingedgeCCairfoilhas
beenthefocusoftheCCWorkshopandotherCFDstudies,3–6 thedualradiusCCairfoildevelopedbyEnglar7 isalso
of interest for aeronautical applications because it has the potential to alleviate the high cruise drag associated with
thecirculartrailingedgeairfoil. ThedualradiusCCairfoilconceptwastestedbyEnglar7 atGeorgiaTechResearch
Institute (GTRI) and more recently by Jones at NASA LaRC as a modification to a General Aviation Circulation
Control(GACC)airfoil.8
Previous computational studies2,6 indicated that current CFD methods did not predict the circular Coanda flows
robustlyoraccurately. TheseparationlocationontheCoandasurfacewasverysensitivetocomputationalparameters
including turbulence models and tended to overpredict the performance (lift) of the airfoil in comparison with the
∗SeniorMemberAIAA,ResearchEngineerNASALangleyResearchCenter(LaRC),Hampton,Virginia.
†SeniorMember,AIAA,SeniorResearchScientistNASALaRC.
‡AssociateFellowAIAA,SeniorResearchScientistNASALaRC.
ThismaterialisdeclaredaworkoftheU.S.GovernmentandisnotsubjecttocopyrightprotectionintheUnitedStates.2006
1of18
AmericanInstituteofAeronauticsandAstronauticsPaper2006-3012
experiment due to a delay in jet separation. In some cases, the CFD solutions predicted a physically unrealistic
wrapping of the streamlines all the way around the airfoil. There were also questions regarding the precise test
conditionsintheexperimentincludingblowingrateandthree-dimensionaleffects.
Withthedual-radiusCCconfiguration,thesharptrailingedgeeffectivelyfixesseparationandavoidsthepotential
for jet wraparound. Numerical simulations of the dual-radius CC airfoils has been limited. Computational analysis
oftheGTRIdualradiusconfigurationbyLuietal. withatwo-dimensionalstructured–gridcodepredictedtheperfor-
mancecharacteristicsoftheairfoilwell.9 InRef.9,Luietal. studiedtheeffectsofleadingedgeblowing,free-stream
velocity,jet-slotheightandblowingfrequencyontheperformanceoftwo-dimensionalsteadyandpulsedCCjets.The
computationsinRef.9accuratelypredictedthesectionalliftcoefficientvariationwithblowingcoefficientsatthelower
geometricanglesofattack. However,atthehigheranglesofattackthecomputationsstalledearly. Thecurrentwork
willalsofocusonthecomputationalanalysisofthedualradiusCCairfoiltestedbyEnglar7aswellastheGACCdual
radiusairfoiltestedbyJones.8 Code–to–codecomparisonsbetweenthreeReynolds-AveragedNavier-Stokes(RANS)
codes are made as well as comparisons with experimental data. The three codes used for the analysis include two
structuredgridcodes,TLNS3DandCFL3D,andoneunstructuredgridcode,FUN3D.Theprimarygoalofthework
is to use multiple codes and multiple configurations to provide an assessment of the capability of RANS solvers to
robustly predict CC dual radius airfoil performance. A second goal is to identify key issues associated with CFD
analysisofthesetypesofflowcontrolconfigurationsthatcanresultindiscrepanciesinthepredictedsolutions.
III. TestConfigurationsandData
A. GTRIDualRadiusAirfoil
The GTRI CC dual radius airfoil is one of several CC airfoil configurations tested in the GTRI Model Test Facility
(MTF)30inchby43inchsubsonicresearchtunnel.7 Theconfigurationanalyzedinthisstudy(GTRI-DR)isa16%
thicksupercriticalairfoilthathasa7.85inchchordanda30inchspan. Mulipledual–radiusflapanglesofδ = 0◦,
f
30◦, 60◦ and 90◦ from the waterline were tested. In the current CFD study, only the δ = 30◦ configuration is
f
analyzed. Figure1(a)showstheprofileofthebaselinesupercriticalairfoilwhichhasan8.0inchchordalongwiththe
dualradiusflapatthe30◦ flapangle. Notethatthebaselineairfoilchordlengthof8.0inchesisusedasthereference
chord length (C = 8.0). The dual-radius flap to airfoil chord radius is c /C = 0.0955, and the jet slot height to
f
chord ratio h/C = 0.00191. The experiment was performed with free transition. Experimental test data include
forcemeasurements,slotmassflow,jetplenumtotalpressure/temperatureandblowingcoefficient,C . Notethatthe
µ
experimentalblowingcoefficientisdefinedas
m˙V
C = j (1)
µ 1ρ U2 A
2 ∞ ∞
whereρ andU arethefreestreamdensityandvelocity,respectively,andAistheareaoftheairfoilreferencechord
∞ ∞
length (C) multiplied by the span. The experimental mass flow rate values (m˙) were measured using a calibrated
meterintheairsupplylineandthejetvelocity,V ,wascalculatedasanisentropicexpansionfromtheductpressureto
j
thefreestreamstaticpressure. Theexperimentalset-upemployedblowingtreatmentalongthewindtunnelwallsfor
alleviating the model/wall juncture vortical flows. Estimates of angle of attack corrections for wall interference and
streamlinecurvaturehavebeenprovided.
B. GACCDualRadiusAirfoil
The GACC model was designed around the mass flow actuation system and sized to fit into the LaRC BART wind
tunnel.10 The model was originally tested with a circular trailing edge configuration. More recently the model was
retested in the BART tunnel with a dual-radius trailing edge.8 The GACC dual-radius airfoil (GACC-DR) is a 17%
thick supercritical airfoil that has a 10.014 inch reference chord and a 28 inch span. The experimental results were
obtainedoveraMachnumberrangeof0.082to0.1correspondingtodynamicpressureof10psfand15psf,respec-
tively. Thedual-radiusflaptoairfoilchordradiusisc /C = 0.088andthedualradiusflapangleisδ = 57◦ from
f f
thewaterline(SeeFig.1(b)). Thejetslotheighttochordratioh/C =0.00099. (ThejetslotheightnotedinFig.1(b)
is for the circular trailing edge configuration.) The experiment was performed with free transition. Experimental
test data include force measurements from a five–component strain gage balance, centerline steady surface pressure
measurements, centerline PIV measurement for steady and pulsed blowing conditions, slot mass flow, jet plenum
total pressure/temperature and blowing coefficient, C . Note that the experimental blowing coefficient is defined
µ
2of18
AmericanInstituteofAeronauticsandAstronauticsPaper2006-3012
as in Eq. 1. The experiment employed no suction or blowing treatment along the wind tunnel walls for alleviating
the model/wall juncture vortical flows. Note that wall interference corrections have not been applied to any of the
GACC-DRexperimentaldatashowninthispaper.
IV. FlowSolvers
A. FUN3DFlowSolver
FUN3D11–13isafinite-volumeRANSsolverinwhichtheflowvariablesarestoredattheverticesofthemesh.FUN3D
solvestheequationsonmixedelementgrids,includingtetrahedra,pyramids,prisms,andhexahedraandhasnowhas
a two-dimensional path for triangular grids. It employs an implicit upwind algorithm in which the inviscid fluxes
areobtainedwithaflux-difference-splittingschemeofRoeandtheviscoustermsareevaluatedwithafinite-volume
formulation,whichisequivalenttoaGalerkintypeofapproximationfortheseterms.Thisformulationresultsinadis-
cretizationofthefullNavier-Stokestermswithoutanythin-layerapproximationsfortheviscousterms. Atinterfaces
delimitingneighboringcontrolvolumes,theinviscidfluxesarecomputedusingasapproximateRiemannsolverbased
on the values on either side of the interface. For second-order accuracy, interface values are obtained by extrapola-
tionofthecontrolvolumecentroidalvalues,basedongradientscomputedatthemeshverticesusinganunweighted
least-squarestechnique. Thesolutionateachtime-stepisupdatedwithabackwardsEulertime-differencingscheme.
At each time step, the linear system of equations is approximately solved with either a point implicit procedure or
an implicit line relaxation scheme.14 Local time-step scaling is employed to accelerate convergence to steady-state.
FUN3D is able to solve the RANS flow equations, either tightly or loosely coupled to the Spalart-Allmaras15 (S-A)
one-equationturbulencemodel. Inthisinvestigationallcomputationsarelooselycoupledandassumefullyturbulent
flow.
B. TLNS3DFlowSolver
TheTLNS3Dcodeisamultiblockstructuredgridsolverthatutilizesthegeneralizedthin-layerNavier-Stokesequa-
tionsasthegoverningequations.16 Thespatialtermsarediscretizedusingthecell-centeredfinitevolumeschemewith
artificialdissipationaddedforstability. TimeisdiscretizedinafullyimplicitsensebyusingeitheramultistepBDF
ormultistageRunge-Kuttascheme. Theresultantnonlinearalgebraicequationsaresolvediterativelyinpseudo-time
withamultigridaccelerationusedtospeeduptheconvergence. TheSpalart-Allmarasturbulencemodelisalsoused
inthisinvestigationandallcomputationsassumefullyturbulentflow.
C. CFL3DFlowSolver
CFL3Disastructured-gridupwindmulti-zoneCFDcodethatsolvesthegeneralizedthin-layerNavier-Stokesequa-
tions.17 Itcanusepoint-matched,patched,oroversetgridsandemployslocaltime-stepscaling,gridsequencingand
multigrid to accelerate convergence to steady stage. A time-accurate mode is available, and the code can employ
low-Machnumberpreconditioningforaccuracyincomputinglow-speedsteady-stateflows. CFL3Disacell-centered
finite-volumemethod.Itusesthird-orderupwind-biasedspatialdifferencingontheconvectiveandpressureterms,and
second-order differencing on the viscous terms; it is globally second-order accurate. Roe’s flux difference-splitting
(FDS)method18 is used to obtainfluxes atthe cellfaces. The solutionis advancedin timewith an implicit approx-
imatefactorizationmethod. Awidevarietyofeddy-viscosityturbulencemodelsareavailableinthecode, including
nonlinearmodels. BoththeSpalart-AllmarasturbulencemodelandtheMenterSSTmodel19 wereusedinthisstudy.
Allcomputationsareassumedfullyturbulent.
V. ComputationalResults
A. DualRadiusAirfoilGridsandBoundaryConditions
The2Dmulti-blockstructuredgridsfortheGTRI-DRandGACC-DRweregeneratedwiththeGridgencommercial
softwarepackage. PartialviewsoftheGTRI-DRairfoilstructuredgridareshowninFig.2. Thegridhas39blocks
andatotalof130,304cells. Thereare529pointsontheairfoilexteriorno-slipsurface. Theouterboundaryislocated
30 chord lengths away from the airfoil. In the wall normal direction, the grid spacing is set such that an average
normalizedcoordinatey+ islessthanoneforthefirstgridcellatthewallforthemainairfoil. Thevaluesofy+ on
thedual-radiusflaparelessthanonefortheno–blowingcasebutincreasesaboveoneatthehigherblowingrates. The
3of18
AmericanInstituteofAeronauticsandAstronauticsPaper2006-3012
effectofthehighery+ valuesonthesolutionwillbediscussedlaterinthepaper. Aclose–upviewofthestructured
gridjetslotregionoftheGTRI-DRairfoilinFig.2(b)showsthatthejetplenumchamberismodeledfarinsideofthe
slot. Totaltemperatureandpressureboundaryconditionsareappliedattheupstreamwallofthechambertomatchthe
measuredexperimental conditions. In thecomputation, given total pressure ratio, total temperature, and flow angle,
the pressure is extrapolated from the interior of the domain, and the remaining variables are determined from the
extrapolated pressure and the input data, using isentropic relations. No-slip boundary conditions are applied to all
otherairfoilsurfacesexcepttheblunttrailingedgeatthejetslotlipandontheflap, whereslipboundaryconditions
wereused.
The2DunstructuredtriangulargridfortheGTRI-DRwasgeneratedwiththeAFLR2advancing-frontgridgener-
ationcode.20 PartialviewsoftheGTRI-DRairfoilunstructuredgridareshowninFig.3. Thegridhas110,164nodes
with935nodesontheno-slipairfoilsurface. Theouterboundaryislocated20chordlengthsawayfromtheairfoil. In
thewallnormaldirection,thegridspacingissetsuchthatonthemainairfoilanaveragenormalizedcoordinatey+ is
lessthanoneforthefirstnodeoffthewall. Aswiththestructuredgridresults,thevaluesofy+onthedual-radiusflap
increaseaboveoneatthehigherblowingrates. Acloseupviewofthejetslotregionintheunstructuredgridisalso
shown in Fig. 3(b). Static temperature and normal mass flow boundary conditions are applied at the upstream wall
ofthechambertomatchtheexperimentaltotalpressureratio,totaltemperatureconditionsintheinletoftheplenum.
No-slipboundaryconditionsareappliedtoallotherairfoilsurfaces.
ThegridsfortheGACC-DRairfoilweredevelopedinasimilarmanner. Thestructuredgridhas34blocksanda
totalof363,150cells.Thereare550pointsontheairfoilexteriorno-slipsurface.Acloseupviewofthestructuredgrid
jetslotregionoftheGACC-DRairfoilinFig.4(a)showsthatthejetplenumchamberismodeledfarinsideoftheslot.
TheGACC-DRunstructuredgridhas142,642nodeswith2,050nodesontheno-slipairfoilsurface. Acloseupview
ofthejetslotregionintheunstructuredgrid,whichextendsfartherupstreamthanthestructuredgrid,isalsoshownin
Fig.4(b). SimilartotheGTRI-DRconfigurations,statictemperatureandnormalmassflowboundaryconditionsare
appliedattheupstreamwallofthechambertomatchtheexperimentaltotalpressureratio,totaltemperatureconditions
intheinletoftheplenum.
B. GTRIDualRadiusAirfoil
The GTRI-DR airfoil was analyzed over a range of blowing coefficients corresponding to the experimental 0◦ ge-
ometric angle of attack. The freestream conditions for this configuration correspond to a Mach number of 0.0842
and a chord Reynolds number of 0.37 million. In this study, all calculations are performed with the SA turbulence
model. The angle of attack corrections range from −0.005◦ at C = 0 to −0.056◦ at C = 0.374. Note that for
µ µ
all computations in this study, the computational blowing coefficient is computed using the same definitions as the
experimentallyderivedvalue(SeeEq.1). Thevariationofexperimentalandcomputationalsectionalliftcoefficient,
C ,withblowingcoefficientfortheGTRI-DRisshowninFig.5(a). ComputationswereperformedwithFUN3Dand
l
CFL3DatthejetplenumconditionscorrespondingtotheexperimentalblowingcoefficientsC =0,0.11,0.22,0.32
µ
and0.37(CFL3Donly). AsindicatedinFig.5(a),boththeFUN3DandCFL3Dcomputationspredicttheexperimen-
taltrendinliftcoefficientvariationwithC verywell. Notethatwhenthetotalpressureandtemperatureconditions
µ
arematchedbetweentheexperimentandcomputation, thecomputationaljetmassflowrateforthatconditionisnot
consistentwiththeexperimentallymeasuredmassflowrate,whichresultsinanoffsetinC forthecomputationaland
µ
experimentaldatapointsinFig.5(a). Thereasonforthisdiscrepancyisnotknown. Figure5(b)alsoshowsC vs.C
l µ
fortheGACC-DRairfoiltomakedirectcomparisonbetweenthetwoairfoilseasier. TheseGACC-DRresultswillbe
discussedinthenextsection.
The good code-to-code comparison of sectional lift coefficient for the GTRI-DR indicates that both codes are
predicting similar performance characteristics for the airfoil. Code-to-code comparisons of chordwise pressure co-
efficients at multiple blowing coefficients in Figs. 6, 7, and 8 indicate how well the detailed flow characteristics are
matchedbetweenthecomputations. AttheC = 0conditionshowninFig.6,theFUN3DandCFL3Dsurfacepres-
µ
sure results compare very well over the entire airfoil. Similarly at C = 0.11 shown in Fig. 7, the computational
µ
surfacepressureresultscompareverywellovertheentireairfoil. AtC = 0.32conditionshowninFig.8,thereare
µ
slightdifferencesintherecompressionareaontheflapuppersurface. TheFUN3Dresultsshowakinkinthesurface
distributionontheupperflapattheinterfaceofthedualradiussurfacesthatisnotpresentintheCFL3Dsolutions.
AstheC valuesincrease,themaximumy+ valuesontheupperflapsurfacealsoincreasedinthejetregionover
µ
theflaptoalmostfourintheCFL3DsolutionandalmostsevenintheFUN3Dsolution. Theeffectofthelargery+
valueswasstudiedbygeneratinganewunstructuredgridwiththesamesurfacepointdistributionsbutwiththewall
spacingreducedbyafactoroffour. FUN3DresultsfortheC = 0.32caseonthenewmeshwithrefinementinthe
µ
normalwallspacingisshowninFig.9incomparisonwiththeCFL3DresultsfromFig.8. Themaximumy+ value
4of18
AmericanInstituteofAeronauticsandAstronauticsPaper2006-3012
ontheupperflapfortheFUN3Dsolutionisreducedtojustbelowtwo. Thesectionalliftcoefficientincreasesslightly
from C = 3.79 to 3.84 with the normal grid spacing refinement. The kink in the pressure distribution is reduced
l
and the code-to-code comparison of surface pressure is improved on the flap. As noted in Fig. 9 the code-to-code
comparisonofsectionalliftcoefficientisalsoslightlyimproved.
Code-to-codecomparisonsofstreamtracesatmultipleblowingcoefficientsinFig.10(a),(b),and(c)alsoindicate
how well the detailed flow characteristics are matched between the computations. In Fig. 10(a), both codes predict
theseparationontheflapatC = 0. AtC = 0.11inFig.10(b), bothcodespredictsimilarturningoftheleading
µ µ
andtrailingedgestagnationstreamlineswithasmallareaofseparationinthelowerflapcove. Atthehigherblowing
coefficient C = 0.32 in Fig. 10(c), both codes predict the increase in the turning of the leading and trailing edge
µ
stagnationstreamlineswhichproducestheincreasedsectionalliftcoefficient.
C. GACCDualRadiusAirfoil
TheGACC-DRairfoilwasanalyzedatthe10psfexperimentalconditionwhereamajorityofthePIVandhotwiredata
weretaken. ThisdynamicpressurecorrespondstoafreesteamMachnumberof0.0824andachordReynoldsnumber
of0.47million.Themajorityofthedataweretakenata0◦geometricangleofattack.Notethatforallcomputationsin
thisstudy,thecomputationalblowingcoefficientiscomputedusingthesamedefinitionsastheexperimentallyderived
value(SeeEq.1). AllcalculationsareperformedwiththeSAmodelunlessotherwisenoted.
Althoughnoangleofattackcorrectionswereappliedtotheexperimentaldata,pastexperiencewithCCairfoildata
ledtheauthorstoinitiallyinvestigatetheeffectsofinducedangleofattackfromthewindtunnelwall/modeljuncture
vortices. Thefollowingformulabasedongeometricangleofattack,α ,andtheexperimentallymeasuredsectional
geo
liftcoefficient,C ,estimatesaneffectiveangleofattack,α =α −1.5C . Figure11showsacomparison
lexp eff geo lexp
ofexperimentalchordwisepressurecoefficientswiththeFUN3Dcalculationsatthegeometricandestimatedeffective
angles of attack for the no blowing condition, C = 0. Figure 12 shows a similar comparison for C = 0.09.
µ µ
Thecorrespondingexperimentalandcomputationalliftcoefficientsarealsoincludedintheplots. Thecomputational
results indicate that the angle of attack correction provides a rational level of adjustment to achieve agreement with
themeasureduppersurfacepressurepeaksandsectionalliftcoefficient.
The variation of experimental and computational sectional lift coefficient, C , with blowing coefficient for the
l
GACC-DR is shown in Fig. 5(b) with the computations being performed at the estimated effective angles of attack.
Evenwiththeestimatedangle-of-attackcorrections,thecomputedsectionalliftcoefficientatC = 0.09aresignifi-
µ
cantlyhigher(byabout30%)thantheexperimentalvalue. Figure 12alsoindicatesthatwithjetblowingtheCFDis
predictingtoohighalevelofcirculation,yieldingpressurelevelsthataresomewhatlowovertheentireupperairfoil
surface.
Computations were made to investigate the sensitivity of the sectional lift coefficient and pressure distributions
tosomeoftheexperimentalparameters. FUN3Dcomputationstoassessthesensitivityoftheliftcoefficienttomass
blowing rate indicate that a drop of 10% in mass flow will result in a 10% decrease in lift coefficient. FUN3D
computations investigating the effects of jet slot expansion under pressure indicate that a 0.003 inch expansion of
the slot width (30% increase) will reduce the lift coefficient to 3.76. This is an 8.25% decrease from the baseline
configuration. Computations modeling the upper and lower wind tunnel walls indicated a negligible effect on the
airfoilliftcoefficientandpressuredistributions.
Computations were also made to investigate the sensitivity of the sectional lift coefficient and pressure distribu-
tions to some of the numerical/computational parameters. Figure 13 shows a comparison of the chordwise pressure
distributionsbetweenthethreeCFDcodeswithnoblowingfortheexperimentalα =0◦angleofattack. Theex-
geom
perimentaldataisincludedforcomparison.Thereislittlecode-to-codevariationforthisconditionexceptontheupper
surfaceofthedual-radiusflap. Allcodestendtoprematurelyseparateonthelowersurfaceoftheairfoil. Thismaybe
duetoadifferencebetweenthelocationoftransitionintheexperimentandthedelayinthedevelopmentofturbulent
eddy viscosity that that occurs for ”fully turbulent” low Reynolds number computations. Figure 14 compares CFD
resultsfromthethreecodesforablowingcoefficientofC = 0.09andacorrectedangleofattackof−4.62◦. Like
µ
theno-blowingcase,the3codespredictveryconsistent,similarresultscomparedtoeachother. Figure15compares
resultsusingtwodifferentturbulencemodelsforthesamecase. AlthoughSSTpredictsasomewhatalowerliftcoef-
ficientthanSA,theoverallcirculationisstillsignificantlyhighcomparedtoexperiment. AswiththeGTRI-DRwhen
theC valuesincrease,themaximumy+ valuesontheupperflapsurfacealsoincreaseinthejetregionovertheflap.
µ
Theeffectofthelargery+valueswasstudiedontheGACC-DRbygeneratinganewunstructuredgridwiththesame
surfacepointdistributionsbutwiththewallspacingreducedbyafactorofsix. Forthenewmeshwithrefinementin
thenormalwallspacing,themaximumy+valueontheupperflapfortheFUN3DsolutionatC =0.09isreducedto
µ
5of18
AmericanInstituteofAeronauticsandAstronauticsPaper2006-3012
justbelowone. ThesectionalliftcoefficientdidnotchangefromC =4.00withthenormalgridspacingrefinement,
l
andtherewerenosignificantchangesinthechordwisepressurecoefficient.
ItisnotclearatthistimewhattherootcauseofthediscrepancybetweenCFDandtheGACCexperimentaldatais.
Althoughsomeoftheexperimentalandnumerical/computationalparametersinvestigatedloweredtheliftcoefficient,
nosingleeffectaccountedforthetotaldifferenceinthecomputationalandexperimentalliftcoefficient.Unfortunately,
therearenoexperimentalvelocityprofilesavailableinthejetexitregiontoserveasacheckonhowwelltheCFDjet
boundaryconditionsarematchingexperiment. However,comparisonscanbemadeofvelocityprofilesinthejetjust
below the trailing edge, as shown in Fig. 16. The CFD solutions predict a core jet velocity nearly twice as high as
experiment. Thus,eitherthejetexitconditionsareverydifferent,orelsethecomputedjetisnotdiffusingasquickly
asintheexperiment.Ifpresent,thislattereffectcouldbecausedbyadeficiencyintheturbulencemodeling.However,
itseemsunlikelythattheprimarycauseofthelargediscrepancyhereisturbulencemodeling,becausetheSAmodel
didanexcellentjobpredictingresultsfortheGTRIairfoil.
VI. Summary
MultiplecodesandmultipleconfigurationswereusedtoprovideanassessmentofthecapabilityofRANSsolvers
topredictCCdualradiusairfoilperformanceandalsotoidentifykeyissuesassociatedwiththeCFDpredictionsof
these configurations that can result in discrepancies in the predicted solutions. The GTRI-DR airfoil was analyzed
overarangeofblowingcoefficientscorrespondingtotheexperimental0◦geometricangleofattack.Two-dimensional
structured and unstructured grid computations predicted the experimental trend in sectional lift coefficient variation
with C very well. Good code–to–code comparisons between the chordwise surface pressure coefficients and the
µ
solution streamtraces also indicated that the detailed flow characteristics were matched between the computations.
Therewasnoexperimentalsurfacepressurecoefficientsoroff–surfaceflowmeasurementsavailableforcomparison
withthecomputations. Althoughthewallnormalgridspacingwassufficientlysmallforthenoblowingcases,asthe
blowingcoefficientwasincreased,thevaluesofy+onthedual-radiusflapincreasedaboveone.However,arefinement
oftheunstructuredgridnormal–wallspacingdidnothaveasignificanteffectonthefinalsolutionintermsofsurface
pressuresorintegratedliftcoefficient.
For the GACC-DR airfoil, code–to–code comparisons between three RANS codes were made as well as com-
parisons with experimental data. Two-dimensional steady blowing and no blowing computations were performed,
and the lift coefficient and chordwise pressure distributions were compared with the experimental data. The effects
ofangleofattackcorrectionswerequantifiedaswellasthesensitivitytoturbulencemodel. Thecomputationaland
experimentalsectionalliftcoefficientandchordwisepressuredistributionscomparedverywellatC =0.Howeverat
µ
C =0.09,althoughthecode–to–codevariationwassmall,thedifferencesbetweenthecomputationsandexperiment
µ
weresignificant. Thecomparisonsofliftcoefficientandchordwisepressuredistributionbetweentheexperimentand
thecomputationsindicatethatthecirculationdevelopedaroundtheairfoilinthecomputationismuchhigherthanthat
of the experiment. Comparisons were made of velocity profiles in the jet just below the trailing edge that indicated
thattheCFDsolutionspredictedacorejetvelocitynearlytwiceashighasthatintheexperiment. Thus,eitherthejet
exitconditionsareverydifferent,orelsethecomputedjetisnotdiffusingasquicklyasintheexperiment.
FortheGTRI-DRairfoil, theRANScodesaccuratelypredictedtheeffectofthejetblowingontheperformance
characteristicsoftheairfoil. However,fortheGACC-DRairfoil,theRANScodesover-predictedthejetblowingef-
fect.ThecauseforthediscrepanciesintheGACC-DRpredictionsisunknown.Computationsweremadetoinvestigate
thesensitivityofthesectionalliftcoefficientandpressuredistributionstosomeoftheexperimentalandcomputational
parameters, but none of these could entirely account for the differences in the experimental and computational re-
sults. Itisdifficulttoensurethatthecomputationismodelingthesamejetexitconditionsastheexperimentwithout
measurementsofthevelocityprofileatthejetexit. Alsotrulytwo-dimensionalexperimentsareextremelydifficultto
performforhigh-liftflows,particularlybecauseoftheeffectsofstrongwall-juncturevorticalstructuresthatinfluence
the entire flowfield circulation. In the GTRI-DR experiment, side–wall blowing was used to alleviate this influence
whereasintheGACC-DRexperimentnoside–wallcontrolwasemployed. Thus,CFDmayindeedbeadequateasa
predictiontoolfordualradiusCCflows,butlimitedanddifficulttoobtaintwo-dimensionalexperimentaldataprevents
aconfidentassessmentatthistime.
6of18
AmericanInstituteofAeronauticsandAstronauticsPaper2006-3012
VII. Acknowledgments
TheauthorswouldliketothankDr. GregoryS.JonesofNASALangleyResearchCenterforprovidingtheGACC
dual-radiusexperimentaldataandProfessorRobertJ.EnglaroftheGeorgiaTechResearchInstituteforprovidingthe
GTRIdual-radiusairfoildefinitionandexperimentaldata.TheauthorswouldalsoliketothankMr.ScottG.Andersof
theNASALangleyResearchCenterforhismanyhelpfuldiscussionregardingthetestingandanalysisofcirculation
controlairfoils.
References
1Rumsey,C.L.,Gatski,T.B.,SellersIII,W.L.,Vatsa,V.N.,andViken,S.A.,“Summaryofthe2004CFDValidationWorkshoponSynthetic
JetsandTurbulentSeparationControl,”AIAAPaper2004–2217,2004.
2Jones,G.S.andJoslin,R.D.e.,ProceedingsoftheCirculationControlWorkshopMarch2004,NASACP-2005-213509,2005.
3Slomski,J.F.,Gorski,J.J.,Miller,R.W.,andMarino,T.A.,“NumericalSimulationofCirculationControlAirfoilsasAffectedbyDifferent
TurbulenceModels,”AIAAPaper2002–0851,2002.
4Paterson,E.G.andBaker,W.J.,“SimulationofSteadyCirculationControlforMarine-VehicleControlSurfaces,”AIAAPaper2004–0748,
2004.
5Chang,P.A.,Slomski,J.F.,Marino,T.A.,andEbert,M.P.,“NumericalSimulationofTwo-andThree-DimensionalCirculationControl
Problems,”AIAAPaper2005–0080,2005.
6Swanson,R.C.,Rumsey,C.L.,andAnders,S.G.,“ProgressTowardsComputationalMethodforCirculationControlAirfoils,”AIAAPaper
2005–0089,2005.
7Englar,R.J.,Smith,J.J.,Kelly,S.M.,andRoverIII,R.C.,“ApplicationofCirculationControltoAdvancedSubsonicTransportAircraft,
PartI:AirfoilDevelopment,”JournalofAircraft,Vol.31,No.5,Sept.1994,pp.1160–1168.
8Jones,G.S.,Yao,C.S.,andAllan,B.G.,“ExperimentalInvestigationofa2DSupercriticalCirculation-ControlAirfoilUsingParticle
ImageVelocimetry,”AIAAPaper2006–3009,2006.
9Lui,Y.,Sankar,L.N.,Englar,R.J.,Ahuja,K.K.,andGaeta,R.,“ComputationalEvaluationoftheSteadyandPulsedJetEffectsonthe
PerformanceofaCirculationControlWingSection,”AIAAPaper2004–0056,2004.
10Jones,G.S.,Viken,S.A.,Washburn,A.E.,andCagle,C.M.,“AnActiveFlowCirculationControlledFlapConceptforGeneralAviation
AircraftApplications,”AIAAPaper2002–3157,2002.
11Anderson,W.K.andBonhaus,D.L.,“AnImplicitUpwindAlgorithmforComputingTurbulentFlowsonUnstructuredGrids,”Computers
andFluids,Vol.23,No.1,1994,pp.1–22.
12Anderson,W.K.,Rausch,R.D.,andBonhaus,D.L.,“Implicit/MultigridAlgorithmsforIncompressibleTurbulentFlowsonUnstructured
Grids,”JournalofComputationalPhysics,Vol.128,No.2,1996,pp.391–408.
13Nielsen, E. J., Aerodynamic Design Sensitivities on an Unstructured Mesh Using the Navier-Stokes Equations and a Discrete Adjoint
Formulation,Ph.D.thesis,VirginiaPolytechnicInstituteandStateUniversity,Blacksburg,VA,1998.
14Nielsen,E.J.,Lu,J.,Park,M.A.,andDarmofal,D.L.,“AnImplicit,ExactDualAdjointSolutionMethodImplicit,ExactDualAdjoint
SolutionMethodforTurbulentFlowsonUnstructuredGrids,”ComputersandFluids,Vol.33,No.9,2004,pp.1131–1155,SeealsoAIAAPaper
2003–0272.
15Spalart,P.R.andAllmaras,S.R.,“One-EquationTurbulenceModelforAerodynamicFlows,”AIAAPaper92–0429,1992.
16Vatsa,V.N.andWedan,B.W.,“Developmentofamultigridcodefor3-DNavier-StokesEquationsandItsApplicationtoaGrid-Refinement
Study,”Vol.18,No.18,Jan.1990,pp.391–403.
17Krist,S.L.,Biedron,R.T.,andRumsey,C.L.,CFL3DUser’sManual(Version5.0),NASATM-1998-208444,1998.
18Roe,P.L.,“ApproximateRiemannSolvers,ParameterVectors,andDifferenceSchemes,”JournalofComputationalPhysics,Vol.43,1981,
pp.357–372.
19Menter, F. R., “Two-Equation Eddy-Viscosity Turbulence Models for Engineering Applications,” AIAA Journal, Vol. 32, No. 8, 1994,
pp.1598–1605.
20Marcum,D.L.,“GenerationofUnstructuredGridsforViscousFlowApplications,”AIAAPaper95-0212,1995.
7of18
AmericanInstituteofAeronauticsandAstronauticsPaper2006-3012
8.0"
CCSloth/C=0.00191
30deg
DualRadiusFlap
BaselineAirfoil
(b)GTRI-DRairfoilprofile
(a)GACC-DRairfoilprofile(notethattheCCslotheightisforthecirculartrailingedgeconfiguration)
Figure1. DualRadiusairfoilprofiles.
8of18
AmericanInstituteofAeronauticsandAstronauticsPaper2006-3012
1
0.5
0
Y
-0.5
-1
-1.5
-0.5 0 0.5 1 1.5 2 2.5
X
(a)Partialviewofgrid
0.1
0
Y
-0.1
-0.2
0.7 0.8 0.9 1 1.1
X
(b)Closeupviewofgridnearjetslotandtrailingedge
Figure2. GTRI-DRairfoilstructuredgrid.
9of18
AmericanInstituteofAeronauticsandAstronauticsPaper2006-3012
(a)Partialviewofgrid
(b)Closeupviewofgridnearjetslotandtrailingedge
Figure3. GTRI-DRunstructuredgrid.
10of18
AmericanInstituteofAeronauticsandAstronauticsPaper2006-3012