NASA Technical Reports Server (NTRS) 20160007544: The NASA Juncture Flow Experiment: Goals

9.4 MB·English

by  NASA Technical Reports Server (NTRS)

The NASA Juncture Flow Experiment: Goals, Progress, and
Preliminary Testing (Invited)
ChristopherL.Rumsey∗,DanH.Neuhart†,MichaelA.Kegerise‡
NASALangleyResearchCenter,Hampton,VA23681-2199
NASA has been working toward designing and conducting a juncture flow experiment on a wing-body
aircraftconfiguration. Theexperimentisplannedtoprovidevalidation-qualitydataforCFDthatfocuseson
theonsetandprogressionofaseparationbubblenearthewing-bodyjuncturetrailingedgeregion.Thispaper
describesthegoalsandpurposeoftheexperiment.Althoughcurrentlyconsideredunreliable,preliminaryCFD
analysesofseveraldifferentconfigurationsareshown. Theseconfigurationshavebeensubsequentlytestedin
a series of “risk-reduction” wind tunnel tests, in order to help down-select to a final configuration that will
attainthedesiredflowbehavior. Therisk-reductiontestingatthehigherReynoldsnumberhasnotyetbeen
completed(atthetimeofthiswriting),butsomeresultsfromoneofthelow-Reynolds-numberexperimentsare
shown.
I. Introduction
Geometric junctures (corners) are common features on most aerospace vehicles. Wing-body juncture flow is of
particular interest to aircraft design engineers, because non-streamlined flow behavior such as separation bubbles in
thisareacanhaveanadverseimpactonaircraftperformance. Thereisconsequentlyaneedtobeabletoaccurately
predictthebehavioroftheflowinjunctureregionswhenusingcomputationalfluiddynamics(CFD).
Inrecentyears,therehasbeensignificantconfusionsurroundingthecomputationofseparatedjunctureflows. In
thethirdDragPredictionWorkshop(DPW-3),1 whichusedtheDLR-F6wing-bodygeometry,participantspredicted
averywiderangeofside-of-bodyseparationbubblesizeswithCFD.ThesameissuealsocameupfortheCommon
Research Model (CRM) configuration at certain higher-than-cruise-angle-of-attack conditions in DPW-4 and DPW-
5.2 Many subsequent studies looked at the issue (see, e.g., Refs. 3–5); the computed juncture bubble appears to
be influenced by a variety of factors, including grid size, grid topology, and numerical treatments. Because of this
highdegreeofuncertaintyintheCFDpredictions,relevantseparatedcornerflowexperimentsfocusedspecificallyon
obtaininghigh-qualitydataforCFDvalidationareneeded.
The flow physics of wing-body juncture flows is quite complex because it can involve several distinct vortical
structures.6,7 Typically a large, unsteady horseshoe vortex as well as a smaller corner vortex are generated near the
wing leading edge at the wing-body intersection, and wrap around the wing. There is also a smaller corner vortex
initiatedbygradientsoftheReynoldsstresses.8 Bordjietal.7 refertothissmallervortexasa“stress-inducedvortex.”
All of these vortices pass downstream, influencing the flow field in the juncture region. Although there is general
awareness that the presence of a larger horseshoe vortex can help suppress corner separation,9 at the current time
theprecisemannerinwhichthevorticalstructuresandassociatedReynoldsstressesinteracttocauseand/orsuppress
cornerseparationisnotfullyunderstood. Alsonotethatmanyfactors—suchasincomingboundarylayermomentum
deficitfactor,wingbluntnessfactor,andwingsweep—playsomeroleinthejunctureflowphysics.10
Over time, the CFD community has come to the consensus that linear (Boussinesq) turbulence models such as
Spalart-Allmaras(SA),11Menter’sshearstresstransport(SST),12andotherswillproduceunrealisticallylargejuncture
separationbubbleswhenusingfinergridsand/orhighlyaccuratenumerics. Inotherwords,whensufficientlyreducing
numericaldiscretizationerrors,alinearturbulencemodelwillpredictjunctureseparationbubblesinaccuratelybecause
of the construction of the turbulence model itself. Yamamoto et al.13 demonstrated that the use of the quadratic
constitutiverelation(QCR)ofSpalart14(whichisanonlinearfixthatcanbeusedinconjunctionwithanyBoussinesq
model)couldimprovecornerflowCFDpredictionsfortheCRM.Similarly, inRumsey,15 SA-QCRaswellasother
∗ResearchScientist,ComputationalAeroSciencesBranch,MailStop128,FellowAIAA.
†ResearchScientist,FlowPhysicsandControlBranch,MailStop170.
‡ResearchScientist,FlowPhysicsandControlBranch,MailStop170,SeniorMemberAIAA.
1of16
nonlinearmodelssuchasexplicitalgebraicstressandsecond-momentReynolds-stresswereshowntobequalitatively
better than linear models for the CRM, in terms of predicting expected bubble size. Dandois16 used SA-QCR to
demonstrateimprovedpredictionsfortheDLR-F6wing-bodyconfigurationaswellasforashock-inducedseparation
in a rectangular duct. Presumably, the nonlinear models perform better because they more accurately represent the
Reynoldsstressesinthecornerregion. GradientsoftheReynoldsstressesleadtosecondaryflowbehavior17 thatmay
play an important role in the formation of the stress-induced vortex, and consequently influence corner separation
bubble formation/suppression. A similar conclusion was made over two decades ago by Chen,18 regarding second-
momentReynolds-stressclosures’advantagesforthistypeofflow.
Experimentally, Simpsonandco-authorshaveinvestigatedjunctureflowsingreatdetail(see, e.g., Refs.19–22).
However, that body of work did not include much focus on trailing-edge corner separation. Instead, the focus was
more on the leading edge region and the horseshoe vortex behavior. For example, they found that there is a low-
frequencyunsteadinessoftheinstantaneousflowstructureassociatedwiththehorseshoevortex,asevidencedbybi-
modalhistogramsofvelocityfluctuations.Thisbi-modalbehaviorhassincebeencomputedwithahybridcombination
ofReynolds-averagedNavier-StokesandLargeEddySimulation(hybridRANS-LES).23
Morerecently,ONERAhasconductedaseriesofimportantseparatedjunctureflowinvestigations,involvingboth
experimentandCFD.Gandetal.7,10,24describedasimplifiedwing-bodyjunctureexperimentthatusedaNACA0012
wingmountedtoaflatplate. Althoughdesignedtoachieveacornerseparation(basedonthemodel’slow“bluntness
factor”),theflowinthewindtunnelexperimentremainedattached. Inthenumericalinvestigation,allRANSmodels
employed—including a Reynolds stress model—produced corner separation, contrary to the experiment. However,
a LES calculation agreed well with the experimental data in this region (with the exception that the convection of
thehorseshoevortexwas“notcompletelysatisfactory”). Subsequently, theinvestigatorsre-designedtheexperiment
with a twisted NACA 0015 wing on a plate.25 This time, the corner flow separated in the experiment. Details were
providedatachordReynoldsnumberof1.2×106andangleofattackof12◦,althoughotherconditionswerealsoin-
vestigated.Experimentaldetailsincludedwallpressurespectraobtainedusingunsteadypressuresensorsandflowfield
measurementsofmeanandturbulentquantitiesfromstereoscopicHigh-SpeedParticleImageVelocimetry(HSPIV).
Theauthorsnotedthatthecornerflowseparationonsetwasdelayedwhentheoncomingboundarylayerwasthicker.
TheONERAgroupconductedCFDanalysisofthetwistedNACA0015configurationinBordjietal.6,26withRANS.
Here, theynotedthattheSAmodelinconjunctionwiththeQCRclosureyieldedimprovedpredictionscomparedto
thelinearSAmodel,whichgrosslyoverpredictedtheextentofcornerseparation.
Asthisisnotasurveypaper,wedonotdescribeotherpastjuncturefloweffortshere. Referencestootherjuncture
flowworkcanbefoundinGandetal.7 andSimpson,19 amongothers. GiventhelargevariationsinpastCFDcompu-
tationsofseparatedjunctureflowfields(attheDPWs,forexample),experimentaleffortstocharacterizeandquantify
the juncture flow physics—like those done by the ONERA group—are extremely important. Although the last sev-
eralyearshaveseenapparentCFDimprovementsthroughtheuseofSpalart’sQCRclosure,agoodunderstandingof
junctureflowphysicsremainselusive,andmanyquestionsstillremain. Forexample,ifbetterpredictionofReynolds
stressesnearthecorneristheprimarydriverforattainingappropriateseparationbubblecharacteristics,thenwhydid
the use of a Reynolds stress closure yield poor results in Ref. 24? In other words, we now presume that nonlinear
turbulence modeling is better for predicting this type of flow field, but existing comparisons are not all consistent.
Furthermore,wehaveonlyverylimitedquantitativedatawithwhichtotakethenextstepstothoroughlyvalidate(and
hopefullyimprove)existingmodels. AdditionaldefinitiveexperimentslikeONERA’sareneededtohelpguideCFD.
TheideafortheNASAJunctureFlowexperimentwasoriginallyconceivedbyafewmembersoftheDPW1steer-
ingcommittee.TheexperimentisintendedtobeprimarilyforthepurposeofCFDvalidationforwing-juncturetrailing
edgeseparationonsetandprogression. UnliketheONERAexperiment, whichusedanunsweptwingmountedona
flatplate,theNASAJunctureFlowexperimentplanstouseasweptwing-bodyconfiguration. Toeliminatequestions
regardingthepossibleinfluenceoftunnelwallboundarylayers, full-spantestingwillbeemployed. Furthermore, as
willbedescribedinSectionIIbelow,asignificantamountofeffortisplannedforattainingflowfielddetailsverynear
the walls in the corner region, and for duplicating measurements via different techniques. Guidelines delineated in
AeschlimanandOberkampf27 andOberkampfandRoy28 fordesignandexecutionofvalidationexperimentswillbe
followed. Althoughnotnecessarilytrustworthy,somepreliminaryCFDanalyseshavealreadybeenmadeforseveral
candidatedesignconfigurations;thesewillbebrieflyoutlinedinSectionIII. Becauseofhighuncertaintiesassociated
withdesigningaconfigurationtoachievethedesiredjunctureflowbehaviorwithCFDtoolsthatmaybeinaccurate,a
cautiousapproachwasadoptedforthisstudy(recallthefirstONERAtestwiththeNACA0012wingthatunexpectedly
failedtoyieldanyseparation). Inthiscautiousapproach,severalpreliminary“riskreduction”experimentshavebeen
conducted, as described in Section IV. The risk reduction experiments allow the team to more confidently select a
finalwingconfigurationthatprovidesthedesiredflowfieldcharacteristics;theyalsohelptodeterminetheappropriate
2of16
placementofinstrumentationforthefinalexperiment. Thispaperendswithabriefdescriptionoffutureplansforthe
NASAJunctureFlowexperiment,andconcludingremarks.
II. GoalsandPurposeoftheJunctureFlowExperiment
AJunctureFlowteam,ledbyNASA,hasbeenactiveforseveralyears. Theteamcurrentlyconsistsofmembers
from NASA Langley, NASA Ames, Boeing, and Penn State, as well as several consultants. In mid-2014, concrete
plans for definitive experimental measurements started to take shape. Several watershed moments occurred early in
thisprocess. First,theteamdecidedtomakeuseofaninternalLaserDopplerVelocimetry(LDV)systemthatwillbe
mounted inside of the fuselage on a movable three-axis traverse system; it will measure the flow field very near the
wing-body juncture through window(s) in the fuselage. See the schematic diagram in Fig. 1. Preliminary estimates
dictatedthatthissystemwillrequireacross-sectionalspaceofatleast13.7incheshighby10.3incheswide, witha
likelyneedforanadditional2inchesvertically. Thisrequirementfixedtheminimumsizeofthejunctureflowmodel.
Complementarymeasurementswithparticleimagevelocimetry(PIV)arealsoplanned.
Figure1. SchematicofinternalLDVsystemmeasuringinthewing-bodyjunctureregion.
Initial efforts were focused on devising a transonic flow application, primarily because the DPWs had used
freestream Mach numbers in this range. However, there were three significant difficulties with this approach: (1)
theuseofCFDtodesignaconfigurationthatattainedallofthedesiredcharacteristicsprovedtobechallenging(the
influence of the wing’s shock wave on the juncture separation bubble was unclear and represented a possible com-
plicatingfactor);(2)thenumberofavailabletransonicwindtunnelfacilitiesofsufficientsizewassomewhatlimited;
and(3)CFDmodelingofthewallsinthetransonicfacilitiesconsideredcouldintroduceahighdegreeofuncertainty.
Theteaminitiallystruggledtotrytoovercometheseobstacles. TheNASAAmes11foottunnelwasconsideredfor
the transonic testing, but in the end the requirements for body size to comfortably fit the internal LDV system in a
full-span model were too big (too much blockage) for that tunnel. Furthermore, some smaller-scale risk-reduction
experimentswereenvisioned,andsubsonictestingvenueswerefareasiertofind. Sothedecisionwasfinallymadeto
trytodesigntheexperimentatsubsonicconditions.
As mentioned earlier, the primary goal for the NASA Juncture Flow experiment is to acquire CFD-validation
quality data for wing-juncture trailing edge separation onset and progression on a full-span wing-body model. By
“CFD-validationquality”wemeanthattheexperimentshouldincludethemeasurementsofallinformation—including
boundary conditions, geometry information, and quantification of experimental uncertainties—necessary for a thor-
oughandunambiguousCFDvalidationstudy.27 TheattainmentofthisdatawillthenenabletheassessmentofCFD
modelsandsub-models. ThemainpurposeoftheCFDmodelassessmentisasfollows:
Assesstheabilityofexistingmodelstopredicttheonsetandextentofthethree-dimensionallyseparated
flowneartheWingJunctureTrailingEdge(WJTE)regionofafull-spanwing-bodyconfiguration,interms
3of16
ofthesurfacetopologyoftheflowfieldstructure. Toprovidearangeofpredictiondifficulty,avariationof
WJTEflowfieldsarerequired,includingtheonsetandprogressionofcornerseparation.
AswillbeshowninSectionIII,todatewithCFDanalysis,theteamhashaddifficultyfindingawing-bodydesignthat
achievesbothfullyattachedflowandseparatedflowviaasimplemechanismsuchasvariationofangleofattack(to
simplifytheprocessofCFDvalidation,thegeometryshouldremainfixed;i.e.,notachieveseparationbychangingor
addingnewgeometricfeatures). However,attainmentofarangeofattachedthroughseparatedflowremainsaprimary
goal. Whetheritisachievableinpracticeremainstobeseen.
OtherexperimentalgoalsincludeadditionaldetailsconsidereddesirableforCFDvalidation. Forexample, time-
averagedvelocities,Reynoldsstressesandtriplemoments,surfacepressures,surfaceshearstress,andtime-dependent
quantitiesaredesired. Thefollowingrepresentsaprioritizedlistofdesireddata(toppriorityfirst):
1. Documentationofthetestarticlegeometryandsurface, asfabricated, asassembled, andastestedinthewind
tunnel.
2. Tunnelboundaryconditionsupstream,around,anddownstreamofthetestarticleforeachangleofattackcon-
ditiontested.
3. Meanvelocitycomponentsthroughtheboundarylayeronthefuselageandwing(bothupstreamandwithinthe
WJTEregion).
4. Second moments (Reynolds stress tensor components) through the boundary layer (same locations as in Item
3).
5. Time-averagedsurfacepressurecoefficientsonthefuselageandwing.
6. Ifboundarylayertrip(s)areused,detailsregardingtheirlocation,type,attachment,etc.
7. Documentationofthesurfacegeometryofthewindtunneltestsectionandthecontractionsectionupstreamof
thetestsection.
8. Documentationofthegeometryofthemountinghardwareandcablingforthetestarticleinthetestsection.
9. Time-seriesmeasurementsofangularpositionandaccelerationon-boardthetestarticle.
10. Surfaceshearstressesonthetestarticle(bothupstreamandwithintheWJTEregion).
11. Spectra of unsteady velocity and pressure at various locations on the fuselage and wing (both upstream and
withintheWJTEregion).
12. Freestreamturbulenceintensityandturbulencelengthscale.
13. Triplevelocitycorrelationsthroughtheboundarylayer(samelocationsasinItem3).
Theaboveitemsmayormaynotbepossibletoattaininthistest;theprioritizationwillbeusedduringtestingto
help make decisions when time is limited or when particular capabilities are not available. All of the mathematical
model input data described above should include a rigorous estimation of total experimental uncertainty, i.e., both
random(precision)uncertaintyandsystematic(bias)uncertainty.
III. PreliminaryConfigurationsandCFDAnalysis
CFD was used to design and analyze a variety of different wing-body configurations. Early in the development
effort, a simple fuselage configuration was settled upon, with the CRM fuselage configuration as a starting point.
However, thisnewJunctureFlowModelfuselagewasmodifiedtoincludetwoplanesofsymmetry(top-bottomand
left-right),anditssidesareverticalandflatinthevicinityofthewings,forgreatereaseinincludingwindowsforthe
internally-mountedLDVlaserbeamstopassthrough. Thefuselageincludesanose, andconstrictsdownintherear
regiontowardwherethestingisattached. Fig.2showsaviewofthefuselageshape. Althoughnotdecidedyet, the
fuselagedesignforthefinalmodelmayincludesomeprovisionforvaryingitslengthupstreamofthewing.
For the wings, several different designs were explored with CFD, by using a realistic-looking planform shape
that was truncated in the spanwise direction in order to allow for an overall larger model. For all of the wings,
4of16
Figure2. Schematicofthecurrently-definedJunctureFlowModelfuselageshape.
predictedtrailing-edgejunctureseparationbubblesizeswerenotedasafunctionofangleofattack. Thegoalwasto
finddesign(s)forwhichtherewouldbeattachedflowatlowanglesofattackandaquickly-growingbubbleathigher
angles of attack. Optimally, a symmetric wing is desired, because one could then take advantage of the top-bottom
symmetrytoexploitdesign-of-experimentideasforattaininguncertaintyinformation. However,theteamalsomade
use of a truncated version of the wing from the DLR-F6 configuration (which is not symmetric). In past transonic
experimentstheDLR-F6wing-bodyhadyieldedcornerseparationbubbles,sotheteambelievedthatthiswinghada
highlikelihoodofachievingcornerseparationinspiteofdifferencesinthefreestream,fuselageshape,andwingspan.
IntheCFD,theSA-RCturbulencemodel,29combinedwiththeQCRclosure14wasprimarilyused. Thereasonfor
thischoicewasundocumentedpositiveexperiencesamongteammembers(alsoseetheIntroductionforadiscussion
ofthepublishedsuccessofSA-QCRforcornerflows). Othermodelswerealsotried. Forexample,thesevenequation
ReynoldsstressmodelSSG/LRR-RSM-w201230wasfoundtoyieldsimilarqualitativeresultstoSA-RC-QCRoverall.
InadditiontothetruncatedDLR-F6wing(hereafterreferredtoassimply“F6”),theteamendedupwithfourother
trialwings. UnliketheF6,whichisnon-symmetricandalsoincludeswingdihedral,theseotherswerealltop-bottom
symmetric. Allwingshadsomewhatsimilarplanformshapes,includinga“Yehudibreak,”orcrank,part-wayoutthe
span, althoughtheF6winghadslightlydifferentleading-andtrailing-edgesweepangles, cranklocation, andcrank
chordthantheothers. Onewing,referredtoas“0015,”usedaNACA0015rootshape. ThiswasblendedtoaNACA
0012atthecrank,andNACA0010atthetip. Thereasonfortheblendingwastoreducethewingthicknessandhence
thelikelihoodofoutboardtrailing-edgeseparation. Asecondwing,referredtoas“0015mod,”usedamodifiedNACA
0015rootshape,forwhichthelast2coefficientsintheairfoilequationwerealteredtoachievegreateradversepressure
gradientovertherearofthewing,whilemaintainingthesametrailingedgethickness. Thiswingwasblendedintothe
sameNACA0012andNACA0010asthe“0015”wing. Athirdwing,referredtoas“F6-S12,”wasdesignedbasedon
asymmetrizedversionoftheF6wing,withincreasedmaximumthicknessandtrailingedgeincludedangleinorderto
achievesignificantadversepressuregradientovertherearpartofthewing,toencouragethelikelihoodoftrailingedge
separationattheroot. RatherthanblendtoaNACA0012and0010,theF6-S12blendedtosymmetricairfoilshapes
derivedfromtheoriginalCRMdesign.Thefourthwing,referredtoas“COCA,”wasdesignedwithCDISC,31withthe
constraintofachievingaspecifiedskinfrictiondistributionattwostationsneartheroot(forexample,C =0.0001at
f
theroottrailingedgeatzerodegreesangleofattack,withspecifiedd(C )/dxupstream).Acurvatureconstraintahead
f
ofthewing’smaximumthicknesswasalsoemployed. Thiswingblendedtothecrankandtipinthesamefashionas
theF6-S12.
Leadingedgefilletsor“horns”(intheleadingedgecornerregionbetweenthewingandfuselage)werealsocreated
foreachwingshape. Onepurposeofthehornistohelpeliminateand/ormitigatethedevelopmentandinfluenceof
thehorseshoevortex. AdiscussionoftheeffectofhornscanbefoundinDevenportetal.21 Notethatthehornsused
hereweregeometricalfairingsonly,andwerenotaerodynamicallydesigned/optimized. Theywerecreatedbyfixing
locationsonthefuselageandwing,andfairingasmoothleading-edgecurvedshapebetweenthem(tangentatthewing
5of16
andapproximately13◦ atthefuselage),whilemaintainingcontinuityeverywherewiththewingslope. Theteamhas
notyetdecidedwhetherornottoincludeahornonthefinalconfiguration. Thevariouswingrootshapesareshown
in Fig. 3 (without horns). Here, the F6 is clearly seen as the only non-symmetric wing shape, and the F6-S12 and
theCOCAarebyfarthethickestwingshapes. Twoofthecompletewing-bodyconfigurations(theF6withhornand
0015modwithhorn)areshowninFig.4. Theotherwings(0015,F6-S12,andCOCA)generallylooksimilarintheir
planformshapestothe0015modinFig.4(b).
Figure3. Wingshapesattheroot(hornsoff).
(a) F6withhorn (b) 0015modwithhorn
Figure4. Viewsoftwoofthewing-bodyconfigurations.
Figure5showsplotsofapproximatecomputedbubblesize(lengthandwidth)asafunctionofangleofattackfor
thevariouswing-bodyconfigurations. Noerrorbarsareincludedintheplots,buttheuncertaintyinthecomputations
isconsideredtobequitehigh. ComputationshavebeenmadebothwithOVERFLOW32 andFUN3D,33 whichhave
provedtobereasonablyconsistentwitheachotherwhenusingSA-RC-QCR.Here,onlyFUN3Dresultsareshown,for
M =0.2andRe =3.5millionbasedonc=crankchord(theparticularseparationbubblesizesaregivenininches,
c
foraconfigurationwithcrankchordofaround22to23inches). Thecrankchordwaschosenforlengthscalebecause
itisaneasilyvisiblereferenceonthemodels,whichishelpfulforthisprojectbecausetheteamisdealingwithmany
differently-sizedscaleversionsoftheconfigurations. LowerReynoldsnumbershavealsobeenrun, yieldingatrend
of somewhat larger bubbles. Note that the details are not important here, only whether or not a bubble is predicted,
andthesizesofthebubblesrelativetoeachother. Basedonpreviousstudies,theteambelievedthattheCFDprobably
hadatendencytopredicttoolargeabubblecomparedtoexperiment. Sothegoalwastoachieveanarrayofdifferent
results,leaningtowardlargerbubblesizes. TheF6wingproducedthelargestbubblelengthandthe0015producedthe
smallest,whilethe0015mod,COCA,andF6-S12eachfellsomewhereinbetween. Bubblewidthtendedtobemore
significantfortheF6-S12andCOCAconfigurations. Althoughnotdetailedhere,theeffectofhornonversushornoff
wasrelativelyminor,withtheformergenerallyproducingslightlylargerseparationbubbles. Figure6showsexample
6of16
(a) Bubblelength (b) Bubblewidth
Figure5. ApproximaterelativebubblesizespredictedbyCFDforthevariouswings(withFUN3D).
CFD-generated streamlines from the five wing designs, in this case at angle of attack of 5 degrees and Re = 3.5
c
million. TheF6andthe0015modeachyielded“traditional-shaped”bubbles(i.e.,whatwouldnormallybeexpected
ontypicalwing-bodyconfigurations),andthe0015modalsoproducedsometrailingedgeseparationslightlyoutboard
ofthebubble. The0015yieldedanextremelytinybubbleatthisangleofattack. TheF6-S12andCOCAdesignsboth
yieldedoddly-shapedbubbleswithatendencytowardlargerbubblewidthsthantheF6and0015mod.
Thissetoffivewingconfigurationswasfelttoencompassareasonablybroadarrayofpossibleoutcomesinterms
of juncture bubble behavior. A few other wing configurations beside these five were also tested, but they either did
not yield the desired behavior or did not add any value beyond what the other five wings provided; so they are not
describedhere. BecauseoflowconfidenceintheCFDpredictionsofcornerseparationsize,theteamdecidedtotest
allfivewingsinaseriesofpreliminary“risk-reduction”windtunneltests. Thesetestsaredescribednext.
IV. RiskReductionWindTunnelTests
Inthissection, anoverviewoftheriskreductionwindtunneltestsisprovided, followedbysomesampleresults
fromoneofthetests.
A. Overview
Forthefinalwindtunneltest,themodelsizeisplannedtobeat8%scale(ascomparedtoarepresentativefull-scale
CRM-likeaircraftconfiguration). Atthisscale,thecrankchordoftheF6wingwillbeapproximately22inches,the
totalwingspanwillbeapproximately131inches(tiptotip),andthefuselageoutermoldlinewillbeapproximately
24.8incheshighby18.6incheswide,largeenoughtoeasilyfittheplannedLDVsystemaswellasanyneededcabling.
Theapproximatelength ofthefuselagewillbe somewhereintheneighborhood of200inches(dependingon where
thedesignshapegetsterminatedtofittoastingmount). TheReynoldsnumberbasedoncrankchordatM = 0.2is
expectedtobeapproximately2.3−2.7million.
However,preliminaryrisk-reductionwind-tunneltestswereconductedfirst(inmid-tolate-2015),inordertohelp
theteamdeterminethefinalwingconfiguration. Table1summarizesalloftheexperiments. Oneoftherisk-reduction
experiments (labeled “Pre 1” in the table) was a limited semi-span low Reynolds number test in the TC2 facility at
NASAAmes. Thisfacilityhasatestsectionsizeof48inchesby32inchesandoperatessubsonically. Itwaschosen
becauseofitslowcostandrelativelyrapidturnaround. However,useofasemi-spanmodelwasconsideredtoberisky
(potentiallyunreliable)duetoinfluencesofthewallboundarylayerandadditionalwall-fuselagejunctureflowonthe
wing-fuselagejunctureflowofinterest,soonlytwowingshapeswerebuilt. Themodelsbuiltforthistestwereat3%
scale. Atthisscale,thecrankchordoftheF6wingwasabout8.2inches,wingsemi-spanwasabout24.6inches,and
Reynoldsnumberwasapproximately600,000–700,000. LimitedtestingwasdonefortheF6andF6-S12wings,and
7of16
bothsurfaceoilflowphotographsandoilfilminterferometrydatawereacquired.
Thesecondwindtunnelexperiment(“Pre2”inthetable)wasafull-spanlowReynoldsnumbertestintheVirginia
TechStabilityWindTunnel. Thistunnelhasa6footby6foottestsectionandoperatessubsonically. Themodelsbuilt
forthistestwereat2.5%scale.Atthisscale,thecrankchordoftheF6wingwasabout6.9inches,wingtip-to-tipspan
(a) F6 (b) 0015
(c) 0015mod (d) F6-S12
(e) COCA
Figure6. ExamplestreamlinesfromCFDforthefiveconfigurations(withFUN3Datα=5◦,M =0.2,andRec=3.5million).
8of16
Table1. Summaryofwindtunneltests(usingF6wingforreference)
Test Tunnel Date Modelsize Re (×106) M Crankchord,c(in) Wingspan(in)
c
Pre1 TC2 Summer2015 3% 0.6–0.7 0.14 8.2 24.6(semi-span)
Pre2 VATech Summer2015 2.5% 0.6–0.7 0.18 6.9 40.9
Pre3 14x22 Fall/Winter2015 6% 2.3–2.7 0.27 16.5 98.3
Final 14x22 TBD 8% 2.3–2.7 0.20 22.0 131.0
was40.9inches,andReynoldsnumberwassimilartothatfortheTC2test,approximately600,000–700,000. Allfive
wing shapes (F6, 0015, 0015mod, F6-S12, and COCA) were tested in this tunnel. The wings were interchangeable
onto the fuselage, and the wings could also be tested with and without a horn. To minimize the model construction
costs, six unique wings were built: F6 (port), F6 (starboard), 0015 (port), 0015mod (starboard), F6-S12 (port), and
COCA (starboard). The following wing pairs were always tested together: F6 with F6, 0015 with 0015mod, and
F6-S12withCOCA.CFDanalysissuggestedthattherewouldbenegligibleinfluenceonseparationbubblebehavior
fromtheuseofslightlydifferentwingsontheportandstarboardside.
BecauseofitslowReynoldsnumber,this“Pre2”testmaynotberepresentativeoftheresultsthatwillbeultimately
obtainedatthehigherReynoldsnumberofroughly2.3−2.7million. Therefore,theresultsfromthistestweretobe
viewed only qualitatively. The transition characteristics, with and without trip dots on the fuselage and wing upper
surface,wereassessedwithinfrared(IR)thermography. Oilflowvisualizationswereusedtoidentifythetrailing-edge
corner separation bubble, to visualize the general flow pattern on the wing upper surface planform, and to visualize
theflowpatternonthefuselageinthevicinityofthewing-fuselagejuncture. Thistestwasaninvaluableprecursorto
the14x22tunnelrisk-reductiontest,givingtheexperimentalistteammembersexperiencewithIRthermographyand
providingpreliminaryresults(albeitatlowReynoldsnumber). AphotographofthemodelintheVirginiaTechtunnel
isshowninFig.7. Tripdotscanbeseenaroundthefuselagenoseaswellasnearthewingleadingedges. Examplesof
IRthermographyimagesaregiveninFig.8. Thedarkershadeindicateslaminarflow,whilethelightershadeindicates
turbulentflow.Here,theuntrippedF6wingisshowntobedominatedbysignificantregionsoflaminarflow(especially
outboard)atα = 0◦,butnaturaltransitiontakesplace(duetoturbulentreattachmentofalaminarseparationbubble
nearthewingleadingedge)somewherebelowα=5◦. AlthoughthequalityofthephotographsinFig.8(c)and(d)is
somewhatpoor,theyconfirmthattheapplicationofthetripdots(atroughly4%chord)successfullyinducedtransition
toturbulence.
Figure7. Photographofthe2.5%scalemodelintheVirginiaTechStabilityWindTunnel.
9of16
(a) Untripped,α=0◦ (b) Untripped,α=5◦
(c) Tripped,α=0◦ (d) Tripped,α=5◦
Figure8. ExampleIRthermographyimages,fortheF6winguppersurfaceintheVirginiaTechtunnel(flowisrighttoleft).
The third risk-reduction wind tunnel experiment (“Pre 3” in the table) was a full-span higher Reynolds number
testintheNASALangley14x22subsonicwindtunnel. Thistunnelhasa14foothighby22footwidetestsection.
Themodelsbuiltforthisrisk-reductiontestwereat6%scale. Atthisscale,thecrankchordoftheF6wingwasabout
16.5inches,wingtip-to-tipspanwas98.3inches,andReynoldsnumberwasexpectedtobeapproximately2.3−2.7
million. Asofthiswriting,the14x22riskreductiontestinghasnotyettakenplace. Allfivewingshapesareexpected
tobetested. Inthismodel,allwingswerebuiltwithahorn,withtheexceptionofoneoftheF6wings,whichcanbe
testedeitherwithorwithoutahorn. Becausethisrisk-reductiontestisgoingtoachievesimilarReynoldsnumbersas
thefinalplannedtest,approximatebubblesizeinformationfromoil-flowvisualizationsonthe6%scalemodelwillbe
usedtohelptheteammakeadecisionaboutthefinalwingconfiguration.
B. Results
AfewsampleresultsfromtheVirginiaTechfull-spanexperimentareshowninthissection. Semi-spanresultsfrom
theTC2tunnelturnedouttobequalitativelydifferentfromthefull-spantests,yieldingsignificantlysmallerseparation
bubbles. Therefore, the TC2 results are not shown here, since they are considered non-representative of a full-span
modelandwillnotbeusedtomakethefinalwingdesigndecision.
Figure 9 shows a sampling of oil flow photographs at α = 5◦. Here, the fuselage and wings were tripped, and
the horns were included. At this angle of attack, all configurations except for the 0015 produced corner separation
bubbles. These results can be compared qualitatively with the CFD predictions in Fig. 6. Although at different
Reynoldsnumbers, therearemanysimilaritiesinshapeandrelativebehavior. OtherCFDresults(notshown)atthe
10of16