wall correction 14x22 wind tunnel

7/30/2019 Wall Correction 14x22 Wind Tunnel

1/15

21stAIAAAppliedAerodynamicsConference,Orlando,FL,23-26June2003.

1

AmericanInstituteofAeronauticsandAstronautics

IMPROVEMENTSTOWALLCORRECTIONS

ATTHENASALANGLEY14X22-FTSUBSONICTUNNEL

VenkitIyer,DavidD.Kuhl,LockheedMartin

and

EricL.Walker,NASALangleyResearchCenterHampton,VA

SeniorMember,AIAA;AerodynamicsSectionSupervisor,LangleyProgramOffice.Member,AIAA;AeronauticalEngineer,LangleyProgramOffice.StudentMember,AIAA;ResearchEngineer,ResearchFacilitiesBranch,AAAC.

ABSTRACT

The new wall pressure measurement system and the

TWICS wall correction system for the 14x22-Ft

subsonic tunnel are described. Results froma recent

semispan test and a full-span test are presented.

Comparison with existing classical methods of

correction is shown. A modification of the TWICS

codetotreattheeffectduetoadeflectedwakefromahigh-lift wing is also discussed. The current

implementation ofTWICS for the 14x22-Ft tunnel is

showntobeanimprovementoverexistingmethods.

1. INTRODUCTION

Windtunnelfacilitiesallovertheworldareconstantly

seekingimprovementsintestingcapabilitiesandinthe

qualityofdataprovidedto thecustomers.Themetrics

ofperformanceisnolongerthenumberofdatapoints

taken,butrathertheefficientmannerinwhichthetest

objectivesareachievedbyacquiringqualitydata.Two

items of importance in the area of data quality are:statistical analysis of repeatability of data based on

periodic check-standard testing1, and correction for

interferenceduetothetunnelwalls.Inparticular,wall

interference appears as a bias error on the measured

data,whichneedstobequantifiedinordertoarriveat

theequivalentfree-airvalues.Itisimportanttodothis

accuratelybecausesubtlephenomenasuchasReynolds

numbereffectscanbemaskedotherwise.

Inthisdrivetoimproveperformanceanddataquality,

wind tunnels are retro-fitted with new hardware, data

acquisition systems and analysis software. The

14x22-Ft facility at the NASA Langley Research

Center, thesubjectof this paper,underwent anumberof modifications during 2001-2002, including a new

drive motor and a new wall pressure measurement

systemamongvariousotherupgrades.Specifically,the

purposeofthenewwallpressuremeasurementsystem

istomeasuremoreaccuratelythepressuresignatureat

the walls with or without the test article. This

measurementinturnisusedinspecifyingtheboundary

condition for a wall interference correction code.

Correctionsto obtaintheequivalent free-airconditions

are therefore computed based on boundary

measurements and forcemeasurements. The general

experience is that this leads to a more accurate

estimation comparedto theclassicalcorrectionsbasedonliftanddragonly.Ofcourse,thisclaimneedstobe

substantiated with extensive cross-tunnel validation

testsusingdifferentsizemodels.

Thispaperisadescriptionoftheimprovementsinwall

correction estimation using the new wall pressure

systematthe14x22-Ftfacility.Section2describesthe

newwallpressuremeasurementsysteminmoredetail;

detailed analysis of the quality of these wall

measurementsisgiveninReference2.Thisisfollowed

bySection 3 which presents wall interference basics,

terminologyanddefinitionofcommontermsused.The

classical methods used previously at the facility arediscussed inSection4. Thewall signaturemethod is

thenintroducedinSection5withadetaileddescription

oftheTWICSwallsignaturemethodnowimplemented

in the tunnel. Sections 6 and 7 discuss the two

importantpartsoftheTWICSimplementation,viz.,the

pre-computedperturbationvelocitydatabase(PVD)and

theemptytunnelcalibrationdatabaseusedasatarein

estimatingtheincrementalwallsignatureforaspecific

testpoint.Sections8and9presentsomeresultsfroma

recentsemispanandafull-spantestatthefacility.As

anexampleofimprovementsmadeonthebasicTWICS

method,theissueofdeflectedwakeanditseffectonthe

wall correction ispresented inSection10. Finally,a

brief summary and concluding remarks are given inSection11.

21st Applied Aerodynamics Conference23-26 June 2003, Orlando, Florida

AIAA 2003-395

Copyright 2003 by the American Institute of Aeronautics and Astronautics, Inc.The U.S. Government has a royalty-free license to exercise all rights under the copyright claimed herein for Governmental purposes.All other rights are reserved by the copyright owner.


2/15

2

2. THENEW14x22-FTTUNNELWALL

SYSTEM

The14x22-Fttunnelunderwentextensivemodifications

inthe Fall 2001 toSpring 2002 time period. A new

main drive motor and a new boundary-layer suction

grating were installed during this time in addition to

varioustunnelcontrolanddataacquisitionupgrades.A

newwallpressuremeasurementsystemwasalsoputin

place. Prior to the upgrade, wall pressures were

measuredinthecenterofthetwosidewalls(calledthe

NorthandSouthwalls)andinthecenteroftheceiling

alongarowof30wallportlocationsdistributedalong

the 40-fttestsection length. However, the quality of

wall pressure measurements was poor due to inferior

measurement instrument accuracy, poor surface and

orifice quality and possibly due to various leakage

flowsintothetestsection3.

Thenewwallpressuremeasurementlayoutisshownin

Figure1.Itconsistsof12rowsof31wallpressuretaps

(4rowsin each ofthe south, ceiling and northwalls;

seeFigure1 forwallandrowdesignationconvention).

Thewallportdistributiondensityin theodd-numbered

rows is selected such that a higher resolution is

obtainedatthefrontmodelcartcenterlocationof17.75

ft.Theportdistributionintheeven-numberedrowsis

weightedsuchthatahigherresolutionisobtainedatan

alternate downstream model centerlocation. Thewall

orificesaredrilledintoplateswhichareaffixedtothe

wallin30 ftsegments.Variousgapsinthewallswere

pluggedorsealedtoreduceleakage.

The pressure lines are connected to electronically

scanned pressure(ESP) moduleswhich aresupplieda

reference pressure from a pressure calibrator unit

(PCU).TheESPmoduleshaveafullscalerangeof10

inchesofwater(0.361psi).Sincethetunneldynamic

pressure(Q)canbeashighas0.95psi,theESPmodule

referencepressureissetdependingonthetunnelQso

thatthewallESPpressuredifferentialiswellwithinits

stated range. Wall measurements are thus made in a

smaller range tuned to the tunnel Q resulting in

improvementsinmeasurementaccuracy.

3. OVERVIEWOFTUNNELWALL

CORRECTIONS

Wallinterferencereferstothechangesinthemeasured

tunnel-stream reference conditions and model

parametersduetotheconstrainingeffectsofthetunnel

walls. Inthe case ofsolid walls, the natural outward

expansionof thestreamlinesisprevented, causingthe

flowaboutthemodeltoaccelerateoverthatinfreeair.

Open jet boundaries allow the flow to over-expand,

causing the flow atthe boundary toslowand balloon

outward. The flow at a ventilated wall is somewhere

between these two extremes. These changes in the

boundaryconditions canmakethe tunnel flow around

themodelsubstantiallydifferentfromthefree-airflow.

Thetunnel measuredvaluesreflect thischangedflow,

and corrections must be applied to the measured and

derivedvaluestogettheequivalentfree-airvalues(i.e.,

whenthewallsareremoved).References4and5givea

comprehensivereviewonthetopicofwallinterference.

To illustrate the idea further consider a three-

dimensional point source in free-stream simulating a

Rankineforebody.Thefree-airsolutionisgivenbythe

sum of the free-stream potential and the source

potential. The perturbation velocities in free-air are

simply the derivatives of thesource potential. When

this singularity is enclosed by four solid walls, flow

tangency is imposed at the boundaries which isobtainedbyplacinganinfinitenumberofsingularities

atreflectionlocationsoffthewalls,followingthewell-

known method of images. Therefore, the additional

perturbation potential introduced by the walls is the

sum of all the image potentials. Correction for wall

interferencethus corresponds to the sumof reflection

potentials, which can be evaluated once the original

singularitystrengthisknown.

Wallinterferencecorrectionisthusaspatiallyvarying

function. The traditional assumption is that the

perturbation velocity field can be approximated by a

change inthe angleof attack and the tunnel velocity.Any left-over differences are usually second-order

effects. However, when they become significantly

largesuchasinthecaseofalargemodel,themeasured

datamaybecomeuncorrectablebytraditionalmethods.

ACFDsolutionofthein-tunnelflowisthenrequired.

Thisishowevercostlyforroutineuseatawindtunnel

facility.

The perturbation velocity field is usually computed

using a wall signature method in which a simplified

representation of the model is made using potential

singularityelementssuchas pointsinks,pointsources,

and point or line doublets. Once the perturbation

velocity solution based on the measured boundary

condition imposed is known, wall corrections can be

quantified by averaging the interference flow field

alongmodelortunnelreferencelines.Thecorrections,

termed primaryor mean wall interferencecorrections,

are given in terms of a blockage correction and an

angleofattackcorrection.


3/15

3

Theblockagecorrectionisobtainedastheaverageofstreamwise perturbation velocities (normalized by

tunnelreferencevelocity)alongthemodelaxis.Thisis

proportionaltotheratioofthemaximummodelfrontal

crosssectionareatothetunnelcrosssectionarea.This

correction is applied to the measured values of MachnumberManddynamicpressureQ.Thecorrectionsfor

MandQareobtainedas

211

2

MM

M

= +

(1)

( )22Q

MQ

= (2)

Theangleofattackcorrectionisobtainedasaweighted

average of the perturbation velocity in the lifting

direction.Thewingchordlineisthereferenceline

customarily used for this averaging, although someclassicalmethodspreferthechordline.Weightingis

done based on the spanwise load distribution. These

primary correctionsleadto correspondingcorrections

onforcecoefficientsCLandCD.Abuoyancycorrection

onCDduetowallinterferenceisalsocomputedfrom

the perturbation velocity gradient averaged along the

model axis. Pitching moment coefficient corrections

are computed based on changes in flow curvature

calculatedfromspatialvariationsintheangleofattack

corrections.

Theinterferenceflowfieldinthevicinityofthe model

isalsoofinterestinordertoassesstheextentofchangein mean corrections with change inreference lines or

pointswheretheyareaveraged.Theselocalvaluesof

corrections are presented as contour plots at different

tunnelsections.

To summarize, wall corrections for a test point are

reportedasfollows:

1. A blockage parameter, and an angle of attack

correction, ; these are the primary correctionsobtainedbyaveragingtheperturbationflowfield.

2. Corrections ontunnelMachnumber and dynamic

pressure, which are functions of the blockageparameter,.3. Corrections on lift, drag and moment coefficients

which result from the changed tunnel reference

velocity,angleofattackandflowgradients. Theseare

derivedcorrectionsobtainedfromtheperturbationflow

field.

4. Contour plotsof perturbation velocityfield inthe

modelvicinity;theseshowtheextentofthedeparture

of local corrections from the averaged correction

values.

Thefree-airorcorrectedvaluesareobtainedbyadding

the corrections to the measured values. For a lifting

modelinasolid-walledtunnel,thefree-airorcorrectedvalues ofM, Q, , CD are larger than the measured

values;free-airCLisusuallydecreased.Dependingon

the model and the test, corrections on roll moment,

sideforceandtailincidencecanalsobecalculated.

4. EXISTINGMETHODSATTHE14x22-FT

TUNNEL

Theexistingwallcorrectionsusedinthefacilityrelyon

standardformulaebasedonmodelandtunnelgeometry,

measuredliftanddragvalues.Thisclassicalblockage

correction is documented in References 4, 6, 7 and

various other references dating back to 1960. It

consists of a fuselage volume blockage term, a wing

blockage term and an attached wake blockage term

based on the apparent drag coefficient. The jet

boundary correction (equivalent to an angle of attack

correction)isalsoappliedbasedontheliftandinduced

dragcoefficients. While this method is of acceptable

accuracyforasmallmodelinattachedflowconditions,

large departures are possible when dealing with large

modelsathigh-liftconditions.Formodelswithalarge

deflectedwake,amethodbasedonHeysonswork8is

applied to calculate the change in the angle of attack

anddynamicpressureinthemodelregion.

Unfortunately, many questions exist regarding thevalidity of the assumptions required for derivation of

theseclassicalmethods.Inaddition,noassessmentof

correction uncertainty is available. Improving the

accuracy of wall corrections involves the use of the

measured wall-pressure boundary condition, which

incorporatesa morerealisticcharacterizationofthe in-

tunnel flow into the correction method and hence a

tightercontrolontheaccuracyofcorrections.

5. THETWICSWALLSIGNATUREMETHOD

TheTransonicWallInterferenceCorrectionSystem(or

TWICS) is a wall correction code based on the wallsignature method developedfor transonic tunnelswith

ventilatedwalls,originallyfortheAmes11-FtTunnel9.

Ithasbeenimplementedforthe14x22-Fttunnelin the

solid-wallorwallsdownconfiguration.TWICSand

its predecessor code WICS10

were developed by

Ulbrichbyusingastrategyofgloballyfittingthewall

signature. A brief summary of the method is given

below.


4/15

4

TWICS uses the pressure signature at the walls

(actuallytheincrementalvaluerelativetotunnelempty

signature)asthebasisforcomputingwallinterference

corrections.Themodelisrepresentedbyanumberof

pointdoublets(tosimulatevolumeandwakeblockage)

andlinedoublets(tosimulateeffectsduetolift).Thefar-field effect due to the assumed singularity

distributionismatchedwiththewallsignature.Thisis

done in a global fitting procedure, which yields the

strengths of the singularities as the solution. The

perturbation velocities are then computed based on

superposition of standard solutions of point and line

doublets which are contained in pre-computed

databases of perturbation velocity solutions.

Corrections for each test point are obtained by

interpolationfromthedatabaseatnear-realtimespeeds.

Compressibility is modeled using Prandtl-Glauert

scaling.Asaresult,thereisanupperlimittotheMach

numberintheapplicationofthismethod.

TheTWICScodeisdesignedtoworkinunisonwitha

panel-method-generated database of wall interference

solutionsbasedonpointdoubletsforblockageandline

doubletsfor lift interference. Byappropriatelysetting

theboundaryconditionsinthepanelmethod,ventilated

wallscanalsobemodeled.Sincethe14x22tunnelis

operatedasasolid-wallfacility,thisfeatureofTWICS

isnotused.Asimplerprocedurebasedonthemethod

ofimagesisusedtogeneratethedatabase.

ThestringofpointdoubletsusedintheTWICSmethod

tomodelvolumeandwakeblockagescanbeshownto

beequivalenttothesource-sinkpairusedin theWICSmethod. The advantage is that the individual point

doubletstrengthscannowbeweightedinproportionto

the cross section area of the test article thereby

increasingthemodelfidelity.InTWICS,theeffectof

the sting is also modeled using a chain of weighted

point doublets, in an analogous fashion to volume

blockage. This shortcut implies the use of a

considerablysmallertunnelcalibrationtestmatrix.

TheinputsforTWICSaredescribedbelow:

1. Tunnel empty signature: The wall signature is

defined in terms of the 12 rows with 30 orifices ineach. This calibration data is required for a specified

range of Mach numbers (or equivalently, dynamic

pressures).Forfullspanmodels,thesignaturewiththe

modelsupport(whichistheverticalpostatX=40ft)at

severalpitchanglesisalsorequiredatvariousoperating

conditions. Several such calibration sets are required

depending on the support system used and the floor

boundarylayersuctionsystem(BLRS)state(onoroff).

SeeSection7fordetailsofthetunnelcalibration.

2. Wallsignatureforagiventestpoint.

3. Testpointvaluesofuncorrectedforceandmoment

values, Mach number, reference velocity at model

centerofrotationandanumberofothertestandmodel

attitudeparameters.

4. Perturbation velocity database (PVD): This is alargetableofpre-computedperturbationvelocitiesused

in signature matching and wall interference

computation.Thedatabasedependsonthewallorifices

layout, tunnel section, Mach number and lift vector

direction. SeeSection6fordetailsofgeneratingthese

databases.

5. Modelsingularitydistributionandgeometrydata.

6. Referencelines alongwhichweightedaveragesof

interferencearetobecomputedandplanesalongwhich

localvaluesofwallinterferencearetobecomputed.

7. Port flags used to de-select specific wall orifices

thatarenottobeusedinthecalculationforagiventest.

In addition to this, the code rejects additional wall

orificesbasedonstatisticsofthefit(seeReference10

for the original rejection criteria). Here an improved

rejection criteria is used based on wall data quality

analysis (see Reference 2 for details). Hence the

original statistical rejection logic used in the code is

turnedoff.

The calculation steps used in TWICS are described

below:

1. Processing of input test data:Foreach test point,

the wall signature is read in. Subsequently, the

corresponding tunnel empty signature is interpolated

fromthecalibrationdatabaseandsubtractedtogettheincrementalortaredwallsignature.

2. Computationoftheequivalentlinedoubletstrength

frommeasuredliftandmodelgeometry:Thestrengthis

thendistributedalongthespanasperinputorcomputed

weights. These weights are based on the estimated

wingloadingdistribution.

3. Interpolation from the perturbation velocity

database (PVD): This is done to estimate the lifting

effectpartofthesignatureateachport.Theliftingpart

of the signature is then subtracted from the tared

signaturetogettheblockageeffectateachport.

4. LeastsquaresfittingandinterpolationfromPVDto

calculatethestrengthsofthepointdoublets:Thetwounknowns computed here are the volume blockage

strength and the wake blockage strength. This step

representsthecoreofthecalculationprocedure.

5. Interpolation from PVD to compute wall

interference at any point in the test section (within

reference grid limits) by superposition of all

singularities: Mean corrections are then calculated

using weighted averaging. Force and moment

coefficient corrections are then computed. The


5/15

5

streamwisedistributionofblockageis usedtoestimate

thebuoyancycorrection.

6. Iteration using corrected tunnel parameters, if

necessary.

Corrections are computed for each point in a polarindependently. The primary mean correction due to

blockageisappliedascorrectionsonMachnumber,M

and dynamic pressure, Q (added to corresponding

measured values). Upwash correction is applied as

correctionontheangleofattack.CorrectionsonCL,CD

and pitching moment coefficient CM are computed

basedontheprimarymeancorrectionsofblockageand

angle ofattack. Inaddition,model-induced buoyancy

correction is also calculated and added to CD. The

method also computes local variations of interference

corrections, which are useful in determining if the

averaging assumption is truly representative of the

interferencefieldinthemodelregion.

6. PRE-COMPUTEDPERTURBATION

VELOCITYDATABASES

The databases for TWICS are large tables of pre-

computed solutions of perturbation velocities induced

by unit singularities (point doublets or line doublets)

enclosedby the tunnel walls. Inthese databaseruns,

thesingularitiesareplacedatdifferentlocationsinthe

tunnel (termed the singularity grid) and solutions are

calculated at different locations inside the tunnel

(termedthereference grid)as wellas atthe wall port

locations.

The database table is thus a function of a number of

variablesasgivenbelow.

Typeofsingularity:Thepointdoubletsingularityisthe

basicbuildingblockforsolidorwakeblockagedueto

amodelinthetunnel.Thisisthesameasapointsource

plus a pointsinkin closeproximity. The line doublet

singularityis thebuildingblock forliftinterference in

the tunnel. This is equivalent to a chain of point

doubletsstartingatthelinedoubletlocationstretching

toinfinity.

Orientationofsingularity:Pointdoubletdirectionused

hereis always pointingin theupstream direction (i.e.,point sink to point source direction is X ). Line

doubletdirectionisoppositetotheliftvectordirection

(i.e.,chainofpointdoubletspointingawayfromthelift

direction). In order to handle a case with model roll

whereliftdirectionchanges,calculationsaredoneat45

degree intervals ofthe line doublet angle inorderto

facilitateinterpolations.

Singularity location: The locationis determined by a

specifiedsingularitygrid;themodelandsupportsystem

shouldbecontainedwithinthisgrid.

Solution location: The location is specified by a

reference grid. Reference lines along which wall

corrections are averaged (such as fuselage centerline,wing chord line) should be contained within this

grid. Solutions are also computed at all pressure

measurement locations on the tunnel walls. The

solutionatatunnelinteriorpointconsistsofthethree

perturbationvelocitycomponents( u,v,w);thesolution

atthewallconsistsofonlythestreamwisecomponent

u.

Model configuration: The database depends on

whether the model configuration is semispan or full-

span. The location of the singularity and reference

grids are dependent onthe model configuration. The

solutionalsodependsonthemodelconfigurationsince

thesemispan model-mounting wall becomesan image

planewiththeeffectivetunnelsectiondoublinginsize.

Mach number: With the use of the Prandtl-Glauert

scalingfactor2

1 M = ,theeffectivetunnelcross

section is reduced by this factor. The singularity

strengths also scale with this factor (2

for point

doublet,3

forlinedoublet).

Note also that the solutions calculated at the wall

include the direct effect of the singularity. This can

thenbecomparedtothewallsignature.Thesolutions

calculatedattheinteriorofthetunneldonotincludethe

direct effect of thesingularity(itis thusequivalent tothesummationofsolutionsfromonlytheimagesofthe

singularity which is the true definition of the

interferencesolution).

InTWICS,oncetheactuallocationofasingularityin

the tunnel is established, multi-parameter linear

interpolation is used to find the corresponding

perturbation velocitysolution. For walllocations, the

actual coordinates of the wall ports are used in the

database. A master databaseis first generated for all

thewallports;acustomizeddatabaseforagiventestis

then derived from it depending on the ports actually

selectedforuseinaparticularTWICSrun.

7. TUNNELCALIBRATIONDATA-BASES

Measuring the tunnel-empty wall signatures

periodicallyisarecommendedqualitycontrolmeasure

for the wall system. It is also required in TWICS to

subtractoutsystematicvariationsfromorificetoorifice

and the effect due tothe tunnel boundarylayer. The

assumptionmadehereis thattheseeffectsareconstant


6/15

6

withorwithoutthemodelinthetunnel.Thecalibration

isdoneforanumberof Qvalues,sothatinterpolation

canbedoneforthetest-pointvalueof Q.Forfull-span

models, the model sting is also considered in the

TWICS simulation, so there is no need to obtain a

calibrationwith thestingor othersupportcomponentswhicharealignedwiththewake.However,otherparts

ofthesupportsystem(suchastheverticalpostforthe

14x22tunnel)shouldbe includedinthecalibration. If

thegeometryofthesecomponentschangewithangleof

attack,forexample,thenthecalibrationshouldbedone

in a range of such parameters. Interpolation for the

tunnel empty signaturewill thenbe a multi-parameter

oneinvolving forexample,Qandtheangleofattack.

Separate databases are required with floor boundary

layerremovalsystemonandoff.

8. EXAMPLERESULTS:SEMISPANMODEL

Wall corrections for a large semispan wing tested

recently at the facility will now be discussed. The

model is a trapezoidal planform high-lift wing with

fixed-setting flaps and slats. A photo of the model

installed inthetunnelisshown inFigure2. Some of

the important model and tunnel geometry parameters

andflowconditionsaregiveninthefollowingtable.

Wingarea 22.028sqft.

Wingchord,

chordsweeps30,26.1

Root,tip,aero.and

geom.chords

4.6,1.7,3.3,

3.1ft

Halfspan,aspectratio 7.088ft,4.561Tunnelcrosssection 14.75by21.75

ft

Modelcenterof

rotation

X=17.75,

Y=7.25ft

TunnelQ 57.8psf

Machnumber,

velocity

0.2,230fps

Angleofattack 4to36

ThetestwasrunwiththeBLRSon.Alltestrunswere

made ata fixedunitReynolds number of1.4 million.

The CLvs. CD curveforthis wingmodel isshown in

Figure 3, which is typicalof a high-lift configuration.

Stallangleis34to35withamaximumCLof3.0anda

CD,stallof0.7.

Thesingularitydistributionisspecifiedaprioriwitha

number of X direction point doublets along the

fuselageaxisandwakeseparationlines.Linedoublets

originatingfromthechordlineareusedtomodelthe

lifting effect with an elliptic loading distribution.

Figures 4a and 4b show the distribution used in

TWICS. The specified locations correspond to zero

angle of attack; for a particular data point, these

locations are displaced as a function of the model

kinematicsandattitude(,inthepresentcase).

Wallcorrectionresultsforarepresentativerun(apolar

of28pointsfrom4to35)isgivenhere.Figure5

showsthemeancorrectionsforthisrun.Thecorrection

on theblockageparameter is shownas well as thecorrectionsonQ.Theblockagestaysbelowthe0.25%

valueforanglesofattackbelow25 .Between25and

34,blockageincreasestodoublethislevel(0.5%)due

toalargercontributionfromthewake.Above34,the

wing begins to stall and the large separated wake

increasestheblockageevenfurther.Alsoshowninthe

plotof Qarethecorrectionvaluescalculatedbytheexisting classical procedure as well as using the

Maskellprocedure11

.Thetwo-stepformoftheMaskellmethod estimates the separated flow effects on

blockageusing a formulation based on flat plate flow

measurements.TheblockagepredictedbyTWICSfalls

inbetweentheclassicalandMaskellresults.

The angle of attack correction shown in Figure 5 is

largelyafunctionofliftandfollowstheCLvs.shape.

Amaximum correctionof1.6isobtainedat =32which means theequivalent free-airangle of attack is

33.6. In TWICS, the correction is obtained byaveraging at the chord line. The classical values

obtainedbythejetboundarycorrectionofReference6,

alsoshowninFigure5,seemtobemoreinagreementwith the TWICScorrections obtained by averaging at

thechordline.

Corrections on the force and moment coefficients are

showninFigure6.TheD

C andL

C valuesfromthe

classical approach are also shown to be in close

agreement, with a slight departure only near stall

conditions.ThebuoyancycorrectiononD

C isshown

to be quite small, varying linearly with the angle of

attack to a maximum value of 25 drag counts. The

changed aerodynamic characteristic due the wall

interferenceintheformofacorrected L DC C curveisalsoshowninFigure6.Overall,asseenfromFigures

5-6,thewallcorrectionsaresubstantialandneedtobe

consideredwhenevaluatingthefree-airperformanceof

thewing.

AgoodwaytoverifyhowtheTWICSmodelsimulates

the real flow is tocomparethe measured incremental

wall signature with the TWICSprediction. Shown in


7/15

7

Figure7aretheincrementalwallsignaturesalongthe

12rowscomparedwiththeTWICSvaluesfor3angles

of attack. Theoverallagreement isremarkably good,

exceptatRows1and12athighanglesofattack.Some

of this difference is due to a deflected wake which

tracks closer the north wall and is not simulated inTWICS.Section10providesmoredetailsoncapturing

thiseffectwhichleadstoabetteragreement.Itisalso

possible that the simple lift model used in TWICS is

also responsible for part of this difference in the

signatures.

9. EXAMPLERESULTS:FULL-SPANMODEL

Wall corrections for a relatively small full span wing

(reference area only 2% of tunnel cross section,

comparedto7%forthetrapezoidalwing)willnowbe

presented.Thewingisanellipticalplanformwingwith

symmetric NACA 0012 sections. This is a more

limitingtestcaseforthewallpressuresysteminthatthe

wall signatures are much weaker (only 10% of the

trapezoidal wing signatures for the same angle of

attack).Someoftheimportantmodelparametersand

flowconditionsaregiveninthefollowingtable.

Wingarea 6.25sqft.

Wingsweep 0

Rootchord 1.2ft

Span,aspectratio 6.75ft,5.625

Modellength 2.8ft

Modelcenterof

rotation

X=17.75,Y=0

ft

TunnelQ 2137psf

Machnumber,

velocity0.050.32,

6350fps

Angleofattack 3to12

Aphotoofthemodelinstalledinthetunnelalongwith

thesupportsystemisshowninFigure8.Thesupport

systemconsistsofasting,arollcoupling,alongsting

(cannon), and a vertical post. The support system

moves in the pitch plane with the model fixed at the

X=17.75, Y=Z=0 position. The aerodynamic

characteristicofthemodelisshowninFigure9.

A tunnel-empty calibration with the support system

aloneinstalledinthetunnelwithrepeatrunsatseveral

Q and values is used here. Generally, an emptytunnelcalibrationcanalsobeusedwithoutanyofthe

wake-oriented support system components (these can

be modeled in TWICS). However, the vertical post

shouldbepartoftheemptytunnelcalibration.Forthe

present case, the only good calibration set that is

availableistheonewiththecompletesupportsystemin

thetunnel. Thisessentiallymeansthat thesingularity

distributionusedherewillcorrespondtoonlythemodel

andwakeandnotanyofthestingcomponents.

Thesingularitydistributionusedfortheellipticalwing

model is shown in Figure 10. Wall correctioncalculationsweredoneforanumberof Qvaluesinthe

range3to12.Resultsindicatethattheblockageisnominallyconstantacrossthisrangeat0.001.Itwas

alsofoundthatforQvalueslessthan10psf,thescatter

in the computed blockage values is large, due to the

wall systemnotableto distinguishthe weaksignature

from other random variations. The corrections onM

and Qarealsoconstantintherangeandtheyscalewith M and Q as given by Equation (1) and (2)

( /M M = ; / 2Q Q = , approximately). The

angle of attack correction scales with with

/ =0.011, approximately. The classical value

calculatedfromchartsinReference7isalso0.011.

Thecorrectionsontheforcecoefficientsareshownin

Figure11.Thecorrectionondragreachesamaximum

of20dragcountsat12 angleofattack;correctionson

liftanddragduetobuoyancyarenegligiblysmall.

Comparison of measured incremental wall signatures

withTWICS-fitvaluesat=3,2and12isshowninFigure12.Agoodglobalfithasbeenobtainedevenat

these low wall signature values. Note that the wall

signaturescaleis0.01comparedtothe0.1scaleused

forthetrapezoidalwingcaseinFigure7.

10. DEFLECTEDWAKEANALYSIS

Thestandardapproachusedinclassicalwallcorrection

methodsas well asin wall signature methodssuchas

TWICSistomodelwall-inducedliftinterferenceusing

a number of line doublets distributed along the wing

span with starting locations along the chord line.

Eachline doublet is alsoequivalent to a semi-infinite

streamwise string of point doublets, pointing opposite

to the lift vector direction, and originating from the

same chord locations of the wing. For a three-

dimensionalwingtherewillbeanumberofsuchpoint

doubletstringsatdifferentspanlocations.

For high-lift wings as well as rotorcraft in forward

flight,thevorticityinthewakeisorientedatanangleto

the freestream in an average sense. In addition, the

correspondingpointdoubletvectorsareinclinedto the

free-streamatanangleprimarilydependentonthelift-

drag ratio, as modeled by Heyson8. Heysons wake

deflectionmodelthusassumesastraight-linetrajectory

for thisvorticity track. This model is equivalent toa


8/15

8

semi-infinite string of point doublets placed along a

deflected wakeline,withthe pointdoubletdirectionat

an angle to the lift vector direction as shown in the

sketchinFigure13a.

The Heyson model calculates thedeflected wake wallinterferencevelocitiesintwosteps.Firsttheeffective

wakedeflectionangleiscalculated.Thisisafunction

of the wing span, area, lift, induced drag, tunnel

geometryandfree-streamvelocity.Thesecondstepin

the Heyson model is to calculate the singularity

strengthsandcorrespondinginterferencevelocities.As

shown Figure 13b for the trapezoidal wing case, the

pointdoubletsaresplitintotwotypes,onepointingto

YproducingthelifteffectandanotherpointingtoX

producingtheinduceddrageffect.

Since the results from the first step of the Heyson

analysisrevealedthatthevorticitymaybedeflectedby

asmuchas10ontheaveragefromthechordplaneat

maximum lift condition, a modified version of the

TWICScodewas created toinclude these effects. In

order to implement the deflected wake, the line

doubletsusedinTWICSwerechangedtoYdirection

pointdoubletstringsnotingthatasummationofachain

ofthesepointdoubletsisequivalenttothelinedoublet

witha90orientation.Itisthenpossibletodeflectthe

vorticity track by specifying the point doublets to be

located along the deflected line. This requires the

generation of a new PVD database with interference

solutionsduetoYdirectionpointdoublets.Thefirst

partoftheHeysonmodelwasaddedtoTWICStofirst

computetheeffectivewakedeflectionangle. Changeswere made in the interpolation scheme in TWICS to

locate the deflected lifting point doublets in the

databasegridfollowedbyinterpolationandsummation.

Changeswerealsomadetodeflectthewakeblockage

pointdoubletsbyacorrespondingangle;thiseffectwas

howeverfoundtobeminor.

Figure 14 shows the wall signature fit from the

deflectedwakeversionofTWICSfor=32(forrows1and12,mostaffectedbythedeflectedwake),which

canbecompareddirectlywithFigure7fromtheregular

TWICScalculation.Itisclearthatthedeflectedwake

modificationhasimprovedthefitinthesetworowsofthe South and North walls. Figure 15 further

demonstratesthatthisistrueforalltherows,showing

thatthestandarddeviationofthefithasdecreasedasa

result of this modification. Note that the global

standard deviation drops by a factor of almost two.

Finally, Figure 16 shows the change in the mean

corrections as a result of the deflected wake

modification. A closer agreement with the classical

valuescanbeobserved.

11. CONCLUSION

Awallsignature-basedcorrectionmethodforthewalls-

down configuration of the 14x22-Ft tunnel has been

implemented based on a new wall pressure

measurement system at this tunnel. Example resultsfromrecentsemispanandfull-spantestsatthefacility

have been presented. Comparison with classical,

closed-formmethodsofwallcorrectionindicatethatthe

wall signature method provides comparable values.

Further, the model-provided wall signatures compare

very well with measured values in a range of model

sizes, angle of attack and Q values. Although true

validation of results require cross-tunnel tests using

scaledmodels,itcanbestatedthatthenewmethodis

an improvement over the existing classical methods

since boundary measurements are used in addition to

forceandmomentinputsandsincemodelshowsclose

agreementwiththemeasuredwallpressures.

REFERENCES1. Hemsch,M,Grubb, J., Krieger,W., andCler, D.,

Langley Wind Tunnel Data Quality Assurance

CheckStandardResults,AIAA2000-2201,June2000.

2. Kuhl,D.D.,andEverhart,J.L.,Measurementand

Control of the Uncertainty of Scanning Pressure

Transducer Measurements, AIAA 2003-3816, June

2003.

3. Iyer, V., and Everhart, J.L., Application of

Pressure-Based Wall Correction Methods to Two

NASA Langley Wind Tunnels, AIAA 2001-2472,

June2001.

4. Garner,H.C.,Rogers,E.W.E.,Acum,W.E. A.,and Maskell, E. C., Subsonic Wind Tunnel Wall

Corrections,AGARDograph109,October1966.

5. Ewald, B. F. R. (Editor), Wind Tunnel Wall

Correction,AGARDograph336,October1998.

6. Quinto,P.F.,andOrie,N.M.,Langley14-by22-

FootSubsonicTunnelTestEngineersDataAcquisition

andReductionManual,NASATM4563,June1994.

7. Barlow,J.B.,Rae,Jr., W.H., Pope, A,Low-Speed

Wind Tunnel Testing, ThirdEd., John Wiley & Sons,

1999,pp.367-427.

8. Heyson,H.H.,LinearizedTheoryofWindTunnel

Jet-Boundary Corrections and Ground Effect for

VTOL-STOLAircraft,NASATRR-124,1962.9. Ulbrich, N. and Boone, A.R., Determination of

theWallBoundaryConditionoftheNASAAmes11ft

TransonicWindTunnel,AIAA2001-1112.

10. Ulbrich, N., The Real-time Wall Interference

CorrectionSystemoftheNASAAmes12-footPressure

Tunnel,NASA/CR-1998-208537,July1998.

11. Maskell,E.C., ATheoryof theBlockageEffects

onBluff Bodiesand StalledWings ina Closed Wind

Tunnel,R&M3400,November1963.


9/15

9

Y

Z

-15-10-5051015

-5

0

5

Cross section view looking downstream

Southwall

Northwall

R

ows

912

Rows 58

Rows14 Z

Y

X

Z

0 10 20 30 40-10

0

10

INFLOW

PLANE

Port distribution on the South Wall

Row 1

Row 2

Row 3

Row 4

X

Y

0 10 20 30 40

-10

0

10

Top view and port distribution on the ceiling

INFLOW

PLANE

Rear modelmounting cart

Front modelmounting cart

Row 5

Row 6

Row 7

Row 8

Figure1.Newwallpressureportsforthe14x22-Fttunnel(ellipticalwingmodeloutlinesshown).


10/15

10

Figure2.Photoofthetrapezoidalwingmountedinthe

14x22-Ft. tunnel. Photo courtesy of NASA Langley

PhotoLab.

Figure3.Aerodynamicscharacteristicsofthe

trapezoidalwing.

X

Y

0 10 20 30

-10

0

10

South wall

North wall

Turn table

Modelcenterof

rotation,

X=17.7

5

Figure4a.Singularitydistributionfortrapezoidalwing,

topview.

X

Z

15 20 25

-5

0

5

Point doublets for body blockage

Line dblt.start locn.s

Pt. doublets forsep. wake

Ceiling

Modelcenterof

rotation,X=17.7

5

Figure 4b. Singularity distribution for trapezoidal

wing,sideview.

CL

CD

0 0.5 1 1.5 2 2.5 3 3.50

0.2

0.4

0.6

0.8

1


11/15

11

ANGLE OF ATTACK (Uncorrected)0 10 20 30 40

0

0.005

0.01

0.015

0.02

BLOCKAGEFAC

OR,

TWICS

ANGLE OF ATTACK (Uncorrected)

0 10 20 30 400

0.5

1

1.5

2

Q,psf

TWICSclassical

Maskell


0

0.5

1

1.5

2

3/4 chord line avg.1/4 chord line avg.classical method

Figure5. Meancorrectionsforthetrapezoidalwingin

the14x22-Fttunnel.


-0.02

0

0.02

0.04

0.06

0.08

0.1

COEFFT.

CORREC

TIONS

CL

CD

CD,Buoy

CL-clas.

CD

-clas.

ANGLE OF ATTACK (Uncorrected)

0 10 20 30 400

0.002

0.004

0.006

0.008

0.01

PITCH.

MOMENTCORRECTION

CL

0 1 2 30

0.2

0.4

0.6

0.8

1

CD

CorrectedUncorrected

Figure 6. Coefficient corrections for the trapezoidal

winginthe14x22-Fttunnel.


12/15

12

0 10 20 30 40-0.1

-0.05

0

0.05

0.1 ROW 1 (South Wall)

0 10 20 30 40-0.1

-0.05

0

0.05

0.1 ROW 9 (North Wall)

0 10 20 30 40-0.1

-0.05

0

0.05

0.1 ROW 5 (Ceiling)

0 10 20 30 40-0.1

-0.05

0

0.05


0 10 20 30 40-0.1

-0.05

0

0.05

0.1 ROW 6 (Ceiling)

X=17.7

5ref.l

ine

0 10 20 30 40-0.1

-0.05

0

0.05


0 10 20 30 40-0.1

-0.05

0

0.05


0 10 20 30 40-0.1

-0.05

0

0.05

0.1 ROW 7 (Ceiling)

0 10 20 30 40-0.1

-0.05

0

0.05


0 10 20 30 40-0.1

-0.05

0

0.05


0 10 20 30 40-0.1

-0.05

0

0.05

0.1 ROW 8 (Ceiling)

0 10 20 30 40-0.1

-0.05

0

0.05


ROW 1 (South Wall) ROW 5 (Ceiling) ROW 9 (North Wall)




Figure 7. Measured and TWICS-fit wall signatures for the trapezoidal wing test. Symbols are incremental

measurements; lines are TWICS-fit values. Three angles of attack 4 (squares, dash-dot line), 10 (triangles,

dashedline),32(circles,solidline)areshown.X-axisisthestreamwisedistanceX;Y-axisistheincremental

normalizedwallvelocity.


13/15

13

Figure 8. Photoof the ellipticalwingmountedin the

14x22-Ft. tunnel. Photo courtesy of NASA Langley

PhotoLab.

CL

-0.25 0 0.25 0.5 0.75 10

0.02

0.04

0.06

CD

CorrectedUncorrected

Figure9.Aerodynamiccharacteristicsoftheelliptical


X

Y

16 18 20 22 24

-4

-2

0

2

4

Tunnel center line

X=17.75 ref. line

Pt. doubletsfor wake

Line dblt.sfor lift

Pt. dblt.s

for vol. blockage

Figure10.Singularitydistributionfortheellipticwing

modelinthe14x22-Fttunnel.

ANGLE OF ATTACK (Uncorrected)0 5 10

0

0.001

0.002

COEFFT.

CORRECTIONS

CD

CL

CD,Buoy

Figure 11. Coefficient corrections for the elliptical



14/15

14

0 10 20 30 40-0.01

-0.0075

-0.005

-0.0025

0

0.0025

0.005

0.0075


0 10 20 30 40-0.01

-0.0075

-0.005

-0.0025

0

0.0025

0.005

0.0075

0.01 ROW 5 (Ceiling)

0 10 20 30 40-0.01

-0.0075

-0.005

-0.0025

0

0.0025

0.005

0.0075


0 10 20 30 40-0.01

-0.0075

-0.005

-0.0025

0

0.0025

0.005

0.0075


0 10 20 30 40-0.01

-0.0075

-0.005

-0.0025

0

0.0025

0.005

0.0075


0 10 20 30 40-0.01

-0.0075

-0.005

-0.0025

0

0.0025

0.005

0.0075


0 10 20 30 40-0.01

-0.0075

-0.005

-0.0025

0

0.0025

0.005

0.0075


0 10 20 30 40-0.01

-0.0075

-0.005

-0.0025

0

0.0025

0.005

0.0075


0 10 20 30 40-0.01

-0.0075

-0.005

-0.0025

0

0.0025

0.005

0.0075


0 10 20 30 40-0.01

-0.0075

-0.005

-0.0025

0

0.0025

0.005

0.0075


0 10 20 30 40-0.01

-0.0075

-0.005

-0.0025

0

0.0025

0.005

0.0075


0 10 20 30 40-0.01

-0.0075

-0.005

-0.0025

0

0.0025

0.005

0.0075


X=17.7

5ref.line





Figure 12. Measured and TWICS-fit wall signatures for the elliptical wing test. Symbols are incremental

measurements;linesareTWICS-fitvalues.Threeanglesofattack3(squares,dash-dotline),2(triangles,dashed

line),12(circles,solidline)areshown.X-axisisthestreamwisedistanceX;Y-axisistheincrementalnormalized

wallvelocity.


15/15

15

X

Y

) wake deflection

=90-

X

Y

) wake deflection

-X dir. pt. dblts for induced drag effect

-Y dir. pt. dblts for lift effect

Figure13.Part(a)(top),Heysonwakedeflectionmodel.Part(b)(bottom),Superpositionusedinthe

Heysonmodel.

X0 10 20 30 40

-0.1

-0.05

0

0.05


X0 10 20 30 40

-0.1

-0.05

0

0.05


Figure 14. Comparison of measuredincrementalwall

velocities (symbols) with computed values from a

modifieddeflectedwakeTWICScode,=32.

ROW NUMBERS

STD.

DEVIATION

1 2 3 4 5 6 7 8 9 10 11 120

0.005

0.01

0.015

0.02

Average value for all rows,TWICS modified for deflected wake

Average value for all rows,TWICS original version

TWICS originalTWICS modifiedfor deflected wake

Figure 15. Improvement in the standard deviation ofthe TWICS fit to the measured wall pressures along

each wall row with the deflected wake modification,

exampleshownaboveisfor=32.


0

0.5

1

1.5

2

Q,psf

TWICS

Classical

TWICS-Defl. Wake


0

0.5

1

1.5

2

1/4 chord (Defl. Wake)classical

Figure16.Improvementinwallcorrectionswiththe

TWICSdeflectedwakemodification.

wall correction 14x22 wind tunnel

Documents