aiaa-2002-0846 numerical viscous flow analysis of an

AIAA-2002-0846Numerical Viscous Flow Analysis of anAdvanced Semispan Diamond-WingModel at High-Lift Conditions

F. Ghaffari, R. T. Biedron and J. M. LuckringNASA Langley Research CenterHampton, Virginia

For permission to copy or republish, contact the American Institute of Aeronautics and Astronautics1801 Alexander Bell Drive, Suite 500, Reston, VA 20191-4344

40th AIAA Aerospace Sciences Meeting & Exhibit14-17 January, 2002

Reno, Nevada

AIAA-2002-0846

1American Institute of Aeronautics and Astronautics

Numerical Viscous Flow Analysis of an Advanced Semispan Diamond-Wing Model at High-Lift Conditions

F. Ghaffari∗, R. T. Biedron§, and J. M. Luckring†

NASA Langley Research CenterHampton, Virginia

AbstractTurbulent Navier-Stokes computational results arepresented for an advanced diamond wing semispan modelat low speed, high-lift conditions. The numericalresults are obtained in support of a wind-tunnel test thatwas conducted in the National Transonic Facility (NTF)at the NASA Langley Research Center. The modelincorporated a generic fuselage and was mounted on thetunnel sidewall using a non-metric/constant-widthstandoff. The analyses include: 1) the numericalsimulation of the NTF empty tunnel flowcharacteristics, 2) semispan high-lift model with thestandoff in the tunnel environment, 3) semispan high-lift model with the standoff and viscous sidewall in freeair, and 4) semispan high-lift model without thestandoff in free-air. The computations were performed atconditions that correspond to a nominal approach andlanding configuration. The wing surface pressuredistributions computed for the model in both the tunneland in free air agreed well with the correspondingexperimental data and they both indicated smallincrements due to the wall interference effects. However,the wall interference effects were found to be morepronounced in the total measured and the computed lift,drag and pitching moment due to standard induced up-flow effects. Although the magnitudes of the computedforces and moment were slightly off compared to themeasured data, the increments due the wall interferenceeffects were predicted well. The numerical predictionsare also presented on the combined effects of the tunnelsidewall boundary layer and the standoff geometry onthe fuselage fore-body pressure distributions and theresulting impact on the overall configurationlongitudinal aerodynamic characteristics.

NomenclatureBC boundary condition

BL boundary layerb/2 reference semispan, 2.6927 ftc local chord, ftCD drag coefficient, Drag/q∞Sref

CL lift coefficient, Lift/q∞Sref

Cm pitching moment coefficient, pitching moment/q∞Sref C

Cp pressure coefficient, p-p∞/q∞

CFD computational fluid dynamicsC-O grid topology, C streamwise and O circumferentialcref cruise wing reference chord at y/(b/2)=0.3, 3.295 ftC wing mean aerodynamic chord, 3.143 ftDERA Defense Evaluation and Research AgencyFP flat plateH-H grid topology, H streamwise an H spanwiseH-O grid topology, H streamwise and O circumferentialLaRC Langley Research CenterLE leading edgeLEF leading-edge flapM∞ free-stream Mach numberMIF Model/standoff-In-Free-airMIT Model/standoff-In-TunnelMNIF Model/No-standoff-In-Free-airNASA National Aeronautics and Space AdministrationNTF National Transonic FacilityOL overlap, ftp local static pressure, psfp∞ free-stream static pressure, psfq∞ free-stream dynamic pressure, psfRANS Reynolds-Averaged Navier-StokesRft unit Reynolds number, per footRC Reynolds number, based on CSref wing reference semi-area, 6.5908 ft2

SA Spalart-AllmarasTTCP The Technical Cooperation ProgramTE trailing edgeTEF trailing-edge flapTS tunnel station, ftU/U∞ ratio of local to free-stream axial velocityWIC wall interference correctionx/c fraction of wing local chordy/(b/2) fraction of model semispanXYZ Reference coordinate systemα angle of attack, degreesαc corrected angle of attack for wall interference, degreesy+ inner law variable

∗ Research Engineer, Senior Member, AIAA

§ Research Scientist† Senior Research Engineer, Associate Fellow, AIAA

Copyright © 2002 by the American Institute of Aeronautics andAstronautics, Inc. No copyright is asserted in the United Statesunder Title 17, U. S. Code. The U. S. Government has a royalty-free license to exercise all rights under the copyright claimedherein for Governmental Purposes. All other rights are reserved bythe copyright owner.

AIAA-2002-0846


IntroductionThe high-lift flow and the resulting aerodynamicsexperienced by an aircraft in take-off and landing aresome of the most complex and difficult phenomenon tosimulate, either experimentally with wind-tunnel tests,or numerically with the computational fluid dynamics(CFD) methods. For an aircraft to achieve the high-liftlevels, required during takeoff and landing, it typicallydeploys mechanical systems that are referred to as high-lift devices. These devices are usually comprised ofleading and trailing edge flaps designed to maximizeattached flow over the vehicle for acceptableaerodynamic lift, efficiency, and handling qualities.However, for high performance military aircraft (subjectof the present study) this must be achieved within thecontext of a relatively low aspect ratio and thin wing,and thus the resultant high-lift flow field can varyconsiderably from that of conventional commercialtransport configurations.The numerical simulation of high-lift flows is verydifficult because of the inherent geometrical complexityas well as the complex flow interactions that can occur.Such geometrical complexity introduced by high-liftdevices includes physical gaps, cavity or cove regions,exposed flap side-edges that are often sharp, flap hinge-lines that may be sharp or rounded, etc. Progress hasbeen made in recent years to numerically simulate thecomplex high-lift flow aerodynamics with ReynoldsAverage Navier-Stokes (RANS) formulations based onmulti-block structured grid technology (Refs. 1-3) withvarious degrees of success. However, the geometricalcomplexity of the high-lift configuration often requiresa tremendous amount of time and resources to be spentin grid generation to set up such a computation. Analternative approach based on the unstructured gridtechnology has received attention in the recent years(Refs. 4, 5), primarily because of its inherent flexibilityin discretizing complex geometry. However, it is alsowidely believed that the existing unstructured gridtechnology, with capabilities to simulate the complexhigh-lift viscous flow characteristics, is still in thedevelopmental stage and may not be ready forapplication by the general user community.The present overall CFD plan, shown schematically inFig. 1, has two main objectives. The first objective isto calibrate a state-of-the-art RANS method forpredicting the low-speed high-lift aerodynamics of anadvanced high performance military wing conceptrecently tested in the NTF (Ref. 6) at the NASALangley Research Center (LaRC). The semispan wind-tunnel model was designed as part of a multi-nationalcollaborative effort (Ref. 7) under the auspices of TheTechnical Cooperation Program (TTCP). TTCPparticipants involved in this effort included the United

States Department of Navy, National Aeronautics andSpace Administration (NASA), and the United KingdomDefense Evaluation and Research Agency (DERA). Themodel incorporated a generic fuselage and was mountedon the tunnel sidewall using a constant width standoff.For the present code calibration purposes, the CFDmodel included not only the semispan configuration butalso the wind tunnel solid walls representing the NTFtest section environment. The focus of the presentnumerical analysis is on the high-lift configuration witha specific rigging arrangement designed for approachlanding conditions. These computations include anempty tunnel simulation to calibrate the viscoussidewall flow (Fig. 1a) and then a simulation of themodel-in-tunnel configuration (Fig. 1b).The second objective of this study is to numericallyassess the interference effects due to the wind-tunnelwalls and the standoff geometry for this configuration.This is accomplished numerically through a systematicremoval of wall interference effects due to solid-wallconfinement (Fig. 1c) and due to viscous sidewall andstandoff combination (Fig. 1d). Experimental wallinterference effects were also obtained, and these are usedto help assess the numerical results. The presentanalysis on this slender vehicle also contributes to theongoing activities, both experimentally (Ref. 8) andnumerically (Ref. 9) at LaRC to develop a semispantest capability at the NTF, which have been primarilyfocused on commercial transport configurations.

Wind-tunnel model and test descriptionThe semispan wind-tunnel model consisted of a genericfuselage with a 1.5-inch non-metric/constant-widthstandoff and a cropped diamond wing planform withleading and trailing edge sweep of 40o and -40o,respectively. The wing was designed for multi-missioninterdisciplinary military requirements for cruise, highangle of attack maneuver, as well as for low-speed,high-lift performance. The photographs in figure 2show the high-lift version of the model from twodifferent perspective views. The wing consisted of a fullspan leading-edge flap, a part-span slotted trailing-edgeflap, and a deflectable shroud ahead of the trailing-edgeflap. The semispan model was mounted on the tunnelsidewall by including a constant width standoff designedto minimize the sidewall boundary layer (BL) effects onthe model aerodynamics. The primary purpose of thetest was to develop an experimental database for fourdifferent variations of the diamond wing with respect tothe flap rigging arrangements (gap and overlap). Theseconfiguration variations included two high-lift riggingsfor approach and landing, one high-lift rigging for take-off, and the baseline cruise-model with no controlsurface deflections. Data from this experiment include

AIAA-2002-0846


static surface pressures, configuration (i.e., wing andfuselage combination) forces and moments, andaeroelastic deformations for many high-lift settings.The semispan model was 7.7 ft long and 2.7 ft wideexcluding the 1.5-inch constant width standoff. Theentire test was conducted in the air mode mainly becausethe model was designed for tests at elevated pressures;the combination of model size and pressure producedfull-scale Reynolds number data. The model wasinstrumented with approximately 450 orifices tomeasure the surface pressures. The majority of thepressure orifices were distributed over the wing and inparticular around the high-lift system. The wingpressure orifices were primarily distributed along sixchord-wise stations located at y/(b/2)=0.15, 0.30, 0.45,0.55, 0.70, and 0.80 (Fig. 3)The high-lift wing configured for approach and landingwas chosen as the baseline configuration for the presentnumerical analysis. The high-lift control surfaces forthis baseline configuration included a 22o deflected full-span leading-edge flap (LEF), 23o deflected shroud, and35o deflected trailing-edge flap (TEF). In addition, theselected baseline model incorporated a 0.5%cref gap and a2%cref overlap rigging (Fig. 4) arrangement.Representative flow conditions for approach and landingwere also selected for the numerical analysis and theyare α=10o, M∞ = 0.2, Rft=7.7x106. For reference, theselected unit Reynolds number of 7.7x106 correspondsto the Reynolds number of 24.2x106 based on the wingmean-aerodynamic chord. Limited aerodynamic analysisof the experimental data has been reported in Ref. 10.

Computations and flow field analysisThe present numerical analysis is performed with theMulti-block, structured-grid based CFD code known asCFL3D (Ref. 11). The code is well documented and hasbeen extensively calibrated for variety of applicationswith different classes of flows and configurations (Refs.12-15). The algorithm is based on the compressible,time-dependent, Reynolds-averaged Navier-Stokesequations that are written in a curvilinear coordinatesystem. A cell-centered, finite volume approach is usedto solve the equations in a conservative form. Anupwind-biased, flux-difference-splitting (Ref. 16) is usedto solve the inviscid terms whereas central differencingis applied to solve the viscous terms. The presentnumerical results are all based on the one-equationmodel of Spalart-Allmaras (Ref. 17). The solutionspresented in this report are all obtained by the use ofmulti-griding and multi-sequencing techniques toaccelerate the convergence characteristics. The variousgrid-block interfaces in the physical domain areconnected to one another either in a two-dimensionalplanar form or a three-dimensional non-planar form. The

flow primitive variables are interpolated across thevarious block interfaces using a searching techniquebased on a combination of linear and polynomialequations as discussed in Ref. 18.

Empty tunnel flow simulationThe empty tunnel flow simulation was conductedprimarily for two reasons. The first reason was toestablish that the tunnel sidewall BL could bereasonably simulated. The second reason was todetermine whether a mixed viscous and inviscidboundary condition for modeling the tunnel walls wasadequate to simulate the test section flow field.The initial numerical model included only the nominalNTF test section and the tunnel sidewall was treatedwith a viscous boundary condition whereas the otherthree walls were simulated with an inviscid boundarycondition (Fig. 5). The characteristic inflow BC and theoutflow BC with specified pressure ratio (p/p∞) of 1.0were imposed at the tunnel inlet and exit plane,respectively. The BL rake data (Ref. 19) measured attunnel sidewall station 13 ft (at the model center ofrotation, see Fig. 5) are used to assess the accuracy ofthe numerical predictions. The rake was 6.25 inches talland incorporated 29 probes with the first 8 probesdistributed uniformly over the first inch to measure thenear field pressure normal to the sidewall. The rake dataanalysis indicated that the edge of the BL (i.e., U/U∞

=0.99) occur at height of ~3.8 inches (~ 96.52 mm).Four sets of Cartesian grids (i.e., H-H topology) weregenerated with different grid resolution to address gridsensitivity effects on the results. Complementary to theBL rake data, computations were performed at M∞= 0.2,and Rft=2x106, based on the Reynolds-averaged Navier-Stokes (RANS) formulation with Spalart-Allmaras(SA) turbulence model. The correlation of thesecomputed results with the measured BL thickness (seefig. 6) indicated an expected disagreement because thenumerical model did not include a proper BL profile atthe inflow plane.Due to the simplicity of the tunnel sidewall BLsimulation, it was decided to exploit flat-plate (FP)theory for estimating the BL thickness growth along thetunnel sidewall. The flat-plate theory estimates (buriedunder the open symbols) of the BL thickness growthwere found to be very close to the turbulent Navier-stokes results computed for the nominal NTF testsection (see Fig. 6). As a result, it was decided to usethe flat-plate BL theory and the existing experimentaldata point (i.e., BL height of 3.8 inches (96.52 mm) atTS-13) to extrapolate a virtual origin to the NTF testsection that would provide a better approximation of theBL thickness at TS-13. The tunnel virtual origin was

AIAA-2002-0846


determined (Fig. 6) from this analysis to beapproximately 15 ft (4600 mm) ahead of the nominalNTF test section.Finally the accuracy of the estimated NTF virtual originby the flat-plate theory was verified by applying theturbulent (SA) Navier-Stokes method. This calibrationwas performed by modifying the volume grid generatedearlier for the computation of the nominal NTF testsection (i.e., 65-axial, 129-normal to the sidewall, 65-lateral, y+ ~ 0.9) to accommodate grids for the upstreamextension of the tunnel geometry. The turbulent Navier-Stokes results for the BL height growth along theextended NTF sidewall is also shown in figure 6. TheNavier-Stokes results for the extended NTF sidewallclearly reaffirm that the flat-plate BL height estimateswere very reasonable. It should be noted that with thisapproach the relatively complex contraction-cone andthe settling-chamber geometry are essentially replacedwith a simple linear upstream-extension of the squarecross-section, tunnel geometry.The measured velocity profile along with the axialvelocity profiles computed for the nominal and theextended NTF test section were also examined. Thoughnot presented here, the comparison clearly illustrated anexcellent agreement between BL rake measurements andthe computed velocity profile with the extended NTFtest section. In addition, as part of the NTF empty-tunnel flow simulation study, turbulent (SA) Navier-Stokes computations were also carried out bysimulating the viscous BL flow on all four walls of theextended tunnel geometry. The results from this studyrevealed that only a slight thinning of the sidewall BLoccurred and that it did not manifest itself until reachingTS 5 ft. This thinning of the BL thickness wascomputed to be approximately 0.1 inch (less than 3% ofthe overall BL thickness) at TS 13 ft. The empty tunnelflow simulation study provided sufficient insight intothe tunnel flow characteristics (i.e., establishment of atunnel virtual origin and the proper resolution ofsidewall BL characteristics) that can directly be appliedto the flow simulation over the semispan diamond wingmodel in the tunnel.

Model/standoff-In-Tunnel (MIT)This section includes discussion on the computationalgrid, numerical solution development and typical flowfield results.A multi-block structured-grid was developed to discretizethe semispan high-lift diamond wing model with thestandoff geometry in the nominal NTF test section (seeFig. 7). The high-lift brackets (Fig. 2) were notmodeled in the present numerical analysis, because of:1) the added grid generation complexity, 2) their

presence should not have a significant impact on theoverall configuration aerodynamics. The semispandiamond wing numerical model is rotated geometricallyand set at 10o angle-of-attack with respect to the tunnelfree-stream. Provisions were made from the emptytunnel flow simulation to properly size the grid over theviscous sidewall and to accommodate the tunnelupstream block extension (not shown in Fig. 7). Sameinflow and out flow BC as those used in the emptytunnel flow simulations were imposed at the tunnelinlet and exit plane, respectively. Viscous BC wasinvoked on all lifting surfaces of the model with theexception of the fuselage base. Inviscid BC wasimposed on the fuselage base to alleviate any possibleconvergence difficulties due to the expected complexityof the flow field (i.e., unsteady, turbulent wake flowfield). In addition, the contribution of the fuselage baseto the overall configuration forces and moment will besubtracted for subsequent analysis; this is consistentwith the experimental data. The volume grid consistedof 39-grid blocks and containing approximately 7million grid points.The XYZ reference coordinate system for the grid isdefined such that: positive X is from upstream todownstream, positive Y is normal to the sidewall, andpositive Z is from tunnel ceiling to floor. The volumegrid is defined in metric units where the viscoussidewall is located at Y= -1.5 inch (i.e., –38.1 mmstandoff width) with the opposing side at Y=8.077 ft(i.e., Y=2461.9 mm). The overall longitudinal lengthof the nominal NTF test section is defined to be 25 ft(i.e., 7620 mm) long.A close-up view of the surface grid for the semispandiamond wing model along with the tunnel sidewallgrids is shown in Fig. 8. The complexity of thegeometry and the care taken to resolve the cove and thegap regions are clearly illustrated. The inboard edge ofthe deflected TE flap is abutted against the fuselagewhereas the outboard side-edge is exposed to the flow.Turbulent Navier-Stokes computations were performedby setting the tunnel free-stream conditions to M∞=0.2, Rft =7.7 x106, and at zero degree angle-of-attack.However, note that the angle of attack (α) for the modelis 10o, because as discussed previously, the geometricalmodel is rotated to 10o angle of incidence relative to thetunnel free-stream condition. A solution convergenceprocedure was developed for this baseline MIT case thatcould be applied in a consistent way to the other casesin the present investigation.With this procedure, the overall solution convergencewas achieved using three grid levels (i.e., coarse,medium, and fine). Over the course of this solutiondevelopment, the overall residuals were reduced by about

AIAA-2002-0846


2.5 orders of magnitude and the oscillations in thecomputed total lift, drag, and pitching moment werereduced to a negligible level. Over the last 500iterations of the fine grid solution, the averagevariations in total lift, drag, and pitching moments werefound to be ±0.07%, ±0.13%, and ±0.3%, respectively.Figure 9 shows the overall solution convergence historythat took about 60 hours of Cray C-90 and requiredabout 300MW of memory. Figure 9 also shows a morequantitative variation of the residuals for each blockduring the course of the solution development at every1000 iteration intervals.A composite image summarizing the computed flowfield in terms of pressure coefficient for the MIT case isshown in Fig. 10. The surface pressure coefficients areshown on the semispan diamond wing model, tunnelsidewall, inflow/outflow and the floor plane of theupstream portion of the extended NTF test section. Inaddition to the surface pressure coefficients, the resultsin Fig. 10 show traces of several particles released inthe field near the tunnel sidewall just ahead of fuselageand around the leading edges of the wing and the trailingedge flap. A few particles were also released close to thefuselage blunt base to highlight the associated wakeflow-field structure. The particle traces are computedwithout any restriction to a particular computationalgrid plane. The computed flow characteristics generallyshow the desired attached flow, from a high-lift designstand point, over the LEF, main wing, shroud, and TEFfor the most part. The close-up image on the upperright corner of the figure shows the low-pressurefootprint associated with the TEF side- flow separationforming a vortex (off surface structure not shown here).

Model/standoff-In-Free-air (MIF)The grid strategy chosen for the MIF computationsutilized the existing MIT grid without any alteration.The MIF grid required six new grid blocks to extend theMIT tunnel walls to the nominal far field (see Fig. 11).The radial extent of the far-field boundary was chosen tobe about five overall fuselage body-lengths (i.e., 12.2C)away from the tunnel centerline. These six blocks addeda total of approximately 160,000 points to the existingMIT volume grid.The identical procedures developed to obtain theturbulent Navier-Stokes solutions for the MIT case wereapplied to acquire the computational results for the MIFcase at the same flow conditions. The MIF solutionconvergence characteristics and the resulting flowfeatures were almost indistinguishable from those of theMIT case shown earlier in Figs. 9 and 10, respectively.As a result, they are not presented here.

Model/No-standoff-In-Free-air (MNIF)The volume-grid blocks associated with the standoffwere extracted from the existing MIF computationalgrid. This modification resulted in a total of 38 gridblocks, and about 6.5 million grid points, to remain forthe numerical representation of the model/no-standoff infree-air (MNIF). The same boundary conditions as theMIF case were applied on all surfaces with theexception of the configuration plane-of-symmetry wherethe general symmetry plane boundary condition wasimposed. The computational procedure established underthe MIT and MIF solution development were applied toobtain the turbulent Navier-Stokes solution for theMNIF case at α= 10o, M∞= 0.2, and Rft =7.7x106.

Predictions and correlation with dataTwo sets of experimental data, referred to as ‘with WIC’and ‘without WIC’, will later be presented in thisreport. While both data sets include all the standardwind-tunnel data corrections, the only difference betweenthem is that one contains the experimentally determinedsolid-wall interference correction (Ref. 20) effect and theother does not. The application of WIC to correct theexperimental data that corresponds to the flowconditions chosen for the numerical analysis resulted inan increase in the model angle of attack, free streamMach number, and dynamic pressure. The correspondingincrements in angle of attack (10o), free-stream Mach(0.2) and dynamic pressure (347.6 psf) were 0.6o, 0.001,and 4.54 psf, respectively. With the exception of theangle of attack, the increments to Mach number anddynamic pressure are considered relatively small in thepresent investigation. To numerically complement thecorrected experimental data, an additional MIFcomputation was performed at M∞= 0.2, Rft =7.7 x106,and a higher free-stream angle-of-attack of 0.6o. As aresult, the corrected angle of attack (αc) for thenumerical model (rotated to 10o angle of incidencerelative to the tunnel free-stream flow) in this newcomputation is 10.6o. The results from this new MIFcomputation at α c =10.6o, along with those originallyplanned will be later correlated with the correspondingexperimental data for surface pressure coefficients andoverall forces and moment.

Surface pressure coefficientsThe computed surface pressure coefficients, the NTFdata (both with and without corrections for wallinterference effects), and the wing geometry sectionalcuts at three span-wise stations are shown in Fig. 12.The experimental data are only shown for the wingmain element because of the inconsistencies found inmapping the experimental pressure-port locations with

AIAA-2002-0846


the corresponding CFD data over the shroud and theTEF. The CFD results are shown for four different casesthat include the MIT, MIF, MNIF, and the MIF at αc =10.6o along with the NTF experimental data. The latterexperimental data are shown for both corrected (denotedby ‘with WIC’) for the tunnel wall effects to representthe free-air results and uncorrected (denoted by ‘withoutWIC’) for the tunnel wall effects. In general, thepressure distributions computed over the main wingcompare well with the measured data for the model inboth the tunnel and free-air, and that they both indicatevery little increment due to the wall interference effects.Similar chord-wise pressure distributions and thecorresponding geometry sectional cuts, as shown in theprevious figure, are presented in Fig. 13 for the threewing outboard-stations. Note that the experimental dataare shown for the entire wing section and do not includeany high-lift element. Similar to the wing inboardstations, the WIC modification to account for the tunnelwall effects on the measured surface pressures appears tobe very small. The CFD results for the MIT and MIFshow small variations in the computed pressures aroundthe leading edge of the deflected flap. The agreementbetween the CFD results and the measured data aregenerally very good at all three stations.

Force and moment coefficientsThe computed overall lift, drag, and pitching momentcoefficients are shown in Fig. 14 along with thecorresponding NTF experimental data with and withoutWIC application. The close-up views of each respectivedata set near the flow conditions of interest to thepresent investigation are shown in the right-handcolumn. Consistent with the experimental data, it isimportant to note that the direct contributions from thestandoff geometry and the blunt fuselage-base areexcluded from all the computed coefficients. Theexperimental lift coefficients indicate a fairly linearvariation with angle-of-attack and the application ofWIC to account for the tunnel wall interference resultsin a decrease in lift curve slope as expected. Thecomputed lift coefficients from the MIT and MIF agreewell with the measured data, especially the incrementsdue to WIC (see the close-up results). The overallmagnitudes are slightly over-predicted by about 0.01.Also the MIF computed lift coefficient at α c = 10.6o

correlates well with the NTF data with WIC. As aresult, the lift curve slope for the free-air computations(i.e., MIF and MIF at αc = 10.6o) is also predicted well.Extracting the standoff geometry physically from thenumerical model causes only a small decrease in theMNIF computed lift coefficient relative to the MIF

prediction and it compares very well with the corrected(i.e., wall interference effects) NTF data.The measured pitching moment characteristics exhibitvery little change at low to moderate CL range, followedby a nose down tendency with increasing CL. Theexperimental data also indicate diminutive effects fromthe application of WIC method on the overall pitchingmoment characteristics. The correlation of the computedpitching moments with experimental data is reasonable,though the plot on the right-hand column may bemisleading because of the expanded scales.The measured drag polar indicates the expected trend andthat the modification due to WIC application appears tobecome more pronounced with increasing CL. Thecomputed drag coefficients are in general agreement withthe experimental measurements in terms of the trendsbut not the magnitude. Also note the increase in thepredicted drag coefficient (~55 counts) with MNIFcomputation compared to the MIF result. In addition tothe obvious geometrical differences between the twomodels (i.e., standoff), there is also the imposedboundary condition on the configuration plane-of-symmetry. In an effort to determine a possible sourcefor this drag change, further diagnostic analysis wasperformed as discussed in the following section.

Sidewall/standoff interference analysisThe predicted drag coefficients from MIF and MNIFcomputations are shown in Fig. 15 along with theresults for the other two CFD cases for completeness.The drag coefficients are plotted against the blocknumber in the computational domain. The resultsindicate that the computed drag remains fairly unchangedfor most the blocks with the exception of the first twothat define the configuration forebody longitudinally upto the fuselage/wing-LE juncture point. The majority ofthe increase in MNIF computed total drag, relative toMIF results, come from the configuration forebody(roughly about 40 counts) and the wing components(roughly about 10 counts). The effect over the forebodyis likely to be associated with the standoff geometry andthe boundary condition imposed on the wall to whichthe model is mounted, i.e., viscous wall for MIF andsymmetry plane for MNIF. As a result, theconfiguration forebody experiences a different flow-fieldsurrounding MIF computations compared to the MNIF.This forebody drag increment could be important to theprocess of estimating full span aerodynamics fromsemispan measurement. Similar data analysis was alsoperformed for the computed lift coefficients thatindicated negligible variations in all blocks with theexception of the wing components, which showedslight variations.

AIAA-2002-0846


Computed pressure distributions along the fuselagecenterline (i.e., fuselage and standoff boundary line, Y=0mm) are shown in Fig. 16. The results clearly indicatevery small differences between all the computationswith the exception of the MNIF solutions over theconfiguration forebody. The majority of the differencesoccur roughly over the front one-third of the fuselagewhere the MNIF predicted pressures appear to exhibitless suction on the upper surface and more compressionon the lower surface. As expected, only the computedpressure coefficients with MNIF indicate a stagnationpoint (i.e., Cp ~ 1.0) on the forebody unlike all theother solutions where their respective pressurecoefficients do not exceed 0.83. The bulging of thepressure distribution in the mid-fuselage region(0<X<1200) is attributed to the wing pressure fieldpropagation inboard onto the fuselage. It is alsointeresting to note that the fuselage pressuredistributions indicate a rapid decrease and increase inpressures, near the fuselage base on the upper (X~1700mm) and lower surfaces (X~1600 mm), respectively.This abrupt change in the fuselage pressure differencenear the base plane is likely caused by the flowseparation at the sharp corners of the base resulting in awake-like flow-field behind the blunt face of thefuselage. The cross sectional pressure distribution onthe forebody is analyzed next to identify thecircumferential extent of the pressure difference betweenMNIF and the other CFD results. The computedcircumferential pressure distribution at two fuselagecross sections of X=-500 mm and X=-200 mm areshown in Fig. 16. The first cross section is very closeto the fuselage nose (see Fig. 16 for relative locationwith respect to the overall fuselage length) and thesecond station is roughly about the mid-forebody. Thefigure also shows the corresponding cross-sectionalgeometry including the standoff component (hash-marked) for reference. The results clearly indicate thatthe pressure difference observed between the MNIF andthe other CFD solutions on the forebody is not confinedto the fuselage centerline but it also manifests itselfcircumferentially at both stations. Similar to the earlierfindings, the computed forebody pressures from theMIT, MIF, and MIF (at α c = 10.6o) cases show verylittle difference between one another. It is important torecall that the computed wing pressures did not changesignificantly between all the examined CFD cases. Thecombination of the viscous sidewall and standoffcertainly has a much larger effect on the forebodypressures compared with the MNIF predictions. Thedifferences in forebody pressures are large enough toimpact the computed drag coefficient (Fig. 16).Complementary to the surface pressure comparison atthe fuselage centerline (i.e., Fig. 16), figure 17 shows

the difference in the computed pressure coefficients atthe plane-of-symmetry between the MIF and MNIFsolutions. The grids are shown for every three points inboth directions for clarity. This result clearly shows thefield effect due to the presence of the combined sidewallBL and standoff geometry across the fuselage centerlineplane. Similar to the earlier findings, this result alsoindicates that the majority of the pressure differencebetween the two solutions is confined only in andaround the forebody region such as the flowcompression ahead of the forebody followed by the flowexpansion around it. These findings are consistent withthe approach taken to design the standoff shape (Ref. 9)around the fuselage forebody to minimize the sidewallBL impact over the configuration flow field and thusmeasured aerodynamic properties.

ConclusionsTurbulent (Spalart-Allmaras) Navier-Stokescomputational results are presented for an advanceddiamond wing semispan model at low speed, high-liftconditions. The numerical results are obtained insupport of a wind-tunnel test that was conducted in theNational Transonic Facility (NTF) at the NASALangley Research Center. The diamond wing modelincorporated a generic fuselage and was mounted on thetunnel sidewall using a non-metric/constant-widthstandoff. The computations are performed at flowconditions of α=10o, M∞ = 0.2, and Rft=7.7x106 thatcorresponded to a nominal flight approach and landingsituation. The overall CFD plan included the numericalsimulation of the NTF empty tunnel flowcharacteristics, semispan high lift model with thestandoff in the tunnel environment, semispan high-liftmodel with the standoff and viscous sidewall in free-air,and semispan high-lift model without the standoff infree-air.At the outset, the numerical analysis demonstrated anapproach to determine an upstream extension (virtualorigin) to the nominal NTF test section for bettersimulation of the empty-tunnel sidewall boundary-layercharacteristics. The lessons learned from the emptytunnel flow simulation were then applied to simulateflow over the complete high-lift model in the tunnel andfree air environment. The numerical predictions showedvery little wall interference effect on the local flowcharacteristics of the model, which exhibited the desiredattached flow over the LEF, main wing, shroud, and theTEF for the most part. The computed wing pressuredistributions are shown to agree well with the measureddata and they both indicate a small effect due to thetunnel wall interference effects. Although themagnitudes of the computed forces and moment were

AIAA-2002-0846


slightly off compared to the measured data, theirincrements due the wall interference effects are predictedwell. Numerical predictions for the combined effects ofthe tunnel sidewall boundary layer and the standoffgeometry are shown to significantly impact the fuselageforebody pressure distribution. As a result, this effect isshown to influence the overall configurationlongitudinal aerodynamic characteristics, particularly thedrag coefficient, for the model with no standoff in freeair.

AcknowledgementThe authors would like to thank Michael Wiese, NormaBean, and Douglas Nark of Analytical Services andMaterial, Inc. for their grid generation effort. Theauthors would also like to thank Chip Bobbitt, Jr. andJoel Everhart of NASA LaRC for providing the NTFempty tunnel boundary layer rake data and theexperimental wall interference effects for the diamondwing test, respectively.

References

1. Takallu, M.A., Laflin, K.R.: Reynolds-AverageNavier-Stokes Simulations of Two Partial-Span FlapWing Experiments. AIAA Paper No. 98-0701,January 1998.

2. Berkman, M.E., Khorrami, M.R., Choudhari, M.,Sadowski, S.S.: Investigation of High-Lift FlowField of an Energy Efficient Transport Wing. AIAAPaper No. 99- 0926, January 1999.

3. Rogers, S.E., Roth, K., Cao, H.V., Slotnick, J.P.,Whitlock, M., Nash, S.M., Baker, M.D.:Computation of Viscous Flow for a Boeing 777Aircraft in Landing Configuration. AIAA Paper No.2000-4221,

4. Mavriplis, D.J., Pirzadeh, S.: Large-Scale ParallelUnstructured Mesh Computations for 3D High-LiftAnalysis. AIAA Paper No. 99-0537, January 1999.

5. Ghaffari, F.: Unstructured Grid Viscous FlowSimulation Over High-Speed Research TechnologyConcept Airplane at High-Lift Conditions.NASA/TP-1999-209718, November 1999.

6. Wahls, R. A.: The National Transonic Facility: AResearch Retrospective (Invited). AIAA Paper No.2001-0754, January 2001.

7. Luckring, J.M., Ghee T.A.: Subsonic ReynoldsNumber Effects on a Diamond Wing Configuration(Invited). AIAA Paper No. 2001-0907, January 2001.

8. Gatlin, G.M., Parker, P.A., and Owens, Jr., L.R.:Development of a Semispan Test Capability at theNational Transonic Facility (invited). AIAA PaperNo. 2001-0759, January 2001.

9. Milholen II, W. E.: A Design Methodology forSemispan Model Mounting Geometries. AIAAPaper No. 98-0758, January 1998.

10. Ghee, T.A. and Taylor, N.J.: Low-speed WindTunnel Tests on a Diamond Wing High-LiftConfiguration. AIAA Paper No. 2000-4504.

11. Krist, S. L., Biedron, R. T., and Rumsey, C. L.:CFL3D User’s Manual (Version 5.0), NASA TM-1998-208444, June 1998.

12. Vatsa, V.N., Thomas, J.L., and Wedan, B.W.:Navier-Stokes Computations of Prolate Spheroidsat Angle of Attack. AIAA Paper No. 87-2627,August 87.

13. Thomas, J.L., Walters, R.W., Taekyu, R.,Ghaffari, F., Weston, R.P., Luckring, J.M.: APatched-Grid Algorithm for ComplexConfigurations Directed Towards the F/A-18Aircraft. AIAA Paper No. 89-0121, January 1989.

14. McMillin, S. N., Thomas, J. L., Murman, E.M.:Navier-Stokes and Euler Solutions for Lee-SideFlows Over Supersonic Delta Wings — Acorrelation With Experiment. NASA TP-3035,1990.

15. Ghaffari, F., Luckring, J. M., Thomas, J. L.: AnOverview of the Navier-Stokes ComputationsAbout the F/A-18 Aircraft With MultiblockStructured Grids. Proceedings from High-Angle-of-Attack Technology, Accomplishments, LessonsLearned, and Future Directions. NASA LangleyResearch Center, Sept. 17-19, 1996. NASA/CP-1998-207676/PT3. pp. 1261 1296, June 1998.

16. Roe, P. L.: Characteristic-Based Schemes for theEuler Equations. Annual Review of FluidMechanics, Volume 18, Milton van Dyke, J. V.Wehausen, and John L.. Lumley, Eds. AnnualReviews Inc., 1986, pp. 337-365.

17. Spalart. P.R., Allmaras, S.R.: A One-EquationTurbulence Model for Aerodynamic Flows. AIAAPaper No. 1992-0439.

18. Biedron, R.T., Thomas, J.L.: A GeneralizedPatched-Grid Algorithm With Application to the F-18 Forebody With Actuated Control Strake.Comput. Sys. Eng., Vol. 1, nos. 2-4, 1990, pp.563-576.

19. Bobbitt, C., Jr., Everhart, J.L., Foster J., Hill, J.,McHatton, R., and Tomek, W.: National TransonicFacility Characterization Status. AIAA Paper No.2000-0293, January 2000.

20. Iyer, V., Everhart, J.L., Bir, P.J., Ulbrich, N.:Implementation of WICS Wall-InterferenceCorrection System at the National TransonicFacility. AIAA Paper No. 2000-2383, June 2000.

AIAA-2002-0846


Figure 1. CFD plan for the NTF diamond wing semispan model.

Figure 2. Diamond wing semispan model description.

Figure 3. Schematic plan-form view of the model.

AIAA-2002-0846


Figure 4. Diamond wing semispan model mounted on the NTF tunnel sidewall and a typical wing high-lift section.

Figure 5. Schematic view of the NTF test section.

AIAA-2002-0846


Figure 6. Computed Navier-Stokes, flat-plate theory and measured boundary-layer thickness-growth along NTF sidewall.

Figure 7. Grid topology for the Model/Stand-off in the nominal NTF test section.

Standoff gridH-H topology

AIAA-2002-0846


Figure 8. Close -up view of the semispan diamond wing numerical model mounted on the NTF tunnel

Figure 9.Solution convergence characteristics for MIT.

AIAA-2002-0846


Figure 10. Turbulent Navier-Stokes flow field simulation for the semispan diamond wing in the extended NTFtunnel test section.

Figure 11. Exterior grid wrapping around the MIT grid for developing the MIF grid.

AIAA-2002-0846


Figure 12. Computed surface pressures and correlation with data.

Figure 13. Computed surface pressure and correlation with data

AIAA-2002-0846


Figure 14. Computed and measured longitudinal aerodynamic characteristics.

AIAA-2002-0846


Figure 15. Computed drag coefficients for configuration various components (blocks).

Figure 16. Computed Cp along the fuselage centerline (Y=0) and two forebody cross sections.

Figure 17. Computed pressure difference between MIF and MNIF along the fuselage center plane (Y=0).

aiaa-2002-0846 numerical viscous flow analysis of an

Documents