AIAA 2000-4221COMPUTATION OF VISCOUS FLOWFOR A BOEING 777 AIRCRAFT INLANDING CONFIGURATION

Stuart E. Rogers and Karlin RothNASA Ames Research CenterMoffett Field, California

Hoa V. CaoThe Boeing CompanySeattle, Washington

Jeffrey P. Slotnick and Mark WhitlockThe Boeing CompanyLong Beach, California

Steven M. Nash and M. David BakerMCAT, Inc.; NASA Ames Research CenterMoffett Field, California

18th AIAA Applied AerodynamicsConference

14 - 17 August 2000 / Denver, ColoradoFor permission to copy or republish, contact the American Institute of Aeronautics and Astronautics1801 Alexander Bell Drive, Suite 500, Reston, VA 20191–4344

AIAA 2000-4221

COMPUTATION OF VISCOUS FLOW FOR A BOEING 777

AIRCRAFT IN LANDING CONFIGURATION

Stuart E. Rogers� and Karlin Rothy

NASA Ames Research Center

Mo�ett Field, California

Hoa V. Cao z

The Boeing Company

Seattle, Washington

Je�rey P. Slotnickxand Mark Whitlock {

The Boeing Company

Long Beach, California

Steven M. Nashk and M. David Baker k

MCAT, Inc.; NASA Ames Research Center

Mo�ett Field, California

Abstract

A series of Navier-Stokes simulations of a completeBoeing 777-200 aircraft con�gured for landing is ob-tained using a structured overset grid process and theOVERFLOW CFD code. At approach conditions, thecomputed forces for the 777 computation are within1.5% of experimental data for lift, and within 4% fordrag. The computed lift is lower than the experimentat maximum-lift conditions, but shows closer agree-ment at post-stall conditions. The e�ect of sealinga spanwise gap between leading edge elements, andadding a chine onto the nacelle is computed at a highangle of attack. These additions make a signi�cant dif-ference in the ow over the wing near these elements.Detailed comparisons between computed and experi-mental surface pressures are shown. Good agreementis demonstrated at lower angles of attack, including aprediction of separated ow on the outboard ap.

Introduction

Calculating the viscous uid ow over a high-lift

�Aerospace Engineer.yChief, Aerospace Operations Modeling O�ce. Senior Mem-

ber AIAA.zPrincipal Engineer.xPrincipal Engineer. Senior Member AIAA.{Engineer/Scientist Specialist. Member AIAA.kResearch Engineer.Copyright c 2000 by the American Institute of Aeronautics and Astro-

nautics, Inc. No copyright is asserted in the United States under Title 17, U.S.

Code. The U.S. Government has a royalty-free license to exercise all rights un-

der the copyright claimed herein for Governmental Purposes. All other rights

are reserved by the copyright owner.

system of a subsonic commercial aircraft is one ofthe most di�cult problems in Computational FluidDynamics (CFD). Even in two-dimensions (2D), state-of-the-art CFD codes fail to consistently predict, withsu�cient accuracy, trends with Reynolds number ortrends with ap/slat rigging changes.1 High-lift ow-�eld analysis is also a very important problem forcommercial aircraft companies; the payo�s for un-derstanding it and designing a more e�cient high-liftsystem for commercial jet transports are quite high.2

Increases in lift coe�cient (CL) and in lift-over-dragcan lead to a simpler high-lift system, resulting in lessweight and less noise, as well as increases in both pay-load and range.

The di�culties in simulating high-lift ows comefrom the severe complexity of both the geometry andthe ow �eld. The complexity of the ow �eld stemsfrom the wing having multi-elements with very smallgaps between them, leading to an interaction of vari-ous viscous ow phenomena. As stated by Meredith,2

these ow phenomena include boundary-layer transi-tion, shock and boundary-layer interactions, viscous-wake interactions, con uent wakes and boundary lay-ers, and separated ows. Since the uid dynamics isdominated by viscous e�ects, only a high-�delity sim-ulation using the Navier-Stokes equations can providethe accuracy necessary to assist in aircraft design.

Under the Integrated Wing Design (IWD) elementof the NASA Advanced Subsonic Technology (AST)Program, a signi�cant e�ort was focused on develop-ing the CFD software tools required to perform pro-

1American Institute of Aeronautics and Astronautics

duction level CFD analysis of three-dimensional (3D)high-lift systems on complete transport con�gurations.One of the program milestones was to perform a CFDanalysis of an entire high-lift aircraft from CAD topost-processed solution in 50 working days. This mile-stone was met with a simulation of a Boeing 777-200landing con�guration using an overset structured gridapproach, and newly developed scripting software.3

This accomplishment is a reduction by an order ofmagnitude in the CFD process time over what waspossible three years earlier.

In addition to cycle time issues, a number of otherchallenges faced by the AST high-lift CFD team wereput forth in a report.4 These issues account for thefact that the current predictive accuracy of 3D Navier-Stokes methods for high-lift ows could not be readilyassessed: there was a lack of su�cient 3D experimen-tal high-lift data; only a limited number of 3D high-liftsimulations had been conducted, and the available sim-ulations had been done on relatively simple geometries;such simulations required signi�cant computationaland labor resources; and most viscous computationalapproaches were not able to simulate the complex ge-ometries found on a high-lift aircraft con�guration.

The success of the development of new overset CFDtools,3 together with the computational and labor re-sources of the AST program, has removed many ofthese obstacles. The result is the ability to performviscous CFD simulations for a number of complexhigh-lift con�gurations and compare them to exper-imental data, thus providing an accuracy assessmentfor Navier-Stokes applications to high-lift aircraft. Intwo companion papers, results are presented for theapplication of the overset CFD method to the ow overa High-Wing Transport aircraft with externally blown aps,5 and the ow over a three-element trapezoidalwing.6 The current work presents new results for thecomputed ow over a over a Boeing 777-200 aircraftcon�gured for landing. The current work attempts tovalidate the overset CFD approach for high-lift aircraftby comparing the computed results to experimentaldata obtained in the NASA Ames 12-Foot PressureWind Tunnel.

In the following sections, this paper presents the ge-ometry and grids used in the current analysis, presentssome initial results for the lift coe�cient, shows thee�ects of sealing a spanwise gap between two leading-edge components and the e�ect of adding of a nacellechine, and presents detailed pressure coe�cient com-parisons between computed and experimental data forseveral angles of attack.

Geometry and Grids

The computations simulate a 4.2%-scale, full-spanmodel of the Boeing 777-200 aircraft as tested in the

NASA Ames 12-Foot Pressure Wind Tunnel. A pho-tograph of this model is shown in Fig. 1. The ma-jor aircraft components included in the computationaland experimental models are the fuselage, the mainwing, the inboard and outboard leading-edge slats, theKrueger slat, the inboard and outboard aps, the ap-eron, the ow-through engine nacelle and core-cowl,the engine strut, and the vertical tail. Although land-ing gear is shown in Fig. 1, the CFD results werecompared with experimental runs with the landing-gear o�. The experimental model also included a chineon the inboard side of the nacelle. Initially, CAD datafor this component could not be found, and so wasnot included. The de�nition of the chine was obtainedlater and then added to the computational model, asdescribed in a later section. Other details of the modelinclude \steps" along the leading-edge of the mainwing: the side-of-body (SOB) step, a step near thestrut, and an outboard (OB) step near the wing tip.The surface grids for the SOB step and the OB stepare shown in Figs. 2a and 2b, respectively.

Fig. 1. Boeing 777-200 wind-tunnel model.

In order to simplify the individual component gridgeneration, the seals between the aps and aperonare omitted. In the experiment, the aps and ap-eron are partially sealed together. This is done usingwax and/or tape after the components are installed foreach rigging. Thus, the CAD de�nition of the compo-nents do not represent the actual wind-tunnel modelin this regard. As shown in Fig. 3, two small spanwisegaps (about 0.5% of the mean aerodynamic chord) arepresent between the ap elements in the computationalmodel. It was anticipated that these small gaps wouldonly have a minor e�ect on the ow solution. The in-board end of the inboard ap is partially sealed againstthe fuselage in the wind-tunnel model. In contrast, Itwas expected that this seal would have a non-negligiblee�ect on the ow. Therefore, the sealing of the inboard ap against the fuselage was modeled in the computa-tion.

At the leading edge, in the vicinity of the strut,

2American Institute of Aeronautics and Astronautics

Fig. 4 shows a close-up of the surface grids on thewing leading edge, the Krueger slat, and the inboardslat. Small gaps exist at either end of the Krueger forthe CFD model whereas for the wind tunnel model,the spanwise ends of the Krueger were partially sealedagainst the strut and inboard slat with wax and tape.These small gaps allow the grid generation for eachcomponent to proceed independent of the placementof neighboring components. The ends of the ap andslat elements are resolved using wingcap grids that inprevious work7 have been shown to adequately resolvethe geometry and near-body ow-�eld.

Slat

Fuselage

Wing

a. Side-of-body step.

Wing

Slat

b. Outboard step.

Fig. 2. Spanwise steps in the wing leading edge.

Both the leading-edge and trailing-edge brackets areomitted from the CFD model. Trailing-edge ap-bracket fairings are expected to have a larger impacton the ow than the brackets, and so the three largest ap-bracket fairings are included in the computationalsimulation: the outboard fairing on the inboard apand the inboard and outboard fairings on the outboard ap. A closer view of the inboard fairing is shown inFig. 5. The fairings are positioned with and sealedagainst the underside of the wing, but are not con-nected to the ap surface.

Flaperon

InboardFlaps

OutboardFlap

Flap−HingeFairing

Fig. 3. Gaps between aps and aperon.

Pylon InboardSlat

Krueger

Fig. 4. Leading-edge spanwise gaps.

Flaperon

InboardFlaps

OutboardFlap

Flap−HingeFairing

Pylon

Fig. 5. Inboard ap-bracket fairing.

The entire grid-system for the Boeing 777-200 wasgenerated on an SGI Octane workstation, with twoR10000 195 MHz processors, 896MB of memory, and13GB of disk space. The execution of the script sys-tem which runs the entire grid generation process fromthe original surface de�nition to the �nal grid systemrequires �ve hours on this machine. The resulting

3American Institute of Aeronautics and Astronautics

grid system for the Boeing 777-200 aircraft con�guredfor landing consists of 22.4 million grid points within79 overset zones. A view of the surface grids on theentire con�guration is shown in Fig. 6, which plotsonly every fourth grid line in each computational di-rection for clarity. An attempt was made to generategrids that would be adequate for all expected ow fea-tures based on previous high-lift CFD problems, mostof which were simulations of two-dimensional multi-element airfoils. Grid spacing of 10�6 times the meanaerodynamic chord is applied normal to the surface.This results in y+ values on the order of 1.0 for the�rst grid point o� the surface. Also, the maximumgrid-stretching ratio in the normal direction is limitedto 1.25. A total of 5617 orphan points (approximately0.02% of the total points) remained within the gridsystem after the overset process; averaging is used toupdate these points within the ow solver. An orphanpoint is a boundary point requiring interpolated solu-tion data from a neighboring grid, but for which thesoftware cannot �nd a neighbor grid with adequateoverlap.

Fig. 6. Surface grids on Boeing 777-200.

B777 Flow Simulation and Analysis

Flow Solver

The OVERFLOW8;9 Navier-Stokes ow solver wasused in all of the current computations. This codeis written to be e�cient for computing very large-scale CFD problems on a wide range of supercom-puter architectures. On vector supercomputers withvery fast secondary memory devices, the OVERFLOWcode includes an out-of-core memory management op-tion, such that the total memory used is a functionof the largest zone in the grid system, not the totalnumber of grid points. The code is e�ciently vec-torized, and is written to execute simultaneously onmultiple shared-memory processors. For cache-based

multiple-processor machines, the code has been paral-lelized using both a shared memory algorithm, andwith a Message-Passing Interface (MPI) library fornon-shared memory systems. For more details, seethe works by Jespersen10 and Taft.11 Approximatelyhalf of the current cases were run using the standardOVERFLOW, version 1.8b, while the rest of the caseswere run using OVERFLOW-MLP version 1.8k. Theformer cases were run on a 16 processor Cray C90computer, and the MLP version was run on an SGIORIGIN 2000 machine with 256 processors.All of the current OVERFLOW computations uti-

lized the third-order Roe upwind-di�erencing12 option,and the Spalart-Allmaras turbulence model13 with the ow assumed to be fully-turbulent. In the experimen-tal investigation of the 777, no boundary-layer tripswere used, and there was no measurement of transitionlocations. The prediction and modeling of transitionfor complex 3D geometries is beyond the capability ofthe OVERFLOW code. The viscous terms in all threedirections are computed, however the cross-derivativeviscous terms were not included. These were not usedbecause they add about 10% to the cost of the com-putation, and because previous test cases have shownthat their use does not a�ect the solution. The multi-grid option9 to the code was used with three levels.Each OVERFLOW case is run using a local time-stepscale of 0.1, a minimum CFL number cuto� for thelocally varying time-stepping (CFLMIN) of 5.0, anda CFL number multiplier for the turbulence model(CFLT) of 4.0. Low Mach number pre-conditioningis not used because this option caused the code tobecome unstable for the 777 grid system. In thesecomputations, the code was considered converged to asteady-state when the L2 norm of the right-hand sidehad dropped at least 2 or 3 orders of magnitude foreach computational grid, and when the variation inthe total lift coe�cient was less than 0.01% over thelast 100 cycles.

Flow Conditions

The simulation conditions for the current analysiscorresponded to data acquired during wind tunnel Run421 in the NASA Ames 12-Foot PressureWind Tunnel.The model was con�gured for landing as de�ned by theFlaps-30 setting. The ow had a free-stream Machnumber of 0.2, a total pressure of 4.5 atmospheres anda Reynolds number based on the mean aerodynamicchord of 5.8 million. The simulation was conducted infree-air; no wind-tunnel mounting hardware was mod-eled. The experimental data used for the comparisonswas corrected for wind-tunnel wall and blockage inter-ference, but excludes tare and interference correctionsfor the bi-pod mounting device.Five solutions were computed using the grid system

for the initial geometry. These were computed at an-gles of attack (�) of 4, 8, 12, 16, and 20 degrees. The

4American Institute of Aeronautics and Astronautics

cases at � = 4 and 8 degrees were restarted from so-lutions obtained on an initial grid system that did notinclude the ap-bracket fairings. Such restarts saveonly a small percentage of the computing time com-pared to the free-stream initial conditions, even thoughthe change from the restarted solution is small. As apoint of reference, the di�erence in CL at � = 8 de-grees with the addition of the ap fairings was 1.8%lower, and the drag coe�cient (CD) was 1.5% lower.Figure 7 shows the convergence history for CL for thecases computed at 4, 8, and 12 degrees angle of at-tack. Due to the proprietary nature of this data, thevalues cannot be included on the y-axis of this andother subsequent plots. These three cases convergedin an average of 2160 cycles, and required an averageof 194 C90 CPU hours per case. This convergence rateis fairly typical of all runs, however the cases at higherangles of attack usually require more cycles.

4 deg rees

8 deg rees

12 deg rees

C y c l e

0 5 0 0 1 0 0 0 1 5 0 0 2 0 0 0

CL

Fig. 7. Convergence of lift coe�cient.

E x p e r i m e n t

O V E R F L O W

A n g l e o f A t t a c k

5 1 0 1 5 2 0

CL

Fig. 8. Lift coe�cient versus angle of attack.

The lift polars for the �rst �ve runs are plotted inFig. 8. The computed CL di�ers from the experimen-tal lift by less than 2 percent for the lower angles ofattack. However, this �gure shows that the computed ow stalls at a much lower angle of attack than the

experiment. Further investigation of the solution at �= 16 degrees shows that the Krueger slat experiencesseparated ow.

Krueger InboardSlat

a. Close view of ow over Krueger.

b. View of inboard portion of wing.

Fig. 9. Mach contours and particle traces at

� = 16 for initial geometry.

Figure 9 shows a close-up of the ow over theKrueger using both particle traces and Mach-numbercontours. The solid Mach-number contours are drawnin the range of 0.0 to 0.1, and show regions of slower ow. This �gure shows how a vortex is formed by owtraveling upward through the small gap between theKrueger and the inboard slat. This vortex ows overthe top of the Krueger, lifting the ow o� the surface ofthe Krueger, causing the ow to separate. In Fig. 9b,it can be seen that as this passes onto the upper sur-face of the wing, the adverse pressure gradient causesrapid expansion, creating a large stall region on top ofthe wing.

5American Institute of Aeronautics and Astronautics

These results suggest that one source of the discrep-ancy between the computations and the experiment athigh angles of attack is the small spanwise gap in thecomputational model between the Krueger slat and theinboard slat. In the experiment, this gap is partiallysealed using tape and wax; the exact experimental ge-ometry is di�cult to duplicate. In an e�ort to sealthis gap, the two elements were combined into a sin-gle inboard slat element, which creates the maximumamount of sealing possible. This modi�cation requireda slight rounding of the wing leading edge step. Fig-ure 10 shows an image of the modi�ed Krueger andinboard slat.

Chine

Modified Krueger+Slat

Fig. 10. Modi�ed Krueger and inboard slat,

and the chine mounted on the nacelle.

Fig. 11. Mach contours and particle traces at

� = 16 for sealed Krueger and slat.

A second grid system was built with this geometrymodi�cation and a ow case was run at an angle of

attack of 16 degrees. The lift coe�cient for this newrun was only about 3% higher than the previous case,and so the result was not as dramatic as expected.At the time this case was run, a CAD representationof the nacelle chine was �nally located, and so thispiece of the geometry was added to the computationalmodel. Figure 10 shows the chine mounted on theinboard side of the nacelle. A third grid system wasbuilt with the chine added to the second grid system,and another case was run at an angle of attack of 16degrees. This resulted in a CL which was about 5%higher than the �rst grid system calculation at thisangle of attack, which is still signi�cantly lower thanthe experimental lift. A single grid was used to add thechine, which provided resolution for the near wake ofthe chine, but did not provide extra resolution for thevortex as it convects downstream over the wing. Whilethe wing grid has adequate o�-body resolution for theslat wake, the chine vortex may not have adequate gridresolution.

Figures 11 and 12 show plots of Mach-number con-tours and particle traces over the inboard region ofthe wing for these two geometry modi�cations. Fig-ure 11 includes just the sealing of the Krueger andslat spanwise gap, and Fig. 12 shows the ow afteraddition of the chine to this geometry. The sealingof the slats does reduce the amount of low-speed owover the wing aft of the strut, however it also increasesthe amount of separated ow at the wing root. Theaddition of the chine further reduces the amount oflow-speed air aft of the strut and nacelle.

Fig. 12. Mach number and particle traces at

� = 16 with addition of chine.

The third geometry with the sealed slats and thechine was also used to compute the ow at angles ofattack of -5.5, -1.1, 4, 8, 12, and 20 degrees. The liftcoe�cient versus angle of attack is plotted in Fig. 13

6American Institute of Aeronautics and Astronautics

for this geometry, together with the experimental re-sults and the results from the �rst geometry. Figure14 plots CD versus �. The results from the third ge-ometry show even better agreement with experimentallift for the lower angles of attack; at � = 4, 8, and 12,the di�erences in lift are all less than 1.5%, and thedi�erences in drag average 4%.

E x p e r i m e n t

OVERFLOW w / ch i ne and sea led K ruege r

OVERFLOW: in i t i a l resu l t s

A n g l e o f A t t a c k

− 5 0 5 1 0 1 5 2 0

CL

Fig. 13. Lift coe�cient versus �

E x p e r i m e n t

OVERFLOW w / ch i ne and sea led K ruege r

A n g l e o f A t t a c k

− 5 0 5 1 0 1 5 2 0

CD

Fig. 14. Drag coe�cient versus �

Pressure Coe�cient Results

The pressure coe�cient (Cp) data is presented herefor the third geometry. The Cp data is plotted forspanwise locations of 13%, 20%, 30%, 60%, and 79%;these locations are illustrated in Fig. 15.

In the next six �gures, Cp is plotted versus thescaled chord-wise coordinate, x=c, where c is the localchord for the particular element. The x-axes are scaleddi�erently for the slat, wing, and aps so that data forthe smaller elements can be seen. The �rst column inthe plot is the slat, the second is the main wing, andthe remaining columns are the aps. The CFD resultsare plotted with solid lines, and the experimental re-sults are plotted with circles. Within each individual�gure, the Cp scale is the same for all of the spanwisecuts. The computational results are compared withexperimental data from the nearest corrected angle ofattack available. The reason that many of the caseswere computed at angles of attack not correspondingto a corrected experimental � is that none of the exper-imental data was made available to the high-lift CFDteam until after several cases were initially computed.Note that the only experimental data available for theleading-edge devices is on the Krueger at the 30% spanstation.

79%

60%

30%20%

13%

Fig. 15. Spanwise locations of the Cp data.

The Cp data is plotted in Figs. 16{21 for anglesof attack of -5.5, 4, 8, 12, 16, and 20 degrees an-gle of attack. In general, good agreement is seenbetween the experiment and the computations. Oneregion in which there is consistently a di�erence is onthe inboard end of the leading inboard ap, on theupper-surface. Here, the computed suction peak isconsistently higher than the experimental results, pro-ducing more lift in the computations on the leadinginboard ap and the main wing than in the experi-ment. This location is very close to the rear bi-podsupport in the experimental model, which includes avery large hole on the underside of the fuselage nearthis ap. This is a possible cause for this discrepancy.Other than this di�erence, the agreement between theexperiment and computation for � = 4, 8, and 12 isexcellent. In particular, the indication of ow separa-tion on the outboard ap is seen at 60% span in boththe experiment and the computation by a at sectionin the pressure curve over the aft portion of the uppersurface at these angles of attack.

7American Institute of Aeronautics and Astronautics

Figure 20 shows the results at 16 degrees angle of at-tack. The inboard Cp data shows a picture consistentwith Fig. 12: the computed ow is stalled at the wingroot. The experimental Cp data shows a much higherleading-edge suction peak, indicating that it has notstalled at the wing root.Figure 21 shows Cp data at 20 degrees angle of at-

tack. This shows that in both the experiment and thecomputed ow, the wing is stalled at all the inboardsections, but still attached at the outboard sections.The computed upper surface pressures are consistentlyhigher than the experiment, and thus the computed liftis too low.The reason for this large discrepancy at higher an-

gles of attack is not known at this time. Possiblereasons for the early stalling of the CFD model in-clude: a discrepancy between the computational andexperimental geometries; inadequacies in the turbu-lence modeling; transition e�ects; insu�cient grid res-olution in the wing-root region or at the inboard endof the inboard slat; and wind-tunnel e�ects, includingboth wind-tunnel walls and bi-pod mounting e�ects.The possibility of a di�erence in the geometry is anissue, even though it is believed that the computa-tional model of the inboard slat is trimmed at thesame inboard plane as the experimental model, a smalldi�erence in the spanwise extent of this element cangreatly e�ect the ow over the wing root. Given addi-tional time and resources, the �rst thing to try wouldbe to extend the inboard slat spanwise so that it sealsagainst the fuselage, which should maximize the liftgenerated by the inboard wing.

Summary and Conclusions

An overset approach has been used to compute the ow over an entire Boeing 777-200 aircraft con�guredfor landing. The computed results have been com-pared with experimental data acquired in the NASAAmes 12-Foot Pressure Wind Tunnel. Good agree-ment between the two is seen for the lift and dragcoe�cients at lower angles of attack: at approachconditions, the computational lift is within 1.5% ofthe experiment, and the computed drag is within 4%.However, the computational model under-predicts thelift at higher angles of attack, and misses maximumlift by nearly 11%. Several di�erences between theexperimental model and the computational geometryexist. Most of these di�erences involve spanwise gapsbetween high-lift elements which are not present inthe wind-tunnel model. The e�ect of completely seal-ing the gap between the inboard slat and the Kruegerslat was demonstrated, as was the e�ect of adding theinboard chine. Both had a dramatic e�ect on the owover the wing aft of the strut and nacelle, but did notdramatically increase the lift as the ow over the wingroot began to stall.

The current work represents a big improvement inthe ability to perform viscous CFD analysis of high-liftaircraft. The use of the overset-grid approach makes itpossible to develop a grid system for a complete high-lift aircraft in several working weeks, which can thenbe used to study design trades-o�s with only a few daysof work. The computational cost of computing numer-ous conditions, however, is substantial. The accuracyof the current approach is excellent at lower angles ofattack, but the inability to compute maximum lift willlimit the usefulness of viscous CFD analysis as a pro-duction design tool. Further work needs to be doneto understand the reason for the poor agreement atmaximum-lift conditions.

Acknowledgments

The authors would like to acknowledge the technicalhelp of Dr. Pieter Buning, NASA Langley ResearchCenter, as well as the AST/IWD program manage-ment for their dedication and insight. They alsowish to acknowledge the helpful comments from Dr.Thomas Coakley and Dr. Scott Lawrence in their re-view of this work. This work was partially funded bythe Advanced Subsonic Technology Program throughNASA contract NAS2-20268.

References

1 Lynch, F. T., Potter, R. C., and Spaid, F. W.,\Requirements for E�ective High Lift CFD," ICASProceedings, 20th Congress, Sept. 1996.

2 Meredith, P. T., \Viscous Phenomena A�ectingHigh-Lift Systems and Suggestions for Future CFDDevelopment," High-Lift System Aerodynamics,AGARD CP-515, Paper No. 19, Sept. 1993.

3 Rogers, S. E., Roth, K., Nash, S. M., Baker,M. D., Slotnick, J. P., Cao, H. V., and Whitlock,M., \ Advances in Overset CFD Processes Applied toSubsonic High-Lift Aircraft," AIAA Paper 2000-4216,Aug. 2000.

4 Bussoletti, J., Johnson, P., Jones, K., Roth, K.,Slotnick, J. P., Ying, S., and Rogers, S. E., \The Roleof Applied CFD within the AST/IWD Program High-Lift Subelement: Applications and Requirements,"AST/IWD Program Report, June 1996. To be pub-lished as a NASA TM.

5 Slotnick, J. P., An, M. Y., Mysko, S. J., Yeh, D.T., Rogers, S. E., Roth, K., Baker, M. D., and Nash,S. M., \Navier-Stokes Analysis of a High-Wing Trans-port High-Lift Con�guration with Externally BlownFlaps," AIAA Paper 2000-4219, Aug. 2000.

6 Rogers, S. E. and Nash, S. M., \CFD Validationof High-Lift Flows With Signi�cant Wind-Tunnel Ef-fects," AIAA Paper 2000-4218, Aug. 2000.

7 Rogers, S. E., Cao, H. V. and Su, T. Y., \GridGeneration For Complex High-Lift Con�gurations,"AIAA Paper 98-3011, June 1998.

8American Institute of Aeronautics and Astronautics

20% Span

13% Span

30% Span

60% Span

x/c

0 0.5 1

x/c

0 0.5 1

-Cp

-Cp

-Cp

x/c

0 0.25 0.5 0.75 1

79% Span

OVERFLOW, Alpha=-5.6Experiment, Alpha=-5.6

x/c

-Cp

0 0.5 1

-Cp

Fig. 16. Cp data for � = -5.6 deg.

8 Buning, P. G., Jespersen, D. C., Pulliam, T. H.,Chan, W. M., Slotnick, J. P., Krist, S. E., Renze, K. J.,\OVERFLOW User's Manual, Version 1.8b," NASALangley Research Center, Hampton VA, 1998.

20% Span

13% Span

30% Span

60% Span

x/c

0 0.5 1

x/c

0 0.5 1

-Cp

-Cp

-Cp

x/c

0 0.25 0.5 0.75 1

79% Span

OVERFLOW, Alpha=4.0Experiment, Alpha=4.2

x/c

-Cp

0 0.5 1

-Cp

Fig. 17. Cp data for � = 4 deg.

9 Jespersen, D. C., Pulliam, T. H., and Buning, P.G., \Recent Enhancements to OVERFLOW," AIAAPaper 97-0644, Jan. 1997.

9American Institute of Aeronautics and Astronautics

20% Span

13% Span

30% Span

60% Span

x/c

0 0.5 1

x/c

0 0.5 1

-Cp

-Cp

-Cp

x/c

0 0.25 0.5 0.75 1

79% Span

x/c

-Cp

0 0.5 1

-Cp

OVERFLOW, Alpha=8.0Experiment, Alpha=7.5

Fig. 18. Cp data for � = 8 deg.

10 Jespersen, D. C., \Parallelism and OVER-FLOW," NAS Technical Report NAS-98-013, October1998. http://www.nas.nasa.gov/Pubs/TechReports/NASreports/NAS-98-013/

20% Span

13% Span

30% Span

60% Span

x/c

0 0.5 1

x/c

0 0.5 1

-Cp

-Cp

-Cp

x/c

0 0.25 0.5 0.75 1

79% Span

OVERFLOW, Alpha=12.0Experiment, Alpha=11.75

x/c

-Cp

0 0.5 1

-Cp

Fig. 19. Cp data for � = 12 deg.

11 Taft, J., \OVERFLOW Gets Excellent Resultson SGI Origin2000," NAS News, Vol. 3, No. 1, Jan.1998. http://www.nas.nasa.gov/Pubs/NASnews/98/01/over ow.html

10American Institute of Aeronautics and Astronautics

20% Span

13% Span

30% Span

60% Span

x/c

0 0.5 1

x/c

0 0.5 1

-Cp

-Cp

-Cp

x/c

0 0.25 0.5 0.75 1

79% Span

OVERFLOW, Alpha=16.0Experiment, Alpha=15.45

x/c

-Cp

0 0.5 1

-Cp

Fig. 20. Cp data for � = 16 deg.

12 Roe, P. L., \Approximate Riemann Solvers, Pa-rameter Vectors, and Di�erence Schemes," J. Comput.

Phys., Vol. 43, pp. 357{372, 1981.

20% Span

13% Span

30% Span

60% Span

x/c

0 0.5 1

x/c

0 0.5 1

-Cp

-Cp

-Cp

x/c

0 0.25 0.5 0.75 1

79% Span

OVERFLOW, Alpha=20.0Experiment, Alpha=19.9

x/c

-Cp

0 0.5 1

-Cp

Fig. 21. Cp data for � = 20 deg.

13 Spalart, P. R. and Allmaras, S. R., \A One-Equation Turbulence Model for Aerodynamic Flows,"AIAA Paper 92-0439, Jan. 1992.

11American Institute of Aeronautics and Astronautics