ebook img

NASA Technical Reports Server (NTRS) 19960022281: A Comparison of Computed and Experimental Flowfields of the RAH-66 Helicopter PDF

12 Pages·0.78 MB·English
Save to my drive
Quick download
Download
Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.

Preview NASA Technical Reports Server (NTRS) 19960022281: A Comparison of Computed and Experimental Flowfields of the RAH-66 Helicopter

NASA-CR-200906 A COMPARISON OF COMPUTED AND EXPERIMENTAL FLOWFIELDS OF THE RAH-66 HELICOPTER Final Report NCC2-5137 : C.P. van Dam A.M. Budge Department of Mechanical & Aeronautical Engineering, University of California-Davis, Davis,CA E.EN. Duque US Army Aeroflightdynamics Directorate, ATCOM, AVRDEC, Moffett Field, CA Abstract of numerical simulations. The flowfield around helicopter fuselages have been This paper compares and evaluates numerical and computed using methods of varying resolution, approxi- experimental /towfields of the RAH-66 Comanche helicop- mations and accuracy. Each methodology has an associ- ter The numerical predictions were obtained by solving ated computational requirement and set-up cost/benefit. the Thin-Layer Navier-Stokes equations. The computa- Generally speaking, as approximations to the actual flow tions use actuator disks to investigate the main and tail field increases, the computation and set-up cost decreases. rotor effects upon the fuselage flowfield. The wind tunnel Therefore, relatively low cost methods such as linear panel experiment was performed in the 14x22 foot facility methods have been used extensively in industry in com- located at NASA Langley. A suite offlow conditions, rotor parison to Navier-Stokes methods. thrusts and fuselage-rotor-tail configurations were tested. The RAH-66 (Comanche) helicopter is an important In addition, the tunnel model and the computational geom- vehicle to the U.S. Army. It has been in development for etry were based upon the same CAD definition. Computa- the past several years and will have its first flight some tions were performed for an isolated fuselage time this Fall. To support its development, the U.S. Army configuration and for a rotor on configuration. Compari- Aeroflightdynamics Directorate formed a collaborative sons between the measured and computed surface pres- Computational Fluid Dynamics (CFD) and wind tunnel sures show areas of correlation and some discrepancies. test program. The CFD and wind-tunnel tests were Local areas of poor computational grid-quality and local designed to provide complementary flow field data. Both areas of geometry differences account for the differences. methods use geometry based on the same CAD database. These calculations demonstrate the use of advanced Com- Tunnel test conditions, pressure tap locations and laser putational Fluid Dynamic methodologies towards aflight velocimeter data were all designed to facilitate CFD verifi- vehicle currently underdevelopment. It serves as an impor- cation and to maximize the use of the two types of flow tant verification for future computed results. field information. CFD results were obtained and reported in an earlier work by Duque and Dimanlig[l]. Their work showed the Introduction use of a Navier-Stokes method with overset grids to com- The flowfield around helicopters are so complicated pute the flowfield of an earlier Comanche fuselage design. that few have attempted to perform Navier-Stokes flow- Although they did not show comparison between compu- field simulations of these aircraft. In comparison, the fixed tation and experiment, they did illustrate the capability to wing community has performed numerous full airframe compute the flowfield of complicated helicopter fuselages. configuration computations over the past few years. The The work presented in this paper will present the first helicopter's more complicated flowfield has limited the use comparison of CFD derived results of the RAH-66 (Comanche) helicopter to wind-tunnel data. The paper first gives a description of the experiment, the flow conditions and the two conditions for comparison. The computational method is then explained including the grid system used, the flow solver and some statistics on computational costs. The comparisons between the measured and computed are then presented. These preliminary comparisons show somefavorablceomparisoannsdsomelocalareas of dis- generated using the HYPGEN[4] code and the surface crepancies. The discrepancies between the computation grids were generated using GRIDGEN2D[3], $3D[2], and and measured pressures can be explained by either poor ICEM/CFD. All grid connectivity information was grid quality or differences in geometry. obtained using the PEGASUS[5] code. The current surface geometry is an accurate representa- tion of the Comanche fuselage's CAD definition and Experiment matches the lines of the wind tunnel 15% scale model. Grid modifications required a complete re-griding of the The wind tunnel tests were performed at the 14x22 foot tunnel located at NASA Langley Research Center. The previous CFD model. Figure 3illustrates the modified sur- face geometry showing an overall representation of the tunnel model, Figure 1, consists of a 15% scale model of surface and field grids and detailed figures of the EOSS the Comanche fuselage with a 4 bladed rotor and powered fan-tail. A suite of flow conditions, rotor thrusts and fuse- and the Fan-Tail assembly. The grid system totaled 30 grids with the rotor actuator disk and approximately 3mil- lage-rotor-tail configurations were tested. The model had approximately 200 surface pressure taps distributed along lion points. Figure 3a shows some overall grid features. the fuselage. The taps were configured at station cuts on The grid around the nose of the aircraft (EOSS) the fuselage, along the symmetry line on both the ventral attempts to resolve all the geometric features as shown in (bottom) and dorsal (top) fuselage surfaces. Additional Figure 3b. Previously, the grid did not accurately represent the EOSS geometry. In addition, the current grid was pressure taps were placed inthe vicinity of the fan-tail. For designed to allow the EO$S to rotate to any arbitrary posi- the fan-tail, flowfield velocity measurements were taken tion. This capability required extremely fine transitions along agrid on the inlet side using Laser Velocimetry (LV) grids between the EOSS components and the fuselage can- and Doplar Global Velocimeter (DGV) techniques as shown in Figure 2. Force balance data was collected inde- opy. The EOSS grids were also designed to facilitate straight forward grid changes in the event of design pendently for the fuselage and the rotor system. Two tunnel runs were chosen from the available data changes in that region of the fuselage. The fan-tail grids were modified to improve its repre- and summarized in the following table. The first case is an sentation as shown in Figure 3c. The vertical tail was isolated fuselage with a corresponding freestream Mach added which required a number of specialized collar grids number of 0.13, a Reynolds number of 989,000 based on to resolve the junctions between the top of the fan-tail and overall fuselage length and at a fuselage angle-of-attack of 0.0 degrees. The fuselage angle-of-attack is based on the vertical tail and the junction between the vertical and hori- CAD geometry definition with 5degrees of rotor shaft tilt. zontal tails. In the previous computation, the vertical tail was ignored. A center-body was added to the fan-tail to The second flow condition was a rotor on configuration. account for the blockage effect and more accurately model The freestream was set to an angle-of-attack of 0.737. The the experimental flowfieid with the fan-tail powered. The Mach number was 0.05 corresponding to a Reynolds num- computational centerbody diameter was scaled from pho- ber of 367,896 based on fuselage length. The rotor thrust tographs of the 15% scale model and the width encom- coefficient (CTRotor) was 0.00509. Both conditions had the passes the fantail width in the region. The resulting fan-tail powered. Estimates of the fan-tail thrust were centerbody is a cylinder with a non-dimensional width of obtained from the fuselage force balance data and were 0.0417, diameter of 0.0471, and the edges rounded to 0.25 approximately 18.35 pounds-force. of the cylinder radius. Table 1: Selected Tunnel Flow Conditions Flow Solver Specifics Fan Thrust = 18.5 lbsf The flowfields were computed using the general com- pressible Navier-Stokes flow solver OVERFLOW version Mach Rey # a(deg.) CTrotor 1.6ax by Buning et.al. [6]. This current version of OVER- FLOW has a number of flow solver options including Case 1 0.13 989,000 0 n/a implicit time stepping options of Block-Tridiagonal, Penta-Diagonai, and LU/SGS. The spatial differencing Case 2 0.05 389,000 -.737 0.00509 options include central differencing with either 2nd and _t=0.076 4th smoothing or various upwind methods. This version is also compatible with various computer platforms, has the ability to perform distributed computations using PVM Computational Method and is multitasked for concurrent processing on Crays computers. The flowfield boundary conditions were set using non- Grid System reflecting boundaries in the far field. At the surfaces, no The grid system is an overset grid system as discussed slip was imposed and the normal momentum equations in the earlier work [1]. The body conforming grids were were used to obtain surface pressures. At the main rotor plane and fan-tail face, a pressure jump was applied that averagethsedensityandvelocitieasttwogridplanetshat sured data can be attributed to either local differences in lieadjacentotthedesirepdressujruemplocationH.alfthe the geometry or local regions of poor computational grid pressurjuempiseitheraddedor subtractetodadjacent quality. The major geometric difference is the main rotor gridpointsandtheenergyvariableappropriateclyom- hub. The main rotor hub is in both the isolated fuselage putedF.orthemainrotor,theCTrotor of 0.00509 corre- and the rotor-fuselages tests. The model support is another major difference and should affect the comparisons along sponds to a non-dimensional Ap of 0.00253. The fan-tail the ventral line and downstream of x/L=0.6. The rotating thrust of 18.35 lbsf corresponds to a non-dimensional Ap components of the main rotor and the fan-tail are another of 0.022045. geometric difference between the model and computed The following computations were run on a Cray YMP- geometry. The main and fan tail impose a non-uniform C90 at NASA Ames Research Center's Numerical Aero- pressure jump at their respective rotor disks while the dynamic Simulation facility. The body forces on the indi- computational geometry uses a uniform pressure jump. vidual grids all converged by 3000 iterations. Each There is also a slight geometry difference at the EOSS. In computation required approximately 20 hrs of CPU time the computational geometry, the EOSS grid was designed on the Cray C90. The computations were typically run to allow the EOSS to rotate to an arbitrary position as in with four concurrent processors. On a non-dedicated job the full scale vehicle. This feature includes a small gap queueing system, each case required 12 hours of wall- between the EOSS and the forward part of the fuselage clock time. The flow solver has a typical compute speed of canopy. The tunnel model has a fixed position EOSS approximately 6 to 7 microseconds per grid-point per iter- which blends in with the fuselage. ation. Each run required a maximum of 22 Mwords of in- Keeping these differences in mind, Figure 6 shows a core memory. comparison between the measured and computed surface pressures along the symmetry line. The dorsal (upper) sur- Results aild Comparisons face pressures show a good comparison between the two. The stagnation at the nose of the fuselage and at the engine pylon leading edge are well predicted. The trends along Flow Fields Particle Traces the upper surface of the fuselage from the canopy, through Figure 4 illustrates computed particle traces released the engine pylon and along the tail boom all correlate well along fuselage sectional locations at the EOSS, canopy with the data. At the leading edge of the fan-tail, the corre- and tail boom locations for the isolated fuselage computa- lation between the experiment and computed results devi- tions. Overall, the particle paths show a predominately ate somewhat. This difference can be attributed to a attached flowfield with some local areas of flowfield sepa- difference in the grid line positions sampled and the loca- ration and vortex formations. At the EOSS location, a tion of the pressure ports. flowfield separation area exists just behind the EOSS The ventral (lower) line surface pressure coefficients do region which results in an unsteady asymmetric flowfield. not compare as favorably. As before, the stagnation pres- From the nose to the back of the tail-boom and before the sures at the fuselage nose and leading edge of the gun tail-fan area, the flowfield remains mostly attached. Along mount are well predicted. The trends of the surface pres- the tail-boom engine exhaust port regions, avortex forms, sures up to the gun mount follow the measured data. How- which gets entrained into the fan-tail inlet on the starboard ever, downstream of the engine mount the surface side of the fuselage. As shown in Figure 4b., a vortical pressures do not correlate well with the experiment. flow pattern forms at the exit side of the fan-tail. Downstream of the gun mount the surface pressures are Figure 5 illustrates the corresponding particle traces for rather fiat with only a slight pressure gradient. The experi- the fuselage-rotor flow condition. Over the forward part of mental measurements shows regions of favorable and neg- the fuselage, the flowfieid stays mostly attached to the ative pressure gradients. fuselage. At the tail-boom location, the flow forms a The turbulence model can also be source of discrep- highly vortical flow pattern as highlighted in both the star- ancy between experimental and computed results. The board and port views. Some of the particles released from Baldwin-Lomax [8] turbulence model was used for the the starboard side of the tail-boom are entrained into the above comparisons. This turbulence model may not be fan-tail inlet and exhausted through its exit plane. Other well suited for separated flow fields and also is not well particles released from the tail-boom starboard side flow suited for overset type grids. The Baldwin-Barth [9] one- downwards, mix with particles released from the port side equation turbulence model is a much better suited model of the tail-boom and then form into awake that flows pri- for such flowfields and grid schemes. Figure 7 shows the marily behind the port side of the fuselage. dorsal surface pressures obtained for the isolated fuselage case using Baldwin-Lomax and Baldwin-Barth. There is little difference between the two results. The major differ- Pressure Comparisons ence exists in the region where the EOSS joins the canopy. The pressure comparisons show areas of correlation The remaining results use the Baldwin-Barth turbulence between the computed and measured surface pressure model. coefficients. Differences between the computed and mea- Figure 8 shows surface pressure comparisons at three keyfuselagecrosssectionsE,OSS(STA4x,/L=0.04), comparetodthemeasureddataT.hecanopcyrosssection Canopy(STAll,x/L=0.256a)ndTail-Boom(STA22x,/ showsa muchmorefavorablceomparisobnetweenthe L=0.693)A.ttheEOSSsectiont,heresultsshowsimilar measureadndcomputerdesultsIn. thelowerpartofthe trendsastheexperimeanbtovethecenterlinBe.elowcen- crosssection(z/L<0.03)t,hepressurceomparewsellin terlinet,hecomputatioonverpredictthsepressurAe.tthe termsofbothmagnitudaendtrendG. reatetrhan0.03the canopysectiontsh,epressureasrealloverpredictedT.he pressurefsollowsimilartrendsbut computerdesults pressurefosllowthesametrendsasin theexperiment underpreditchtetest.Thetailboomsectionshowsvery abovez/L=0.05B.utbelowthatline,thepressurdeonot goodcorrelatioTn.hepressubreelowz/L=0.04allmatch. eithefrollowthemeasuretrdendsnordotheyhavesimilar Abovethisheightt,hecomputerdesultsareagainunder- magnitudeAst.thetail-boomsectionth,esurfacperessure predicted. followsin magnitudtehemeasurepdressurvealuesat pressupreortlocationasbovethebottomsurfaceT.owards Summary and Conclusions thebottomsurfactehepressuredseviatgereatlyT.heports atthissectionaregreatlyaffectebdytheupstreammodel In closing, this paper presents a unique comparison post. between Navier-Stokes solutions and wind-tunnel tests for Notethatallthecross-sectipolnotshaveoscillationins helicopter fuselage. Comparisons for an isolated fuselage thesurfacepressurewshenthesurfacegeometrhyasan case and for a powered rotor case revealed areas of corre- abrupcthangeintheslopeT. heseabrupcthangeesxist lation between the computation and experiment. Geomet- throughotuhtefuselagleengthF.orthestationsshownth,e ric differences account for some of the discrepancies EOSSsectionhasabrupcthangeastroughlycenterline between the measured and computed results. In addition, andit exhibitsacharacteristpicressurfeluctuationT.he local regions of poor grid quality also added to the errors. oscillatioinsalsoshownintheCanopsyectioantroughly Overall, this paper gives confidence to the computational z/L=0.03A.llcrosssectionplotsshowthisbehavioarnd method used and provides flowfield information that can canbecontrollebdylocalsmoothinogfthegeometrayt be used to complement wind tunnel data. Further compar- theseabrupgteometrloycations. isons are required to further investigate the fidelity of the Inadditiont,herearesomeslightpressuorescillations solutions. showninthedorsaalndventraslurfacperessurdeistribu- tionsC. loseirnspectioonftheflowfieldsolutionrevealed References eithermismatcheins gridresolutionast oversegtrid boundarieosroversegtridholeboundarietosocloseto highgradienfltowregionsB.othconditionasrecontrolla- Ill Duque, E.P.N, and Dimanlig, A.C.B, "Navier- bleandrequirefurtheardjustmentotsthegridconnectivity Stokes Simulation of the AH-66 (Comanche) Heli- inordetroeliminatteheoscillations. copter", 1994 AHS Aeromechanics Specialists Theresultsfortherotor-fuselacgaeseisshowninFig- Conference, San Francisco, CA, Jan. 1994. ure9andFigureI0.Asintheisolatefduselagceaset,he [2] Luh, R., "Surface Grid Generation for Complex pressurdeistributioncsorrelatweelltotestdatainsome Three-Dimensional Geometries" NASA Technical locationas,ndfailstocorrelatienothersT.hemidlinedor- Memorandum 101046. salsurfacepressuresshowninFigure9 showthecom- [31 Steinbrenner, J.P, Chawner, J.R. and Fouts, C.L., putedpressurefsollowingthe measuretdrends.The "The GRIDGEN 3-D Multiple Block Grid Genera- stagnatiopnressuraetsthefuselagleeadinegdgeandatthe tion System, User's Manual", WRDC-TR-90-3022, leadingedgeoftheenginemounat rewellcapturedA.t boththecanopyregionandaftoftheenginepylon,the Wright Research and Development Center. pressurehsavesimilartrendsbutthecomputepdressures [4] Chan, W.M. and Steger, J.L., Enhancements of a underpredtihctepressurmeagnitudeTsh.edifferenceasft Three-Dimensional Hyperbolic Grid Generation oftheenginpeyloncanbeattributetdothemainrotorhub. Scheme, Applied Math and Comp., 51, 181-205, Asbeforet,hecomputevdentrallinesurfacepressure 1992. coefficiendtsonotcomparweellagainstht eexperiment. [51 Suhs, N.E., and Tramel, R.E., PEGSUS 4.0 Users Thestagnatioantthefuselagneoseandattheleadinegdge _, Arnold Engineering Development Center, ofthegunmounatrebothwellcaptureadlongwithtrends AEDC-TR-91-8, June 1991. inthesurfacperessurfersomthenoseuptothegunmount. [61 Buning, P.,Chan, W., Renze, K., Sondak, K., Chiu, Howeveor,ncepastthegunmountthecomputesdurface I., and Slotnick, J., OVERFLOW Users Manual pressuriserathefrlatwithaslightpressurgeradienTt.he Version 1.6ax, May 1995. measureddataontheotherhandhasrelativellyargevaria- tions. [71 Pulliam, T.H. and Chaussee, D.S., A diagonal Threekeyfuselagecross-sectionparelssurdeistribu- Form of an Implicit Approximate-Factorization tionsatthesamelocationassintheisolatefduselagcease Algorithm, Journal of Computational Physics, Vol- arecompareadndshowninFigure10.TheEOSScross ume 39, Number 2, February 1981. sectionshowasnundeprredictioonfthesurfacperessures 4 [8] Baldwin, B.S. and Lomax H., "Thin Layer Approx- imation and Algebraic Model for Separated Turbu- lent Flows," AIAA-78-0257, AIAA 16th Aerospace Sciences Meeting, Huntsville, AL, Jan. 1978. [9] Baldwin, B.S. and Barth T.J, " A One-Equation Turbulence Transport Model for High Reynolds Number Wall Bounded Flows," NASA TM 102847, Aug. 1999 Figure 1Comanche 15% Scale Model Figure 2Laser Velocimeter At Fan-Tail a) Sample Overall Grids Vertical Tail Gap / Collar Grids \ \,," _\_"(",3, "'_, " /,al Center Body .,# ),_'_/',<h;S " • _XT. e- 257 x_ •C:,".'?v",,:_a:,¢, ;", _'! !N ..... ,5 b) EOSS Detail c) Fan-Tail Detail Figure 3 :Grid Systems Details a) Starboard View b) Port View c) Fan Inlet Figure 4 Flow Field Particle Traces, Run 245, Pt 18 Mach =0.13, Reynolds #= 989,000, ct--0°,,Fan Thrust = 18.5 lbsf., Isolated Fuselage a) Starboard View b) Port View c) Fan Outlet Figure 5 Flow Field Particle Traces, Run 153 Pt 2 Mach = 0.05, Reynolds # = 367,896, ot=-0.737, CTRotor = 0.00509, Fan Thrust = 18.5 lbsf O 3.0- 2.5- 2.0- 1.5 1.0 0.5 I 0.0 -0.5 -1.0 -1.5 a) Dorsal -2.0 0.0 0.2 0.4 0.6 0.8 1.0 3.0- Post Location 2.5 2.0 1.5 1.0 0.5 0.0 " -0.5 -1.0 w • Experiment -1.5 b) Ventral Computation -2.0 0.0 0.2 0.4 0.6 0.8 1.0 x/L Figure 6 Comparison of Dorsal and Ventral Midline Pressure Coefficients, Run 245, Pt 18, Mach = 0.13, Reynolds # = 989,000, _=0°,, Fan Thrust = 18.5 Ibsf., Isolated Fuselage 10

See more

The list of books you might like

Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.