IACC Appendage Studies

Transcription

1 - THE ACC Appendage Stdies ELEVENTH CHESAPEAKE SALNG YACHT SYMPOSUM E. N. Tinoco, The Boeing Company, Seattle, Washington, USA A. E. Gentry, The Boeing Company, Seattle, Washington, USA P. Bogataj, The Boeing Company, Seattle, Washington, USA E.G. Sevigny, The Boeing Company, Seattle, Washington, USA B. Chance, Chance & Company, nc., Essex, Connectict, USA ABSTRACT Experimental and comptational stdies of several representative ACC appendage geometries were carried ot to establish baseline data and verify comptational models and methods. Wind tnnel tests of an nheeled, nswept, constant section, rectanglar planform keel monted on a grond plane inclded force and moment measrements, and wake srveys at varios angles of attack. Test configrations (all at constant draft) inclded the addition of ballast blbs and winglets. v w Wl.x W2.x x,y,z Total velocity Pertrbation velocity in x direction Pertrbation velocity in y direction Pertrbation velocity in z direction Short winglet Long winglet Cartesian coordinates angle of attack (leeway) density vorticity Streamfnction Correlations of comptational reslts with experimental wind tnnel data were made. A52/PAN AR potential flow indced drag predictions proved to be in good agreement with the wind tnnel data. Comparisons are also presented between A598 (A52 +bondary layer), wind tnnel reslts and empirical predictions. Again good agreement was shown for cases within the limitations of the bondary layer method. NOMENCLATURE Al Bl B2 b Co Cd co Coo Cop CL Cl CM25c dco e h p Sl s Keel-blb fairing Rond blb Beavertail blb wing or keel span Drag (total) coefficient Section drag coefficient ndced drag coefficient Zero-lift drag coefficient Profile drag coefficient Lift coefficient Section lift coefficient Pitching moment abot 25% chord line of the keel Wind tnnel wall correction to indced drag indced drag efficiency factor vertical distance between wings Pressre Keel distance NTRODUCTON Sailboats operate at the air/water interface, a free srface, permitting the formation of waves de to pressre distrbances - inclding lift - with conseqential loss of energy downstream. The free srf ace is of particlar importance in the ACC (nternational America's Cp Class) where performance and heel are high de to generos sail area, relatively low displacement and stability, and the high speeds developed. deally, fll consideration shold be given to this air/water interface in the development of the wetted portion of an ACC yacht. Free srface comptational modeling codes are beginning to emerge from the developmental stage. However, they are costly to panel and rn and may still be lacking in the reliability and accracy reqired for modelling lifting conditions. Cost effective, reliable, and timely comptational modelling decisions wold have to accept some simplification. Fortnately, most design interest was directed toward the stdy of representative ballast packages (blbs) with and withot winglets. t was reasoned that with the increased draft of the ACC class that the combined flow fields from the hll and blb wold allow the simplification of reflection plane models of keel/ballast packages alone. Reflection plane models are strictly 51

2 valid only at zero Frode nmber (zero speed) or at infinite Frode nmber when a plane of anti-symmetry is sed. This approach is frther reinforced by Reynold's nmber and blockage of large hll models, and cost considerations in available wind tnnels. Attention was therefore directed at the development of CFD (Comptational Flid Dynamics) lift and indced as well as viscos, drag methodologies a~d prediction techniqes. The comptational effort was combined with wind tnnel validation of the effects of tip tank like blbs on a typical ACC keel and the effects of winglets on the keel/blb combinations. Global optimization was not ndertaken de to the free srface and other modelling isses. However, wind tnnel tests confirmed CFD predictions and formed a sefl basis for the development of Stars & Stripes final keel and that on the sccessfl defender America3. The reslts of a joint CFD (Comptational Flid Dynamics) and wind t~nel based stdy on generic appendages sitable for the new class of ACC yachts are presented. This work was sponsored by The Boeing Company in partnership with PACT (Partnership for America's Cp Technology). Compting resorces necessary to spport this work were made available throgh PACT by Cray Research and BM. The stdy focsed on establishing a parametric experimental data base of keel, ballast blbs and winglet configrations, and the ' validation of CFD tools for their design and analysis. A wind tnnel test was condcted of configrations consisting of combinations of a single keel configration, two blb configrations and two winglet configrations, each a~ two longitdinal positions on each blb. Both force and wake srvey data were obtained. The wake srvey data was sed in a Boeing-developed code that provided a mapping of spanwise lift, profile drag and indced drag. The wake and force ' data were available to both PACT and the American syndicates, and as a validation data base for Comptational Flid Dynamics work. GEOMETRY A total of thirteen configrations were wind tnnel tested. These configrations were made p from the following basic parts illstrated in Fi~7e l; one keel, two blbs, one fairi~g, and two sets of winglets. PACT spplied Boeing with the baseline geo~etry for keel Sl, and blb Bl. Boeing then designed the additional parts that ~ere then manfactred by PACT and spplied back to Boeing for wind tnnel testing. KEEL ALONE KEEL+ BLUB Baseline :i1:j:':l~'!.\~ Fmrtng A1 KEEL+ BEAVERTAJL BLUB KEEL+ BULB + WNGLETS Effect~blbJl end blllortng Fiiiing A1 51 B =of~ Wlnglels Figre 1. Wind Tnnel Test Configration Bx 51 W1.x W2.x The keel Sl, has a 4.5 inch maximm span and a inch chord with a NACA 63A9 section airfoil, Figre 2. The tip is finished off with circlar arcs sch that the diameter matches the keel thickness along the tip chord. The blb Bl, is axisymmetric with an airfoil section of a NACA 64A15 scaled p to 16.7% maximm thickness, a length of 5693 inches, and a maximm diameter of 9.41 inches, Figre 3. Blb B2, from the nose to 22.5 inches aft is identical to blb Bl, and has a total length of inches. The aft section is two dimensional in character with a constant width eqal to the maximm diameter, Figre 4. Both blbs had the same volme. Two sets of winglets were made, both having root chords of 6.5 inches, tip chords of 2. inches, Figre 5. Winglet airfoils are symmetric NACA 64A1 sections. No twist was incorporated into either set of winglets. When attached to the blb Bl in the forward position, winglets Wl, have a planview semispan of 17.4 inches, while winglets W2, have a semispan of 25. inches. Both pairs of winglets have anhedral sch that the tip reference chord coincides with the maximm keel span of 4.5 inches. The forward winglet positions, Wx.l, were positioned sch that the trailing edges of the winglets were aligned with the trailing edge of the keel. The aft winglet positions were constrained by strctral monting reqirements bt were as far aft as practical. A wing-body strake fairing Al, was designed to redce the classic vortex which forms at that intersection and thereby redce the associated drag, Figre 6. Bondary layer.trips were sed on the model to stabilize the position of t7bl7nt transition. The rond epoxy trip disks had a height of.1 inches The trip disks were placed on both sid~s of the keel and winglets 1. inch aft of the leading e?ge in plan view, and the blbs had a ring of disks 2. inches aft of the nose in side view. Sblimation type flow visalization was sed to 52

3 verify that the trips were effective. There were no corrections to the force data for the drag of the trips Tnnel nstallation nverted c:::: L~., l 25. ~ Figre 5. Positioning of Winglets on Blbs Bl and B2 W1.1 W1.3 ~~ll ' 4.S ' ' lorow>d Plane ("H D") Figre 2. Configration Sl - Keel Alone ~ '\ 4.5 rond Plane ("Hll1 Figre 3. Configration SlBl - Keel + Rond Blb <El 11T1 jlljj~ Figre 6. Strake Fairing Configration on Either Blb WND TUNNEL TEST DESCRPTON 4.5 Figre 4. Configration SlB2 Beavertail Blb + The wind tnnel facility sed was the University of Washington Aerodynamics Laboratory (UWAL) low speed wind tnnel. This tnnel is a continos, doble retrn, atmospheric type tnnel. The test section and model installation is illstrated in Figre 7. The test section is rectanglar with trianglar fillets in all for corners. The width is 12 feet, the height is 8 feet, and the fillets are 1.5 feet high and wide 53

4 providing a cross sectional airflow area - of 91.5 sqare feet. The six component force balance sed is monted below the tnnel floor. An eight foot sqare grond plane was installed which is centered on the balance with the top of the grond plane inches above the tnnel floor. A trntable three feet in diameter is attached at it's center to the balance, and is centered on the grond plane. This trntable is the strctre to which the inverted model is attached. A force drag tare for the trntable which inclded the effect of temperatre on the skin friction was sbtracted from the measred balance drag. The nominal test conditions were a dynamic pressre of 65 ponds per sqare foot, 1 Mach nmber, and a Reynolds nmber per foot of 1.35 million. Tnnel- Model Splltlar Plate Flow arvey noglon iexu!l1emii!:iiia1li._m:=::55aaa:t::::!::::.::::t::::::nne1 floor.a.dim t Figre 7. Wind Tnnel nstallation The wake srvey analysis procedre reqired three component velocity data and axial total head data in a plane perpendiclar to the reference free stream flow and immediately behind the model. To obtain this flow data a traversing five-hole pitot static cone head probe is sed. The wake srvey mechanism and an installed model are shown in Figre 8. The vertical strts penetrating the ceiling are attached to a motor drive sch that they can be moved to any desired vertical position. The cross bar attached at the bottom of the vertical strts spports a motor driven radial arm to which is attached a fivehole pitot static probe. The arm is programmed to sweep ot an arc, stop while the vertical mechanism increments a step, and then sweep back, etc, etc, all the while probe data is being taken on the fly while the radial arm is sweeping. The coverage area need only be large enogh to captre the velocity deficit and the non-zero vorticity field. t is not necessary to sweep a very large region that has far field total pressre and zero vorticity even thogh the flow is deflected. Figre 8. Wind Tnnel Model nstallation with Wake Srvey Rig TEST RBStJLTS Thirteen configration combinations were tested measring force data and eight of these were selected for wake srvey analysis. Representative reslts from the test follow. force pata Res1t Mltiple rns of most configrations were done to check repeatability. The one sigma drag repeatability is 1 1/4 conts. One drag cont is.1 drag coefficient or.47 sqare feet of drag area. The angle of attack (leeway angle) range is -2 degrees to 12 degrees if balance limits permitted. Both longitdinal and lateral force/moment data were obtained for the thirteen configrations. The trends observed were as expected. That is, zero lift drag varies with wetted area inclding the beavertail blb B2 drag being greater than the axisymmetric Bl by approximately the 3% wetted area increment. ndced drag is increased by the addition of either blb to the keel. The beavertail blb B2, redces indced drag relative to the axisymmetric blb Bl. Winglets frther redce indced drag as expected. The keel lift crve slope for keel alone is increased abot 5% with either blb added, and increased nearly 9% to 12% with the addition of the winglets. The lift center of pressre locations, both longitdinal and lateral were also obtained in the linear region. Smmary comparison plots of the force data for several of the configrations tested are presented to illstrate typical reslts and 54

5 comparative performance. Data presented incldes longitdinal lift vs. angle of attack (leeway) and pitching moment, the classic drag polar, and a profile drag polar where an ideal reference span indced drag has been sbtracted from the total drag level. The profile drag polar format allows increased scale sensitivity for easier comparison of indced and profile drag differences between varios configrations. Figres 9 thorogh 11 illstrate the effects of adding the axisyrnmetric blb Bl to the keel, the strake fairing Al, and then the smaller winglet Wl in the forward position to the blb. The addition of a blb to the keel is detrimental to performance from both a wetted area/profile drag standpoint and indced drag standpoint. Obviosly, the blb mst increase sailing efficiency by redcing the boat's heeling angle if it is to be a benefit. The addition of the smaller winglet Wl, to the blb is seen to be beneficial above lift coefficients of 5 or angles of attack of 3.5 degrees relative to the blb withot the winglet. The addition of the strake fairing indicates p to an eight cont benefit or 3% drag redction for blb Bl. Figre 12 shows the profile drag polar effects of the two blbs with and withot winglets. Withot winglets the beavertail blb, in comparison to the axisyrnmetric blb, redces indced drag significantly, at the cost of slightly increased zero lift drag corresponding to its higher wetted area. Ths, the beavertail blb is better than the axisyrnmetric blb above lift coefficients of.17 or angles of attack of 3 degrees. However, with the addition of the winglets, the axisyrnmetric blb has the best drag performance. c.8.7.!!.6 S1 = d J& Jl4 JJ2.112 ' a- Angle of Attack C1129c - Pitching Moment Figre 9. Comparison of Lift and Pitching Moment Characteristics: Sl, SlBl, SlBlAl, SlBlWl.l.8.7 c D..! i.5 8 = :::; OA <i.3.1 eo- ng Coelflc1en1 ~! \ D36.G D52.D56 Figre 1. Comparison of Drag Polar.8 i.7 ~ = :::J S1 ' Characteristics: Sl, SlBl, SlBlAl, SlBlWl.l S1B1.6.DO D2 2 4 Co~ - ~Jag Coefficient Figre 11. Comparison of Profile Drag o.a i 11.7 "'.6 i Charastics: Sl, SlBl, SlBlWl.l, SlB2Wl.l Ji.m,.,,,.,,,,, /, /.5, / D.1 SB2W1.1 /~~ '/,. ' \ S1B1 Figre 12. Comparison of Profile Drag Characteristics: SlBl, SlB2, SlBlWl.l, SlB2Wl.l 55

6 Wake Sryey pteqratigna A wake srvey is the collection of three component velocity, and pitot static data in a plane immediately behind the model and perpendiclar to the axial centerline of the tnnel. These data are then processed by the wake srvey analysis code to prodce wake mappings of vorticity, profile drag, and indced drag. These mappings can then be integrated to determine the total lift profile drag, and indced drag, or ' partially integrated to obtain spanwise effects. The wake srvey analysis code is gronded on momentm theory. The basic momentm eqation is rearranged to be a fnction of total head deficit, axial velocity deficit, and cross flow velocity terms. The first two terms are essentially profile drag and the third term is indced drag. The profile drag terms are maniplated relative to freestream to indicate zero as soon as the energy level reaches that of freestream. The indced drag crossflow terms are then reworked into fnctions of vorticity and potential stream fnction. These are fnctions of parameters that are non-zero only in the immediate wake region of the flow. The otline of the eqation transformation is indicated in Figre 13. The physical/mathematical backgrond and details are described in the references 1 to 5. The processed data from the wake srvey analysis code can be provided in a variety of formats. One format provides contor maps of profile drag, indced drag, and the angle the free air is deflected by the presence of the configration. Keep in mind that these data are compted from data measred in a plane downstream of the model; ths, these distribtion vales are in this plane also. The conservation of momentm does imply that the integrated vales of lift and drag will not change with downstream movement, only that the distribtion can change with the deflection of the streamlines. Profile drag contors for a winglet configration, SlBlWl.1, are shown in Figre 14. Another format provides spanwise integrations indicating the section lift of the lifting srface and both the indced drag and profile d~ag spanwise distribtions. The integrated keel data for the keel alone, Sl, is shown in Figre 15. For contrast, Figre 16 shows integrated keel data and Figre 17 provides integrated winglet data for a configration with blb and winglet, SlBlWl.l. + < ce :.;Jllilll~f'!<~ : ~ C-;;-ntrol vol~ i S P oo,.,=< =:=. ::i? P oo Viscos wake Poo L J P Wake srvey station j. DRAG Drag pfj(u'- -U')ds+(P.-P)ds {momentm} Jf _.,v:_ - P. )ds + f J,_,(U' - - U' )ds {profile drag} +f J,_,(v' +w')ds {indced drag} ' e U' +~(P,. - P,) U, e.g(u" -U)ds {blockage} U'=U" -U. U, eu.+u, Drag profile J _,,[P,_ -P. +HU" -U)(U" +U-2U,)]ds Drag indced where Hf _.,( V'~ )ds; yt = srramfnction ~ =vorticiry V' solved for by -~ =~+~ V' Oat walls LFT (bondary condition) c(y)c,(y)=-2jf _._~ds C, =~jc(y)cl(y)dy Figre 13. Otline of Wake Srvey Theoretical Basis Figre 14. Profile Drag sobars in Wake SlBlWl.l Looking first at the distribtions from the keel alone in Figre 15, note that the spanwise section lift distribtion is relatively constant. This is becase the keel is not twisted since it mst operate symmetrically (sail on both tacks). Also note that the tip vortex has moved inboard from the physical 4.5 span to nearly 39 inches with the downwind movement to the wake srvey plane. The indced drag is seen to be concentrated in the tip vortex and the profile drag is distribted niformly along the span. For most cases there is an increase in the profile drag at the tip de to the vortex itself. This is probably de becase the vortex sweeps the flow jst inboard of the tip to captre some of the keel profile drag loss. At the "hll" end of the keel G'. 56

7 there is a slight drop off de to not measring the first 1 inches of keel span to prevent possible damage to the probe by contact with the grond plane Keel Lift Distribtion n lle Wllka CLwAKE.41 Ref. Chord a n. j s Wlnglet Span Lift Distribtion n the Wake CLwAKE = -.42 LEFTWNGLET g ~t ~ "--~ 1 Ref. Chord = 4.65 n.4, , 3Q so. i.12 KM 1nc1cec1 Drag Distribtion n the Wake ~ COWAKE,..112 ~.8 l j l ". "E..12 J Keel Proflle Dnlg Dlstrlb11on n lle Wllka. COPw..95 r c ' " S G. 3. CM!!pen. na ao. Figre 15. Keel Span Wake Smmary: i..12.oa e " g 1.D4 i!. o.. Figre SlBlWl.l Keel Lilt Distribtion n the Wake CLwAKE =.427 Ref. Chord= n. Keel ndced Drag Distribtion n the Wake ColwAKE = so. Keel Profile Drag Distribtion n the Wake J< CDPwAKE =.156 S1B1W Q. 3Q. 4Q. 5. Keel Splln, lnchn 16: Keel Span Wake Smmary: sa. 6. Sl!.12.8 f ~.G4 ell ij..12 Oi.8 : g 1l.D4 ell i! Wlnglet ndced Drag Distribtion n the Wake CDlwAKe = Wlnglet Profile Drag D111rltJ11on n the Wake J< CDPwAKE =.156 S181W1.1 lncld km and. blb prollle chg Wlnglet Spon, nches Figre 17. Winglet Span Wake Smmary: SlBlWl.l Figre 16 shows that the addition of the blb and winglets retrns the keel span loading nearly to the geometric span. Ths, the span efficiency is significantly improved, and the indced drag redced. The keel span load is nearly constant along the keel span with significantly less loading across the blb. Figre 17 illstrates the contribtion from the winglets and blb in the winglet spanwise direction. The pper figre shows the span loading of the winglet and blb referenced to the winglet mean geometric chord. The center plot shows the indced drag de to the winglets except between the winglet roots where the keel and blb effects are inclded. Finally the lower plot shows the profile drag de to the winglets inclding the contribtion de to the keel and blb. The integrated reslts for both profile drag and the sm of profile pls indced drag have been compared to the force data polars. Comparisons for three of the configrations are shown in Figres 18 thorogh 2. The wake srvey analysis techniqe did accomplish the fndamental goals of determining lift, indced drag, and profile drag with reasonable accracy. The one sigma variance for lift was 8.5 % and for drag was 7.8%, and the mean when compared to the force data reslts was within 2. %. The significance of this techniqe is the diagnostic capability it provides by being able to generate wake maps and 57

8 spanwise distribtions of lift, indced drag, and profile drag. Ths, it can pinpoint the location of significant flow characteristics both good and bad. c 1..8 ;g.6 Gi 5.4 c = "ij Wake Srvey -Cop Lift Comparison cop+cot a-angle of Attack, deg. Drag Comparison A. - Cop +Cot + dc1 wa enact 1 Force Measrement, Rns 16, 22, Tl. ~ m m ~ ~ ~ Co - Drag Coefficient Figre 18. Wake Smmary - Force Comparison: Sl COMPUTATONAL STUDES Comptational Flid Dynamics (CFD) stdies were condcted as a complementary and concrrent approach to the ACC appendage stdies along with the wind tnnel testing. These stdies focsed on the hydrodynamics of the keel and associated appendages. Analyses inclded modeling of the hll and rdder in some instances. The free srface was for most of the stdies treated as a solid reflection plane simlating a Frode nmber of zero. Capability also existed for sing an anti-reflection plane simlating infinite Frode nmber. The objective of the CFD stdies was not jst to demonstrate an ability to reasonably match experimental data afterthe-fact. That is necessary to establish creditability, however. The objective was to develop a process by which key hydrodynamic performance characteristics of the appendages cold be qickly and reliability calclated. These ingredients are necessary for effective se of CFD in developing new designs for 12 c 1..8 Ci = 8 = (,).6 :J.4 (,) c i.6 Ci (,) 5.4 (,) Wake Srvey -CoP > Lift Comparison S1B1 ForceUUrement, Rna 31, CoP+Cot a - Angle of Attack, deg. Drag Comparison A.-CoP+Cot +dc1 - Force llhanment, Co - Drag Coefficient Rna31, 35 Figre 19. Wake Smmary - Force Comparison: SlBl which experimental data do not yet exist. This reqires not only robst and reliable CFD methods bt also the spporting geometry tools for bilding the comptational models. As a conseqence the focs of these stdies was not on the ctting edge of CFD technology, bt on the more matre technologies that cold be relied pon in the heat of battle (i.e. a design project). The primary comptational tools sed for these stdies inclded a very general three-dimensional panel method, A52/PAN AR. This method was copled to a bondary layer analysis code, A598 for viscos analysis. A proprietary geometry system, AGPS, was sed to define and constrct the comptational models. A description of these tools follows. AS2/PNJ A:R Papal Method A52/PAN AR is a general threedimensional bondary vale problem solver (panel method) for the Prandtl-Glaert eqation, references 6 and 7. This eqation governs incompressible and linear compressible flows in both the sbsonic and spersonic flow regimes. Several important featres distingish A52/PAN AR from other panel methods. These inclde: higher order nmerics, 58

9 c = CD CD J c.!!.o 1..8 ~.6 li J.o WekaSUrvay Lift Comparison S1B1W e-cop -Cop +Cot a-angle of Attack, deg. Drag Comparison.&-Cop+Co1 +dcot wall effect Force Mearament, Rn6,65 1,. Force Measrement, Co - Drag Coefficient Rns6,65 Figre 2. Wake Smmary - Force Comparison: SlBlWl.l These inclde: higher order nmerics, continity of doblet strength, continity of geometry, and very general bondary condition options. A52/PAN AR is a "higher order" panel method, i.e., the singlarity strengths are not constant on each panel. n particlar, the formlation allows a linear sorce and qadratic doblet variation on each panel. Continity of geometry is accomplished by sing trianglar sbpanels. This approach eliminates gaps in the configration srfaces, and allows continos vorticity over the entire configration. The combination of linear sorces, continos qadratic doblets, continos srface geometry, and a very robst fll direct matrix solver reslts in a very reliable method for analysis. As a reslt of these nmerical featres A52/PAN AR is very forgiving of irreglar paneling, a featre that greatly enhances its practical sability for applications involving complex configrations, sch as a hll, keel, rdder and assorted other appendages. For indced drag calclations a farfield Trefftz plane formlation very similar to that sed for the experimental 12 )( wake analysis is sed. The A52/PAN AR code ses the vortex strengths directly rather than differentiating the lifting srface spanwise loading that can lead to inaccracies. Reference 8 indicates that the application of the classical Trefftz plane formlation will yield accrate drag vales only if there are no drag forces on the wake itself. n the physical world this is atomatically taken care of by the wake roll-p. Comptationally, it is a mch more difficlt problem. The rotine calclation of wake roll-p from complex geometries to sfficient accracy for reliable indced drag prediction has not been demonstrated by any method! Fortnately, reference 8, sggests an approach to avoid the problem of drag forces on the wake itself. f the wakes are shed in the freestream direction they may not be force free bt they will be drag free. Any drag calclated in the Trefftz plane will then be that of the configration only, and will not inclde drag from the wake. This alleviates the need for wake contraction in the comptational model and reslts in a very reliable method for indced drag calclations from complicated geometries. This approach works as long as the configration lift distribtion is not ndly sensitive to the presmed wake trajectory. Lifting srfaces can be modeled as either "thick" (external srface paneling) or "thin" (camberline paneling) srfaces withot loss of accracy in calclating drag. The A52/PAN AR code is available throgh the Advanced Aerodynamics Concepts Branch at the NASA Ames Research Center. Analysis of cases with over 1, panels is possible depending on available scratch disk space. Typical cases reqire between 2 to 3 panels depending on configration complexity. Versions of the code have recently been ported to a variety of UNX based workstations inclding the BM R/S 6. A598 Vico Ana1y1i1 Code A598, references 9 and 1, is the A52 inviscid panel code copled with a finite difference bondary layer program. 3-D bondary layer analysis over bodies that lie in the x-z plane of symmetry, and 2-D (with sweep and taper) bondary layer analysis over wing-like srfaces is possible. The application of the strip bondary layer analysis which incldes sweep and taper effects to wing like srfaces is rather straight forward as long as the flow is attached, i.e., no separation. The application of the 3-D bondary layer analysis to bodies is more limited The code is very particlar abot the types of geometries it will analyze. t works best with bodies that have two planes of symmetry, or can be 59

10 analyzed as if they did. This can be applied to canoe hlls and ballast blbs. For most bodies bondary layer theory will not stay attached for their fll length. Sch separation prevents,. calclation of the profile (skin friction + pressre) drag. f the separation is very close to the end of the body, the skin friction calclations are probably valid. The code has been able to calclate attached flow for certain type bodies, inclding beavertail ballast blbs. The A598 code is also available throgh the Advanced Aerodynamics Concepts Branch at the NASA Ames Research Center. AGES Aero Grid And Pnne1jpg System The key to effective CFD applications is the ability to create comptational models in a.timely mann7r. AGPS, Aero Grid and Paneling System, is the key element of this process, references 11 and 12. The se of CFD involves several steps: - Representation of the vehicle shapes within the compter. This is traditionally called "lofting." This may be done within AGPS or the srfaces may be read in from an external compter aided design (CAD) system. Otpt srface definitions in the form of GES files also can be generated for nmerical control milling operations of the yacht components. - Check of the qality of the srfaces. Smoothness of the srface contors is very important from the hydrodynamic standpoint. - The discritization of the srfaces into a form sable by the particlar CFD code. This generates the shape definition to be sed by the CFD code. For the A52 program this is called "paneling" and prodces the srface network of points sch as is shown in Figre Analysis of the CFD geometric definition on the spercompter or the workstation. This prodces files containing the detailed srface pressre distribtions pls lift and drag data. - Retrn of the CFD reslts to the workstation for review and analysis. This sally involves the plotting of the srface pressre distribtions in the form of threedimensional color contor pictres, and two-dimensional airfoil pressre distribtion, spanload, and force plots. Keel - Blb with Straight Winglets Paneling with wakes PANELNG WTH WAKES MX1 Upright Figre 21. Paneling With Wakes n the spport of the work described in this paper all of the above steps were accomplished by the AGPS program, except for the 2-D plots and, of corse, for the CFD soltion itself. AGPS has been in se at Boeing for over ten years. n addition to being a srface lofting system AGPS is a geometry programming langage. t is the geometry programming aspects of AGPS that make it sch a powerfl tool for solving new CFD problems. Becase the programming can be done in a niqe interactive manner, procedres for generating complex srface lofts, and the sbseqent paneling procedres for se in CFD codes can be accomplished in a fraction of the time that it wold normally take sing a langage sch as FORTRAN. Once the paneling procedres have been written the paneling of complete complex configrations is accomplished in a few mintes on the workstation. The reslt is a file that is ready to sbmit to the CFD analysis code. These paneling procedres atomatically handle all of the srface-srface intersections and the preparation of the srface and wake networks with all of the abtments matching exactly. The paneling procedres developed for the PACT stdies inclded complete yacht configrations consisting of hll, keel, blb, winglet, rdder, and trailing wake systems. The paneling cold be accomplished atomatically for both pright and heeled sailing conditions. 6

11 COMPUTATONAL RESULTS The initial focs of the comptational stdies was aimed at the se of A52/PANAR in assessing its applicability to yachting type of geometries, and to evalating its ability to calclate indced drag. Procedres were developed in AGPS to qickly allow the generation of panel models for inpt into A52. Paneling models, illstrated in Figre 21, of complete hll-keelrdder and isolated appendages were spported. Stdies to evalate the accracy of the Trefftz plane indced drag formlation in A52 were also ndertaken. Figre 22 shows a comparison of A52 indced drag reslts in terms of "e", the span efficiency factor, with classical theory for a bi-plane geometry in which both gap and pper and lower wing span are varied. The wings were modeled as simple rectanglar thin srfaces in A52. Excellent agreement with theory is seen for the case where the pper and lower wing span are the same. The agreement deteriorates for wings of neqal span. One sorce of error is that the spanload distribtion in the A52 soltion was not constrained to be elliptic as assmed in the theoretical soltion. n the Trefftz plane the bi-plane arrangement is very similar to that of a keel and rdder or even a tandem keel. A52 stdies were sed to determine the optimm rdder and flap settings for minimm indced drag on a hll with keel and rdder at varios heel angles. Similar type of calclations have also been carried ot for an aircraft configration and shown to be in good agreement with wind tnnel test data. 1. r /e '"'-5:. CLASSCAL '',.. / LFTNG LNE,'~ :::- ~. THEORY '',, ' ' m. '". "''-.6 _ ii ' ',.T. ' _ ~9.8',. ' '. ' o~.,.., hlb, OA o~ Figre 22.ndced Drag Bi-Plane Gap Stdy Another check of the Trefftz plane drag calclation was to compare reslts with those obtained by integrating pressre over the paneled srface of the A52 soltion with wake in the freestream direction: Leeway CL Co Co Effective ntegrated Pressre Trefftz plane Span Ratio 6 degrees o degrees..5. Figre 23. Hll +Keel ndced Drag Test Case configration. A comparison of reslts for a simple hll-keel configration are shown in Figre 23. Good agreement is seen between the two methods. ntegrated pressre calclations reqire reasonable panel density for accracy and are not valid for thin srf aces nless the leading edge sction singlarity is properly evalated. The acqisition of wind tnnel data on the varios appendage configrations tested allowed a more throgh evalation of the predictive capabilities of A52. Figres 24 to 27 show comparisons of effective draft, keel lift crve slope, and chordwise and spanwise locations of center of pressre. The comparisons are for six configrations that inclded: keel alone, keel with axisymmetric and beavertail blb, and both sizes of winglets. Most of the comptational reslts had been generated prior to the wind tnnel test. The agreement with experimental data is excellent in terms of predicting trends and increments between configrations, and very good on predicting performance levels. As is typical of an inviscid method, predicted levels tended to be greater than measred by experiment. t: o( a: Q 4 : w 3 > i3 25 w ~ ~ w 3 CONRGURATON Figre 24. A52 Comparisons with Wind Tnnel Data - - Effective Draft 61

12 .11.7 ~ ~ d ~ ~.3 ". i:.1 S S181W1.1 S181W2.1. S182 S182W1.1 CONFGURATON Figre 25. A52 Comparisons with Wind Tnnel Data - Lift Crve Slope.18!!.1' i. 8 r.1. E i.. cf'.mt l11t1w1.1 S181W:l.1.,. 9192W1.1 CONFGURATON Figre 28. A52 Comparisons with Wind Tnnel Data - Zero Lift Drag S1 S181 S181W1.1 S181W2.1 S182 S182W1.1 CONFGURATON Figre 26. A52 Comparisons with Wind Tnnel Data - Keel Center of Pressre S S181W1.1 S181W S1"12W1.1 CONFGURATONS Figre 27. A52 Comparisons with Wind Tnnel Data - Lateral Center of Pressre Semi-empirical "handbook" methods were sed to estimate zero lift drag for the six configrations The comparison with with experimental data are shown in Figre 28. n general, the trends are well predicted, althogh the estimated drag levels are somewhat higher than the experimental measrements. A598 was also sed to analyze the viscos characteristics of the wing-like srfaces of the keel and winglets tested at the University of Washington wind tnnel. Attempts to se the code to analyze the viscos characteristics of the two ballast blbs tested was not sccessfl. "Handbook" estimates were instead made for the two blbs. Reslts of these analyses were sed to bild drag estimates of the configrations. The estimated drags were then compared to force data from the wind tnnel test. The drag bild-p consists of the profile drag of the keel and winglets (if present) from the A598 analysis, a "handbook" estimate of the blb profile drag, and the indced drags calclated by A52. The indced drags and the profile drags from A598 vary as a fnction of leeway (angle-of-attack). The "handbook" estimate for the blb is invariant with angle of attack. A comparison of the total drag bild-p with experimental force data for the keel alone is shown in Figre 29. The agreement is very good. Drag bildps for three other configrations are compared with their respective force data on Figres 3 to 32. For these configrations the agreement is not as good as it was for the keel alone. The bild-ps all indicate too high a drag level a zero lift. Shifting the drag bild-ps to match at zero lift improves the agreement with the force data. We believe the discrepancy is mainly de to the inaccracy of the constant "handbook" estimate for the blb drag. mproving or ability to predict the installed drag of the large ballast blbs will contine to be an area of focs. t is nlikely that bondary layer methods will be of significant vale in this endeavor. We may have to depend on Navier-Stokes methods to deal with these flows. The information available from the wake srveys shold help to develop 62

13 and calibrate CFD methods for these applications. TOTAL DRAG CONFGURATON: S1 B1 W2.1.7 TOTAL DRAG CONFGURATON: S1 ////' - Cl WAKE SURVEY '"FORCE DATA Cl '.1 FORCE DATA 'x,/,l if + /:/' /~-::.-A598 KEEL ~ W NG LETS COP / + A52 CD BULB DRAG ESTMATE ii SHFTED TO MATCH COPO f ~A598+A52+BULB EST. / // 'A59B KEEL COP + A52 CD CD.4.8.o.1 Cl.1.12.Cle.o 4.o a.o 2.o co Figre 29. Drag Polar Bild-Up: Sl FORCE DATA TOTAL DRAG CONFGURATON: S1 B1 -f // /;$-" '/.?:;-,,,,,.~ ~,,,,,.;; "' ::~,rj /, 598 KEEL CDP + A52 CD BULB DRAG EST MA TE /~l''--.a598+a52+bulb EST. ~/ SHF'TED TO MATCH COPO.8.o.11 2.O 4.O.O 2.O CD Figre 3. Drag Polar Bild-Up: SlBl Cl.1 TOTAL DRAG CONFGURATON: S1 B1 W1. 1 FORCE DATA -~--.12 co " / '\\, ~:,,,' //./ /~-/ A598 KEEL ~ WNCLETS CDP // + A52 COi o! + BULB DRAG EST MATE A598+A52+BULB EST. SHFTED TO MATCH COPO.o 4.o Figre 31. Drag Polar Bild-Up: SlBlWl.1 Figre 32. Drag Polar Bild-Up: SlBlW2.l CONCLUDNG REMARKS These stdies proved sefl in advancing the state of the art in ACC appendage design and analysis. Late in the America's Cp Defender elimination ronds these methods were sed to design a new keel and associated appendages for one of the Defense syndicates. On-thewater racing demonstrated a dramatic performance improvement. The existence of the experiment data base and the development of the analysis processes allowed the design work, which involved a series of seqential CFO analyses and geometric refinements, to be accomplished in considerably less than one week! A52/PAN AR has proven to be a general and robst panel method that is applicable to the analysis of a variety of aerospace vehicles as well as ACC yachts. The indced drag calclations procedres within the method are well sited for appendage analysis. The bondary layer analysis capability of A598 was applicable for analysis of winglike components sch as keel and winglets, bt was not always sable on ballast blb geometries. Reliable viscos analysis methods on threedimensional bodies sch as blbs and hlls are definitely needed and reqire frther development. AGPS was invalable in making possible the speedy generation of comptational models. This was essential dring the very short-fsed propriety design work that was ndertaken late in the America's Cp campaign. ACKNOWLBDGMENTS This work wold not have been possible withot the spport of The Boeing Company in partnership with PACT (Partnership for America's Cp Technology), and the compting resorces provided by PACT sponsors; Cray Research, nc. and BM. Cray Research provided spercompter access to the Boeing Compter Services CRAY Y-MP 63

14 spercompter. BM provided an R/S 6 workstation to spport se of AGPS and some of the CFD comptations. nsight gained at the reglarly held review meetings with other PACT sponsored researchers was also very beneficial. n addition to the athors listed in this paper, William Herling and Winfried Feifel of The Boeing Company were also involved in related facets of this work not reported in this paper. REFERENCES 1. Maskell, E. C. "Progress Towards a Method for the Measrement of the Components of the Drag of a Wing of Finite Span," R.A.E. Technical Report 72232, December 14, McLean, J. D. and Randall, J. L., "Compter Program to Calclate Three Dimensional Bondary Layer Flows over Wings with Wall Mass Transfer," NASA CR- 3123, Snepp, D. K. and Pomeroy, R. C., "A Geometry System for Aerodynamic Design," AAA , September Gentry, A. E., "Reqirement for a Geometry Programming Langage for CFD Applications," NASA CP-3143, Hackett, J. E., W, J. C., "Drag Determination and Analysis from Three Dimensional Wake Measrements," 13th Congress of nternational Concil of the Aeronatical Sciences, Seattle, Agst W, J. C., Hackett, J. E., and Lilley, D. E., "A Generalized Wake ntegral Approach for Drag Determination in Three-Dimensional Flows," AAA Paper , Janary Brne, G. W., "Qalitative Three Dimensional Low-Speed Wake Srveys," Fifth Symposim on Nmerical and Physical Aspects of Aerodynamic Flows, California State University, Long Beach, California, Janary Brne, G. W. and Bogataj, P. W., "ndced Drag of a Simple Wing from Wake Measrements," SAE 91934, October Magns, A. E. and Epton, M. A., "PAN AR - A Compter Program for Predicting Sbsonic or Spersonic Linear Potential Flows Abot Arbitrary Configrations Using A Higher Order Panel Method," Vol. 1. Theory Docment (Version 1.), NASA CR-3251, Saaris, G. R., et al, "A52 User's Manal - PAN AR Technology Program for Solving Problems of Potential Flow Abot Arbitrary Configrations," Boeing Docment D6-5473, 1989, (also available throgh NASA Ames Research Center, Advanced Aerodynamics Concepts Branch) 8. Kroo,., and Smith, S., "The Comptation of ndced Drag with Non- P lanar and Deformed Wakes," SAE 91933, October Sllivan, P. P., et. al., "User's Manal for the Viscos Panel Analysis System - Program A598," Boeing Docment , 1989, (also available throgh.nasa Ames Research Center, Advanced Aerodynamics Concepts Branch) 64