[Reprinted from TW: AERONAUTICAL JOURNAL OF THE ROYAL AERONAUTICAL SOCIETY. Aw 1993} An aerodynamic model for flapping-wing flight J.D. DeLaurier Institute of Aerospace Studies, University of Toronto, Downsview, Ontario, Canada PRINTED BY MANOR PARK PRESS LIMITED. EDISON ROAD, HAMPDEN PARK. EASTHOURNE, SUSSEX. ENGLAND. AND PUBLISHED BY THE ROYAL AERONAUTICAL SOCIETY. $ HAMILTON PLACE LONDON WIV BO. ENGLAND.An aerodynamic model for ABSTRACT ‘A design-oriented model for the unsteady aerodynamics of @ fapping wing has been developed using a modified strip theory approach. Within this constraint, vortex-wake effects are account- ced for as well as panial leading edge suction and post stall behaviour. Also, the contributions of sectional mean angle of attack, camber, and friction drag are_ added, which allows this ‘model to be used for the calculation ofthe average lift as well ax the thrust, power required, and propulsive efficiency of a Napping wing in equilibrium fight. An example of such calculations is pre- Sented in the performance prediction of a mechanical flying pterosaur replica NOMENCLATURE AR Wing aspect ratio 2 Semispan length © ‘Acrofoil chord h Plunging displacement of leading edge in flapping direction (lyn Finite-wing Theodorsen function Ci “pendant dette y Eaton) C” Drecocticen Neral ec ofc ob F, Nachos fc deine by Eaton 21) Fe) ample components oC) gen hy Equation) £0 © eae eueny dfn by Eqn) ttn fi Pihing momen Nosema the wags hod tower Rn Reyos umber rte F, eating edge suction fre G Righrace V—Reltve ow vliy tector aeion, given by Eatin) vy awn vl tte esto oon 3° Gente dng npn Manuscript received 7 October 1991, acepted 22 January 1993, Paper No. 1853 flapping-wing flight J.D. DeLaurier Institute of Aerospace Studies, University of Toronto, Downsview, Ontario, Canada & Relative angle of attack at the é-chord location du to the wing’s motion The flow's relative angle of atack at the 2Je-chord location, given by Equation (1) Angle of section’ eri line By Magnitude of the dynamic tvist’s ner variation 88 ——_Dynamically-varying pitch angle, (8-0) © Cheleansle, defined by Equation (33) 1) Propusive efficiency 1, Leading edge suetion efficiency ©) Ptchangle of chord with respect to U 8, Pitchangle of Napping axis with respect o U © Mean pitch ange of chord with respect to Mapping axis "Atmospheric density © Flapping frequen, rads Subsoripts 2 Apparent mass @e Aerodynamic centre © Gireulation f Crossflow fF Fretion in Input out Oxtput Sep Separated flow Superscripts are Mean value Time derivative INTRODUCTION ‘The motivation for this work is based onan interest in mechanical flapping-wing fight. For that reason the analysis is. very design oriented, capable of being readily implemented for the performance prediction of a variety of candidate configurations. Most previous work seems to fall, roughly, into two categories, ‘The first, and most common, is the quasi-steady model where un steady wake effects are ignored. That is, flapping Frequencies are assumed to be slow enough that shed wake effects are negligible. ‘Although such an assumption gives a great simplification to the aerodynamic modelling, this category can still contain a wide ange of sophistication in its detailed approaches, One ofthe sim plest examples is given by Kichemann and von Holst where a rigid elliptical-planform wing is assumed to be performing. span ‘wise uniform motions, whereas Schmeidler®.» presents a much ‘errata sounal A 869 Otamore detailed analysis using lifting-line theory © predict the Performance of a root-lapping wing. One of the most refined Versions of the lifting-line approach is offered by Betteridge and Archer where they use their analysis 10 investigate the possibility of optimised flapping behaviour ‘The quasi-steady approach also includes models of intermediate complexity, where the aerodynamic effects are concentrated at certain representative spanwise points on the wing. Walker) chooses three points along the semispan of a oot-lapping wing, and assumes the motion to be such that the lift and drag are constant values on the downstroke and different constants on the ‘upstroke, Norberg® chooses a single representative point, at 70% fof the semispan, performing sinusoidal motion with constant lift-curve slope coefficients throughout the fapping cycle. ‘The second category accounts for the unsteady aerodynamic effects by modelling the wake in a variety of ways. Among those analyses that include the mean lift required for equilibrium fight (s compared with studies of animal swimming), Philps, Est, and Pratt”) represent the unsteady wake of a root-flapping non twist ing rigid wing with diserete nonplanar vortex elements which include spanwise vortices spaced one per half cycle aft of the ‘quarter-chord bound vortex. A similar model was developed by Blackwell and Archer! for their study of the propulsive charac teristics of a twisting wing, root flapping with constant, but ‘unequal, upsroke and dosenstroke motions ("savvtooth motion”) All of the above analyses assume that the wing is spanwise rigid. That is, the wing’s semispan length (measured along 2 span- wise axis moving with the wing) is assumed to stay constant {hroughout the flapping motion’s full cycle. A remarkable departure from this time-honoured assumption was offered by Rayner when, upon noting that a lifting wing can produce nega tive thrust on the upstroke, he formulated a model assuming the wing to be aerodynamically active only on the dowastroke. Thus, the vortex wake is a series of closed rings. However, Lighthill!® noted how the lift requirement for most bird flight compels some serodynamic activity on the upstroke, and described an extension ‘of Rayner’s model") in which upstroke lift is allowed, but that a span difference between upstroke and downstroke produces the ret thrust. ‘The present analysis does not assume a variable span. Since the ‘motivation was 10 study the feasibility of mechanical fapping- ‘Wing igh, it was fle that an important first step was to see if this \was achievable without having to envision a span variation mech. anism, However, the kinematics do allow for spanwise bending land twisting. Further, this model assumes a continuous sinusoidal motion, with equal times between the upstroke and downstroke. This, along with the high aspect ratio envisioned for the wing, justified the assumption of a modified strip theory where the finite span unsteady-wake effects are accounted for by modified Theodorsen functions. This analysis also differs from previous work in that camber and partial leading edge suction effects are accounted for. Too often, researchers interested in animal fight have chosen, as their starting point, an inviscid flow theory which assumes 1008 leading edge suction. However, the realty is tht the wings of fly ing animals can be highly cambered with litle, if any, leading edge suction. Post stall behaviour is accounted for in this analysis. It may ‘wel be that the variable span model for lapping Right can reais. tically allow totally attached flow. However, flapping wings with the presently assumed kinematics appear to be characterised by significant flow separation over potions of the cycle. Infact, such behaviour may not be undesirable for producing the average lift and thrust required for sustained flight. An example is shown of an efliciently designed mode! pterosaur wing for which this theory predicted, and experiment showed, significant outer panel flow separation, METHOD OF ANALYSIS ‘The Kinematics for each section of the wing ate ilusiated in Figure 1. Upon using the leading edge as a reference point, the section’s motion consists of a plunging velocity, fi, and a pitch angle, 8. Note that fis not necessarily perpendicular to the mean- stream velocity, U. IF the wing is root flapping, as shown in Fig. 2, then f would be perpendicular tothe flapping axis. ‘The wing's aspect ratio is assumed to be large enough thatthe flow over each section is essentially chordwise (in the mean- stream direction). Therefore, the section's circulatory” normal force is piven by oy, Pe pay V is the flows relative velocity at the Yechord. location, and ,(0)=2x{ar-+e4 +8) ® The parameters in Equation (2) ae illustrated in Fig. 1, where itis, seen thatthe angle of the zero lift Line, dy, isa fixed value forthe serofoil, and 8 is the section's mean pitch angle. Further, 8 is given by the sum: 5-5,48, ° where 8, isthe anele of the apping axis with respect tothe tncan-steam velocity, U, and 8, the man angle ofthe chord ith respect othe Rapping axis: Note that if the wing doesnot havea Mapping as (sch for whoe-wing mations). then Os the wings mean pitch ange of Figure 1. Wing section aerodynamic forces and motion variables. Figure 2. Assumed stip theory equivalence to whole wing motion DeLawier Ara! A 1988“The remaining angle in Equation 2), is given by +f ance) e-[SSak where ais he eave angle of atack atthe chord location due {ote wing's maton )) © “ u fico ca+u(o 5 u ‘The coefficient ofa: in Equation (4) accounts forthe wing's finite- span unsteady vortex wake by means of a strip theory model. As illustrated in Fig. 2, cach chordwise strip on the wing is assumed {oct as if it were part of an elliptical planform wing, ofthe same aspect ratio, executing simple harmonic whole wing motions iden- tical to tha ofthe strip’s. For such a wing, Jones! derived that the unsteady normal-force coefficient, BC, is given by ac, RUE ie o Where C(k)jyus iS a modified Theodorsen function for finite AR wings and isthe duced frequency, given by ne = o C18 8 complex function, and it was Found convenient to use Scherer's altemative formulation: ARC) CU sas = Bea) cee ree) o ce aj ® Fk) 86k (+c) o-sar Ba2+AR) om aR {Upon noting that te assumed motion is given by a) IBL+ a=ac™ (10) ‘one obtains, when Equations (7), (9), and (10) are substituted into Equation (4), that [reas + ay 1(U,is due to the micun lift produced by 4, and 8, and it may be calculated in a variety of ways. If one wished to stay’ consistent with the strip theory model assumed for the unsteady aerodynamic terms, then w, /U could be approximat- ced by the downwash for an untwisted ellipical-planform wing, ‘obtained from Kuethe and Chow e «2 2+AR However, ifthe wing has significant spanwise variation of og 8, then one may wish to calculate w, /U by a more accurate method, such as the extended lifing-line theory for twisted wings described in Ref. 14 Returning to Equation (1), note thatthe flow velocity, V, must incude the downwash as well asthe wing's motion relative to U. This is accomplished by including a” along with the kinematic parameters )- sa] « ‘An additional normal force contribution comes from the appar- tent mass effect, which acts at the midchord (see Fig. 1) and is sven by v= tes a ole prc’ 4 indy as ‘where »'3 is the time rate of change of the midchord normal velocity Component due tothe wing's mation: as) ‘Therefore, the section’s total at av = aN, +€N, 06) ‘The section’s circulation distribution likewise generates forces in the chordwise direction, a illustrated in Fig. 1. From DeLaurier', the chordwise force due to camber is given by pov. 2 Gartick's® expression for the leading edge suction of a two dimensional aerofoil may be applied to the present strip theory model by extending it with Equation (4) to obtain +3) 2M ety as) Shed low normal force is Dry =-240(0+8) ody an) qu) 2 ‘The efficiency term, ty. accounts forthe fact that most aerofoils, due to viscous effects, have less than the 100% leading edge suetion predicted by potential-low theory. Viscosity also gives a chordwise friction drag er, nae ,) Buea “ where Vs the flow sped tangential tothe setion, approximated by V, =UCos0-isin(0-8,) 20) and (Cos the drag coeticient due o skin fiction, for which ex- ressions may be found in Hoerner!. Thus, the total chordvise force is AF, = 67, ADae “AD, en ‘An advantage of the strip theory mode! is that it allows for an approximation to localised post stall behaviour. The dynamic stall lange i obtained fom Prout") (ara *8| 4] en and is chosen to apply atthe leading edge. Therefor, the criterion for atached flow over the section is (alan | 68 (2) smn es When the attached flow range is exceeded, totally separated flow is assumed to abruptly occur, for which condition all chordwise forces are negligible: ‘Arnett buna! Aa 1958 OtaDara Dy =0 om tod he oma ace igen by ev=(a¥),, (0), 08) (9p Sn si a et colo eg (ay =(Cay ee 26) ter ve(e4ug)t @ and Vis the midchord normal velocity component due 1 the wing's motion (note that Wy. in Equation (15), i the linearised time derivative of V5) ¥,=hcox(0-B,)+tcd+usino -28) ‘Also, (dN) i due to apparent-mass effects, assumed to be half ‘that ofthe’ attached flow value in Equation (14) (0), Now, the equations for the segment’s instantaneous lift and thrust are 29) aL = dNCos0+ AF, Sind G0, AT = dF, Cos8—ANSin@ on ‘These may be integrated along the span to give the whole wing's instantaneous lift and thrust: iG @) ‘where (tis the section's dibedral angle at that instant in the Napping cycle. ‘The wing's average Iift and thrust are obtained by integrating Lt) and T(0) over the cycle. To do ths, it was found most conve- rent to perform the integration with respect to cycle angle, 4, instead of time, t, where &) so thal the average i and thrust are expressed as Lp 1 f"utone i on G4) TE free ‘One may also obtain the instantaneous power required to move the section against its aerodynamic loads. For attached flow, this isgiven by A 1 an, = AF isin(a-B, +a icogo-0,)+ 4] 6s) vay, [ ted] -am,0-a0,6 ‘where dl isthe section's pitching moment about its aerodynamic cen- tre and dl, includes apparent-camber and apprent-ineria moments: L 1 mc" + pct 36) au, =-{ Loncu + So ‘a> 36) For separated flow, the input power expression becomes ap, coo The instantaneous aerodynamic power absorbed by the whole wing i found from 3) +44] en na=2fa, an and be average input power, throughout the cyl, is given by Fe=[" (aoe cc) Upon noting that the average ouput owe from the wings 2 =Tu 0, the average propulsive effciency maybe clued fom Fw i @ NUMERICAL EXAMPLE Jn the mid 1980s, the National Air and Space Museum in Washington DC contracted the AeroVironment Corporation of Monrovia California to design and construct a giant ying robot pterosaur, This challenging task produced remarkable research in several areas, including computerised stability augmentation and fight contol systems, as well as providing an engineering reinterpretation of paleontological evidence. ‘The author was requested by AeroVironment to apply the flap- ping-wing analysis deseribed above to a performance prediction of the completed 18 ft-span model. The 40 Tp total weight (mainly {due to onboard batteries) plus the drag of the fur-like cover ‘was evidently preventing. sustained lapping flight. Although AcroVironment had done its own aerodynamic predictions, an independent check was desired. The author and his associate, Jeremy Harris of Battelle Memorial Institute, were pleased to perform these calculations for this unusual aircraft. ‘The wing’s planform geomety is illustrated in Fig. 3, where the inflight static twist distribution, Bis designed to be zero. The aerofoil along the span is the Liebeck LPT 110A, for which it ‘was calculated that os Figure 3. Wing plantorm of the ON pterosaur mod Detar Anant! Sour Aon 1288It was assumed, for this example, that negative-o” stalling would not occur. Therefore, (Ayi)pin Wa NOt specified. Also, at the time this work was performed, the dynamic-stall criterion had ‘not yet been incorporated into the analysis. Thus from Equation (22), was effectively zero, ‘The texture of the wing's surface is such as to produce a full chord turbulent boundary layer, so that the frietion drag. was ‘oblained from Ref. 17 as: 0.89) Oy TegtRaye™ «) where Rn is based on the local chord. Further the post stall no mal force coeficent, (Cy. was chosen to be that for a high-AR fiat pate, given by Ref. 20.85 198 Root flapping kinematics are assumed, with no spanwise bend ing. Therefor, the plunging motion is given by h=-{Ty)Cose 44) Where Tis the maximum flapping angle magnitude, given as 03491 rad 20. ‘The dynamic twisting, 80, i ikewise assumed to vary linearly slong the span so that 50=-(Bqy)sing «s) For this study, was chosen as variable. Funher, note that the above equations fix the phasing between plunging and piteh- ing at ~ 90", Diserete clement integrations of Equations (32), (34), (38), and (G9) were performed. For example, by assuming the time integra- tion to be approximated by a summation at discrete cycle increments, ¢; = 1 to m), Equation (34) becomes, 46), « where {refers to station locations along the wing's semispan, y, (= 1 10 n), This means, for the equations presented in the previ- ‘us Section, that dy is replaced by a finite increment Ay, which is ‘ented about y, wth corresponding values fr 6, et. For the example, equal Ay increments of 6 in were chosen giving = 1. ‘Also, m= 20 was selected for te tine intervals The results are shown in Figs 4 and 5 where, forthe given fap- ping frequency of 12 Ha, its seen thatthe wings wil it the Todel’s 40 1b weight i the Rapping-axis angle, 6, equals 75°, the ight speed, U, equals 44 10s, and the dynamic ewist magaic tudes, By a in excess of 22°. Beyond thi, the lift stays fly constant upto the highest, value considered (30°) “The average thrust, however, achieves ils maximum value of 1.2 Tb within a very narow range of By. namely = 225%/ Beyond this, the thrust fll ff rapidly to zero at B, = 28°. This is also seen in the propulsive effciency curve, where a distinct ‘maximum of 42% i reached athe = 2:25 valve t should be mentioned that the flapping wing propulsive eff- ciency, as defined in this ancl, is not diectly comparable to Dropelier efficiencies unless, for a given fight situation, the Wing's drag were subirated fom the propellers thrust. There- fore te example lapping wing is « more efficient tte than the 42% figure would indicate. “The average input power, Py decreases steadily with Bi, which is simply an indication of how the wing requires less work 0 Nap 4 the dynamictwistng magnitudes increase. In fat, i By ine creased much beyond. 30", P,, would become’ negative, Flapping Axis Angle, 6,-7.5 deg Un 46 fs00. Freq =.2 He a SC SEER ERES "Tun ay os Dynamic Twist, Bo, (deg) Figure 4 QN average it and trust performance. indicating a “wingniling” situation where energy is taken from the 44s Mow. Its seen that this occurs quite abruptly beyond the optimum B, contin, Te should be noted tha, for engine sizing purposes, isthe peak value of input power that i required, not the average. For this ample, the peak input power required to overcome the ser0- dynamic reaction forees and moments was aleulated 0 be 500 wats. For the B, = 225%/ft maximum efficiency case, the analysis predicted considerable stalling on the outboard panels during the downstroke. This was duc, in large measure, tothe high value of 6, required for equilibrium fight. Such behaviour was also ‘observed during) flow visualisation experiments by ‘AeroVironment. Finally, although the optimum By = 225%/t dynamic wis distribution appeared to identify condition for sustained fight, tests ofthe QN airraft showed only powered glides. One at may be that the optimum B, requiernent allows vitally no error, Which would be difficult to implement in practice. Additonal. thee may have been significant diferences between the acta and predicted behaviour ofthe acrofoi. For example, fight tests shoved thatthe furlike covering evidently gave more drag than & Smooth skin, This would require expression by an increase inthe (Co), values used in this analysis, hus redvcing thrusting ability Also, when the author inspected the airralt atthe National i and Space Museum, it appeared tha the ireglaity ofthe cover- ing around the leading edge would significantly decease the lead ing edge suction efficiency from it predicted vale of, = 098. It eines Pee 22h x0 ny (¢ 20 € 5% & 10 Bw go °F 05 1 1S 2 25 Draw eo Figure 6.0N average input power and propulsive efficiency. Arras us! Apt 959 Dauneris difficult to see how this could have been avoided, considering the constraints imposed by the requirements for realism in appeat~ tance and kinematics. However, a reduction in 1, likewise reduces the wing's thrusting ability. Considering thatthe margin for sustained flight was fairly narrow to begin with, any of the above reasons could have sufficiently compromised the aircraft's performance, CONCLUSIONS ‘This article has presented a design-oriented method by which one ‘may predict the fight performance of harmonically flapping wings. The major assumptions are that frst the semispan remains constant throughout the motion; and second, a modified strip theory is used to model the aerodynamics. However, general dis- tributions of spanwise twisting and first order bending may be specified. Also, certain important real fluid effects are accounted for, such as post stall behaviour and partial Teading edge suction. ‘These are features which should be included in any accurate flap- ping-wing. analyses, especially when applied to flying animals ‘which usually have sharp edged wings with litle leading edge suction When the analysis was applied to the Project QN mechanical pterosaur, it was found that sustained fight was only possible for 4 vinually singular dynamic twist distribution. The difficulty of ‘mechanically implementing this, along with other considerations stich as the aerofoil’s “as built” us theoretical performance, Appeared to confirm the aircraft's inability to climb or sustain, It should be noted, though, that without the drag of the furlike covering and the weight of the batteries for electric propulsion, the analysis indicates that such a flapping wing aircraft could really achieve sustained flight. That is, the required lift and thrust ‘may be obtained from a harmonically flapping constant semispan ‘wing with localised flow separation. “This should be compared withthe attached flow variable span ‘model discussed inthe Introduction, which appears to be an accu- rate representation for a spectrum of animal flight, especially at low speeds. The present analysis, however, shows that the con- stantsemispan model is also capable of efficient flapping-wing Aight for certain conditions, such as may be experienced by large animals a high speeds, or omithopter aircraft. Tn Fact, for the purposes of achieving mechanical lapping-wing flight, constant semispan motion has been the traditions approach by experimenters. Even if the designers had been aware of any variable span concept, the difficulty of its mechanical implemen- tation (in addition to all the other challenges of omithopter design) ‘may well have discouraged its use. What is seen, however is that ‘ constant semispan wing holds the promise i properly designed and incorporated, of producing the thrust and lift required for successful flapping-wing Right. ‘ACKNOWLEDGEMENTS “This work was supported by an Operating Grant from the Natural Sciences and Engineering Research Council of Canada. The author appreciates the co-operation of the AeroVironment Corporation in making information on the QN Project available for this publication. Further, this work benefited greatly from the careful study and helpful suggestions of .M. Harris, Principal Research Engineer at Battelle Memorial Institue, who also produced Figs 4 and 5 REFERENCES Tr KOciiEMAN,D. and VON Hous, E, Aerodynamics of animal igh, Lafevissen, September 1941, 8,8), pp 277-282, Translated by LJ. Baker forthe Ministry of Aicraft Production, R-T.P. Translation No. 1672 2. SchowrbteR, W. Dynamik des Schwingentuges, Lutflriforschuns, 1935. (4). pp 128-133. 3. Scumiousk, W. Flugzeuge mit Flugelanieb, Luffalvsorschun, 1936, (4), pp 111-117 4. Berrexbcr, DS. and ARCHER, RID. A study ofthe mechanics of flapping wings, Aeronaut Q. May 1974, pp 129-142 5. WALKER, G. The lapping Right of birds. Aeronau Soc, 1927, 3, pp 59094, 6. Noxeeto, UM, Evolution of vertebrate fight: an_ aerodynamic model for he transition from ling to active Might, Am Naturalist Seprember 1985, 126 (3, pp 303-327 7. Pats, DJ. Easr, RA. aid PRATT, NIH. An unsteady fiting-tine theory'of flapping wings with application tothe forward fight of binds J Fluid Mech 1981, 112, pp 97-125. 8. Buackweil, J and AncuER, RID. Peformance Characteristics of Simple Flapping Motion With Application tothe Cruising Fight of Birds, Univerty of New South Wales Repor, IS85/FMT/1 ISSN, 156 3068 9. RAYMER.IM.A vores theory of animal Right Pat 2 The forwant Aigh of bids, J Flud Mech, 1979, 91 (8), pp 731-763 10 Lovin, Si ames, Some challenging new applications for basic ‘mathematical methods inthe mechanics of ude tht were cgially pursued. with aeronautical sims, Aeronaut J. February 1990, 54.032), pp 41-32 11, Ravan, J Vertebrate lapping Bight mechanics and aerodynamics, and the evoltion of flight in tts, Biona-report 8, Gustav Facer, ‘Stuttg 1986 9p 27-74, 12, Jones, RT. The Unsteady Lift of « Wing of Finite Aspet Ratio, NACA Report No. 681, 1940. 13, Sctteken, 10. Experimental and theoretical investigation of large amplitude osilaing fit propulsion systems, Hydronauts, Laurel, (Ma, December 1968, M4, Kuemie, AM. and Cuow, C-Y, The faite wing, Foundations of ‘Aerodynamics, the, Sohn Wiley, New York, 1986, pp145-165 15, DeLabnien, J. Drag of wings with cambered serofois and partial Jeading-oge suction. J Aver, October 1983, 20, (10 pp 882-886 16, Gaeric, LE, Propulsion of « Fspping and Osclating Aerofil [NACA Report No 567, 1936 17, Hoeven, .F. Skin-friction drag, Fluid-Dsnamic Drag, Publishes bythe Author, Brick Town, NI, 1965, pp 21 fo 2-16, 1 Poury, RAW, Airflls for rotor blades, Helicopter Performance, Stability, and Consol, PWS Engineering, Boston, 1986, pp 397-409. 19 Brooks” AN. MacCnrby, PB, LissaMay, PIS. and! MORGAN, WER. Development of a wine flapping fying replica of the largest Prosar, AIAA Paper 85-1446, 1985 20. Hours. S.F_ Pressure drag, Fluid-Dynamic Dra, Published by the ‘Author, Brick Town, NJ, 1965, pp 3-16. Eaivors note: Two further papers on flapping-wing flight by Professor DeLaurier are de for publication in The Aeronautica Jour Datsun Anon! Jour) A 1989