Escolar Documentos
Profissional Documentos
Cultura Documentos
Abstract
2
dynamics of flapping wings have examined as well as some preliminary insight into the
either rigid wings or wings with a prescribed flow field using strobed laser sheet flow vi-
motion.6, 7 Some of these studies look at sualization.
ornithoptic or bird-like flapping, i.e. flap-
ping without the pronation and supination
phases of insect-like flapping. Some are re- Experimental Setup
stricted to small disturbances while others
are computationally intensive CFD simu-
lations. DeLaurier8 developed an aerody-
namic model for ornithoptic flapping which
has been applied to the aeroelastic analy-
sis of a large-scale ornithopter.9 Walker10
recently developed a simple analysis that
can predict the translational and rotational
components of the airloads on the Robofly
wings. The development of a comprehensive
theory for unsteady force generation by in-
sect wings is partly hindered by a lack of ex-
perimental data at the chord Reynolds num-
bers of interest (103 − 104 ).
An important feature of insect wings is Figure 2: Flapping wing mechanism (Con-
that they can deform greatly during flight. cept by M.J. Tarascio13 )
Also, unlike birds or bats, insect muscles
stop at the wing base so any active con-
trol of the wing shape is not likely.11, 12 Flapping Wing Mechanism
Passive aeroelastic design is hence very im-
portant for insect wings. The Robofly The flapping wing test apparatus is a
measurements are based on very low fre- passive-pitch, bi-stable mechanism capa-
quencies of motion because the fluid used ble of emulating insect wing kinematics13
had a high viscosity. Thus wing bending (Fig. 2). The desired flapping and pitching
and passive aeroelastic effects are likely to motion is produced by a Hacker B20 31S
be very small in the Robofly experiment. brushless motor, which is controlled by a
Tarascio and Chopra13 presented experi- Phoenix PHX-10 sensorless speed controller
mental results for a flapping wing proto- in combination with a GWS microprocessor
type that operated in air at high flapping precision pulse generator. The motor shaft
frequencies. Recently, the present authors is rigidly attached to a rotating disk, which
measured the thrust generated by insect- in turn is attached to a pin that drives a
like flapping wings mounted on this flap- Scotch yoke. The Scotch yoke houses ball
ping wing prototype.14 In this paper, an ends, which are attached to shafts that are
improvement of the thrust measurement free to flap with the motion of the yoke. As
methodology is described and thrust mea- the shaft is actively flapped, pitch actuators,
surements for a number of wing and stroke which are rigidly attached to the shaft, make
parameters are presented. In addition, a few contact with delrin ball ends at the end of
vacuum chamber test results are presented each half-stroke, causing the shaft to pitch
3
Figure 3: Pitch assembly
4
on the pitch housing. A pitching motion a number of flapping cycles.
of the shaft causes the magnets to move in
relation to the Hall effect sensor, producing Vacuum Chamber
a change in its output. To quantify the inertial forces acting on the
wing, a small vacuum chamber has been de-
signed and built using a 1600 diameter, 1/200
thick acrylic cylinder (Fig. 6). At the two
ends of this cylinder 100 thick acrylic plates
are held by screws. The upper plate is fit-
ted with a valve to connect to a vacuum
pump. In addition, this plate also has a vac-
uum gauge and two electrical feedthroughs
for connecting the motor, pitch sensor and
force sensors. All vacuum chamber tests are
conducted at a gauge pressure of 2700 of Hg,
which corresponds to a 90% vacuum.
Flow Visualization
The flow visualization test stand consists of
a steel frame bolted to the ground, on which
the flapping wing mechanism is mounted ap-
proximately 40 above ground (Fig. 7). Alu-
minum plates extend from ground level to
approximately 30 above the mechanism to
provide an image plane for the single wing.
At the top of the aluminum plates, an alu-
Figure 6: Vacuum Chamber minum honeycomb extends 20 horizontally.
The seed for the flow visualization is pro-
Because strain gauges are used on the load duced by vaporizing a mineral oil into a
cell, only the moment acting at the base of dense fog, which passes through a series of
the wing is measured. To convert this mo- ducts before reaching a diffuser mounted on
ment into an equivalent force, the distance top of the honeycomb. The diffuser reduces
from the wing base at which this force acts the vertical velocity of the fog, while the
must be known. Because the forces on a honeycomb helps to eliminate any swirl or
flapping wing are predominantly inertial in turbulence in the flow.
nature,15,16 the point on the wing at which Flow visualization images are acquired by
these forces act is calculated analytically, strobing the flow with a laser sheet gen-
and is used to determine the forces acting erated by a dual Nd:YAG laser, as shown
on the wing. These forces are then trans- in Fig. 8. This laser is triggered once
formed into vertical and horizontal compo- every flapping cycle by a Hall effect switch
nents using the measured pitch angle. The mounted on the flapping wing mechanism.
mean aerodynamic thrust is calculated by A charge coupled device (CCD) camera is
taking the average of the vertical force over used to capture the images.
5
Analysis
Experiments have shown that the lift and
drag coefficients on flapping wings are
higher because of the leading edge vortex.3
Previous quasi-steady analyses, such as
Ref. 2, did not account for this increased
performance and hence failed to accurately
predict the lift generating capacity of in-
sect wings. However, quasi-steady analyses
can explain the lift produced by an insect
wing if the effects of a leading edge vor-
tex, on the lift and drag coefficients, are ac-
counted for. This has led to a revival of
quasi-steady models in recent years.5 How-
ever, such models cannot account for the
force peaks resulting from wing wake inter-
actions because these effects are unsteady in
nature. A blade element model developed
by Walker10 is used to predict the airloads
Figure 7: Flow Visualization Test Setup
on the flapping wings. In this analysis, the
wing is assumed to be rigid, i.e. the effects
of elastic bending and torsion are ignored.
Z i z1
z
y
Laser Sheet
θ y
1
Wing Flapping
Axis Yi
Camera
ζ
Xi
Laser
x1 x
6
has its origin at the center of rotation. The dm, located at radius r, and at a distance y
flapping angle ζ denotes the rotation of the in front of the pitching axis, are given by,
flapping reference frame x1 y1 z1 about the T
Zi axis as shown. The wing reference frame y cos θ ζ̈ + ζ̇(r ζ̇ − 2y θ̇ sin θ)
xyz is obtained by rotating the flapping ref- Fi = dm 2 2 2
y(θ̇ + ζ̇ cos θ) − r cos θζ̈
erence frame by the wing pitch angle θ, rζ̈ sin θ − y(θ̈ + sin θ cos θζ̇ 2 )
α = tan−1
v (r, t)
n
(3)
Results and Discussion
vx (r, t)
Thrust Measurement
and where vx (r, t) and vn (r, t) are the veloc-
ities parallel and perpendicular to the wing The thrust generated by two aluminum-
chord, respectively. As shown in Ref. 10, mylar wings has been measured for a num-
these velocities are determined at the 3/4 ber of stroke and wing parameters. The
chord location, which was found to give wing planform is based on a scaled-up fruit
good agreement with experimental results fly wing similar to the Robofly3 wings.
These wings, made from 0.0200 thick alu-
for the Robofly wings (for lift due to transla-
tion and rotation). The forces dFn and dFx minum frames, are shown in Fig. 10. In
were transformed to the flapping reference Ref. 14, it was shown that Wings I and
frame through the pitch angle θ to deter- II produce the same amount of thrust but
mine the vertical and horizontal circulatory Wing II can attain higher frequencies on
forces. Non-circulatory forces generated by the flapping wing mechanism because of its
the acceleration of the wing in a direction lower mass. In this paper, results are pre-
perpendicular to the chord were calculated sented for Wings II and III only. All the
and added to the circulatory forces. results are based on a flapping stroke angle
of 80◦ , i.e. the angle ζ varies from −40◦ to
+40◦ . Each wing was tested at two pitch an-
Inertial Forces
gles of 30o and 45o . A pitch angle of 30◦ im-
The forces acting on a flapping wing are pre- plies that the pitch is 30◦ during the down-
dominantly inertial in nature.15,16 The in- stroke and then changes to −30◦ (150◦ ) dur-
ertial forces acting on an infinitesimal mass ing upstroke. Similarly for the 45◦ case, the
7
3
Experiment
(Circular
2.5 o load cell)
Stroke : 80
Pitch : +30o/−30o
Thrust (grams)
2
Experiment
0.5 (Square load cell)
0
3 4 5 6 7 8 9 10 11
Frequency (Hz)
8
section, these spurious strains are mini- tal change of 90◦ . However, when the pitch
mized. Figure 12 shows the thrust measured angle is 30◦ , the total change in pitch is 120◦
using this redesigned load cell. Although as the wing flips from 30◦ to 150◦ . The re-
there still remains a discrepancy between duced flip angle at 45◦ pitch, reduces the
the analysis and experiment, the cylindri- rotational circulation for Wing III. Because
cal load cell shows a quadratic increase in Wing III generates a significant amount of
thrust as the frequency is increased. All re- lift from rotational circulation, the net in-
maining thrust measurements in this paper crease in thrust is very small at 45◦ pitch
have been made using the cylindrical load angle. Also, at low frequencies, the mea-
cell. Although a cylindrical cross-section sured increase in thrust is smaller. This may
provides good results, it suffers from the be caused by a weaker leading edge vortex
drawback that the strain gauges do not have at slower speeds.
a flat surface to bond with. This leads to a
short useful life of these load cells before the 6
strain gauges need to be replaced, thus in-
5 Stroke : 80o
creasing the overall testing time. Pitch : +30o/−30o
Experiment (Wing III)
Figure 13 shows a comparison of the
Thrust (grams)
4
experimental measurements and analytical
3
thrust predictions for the two wings at a
pitch angle of 30◦ . Wing III pitches about 2
the 20% chord location, compared to Wing Analysis (Wing III)
1
II, which pitches about the 50% chord loca- Experiment (Wing II)
9
6 6
o
Experiment (45 )
5 5
Wing III Wing II
o
Stroke : 80o Stroke : 80
Thrust (grams)
Thrust (grams)
0 0
3 4 5 6 7 8 9 10 11 3 4 5 6 7 8 9 10 11
Frequency (Hz) Frequency (Hz)
Figure 15: Effect of wing pitch angle on Figure 17: Effect of early rotation on thrust
thrust (Wing III) (Wing II)
6 6
Pitch: Baseline
5 5
Wing III Wing II
Stroke : 80o 4 Stroke : 80o
Thrust (grams)
Thrust (grams)
0 −1
3 4 5 6 7 8 9 10 11 3 4 5 6 7 8 9 10 11
Frequency (Hz) Frequency (Hz)
Figure 16: Effect of early rotation on thrust Figure 18: Thrust in air and vacuum for
(Wing III) Wing II
Figure 16 shows the effect of a slight in pitch phase to change the lift generated
change in pitch phase on the thrust gener- by their wings. Figure 17 shows that the
ated by Wing III at a pitch angle of 45◦ . thrust reduces when Wing III pitches early.
To change the pitch phase, the ball ends An interesting observation is the nearly lin-
shown in Fig. 2 are moved slightly toward ear variation of thrust with frequency for
each other, thus causing the pitch actuator the case of early pitching. The reduction in
to hit them early, producing an early pitch, lift is unexpected since it has been reported
(i.e. the wing flips over earlier in the flap- elsewhere3 that early pitching may produce
ping cycle as compared to the baseline case). a positive wake capture, i.e., when the wing
For this case, the wing starts pitching 0.04T flips early in the flapping cycle, its interac-
earlier than the pitch starting point for the tion with the wake created during the pre-
baseline case, where T is the time period of vious cycle increases the total thrust. Fig-
one flapping cycle. Insects use such changes ure 17 shows the effect of early pitching on
10
the thrust generated by Wing II. Although a anism. Also, the temporal variation of air-
limited amount of data is available for this loads contains frequencies higher than the
case, the thrust is nearly unchanged when flapping frequency, which may be caused by
compared to the baseline case. the elastic bending and twisting of the wing.
It can be seen from Fig. 19 that during the
translational phases (i.e., the portion of time
Vacuum Chamber Tests
when the pitch angle is nearly constant),
Inertial forces are a large part of the to- there are significant durations of negative
tal forces measured using the load cell. To thrust. For example, for the time period
eliminate these inertial loads from the total 0.2T-0.3T, there is a nearly constant nega-
measured loads, the flapping wings must be tive thrust on the wing, although the wing
tested in a vacuum. Tests conducted in a does produce significant positive thrust from
90% vacuum show that, as expected, Wing time 0.1T to 0.2T.
II generates a very small thrust at a pitch
angle of 30◦ , as shown in Fig. 18. However, Flow Visualization
the measurement error in the vacuum cham-
ber data is larger as compared to the mea- Preliminary flow visualization results are
surement error in air. presented here to show the differences be-
Vacuum chamber tests were used to sub- tween Wing II and Wing III at 45◦ pitch
tract the inertial forces from the total forces angle and the baseline case of pitch phase.
measured in air. When the wing was tested One of the reasons for the high lift generat-
in vacuum, the frequency attained by the ing capability of insects, even at large pitch
mechanism was not the same as the fre- angles, is the presence of an attached leading
quency in air at the same motor supply volt- edge vortex on top of the wing. Figures 20
age. However, to subtract the inertial forces and 21 show such a leading edge vortex on
from the total loads, the test frequencies in Wings II and III, respectively. In these pic-
air and vacuum must match closely. This tures, the wing is close to mid-stroke, the
was done by adjusting the motor supply laser sheet is at mid-span of the wing, and
voltage during the vacuum chamber tests the camera is placed perpendicular to the
to change the frequency. Figure 19 shows laser sheet, as shown in Fig. 8.
the thrust generated in one flapping cycle by Figures 22 and 23 show flow visualiza-
Wing II, in air and in vacuum at a frequency tion images of Wings II and III, respec-
close to 10.7 Hz. The frequency for the vac- tively, at a point in the flapping cycle when
uum test was 10.71 Hz, while the frequency the wing was midway through the prona-
in air was 10.65 Hz. Because these frequen- tion phase. The pitch angle at this point
cies are slightly different, the results were is nearly 90◦ , as can be seen from the im-
plotted against non-dimensional time in the ages. Figures 24 and 25 show images taken
flapping cycle. Figure 19 also shows the air- when the wings were slightly beyond the
loads obtained after subtracting the inertial mid-pronation point. As mentioned ear-
forces from the total forces, and the pitch lier, Wing III generates substantially greater
angle θ measured both in air and in vacuum. lift compared to Wing II because of rota-
It is evident from this figure that the pitch tional circulation. The flow visualization
angle varies slightly in vacuum because of a images show the presence of a strong circu-
change in the dynamics of the drive mech- lation around Wing III during the pronation
11
Airload (grams) Thrust (grams)
200
Mean (Air)
2.4503
0
Mean (Vacuum)
0.42508
−200
0
200
Mean airload
2.0253
0
−200
200
θ (degrees)
Vacuum
100 Air
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Non−dimensional Time (t/T)
Figure 20: Leading edge vortex (Wing II) Figure 21: Leading edge vortex (Wing III)
12
Figure 25: Post mid-pronation (Wing III)
Figure 22: Mid-pronation (Wing II)
phase. For Wing II, two small vortices were
visible on the top surface of the wing. How-
ever, Wing III shows a single strong vortex
at the same point in the flapping cycle.
13
line case. Research Projects Agency TTO Docu-
The inertial forces produced by Wing II ment, 1996.
have been measured by testing it at a vac-
[2] Ellington, C. P., “The Aerodynamics
uum pressure of 2700 of Hg (90% vacuum).
of Hovering Insect Flight,” Philosoph-
These forces were then subtracted from the
ical Transactions of the Royal Society
total measured loads to obtain the aero-
of London Series B , Vol. 305, No. 1122,
dynamic forces on the wings. The resul-
Feb. 1984, pp. 1–181.
tant aerodynamic force showed high fre-
quency oscillations which were suspected to [3] Dickinson, M. H., Lehmann, F., and
be caused by the elastic bending and twist- Sane, S. P., “Wing Rotation and the
ing of the wing. The measured pitching mo- Aerodynamic Basis of Insect Flight,”
tion of the wing showed a small change in Science, Vol. 284, June 1999, pp. 1954–
vacuum as compared to the pitching motion 1960.
measured in air at the same frequency. This
may cause some error in the aerodynamic [4] Liu, H., Ellington, C. P., and Kawachi,
force calculations. K., “A Computational Fluid Dynamic
Preliminary flow visualization tests were Study of Hawkmoth Hovering,” Jour-
conducted on the two wings at a wing pitch nal of Experimental Biology, Vol. 201,
angle of 45◦ . An attached leading edge vor- No. 4, 1998, pp. 461–477.
tex, which enables the wing to generate lift [5] Sane, S. P., “The Aerodynamics of In-
at such a high pitch setting, was observed sect Flight,” Journal of Experimental
on both wings. Qualitatively, the leading Biology, Vol. 206, 2003, pp. 4191–4208.
edge vortex seems to be larger for Wing III.
Also, the circulation around the wing dur- [6] Lan, C. E., “The Unsteady Quasi-
ing pronation is greater for Wing III, which Vortex-Lattice Method with Applica-
generates greater lift from rotational circu- tions to Animal Propulsion,” Journal of
lation. Fluid Mechanics, Vol. 93, No. 4, 1979,
pp. 747–765.
14
Progress in Aeronautics and Astronau-
tics, Chap. 18, AIAA, 2001, pp. 399–
428.
15