Journal of Bionic Engineering 7 (2010) 102–111

Ornithopter Flight Simulation Based on Flexible Multi-Body Dynamics

Andreas T. Pfeiffer1,2, Jun-Seong Lee1, Jae-Hung Han1, Horst Baier2

1. Department of Aerospace Engineering, KAIST, Gwahangno 335, Yuseong-gu, Daejeon 305-701, Republic of Korea
2. The Institute of Lightweight Structures, Aerospace Department, Technical University Munich,
Boltzmannstr. 15, 85747 Munich, Germany

Abstract
This paper introduces a flight simulation of an ornithopter (flapping-wing air vehicle) based on the flexible multi-body
dynamics, a refined flapping-wing aerodynamic model and the fluid-structure interaction approach. A simulated ornithopter was
modeled using the multi-body dynamics software, MSC.ADAMS, where the flexible parts can be included by importing a finite
element model built in the finite element analysis software, ANSYS. To model the complex aerodynamics of flapping-wing, an
improved version of modified strip theory was chosen. The proposed integrative simulation framework of ornithopter was
validated by the wind tunnel test data reported in the literature. A magpie-sized model ornithopter was numerically designed and
simulated to have the longitudinal trim flight condition. We observed a limit-cycle-oscillation of flight state variables, such as
pitch attitude, altitude, flight speed, during the trimmed flight of the model ornithopter. Under the trimmed condition of free
flight of the model ornithopter, we fixed all the degrees of freedom at the center of gravity to measure the constraint forces and
moment. The concept of the “zero moment point” is introduced to explain the physics of ornithopter trimmed longitudinal flight.
Keywords: ornithopter, flexible multi-body dynamics, flight stability, trim flight
Copyright © 2010, Jilin University. Published by Elsevier Limited and Science Press. All rights reserved.
doi: 10.1016/S1672-6529(09)60189-X

When dealing with a conventional aircraft, we start the

1 Introduction design with simple assumptions, such as a cruising flight
Birds, bats and insects are the most efficient and or a trimmed flight condition, which requires an aircraft
outstanding flying objects, resulting from over 150 mil- to have the forces and moments equilibrium with respect
lion years of evolutionary process. In nature flyers, the to the center of gravity without any continuous control
whole flight systems such as wing structure, wing efforts. However, in case of ornithopters, there is no such
kinematics, and flight controller are well-optimized and equilibrium point due to the continuous flapping-wing
harmonized. An ornithopter is the engineering realiza- motion during the cruising flight. There have been nu-
tion of such birds, bats and insects, to imitate the flap- merous attempts to design and develop various dimen-
ping-wing flight. The ornithopter seemed to be forgotten sions and types of flapping-wing air vehicles[3–5].
with the great success of fixed-wing aviations, however However, most of them were based on intuitive design or
recently we have seen the rebirth of ornithopters as the trial-and-error approach to mimic certain types of
next generation of Unmanned Air Vehicles (UAVs), or creatures[6]. The main difficulty of ornithopter design
bio-inspired flapping UAVs[1,2]. Equipped with an elec- comes from the fact that there is almost no knowledge
tro-optical camera or chemical detection sensors, they about what the most important design parameters are
could be used for typical UAV missions, such as sur- and how they affect each other to flapping-wing flight
veillance and reconnaissance tasks, with quite perfect dynamics. In order to develop a more systematic design
camouflage due to their inherent nature-like shapes. process, and to improve the vehicle performances, such
The preliminary design of ornithopters is quite as the payload capability and endurance, it is necessary
different from that of fixed- and rotary-wing aircrafts. to establish a realistic simulation model with high

Corresponding author: Jae-Hung Han

E-mail: jaehunghan@kaist.edu
 Pfeiffer et al.: Ornithopter Flight Simulation Based on Flexible Multi-Body Dynamics 103
computational efficiency[7–8]. sub-structure. Except the main-wing, all other bodies of
The ornithopters have nonlinear, time-varying, and the ornithopter such as fuselage and tail-wing, are as-
periodic flight characteristics due to the flapping-wing sumed to be rigid for simplicity in this study. In case of
aerodynamics governed by unsteady low Reynolds the flapping-wing aerodynamics, a reduced-order aero-
number flow, and very thin flexible membrane wing dynamic model is used for high computational effi-
structures. Moreover, the wing kinematics known as ciency. The computational fluid dynamics techniques
flapping, pitching, twisting, and lagging also affects to are used to solve the flapping-wing aerodynamics in low
both aerodynamics and structural dynamics, namely, the Reynolds number regime. However, it takes 10 hours to
fluid-structure interaction[9]. Recently, some studies for simulate only 4 flapping cycles of hawkmoth wing-body
flight simulations of flapping wing vehicle have been model[13]. In the preliminary design and analysis of or-
reported using insect (e.g. Drosophila) modelings due to nithopter, such computational burden might be a serious
the plenty of experimental and numerical aerodynamic drawback. Our aerodynamic model used in this study is
data. However, they have mostly utilized the rigid wing originated from the Modified Strip Theory (MST) pro-
not a flexible wing, and have calculated the aerodynamic posed by DeLaurier[14], which has been further improved
loads simply in quasi-steady sense, not based on by taking a high resultant angle of attack (i.e., large
fluid-structure interaction[10–12]. flapping and twisting motions in unsteady flow) and
In this study, we proposed the time-efficient inte- dynamic stall effects for plunging-wing motion into
grative simulation framework for the trimmed longitu- considerations[9]. The improved MST was also experi-
dinal flight of model ornithopter considering fluid- mentally validated[15]. More detailed mathematical
structure interaction. We targeted not an insect but a bird, formulations of the MST can be found in Refs. [9] and
such as magpie, which has a single flapping frequency to [14]. Here, the aerodynamic model was programmed by
make simple and dominant flapping-wing motion, and a FORTRAN, and included as a global- force-subroutine
tail-wing to improve longitudinal flight stability. (gfosub.f) in MSC.ADAMS. The wing sectional kine-
matic variables, such as plunging, and pitching motion,
2 Efficient flight simulation methodology and
simulated in MSC.ADAMS are used as the input pa-
validation rameters of our aerodynamic model to calculate the
To successfully predict the flight characteristics of instantaneous sectional aerodynamic loads. The
ornithopters, the following three items need to be con- fluid-structure interaction analysis was conducted by
sidered: (1) The structural wing deformation (finite combining the flexible flapping-wing parts, the im-
element model of the flexible wing) and the kinematic proved MST in MSC.ADAMS. Fig. 1 shows the flow-
motion of the ornithopter wing (flexible multi-body chart of proposed efficient ornithopter flight simulation
dynamics), (2) The flapping-wing aerodynamics, and (3) framework.
The fluid-structure interaction approach.
In the phase of preliminary sizing and design of
ornithopter, we need a time-efficient design tool which
provides the flapping-wing aerodynamics considering
the fluid-structure interaction and flight dynamic be-
havior with reasonable accuracy.
For the efficient analysis of the flexible multi-body
system such as ornithopter, the commercial software
MSC.ADAMS can be used to model the flexible-wing
kinematic motion with the finite element analysis soft-
ware such as ANSYS in the form of modal neutral file.
The theoretical background of this interface is the
Craig-Bampton’s component mode synthesis (also
known as dynamic sub-structuring), which uses a Fig. 1 Simulation flowchart of proposed efficient flapping-wing
modal approach to represent the flexibility of each flight simulation framework.
 104 Journal of Bionic Engineering (2010) Vol.7 No.1

The proposed efficient integrative simulation motion can be simultaneously measured. It is worth
framework was validated by the experimental wind saying that the constraint force is the sum of the aero-
tunnel testing data in Ref. [9]. The dynamic forces gen- dynamic forces generated by the interaction between
erated from the Rectangular Flapping-Wing (RFW) airflow and flapping-wing motion, and the inertial forces
model[9] were compared with the simulation results in of the flexible flapping-wing.
this study. The RFW model has the wingspan of 540 mm Fig. 2 shows the comparisons of the vertical and
and the aspect ratio of 5.58. To model the RFW wing in horizontal constraint forces at the flapping frequency f =
the ANSYS, the carbon fiber beams were modeled using 4 Hz, the flow speed U∞ = 6 m·s−1, and the mean angle of
BEAM4 and PIPE16 elements (the number of elements: attack i.e. the mean pitch attitude θMEAN = 0˚ and 20˚, for
118), and the influence of the skin was considered by the following three cases: (1) the experimental data
adding point masses for the natural frequency adjust- (Ref.[9]-EXP), (2) in-house code simulation (Ref.[9]-
ment (MASS21 elements, the number of elements: 100) SIM), and (3) the calculated results in this paper by using
to the structure, and the structural model was also ex- the proposed Efficient Fluid-Structure Interaction
perimentally confirmed by matching the natural fre- (E-FSI). The vertical constraint force, FV, is generated in
quencies, the 1st, 2nd, and 3rd mode shapes and the the form of a sinusoidal function with the same fre-
frequency response functions at the wing tip. The de- quency as that of the flapping-wing motion, δM. The
tailed material properties and dimensions are explained maximum value of the vertical constraint force lies at the
in Ref. [9]. middle of the downstroke (shaded area) where the flap-
In the wind tunnel test described in Ref. [9], the ping velocity is maximized. The horizontal constraint
flapping-wing motion generator device is fixed on a test force, FH, also has harmonic wave form. However, the
stand consisting of two single-point bending beam type horizontal constraint force has the twice dominant fre-
load cells which are perpendicular and parallel to the quency with respect to the flapping wing frequency.
flow direction, respectively, so that the vertical and Incase of θMEAN = 0˚, the maximum of horizontal con-
horizontal constraint forces produced by the flapping straint force occurs approximately at the middle phases

60 60
Ref. [9]-EXP Ref. [9]-SIM E-FSI Ref. [9]-EXP Ref. [9]-SIM E-FSI
40 40

20 20

0 0

−20 −20

−40 −40

2 2

1 1

0 0

−1 −1

0.3
0.2
0.2

0.1 0

0 −0.4

−0.1
−0.4
−0.2
−0.6
0.00 0.15 0.30 0.45 0.60 0.75 0.00 0.15 0.30 0.45 0.60 0.75
Time (s) Time (s)
(a) θMEAN = 0˚, f = 4 Hz, U = 6 m·s−1 (b) θMEAN = 20˚, f = 4 Hz, U = 6 m·s−1
Fig. 2 Validation of proposed efficient flapping-wing simulation framework.
 Pfeiffer et al.: Ornithopter Flight Simulation Based on Flexible Multi-Body Dynamics 105
of the downstrokes and upstrokes (without shaded area). The applied forces and moments consist of the
And for θMEAN = 20˚, the maximum values of the hori- aerodynamic and inertial loads as described in Eq. (2),
zontal constraint forces were generated at the middle where the subscripts of F, XB, and YB indicate the forces
stroke like the symmetric flapping condition but during acting on the body-fixed coordinates, and θ denotes the
the upstroke, there are distortions due to the wing de- Euler angle. The aerodynamic loads are calculated from
formation. the flapping-wing aerodynamic model by using the in-
The proposed E-FSI simulation shows reasonable stantaneous main-wing sectional velocity and the flow
accuracy compared with the experimental data in Ref. speed, and the inertial loads are computed inside the
[9]. And in the point of computational efficiency, our MSC.ADAMS.
integrative simulation takes approximately two minutes
for the calculation of fluid-structure interaction of 10 ⎪⎧ FX B ⎪⎫ ⎡ cos θ sin θ ⎤ ⎧ FH ⎫ G G G
⎨ ⎬= ⎢ ⎥ ⎨ ⎬ =FAerodynamics + FInertia =F (2)
flapping cycles with 0.005 second integration time step ⎩⎪ FZ B ⎭⎪ ⎣sin θ − cos θ ⎦ ⎩ FV ⎭
(AMD Phenom 9500 Quad-core Processor, 4 GB RAM). Our ornithopter model has the center of gravity (or
In the following section, we extended this E-FSI the center of mass) at point P1 on the rigid fuselage, as
simulation framework of flexible flapping-wing to the shown in the Fig. 3. Most of the mass are concentrated in
flight dynamics simulation of model ornithopter which the fuselage. During the continuous flapping-wing mo-
has fuselage and tail-wing. tion, the change of the point P1 with respect to the point
3 Ornithopter system modeling P2 in the body-fixed coordinate is negligible due to the
light flexible flapping-wing structure. The flexible
A flight simulation of ornithopter can be performed flapping-wing, so called main-wing is attached to the
by analyzing the flexible multi-body dynamics i.e. rigid fuselage at the points P2 and P3. The subscript W
aeroelastic analysis of the flexible flapping-wing in indicates the main-wing-fixed coordinate. The origin of
relation to the rigid body dynamics of the ornithopter the coordinate is located at the point P2, and has the
body, or more precisely the center of gravity of entire flapping-wing motion, δM. The main-wing motion has
vehicle system. For this purpose a simplified tail-wing the sine wave form with the amplitude 33.06˚ and the
model is constructed and the flight simulations are per- single flapping frequency, f. The point P3 is the universal
formed using the proposed method. joint to make the proper passive wing deformation for
In general, the longitudinal flight dynamics in the positive thrust, and the point P4 has the only one degree
body-fixed coordinate can be expressed as Eq. (1)[16], of freedom, δT, which indicates the tail-wing elevation
where the flight state variables U, V, Q, and θ, are de- angle.
fined as the forward, and the vertical velocity, the pitch YW
rate, and the pitch attitude, respectively, in the
body-fixed coordinate described in Fig. 3. M YB , and IYB
YB
are the moment and the moment of inertia in the direc-
δM
tion of YB. xI and zI are the forward and vertical position XW
XB
P2
of the center of gravity in the reference frame, as de- P5
FV
U∞ θ
scribed in Fig. 3. For a given initial condition, we can FH P1 P3
YT
integrate Eq. (1) to find the flight trajectory at time t > 0. xCG P4 (θ−δT)
h mg XT δT
⎧U = −QW − g sin θ + FX B m YI ZT
ZB
⎪
⎪W = QU + g cos θ + FZ B m XI
⎪ ZI
⎪⎪θ = Q
⎨ (1) Fig. 3 Schematic of the model ornithopter.
⎪Q = M YB IYB
⎪ 3.1 The flexible part – Elliptic Flapping Wing (EFW)
⎪ X I = U cos θ + W sin θ Most of the flapping-wing vehicles in nature use
⎪ 
⎩⎪ Z I = −U sin θ + W cos θ = h the quarter-ellipse or ellipse wing shape rather than a
 106 Journal of Bionic Engineering (2010) Vol.7 No.1

rectangular wing shape in order for the induced drag

270 mm
minimization. Our ornithopter model also adopts the P2
YW LE1 LE2
nature-like quarter-ellipse wing shape with a wingspan
2 STIF4
of 540 mm and an aspect ratio of 5.58, approximately −XW DIAG
P3 DIAG1
similar to the RFW model described in Section 2. By STIF3
performing a heuristic optimization process, design
STIF2
parameters of the wing structure such as Young’s STIF1
Aerodynamic surface
modulus and geometry (e.g. location of stiffeners) were (Strip)

swept to obtain an efficient wing, which is able to gen- Fig. 4 Structural model of EFW with overlapped aerodynamic
erate sufficient positive mean lift and thrust for the sus- surfaces (strips).

tained flight. The material properties, such as Young’s

modulus, density and thickness of EFW structural θMEAN ˚
3.0
components, are listed in Table 1.
2.5
2.0
Table 1 Material properties and dimension of EFW structure θMEAN ˚

FV,MEAN
1.5
E Poisson ρ Thickness 1.0
Part name
(GPa) Ratio (kg·m−3) (mm) 0.5
LE1 250 0.31 1800 2 0.0
LE2 100 0.31 1800 1 −0.5
20.0
DIAG1~2 175 0.31 500 0.4 17.5 20θ
18 MEAN ˚
STIF1~4 175 0.31 500 0.4 15.0 16
12.5 14
SKIN 100 0.31 500 0.04 U∞ (m·s−1) 12
10.0 10 f (Hz)

We applied the BEAM4 elements for the lead-

ing-edge and stiffeners which were made of carbon-fiber
rod. For the skin of flexible flapping-wing, SHELL63
elements were used instead of the membrane elements.
FH,MEAN

That is why we used rather highly stiff materials with

very thin thickness in the skin modeling. In our analysis
framework, the only linear elements were used to model
the flexible flapping-wing ornithopter. Total number of
finite elements was 1857, and it was sufficient to rep-
resent the structural dynamic response of the flexible
flapping-wing.
Fig. 5 Mean values of vertical and horizontal constraint forces vs.
As shown in Fig. 4, the quarter-ellipse wing area flight speed, flapping frequency, and mean pitch angle.
was divided into 27 aerodynamic surfaces. The aero-
dynamic loads, such as section normal and tangential pitch angle, and is linearly proportional to these two
forces and moments, were applied at the section quarter parameters. In comparison, the flapping frequency has a
chord which is closed to the location of elastic axis. smaller influence but a linear relationship as well. The
Fig. 5 shows the overall performances of the mean thrust increases linearly with the flapping fre-
main-wing in terms of mean value of vertical and hori- quency and decreases linearly with the flow speed in all
zontal constraint forces against flight speed, flapping cases of mean pitch angle. The mean pitch angle has
frequency, and mean pitch angle. For each case, the only a minor influence and when it was increased a
mean values of constraints forces were obtained for the slight decrease of the mean thrust was seen. This aero-
10 flapping cycles. dynamic map can be used to find the trim flight condi-
It can be seen that the amount of mean vertical tion that and it will be introduced in the following sec-
force depends mostly on the flow speed and the mean tion.
 Pfeiffer et al.: Ornithopter Flight Simulation Based on Flexible Multi-Body Dynamics 107
3.2 The rigid part – The fuselage and the tail-wing conditions in terms of θMEAN, f, and U∞ from the aero-
To perform a complete ornithopter flight simulation, dynamic map of main-wing as shown in Fig. 5.
a tail-wing aerodynamic model is necessary to balance When we install the EFW that was designed in 3.1,
the non-zero flapping-wing excitation moment for our to the fuselage, we should consider the location of the
model ornithopter system. We use a simple tail-wing center of gravity, and the center of pressure satisfying
aerodynamic model, the thin-airfoil theory with the the vehicle longitudinal stability, the pitching moment
finite-wing consideration, as shown in Eq. (3). The equilibrium at the point P1. To do this, we should design
flapping-wing excitation moment can be calculated im- the proper tail-wing to hold the vehicle pitching attitude
plicitly by the cross-product of time-varying aerody- near the pre-described trimmed θMEAN. Table 2 shows
namic forces generated by continuous flapping-wing the specification of our model ornithopter. The tail-wing
motion, and moment arm which is also time-varying is assumed to be a thin rigid plate and approximated by a
distance between the center of pressure and the center of point-mass of 1.5 g. The lift force of tail-wing depends
gravity. We assume that the difference between the ve- on such parameters as the mean pitch attitude θMEAN, the
hicle pitch attitude, θ, and the tail-wing elevation angle, tail-wing elevation angle δT and the flight speed U∞.
δT, is proportional to the tail-wing lift coefficient, CL,T.
The tail-wing drag coefficient, CD,T, or −CT,T (CT,T, Table 2 Specification of the model ornithopter

tail-wing thrust coefficient) is consists of two parts. (1) Total mass Length Mean chord Wing span
Zero-lift drag or friction drag, and (2) Induced drag. The 0.118 kg 0.18 m 0.05 m 0.54 m

tail-wing geometry is represented by the wing aspect Main-wing Tail-wing Main-wing Tail-wing
area area aspect ratio aspect ratio
ratio, ART.
0.044 m2 0.007 m2 5.58 2
⎧ ART
⎪CL,T = 2 + AR ⋅ 2π (θ − δ T ) In case of the fixed-wing aircraft at the trimmed
⎪ T
⎨ (3) point, most of the vehicle flight state variables should
⎛ 2
CL,T ⎞
⎪C = −C = − ⎜ C + ⎟⎟
⎪ T,T D,T ⎜ D0 ,T πeAR have zero or very small fluctuation due to the wind gust
⎩ ⎝ T ⎠
or turbulence, and the external applied forces and mo-
When we search the trim conditions of free flight of ments are also zero. However, ornithopter’s large flap-
model ornithopter, we should consider the force and ping-wing motion continuously excites the vehicle flight
moment equilibriums at the center of gravity. Due to the state variables. Moreover, the time-varying forces and
continuous dynamic motion of main-wing, all the forces moments with non-zero mean value, which are applied
and moments are time-varying. So, we should construct to the ornithopter dynamics, can cause the divergence of
the force and moment equilibrium in the time-average vehicle state variables during the numerical integration.
sense by Eq. (4). The subscript, M and T, represent the We heuristically changed the tail-wing area, aspect ratio,
main-wing and the tail-wing, respectively, and the sub- the tail-wing elevation angle, and the distance between
script CG represents center of gravity. points P1 and P4 (Fig. 3), so that the vehicle states were
bounded or converged to zero. The ornithopter trimmed
⎧∑ FV = ( FV, M ) + ( FV, T ) − mg = 0
⎪⎪ MEAN MEAN flight criterion shown in Eq. (5) is weaker than the
⎨∑ FH = ( FH, M )MEAN + ( FH, T )MEAN = 0 (4) fixed-wing aircraft one. In other words, all the longitu-
⎪
⎪⎩∑ M CG = ( M M )MEAN + ( M T )MEAN = 0
dinal flight state variables vector xlongitudinal should have a
finite value of 2-norm bounded by arbitrary finite value,
The mean vertical force of tail-wing is small R. The value of R is determined by the observation of
enough compared with that of the main-wing. Most of real physics of ornithopter flight, for example, R(θ) = 40˚,
the weight, mg, of the ornithopter is compensated by the in this study. The time history of state variables in orni-
mean aerodynamics of main-wing. Especially, as shown thopter trimmed flight can be found in the next section.
in Fig. 5, the mean pitch angle θMEAN is the most effec-
tive control parameter to resist its gravity. In the pre- ⎧ xlongitudinal = [U W θ Q h]
T
⎪
⎨ (5)
liminary design of ornithopters, once the total mass
⎪⎩ xlongitudinal 2 ≤ R < ∞
budget is fixed, then we can find several possible trim
 108 Journal of Bionic Engineering (2010) Vol.7 No.1

4 Flight simulation results and discussions Table 3 Trimmed flight condition of the model ornithopter
For any flying object, the flight stability is as es- Flapping frequency Tail-wing angle Flight speed
sential as lift and thrust. It stands for the tendency of the 12.0 Hz 11.85˚ 10.8 m·s−1
air vehicle which returns to its original equilibrium po-
sition after it has been perturbed. The flapping-wing Fig. 6 illustrates the trim flight characteristics of the
flight inherently produces the time-varying forces and ornithopter; the change of the pitch attitude θ with re-
moments due to their continuous wing motion. In order spect to the main-wing flapping motion can be clearly
to maintain a stable flapping-wing flight, the tail needs observed. During the downstroke, the pitch attitude
to control the mean pitching moment at point P1, re- starts to change when the wing passes approximately
sulting from the aerodynamic forces acting on the through the neutral position, since the total aerodynamic
main-wing. force, which acts in front of the center of gravity, shows
One of the trim conditions for the ornithopter a maximum at this point. During the upstroke the value
model is specified in Table 3. The initial guess of trim θMEAN is reduced again to its initial position. Fig. 7
condition for 0.118 kg model ornithopter, are approxi- shows the flight trajectory with the given trim configu-
mately (θMEAN, f, U∞) = (10˚, 12 Hz, 10 m·s−1) from the ration in Table 3.
aerodynamic map in Fig. 5. During the heuristics flight The body oscillates vertically (parallel to the grav-
simulation constrained to Eq. (5), we find the trimmed ity) with the peak-to-peak value of approximately 1 cm
tail-wing elevation angle δT = 11.85˚, and the initial trim at the same flapping frequency of main-wing motion,
conditions are refined as (θMEAN, f, U∞) = (9.2˚, 12 Hz, namely short period mode, and the long periods mode
10.8 m·s−1). also can be found with approximately 3 cm peak-to-peak

Fig. 6 Longitudinal flight of the model ornithopter at trim condition for one flapping cycle.
|δM(f)| (degree)

30 30 12 Hz
δM (degree)

0
15
−30
δM, MEAN = 0˚
0
12 Hz
|θ(f)| (degree)
θ (degree)

10
15
5
0
θMEAN = 9.2˚ 0
2.16 1
1×10−1
|h(f)| (m)
h (m)

1×10−2 12 Hz
2.13 1×10−3
1×10−4
hMEAN = 2.14 m
1×10−5
0 1 2 3 4 5 0 10 20 30 40
Time (s) f (Hz)
Fig. 7 Flight trajectory and oscillating vehicle pitch attitude with trim condition. θMEAN = 9.2˚, f = 12 Hz, U = 10.8 m·s−1.
 Pfeiffer et al.: Ornithopter Flight Simulation Based on Flexible Multi-Body Dynamics 109
value. The vehicle pitch attitude, θ is also harmonically To find the source of the limit-cycle-oscillation of
fluctuating from 0˚ to 20˚. The Fast Fourier Transform pitching dynamics of model ornithopter quantitatively,
(FFT) of main-wing motion, δM, the pitch attitude, θ, and we introduce the Zero Moment Point (ZMP), MZMP,
the altitude, h, was conducted, and each frequency which implies the point which has zero moment and all
spectrum has the dominant frequency at 12 Hz. Similar external forces acting on that point as a resultant force.
oscillating tendencies are found in the real flight data of To do this, we constrain the translational and rotation
ornithopter[17] and the numerical analysis results[16]. degrees of freedom of point P1 in Fig. 3, and fix the
mean pitch angle θMEAN = 9.2˚. We name this numerical
Limit-cycle-oscillation and the zero moment point experiment as a constrained flight test or a tethered flight
In Eq. (1), the time derivative of pitch attitude is the test. During the constrained flight test, we can acquire
pitch rate, and the time derivative of pitch rate is equal to the vertical and horizontal force coefficients, CF,V, and
the pitching moment normalized by the moment of in- CF,H, and the pitching moment, CM,CG at point P1 under
ertia. To make the pitch attitude time-invariant, the pitch the trimmed free flight conditions such as (θMEAN, f, U∞,
rate and the pitching moment should be zero during the δT) = (9.2˚, 12 Hz, 10.8 m·s−1, 11.85˚).
continuous flapping-wing motion. However, we ob- The moment with respect to point P5 in Fig. 3 can
served the large pitch attitude oscillation with the flap- be expressed as Eq. (6). In Eq. (7), we recollect the terms
ping frequency, and the relationship between the pitch in Eq. (6) divided by 0.5ρ∞U∞2SMcMEAN where cMEAN =
attitude and the pitch rate makes the closed circular phase (main-wing area, SM)/(main-wing-span, bM). The dis-
portrait as shown in Figs. 7 and 8. This kind of closed tance between the center of gravity and the zero moment
curve represents so called the limit-cycle-oscillation, point, ε in Eq. (7), is calculated based on the result in
which is one of the nonlinear phenomena of nonlinear Fig. 9a.
dynamic systems. It was observed that with the trim
M P5 = M ZMP − xZMP ⋅ ( FV cos θ − FH sin θ )
condition of the ornithopter the flight states showed the (6)
same steady state oscillation patterns regardless of the = M CG − xCG ⋅ ( FV cos θ − FH sin θ )
perturbation of initial conditions within certain ranges.
( xZMP − xCG )
We cannot make the trimmed flight of our model ε= = ( xZMP − xCG )
cMEAN
ornithopter have the constant pitch attitude due to the
CM, CG
continuous flapping wing motion. To reduce the oscil- =− (7)
CF,V cos θ MEAN + CF,H sin θ MEAN
lation amplitude or to have time-invariant pitch attitude,
we need a damping component in the system model, or
time-varying control input by means of the tail-wing 30

elevation angle motion, or the flapping motion. 0

−30
0.5
0.0
−0.5
0.15
Pitch rate (degree·s−1)

0.00
−0.15
0.2
0.0
−0.2

0.0 0.1 0.2 0.3 0.4

Time (s)
(a)

Fig. 9 Coefficient of constraint forces and moment with respect to

the flapping-wing motion (a), main-wing motion vs. the
Fig. 8 Limit-cycle-oscillation of pitching dynamics under non-dimensional distance between xZMP and xCG for upstroke and
trimmed flight. downstroke motion (b). θMEAN = 9.2˚, f = 12 Hz, U = 10.8 m·s−1.
 110 Journal of Bionic Engineering (2010) Vol.7 No.1

1 limit-cycle-oscillations of pitching dynamics.

Static moment equilibrium
Downstroke
Acknowledgement
(xZMP − xCG) / CMEAN

0
This work was supported by the Defense Acquisi-
tion Program Administration and Agency for Defense
−1 Upstroke Development under the contract UD090082JD,
Jun-Seong Lee thanks the support of the Brain Korea 21
−2 Project in 2010.
−40 −30 −20 −10 0 10 20 30 40
δM (degree) References
(b)
[1] Han J H, Lee J S, Kim D K. Bio-inspired flapping UAV
Fig. 9 Contiuned. design: A university perspective. Proceedings of SPIE, San
Diego, USA, 2009, 7295I-1–7295I-12.
Let ε be the stability residual for ornithopter lon-
[2] Pines D J, Bohorquez F. Challenges facing future mi-
gitudinal flight. If ε approaches to zero, then the zero
cro-air-vehicle development. Journal of Aircraft, 2006, 43,
moment point, xZMP, and the center of gravity, xCG, are
290–305.
the same, called the static moment equilibrium at each
[3] Jones K D, Bradshaw C J, Papadopoulos J, Platzer M F.
time step. Moreover, ε has zero value at a certain phase
Improved performance and control of flapping-wing pro-
of flapping-wing motion; the moment at the center of
pelled micro air vehicles. Proceedings of the 42nd Aero-
gravity is also zero. However, our constrained flight test space Sciences Meeting & Exhibit, Reno, USA, 2004,
under free flight trim condition shows that the behavior AIAA-2004-0399.
of ε depends on the flapping wing motion, and not ex- [4] Wood R J. Liftoff of a 60mg flapping-wing MAV. Proceed-
actly zero. This causes the limit-cycle-oscillation of ings of the 2007 IEEE/RSJ International Conference on In-
longitudinal trim flight of model ornithopter, thus there telligent Robots and Systems, San Diego, USA, 2007,
should be an additional moment, such as active tail-wing 1889–1894.
motion, to make ε zero for time-invariant pitch attitude, [5] Park J H, Yoon K J. Designing a biomimetic ornithopter
at the end of upstroke and the downstroke motion. capable of sustained and controlled flight. Journal of Bionic
Engineering, 2008, 5, 39–47.
5 Conclusions
[6] Liu T. Comparative scaling of flapping- and fixed-wing
To perform a simulation of a trimmed flap- flyers. AIAA Journal, 2006, 44, 24–33.
ping-wing flight, it is necessary to consider flexible [7] Han J H, Lee J Y, Kim D K. Ornithopter modeling for flight
multi-body dynamics as well as the fluid-structure in- simulation. Proceedings of the International Conference on
teraction. We have established an efficient integrative Control, Automation and Systems, Seoul, Korea, 2008,
simulation framework for the flapping-wing flight using 1773–1777.
MSC.ADAMS and ANSYS in combination with a re- [8] Drela M. Integrated simulation model for preliminary
duced-order MST model. The proposed method has aerodynamic, structural, and control-law design of aircraft.
been verified by comparing experimentally measured Proceedings of the 40th AIAA SDM Conference, St Louis,
forces of a flapping-wing and the corresponding simu- USA, 1999, AIAA-99-1394.
lation results. Based on the aerodynamic map of the [9] Kim D K, Lee J S, Lee J Y, Han J H. An aeroelastic analysis

main-wing, we successfully designed a model orni- of a flexible flapping wing using modified strip theory.
Proceedings of SPIE, San Diego, USA, 2009, 69281O-1–
thopter which was capable of flying in trim. We ob-
69281O-10.
served the limit-cycle-oscillation during the trimmed
[10] Smith m J C. Simulating moth wing aerodynamics: Towards
longitudinal flight of the model ornithopter, and we tried
the development of flapping-wing technology. AIAA Journal,
to interpret the physics by introducing the distance be-
1996, 34, 1348–1355.
tween the zero moment point and the center of gravity, ε,
[11] Grauer J A, Jr Hubbard J E. Multibody model of an orni-
so called the stability residual. The present simulation
thopter. Journal of Guidance, Control, and Dynamics, 2009,
framework provides us an efficient way to study com-
32, 1675–1679.
plex flights dynamics of ornithopters and to reduce
 Pfeiffer et al.: Ornithopter Flight Simulation Based on Flexible Multi-Body Dynamics 111

[12] Dickson W B, Straw A D, Dickinson M H. Integrative model tion of a flapping-wing aerodynamic model for MAV ap-
of Drosophila flight. AIAA Journal, 2008, 46, 2150–2164. plications. Proceedings of SPIE, San Diego, USA, 2008,
[13] Liu H, Aono H. Size effects on insect hovering aerodynam- 69282M-1–69282M-8.
ics: An integrated computational study. Bioinspiration and [16] Dietl J M, Garcia E. Stability in ornithopter longitudinal
Biomimetics, 2009, 4, 015002-1–015002-13. flight dynamics. Journal of Guidance, Control, and Dy-
[14] DeLaurier J D. An aerodynamic model for flapping-wing namics, 2008, 31, 1157–1162.
flight. Journal of the Royal Aeronautical Society, 1993, 97, [17] Grauer J A, Jr Hubbard J E. Inertial measurements from
125–130. flight data of a flapping-wing ornithopter. Journal of Guid-
[15] Lee J S, Kim D K, Lee J Y, Han J H. Experimental evalua- ance, Control, and Dynamics, 2009, 32, 326–331.