52nd AIAA/ASME/ASCE/AHS/ASC Structures, Structural Dynamics and Materials Conference<BR> 19th AIAA 2011-2008

4 - 7 April 2011, Denver, Colorado

Approximate Aerodynamic and Aeroelastic Modeling of

Flapping Wings in Hover and Forward Flight

Abhijit Gogulapati ∗ and Peretz P. Friedmann †

Department of Aerospace Engineering, The University of Michigan, Ann Arbor, 48109, USA

Results generated from an aeroelastic model that combines a nonlinear structural dynamic
model based on MARC with an approximate aerodynamic model that incorporates leading edge
vortices and a wake model are presented. The aerodynamic model, used in our earlier studies, is
extended to forward flight, and the effect of fluid viscosity is incorporated into the formulation
in a partial manner. Results presented describe aerodynamic and aeroelastic studies conducted
on airfoils and flapping wings in hover and forward flight. For the rigid cases considered, the
approximate model shows reasonable agreement with CFD based results and predicts the trends
accurately. The aeroelastic results presented indicate that wing flexibility has beneficial effects
in both hover and forward flight. Comparisons show that aerodynamic loads are comparable to
inertia loads for the cases considered.

Nomenclature
A 1 − A 11 Coefficients that are computed from the airfoil degrees of freedom
Af ,Bf Filter coefficients
Aw Wing area
d A sp Instantaneous area vector in the ( X SP , YSP , Z SP ) coordinate system
a Distance between pitching axis and mid-chord of the airfoil
CD Drag coefficient
CL Lift coefficient
CT Thrust coefficient
c airfoil chord
cr Root chord
D Drag
E Elastic modulus
e Unit vector; subscript identifies the direction.
F sp Force vector in the ( X SP , YSP , Z SP ) coordinate system
F h , Fv Components of the aerodynamic forces in the horizontal and vertical directions
f Flapping frequency
h Plunge degree of freedom of the airfoil
j, k Indices
L Lift
LC Circumference of the airfoil
l Lead-lag degree of freedom of the airfoil
l0 Lead-lag amplitude
M User defined number of sections
mf Order of the filter
Nsections Total number of aerodynamic span stations on the wing
Nθ Discretization of the circle
n wksubit Number of wake sub-iterations
p Local static pressure on the airfoil
∗ Ph.D. candidate, Student Member AIAA.
† François-Xavier Bagnoud Professor, Fellow AIAA.

1 of 31

Copyright © 2011 by Abhijit Gogulapati and Peretz Friedmann. American Institute

Published by of Aeronautics
the American Institute ofand Astronautics
Aeronautics and Astronautics, Inc., with permission.
p∞ Free stream pressure
p ap pl ied Pressure that is applied on the wing
p computed Pressure that is computed using the unsteady Bernoulli principle
p l imit An upper bound value for pressure
p re f Reference pressure
q Velocity vector on the normal cylinder
Re Reynolds number
Rj Radial location of the j th spanwise section with respect to the wing root
R span Span of a finite wing, measured from root to the tip
R0 Radial coordinate of a point on the circle in the complex plane
rc Vortex core radius
rv Distance
rv Non-dimensional distance
s Arc coordinate
s0 Origin of the curve that is used to integrate along the airfoil
s wk1 , s wk2 Arc coordinates of the shear layers on the airfoil
T Thrust
t Time
∆t Time increment
tv Age of a vortex
tv Non-dimensional time
Ure f Reference speed
U ti p Maximum tip speed
u ∞ , v∞ Components of the free stream velocity resolved parallel and normal to the stroke plane
U∞ Free stream velocity vector
U∞ Magnitude of U ∞
UI Free stream velocity vector in the ξ I − η I coordinate system
UI Magnitude of U I
u E , vE Components of free stream in the vortex wake model
Vw Wing volume
v ind Velocity induced by a vortex
vξ , vη Airfoil velocities, measured in the ξ − η coordinate system
f l ex f l ex
∆ vξ , ∆ vη Incremental components of velocity due to airfoil flexibility, measured in the ξ − η coordinate system
vθ Tangential component of fluid velocity on the circle
( X SP , YSP , Z SP ) Coordinate system fixed to the stroke plane
( X w , Yw , Z w ) Wing fixed coordinate system
x sp Instantaneous position vector of a point on the wing in the ( X SP , YSP , Z SP ) coordinate system
y f iltered , yun f ilteredFiltered and unfiltered signals
z vwk Coordinate of a shed vortex on the complex plane

Greek Symbols
α Pitch angle
α0 Pitch amplitude
αf s Angle between free stream velocity vector and eξ I
αr Feathering angle in hawkmoth kinematics
αs j , α c j Fourier coefficients in hawkmoth kinematics
α tot Instantaneous angle of attack
β Flap angle
βsp Angle between the stroke plane and the free stream velocity vector
β0 Flap amplitude
Φr Sweep angle in hawkmoth kinematics
Φ0 , Φ s j , Φ c j Fourier coefficients in hawkmoth kinematics
ϕ̃ Angular position of a vortex or wing section on the normal cylinder, shown in Figure 7
ϕ Angular position of a vortex in an airfoil coordinate system
ϕα Phase

2 of 31

American Institute of Aeronautics and Astronautics

φ Velocity potential
p
ı −1
ψ Rotation vector
ζ Coordinate of a point in an airfoil coordinate system on the normal cylinder
Γf s Circulation due to free stream
Γ0 Quasi-steady circulation
γb Bound vorticity
γf s Component of vorticity due to free stream
γus Component of vorticity due to airfoil velocities
γwi Component of vorticity on the circle that is induced by the wake
dΓvwk Circulation a vortex
Θr Elevation angle in hawkmoth kinematics
Θ0 , Θs j , Θ c j Fourier coefficients in hawkmoth kinematics
θ Angular coordinate on the circle in the complex plane
τ, σ Thickness and camber parameters in the aerodynamic formulation
µ Advance ratio
ν∞ Kinematic viscosity of the fluid
ν Poisson’s ratio
ω Circular frequency
ρ Density of the material
ρ∞ Free stream density of the fluid
(ξ, η) Coordinates in an airfoil coordinate system; superscript identifies the coordinate system
ς Dummy variable, used in integration

Additional Subscripts and Superscripts

| ·| Absolute value or magnitude
( ¨· ) Double derivative with respect to time
( · ) | inviscid Inviscid quantity
( · ) |viscous Viscous quantity
( · )∗ Complex conjugate
( · ) f l ex Quantity corresponding to a flexible airfoil

I. Introduction
During the last fifteen years there has been increasing interest in micro air vehicles (MAVs) for both mil-
itary and civilian missions that involve confined spaces, such as buildings, or short distances. These vehicles
typically have maximum geometric dimensions of 15 cm, maximum weight of 100 grams, and are expected
to operate at low Reynolds number (102 < Re < 104 ) and low forward flight speed (< 15 m/s).1 In particular,
flapping wing designs, inspired by hover-capable biological flyers such as insects, bats, and hummingbirds, are
particularly interesting due to the exceptional flight capabilities observed in the biological counterparts.1
A significant portion of the research on flapping wing vehicles has focused on understanding the mech-
anisms that generate unsteady aerodynamic forces. This research1–7 has identified leading edge vortices
(LEVs), wake capture, and tip vortices, as the primary force generating mechanisms. Recent emphasis has
been on investigating the dynamics of LEVs, and its interaction with kinematics.8 Water tunnel experiments,
conducted using airfoils undergoing combined pitch-plunge motion, showed that the behavior of LEVs was
dependent on the reduced frequency for a given effective angle of attack.8 In these tests, flow field mea-
surements were done using particle induced velocimetry (PIV), and the effective angle of attack was defined
based on the combined pitch-plunge motion. Attempts to model the aerodynamic environment in a quantita-
tive manner have been based on two approaches: (1) computational fluid dynamics (CFD) simulations based
on the solution of the Navier Stokes (NS) equations and (2) approximate aerodynamic models based on po-
tential flow solutions. Simulations using CFD yield the best resolution of the unsteady flow field. However,
such approaches require significant amounts of computer time that prevent parametric studies. Approximate
aerodynamic models offer a compromise between accuracy and computational efficiency and thus are suitable
for trend and design studies. It is important to emphasize that the approximate models that have practical

3 of 31

American Institute of Aeronautics and Astronautics

applications have to be able to model the effect of LEVs and wake capture.
The approximate unsteady aerodynamic theories used for flapping wing problems can be classified as as-
sumed (or prescribed) wake and free wake models. Assumed wake models are classical unsteady models such
as Theodorsen’s theory.9 Reference [10] incorporated the effect of the LEVs in Theodorsen’s theory by modify-
ing the unsteady aerodynamic lift and moment expressions using the Polhamus leading edge suction analogy.11
This model10 was compared to experiments and was capable of predicting the trends in aerodynamic forces.
Free wake models account for evolution of the wake, thereby providing a reasonable approximation to the
development of the unsteady wake during a flapping cycle. In particular, free wake models that account for
LEVs are two-dimensional formulations that use a discrete vortex representation of the wake.12–14 These
formulations are suitable for flapping wings in hover following simplifying assumptions on the geometry of
the shed wake. The model developed in Ref. [12], which accounts for separation close to the leading edge,
compared well with experimental data for airfoils in steady flow. In this approach the chordwise location of
the separation point, which may be obtained using independent computations or experiments, is explicitly
incorporated into the formulation. Reference [13] described a vortex blob based formulation that simulates
the unsteady flow field around flexible thin airfoil that is undergoing prescribed rigid body motion as well as
prescribed deformation. The model displayed good correlation with previously published data for rigid air-
foils; however, the study indicated that additional work was required before the formulation could be used for
aeroelastic studies. References [14, 15] presented the development and implementation of a discrete vortex
model that is applicable to insect-like flapping wings in hover. The model was used to simulate rigid wings un-
dergoing hover kinematics, and for the limited number of cases considered, compared well with experimental
data.15
For the case of forward flight, the presence of a free stream alters the shed wake geometry and wing-wake
interaction. Therefore, additional modifications are needed before two dimensional aerodynamic formula-
tions12–14 can be used to model forward flight. A review of literature indicates that the application of free
wake models that incorporate LEVs to the problem of flapping wings in forward flight has received limited at-
tention. Recently, Ref. [16] extended the aerodynamic formulation developed in Refs. [14, 15] to forward flight
using the concept of prescribed wake adopted from rotary wing aerodynamics. Preliminary results indicated
that forward flight increased the peak force during one half of the stroke and reduced the force during the
other half of the stroke; however, comparisons with CFD based or experimental results were not presented.
The importance of wing flexibility in enhancing the lift producing capability of flapping wings has been
considered in a number of studies. Studies that have attempted to examine this issue in a systematic manner
include Refs. [10, 16–25]. An important finding of Refs. [17, 18], which was later corroborated by experiments
conducted in Ref. [10], was that a dominant component of the loading on flapping wings is due to inertia
loads. Reference [10] examined wing models based on membranes reinforced by metal beams, using a linear
structural model. It concluded that wing flexibility changed the aerodynamic loads and therefore should
be included in the model. Reference [19] examined the influence of flexibility in insect wings by using a
linear finite element model of a dragonfly wing that was combined with CFD based aerodynamic loading
and found that spanwise flexibility had a favorable impact on thrust generated and power consumed by the
flapping wing. Reference [16, 20, 21] described a nonlinear aeroelastic model that is obtained by coupling
a nonlinear finite element model of the wing with an approximate model. These studies, which examined
MAV wings undergoing prescribed motion found that the effect of aerodynamic loads was small compared
to effect of inertia loads for the cases considered; wing flexibility had a small but favorable impact on lift
generation. References [22, 23] describe the development of a computational aeroelastic framework obtained
by combining geometrically nonlinear beam and shell based structural dynamic models with a CFD based
flow field. The framework was used in Ref. [24] to simulate isotropic Zimmerman wings in hover. The results
indicated that the behavior of peak lift was non-monotonic with decrease in elastic modulus, and concluded
that only a limited amount of flexibility was beneficial for thrust generation. Plunging isotropic plunging
wings, offset by a constant flap rotation, in hover for R e = 100 were studied in Ref. [25]. The aeroelastic model
was obtained by coupling a viscous incompressible fluid description based on a lattice Boltzmann model to
a structural description based on a lattice spring model. This study,25 in which flexibility was incorporated
using linear springs, found that a peak in lift coefficient coincided with an excitation frequency that was close
to the resonant frequency of the wings. Combinations of elastic modulus and flapping frequency were found
for which the lift generated was sufficient to support the weight of a realistic insect.
The overall objective of the current paper is to develop an understanding of the effect of flexibility on the
performance of anisotropic flapping wings in hover and forward flight using an aeroelastic model, initially

4 of 31

American Institute of Aeronautics and Astronautics

described in Refs. [16, 20, 21], which has been extended to forward flight. The specific objectives of the paper
are:

1. Extend the aerodynamic formulation to incorporate forward flight.

2. Include the effect of Reynolds number in a partial manner by incorporating the effect of temporal decay
of vorticity once it is shed into the wake.

3. Compare results obtained using the modified aerodynamic model with those obtained using CFD for
airfoils and wings in hover and forward flight.

4. Conduct trend type studies for flexible wings in hover and forward flight.

II. Nonlinear Aeroelastic Model

The aeroelastic model is obtained by coupling a nonlinear finite element model of the wing with an approx-
imate aerodynamic model that incorporates LEVs and a free wake; a detailed description of the individual
components and formulation of the aeroelastic equations of motion for flapping wings in hover was presented
in Refs [20,21]. In this section, a summary of the aeroelastic model including modifications to the aerodynamic
formulation to incorporate flapping wings in forward flight are presented.

A. Structural Dynamic Model and Wing Kinematics

The structural dynamic models for MAV wings are based on the MARC code26 using shell elements that are
capable of undergoing large amplitude rigid body motion as well as moderate-to-large flexible deformation.
A variety of constitutive laws are available so that isotropic as well as anisotropic wings may be modeled.
Wing kinematics, which consist of large amplitude rigid body rotations prescribed at the root, are applied as
displacement boundary conditions at one or more nodes.

B. Approximate Aerodynamic Model

The unsteady aerodynamic model that was selected for use and modification was developed in Refs. [14, 15].
This formulation, originally derived for the case of hover, is based on two-dimensional potential flow and uses a
vorticity/circulation approach to compute the aerodynamic loads. Several modifications were introduced by us.
First, the model was modified to account for spanwise and chordwise flexibility of flapping wings as described
in Refs. [20, 21]. In the current paper, the aerodynamic formulation is extended to incorporate forward flight.
Moreover, the effect of Reynolds number is also incorporated into the calculation of shed vorticity. Details of
these extensions are provided next.

Overview of the Aerodynamic Formulation

The overall approach, which retains the essential aspects of the hover formulation for rigid and flexible
wings,14, 20, 21 is summarized in Figure 1. First, the wing is divided into several spanwise stations, as shown in
Figure 2, where each section is represented as an airfoil. For each airfoil, an airfoil-wake surface that captures
the airfoil degrees of freedom (DOF), and approximates the geometry of the shed wake, is identified. Next, the
airfoil and the airfoil-wake surface are transformed to a circle in the complex plane using a conformal map-
ping. Thus, the airfoil bound and shed wake vorticity are computed on the complex plane. The quasi-steady
component of vorticity is obtained by neglecting the effect of the shed wake. The strength of shed vorticity is
computed by enforcing a stagnation condition at the leading edge (LE) and a Kutta condition at the trailing
edge (TE). The airfoil bound vorticity is obtained as a sum of the quasi-steady and wake-induced vorticity on
the airfoil. Next, the vorticity in the complex plane is transformed back to the airfoil-wake surface (physical
plane) using an inverse transform. The unsteady loads acting on the airfoil are obtained from the total vor-
ticity using the unsteady Bernoulli equation. Finally, the shed vorticity is convected using the Rott-Birkhoff
equation, which is derived from Biot-Savart law for two dimensional flow.

5 of 31

American Institute of Aeronautics and Astronautics

 Divide wing into Identify suitable
spanwise stations wake surface

For each spanwise station

Airfoil-wake Generalized Circle in a

surface Joukowski transform complex plane

Compute quasi-steady
vorticity and circulation
YW
Wing section used in
Rigid root aerodynamic analysis
Compute strengths of vortices
shed into the wake
Stagnation condition at leading edge XW

Flapping axis
Kutta condition at trailing edge Feathering axis

Compute bound vorticity

Circle in a Inverse transform Airfoil-wake Rj

complex plane surface
Figure 2. Spanwise sections on a flapping wing. R j denotes radial
location of the section.
Compute unsteady loads
Pressure using unsteady Bernoulli equation
Force/moment using vortex impulse method

Wake evolution using

the Rott-Birkhoff equation

Figure 1. Schematic of the aerodynamic formula-

tion.

1. The Airfoil-Wake Surface

Wing kinematics of biological flapping wing flyers, in both hover and forward flight, consists of a predominant
sweep or flap motion in the stroke plane (SP), pitching about the feathering axis, and a comparatively small
elevation angle, as indicated in Figure 3.27 Therefore, the feathering axis of the wing is assumed to move on
the stroke plane (SP).14 Consequently, the surface described by the airfoil motion is a cylinder that is normal
to the stroke plane; this normal cylinder (NC) is shown in Figures 4 and 5.
For hover,14 the shed wake is assumed to be confined to the NC; therefore, NC is a convenient choice for
the airfoil-wake surface. For the case of forward flight, the vortices shed into the wake are carried away with
the free stream due to velocity of forward flight and therefore a suitable approximation to the wake surface
has to be identified.

2. Extension to Forward Flight

Forward flight is characterized by a free stream velocity vector that is assumed to lie in the YSP − Z SP plane as
depicted in Figure 4. A complete description of the free stream velocity involves two independent parameters
u ∞ and v∞ . These quantities are determined from the tip speed, advance ratio, and stroke plane inclination
angle, using Eq. (1).
−u∞
µ= so that u ∞ = −µ U ti p and v∞ = u ∞ tan βsp
¡ ¢
(1)
U ti p
where u ∞ , v∞ , and βsp are shown in Figure 4.
In the present analysis, the NC, which is used for hover, is also used to approximate the geometry of the
shed wake in forward flight. The effect of free stream due to forward flight is incorporated by modifying the

6 of 31

American Institute of Aeronautics and Astronautics

 )
(SP
ne
e pla
k
S tro

pronation xis
in ga
pp
Fla
Feathering axis
axis
Body

Hor
izo

up
n

do
tal
Bod p

str
lan

wn
y in e

o
clin

ke
tro
atio Vertical
n

e k
Stro
ke p Direction
lan of flight
e inc supination
lina (Horizontal)
tion

Figure 3. Insect flapping stroke

expressions used to determine the quasi-steady component of vorticity and the vortex wake model as described
later.

3. Coordinate Systems
Several rectangular coordinate systems, which are defined on the flattened NC, are used for each airfoil sec-
tion. These coordinate systems, shown in Figure 6, are listed below, where the unit vectors corresponding to
the axes are denoted using e with an appropriate subscript:

1. ξ, η : Zero lift coordinate system. Origin at the mid-chord; eξ A along the zero lift line (ZLL); eη A normal
¡ ¢

to ZLL.

2. ξ I , η I : Stationary coordinate system fixed to the NC; also shown in Figure 5. eξ I is parallel to the stroke
¡ ¢

plane; eη I is parallel to Z SP . Identified by superscript I .

3. ξnr , ηnr : Non-rotating coordinate system that translates with the airfoil. Origin at the intersection of
¡ ¢

the feathering axis of the wing and the normal cylinder. eξnr and eηnr are parallel to the stroke plane
and eη I respectively. Identified by superscript nr .

4. ξ f a , η f a : Coordinate system that is fixed to the center of rotation of the airfoil. Origin at the intersection
¡ ¢

of the feathering axis of the wing and the normal cylinder. eξ f a and eη f a are parallel to eξ and eη
respectively. Identified by superscript f a.

Each airfoil has three degrees of freedom (DOF) that are defined in ξ I − η I on the flattened NC as shown
in Figure 6; these are lead-lag ( l ), plunge ( h), and pitch (α) respectively. The airfoil DOFs and velocities are
obtained from the structural dynamic model at each time step; consequently, these quantities include the effect
of wing deformation in addition to the effect of the wing kinematics. The effect of spanwise deformation on
the radius of the NC is incorporated by using an average time dependent radius as follows: Let R 0j and R tj
denote the radial locations of a wing section at the start of the motion and ³ at some
´ subsequent time t. Then,
the average radius of the normal cylinder at time t is given by R̄ tj = 0.5 R 0j + R tj . Subsequently, R̄ tj is used to
calculate the distances on the NC.

4. Quasi-steady Vorticity
Quasi-steady vorticity,14 computed on the circle, is obtained as a sum of two components: (a) a free stream com-
ponent that is computed from the instantaneous angle of attack and free stream velocity, and (b) an unsteady
component that is computed from airfoil velocities.

7 of 31

American Institute of Aeronautics and Astronautics

 Zsp U∞ v∞
Rj
Ysp u ∞ βsp

X sp

)
(SP
ane
k e pl
Stro
face
l sur
rica
Cy lind

Figure 4. Coordinate system that is fixed to the stroke plane. Free stream velocity vector and cylindrical surface are also
shown.

Z sp

Stroke p
plane ((SP)) Rj Cylindrical surface
normal to the
stroke plane
Csp
Ysp
X sp

I
η
I
ξ

Figure 5. Normal cylinder described by the wing section.

Free stream component of vorticity

The vorticity and circulation due to a free stream, derived in Ref. [14], are
h i
γ f s (θ , t) = −2 u I sin(θ − α) + sin (α) (2)
Γ f s ( t) = −4πR 0 u I sin (α)

where the vorticity and circulation are positive in the counter-clockwise direction as indicated in Figure 6.
In Ref. [14], the analysis was limited to hover; therefore, it was assumed that u I ≡ 0 and Eq. (2) did not
contribute to the quasi-steady component of vorticity. In the current study, the effect of free stream velocity is
incorporated into Eq. (2) as follows.
The free stream velocity vector, and its magnitude, at each wing section are given by

UI = − u ∞ cos ϕ̃w eξ I + v∞ eη I
q
¢2
u ∞ cos ϕ̃w + v∞
2
¡
UI = kU I k = (3)

where ϕ̃ is depicted in Figure 7 and subscript w indicates wing section. The instantaneous angle of attack
of the airfoil, which is equal to the angle between the free stream velocity vector and the ξ axis as shown in
Figure 6, is:
vI v∞
µ ¶ µ ¶
α tot = α + α f s where α f s = tan−1 = tan−1 −
uI u ∞ cos ϕ̃w

8 of 31

American Institute of Aeronautics and Astronautics

Therefore, the modified expressions for vorticity and circulation due to the free stream velocity and angle of
attack are given by
h ¢i
γ f s (θ , t) = −2U I sin(θ − α − α f s ) + sin α + α f s
¡
(4)
Γ f s ( t) = −4πR 0U I sin α + α f s
¡ ¢

YSP

-u cos ~
ϕ

8
-u

8
ηnr Γ, γ are positive
η I ηfa
counter-clockwise
Wing section
α ξnr Rj or shed vortex
UI η
l
vI a
~
ϕ
αfs
b X SP
uI h

I fa
ξ ξ, ξ
Figure 6. Degrees of freedom of the airfoil and coordinate systems
used. Stroke Plane

Figure 7. Component of u ∞ normal

to the instantaneous position of the
wing. .

Unsteady component of vorticity

The derivation of the unsteady component is identical to the one presented in Ref. [21]. The expression for the
tangential velocity of the fluid on the surface of a flexible airfoil is given by
¯ f l ex 1 h 1 1 1
vθ (θ , t)¯us = − A 1 cos θ − ( A 2 + A 7 ) cos 2θ + A 3 sin θ + ( A 4 − A 5 + A 6 ) sin 2θ
R0 2 2 2
i
− A 8 sin θ cos 2θ + A 9 sin θ sin 2θ − A 10 cos θ cos 2θ + A 11 cos θ sin 2θ
f l ex
∆ vξ h i
+ τ (sin θ − sin 2θ ) + σ (cos θ − cos 2θ ) + (5)
R0
f l ex h
∆ vη i
+ − 2R 0 cos θ + τ (cos θ − cos 2θ ) − σ (sin θ − sin 2θ )
R0

and the vorticity on the circle is obtained as

f l ex ¯ f l ex Γ0
γus (θ , t) = vθ (θ , t)¯us + (6)
2πR 0

where
1 2 1 2
· µ ¶¸
Γ0 = 2π 2R 0 l̇ sin α + ḣ cos α + α̇ τ + σ − 2R 0 (R 0 + a)
¡ ¢
2 2

5. Effect of Fluid Viscosity in Shed Vorticity

In a viscous fluid, the influence of shed vorticity at a point of interest decays with increase in distance as well
as the age of the vortex. At sufficiently large Reynolds numbers, the temporal decay of the vortex strength

9 of 31

American Institute of Aeronautics and Astronautics

is slow enough compared to the time of simulation so that one may assume that the strength of the vortex is
constant with respect of time. However, this effect needs to re-examined at lower Reynolds number.
In this study, the decay of vortex strength is incorporated in the expression of induced velocity due to shed
vorticity. From Ch. 13, Ref. [28], the induced velocity due to a viscous shed vortex is given by the following
expression:

r2
à !
r2
µ ¶
− 4ν v t −R e 4 tv
¯ ¯ ¯
v ind ¯ = v ind ¯ 1− e = v ind ¯ 1− e (7)
¯ ¯ ∞ v
¯ v
viscous inviscid inviscid

where r v denotes the distance between the vortex and the point at which induced velocity is computed, t v
denotes the age of the vortex, and

Ure f c rv t vUre f
Re = ; rv = ; tv =
ν∞ c c
Incorporating this effect into the aerodynamic model modifies the constraint conditions used to determine
the shed vorticity, the expressions used to computed wake induced and bound vorticity, the evolution of the
wake, and the calculation of the pressure using the unsteady Bernoulli equation; ingredients of the model are
discussed next.

6. Determination of Shed Vorticity

shear layer 2
iy
shear layer 2
zwkv
rv
η v
ζ
rv iθ
R0e
ϕ shear layer 1
ζ θ
shear layer 1 x
ξ
Circle on
Airfoil-wake surface complex plane

Figure 8. Distances and angles that are used in the computation of velocity potential.

Following Ref. [14], the induced velocity on the circle due to an inviscid 2D vortex shed into the wake is
given as follows
z wk + R 0 e ıθ
à v !
1
¯ I
v ind (θ , t) ¯ =− R v dΓvwk (8)
¯
inviscid 2πR 0 z wk − R 0 e ıθ
wk

where, R denotes real part, wk = wk1 + wk2, and wk1 and wk2 denote shear layers emanating from the airfoil
shown in Figure 8. Combining Eqs. (7) and Eq. (8), the induced velocity becomes

I " z vwk + R 0 e ıθ
à ! ¶#
1 r2
µ
− 4ν v t
¯ ¯ ¯
v ind (θ , t) ¯ =− R dΓvwk 1− e where r v = ¯ z vwk − R 0 e ıθ ¯ (9)
¯ ∞ v
¯ ¯
viscous 2πR 0 z vwk − R 0 e ıθ
wk

where, r v is indicated in Figure 8, and the effect of viscosity is included inside the integral on the right hand
side (RHS). At each time step, the strengths of the vortices that are shed from the edges of the airfoil are
determined by imposing a Kutta condition at the TE and a stagnation condition at the LE on the circle in
the complex plane.14 These constraint conditions on induced velocity were originally derived in Ref. [14] for
inviscid flow. Including the effect of viscosity, the modified expressions are as follows.

10 of 31

American Institute of Aeronautics and Astronautics

 The stagnation condition at the LE:
I " Ã zv − R ! r2
¶#
1 1 1
· µ ¶ ¸ µ
wk 0 v − 4ν v t
A 1 − A 2 + A 7 + A 10 − 2U I sin α + α f s = R v dΓwk 1 − e
¡ ¢
∞ v (10)
R0 2 2π R 0 z wk + R 0
wk
r v = ¯ z vwk + R 0 ¯
¯ ¯
where
The Kutta condition at the TE:
I " Ã zv + R ! µ r2
¶#
wk 0 v − 4ν v t
Γ0 = − R v dΓwk 1 − e ∞ v r v = ¯ z vwk − R 0 ¯
¯ ¯
where (11)
z wk − R 0
wk

where, the effect of viscosity is included inside the integrals on the RHS of Eqs. (10) and (11). At each time
step, the latest shed vortices are placed at pre-determined distances from the LE and TE, and their inviscid
strengths are computed using Eqs. (10) and (11) as described in Ref. [14]. Subsequently, the wake induced
vorticity on the circle is given as follows:

I " Ã z v + R e ıθ ! ¶#
Γ0 1 r2
µ
0 − 4ν v t
¯
wk v
γwi (θ , t)¯ =− − R v dΓwk 1 − e ∞ v (12)
¯
viscous 2πR 0 2πR 0 z wk − R 0 e ıθ
wk
¯ ¯
where r v = ¯ z wk − R 0 e ıθ ¯
¯ v ¯

7. Calculation of the bound vorticity

The airfoil bound vorticity on the circle is obtained by combining Eqs. (4), (6), and (12) as follows.
¯
f l ex
γb = γ f s + γus + γwi ¯ (13)
¯
viscous

8. Wake evolution
Vortices on the flattened NC are convected using the Rott-Birkhoff equation,28 which is derived from Biot-
Savart law for two dimensional flow. The Rott-Birkhoff equation yields induced velocity at any point due to a
vortex and it is implemented in the ξ I − η I coordinate system as shown in Eq. (14).
∞ h dΓk i
q∗ (ζ I ) =
X
¡ I I
+ ( u E − ıv E ) (14)
k=1 2π ı ζ − ζ k
¢

where, ∞ indicates that the summation includes airfoil bound and all shed vortices in the airfoil-wake system,
and u E , vE are components of the free stream velocity at each shed vortex. The effect of free stream velocity
due to forward flight is incorporated by using
u E = − u ∞ cos ϕ̃Γk and v E = v∞ (15)
where ϕ̃Γk denotes the instantaneous sweep angle of the dΓk , as depicted in Figure 7. Therefore, Eq. (14)
yields
∞ h dΓk ¢i
q∗ (ζ I ) =
X
ϕ̃
¡
¡ I I
+ − u ∞ cos Γ k
− ıv ∞ (16)
k=1 2π ı ζ − ζ k
¢

Note that Eq. (16) becomes singular as ζ I − ζkI approaches zero. Therefore, numerical implementation of Eq.
¡ ¢

(16) requires de-singularization of the vortex core, in which each discrete vortex is assumed to have a finite
core radius. In this paper, the de-singularization procedure adopted in Ref. [13] is used. Consequently, Eq.
(16) is modified as follows.
¡ I ¢∗
ζ − ζkI
" #
∗ I
X∞ dΓk
q (ζ ) = ¢ + − u ∞ cos ϕ̃Γk − ıv∞
¡ ¢
I ∗
¡ I I ¡ I
k=1 2π ı ζ − ζ k ζ − ζ k
¢
¢∗
dΓk ζ I − ζkI
" ¡ #
X∞
+ − u ∞ cos ϕ̃Γk − ıv∞
¡ ¢
= I I 2
k=1 2π ı||ζ − ζ k ||
¢∗
dΓk ζ I − ζkI
" ¡ #
X∞
¢ + − u ∞ cos ϕ̃Γk − ıv∞
¡ ¢
≈ ¡ 2 I I 2
(17)
k=1 2π ı r c + ||ζ − ζ k ||

11 of 31

American Institute of Aeronautics and Astronautics

Incorporating the effect of viscosity, Eq. (17) yields

r2
 ¢∗   
dΓk ζ I − ζkI
¡ v
¯ ∞ − 4ν kt
q∗ (ζ I )¯
X
¢ 1 − e ∞ vk  + − u ∞ cos ϕ̃Γk − ıv∞ 
¡ ¢
≈ (18)
¯ 
I I 2
¡ 2
k=1 2π ı r c + ||ζ − ζ k ||
viscous
¯ ¯
where r vk = ¯ζ I − ζkI ¯
¯ ¯

It was noted in Ref. [15] that the use of wake sub-iterations improves quality of the solution, especially
when wake distortion due to wing-wake interaction is expected. Therefore, the Euler scheme, which was used
in Ref. [15], is used in the current study. In this scheme, the positions of the wake vortices are computed by
performing several sub-iterations within each time step. Then,
n wksubit
∆t
ζ I ( t + ∆ t) = ζ I ( t) +
X
q j ( t)
j =1 n wksubit

9. Calculation of Aerodynamic Loads using the Unsteady Bernoulli Equation

Shear layer 2

C2
swk2
fa
η
η
C1 Shear layer 1
C3 s
s0
ηI fa
ξ,ξ C4
ξ
I swk1

Figure 9. Contours of integration for computing velocity potential from bound and shed vorticity from an airfoil. For
clarity, shed vortices are indicated by solid circles.

The unsteady Bernoulli principle relates the local pressure at any point to the velocity potential at that
point. The expression derived in the ξ f a − η f a coordinate system21 is as follows:
¶2 ¶2 ¸ µ
p ∞ − p ∂φ 1 ∂φ ∂φ f a ∂φ f a ∂φ
·µ µ ¶
= + + − vξ + vη (19)
ρ∞ ∂t 2 ∂ξ f a ∂η f a ∂ξ f a ∂η f a

fa fa
where vξ and vη are the velocities of the airfoil resolved in the ξ f a − η f a coordinate system. It is important
to note that the velocity potential is discontinuous across shear layers, i.e. the wakes shed from the airfoil;
therefore, it has to be computed along piecewise continuous contours.29 Vorticity and circulation are defined as
positive when counter-clockwise; therefore, integration along the contours is performed in a counter-clockwise
manner. The origin of integration is selected to be a point close to the trailing edge on the upper surface of the
airfoil as shown in Figure 9. Furthermore, let s wk1 and s wk2 denote the arc coordinates at which the shear
layers are formed, shown in Figure 9, and these denote the location at which φ is discontinuous. For an airfoil
with a sharp trailing edge, a shear layer emanates from the trailing edge; therefore, s wk1 = L C , and the total
velocity potential at any point in the airfoil wake system is given by

φ( s, t) = φb ( s, t) + φwk1 ( s, t) (20)

For attached flow, φb is obtained by integrating the vorticity along contour C 1 shown in Figure 9. For separated
flow, φb is obtained by integrating the vorticity along contours C 1 − C 2 − C 3 . In Ref. [21], the expressions of φb
and φwk1 were obtained for inviscid flow. Incorporating the effect of viscosity, the modified expressions are as
follows:

12 of 31

American Institute of Aeronautics and Astronautics

 For 0 < s < s wk2
Zs
φb ( s, t) = φ0 + γb ds (21)
0

For s wk2 < s < L C

s wk
Z 2− Zs
r2
Z n µ ¶o
− 4ν v t
φb ( s, t) = φ0 + γb ds + γwk2 (ς, t) 1 − e ∞ v dς + γb ds (22)
0 L wk2 s wk2 +

The velocity potential due to shear layer 1 is given by

γ r2
wk1 (ς, t)
Z n µ ¶o
− v
φwk1 ( s, t) = ϕ(ς, s, t) 1 − e 4ν∞ t v dς (23)
2π
L wk1

where, in Eqs. (22) and (23), the effect of viscosity is included inside the integrals on the RHS and
¯ ¯
r v = ¯ζv, f a (ς, t) − ζ f a ( s)¯ (24)
¯ ¯

Moreover ϕ ∈ [0, 2π), depicted in Figure 8, is given by

³ ´
ϕ(ς, s, t) = argument ζv, f a (ς, t) − ζ f a ( s) (25)

C. Fluid-structure coupling

Initialize system
variables

Prescribe
rigid body motion
Equations of motion (EOM) in
integral form (Principle of virtual work)
Newton-Raphson loop
Use previously computed
Time integration loop

equilibrium configuration as Compute Compute

the reference configuration
aerodynamic flexible
loads deformation
Obtain approximate EOM
Quantities (stress, strain, displacements, etc) in
the deformed configuration are approximated
by quantities referred to the reference configuration

Convect wake

Linearize the approximate

EOM in each time step

NO
Total
simulation
Finite element approximation time?
Converts EOM to matrix form

YES

Time integration using a

suitable numerical scheme End

Figure 10. Formulation of the aeroelastic equations (left) and implementation of the aeroelastic model in MARC (right).

The equations of motion representing the aeroelastic response problem are obtained from an updated La-
grangian (UL) approach.30, 31 An approximate solution is obtained by referring all the quantities (stress, strain
and displacements) of the deformed configuration to the equilibrium configuration obtained in the previous
time step, and linearizing the resulting equations of motion (EOM). Implementation of the UL formulation in

13 of 31

American Institute of Aeronautics and Astronautics

MARC30, 31 is illustrated by Figure 10 and was summarized in Ref. [20]. The implementation of the coupled
problem is shown in the block diagram given in Figure 10. At each time step, rigid body motion is prescribed
as displacements at specified nodes. The aerodynamic loads, computed based on the wing motion at the begin-
ning of each time step, are applied to the structure via FORCEM, a user defined subroutine in MARC. This
subroutine is called from the main program for each step of the Newton-Raphson iteration within a time step
to ensure convergence of the structural displacements for the applied loads. Finally, the vortices shed into the
wake are convected at the end of the time step.

III. Results and Discussion

The results are presented for rigid and flexible configurations: (1) aerodynamic comparisons for rigid air-
foils and wings in hover and forward flight, and (2) aeroelastic results for anisotropic wings in hover and
forward flight.

10
Original
Filtered
CL

−10
0 1 2 3 4 5
Number of cycles

Figure 11. Sample comparison of original and filtered load signals

When using the approximate aerodynamic model, leading edge separation is assumed unless specified
otherwise and the transient aerodynamic loads are obtained using the unsteady Bernoulli equation. A previ-
ous study21 noted that the interaction of the airfoil with previously shed vortices generated large amplitude
numerical oscillations in the aerodynamic loads that had to be eliminated before the unsteady loads can be
applied on the flexible structure in aeroelastic simulations. The numerical oscillations are not critical to sim-
ulations involving rigid airfoils and wings; however, smoothing the signals improves clarity when comparing
the time histories of the aerodynamic loads with CFD based results. In this paper, the load signals, which
are obtained from the approximate aerodynamic model for rigid cases as well as the flexible cases, are post-
processed using zero-phase digital filters available in MATLAB® (version 8.0). The filters are implemented in
MATLAB® using the filtfilt command as shown in Eq (26). A sample comparison of the original and smoothed
signals is shown in Figure 11. ¡ ¢
y f iltered = filtfilt B f , A f , yun f iltered (26)
¡ ¢
where B f and A f are filter coefficients based on Chebyshev functions that are arrays of length m f + 1 . The
implementation of the zero-phase digital filters may be obtained from the extensive documentation provided
in MATLAB® .

A. Aerodynamic Comparisons
Results are presented for the following cases: (1) flat plate airfoils undergoing prescribed motion, (2) rigid
Zimmerman wings in hover, and (3) rigid Zimmerman wings in forward flight.

1. Airfoil Cases
The forces generated by airfoils undergoing prescribed motion are presented in Figures 12 and 13 for the case
of hover, whereas Figure 14 includes a free stream velocity and therefore corresponds to the case of forward
flight. The lift and drag are components of the aerodynamic force along the η I and ξ I axes respectively. The
airfoil kinematics are described by Eq. (27) and the parameters considered are shown in Table 1.
l ( t) = l 0 sin (2 π f t)
π
α( t) + α0 sin 2 π f t + ϕα
¡ ¢
= (27)
2

14 of 31

American Institute of Aeronautics and Astronautics

 Case ID Name f , Hz l0 α0 ϕα
1 π π
1 Delayed rotation 2π c 4 3
1 4π π
12 Synchronized rotation 3π 1. 5 c 9 2

Table 1. Amplitudes and phase for airfoil kinematics. The case ids are obtained from Ref. [32]

where, the pitching is about the mid-chord. The results are obtained for R e = 100, c = 1m, ρ ∞ = 1 kg/m3 ,
ν∞ = 0.01 m2 /s, Ure f = 1.0 m/s, where R e and Ure f are defined as follows.

Ure f c
Re = and Ure f = 2π f l 0 (28)
ν∞

The non-dimensional forces are defined as follows:

L D
CL = 1 2
and CD = 1 2
(29)
2 ρ ∞U re f 2 ρ ∞U re f c

The CFD based results are taken from Ref. [32]. The approximate results were obtained for the following
parameters: Nθ = 200, r c = 0.1 c, n wksubit = 4, and circulation limit21 was fixed at 2.0. For both cases, CFD
simulations indicated that there was leading edge separation. However, for case 12, the vorticity shed from the
LE was substantially weaker when compared to that shed from the TE. Therefore, the approximate model was
also used with the attached flow assumption. The simulations using the approximate model were conducted
assuming both inviscid as well as viscous flow.

2 2
CL

CL

0 0

−2 −2

0 1 2 Approx3 4 5 0 1 2 Approx
3 4 5
CFD CFD

5 5
CD

CD

0 0

−5 −5
0 1 2 3 4 5 0 1 2 3 4 5
Number of cycles Number of cycles

(a) Inviscid separated flow (b) Separated flow with viscous effects

Figure 12. Force coefficients for Case 1

The force coefficients obtained for case 1 are shown in Figure 12; results obtained by assuming inviscid
flow and incorporating viscous effects are shown in Figures 12(a) and 12(b) respectively. Simulations using
the approximate model were coducted using 500 time steps per cycle, and the load signals were smoothed using
the filter described in Table 2. The comparisons illustrate that incorporating the effect of viscosity improves
correlation with CFD based results; in particular, the improvement is noticeable for C D . These results suggest
that the influence of fluid viscosity in the interactions involving shed vorticity is important for this case.
The force coefficients obtained for case 12 are shown in Figure 13. Simulations assuming separated and
attached flows were conducted by using 800 and 500 time steps per cycle respectively, and the loads obtained
were smoothed using filters described in Tables 2 and 3 respectively. The results for separated flow, shown
in Figures 13(a) and 13(b), indicate that the inviscid approximate model does not compare well with CFD for
this case. Incorporating viscous effects improves correlation slightly, as shown in Figure 13(b). However, the

15 of 31

American Institute of Aeronautics and Astronautics

 2 2
CL

CL
0 0

−2 −2
0 1 2 Approx
3 4 5 0 1 2 Approx
3 4 5
CFD CFD

2 2
CD

CD
0 0

−2 −2

0 1 2 3 4 5 0 1 2 3 4 5
Number of cycles Number of cycles

(a) Inviscid separated flow (b) Separated flow with viscous effects

2 2
CL

CL
0 0

−2 −2
0 1 2 Approx
3 4 5 0 1 2 Approx
3 4 5
CFD CFD

2 2
CD

CD

0 0

−2 −2

0 1 2 3 4 5 0 1 2 3 4 5
Number of cycles Number of cycles

(c) Inviscid attached flow (d) Attached flow with viscous effects

Figure 13. Force coefficients for Case 12

improvement is most noticeable in the first and second cycles of C D . Results obtained by assuming attached
flow, shown in Figures 13(c) and 13(d), indicate that the approximate results show reasonable correlation with
CFD based results; however some of the peaks in C D are not adequately captured. Incorporating viscous
effects improves correlation slightly; however, the difference is small. These results imply that when the
vorticity generated from the LE is weak, the principal contribution to the aerodynamic loads is from the TE
vortices.
The forces generated by the airfoils in the presence of a free stream are shown in Figure 14. The kinematics
employed are the same as those for cases 1 and 12, where a free stream velocity vector that is parallel to ξ I is
now introduced; the free steam velocity is equal to 0.2Ure f , where Ure f is given in Eq. (29). Simulations with
the approximate model were carried out assuming separated flow for case 1, attached flow for case 12, and
including effect of viscosity. The results were obtained using 500 time steps per cycle, and smoothed using the
filter described in Table 2. Comparisons indicate that the approximate model produces reasonable agreement
with CFD for the cases considered. In particular, comparisons for case 12, shown in Figure 14(b), indicate that
the discrepancy in the peaks of C D is larger when compared to the discrepancy observed for case of hover; this
may be attributed to the increased contribution of vorticity shed from the LE.

16 of 31

American Institute of Aeronautics and Astronautics

 L 2 2

CL
0
C

0
−2
−2
0 1 2 Approx
3 4 5 0 1 2 Approx
3 4 5
CFD CFD

5 2
D

CD
0 0
C

−5 −2

0 1 2 3 4 5 0 1 2 3 4 5
Number of cycles Number of cycles

(a) Case 1: separated flow with viscous effects (b) Case 12: attached flow with viscous effects

Figure 14. Force coefficients for Cases 1 and 12 with free stream

Af 1.0000 -3.5379 4.7367 -2.8414 0.6439

−3
B f (×10 ) 0.0797 0.3186 0.4779 0.3186 0.0797

Table 2. Filter coefficients

2. Rigid Wings in Hover

The results for the rigid wings are based on a Zimmerman planform, shown in Figure 2, that have an aspect
ratio of 7.65, c r = 25 mm, R span = 75 mm respectively.33 The CFD simulations were conducted using the
numerical framework described in Ref. [34, 35] that solves the finite-volume-based Navier- Stokes equations
using a pressure-based algorithm. The aerodynamic loads obtained using the approximate aerodynamic model
were computed for the following set of parameters: Nsections = 59, Nθ = 100, r c = 0.1 × chord , and n wksubit = 4
respectively. The numerical experiments were conducted in air (ρ ∞ = 1.209 kg/m3 , ν∞ = 1.568 × 10−5 m2 /s).
The lift and thrust generated by rigid wings are shown in Figures 15, 16, and 17. The lift and thrust are
the components of the aerodynamic force resolved along YSP and Z SP , shown in Figure 4, respectively. The
corresponding non-dimensional quantities are defined in Eq. (30).
L T
CL = 1 2
and CT = 1 2
where Ure f = 4πβ0 f (30)
2 ρ ∞U re f R span c r 2 ρ ∞U re f R span c r

Furthermore, X W and YW , shown in Figure 2, coincide with X SP and Z SP , shown in Figure 5, respectively
at the start of the motion. The flapping motion, which corresponds to a simple rotation about YW , given by Eq.
(31).

β( t) = β0 sin(2π f t) (31)
The non-dimensional lift and thrust for β0 = 35◦ and various values of f are shown in Figures 15 and 16.
Simulations using the approximate and CFD based models were conducted using 300 and 500 time steps per
flapping cycle to discretize the motion respectively. Subsequently, a fourth order filter ( m f = 4) described in
Table 4 was used to smooth the signals obtained using the approximate model.
Figure 15 shows the force coefficients obtained for β0 = 35◦ and f = 10 Hz. The result indicates that the
approximate model over-predicts the lift compared to the CFD based result; the mean and maximum errors
in the peaks are 50% and 100% respectively. The discrepancy may be attributed to the spanwise flow and tip
vortices that may be important for the combination of amplitude and flapping frequencies that are considered,
which are not incorporated in the approximate aerodynamic model. Due to the lack of wing pitch or twist, the
thrust corresponds to aerodynamic force parallel to the chordwise direction of the wing. In the approximate
computations, the wing is assumed to have zero thickness; consequently, the force parallel to the wing chord is

17 of 31

American Institute of Aeronautics and Astronautics

 Af 1.0000 -3.9109 5.7376 -3.7423 0.9156
B f (×10−6 ) 0.1517 0.6067 0.9101 0.6067 0.1517

Table 3. Filter coefficients

−3
x 10
4
2
0

CT
−2
−4
−6
0 1 2 CFD 3 4 5
Approx
5
CL

−5
0 1 2 3 4 5
Number of cycles

Figure 15. Force coefficients generated by rigid wings for 35◦ and f = 10 Hz

equal to zero. In the CFD based computations, this quantity is equal to the sum of viscous forces on the wing
surface and the suction pressure on the edges of the wing; the result in Figure 15 indicates that the thrust
obtained from CFD based calculations is several orders of magnitude lower than the lift.
The time histories of C L for all the flapping frequencies considered is shown in Figure 16, wherein the CFD
based and approximate results are shown on separate plots. The thrust was several orders of magnitude lower
than lift; consequently comparisons involving this quantity are not shown. It is interesting to note that the
lift coefficients, computed for various frequencies, are very similar. This trend indicates that C L is somewhat
insensitive to Reynolds number. Furthermore, based on the definition of C L in Eq. (30), it follows that the
peak lift is proportional to the square of the flapping frequency. Similar trends are obtained from simulations
using the approximate model for β0 = 18◦ , as shown in Figure 17, indicating that the relation between the lift
and flapping frequency is insensitive to flapping amplitude.

Af 1 -3.0456 3.6243 -1.9780 0.4158

Bf 0.0010 0.0041 0.0062 0.0041 0.0010

Table 4. Filter coefficients

The forces generated by rigid wings undergoing prescribed combined flap-pitch motion are shown in Figure
18. The wing kinematics are described by the following Euler rotations: a flapping rotation about YW described
by Eq. (31) followed by a feathering motion about X W described by Eq. (32); where, X W , indicated in Figure
2, originates at the quarter-chord point of the root. The parameters for which the results are obtained are
summarized in Table 5. The CFD and approximate simulations were conducted by using 500 and 200 time
steps per flapping cycle respectively. The results, in Figure 18, indicate that the approximate model shows
reasonable correlation with CFD for the cases considered.

α( t) = −α0 sin(2π f t) (32)

3. Rigid Wings in Forward Flight

The forces generated by rigid wings in forward flight, shown in Figures 20 and 21, are obtained for combi-
nations of µ and βsp that are based on experimentally obtained data for hawkmoths in forward flight.27 The

18 of 31

American Institute of Aeronautics and Astronautics

 CFD
5

CL
0

−5 10 Hz
0 1 2 3 4
20 Hz 5

Approx 30 Hz
5 40 Hz
CL

−5
0 1 2 3 4 5
Number of cycles

Figure 16. Lift coefficients generated by rigid wings for β0 = 35◦ .

Figure Number Flap amplitude Pitch amplitude f

◦ ◦
18(a) 10 5 10 Hz
18(b) 15◦ 5◦ 10 Hz
◦ ◦
18(c) 15 10 10 Hz

Table 5. Force coefficients generated by wings undergoing combined flap-pitch rotation.

specific combinations used in the current study, which are based on the data obtained for Moth M1 in Ref. [27],
are shown in Figure 19 and listed in Table 6. The components of free stream velocity are calculated using Eq.
(1) where U ti p = 2πβ0 f R span .

µ 0 0.05 0.10 0.15 0.20 0.25

◦ ◦ ◦ ◦ ◦
βsp 0 14 21.2 27.7 34 40◦

Table 6. Combination of advance ratios and stroke plane inclinations used in the current study

The non-dimensional lift generated by rigid wings for β0 = 35◦ , f = 10 Hz, and various advance ratios, are
shown in Figure 20. The thrust generated was several orders of magnitude lower than lift; therefore results
for this quantity are not presented. The CFD and approximate simulations were conducted by using 500 and
300 time steps per flapping cycle respectively; results obtained using the approximate model were smoothed
using the filter described in Table 4. The comparisons indicate that the approximate model shows reasonable
agreement with CFD based results for the cases considered. The correlation between the results improves
with increasing advance ratio. The mean lift coefficients, which represent the time averaged values of C L , are
shown in Figure 21. This result shows that the mean lift decreases (becomes more negative) with increasing
forward flight speed, and that the approximate model over-predicts the value but captures the trend. The
results shown in Figures 20 and 21 demonstrate that the modified aerodynamic model can be used to conduct
trend type studies for wings in forward flight despite the simplifying assumptions in the formulation.

19 of 31

American Institute of Aeronautics and Astronautics

 −7
x 10
1

0.5

CT
0

−0.5

−1 10 Hz
0 1 2 203Hz 4 5
30 Hz
2 40 Hz

1
CL

−1

−2
0 1 2 3 4 5
Number of cycles

Figure 17. Lift and thrust coefficients, calculated using the approximate model, generated by rigid wings for β0 = 18◦ .

0.2
0 0 0
T

CT

CT
−0.2 −0.2
C

−0.1
−0.4 −0.4
−0.2
0 1 2 Approx
3 4 5 0 1 2 Approx
3 4 5 0 1 2 Approx
3 4 5
CFD CFD CFD
2 5 5
CL

CL

0 0 0
C

−2 −5 −5
0 1 2 3 4 5 0 1 2 3 4 5 0 1 2 3 4 5
Number of cycles Number of cycles Number of cycles

(a) β0 = 10◦ , α0 = 5◦ (b) β0 = 15◦ , α0 = 5◦ (c) β0 = 15◦ , α0 = 10◦

Figure 18. Force coefficients generated by a rigid wing undergoing combined pitch-flap motion.

B. Aeroelastic Results
Aeroelastic simulations were conducted using anisotropic wings that are based on an unstressed CAPRAN film
(membrane) that is supported by a carbon fiber based spar-batten skeleton. Following the notation used in Ref.
[33], the wings are labeled as LiBj where i and j denote the number of prepreg layers in the LE spar and the
battens, as shown in Figure 22. Structural dynamic modeling of the anisotropic wings in MARC was presented
in Ref. [16]. The finite element models of the wings, shown in Figure 2, are composed of 1263 shell elements
(Element type 75 in MARC) that are capable of modeling large rigid body motion as well as moderate-to-large
flexible deformation. The geometric and material properties of the prepreg and the membrane16 are provided
in Table 7. Wing kinematics are implemented as displacement boundary conditions at the nodes that form the
triangular root. Furthermore, the EOM obtained using the UL method are integrated forward in time using a
single step Houbolt numerical scheme.36
During the simulations, a pressure based filter that is described by Eq. (33) was used to limit the magni-
tude of numerical noise that is transmitted to the flexible wing; note that the limit should be sufficiently high

20 of 31

American Institute of Aeronautics and Astronautics

 60
Expt
Data used in current study
50

SP inclination, in degrees
40

30

20

10

0
0 0.05 0.1 0.15 0.2 0.25 0.3
Advance ratio

Figure 19. Combinations of advance ratio and SP inclination: Expt - Experimental data.27

so as not to introduce significant errors during calculations.


 − p l imit
 if p cal culated ≤ − p l imit
p ap pl ied = p cal culated if − p l imit ≤ p cal culated ≤ p l imit (33)

p l imit if p cal culated ≥ p l imit


For all the cases considered in this study, a limiting pressure given by p l imit = 36 p re f , where p re f as defined
in Eq. (34), was found to be adequate.

1 2
p re f = ρ ∞U ti p where U ti p = 2π f β0 R span (34)
2

Material Properties
Carbon fiber prepreg E 11 = 233 GPa
E 22 = 23.1 GPa
(Properties of one layer) E 12 = 15.5 GPa (L1B1, L1B2)
E 12 = 10.5 GPa (all other configs)
ν12 = 0.05
ρ = 1740 kg/m3
Thickness = 0.1 mm
Capran membrane E = 2.76 GPa
(From experiments) ν12 = 0.489 (Incompressible)
ρ = 1384 kg/m3
Thickness = 15 microns

Table 7. Material properties of the composite and membrane

1. Anisotropic Wings in Hover

The computed and experimentally measured magnitudes of thrust generated by anisotropic wings are com-
pared. The results, which are obtained for one-layer batten configurations and β0 = 35 o , for a range of flapping
frequencies, are shown in Figures 23 and 24. The experimental results were obtained from Ref. [33] and com-
putations are based on the approximate aeroelastic model. Note that the experiments were conducted using
two wings undergoing symmetric flapping motion, whereas the computations were performed by considering
a single wing; therefore, the thrust obtained from the simulations was multiplied by a factor of two to allow
comparison with the experimental result. For the approximate model, 300 time steps per flapping cycle were
used to discretize the motion. The simulations, conducted for a total of 5 cycles, indicated that an approximate

21 of 31

American Institute of Aeronautics and Astronautics

steady state was achieved after two cycles. Therefore, the mean thrust was calculated by time averaging the
thrust generated over cycles 2 through 5.
The thrust predicted using the approximate aeroelastic model shows reasonable agreement with experi-
mentally measured thrust for all the cases considered. An accurate assessment of the error is complicated
because due to the limited amount of experimental data available, the sensitivity of the thrust to variability
in material, geometric properties, and wing construction cannot be determined. Therefore, the effect of dif-
ferences in structural dynamic modeling on the thrust generated has to be determined before quantifying the
differences in the aeroelastic results.
Figure 24 shows the thrust generated by one-layer batten configurations. These results indicate that the
approximate model predicts the trends that are observed from the experimental measurements. Different
wing configurations produce maximum thrust at different ranges of flapping frequencies. Configuration L1B1
produces maximum thrust at frequencies between 5 Hz and 18 Hz, L2B1 for frequencies between 18 Hz and
35 Hz, and L3B1 for frequencies between 35 Hz and 40 Hz respectively.
The thrust generated by L1Bj configurations is shown in Figure 23(a). Experimental measurements as
well as the computations for L1B1 show a peak in thrust for a flapping frequency in the range of 20 − 25 Hz;
the computations for L1B2 show a similar peak. In Ref. [16], the fundamental frequency of L1B1 was found
to be approximately 21 Hz. Therefore, it appears that the peaks in Figure 23(a) occur when the frequency of
wing excitation is close to the fundamental frequency of the wing. Additional simulations using L1B1 were
conducted in order to examine the dependence of the peak in thrust on the flapping amplitude. The results,
shown in Figure 25, confirm that a peak in thrust is obtained when the excitation frequency is close to the
natural frequency of the wing for all the amplitudes considered. This is an observation that might be valuable
for wing design and selection of wing kinematics.
The aerodynamic and inertia loads acting on the wings were compared. Sample results for a limited
number of cases are shown in Figures 26 and 27. The aerodynamic and inertia loads are calculated using Eq.
(35). These results, which resemble similar comparisons for other cases, indicate that the aerodynamic loads
are comparable to inertia loads. This is contrary to the observations of Refs. [10, 17, 18] and suggests that
the relative importance of aerodynamic to inertia loads is configuration dependent, thereby highlighting the
importance of aeroelastic effects in flapping wings.
¯ Ï ¯ Ñ
F sp ¯ = p d A sp and F sp ¯ = dm ẍ sp (35)
¯ ¯
aero inertia
Aw Vw

2. Wings undergoing insect-like kinematics for the case of hover

Results were obtained for rigid and flexible wings that are actuated using insect-like kinematics. The kinemat-
ics, based on the hover kinematics of hawkmoths,27 are obtained by scaling the stroke amplitudes by a factor
of two. The time dependent rotations, prescribed about the Yw , X w , and Z w axes respectively, are described by
the Euler angles in Eq. (36) and shown in Figure 28.

Φ0 1 X 3 £
Φ r ( t) Φ c j cos(2π j f t) + Φs j sin(2π j f t)
¤
= +
4 2 j=0
α0 1X 3 £
α r ( t) α c j cos(2π j f t) + αs j sin(2π j f t)
¤
= − − (36)
4 2 j=0
Θ0 1 X 3 £
Θ r ( t) Θ c j cos(2π j f t) + Θs j sin(2π j f t)
¤
= +
4 2 j=0

where αr1 and Θr are prescribed about the quarter-chord point at the root. The fourier coefficients of the Euler
angles are given in Table 8. The simulations were conducted using 300 time steps per flapping cycle, and the
results obtained from the aeroelastic model were smoothed using a filter described by Eq. (2).
The lift and thrust, in grams, generated by rigid and anisotropic wings for f = 10 Hz, and f = 20 Hz, are
shown in Figure 29. The results correspond to the forces generated by an isolated flapping wing. For f = 10
Hz, L1B1 generates higher thrust compared to the other wings, as shown in Figure 29(a); however, the lift
generated by all the configurations is somewhat similar. For f = 20 Hz the results in Figure 29(b) indicate
that L1B1 generates less thrust and lift compared to L3B1 and the rigid wing whereas L3B1 produces the

22 of 31

American Institute of Aeronautics and Astronautics

 Φ0 Φ c1 Φ c2 Φ c3 Φ s1 Φ s2 Φ s3
8.1074 52.9298 0.5959 2.2231 -6.9385 -2.6643 1.5011

Θ0 Θ c1 Θ c2 Θ c3 Θ s1 Θ s2 Θ s3
-1.9395 1.4152 6.1994 0.4469 2.4752 0.5730 -0.7162

α0 α c1 α c2 α c3 α s1 α s2 α s3
9.6698 -14.9886 0.5099 -0.9282 -58.9230 -0.2521 -7.6834

Table 8. Fourier coefficients in the hawkmoth kinematics. Values are given in degrees.

maximum thrust. Recall that for this frequency, the L1B1 and L3B1 configurations generated the maximum
and minimum thrust when using a simple flap actuation. This suggests that the impact of wing flexibility may
be kinematics dependent.

3. Anisotropic Wings in Forward Flight

The effect of forward flight on the aerodynamic forces generated by flexible wings is examined using anisotropic
wings undergoing prescribed flap motion for a range of forward flight conditions. The results, shown in Figures
30 and 31, are obtained for L1B1 and L3B1, β0 = 35◦ , f = {10, 40} Hz, and the forward flight conditions corre-
sponding to the values given in Table 6. The flapping motion is described by Eq. (31) and the magnitude of the
free stream velocities were determined based on the tip speed of a rigid wing undergoing the same kinematics.
The mean forces are obtained by time averaging the transient forces over cycles 2 through 5. The mean forces
computed for one wing were multiplied by a factor of two so as to approximate the force generated by a vehicle
that has a pair of wings.
The mean lift and thrust generated by rigid and flexible wings are shown in Figure 30. The results indi-
cate that the lift decreases (becomes more negative) with increase in forward flight speed; but the change in
thrust is somewhat small. Also, different wings produce maximum mean thrust at different frequencies: L1B1
produces maximum thrust at f = 10 Hz, whereas L3B1 produces the maximum thrust at f = 40 Hz; similar
trends were obtained for the case of hover.
The mean horizontal and vertical forces generated by the wings, indicative of the propulsive and payload
capacity of the wings, are shown in Figure 31. The horizontal and vertical directions are shown in Figure
3. Note that both F h and Fv are positive in an actual vehicle; a negative value of F h denotes drag. The
results show that Fv increases and F h decreases with increase in forward flight speed, and wing flexibility
has a beneficial influence. The flexible configurations have higher payload capacity and lower drag than rigid
wings. Configurations L1B1 and L3B1 have the largest payload capacity and least drag at 10 Hz and 40 Hz
respectively. These results indicate that the trends in force generation obtained for wings in hover also apply
to forward flight.
Of the various wings and flapping frequencies considered in this study, positive values of both F h and Fv
were obtained only in the case of L3B1 at f = 40 Hz, as shown in Figure 31(b). Thus, this case seems to
be the most effective combination of wing flexibility and kinematics encountered in this study. The payload
capacity for this case, which is 12 grams, is inadequate to support the expected weight of a potential vehicle
(approximately 50 grams). This implies that further exploration of the parameter space is needed. Note that
typical insect kinematics include a large pitching motion during and at the ends of each flapping stroke. In
the current study the wing pitch was due to wing torsion. Therefore, the interaction of wing flexibility and
kinematics that include active pitching of the wing has to be explored in greater detail.

23 of 31

American Institute of Aeronautics and Astronautics

 4 4
Approx Approx
3 CFD 3 CFD

2 2

1 1
L

CL
0 0
C

−1 −1

−2 −2

−3 −3

−4 −4
0 1 2 3 4 5 0 1 2 3 4 5
Number of cycles Number of cycles

(a) µ = 0.0 (b) µ = 0.05

4 4
Approx Approx
3 CFD 3 CFD

2 2

1 1
L

CL

0 0
C

−1 −1

−2 −2

−3 −3

−4 −4
0 1 2 3 4 5 0 1 2 3 4 5
Number of cycles Number of cycles

(c) µ = 0.10 (d) µ = 0.15

4 4
Approx Approx
3 CFD 3 CFD

2 2

1 1
L

CL

0 0
C

−1 −1

−2 −2

−3 −3

−4 −4
0 1 2 3 4 5 0 1 2 3 4 5
Number of cycles Number of cycles

(e) µ = 0.20 (f) µ = 0.25

Figure 20. Force coefficients generated by rigid wings in forward flight: β0 = 35◦ , f = 10 Hz

24 of 31

American Institute of Aeronautics and Astronautics

 CFD
0 Approx
−0.05

Time averaged CL
−0.1

−0.15

−0.2

−0.25

−0.3

−0.35

−0.4
0 0.05 0.1 0.15 0.2 0.25
Advance ratio

Figure 21. Mean lift generated by rigid wings in forward flight.

Figure 22. Anisotropic wing configurations, from Ref [33].

7
L2B1 Expt L3B1 Expt
1 6 L2B1 Comp L3B1 Comp
8
0.8 5
6
Thrust, g

4
0.6
L1B1 Expt 3 4
0.4 L1B1 Comp
2
0.2 2
1
0
5 10 15 20 25 30 35 40 5 10 15 20 25 30 35 40 5 10 15 20 25 30 35 40
Frequency, Hz Frequency, Hz Frequency, Hz

(a) L1B1 (b) L2B1 (c) L3B1

Figure 23. Comparison of thrust generated by anisotropic wings: ‘Expt’ - Experiments,33 ‘Comp’ - computations, current
study

25 of 31

American Institute of Aeronautics and Astronautics

 Experiments Computations
10 10

8 8

6 6
Thrust, g

L1B1
4 L2B1
4
L3B1

2 2

0 0
10 20 30 40 10 20 30 40
Frequency, Hz Frequency, Hz

Figure 24. Thrust generated by one-layer batten configurations: Experiments,33 Computations - current study

1.4
β0 = 9°
1.2 β0 = 18°

1 β0 = 35°
β0 = 35°
0.8
Thrust, g

0.6

0.4

0.2

−0.2
5 10 15 20 25 30 35 40
Flapping frequency, Hz

Figure 25. Thrust generated by L1B1 for various flapping amplitudes: Dashed line - Experiments,33 Solid lines - computa-
tions, current study. The vertical green line indicates the natural frequency of L1B1

26 of 31

American Institute of Aeronautics and Astronautics

 L1B1, 10 Hz L1B1, 40 Hz
1 5
FX , in g

FX , in g
0 0
SP

SP
−1 −5

−2 −10
0 1 2 3 Aero 4 0 1 2 3 Aero 4
Inertia Inertia
2 10
FY , in g

FY , in g
0 0
SP

SP
−2 −10
0 1 2 3 4 0 1 2 3 4

1 10
FZ , in g

FZ , in g
0.5 5
SP

SP
0 0

−0.5 −5
0 1 2 3 4 0 1 2 3 4
Number of cycles Number of cycles

(a) f = 10 Hz (b) f = 40 Hz

Figure 26. Aerodynamic and inertia loads acting on L1B1 in hover: β0 = 35◦

L3B1, 10 Hz L3B1, 40 Hz
1 20
FX , in g

FX , in g

0
0
SP

SP

−1

−2 −20
0 1 2 3 Aero 4 0 1 2 3 Aero 4
Inertia Inertia
5 50
FY , in g

FY , in g

0 0
SP

SP

−5 −50
0 1 2 3 4 0 1 2 3 4

1 40
FZ , in g

FZ , in g

0.5 20
SP

SP

0 0

−0.5 −20
0 1 2 3 4 0 1 2 3 4
Number of cycles Number of cycles

(a) f = 10 Hz (b) f = 40 Hz

Figure 27. Aerodynamic and inertia loads acting on L3B1 in hover: β0 = 35◦

27 of 31

American Institute of Aeronautics and Astronautics

 Hawkmoth Kinematics
150
Φr
αr
100
Θr

Angles, degrees
50

−50

−100

−150
0 0.2 0.4 0.6 0.8 1
Number of cycles

Figure 28. Hawkmoth kinematics: thin lines denote actual kinematics;27 thick lines denote scaled kinematics that are
used in the current study.

f = 10 Hz f = 20 Hz

Rigid 2 Rigid
0.5
L3B1 L3B1
0.4 L1B1 1.5 L1B1
Thrust, g

Thrust, g

0.3
1
0.2
0.5
0.1
0 0

2
0.4
0.2 1
Lift, g

Lift, g

0 0
−0.2
−1
−0.4
−2
−0.6
0 0.5 1 1.5 2 2.5 3 3.5 4 0 0.5 1 1.5 2 2.5 3 3.5 4
Number of cycles Number of cycles

(a) (b)

Figure 29. Lift and thrust generated by rigid and flexible wings undergoing scaled hawkmoth kinematics

28 of 31

American Institute of Aeronautics and Astronautics

 f = 10 Hz f = 40 Hz
0.4 15

0.3
10
mean thrust, g

mean thrust, g
0.2
5
0.1
0
0
L1B1 L1B1
−0.1 L3B1 −5
L3B1
0.2 Rigid Approx 5 Rigid
Rigid CFD
0
0
mean lift, g

mean lift, g
−0.2
−5
−0.4
−10
−0.6

−0.8 −15
0 0.05 0.1 0.15 0.2 0.25 0 0.05 0.1 0.15 0.2 0.25
Advance ratio Advance ratio

(a) (b)

Figure 30. Mean lift and thrust, in grams, generated by rigid and flexible wings

f = 10 Hz f = 40 Hz
0.2 5

0 0
mean Fh, g

mean Fh, g

−0.2 −5

−0.4 −10

L1B1
−0.6 −15 L1B1
L3B1
L3B1
0.8 Rigid Approx 15 Rigid
Rigid CFD
0.6
10
mean Fv, g

mean Fv, g

0.4

0.2
5
0

−0.2 0
0 0.05 0.1 0.15 0.2 0.25 0 0.05 0.1 0.15 0.2 0.25
Advance ratio Advance ratio

(a) (b)

Figure 31. Mean horizontal and vertical forces, in grams, generated by rigid and flexible wings

29 of 31

American Institute of Aeronautics and Astronautics

 IV. Concluding Remarks
The approximate aerodynamic model for flapping wings, originally developed for hover, has been extended
to forward flight. The effect of viscosity was also incorporated in an approximate manner by including the
temporal decay of shed vorticity into the calculation of induced velocity. The results were obtained for rigid
airfoils, wings, and aeroelastic behavior of anisotropic wings.

1. Comparisons of forces generated by airfoils undergoing prescribed motions at R e = 100 show that incor-
porating the effect of viscosity in the approximate model improved correlation with CFD based results.

2. The forces computed using the modified approximate aerodynamic model were compared with those ob-
tained from CFD simulations for rigid Zimmerman wings in hover and forward flight. The approximate
model shows reasonable agreement with CFD results and it captures the trends accurately.

3. Forces computed from the approximate aeroelastic model were compared with those obtained from exper-
iments conducted on several Zimmerman wing configurations in hover. The approximate model shows
reasonable agreement for the cases considered. In particular, a peak in thrust was obtained when the
excitation frequency was close to the natural frequency of the wing, and the location of the peak is
independent of the flapping amplitude.

4. Comparisons of inertia and aerodynamic loads acting on the wings indicate that the aerodynamic loads
are comparable to inertia loads for the cases considered. This is contrary to what was found in previous
studies. Therefore, it appears that the relative importance of aerodynamic and inertia loads in flapping
wings is dependent on the configurations considered.

5. The payload capacity and propulsive capability of rigid and flexible wings in hover and forward flight
were examined, and it was found that flexible wings have larger payload capacity and lower drag com-
pared to rigid wings for the cases considered. Different flexible configurations perform better at differ-
ent flapping frequencies. Thus, the choice of the ‘best’ flexible configuration depends on the flapping
frequency. This behavior was also noted in hover.

6. Results for rigid and flexible wings undergoing insect-like kinematics suggest that the impact of wing
flexibility on the force generation capacity of flapping wings may be dependent on the kinematics used.

Acknowledgments
This work was supported by the Air Force Office of Scientific Research’s (AFOSR) MURI with Dr. Douglas
R. Smith as Program Director. The first author would like to thank Patrick Trizila and Dr. Hikaru Aono for
generating the CFD based results for the rigid airfoils and wings respectively.

References
1 Mueller, T. J., Fixed and Flapping Wing Aerodynamics for Micro Air Vehicle Applications, Vol. 195, Progress in Aeronautics and
Astronautics, published by AIAA, 2001.
2 Platzer, M. E. and Jones, K., “Flapping Wing Aerodynamics - Progress and Challenges,” 44 th AIAA Aerospace Sciences Meeting and
Exhibit, Reno, Nevada, January 2006, AIAA Paper Number 2006-500.
3 Sane, S. P., “The Aerodynamics of Insect Flight,” The Journal of Experimental Biology, Vol. 206, 2003, pp. 4191 – 4208.
4 Ansari, S. A., Żbikowski, R., and Knowles, K., “Aerodynamic Modelling of Insect-like Flapping Flight for Micro Air Vehicles,”
Progress in Aerospace Sciences, Vol. 42, 2006, pp. 129 – 172.
5 Shyy, W., Lian, Y., Tang, J., Liu, H., Trizila, P., Stanford, B., Bernal, L., Cesnik, C., Friedmann, P., and Ifju, P., “Computational
Aerodynamics of Low Reynolds Number Plunging, Pitching, ad Flexible Wings for MAV Applications,” 48 th AIAA Aerospace Sciences
Meeting and Exhibit, Reno, Nevada, January 2008, AIAA Paper No. 2008-523.
6 Shyy, W., Lian, Y., Tang, J., Viieru, D., and Liu, H., Aerodynamics of Low Reynolds Number Flyers, Cambridge University Press,
2008.
7 Shyy, W., Aono, H., Chimakurthi, S., Trizila, P., Kang, C.-K., Cesnik, C., and Liu, H., “Recent Progress in Flapping Wing Aerody-
namics and Aeroelasticity,” Progress in Aerospace Sciences, Vol. 46, No. 7, 2010, pp. 284 – 327.
8 Baik, Y. S., Rausch, J. M., Bernal, L. P., Shyy, W., and Ol, M. V., “Experimental Study of Governing Parameters in Pitching and
Plunging Airfoil at Low Reynolds Number,” 48th AIAA Aerospace Sciences Meeting Including the New Horizons Forum and Aerospace
Exposition, No. AIAA 2010-388, 4 - 7 January 2010, pp. 1–27.
9 Bisplinghoff, R. L., Ashley, H., and Halfman, R. L., Aeroelasticity, Addison Wesley Co., 1955.

30 of 31

American Institute of Aeronautics and Astronautics

 10 Singh, B. and Chopra, I., “Dynamics of Insect-Based Flapping Wings: Loads Validation,” 47th AIAA/ASME/ASCE/AHS/ASC
Structures, Structural Dynamics, and Materials Conference, No. AIAA 2006-1663, Newport, Rhode Island, 1-4 May 2006, AIAA Paper
Number 2007-1757.
11 Polhamus, E. C., “A Concept of the Vortex Lift of Sharp Edge Delta Wings Based on a Leading-Edge Suction Analogy,” Tech. Rep.
NASA TN D-3767, National Aeronautics and Space Administration, 1966.
12 Katz, J., “A Discrete Vortex Method for the Non-steady Separated Flow over an Airfoil,” Journal of Fluid Mechanics, Vol. 102, 1981,
pp. 315–328.
13 Shukla, R. K. and Eldredge, J. D., “An Inviscid Model for Vortex Shedding from a Deforming Body,” Theoretical and Computational
Fluid Dynamics, Vol. 21, 2007, pp. 343–368.
14 Ansari, S. A., Żbikowski, R., and Knowles, K., “Non-linear Unsteady Aerodynamic Model for Insect-like Flapping Wing in the
Hover. Part 1: Methodology and Analysis,” Proceedings of the I MECH E Part G Journal of Aerospace Engineering, Vol. 220, No. 2, 2006,
pp. 61–83.
15 Ansari, S. A., Żbikowski, R., and Knowles, K., “Non-linear Unsteady Aerodynamic Model for Insect-like Flapping Wing in the
Hover. Part 2: Implementation and Validation,” Proceedings of the I MECH E Part G Journal of Aerospace Engineering, Vol. 220, No. 2,
2006, pp. 169–186.
16 Gogulapati, A., Friedmann, P. P., Kheng, E., and Shyy, W., “Approximate Aeroelastic Analysis of Flapping Wings in Hover: Compari-
son with CFD and Experimental Data,” 51st AIAA/ASME/ASCE/AHS/ASC Structures, Structural Dynamics, and Materials Conference,
No. AIAA-2010-2707-967, Orlando, Florida, 12-15 April 2010.
17 Combes, S. A. and Daniel, T. L., “Into thin air: contributions of aerodynamic and inertial-elastic forces to wing bending in the
hawkmoth Manduca sexta,” The Journal of Experimental Biology, Vol. 206, 2003, pp. 2999–3006.
18 Daniel, T. L. and Combes, S. A., “Flexible Wings and Fins: Bending by Inertial or Fluid-Dynamic Forces,” Integrative and Compar-
ative Biology, Vol. 42, 2002, pp. 1044 – 1049.
19 Hamamoto, M., Ohta, Y., Hara, K., and Hisada, T., “Application of Fluid-Structure Interaction Analysis to Flapping Flight of
Insects with Deformable Wings,” Advanced Robotics, Vol. 21, No. 1-2, 2007, pp. 1–21.
20 Gogulapati, A., Friedmann, P., and Shyy, W., “Nonlinear Aeroelastic Effects in Flapping Wing Micro Air Vehicles,” 49th
AIAA/ASME/ASCE/AHS/ASC Structures, Structural Dynamics, and Materials Conference, No. AIAA Paper Number 2008-1817,
Schaumburg, IL, April 7 - 10 2008.
21 Gogulapati, A., Friedmann, P., and Shyy, W., “Approximate Aeroelastic Analysis of Flapping Wings in Hover,” International Forum
for Aeroelasticity and Structural Dynamics, No. IFASD Paper Number 2009-143, Seattle, WA, June 22-25 2009.
22 Chimakurthi, S. K., Tang, J., Palacios, R., Cesnik, C. E. S., and Shyy, W., “Computational Aeroelasticity Framework for Analyzing
Flapping Wing Micro Air Vehicles,” 49th AIAA/ASME/ASCE/AHS/ASC Structures, Structural Dynamics, and Materials Conference,
Schaumburg, Illinois, April 2008, AIAA Paper Number 2008-1814.
23 Chimakurthi, S. K., Stanford, B. K., Cesnik, C. E. S., and Shyy, W., “Flapping Wing CFD/CSD Aeroelastic Formulation Based on a
Co-rotational Shell Finite Element Formulation,” 50th AIAA/ASME/ASCE/AHS/ASC Structures, Structural Dynamics, and Materials
Conference, Palm Springs, California, May 2009, AIAA Paper Number 2009-2412.
24 Aono, H., Chimakurthi, S., Wu, P., Sallstrom, E., Stanford, B., Cesnik, C., Ifju, P., Ukeiley, L., and Shyy, W., “A Computational and
Experimental Study of Flexible Flapping Wing Aerodynamics,” 48th AIAA Aerospace Sciences Meeting Including the New Horizons Forum
and Aerospace Exposition, No. AIAA 2010-554, Orlando, Florida, 4-7 January 2010.
25 Masoud, H. and Alexeev, A., “Resonance of flexible flapping wings at low Reynolds number,” Physical review. E, Statistical, nonlin-
ear, and soft matter physics, Vol. 81, No. 5, 2010, pp. 1–5.
26 MSC. MARC, Volumes A - D, 2005.
27 Willmott, A. P. and Ellington, C. P., “The Mechanics of Flight in the Hawkmoth Manduca Sexta: I. Kinematics of Hovering and
Forward Flight,” The Journal of Experimental Biology, Vol. 200, 1997, pp. 2705 – 2722.
28 Saffman, P. G., Vortex Dynamics, Cambridge University Press, 1992.
29 Gorelov, D. N., “Calculation of Pressure on an Airfoil Contour in an Unsteady Separated Flow,” Journal of Applied Mechanics and
Technical Physics, Vol. 49, No. 3, 2008, pp. 437–441.
30 Bathe, K., Ramm, E., and Wilson, E. L., “Finite Element Formulations for Large Deformation Dynamic Analysis,” International
Journal for Numerical Methods in Engineering, Vol. 9, 1975, pp. 353 – 386.
31 Bathe, K., Finite Element Procedures, Prentice-Hall Inc., 1996.
32 Trizila, P., Kang, C.-K., Aono, H., Visbal, M., and Shyy, W., “Fluid Physics and Surrogate Modeling of a Low Reynolds Number
Flapping Rigid Flat Plate,” 28th AIAA Applied Aerodynamics Conference, No. AIAA 2010-5081, Chicago, Illinois, 28 June - 1 July 2010,
pp. 1–46.
33 Wu, P., Ifju, P., Stanford, B., Sallstrom, E., Ukeiley, L., Love, R., and Lind, R., “A Multidisciplinary Experimental Study of Flap-
ping Wing Aeroelasticity in Thrust Production,” 50th AIAA/ASME/ASCE/AHS/ASC Structures, Structural Dynamics, and Materials
Conference, No. AIAA-2009-2413, Palm Springs, California, 4-7 May 2009.
34 Aono, H., Kang, C., Cesnik, C. E. S., , and Shyy, W., “A Numerical Framework for Isotropic and Anisotropic Flexible Flapping Wing
Aerodynamics and Aeroelasticity,” 28th AIAA Applied Aerodynamics Conference, No. AIAA 2010-5082, Chicago, Illinois, 28 June - 1 July
2010, pp. 1–25.
35 Kang, C.-K., Aono, H., Cesnik, C. S., and Shyy, W., “A Scaling Parameter for the Thrust Generation of Flapping Flexible Wings,”
49th AIAA Aerospace Sciences Meeting Including the New Horizons Forum and Aerospace Exposition, No. AIAA 2011-1313, Orlando,
Florida, 4-7 January 2011.
36 Chung, J. and Hulbert, G. M., “A Family of Single-Step Houbolt Time Integration Algorithms for Structural Dynamics,” Computa-
tional Methods in Applied Mechanics and Engineering, Vol. 118, 1994, pp. 1 – 11.

31 of 31

American Institute of Aeronautics and Astronautics