Flutter of subsonic wing

M.R. Moosavi
a
, A.R. Naddaf Oskouei
a
, A. Khelil
b,
*
a
Department of Mechanical Engineering, Imam Hossein University, P.O. Box 16535-187, Tehran, Iran
b
Departement Genie Civil, ERIN IUT Nancy Brabois, 54601 Villers Les Nancy cedex 1, France
Received 15 July 2003; received in revised form 16 August 2004; accepted 14 October 2004
Available online 8 December 2004
Abstract
In this paper, a procedure is developed based on Galerkin method to predict the speed and
frequency in which utter occurs. The nite element structural model used for the wing is a three-
DOF cantilever beam, in which one coordinate is related to the vertical displacement and the other
two are corresponding to bending and rotation. This beam element has Hermit-cubic-type in bending
and linear in rotation characteristics. Consequently, an eigenvalue problem with non-symmetric
matrix coefcients was derived. It was found that as free stream velocity increases from zero up to
0.554 (for incompressible ow) and 0.526 Mach (for compressible ow), the real parts of the
eigenvalues have negative signs and the system become stable. Further increasing of free-stream
velocity causes the amplitude of the frequencies approach zero and become positive, which indicates
dynamic instability, or utter of the system.
q 2004 Elsevier Ltd. All rights reserved.
Keywords: Aeroelasticity; Subsonic wing utter; Quasi-steady aerodynamic; Galerkin method; Eigenvalue
problem
1. Introduction
One of the most critical multidisciplinary activities in aircraft, reusable launch vehicle,
and aerospace plane design is aeroelastic instability predictions and avoidance. In order to
achieve desired minimum weight design of ight vehicles, the aeroelastic instabilities,
such as utter and divergence of lifting surfaces, must be included before the initial ight
tests.
0263-8231/$ - see front matter q 2004 Elsevier Ltd. All rights reserved.
doi:10.1016/j.tws.2004.10.001
Thin-Walled Structures 43 (2005) 617627
www.elsevier.com/locate/tws
* Corresponding author. Tel.: C03 83 68 25 36; fax: C03 83 68 25 32.
E-mail address: khelil@iutnb.uhp-nancy.fr (A. Khelil).
Early pioneering work on utter was utter suppression, which was based on the
aerodynamic energy concept [1]. This approach is based on the energy required to
sustain simple harmonic motion in a typical section, having pitch, plunge, and control
surface generalized degrees of freedom. When the sign of energy is positive, energy
must be supplied, and the system is stable. General criteria and control laws for
system stability were derived and demonstrated in wind-tunnel tests. Other work
pursued in parallel was on frequency-domain aerodynamics and classical control
techniques [2].
Aeroelasticity is the science of analyzing the mutual interaction between the
aerodynamics and structure dynamics of a moving system such as aerospace vehicles.
Examples of some aeroelastic problems of aerospace vehicles are:
Panel and wing utter of space vehicles
Limit cycle oscillations of ghter aircraft with external stores
Vertical tail buffeting of twin-tailed ghter aircraft
Helicopter/shipboard dynamic interface
Fluidstructure interaction in multi-blade rotorcraft
Active aeroelastic wings
Higher harmonic control of helicopter rotor blades.
Although the physical aeroelastic problem is an interaction between the uid and
structure, the numerical aeroelastic problem is a complex multidisciplinary problem that
involves strong interaction between several physical and numerical disciplines. The
physical disciplines are the uid dynamics around the uid dynamics of the aerospace
vehicle and the structure dynamics of the exible bodies. The numerical disciplines are the
numerical interfacing between the uid and structure interfaces and the grid motion due to
the exibility of the structure [3].
One way to solve the aeroelastic problem is to combine all the modules of the
physical and numerical disciplines into one monolithic code. However, such giant
monolithic codes are difcult to develop and maintain, and their nature, cannot
contain up-to-date technology. Another way to solve the problem is to integrate the
individual analysis modules into a multidisciplinary computing environment in which
the analysis modules can run concurrently with synchronized data transfer between
the analysis modules [4,5].
A major breakthrough in aeroelasticity has been the development of time-domain
aerodynamics, based on rational function approximation to the unsteady aerodynamic
loading in the frequency domain. These approaches add a considerable number of
augmented states to the equations of motion. The number of these augmented states can he
reduced using the minimum state method [6,7].
Later, alternative two-dimensional compressible aerodynamic tools in the time
domain were developed by Leishman and Nguyen [8]. This model is suitable for both
incompressible and compressible cases. These models are based on indicial
aerodynamics and require a fairly large number of augmented aerodynamic states.
Recently, these time-domain unsteady aerodynamic models have also been extended
to two-dimensional wing/control surface combinations. A very signicant portion of
M.R. Moosavi et al. / Thin-Walled Structures 43 (2005) 617627 618
the experimental research on utter suppression has been conducted on aeroelastically
scaled wind-tunnel models, with an emphasis on wing/store utter suppression.
Because of the complexity of the problem, a large number of studies have been
theoretical in nature, aimed at understanding the basic problems, or supporting the
experimental work conducted. Almost 25 years have gone by since the eld has
started [9].
2. Flutter of cantilever wings
A classical example of non-self-adjoint problems consists of the combined bending and
torsional vibration of a cantilever aircraft wing in steady air ow shown in Fig. 1. Before
we begin describing the problem, we should dene the two axes shown in Fig. 1a, the
inertia axis and the elastic axis. The inertia axis is dened as the locus of the mass centers
of the cross sections and the elastic axis is the locus of the shear centers, whereas a shear
center is a point at which shearing force produces a pure bending and a moment about it
produces a pure torsion. We denote the bending deection of the elastic axis by w(x, t) and
the torsional rotation about the elastic axis by q(x, t), where w is positive if it acts
downward and q is positive if the leading edge is up (Fig. 1b). The angle q is referred
to the local angle of attack. We take the x-axis to coincide with the elastic axis, which is
assumed to be straight, and denote the distance between the leading edge and the elastic
axis at any point x by y
0
(x), the distance between the elastic axis and the inertia axis by
y
q
(x) and the chord length by c(x). The bending deection of the elastic axis is shown in
Fig. 1c. The speed of the air ow relative to the wing denoted by U is assumed to be
constant.
The boundary-value problem for the free vibration of the wing in the presence of
aerodynamic forces is described by the following differential equations [57]
Fig. 1. (a) Elastic axis and inertia axis for a cantilever aircraft wing in steady air ow; (b) wing cross section;
(c) bending deection of the elastic axis.
M.R. Moosavi et al. / Thin-Walled Structures 43 (2005) 617627 619
v
2
vx
2
EI
v
2
w
vx
2
_ _
Cm
v
2
w
vt
2
Cmy
q
v
2
q
vt
2
C
rU
2
2
c
!
dC
L
dq
q C
1
U
vw
vt
C
c
U
3
4
K
y
0
c
_ _
vq
vt
_ _
Z0; 0!x!L (1a)
K
v
vx
GJ
vq
vx
_ _
Cmy
q
v
2
w
vt
2
CI
q
v
2
q
vt
2
C
rU
2
2
c
2
cp
8U
vq
vt
C
1
4
K
y
0
c
_ _
dC
L
dq
q C
1
U
vw
vt
C
c
U
3
4
K
y
0
c
_ _
vq
vt
_ _ _ _
Z0; 0!x!L (1b)
which, the boundary conditions are
w Z0;
vw
vx
Z0; q Z0; at x Z0; EI
v
2
w
vx
2
Z0;
v
vx
EI
v
2
w
vx
2
_ _
Z0; GJ
vq
vx
Z0; at x ZL (2)
where
EI bending stiffness,
GJ torsional stiffness,
m mass per unit length,
I
q
mass moment of inertia per unit length,
r air density and
C
L
local lift coefcient.
The aerodynamic forces and moments were derived by means of the so-called quasi-
steady strip theory whereas the local lift coefcient C
L
is proportional to the
instantaneous angle of attack q. The derivative dC
L
/dq is assumed to be constant, with
a theoretical value of 2p for incompressible ow and an experimental value of a little bit
less than 2p. The quasi-steady assumption implies that the aerodynamic forces and
moments depend only on the instantaneous deformations; and prior history of the motion
can be ignored, which simplies the equations of motion greatly [3]. In fact, the resulting
equations of motion and boundary conditions are linear and homogeneous. Still, the
system is non-self-adjoint.
The boundary-value problem admits a solution in the exponential form
w(x; t) ZW(x)e
lt
; q(x; t) ZQ(x)e
lt
(3)
where W(x), Q(x) and l are in general complex. Inserting Eq. (3) into Eqs. (1a), (1b) and
(2) and dividing through by e
lt
, we obtain the differential eigenvalue problem consisting of
M.R. Moosavi et al. / Thin-Walled Structures 43 (2005) 617627 620
the differential equations
d
2
dx
2
EI
d
2
W
dx
2
_ _
C
rU
2
2
c
dC
L
dq
QCl
rU
2
c
dC
L
dq
WCc
3
4
K
y
0
c
_ _ _ _
Q
Cl
2
m(W Cy
0
Q)
Z0; 0!x!L (4a)
K
d
dx
GJ
dQ
dx
_ _
C
rU
2
2
c
2
1
4
K
y
0
c
_ _
dC
L
dq
Q
Cl
rU
2
c
2
1
4
K
y
0
c
_ _
dC
L
dq
WCc
1
4
K
y
0
c
_ _
3
4
K
y
0
c
_ _
dC
L
dq
C
p
8
_ _
Q
_ _
Cl
2
(my
q
WCI
q
Q)
Z0; 0!x!L (4b)
and the boundary conditions
W Z0;
dW
dx
Z0; QZ0; at x Z0; EI
d
2
W
dx
2
Z0;
d
dx
EI
d
2
W
dx
2
_ _
Z0; GJ
dQ
dx
Z0 at x ZL (5)
The differential eigenvalue problem, Eqs. (4a), (4b) and (5), has no closed-form
solution, so that we consider an approximate solution by means of Galerkins method. To
this end, we assume a solution in the form
W(x) Zf
T
1
(x)a
1
; Q(x) Zf
T
2
(x)a
2
(6)
in which f
1
and f
2
are vectors of comparison functions and a
1
and a
2
are vectors of
undetermined coefcients; where f
1
and a
1
are of dimension n
1
and f
2
and a
2
of
dimension n
2
, and n
1
Cn
2
Zn. The vector f
1
satises the boundary conditions are follows
f
1
(0) Z0; f
/
1
}
xZ0
Z0
EIf
//
1
}
xZL
Z0; (EIf
//
1
)
/
}
xZL
Z0
(7a)
and the vector f
2
satises the following boundary conditions
f
2
(0) Z0; GJf
/
2
}
xZL
Z0 (7b)
where primes denote the ordinary derivatives w.r.t. x. Inserting Eq. (6) into Eqs. (4a) and
(4b), premultiplying Eq. (4b) by f
2
and integrating over the length of the beam, we obtain
the following algebraic eigenvalue problem
K CU
2
H ClUL Cl
2
M|a Z0 (8)
M.R. Moosavi et al. / Thin-Walled Structures 43 (2005) 617627 621
where aZ[a
T
1
a
T
2
]
T
and the various matrices have the submatrices
K
11
Z
_
L
0
f
1
(EIf
//
1
T)
//
dx Z
_
L
0
EIf
//
1
f
//
1
T dx
K
12
Z0 K
21
Z0
K
22
ZK
_
L
0
f
2
(GJf
/
2
T)
/
dx Z
_
GJf
/
2
f
/
2
T dx H
11
Z0
H
12
Z
r
2
dC
L
dq
_
L
0
cf
1
f
T
2
dx
H
21
Z0
H
22
Z
r
2
dC
L
dq
_
L
0
c
2
1
4
K
y
0
c
_ _
f
1
f
T
2
dx
L
11
Z
r
2
dC
L
dq
_
L
0
cf
1
f
T
1
dx
L
12
Z
r
2
dC
L
dq
_
L
0
c
2
3
4
K
y
0
c
_ _
f
1
f
T
2
dx
L
21
Z
r
2
dC
L
dq
_
L
0
c
2
1
4
K
y
0
c
_ _
f
2
f
T
1
dx
L
22
Z
r
2
_
L
0
c
3
1
4
K
y
0
c
_ _
3
4
K
y
0
c
_ _
dC
L
dq
C
p
8
_ _
f
2
f
T
2
dx
M
11
Z
_
L
0
mf
1
f
T
1
dx
M
12
Z
_
L
0
my
q
f
1
f
T
2
dx
M
12
Z
_
L
0
my
q
f
2
f
T
1
dx
M
22
Z
_
L
0
I
q
f
2
f
T
2
dx
(9)
The eigenvalue problem (8) can be expressed in the standard form
Ax Zlx (10)
in which xZ[a
T
la
T
]
T
and
A Z
0 I
KM
K1
(K CU
2
H) KM
K1
UL
_ _
(11)
The eigenvalue l is a continuous function of the air speed U. When UZ0, the system is
conservative and l is pure imaginary. For Us0, lZaCiu. It can be shown that [1] for
sufciency small U and for dC
L
=dq!2p the wing is losing energy to the surrounding air,
so that the motion represents damped oscillation. This implies asymptotic stability, so that
a!0. At some point, as U increases, a turns from negative to positive, as shown in Fig. 2,
so that the motion turns from asymptotically stable to unstable. At point aZ0, at which the
motion is merely stable and ready to become unstable, the air speed reaches the critical
M.R. Moosavi et al. / Thin-Walled Structures 43 (2005) 617627 622
value U
cr
. There can be more than one critical point but the lowest one is the most
important. Because in actual ight, U increases from all initial zero value. There are two
types of critical values, depending on the imaginary part u. When aZ0 and uZ0, so that
lZ0, the wing is said to be in critical divergent condition. When aZ0 and us0, the wing
is said to be in critical utter condition. To calculate U
cr
, it is necessary to solve the
eigenvalue problem repeatedly by increasing the value of U, beginning with a small
magnitude. In the beginning all the eigenvalues will have negative real part. Then the rst
value of U for which the real part of eigenvalue becomes zero is U
cr
.
At rst estimate of U
cr
can be obtained by approximating W and Q by means of a single
term each, n
1
Zn
2
Z1. Then, letting lZiu in Eq. (10) and pre-multiplying by block-diag.
[I M], we can obtain U
cr
from the following determinantal equation
det
Kiu 0 1 0
0 Kiu 0 1
Kk
11
KU
2
cr
h
12
K(ium
11
CU
cr
l
11
) K(ium
12
CU
cr
l
12
)
0 K(k
22
CU
2
cr
h
22
) K(ium
12
CU
cr
l
21
) K(ium
22
CU
cr
l
22
)
_

_
_

_
Zu
4
(m
11
m
22
Km
2
12
) Kiu
3
U
cr
[m
11
l
22
Cm
22
l
11
Km
12
(l
12
Cl
21
)]
Ku
2
[U
2
cr
(l
11
l
22
Kl
12
l
21
Kh
12
m
12
Ch
22
m
11
) Ck
22
m
11
Ck
11
m
22
]
KiuU
cr
[U
2
cr
(h
12
h
21
Ch
22
l
11
) Ck
22
l
11
Kk
11
l
22
] Ck
11
(U
2
cr
h
22
Ck
22
)
Z0 (12)
Eq. (12) is a complex variable equation. Therefore, its satisfaction requires that both the
real and imaginary part to be zero, which derives a solution for both u and U
cr
. Indeed,
equating the imaginary part to zero, we obtain
u
2
Z
U
2
cr
(h
12
l
21
Ch
22
l
11
) Ck
22
l
11
Kk
11
l
22
m
12
(l
12
Cl
21
) K(m
11
l
22
Cm
22
l
11
)
(13)
Then, inserting Eq. (13) into the real part of Eq. (12), we obtain the quadratic equation in
U
2
cr
aU
4
cr
CbU
2
cr
Cc Z0 (14)
where
Fig. 2. The real part of eigenvalue vs. the air speed.
M.R. Moosavi et al. / Thin-Walled Structures 43 (2005) 617627 623
a Z(h
12
l
21
Ch
22
l
11
)(h
12
l
21
Ch
22
l
11
) K(m
11
m
22
Km
2
12
) K(l
11
l
22
Kl
12
l
21
Kh
12
m
12
Ch
22
m
11
)[m
12
(l
12
Cl
21
) K(m
11
l
22
Cm
22
l
11
)] (15)
b Z2(h
12
l
21
Ch
22
l
11
)(k
22
l
11
Kk
11
l
22
)(m
11
m
22
Km
2
12
) K[(h
12
l
21
Ch
22
l
11
)(k
22
m
11
Ck
11
m
22
) C(l
11
l
22
Kl
12
l
21
Kh
12
m
12
Ch
22
m
11
)][m
12
(l
12
Cl
21
) K(m
11
l
22
Cm
22
l
11
)] Ck
11
k
22
[m
12
(l
12
Cl
21
) K(m
11
l
22
Cm
22
l
11
)]
2
c Z(k
22
l
11
Kk
11
l
22
)
2
(m
11
m
22
Km
2
12
) K(k
22
l
11
Kk
11
l
22
)(k
22
m
11
Ck
11
m
22
)[m
12
(l
12
Cl
21
) K(m
11
l
22
Cm
22
l
11
)] Ck
11
k
22
[m
12
(l
12
Cl
21
) K(m
11
l
22
Cm
22
l
11
)]
2
Consequently, the solution of Eq. (14) simplied as follows:
U
2
cr
ZK
b
2a
G
1
2a

b
2
K4ac
_
(16)
Hence, there are four values for U
cr
. For utter to occur, at least one of these values must
be real and positive. Then, an approximation for the critical air speed U
cr
is given by the
smallest real positive value.
3. Validation
In order to validate the accuracy of the model developed so far, Golands wing was
used. The utter result in incompressible ow by using the quasi-steady method was
calculated, which is compared with the exact one [5]. It can be seen in Table 1 that the
correlation is excellent for the onset of utter speed and utter frequency. Followed by the
rst step, compressibility effect on Golands wing was investigated and the comparison
results are also displayed in Table 1. It can be seen that the compressibility only causes
about 5.0% decrease of the utter speed and 2.2% decrease of the utter frequency
compared with the incompressible model predicted. This is consistent with the well-
known fact that at the lower range of the compressible subsonic speeds, the effect of
compressibility on utter is quite small.
Table 1
Comparison of the calculated utter results
Method Description Flutter speed (Mach#) Flutter frequency (Hz)
Exact 2-D incompressible ow 0.554 11.25
Galerkin 2-D incompressible ow 0.554 11.15
Galerkin 2-D compressible ow 0.526 10.90
M.R. Moosavi et al. / Thin-Walled Structures 43 (2005) 617627 624
4. Numerical example
Consider a wing with constant chord and the following characteristics (Fig. 1):
L Z10 m; y
cg
Z0:6c; m Z200 kg; r Z1:220 kg=m
3
; c Z3 m;
y
0
Z0:5c;
dC
L
dq
Z2p
A nite element model of the wing gives natural frequencies and mode shapes similar to
those derived experimentally. The rst four modes represent the rst and the second
bending and torsion, with corresponding frequencies as 9.60, 38.20, 48.35, and 91.54 Hz,
respectively.
Applying the approach developed in this paper it was found that
u
F
Z96 rad=s and V
F
Z191 m=s
as shown in Figs. 35.
We also applied the Galerkin method to both incompressible and compressible ow in
two dimensions. The results are then compared with the exact values obtained from [5].
Comparison results show good agreement between our applied method and the exact one.
See Table 1.
5. Conclusions
A new approach has been introduced for calculating the utter condition of a linear
3-DOF aeroelastic system. By using this approach with FEM, it can be employed for any
Fig. 3. Velocity vs. damping.
M.R. Moosavi et al. / Thin-Walled Structures 43 (2005) 617627 625
higher order aeroelastic system. Based on the model developed here, the effect of
aerodynamic compressibility on utter has been considered. The major conclusions are:
Compressibility at high subsonic speeds has a signicant effect on utter speed of the
wing. The compressibility causes a drop of the utter speed and utter frequency as
compared to the incompressible ow. In addition, the mechanism of utter may be
dramatically changed by compressibility.
Elastic tailoring can effectively change the onset of utter. However, this may be
achieved at the cost of dramatically increasing the response intensity.
Fig. 4. Damping vs. frequency.
Fig. 5. Velocity vs. frequency.
M.R. Moosavi et al. / Thin-Walled Structures 43 (2005) 617627 626
References
[1] Nissim E. Flutter suppression using active controls based on the concept of aerodynamic energy. NASA TM
6199; June 1971.
[2] Friedmann PP. The renaissance of aeroelasticity and its future. J Aircraft 1999;36(1):10521.
[3] Bisplinghoff RL, Ashley H, Halfman RL. Aeroelasticity. Reading, MA: Addisson-Wesley; 1955.
[4] Dowell EH, et al. A modern course in aeroelasticity, 3rd ed. Norwell, MA: Kulwer Academic; 1995.
[5] Lin J, Iliff KW. Aerodynamic lift and moment calculations using a closed-form solution of the Possio
equation. NASA/TM-2000-209019; 2000.
[6] Karpel M. Design for active and passive utter suppression and gust alleviation. NASA CR3482; June 1981.
[7] Karpel M, Hoadley ST. Physically weighted approximations of unsteady aerodynamic forces using the
minimum-state method. NASA TP 3025; March 1991.
[8] Leishman JG, Nguyen KQ. State space representation of unsteady airfoil behavior. AIAA J 1990;28(5):
83644.
[9] Hariharan N, Leishman JG. Unsteady aerodynamics of a apped airfoil in subsonic ow by indicial concepts.
J Aircraft 1996;33(5):85568.
M.R. Moosavi et al. / Thin-Walled Structures 43 (2005) 617627 627