See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/238660518

Aeroelastic Analysis of Composite Wings

Conference Paper · April 1996

DOI: 10.2514/6.1996-1444

CITATIONS READS
34 437

3 authors:

Carlos E. S. Cesnik Dewey H. Hodges

University of Michigan Georgia Institute of Technology
277 PUBLICATIONS   7,114 CITATIONS    378 PUBLICATIONS   9,771 CITATIONS   

SEE PROFILE SEE PROFILE

Mayuresh J Patil
Virginia Polytechnic Institute and State University
136 PUBLICATIONS   2,105 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Damage Tolerant Design Strategies for Wind Turbine Blades View project

X-HALE View project

All content following this page was uploaded by Dewey H. Hodges on 31 May 2014.

The user has requested enhancement of the downloaded file.

 Aeroelastic Analysis of Composite Wings
Carlos E. S. Cesnik,∗ Dewey H. Hodges† and Mayuresh J. Patil‡
Georgia Institute of Technology, Atlanta, Georgia

Abstract the aerodynamics is appropriate. In summary, to ob-

tain an optimum design, we need an accurate as well as
An aeroelastic stability analysis is presented for efficient model to assess the aeroelastic stability.
high-aspect ratio composite wings. The structural
model is based on an asymptotically correct cross- Previous Work
sectional formulation and a nonlinear geometric ex-
One can find several papers published in the liter-
act beam analysis, both derivable from 3-D elastic-
ature related to the aeroelastic problems of aircraft. In
ity. A new 2-D unsteady inflow finite-state theory is
order to have an overview of tendencies and methods
considered for the aerodynamic part of the solution.
in current practice, a sample of the significant work is
Theodorsen theory is also implemented and used for
summarized in what follows.
most of the preliminary tests. The paper discusses,
among other things, the importance of using the right Ref. 1 presents a historical background of aeroe-
stiffness formulation in order to model material cou- lastic tailoring and the theory underlying the technol-
plings, the variations of divergence and flutter speeds ogy. The paper provides a good definition and historical
with the changes in the lamination angle of a box-beam evolution of codes and the activities of various research
model of a wing cross section, and some of the effects of groups. In many studies, the structural deformation
a nonlinear structural model on the aeroelastic stability model used is that of a beam-like wing, since tailoring
of a slender wing. focuses on bend-twist deformation coupling. Restrain-
ing the freedom of the chordwise bending mode can re-
Introduction sult in substantially different natural frequencies and
mode shapes for highly coupled laminates. The pru-
dence of retaining rigid-body modes in flutter analysis
Aeroelastic stability is an important factor in the
during design iterations was pointed out.
design of modern flexible-wing aircraft. The increasing
use of composite materials in the main structural com- In Ref. 2, static aeroelastic problems such as span-
ponents opens a wide range of options for the designer. wise lift redistribution, lift effectiveness, and aileron ef-
The tailoring characteristics not only improve the struc- fectiveness are discussed. Two theoretical models are
tural performance, thus reducing the final structural commonly used: (1) laminated plate theory with el-
weight, but allow for possible material couplings. Both ementary strip theory airloads and (2) a more gen-
static and dynamic aeroelastic stability can be altered eral representation of the laminated wing structural,
by those couplings. Thus, an analysis tool is needed to in matrix form, with a discrete-element aerodynamics
accurately model the composite structure and still be (Weissinger L method). In the latter case the box beam
computationally efficient enough to be used in prelimi- is characterized by bending stiffness EI, torsional stiff-
nary design. Moreover, the increasing interest in using ness GJ, and bending-twist coupling K. These are
active controls to change the stability characteristics of derived from classical plate theory applied to the top
the system indicates that a finite-state formulation for and bottom flanges. The example wing has a layup of
∗ Post Doctoral Fellow, School of Aerospace Engineering.
some plies of 0◦ , some of 45◦ , some of −45◦ , and some
Member, AIAA, AHS.
with a variable angle θ. It also discusses the concept
† Professor, School of Aerospace Engineering. Fellow, AIAA. of “aeroisoclinic” as the flexible wing deforms in such a
Member, AHS. way that the spanwise center of pressure does not move,
‡ Graduate Research Assistant, School of Aerospace Engineer-
nor is the wing lift-curve slope changed from its rigid-
ing. Member, AIAA.
Copyright°1996
c by Carlos E. S. Cesnik, Dewey H. Hodges,
wing value. The author concludes that “tailoring in-
and Mayuresh J. Patil. Published by the American Institute of volving bending-torsion coupling is seen to be effective
Aeronautics and Astronautics, Inc., with permission. for high-aspect-ratio wings as well as low-aspect-ratio

Presented at the 37th Structures, Structural Dynamics and Materials Conference, Salt Lake City, Utah, April 15 – 17, 1996
wings. However, more severe design requirements may for the different structural models and care should be
negate this effectiveness if the span is too large.” taken when modeling a wing.

Ref. 3 investigates critical flutter and divergence Ref. 6 tries to incorporate restrained warping ef-
velocities for a swept wing as influenced by the bending- fects in the analysis of static aeroelasticity. The struc-
torsion stiffness coupling of a composite cantilever wing. tural model is still a simplified plate-beam with all its
It assumes a plate-beam model without chordwise de- bending and torsional stiffness derived from the lam-
formation, where stiffness from spars, ribs, webs could inate composite upper and lower skins. The Laplace
be added algebraically to those of the plate. Coupling method is used to solve exactly the static equations.
terms (denoted by d22 and d26 ), due to the offset be- As concluded by the authors, “However, the ab initio
tween the rotation axis and the reference axes, were con- prediction of the character (beneficial or detrimental) of
sidered. Also, there is the presence of a coupling term the warping inhibition effect as a function of the given
(S) associated with torsional stiffness due to rigidity in composite cantilevered structure constitutes, neverthe-
tension (warping effect). The aerodynamics was based less, a task that, in spite of its importance, could not
on strip theory (“noting that the inaccuracy increases be clarified within this study.”
as the sweep angles of the wing increases”). The main
effect of including the d22 , d26 , and S terms is the in- Ref. 7 continues the work of the first author where
crease in the frequency of the torsion mode (due to the the effects of restrained warping, transverse shear ef-
warping effect). The flutter speed for Goland’s wing fects, and sweep angle are studied in the presence
is almost twice the one originally calculated, while the of a more sophisticated structural model. This work
opposite trend was observed for divergence. Passive sta- uses a thin-walled beam model based on the follow-
bility enhancement by aeroelastic tailoring of the ratio ing premises: i) cross sections do not deform in their
K own planes; ii) transverse shear measures are included;
EI involves a compromise, since maximizing divergence
speed tends to minimize flutter speed and vice versa. iii) the warping restraint effect is taken into account
through the non-uniform torsion (constrained torsion);
Ref. 4 concentrates in the aeroelastic problem of iv) the primary warping is associated with uniform-
transport aircraft with aft swept wings. The solution torsion only; and v) the secondary warping is based on
of the aeroelastic problem is based on the “integrating the work of Ref. 8. Some numerical results for diver-
matrix technique,” and the structural model is based gence of a box beam as function of sweep angle, ply
on Classical Plate Theory, which contains information angle, and the presence or absence of transverse shear
about compliances required for bending, torsion, and effects and restrained torsional warping are presented.
extension of an equivalent beam. The aerodynamics is In this paper we attempt to reproduce some of the re-
derived from a modified strip theory based on Jones’ sults of Ref. 7 to illustrate the importance of careful
approximation of the Theodorsen function. For the structural modeling.
numerical tests, the author considers both symmet-
ric (bending-twist) and antisymmetric (extension-twist) Refs. 9 and 10 are based on a shear deformable
plate layups. Among the main conclusions, “extension- plate-beam model with chordwise non-deformability.
−D16
torsion coupling ( √D ) caused some degradation of
The flutter solution is sought in closed form and this
11 D33
the flutter boundary; extension-bending coupling was limits the applicability to general configurations. Ba-
not as damaging as the extension-torsion. When both sically no results were presented. Ref. 11 further ex-
were present, the performance of the wing was degraded plores the solution and checks for the flutter instability
to that of the symmetric reference laminate.” But due of both straight and swept wing aircraft. Still, the sim-
to the way the structural model was defined, the authorplified shear deformable plate-beam model with chord-
wise non-deformability is used as the structural model
concluded that “it was possible to show the results only
for a limited number of laminates, and more investiga- and the Theodorsen’s functions for aerodynamics. A
tion would be necessary.” parameter included in the formulation allows one to
trigger whether the warping inhibition is incorporated
Ref. 5 shows another historical background and or discarded in the analysis. The main conclusion is
techniques involving composite materials for aeroelas- that “. . . transverse shear flexibility has, in general, a
tic analysis and design. One of the most interesting deleterious effect on the aeroelastic response.”
points made by the author is related to the different
techniques used in the literature to reduce the struc- To put this work in perspective, concerning struc-
tural model from the laminated plate to an equivalent tural modeling, many improvements have occurred dur-
beam. As shown there, the results are quite different ing the past several years. Starting from the crude

2
American Institute of Aeronautics and Astronautics
isotropic Euler-Bernoulli beam approach, the represen-
tation of composite wings improved to box-beam mod-
eling and full finite element discretization of the wing.
Unfortunately, even though very powerful, the full finite
element method is still too expensive to be coupled with
other disciplines with a similar level of details in the pre-
liminary design phase. Therefore, a reduced approach
(plate or beam) should be used and the total number
of states kept as low as possible. Regardless the kind of
modeling used, it has to take into account the complex
geometry and material distribution of a lifting surface.
The most popular method of composite wing model-
ing found in the literature is the plate-beam approach,
where upper and lower flanges of the wing box are mod-
eled as laminated composite plates. Then, those plates
are reduced to beams that are usually characterized by
three parameters: EI, GJ, and K, corresponding to
the beam bending stiffness, torsional stiffness, and the
bending-torsion coupling stiffness. Unfortunately, this
reduction is sometimes done improperly, and final re-
sults may vary quite significantly among the processes.5
This approach, even though very efficient, lacks the in-
clusion of other kind of couplings that might be present
due to changes in the material distribution. Moreover,
it is seldom representative of the shape and material
distribution for an actual wing.
From the subsonic aerodynamics point of view, el-
ementary strip theory is still in use. That is basi-
cally due to its simplicity and closed-form expressions
for the airloads. Also found is the Doublet Lattice
method, as used in NASTRAN,12 for example. More de-
tailed formulations, like the three-dimensional unsteady
Euler/Navier-Stokes equations, are still too expensive
for preliminary design. As an example, the Navier-
Stokes version of ENSAERO13 for “a typical dynamic
aeroelastic response . . . requires about 4 CPU hours
and 8 million words of central memory” on the 380
MFLOPS CRAY C90 at Ames Research Center.
Present Approach
This paper will address several closely related as-
pects of composite wing aeroelasticity. First, using re-
cently developed tools14,15,16 it is now possible to rigor-
ously collapse a three-dimensional (3-D) built-up struc-
tural model to either one- (1-D) or two-dimensional (2-
D) models. Since at least the trends of all phenomena
known to be significant in wings of moderate to high
aspect ratio can be represented in terms of 1-D models
(i.e., refined beam theory), the geometrically-exact in-
trinsic beam theory is used, along with the correspond-
ing asymptotically correct cross-sectional analysis. This
3
American Institute of Aeronautics and Astronautics
will allow for rigorous treatment of drag effects, steady-
state elastic twist, sweep, etc., in a model that requires
very few degrees of freedom relative to a 2-D or 3-D
model. It is noted that the present structural model
has been thoroughly validated for many different types
of cross sections. This approach, in spite of its sim-
plicity, is very faithful to 3-D elasticity for the global
response of the structure.
Theodorsen’s theory has been widely used to get
the unsteady aerodynamic forces and moments. But
the theory is restricted to harmonic motion and thus
cannot give the damping for cases other than flutter.
It is used only to get the flutter/divergence trends for
various kinds of structural coupling. To get a model
which is simple yet accurate and gives an aerodynamic
representation that is amenable to state-space analy-
sis, a recently developed 2-D finite-state aerodynamic
model17 is used. This kind of simulation could later be
used to design control systems or analyze an aeroser-
voelastic problem.
Structural Model
During the last seven years, a comprehensive beam
modeling framework has been developed by the first two
authors and others.14,18,19,15,20,,16 With the modeling
power of the finite element method, it takes a two-step
approach and, therefore, facilitates the keeping of a very
small number of states. It is based on 3-D elasticity and
is capable of modeling complex cross-sectional geome-
tries (solid, built-up, or thin-walled; open or closed; air-
foil shaped if necessary), including all possible couplings
and deformation. The resulting beam equations are ge-
ometrically exact (and therefore nonlinear). Asymp-
totically correct 3-D strain/stress can be recovered at
any point within the structure. It has been successfully
applied to rotary-wing static and dynamic aeroelastic
stability problems.
The mixed variational formulation for dynamics of
moving beams14 is applied here. A global frame a is
fixed to the aircraft body and the formulation is de-
veloped in this frame. The undeformed wing cross-
sectional reference frame is denoted by b, whereas B
denotes the deformed wing cross-sectional frame. The
expressions for strain and kinetic energies as well as
their variations can be derived. Using these expressions
in Eq. (24) of Ref. 14 and applying Lagrange’s multi-
pliers on the strain and velocity constraints one gets, With the transformation we can express displacement
Z Z lh and velocity components in a which are independent of
t2
beam geometry and deformation, and still express inter-
δV ∗ TB PB + δΩ∗ TB HB − δγ ∗ T FB
t1 0 nal force and momentum components in B which allows
to use the simplest form of constitutive laws. Thus we
−δκ∗ T MB + δFBT (γ − γ ∗ ) + δMBT (κ − κ∗ ) get the global frame version of the mixed variational
¤ formulation as21
(1) Z t2
−δPBT (VB − VB∗ ) − δHB
T
(ΩB − Ω∗B ) dx1 dt
δΠb dt = 0 (7)
Z t2 t1
+ δW dt = δA
t1
where
Z l½
δΠb = δu0T T ab
a C C FB
where column matrices γ, κ are the generalized strain 0
measures, VB , ΩB are the generalized velocity measures, £ ¤
FB , MB are the generalized force measures, PB , HB are ea C T C ab PB
+δuTa (C T C ab PB )˙ + ω
the generalized momentum measures in frame B, and 0T T
δW and δA are the virtual work of applied loads and +δψ a C T C ab MB − δψ a C T C ab (e e) FB
e1 + γ
virtual action at the ends of the beam and at the ends of h i
ea C T C ab HB + C T C ab VeB PB
T
the time interval, respectively. The strain-displacement +δψ a (C T C ab HB )˙ + ω
relations γ ∗ , κ∗ and velocity-displacement relations VB∗ ,
Ω∗B satisfy the following kinematic relation T£ ¤ 0T
−δF a C T C ab (e1 + γ) − C ab e1 − δF a ua
¡ ¢
γ ∗ = C Ba C ab e1 + u0a − e1 (2) Ã !
T θe θθT 0T
  −δM a ∆+ + C ab κ − δM θ
θe 2 4
 ∆− 2  0
κ∗ = C ba 

θ (3) T ¡ ¢
θT θ  +δP a C T C ab VB − va − ω
T
ea ua − δP a u̇a
1+
4
à !
VB∗ = C Ba (va + u̇a + ωe a ua ) (4) T θe θθT ¡ T ab ¢
+δH a ∆ − + C C ΩB − ω a
  2 4
θe
 −
2 
∆ ¾
Ω∗B = C ba 
 T 
 θ̇ + C Ba ωa (5) T T
θ θ −δH a θ̇ − δua fa − δψ a ma dx1
T
1+
4
³ ´ ¯l
where ua , va , ωa are column matrices that contain the T T T ¯
− δuTa F̂a + δψ a M̂a − δF a ûa − δM a θ̂ ¯
displacement, initial velocity, and initial angular veloc- 0
ity measures in frame a and C ab , C Ba are the transfor- (8)
mation matrices. In the above equation, strain and force measures and
velocity and momentum measures are related through
Each term in Eq. (1) is transformed to frame a. the constitutive laws in the following form
The main purpose of the transformation is increasing ½ ¾ ½ ¾
the versatility without destroying the compactness of FB γ
= [S ] (9)
the variational formulation. In the transformation, all M B κ
the virtual quantities are depend only on Rodrigues ½ ¾ · ¸½ ¾
PB m∆ 0 VB
parameters (θ) measured in frame a at the point of = (10)
HB 0 I ΩB
interest.14 Now the rotation matrix C is defined as
C ab C Ba and expressed in terms of θ as where S is the cross-sectional stiffness matrix that can
µ ¶ be generally obtained by the analysis of Ref. 22. That
θT θ θθ T
takes into account the material distribution and actual
1− ∆ − θe +
4 2 geometry for an initially curved and twisted beam. For
C= T
(6) thin-walled, closed cross sections, the simple yet asymp-
θ θ
1+ totically correct theory of Ref. 16 may be used.
4

4
American Institute of Aeronautics and Astronautics
 The above equations represent a fully nonlin- Aeroelastic system of equations
ear model for initially curved and twisted anisotropic
beams. Its main features include geometrical exactness Due to the formulation’s weakest form, the simplest
and mathematical elegance.14 Written in the above form, shape functions can be used.14 Thus,
it expresses the variation of the energy terms with com-
ponents in the global frame even though the stress and δua = δui (1 − ξ) + δuj ξ ua = ui
strain fields are still in the local beam frame; thus, the
unknowns can be measured in a very convenient way. δψ a = δψ i (1 − ξ) + δψ j ξ θ = θi

Aerodynamic Model δF a = δF i (1 − ξ) + δF j ξ FB = Fi
(17)
δM a = δM i (1 − ξ) + δM j ξ MB = Mi
A 2-D finite-state inflow model was developed by
Peters et al.17 that is derived from first principles. As
δP a = δP i PB = Pi
shown in Ref. 17, this method gives excellent correlation
with Theodorsen and Wagner functions with few inflow
states. δH a = δH i HB = Hi
L
The non-dimensional lift (L̄ = 2πρbV 2 ) and mo-
M
ment about the midchord (M̄ = 2πρb2 V 2 ) expressions17 With these shape functions, the spatial integration in
for an airfoil are given by Eq. (8) can be performed explicitly to give a set of non-
linear equations as described in Ref. 23. These equa-
µ ¶
1 1 1 tions can be separated into structural (FS ) and aerody-
L̄ = wa + w1 − λa + ẇa − ẇ2 (11) namic (FL ) terms and written as
2 2 2

1 1 1 FS (X, Ẋ) − FL (X, Y, Ẋ) = 0 (18)

M̄ = (wa − λa ) − w2 − (ẇ1 − ẇ3 ) (12)
2 4 16
where X is the column matrix of structural variables
where λa denotes the inflow and wn ’s are the coefficients
and Y is a column matrix of inflow states. Similarly we
of the total downwash expansion.
can separate the inflow equations into an inflow compo-
The total downwash on a airfoil is completely de- nent (FI ) and a downwash component (FW ) as
termined by the non-penetration boundary condition.
For small airfoil deformation y(x, t) (the airfoil lies on − FW (Ẋ) + FI (Y, Ẏ ) = 0 (19)
−1 < x < +1), one gets
µ ¶
∂y ∂y
w=− + (13) Solution Process
∂x ∂t
The solutions of interest for the two coupled sets
of equations (Eqs. 18 and 19) can be expressed in the
The total downwash is expanded in terms of the form      
velocity potentials given in Ref. 17. The inflow (λa ) is  X   X̄   X̌(t) 
represented in terms of N states λ1 , λ2 , . . ., λN as = + (20)
     
Y Ȳ Y̌ (t)
1X
N
λa ≈ bn λ n (14) where (¯) denotes steady-state solution and (ˇ) denotes
2 n=1 the small perturbation on it.
where the bn are found by least square method, and For the steady-state solution one gets Ȳ identically
the λn are obtained by solving a set of N first-order equal to zero (from Eq. 19). Thus, one has to solve a
differential equations17 set of nonlinear equations given by

1 FS (X̄, 0) − FL (X̄, 0, 0) = 0 (21)

λ̇a − λ̇2 + λ1 = 2Γ̄˙ (15)
2
1 ³ ´ 2 The Jacobian matrix of the above set of nonlinear
λ̇n−1 − λ̇n+1 + λn = Γ̄˙ (16) equations can be obtained analytically and is found to
2n n
be very sparse.23 The steady-state solution can be found

5
American Institute of Aeronautics and Astronautics
very efficiently using Newton-Raphson method. Once a are also used in some of the studies presented, allowing
steady-state solution is obtained, one can either do a us to compare our results with published ones.
static stability (divergence) analysis or a dynamic sta-
bility (flutter) analysis about the steady state.
To get the divergence solution, Eq. 21 is trans-
Static Aeroelastic Tailoring
formed into an eigenvalue problem, the eigenvalues of
which give the divergence dynamic pressure:
· ¸ · ¸
∂FS 1 ∂FL Given a geometry and material distribution for the
{X} = qdiv {X} (22) wing cross section, the aeroelastician must have at hand
∂X X=X̄ q ∂X X=X̄
consistent stiffness constants to be used in the beam
where qdiv is the divergence dynamic pressure and both analysis. Up to now, few asymptotically correct cross-
the Jacobian matrices are obtained at the calculated sectional analysis formulations are available. Among
steady state. them, the most general formulation has been imple-
mented in VABS,24 a finite-element based computer
By perturbing Eqs. 18 and 19 about the calculated
code that handles arbitrary geometry, composite ma-
steady state using Eq. 20, the transient solution is ob-
terials, and effects such as initial twist and curvature
tained from
at the cross-sectional level. For thin-walled single-cell
 
· ∂FS ¸   X̌ composite beams, the work of Ref. 16 is a good alter-
∂X − ∂X − ∂F
∂FL L
∂Y + native. Both are used in this work. Even though not
0 ∂FI X = X̄  
∂Y Y̌ asymptotically correct, the work of Ref. 25 is also used
Y=0
    (23) in this paper as a simple way to get approximate trans-
· ∂FS ¸ ˙  0
 X̌
∂ Ẋ
− ∂FL
∂ Ẋ
0 verse shear stiffness constants analytically.
=
− ∂F W ∂FI X = X̄  ˙   
∂ Ẋ ∂ Ẏ Y̌ 0

AAAA
Y=0

Now assuming the dynamic modes to be of the form

est , the above equations can be solved as an eigenvalue
AAAA
h = 0.4 in

AAAA A
c = 10.0 in
b = 2.0 in
L = 80.0 in L
problem to get the modal damping, frequency and mode
shape of the various modes. Thus one can get the sta-
bility condition of the aeroelastic system at various op-
erating conditions (steady states). AAAAAAA
AAAAAAA
Material: Graphite/Epoxy
E1 = 30 Msi

AA AAAAAAA
h
E2 = E3 = 0.75 Msi
Numerical Results
AA AA AAAAAAA
G13 = G23 = 0.37 Msi
b

AAAAAAA
G12 = 0.45 Msi
c ν12 = ν23 = ν13 = 0.25
In what follows, different aspects of the aeroelas-
tic stability of a high-aspect ratio wing model are dis-
cussed. Among them, the importance of using the right
stiffness formulation in order to model material cou- Fig. 1. Geometry of the box beam
plings, the variations of divergence and flutter speeds
with the changes in the lamination angle of a box-beam
model of a wing cross section, and some of the effects of
As a first configuration test, the wing used in
a nonlinear structural model in the aeroelastic stability
Ref. 7 is considered for the present numerical study.
of a slender wing.
The wing is prismatic, and the planform is shown in
A modular computer code has been developed Fig. 1. The structural model is a box beam made of
which is a direct implementation of the aforementioned Graphite/Epoxy (properties described in Fig. 1) and
structural and aerodynamic theories. For its valida- the ply angle can be varied from −90◦ to 90◦ , depend-
tion, a simplified code was written that implements a ing on the configuration of the cross section (Fig. 2).
linearized beam model (solution based on the Rayleigh- The sweep angle Λ is allowed to vary. The linear code
Ritz method) and 2-D Theodorsen’s unsteady aerody- is used for this study in order to allow a direct compar-
namic formulation (strip theory). Solutions from this ison with the results presented in Ref. 7.

6
American Institute of Aeronautics and Astronautics
 +θ +θ
4000
Classical 4x4 Stiffness
Overconstrained 4x4 Stiffness

Normalized Divergence Dynamic Pressure

J Theory of Librescu & Song
J
3000
Λ=0
+θ −θ J J
J
2000
JJ Λ = −2 J J
Fig. 2. CUS (left) and CAS (right) configurations (Ply J J
angle θ measured about the outward normal axis) J J
Λ = −15 J
J
J J
1000 J
J
Fig. 3 shows the variation of divergence dynamic
Λ = −60 J
pressure with ply angle for a circumferentially uniform J J J J
stiffness (CUS) configuration, including different values J J J J
0
of the sweep angle Λ. The CUS configuration produces
0 15 30 45 60 75 90
extension-twist coupling and the fiber orientation in the
Ply angle (degree)
cross section is represented in Fig. 2 (left). When the
authors of Ref. 7 studied this configuration, they were Fig. 3. Variation of divergence dynamic pressure with
interested in the effects of transverse shear in the diver- ply angle for a CUS configuration (normalized with re-
gence speed. The symbols showed in Fig. 3 are samples spect to the divergence dynamic pressure for 0◦ ply an-
of their numerical results without the inclusion of trans- gle)
verse shear. As discussed in Ref. 19, there are basically
two ways to get a 4 × 4 stiffness model from a 6 × 6 stiff-
ness formulation for the anisotropic beam. The first is
achieved by just neglecting the transverse shear effects Flutter Tailoring
all together from the stiffness matrix. This does not lead Again here linear structural theory is used to cal-
to a correct 4 × 4 matrix, over-estimating some of the culate flutter speeds. Also, strip theory is used to ex-
stiffness constants (see dotted lines in Fig. 3). The sec- trapolate the 2-D unsteady Theodorsen’s theory to a
ond approach is the consistent one, done by minimiza- finite wing. Artificial damping is added and a flutter
tion of the strain energy with respect to the transverse determinant is created. The V-g method is used to get
shear measures (solid lines in Fig. 3). By doing so, the the flutter speed.
important contribution of the coupling terms between
transverse shear and the classical measures are correctly For lack of published flutter results for composite
accounted for. This result can be directly achieved by box beams, the code is validated by comparing the flut-
using an asymptotically correct classical formulation, as ter speed of Goland’s typical wing.26 Fig. 4 shows the
done in Refs. 16 and 24. V-g plot obtained for this case. The flutter and diver-
gence point can be easily spotted. The present theory
gives a flutter speed of 445 fps as compared to the exact
For this particular example, the 6×6 cross-sectional
flutter speed of 450 fps, and the flutter frequencies are,
stiffness constants were obtained by using Ref. 25. Even
respectively, 70 rad/s and 70.7 rad/s (both with 1.0%
though not described in Ref. 7, the results suggest that
relative error).
the authors used the first method of disregarding trans-
verse shear effects. As one can see from Fig. 3, when the Now for the composite box beam wing, consider the
bending stiffness starts dominating the behavior of the variation of the divergence and flutter speeds with ply
wing (large sweep angles and 15◦ ≤ θ ≤ 60◦ ), even the angle for a circumferentially asymmetric stiffness (CAS)
qualitative behavior changes from the two stiffness mod- configuration. The CAS configuration produces vertical
els. The missing effects that are totally associated with bending-twist coupling, the fiber orientation in the cross
transverse shear7 are in part caused by the reduction of section is represented in Fig. 2 (right). As for diver-
the effective bending stiffness due to the bending-shear gence, positive ply angles produce a favorable bending-
coupling (present in a CUS configuration). This just twist coupling, leading to a very high divergence speed,
reinforces the fact that the aeroelastician has to have whereas a negative ply angle shows lower divergence
available a consistent stiffness model to be used in the speed (see Fig. 5). The flutter results are more inter-
analysis. esting and thought provoking. Flutter involves dynamic

7
American Institute of Aeronautics and Astronautics
interaction of various modes. Also, the normal modes
of vibration of the composite beam change with ply an-
gle, thus leading to a change in the flutter mode shape.
In Fig. 5, only the lowest flutter speed is represented.
The plot is not smooth due to the changes of the lowest
flutter mode shape. Future work will include examining
these flutter mode shapes and its variation with ply an-
gle, which should provide a better understanding about
the phenomenon.
100
frequency
80 damping (g)
frequency (rad/s)-->
Effects of Geometric Nonlinearity
The effects of geometric nonlinearities on the
steady-state solution, and the variation of divergence
speed with it, are calculated using the geometrically
exact formulation.
First, the effect of geometric nonlinearity on
Goland’s typical wing is illustrated. Fig. 6 and 7 com-
pare the variation of the ratio of total to rigid lift, and
the variation of tip displacement, respectively, with the
dynamic pressure obtained by the goemetrically exact
theory and its linearized form. As one can see, both

total lift and tip displacement are overpredicted by the

60
linear theory, and the difference increases with the dy-
40
namic pressure. At 80% of the divergence dynamic pres-
sure expected based on the solution about the unloaded
20 state, there is an error of about 100% for the total lift
ratio, and of more than that for the tip displacement.
0 These point to the fact that geometrically exact theory
becomes increasingly important as one increases the dy-
-20 namic pressure, and, therefore, the loads.
<-- g

-40
5
linearized theory
-60
0 200 400 600 800 1000 geometrically exact theory
Velocity (fps) 4
(Total Lift)/(Rigid Lift)

Fig. 4. V-g plot for Goland’s wing 3

2
2

1.5 1
Normalized Velocity

0
1 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
q/q_div

Fig. 6. Variation of lift with the nonlinear steady state

0.5 for Goland’s wing
Divergence boundary
Flutter boundary
Divergence is a static instability that is calculated
0 about a given equilibrium state. The linear approach
-90 -60 -30 0 30 60 90
takes the unloaded state and calculates its correspond-
Ply angle (degree)
ing divergence dynamic pressure. But as the dynamic
pressure increases (in the subcritical range), the wing
Fig. 5. Variation of flutter and divergence velocities with deforms and different equilibrium positions are reached
ply angle for a box beam wing (CAS configuration) – by the wing. From this loaded equilibrium position, a
normalized with respect to the divergence velocity for new critical dynamic pressure may be evaluated. Fig. 8
0◦ ply angle shows the change in the divergence dynamic pressure

8
American Institute of Aeronautics and Astronautics
based on the loaded equilibrium state with the one from fore. The results are presented in Figs. 9 – 11. For this
the unloaded state, as function of the loading level (dy- numerical test, a CAS configuration with θ = −45 is
namic pressure). After almost no change on the critical chosen. The trends are similar to the one before, but
dynamic pressure up to 50% of the unloaded divergence as one can see from Figs. 9 – 11, the effects are more
dynamic pressure, the difference starts increasing and pronounced in the composite wing. This is due to the
reaches more than 9% at 80% of the unloaded diver- higher flexibility of the composite setup and the pres-
gence dynamic pressure. ence of bending-twist coupling at the constitutive-law
level. Particularly, Fig. 11 shows a large increase in the
8 divergence dynamic pressure when the loaded equilib-
linearized theory rium state is taken into account.
7 geometrically exact theory 4
6 linearized theory
tip displacement (ft)

geometrically exact theory

5
3

(Total Lift)/(Rigid Lift)

4

3
2
2

1
1
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
q/q_div 0
Fig. 7. Variation of tip displacement with the nonlinear 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
steady state for Goland’s wing q/q_div

Fig. 9. Variation of lift with the nonlinear steady state

10 for the composite box beam (CAS θ = −45◦ )
% change in q_div with steady state

8 120
linearized theory
6 100 geometrically exact theory
tip displacement (in)

4 80

2 60

0 40

-2 20
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
q/q_div
0
Fig. 8. Variation of divergence dynamic pressure with 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
the nonlinear steady state for Goland’s wing q/q_div

Fig. 10. Variation of tip displacement with the nonlinear

A similar study to the one just described above steady state for the composite box beam (CAS θ =
was done for a composite box beam wing described be- −45◦ )

9
American Institute of Aeronautics and Astronautics
 These results show that for a highly flexible wing,
the flying load conditions and the associated equilib-
rium state should be taken into account in order to
estimate the divergence speed. Since in most cases
large displacements and rotations are involved, a nonlin-
ear structural model is required. Some extra attention
should be given to the effects of those new equilibrium
position on the applicability of the aerodynamic and in-
flow models. This and a similar study for the nonlinear
effects on the flutter speed will be presented in a later
paper.
70
60
50
% change in q_div
40
30
20
10
0
0 0.1 0.2 0.3 0.4 0.5
q/q_div
Fig. 11. Variation of divergence dynamic pressure with
the nonlinear steady state for the composite box beam
(CAS θ = −45◦ )
Conclusions
Aeroelastic stability is an important factor in the
design of modern flexible wing aircraft. The possibility
of using material couplings in the structural tailoring
process opens new frontiers to the design of a compos-
ite wing. As shown, both static and dynamic aeroe-
lastic stability can be altered by those couplings. The
present work discussed state-of-the-art analysis tools to
be used in preliminary design of high-aspect ratio com-
posite wings.
Aeroelastic analysis has been implemented for lin-
ear and nonlinear models of anisotropic closed section
beams using Theodersen’s function and 2-D finite-state
aerodynamics. The later aerodynamic model has not
been exercised for dynamic stability analysis in the
present paper. Goland’s typical wing is used to verify
the procedure and to show the presence of some non-
linear effects even in a metalic structure. Box beam
American Institute of Aeronautics and Astronautics
configurations with different levels of vertical bending-
twist and extension-twist couplings are used to repre-
sent composite wings. Results show the need for the use
of the geometrically exact theory for aeroelastic analy-
sis, especially for composites. Due to the complexity of
the possible couplings present in the structure, special
attention must be given to the cross-sectional stiffness
constants used in the analysis and their effects in the
aeroelastic response.
For future study, the 2-D finite-state theory will be
used for dynamic analysis. This will enable the calcu-
lation of flutter speeds, real damping coefficients, effect
of steady-state lift on flutter, etc.
References
[1] Shirk, M. H., Hertz, T. J., and Weisshaar, T. A.,
“Aeroelastic Tailoring – Theory, Practice, and
Promise,” J. Aircraft, Vol. 23, No. 1, Jan. 1986,
pp. 6–18.
[2] Weisshaar, T. A., “Aeroelastic Tailoring of For-
ward Swept Composite Wings,” J. Aircraft,
Vol. 18, No. 8, Aug. 1981, pp. 669–676.
[3] Lottati, I., “Flutter and Divergence Aeroelastic
0.6 0.7 0.8
Characteristics for Composite Forward Swept Can-
tilevered Wing,” J. Aircraft, Vol. 22, No. 11, Nov.
1985, pp. 1001–1007.
[4] Green, J. A., “Aeroelastic Tailoring of Aft-Swept
High-Aspect-Ratio Composite Wings,” J. Aircraft,
Vol. 24, No. 11, 1987, pp. 812–819.
[5] Weisshaar, T. A., “Aeroelastic Tailoring – Creative
Uses of Unusual Materials,” In Proceedings of the
28th Structures, Structural Dynamics, and Mate-
rials Conference, Monterey, California, April 6–8,
1987, AIAA-87-0976-CP.
[6] Librescu, L. and Thangjitham, S., “Analytical
Studies on Static Aeroelastic Behavior of Forward-
Swept Composite Wing Structure,” J. Aircraft,
Vol. 28, No. 2, Feb. 1991, pp. 151–157.
[7] Librescu, L. and Song, O., “On the Static Aeroe-
lastic Tailoring of Composite Aircraft Swept Wings
Modelled as Thin-Walled Beam Structures,” Com-
posites Engineering, Vol. 2, No. 5–7, 1992, pp. 497–
512.
[8] Gjelsvik, A., The Theory of Thin-Walled Bars,
John Wiley, New York, 1981.
10
 [9] Karpouzian, G. and Librescu, L., “Exact Flutter [18] Atilgan, A. R. and Hodges, D. H., “A Unified Non-
Solution of Advanced Anisotropic Composite Can-
tilevered Wing Structure,” In Proceedings of the
34th Structures, Structural Dynamics, and Mate-
rials Conference, La Jolla, California, April 1993,
pp. 1961–1966, AIAA-93-1535-CP.
[10] Karpouzian, G. and Librescu, L., “A Refined
Structural Model of Advanced Composite Aircraft
Wings and Its Use in Aeroelastic Analyses,” In
Librescu, L., editor, Non-Classical Problems of
the Theory and Behavior of Structures Exposed
to Complex Environmental Conditions, pp. 83–94.
ASME, AMD-Vol. 164 1993.
[11] Karpouzian, G. and Librescu, L., “Three- [21] Shang, X. and Hodges, D. H., “Aeroelastic Sta-
Dimensional Flutter Solution of Aircraft Wings
Composed of Advanced Composite Materials,” In
Proceedings of the 35th Structures, Structural Dy-
namics, and Materials Conference, Hilton Head,
South Carolina, April 1994, pp. 2851–2857, AIAA-
94-1419-CP.
[12] Anon., MSC/NASTRAN Handbook for Aeroelastic
Analysis, The Macneal-Schwendler Corporation,
Los Angles, California, 1987.
[13] Guruswamy, G. P. and Byun, C., “Fluid-Structural
Interactions Using Navier-Stokes Flow Equations
Coupled With Shell Finite Element Structures,”
In Proceedings of the AIAA 24th Fluid Dynamics
Conference, Orlando, FL, July 6–9, 1993, pp. 1–10,
AIAA-93-3087.
[14] Hodges, D. H., “A Mixed Variational Formulation
Based on Exact Intrinsic Equations for Dynamics
of Moving Beams,” International Journal of Solids
and Structures, Vol. 26, No. 11, 1990, pp. 1253 –
1273.
[15] Cesnik, C. E. S. and Hodges, D. H., “Stiffness Con-
stants for Initially Twisted and Curved Composite
Beams,” Applied Mechanics Reviews, Vol. 46, No.
11, Part 2, 1993, pp. S211 – S220.
[16] Berdichevsky, V., Armanios, E., and Badir,
A., “Theory of Anisotropic Thin–Walled Closed–
Cross–Section Beams,” Composites Engineering,
Vol. 2, No. 5–7, 1992, pp. 411–432.
[17] Peters, D. A., Karunamoorthy, S., and Cao, W.-
M., “Finite State Induced Flow Models; Part I :
Two-Dimensional Thin Airfoil,” Journal of Air-
craft, Vol. 32, No. 2, Mar.–Apr. 1995, pp. 313–322.
11
American Institute of Aeronautics and Astronautics
View publication stats
linear Analysis for Nonhomogeneous, Anisotropic
Beams with Closed Cross Sections,” AIAA Jour-
nal , Vol. 29, No. 11, 1991, pp. 1990 – 1999.
[19] Hodges, D. H., Atilgan, A. R., Cesnik, C. E. S.,
and Fulton, M. V., “On a Simplified Strain Energy
Function for Geometrically Nonlinear Behaviour
of Anisotropic Beams,” Composites Engineering,
Vol. 2, No. 5 – 7, 1992, pp. 513 – 526.
[20] Cesnik, C. E. S., Sutyrin, V. G., and Hodges, D. H.,
“Refined Theory of Composite Beams: The Role of
Short-Wavelength Extrapolation,” Int. J. Solids
and Structures, 1996, To appear.
bility of Composite Rotor Blades in Hover,” In
Proceedings of the 36th Structures, Structural Dy-
namics, and Materials Conference, New Orleans,
Louisiana, April 1995.
[22] Cesnik, C. E. S., Cross-Sectional Analysis of
Initially Twisted and Curved Composite Beams,
Ph.D. Thesis, Georgia Institute of Technology,
May 1994.
[23] Hodges, D. H., Shang, X., and Cesnik, C. E. S.,
“Finite Element Solution of Nonlinear Intrinsic
Equations for Curved Composite Beams,” In Pro-
ceedings of the 36th Structures, Structural Dy-
namics, and Materials Conference, New Orleans,
Louisiana, April 1995.
[24] Cesnik, C. E. S. and Hodges, D. H., “VABS: A
New Concept for Composite Rotor Blade Cross-
Sectional Modeling,” In Proceedings of the Amer-
ican Helicopter Society 51st Annual Forum, Fort
Worth, Texas, May 9–11, 1995.
[25] Rehfield, L. W., Atilgan, A. R., and Hodges, D. H.,
“Nonclassical Behavior of Thin-Walled Composite
Beams with Closed Cross Sections,” Journal of the
American Helicopter Society, Vol. 35, May 1990,
pp. 42–50.
[26] Goland, M., “The Flutter of a Uniform Cantilever
Wing,” Journal of Applied Mechanics, Vol. 12,
No. 4, December 1945, pp. A197 – A208.