Beruflich Dokumente
Kultur Dokumente
Composite Structures
journal homepage: www.elsevier.com/locate/compstruct
Department of Mechanical and Aerospace Engineering, Science and Research Branch, Islamic Azad University, Tehran, Iran
Department of Aerospace Engineering, Center of Excellence in Computational Aerospace, Amirkabir University of Technology, 424 Hafez Avenue, Tehran 15875-4413, Iran
c
Aerospace Engineering Dept., Sharif University of Tech., Azadi Ave., PO Box 11155-8639, Tehran, Iran
b
a r t i c l e
i n f o
Article history:
Available online 17 March 2014
Keywords:
Aeroelasticity
Composite wing
FEM
Nonlinear utter
ONERA aerodynamic
VABS
a b s t r a c t
The aim of this paper is to develop a modied 1D structural dynamics model for aeroelastic analysis of a
composite wing under large deformations. To attain this goal, an accurate available mechanical beam
model of a composite wing was considered and improved to simulate large deformation behavior. Also,
in aerodynamic aspect of view, a semi-experimental unsteady aerodynamic (ONERA dynamic stall)
model has been incorporated to construct the aeroelastic model. To set up a utter determination tool
based on the eigenvalue analysis, Finite Element Method (FEM) has been implemented to discretize
the aeroelastic equations. Also, a nite element cross-sectional analysis code VABS (Variational Asymptotical Beam Sectional Analysis) has been applied to determine composite cross-sectional properties
across the wing span. Because of the existence of nonlinear terms in the aeroelastic equations, due to
the large deformation behavior, the perturbed dynamic equations have been established about the nonlinear static equilibrium to capture the utter boundaries. The obtained results are in good agreement
with the available experimental data. It is found that the present aeroelastic model is appropriate for
analysis of composite wings with arbitrary cross-sections.
2014 Elsevier Ltd. All rights reserved.
1. Introduction
Aeroelastic instability is an important concept in an air vehicle
design process that may be lead to a catastrophic failure. Many
accidents due to this phenomenon have been reported yet [1].
Nowadays, the demands for high maneuverability, performance
and speed air vehicles as well as agility are increasing with application of composite materials in aerospace industries. To meet the
above characteristics, lightweight and therefore more exible
structures have been developed. This will result in signicant
structural nonlinearity specically for wings.
In the structural dynamic and aeroelastic analyses of wings, for
sake of simplicity, 1D beam models are always used. In a 1D model,
the 3D problem is reduced to a set of variables that only depends on
the beam-axis coordinate. 1D structural elements (beams) are simpler and computationally more efcient than 2D (plate/shell) and
3D (solid) elements. This feature makes beam theories still very
attractive for the static, dynamic and aeroelastic analysis of structures. The famous classical beam models have been constructed
based on the EulerBernoulli or Timoshenko theory. But, they have
Corresponding author. Tel.: +98 3113660011.
E-mail addresses: koohi@iaukhsh.ac.ir (R. Koohi),
(H. Shahverdi), haddadpour@sharif.edu (H. Haddadpour).
http://dx.doi.org/10.1016/j.compstruct.2014.03.012
0263-8223/ 2014 Elsevier Ltd. All rights reserved.
h_shahverdi@aut.ac.ir
119
120
3 2
rxx
Q 11
6
6
r Q e ) 4 rxf 7
5 4 Q 15
rxg
Q 16
Q 15
Q 55
Q 56
Q 16
32
exx
76
7
Q 56 54 cxf 5
cxg
Q 66
dU
Z
0
le
8
9T 2
ZZ >
< dexx >
= Q 11
6
dcxf
4 Q 15
A>
: dc >
;
Q 16
xg
Q 15
Q 55
32
3
exx
Q 16
76
7
Q 56 54 cxf 5dgdfdx:
Q 56
Q 66
cxg
cxg
cxf cxf g/;x /0
The main deference between the present study and Ref. [22], is that
in the present study, Eq. (4) is implemented after taking variation of
axial strain, exx , that results in Eq. (5) and keeps higher order terms,
which is important in large deection computations, but in Ref. [22]
these terms did not appear because Eq. (4) has been used before
taking variation of exx that yields to Eq. (6).
Integrating Eq. (3) over the cross-section gives the modulus
weighted section constants, which are presented in Ref [22]. These
section constants can be calculated using a separate, two-dimensional linear FEM analysis of an arbitrarily shaped composite crosssection which is decoupled from the nonlinear, one-dimensional global analysis for the beam. However, in this study, an improved nite
element cross-sectional analysis code (VABS) [25] is used. Heres
how to use it to provide the required cross-sectional properties for
the beam analysis. The proper usage of VABS outputs in the structural
model is explained here concisely (for more details see Ref. [28]).
From Ref. [26], the VABS strain energy is given by
8 9T 2
exx 9
exx > S11 S12 S13 S14 S15 S16 38
>
>
>
>
>
>
>
>
>
>
c > 6 S21 S22 S23 S24 S25 S26 7>
c >
>
>
Z le >
< xg >
< xg >
= 6
=
7>
c
S
S
S
S
S
S
c
6 31
V
xf
32
33
34
35
36 7
xf
2U
dx
7
6
S41 S42 S43 S44 S45 S46 7>
jx >
0 >
> jx >
> 6
>
>
>
>
>
5>
4
>
>
>
>
j
S
S
S
S
S
S
j
>
>
>
>
51
52
53
54
55
56 : g ;
: g;
jf
jf
S61 S62 S63 S64 S65 S66
8 9T
8 9
exx >
exx >
>
Z le >
< =
jx e A j B j C j D< jx =dx
2
xx
x
g
f
>
0 >
: jg >
: jg >
;
;
jf
jf
121
where the elastic twist is given by jx, while jg and jf are the moment strains corresponding to bending. The S, A, B, C, and D matrices are in the output list of VABS.
Using the strain energy relations given in Eq. (3) for the present
formulation, and the corresponding Eq. (7) for VABS, a direct comparison of the cross-sectional constants associated with both equations can be conducted. In order to couple VABS to the present
model, the cross-sectional parameters in the present strain energy
formulation are replaced with their VABS counterparts. Instead
VABS accounts for in-plane stresses and out-of-plane warping in
the H, A, B, C, and D matrices and so this hybrid strain energy will
be accurate for modeling of composite beams (for more details see
Ref. [28]).
n
o
2
L 0:5aqA b U_ f0 xA 0:5bh aqA bU g0
(
)
n
X
_
U f0 h 0:5U f0 b xA h
c Bi
i
i1
M 0:5aqA b
dT
le
ZZ
qV dVdgdfdx
V R_
10
^e0x
^e0g
^e0f
T
Tde ^ex
^ey
^ez
1=8b xA 0:5bh
aqA bxA U g0
(
_
U f0 h 0:5U f0 b xA h
)
2
T
11
n
X
ci Bi
D C d qA bU R C d qA bU 2g0 U 2f0
v ;x
w;x
where
13
dW e
le
P du Q d~
hdx
14
17
6
7
cosh0 /
sinh0 /
Tde 4 v ;x cosh0 / w;x sinh0 /
5
0
0
v ;x sinh0 / w;x cosh0 / sinh0 / sc cosh0 cosh0 / sc sinh0
16
15
i1
12
122
8
9 8 9
EA
>
> >0>
>
<Vx >
= < =
v_ ;
V EA
y
>
>
> >
>
: EA >
; :w
_;
Vz
8 9 8
8
9
8
9
9
A
EA
A
>
>
> > 0 >
>
>
>
>
< U x0 >
<Vx >
<Vx Vx >
=
=
= <
=
A
U g0 Tde V EA
and
V Ay V F
V
y
y
>
>
>
>
>
>
>
>
: 0;
>
>
>
: A>
: EA
; : 0 ;
;
Uf
Vz
V z V Az
v fUc gT fVg;
18
B_ i bi V F =bBi U f0 b xA h_
19
where ci = VFbiai/b.
The constants ai and bi are the coefcients used in the quasipolynomial approximation of the Wagner function that for the rst
and second states are
LT L Lstall ;
Lstall bqu2 C L2
2
L2 at s C_ L2 rC L2
t 2s C
21
where ts = b/U.
The parameters DCL and DCM are the deviation from the extended linear force curve (Fig. 4). Nonlinearity in the ONERA model
arises from Eq. (21) due to the dependence of its coefcients
(a, r, e) on DCL. These parameters must be identied for a special
airfoil.
3. Solution methodology
As it mentioned before the nite element method is implemented in this study for solving the system of aeroelastic equations. Therefore, the wing is divided into several beam elements.
The discretized form of the Hamiltons principle is written as
t2
t1
n
X
dU i dT i dW ei dt 0
23
Each beam element consists of two end nodes and one internal node
at its mid-point, which results in 17 nodal degrees of freedom, as
shown in Fig. 5. Thus,
T
T
fVg V 1 V 1;x V 2 V 2;x ; fWg W 1 W 1;x W 2 W 2;x ; fUg /1 /2 /3 T
xg1
fCg g c
24
h
q fVgT
fWgT
fUgT
fCg gT
fCf gT
iT
25
Since
the
variation
of
the
generalized
coordinates
xg ; dc
xf are arbitrary over the time interval, therefore
dv ; dw; d/; dc
dq is also arbitrary; and this results in the nite element equations
of motion for the ith beam element, which is written as
g Ki fqg fFi g 0
Mi fq
C M2 DC M
/ fUq gT fUg
20
where CL2 and CM2 are additional 2-dimensional lift and moment
coefcients due to stall which satisfy
@ DC L
r DC L t s e
a_
@a
w fUc gT fWg;
22
i1
where n is the total number of beam elements and dU, dT and dWe
are the variation of strain energy, kinetic energy, and virtual work of
external loads, respectively. The Hermitian shape functions are used
to discretize the space dependence: cubic polynomials for v and w;
26
where [M] is the structural mass matrix, [K] is the stiffness matrix
including linear structural stiffness matrix, nonlinear structural
stiffness matrix and the nonlinear aerodynamic stiffness matrix that
also is a function of the aerodynamic states. Also, the applied aerodynamic force vector, {F} is a nonlinear function of deections and
theirs derivatives with respect to time. So, it includes the aerodynamic damping terms.
After computing and assembling the mass, stiffness matrices
and force vector, the natural frequencies and related mode shapes
of the wing are rstly calculated. Hence, for the free vibration analysis, the equations of motion for total elements are
KS q 0
Mq
27
qi Q i y
28
The new unknowns of the problem, y, is the vector of the generalized modal coordinates and has a size of Nm, where Nm is the number of modes used to perform the modal coordinate transformation.
The columns of [Qi] correspond to the portions of the normal mode
eigenvectors for the ith element. The assembled matrices and load
vector of the wing are obtained as follows:
K
n
X
Q i T Ki Q i ;
C
n
X
Q i T Ci Q i ;
i1
M
n
X
Q i T Mi Q i ;
i1
29
i1
F
n
X
Q i T Fi ;
i1
where
30
123
Fig. 5. Wing nite element model and related nodal degrees of freedom.
33
{FB} is the additional force vector for modeling the unsteady aerodynamic (Eq. (19)) and stall aerodynamic (Eq. (21)) .The solutions of
Eq. (30) can be expressed in the form
X X 0 DX
34
Meq
My 0
0
(
1emfFeq g
;
"
Keq
_
_ y
; B; B
Ky; y;
0
_
_ y
; B; B
Fy; y;
_ C
_ C; C;
_ B; B;
FB y; y;
where
M @f=@ X
X0 ;0;0
;
31
B C gT
fBg
K @f=@XX0 ;0;0
36
Eq. (35), can be expressed in the rst order state variable form after
neglecting the higher order terms by
37
32
z
_
C @f=@ X
X0 ;0;0
z_ Az
fXg f y
35
and {C} is the ONERA stall aerodynamic states that has a size of n
and is dened as
38
"
T
. . . Bn1 Bn2 g
DX
DX_
0
1
I
1
M K M C
#
39
124
Table 1
Material Properties [39].
Parameter
Graphite/Epoxy
Styrofoam
97.3 Gpa
6.3 Gpa
5.3 Gpa
0.28
1540 kg/m3
0.135 mm
15 MPa
15 MPa
8 MPa
0.28
35 kg/m3
The stability of the system can be investigated through the eigenvalue analysis of A. Of course, these eigenvalues are complex conjugate pairs
kj fj ixj ;
j 1; . . . ; Nm
40
The wing is stable if all eigenvalues have the negative real parts.
Table 2
Comparison of the linear modal frequencies (Hz).
Composite layup
Mode number
Experiment [39]
Present analysis
Ref. [39]
Value
Value
FEM (NASTRAN)
[03/90]S
1st Bending
2nd Bending
1st Torsion
4.0
27.1
21.4
4.2
27.3
20.1
5.0
0.7
6.1
4.3
27.2
24.6
7.5
0.4
15
4.2
26.7
21.8
[152/02]S
1st Bending
2nd Bending
1st Torsion
3.6
27.1
22.7
3.8
27.4
21.8
5.5
1.1
4.0
3.9
28.6
23.5
8.3
5.5
3.5
3.8
26.0
22.0
Table 3
Comparison of linear aeroelastic results.
Composite layup
Present analysis
Ref. [39]
% Error
[03/90]S
28.40
11.21
28.45
28.2
11.86
28.20
0.5
5.5
0.9
[152/02]S
28.32
12.52
26.89
11.64
5.3
7.5
125
Fig. 11. Variation of utter speed with root angle of attack for [03/90]S wing.
Fig. 10. [03/90]S Wing torsional natural frequencies versus tip displacement.
experimental results in Ref. [39] and the nite element results obtained by NASTRAN. The obtained results, including the natural
frequencies of the rst two out of plane bending modes and the
rst torsion mode, show good agreement in comparison with available experimental data and FEM (NASTRAN). Table 3 compares the
obtained results for linear aeroelastic analysis including utter
speed, utter frequency and divergence speed with those exist in
Ref. [39]. It should be noted that the reported results in Ref. [39]
were obtained by applying the HodgesDowell structural model
with three bending and three torsion beam mode shapes and
implementation of the unsteady ONERA aerodynamic model.
4.2. Nonlinear results
To construct the eigenvalue analysis of the considered nonlinear aeroelastic model, in the rst stage a concentrated force and
moment is applied to the tip of the each aforementioned composite
wing which causes deformation of the wing. The deformation results have been shown in Figs. 8 and 9.
Fig. 10 reveals that the rst three torsional natural frequencies
obtained at these deformations for [03/90]S wing. In the second
stage, a steady angle of attack is added to the root of [03/90]S and
[152/02]S wings which causes deformation of the wings due to
aerodynamic loads. Thus, the aeroelastic system can be expressed
in the perturbed form about this deformation state (linearization
method). The favorite results, including nonlinear utter speed of
aeroelastic model obtained by the solution of the perturbed eigenvalue problem, are shown in Figs. 11 and 12. In order to comparison, the reported results by Dunn and Dugundji [39] are also
presented in these gures. It should be noted that Dunn and
Dugundji [39] used the HodgesDowell [2] equations that did
not include transverse shear deformations and some anisotropic
material coupling terms. These gures show good agreement
between present analysis and experiment results by [39]. Figs. 9
and 10 also show a better agreement between the present result
and the experimental data than that for Ref. [39]. It can be noted
that the effects of transverse shear deformations and composite
126
Fig. 12. Variation of utter speed with root angle of attack for [152/02]S wing.
References
[1] Hodges DH, Pierce GA. Introduction to structural dynamics and aeroelasticity,
Cambridge aerospace series book; 2011.
[2] Hodges DH, Dowell E. Nonlinear equations of motion for the elastic bending
and torsion of twisted non uniform rotor blades. NASA TN D-7818; 1974.
[3] Rosen A, Friedmann PP. The nonlinear behavior of elastic slender straight
beams undergoing small strains and moderate rotations. J Appl Mech
1979;46:1618.
[4] Dowell E, Traybar J, Hodges DH. An experimentaltheoretical correlation study
of nonlinear bending and torsion deformations of a cantilever beam. J Sound
Vib 1977;50:53344.