You are on page 1of 15

Journal of Fluids and Structures ( ) –

Contents lists available at ScienceDirect

Journal of Fluids and Structures


journal homepage: www.elsevier.com/locate/jfs

Structural dynamic reanalysis method for transonic


aeroelastic analysis with global structural modifications
Dongfeng Li, Qiang Zhou, Gang Chen *, Yueming Li
State Key Laboratory for Strength and Vibration of Mechanical Structures, School of Aerospace, Xi’an Jiaotong University, Xi’an, 710049,
China
Shaanxi Key Laboratory for Environment and Control of Flight Vehicle, School of Aerospace, Xi’an Jiaotong University, Xi’an, 710049,
China

article info a b s t r a c t
Article history: Transonic aeroelastic analysis is an important aspect of modern aircraft design. When
Received 24 October 2016 the modeshapes changed for the structural stiffness or/and mass matrix variation, the
Received in revised form 17 April 2017 structural model and aerodynamic model could be reconstructed to maintain accuracy.
Accepted 7 June 2017
However, these reconstruction procedures have to be repeated every time when there
Available online xxxx
is a change in the aircraft structure design, where each of these reconstructions takes a
Keywords: considerable time. In this paper, an automated procedure for transonic aeroelastic analysis
Transonic aeroelastic method was proposed when the structural model parameters changed at global level. To
Flutter prediction manage thousands of calculations and makes more efficient and reliable calculations, the
CFD/CSD coupled method structural dynamic reanalysis method was employed to set up an automated procedure to
Structural dynamic reanalysis link the aeroelastic solvers and structural design parameters without changing structural
Global structural modification modal coordinates. The proposed automated procedure was illustrated and verified by
the AGARD 445.6 wing aeroelastic model with different global level structural parameter
variations. The results show that the automated procedure can accurately produce pulse
responses corresponding to the mode excitations and the flutter characteristics of AGARD
445.6 wing in a very large modification range. The whole automated procedure is effi-
cient and reliable in managing transonic aeroelastic analysis when structural parameters
changed at global level.
© 2017 Published by Elsevier Ltd.

1. Introduction

High-performance modern aircrafts often cruise in transonic regime in order to achieve high efficiency (Rizk, 1981).
The transonic aeroelastic analysis is an important part and critical engineering challenge of modern aircraft design process
because wing flutter may lead to the catastrophe during flight. Flutter is a dynamic instability aeroelastic phenomenon
resulting from a coupling between unsteady aerodynamic forces and structural vibrations (Schuster et al., 2003). Aeroelastic
analysis is a multi-disciplinary problem which involves unsteady aerodynamic loads, inertial, and elastic forces of structures.
High-fidelity fluid model is the basic of accurate modeling of aeroelastics. Linear aeroelastic analysis methods for aircraft
aeroelastic stability analysis are usually based on the potential flow theory. In subsonic flow, most of them rely on the
Doublet-Lattice method (Albano and Rodden, 1969) for computing the unsteady aerodynamic response to modal motions.
Till now, it is still the most used method for aeroelastic design in subsonic unsteady flows. In supersonic aeroelastic analysis,
various methods associated with piston the theory (Ashley and Zartarian, 1956) are commonly used. However, in transonic

* Corresponding author.
E-mail address: aachengang@xjtu.edu.cn (G. Chen).

http://dx.doi.org/10.1016/j.jfluidstructs.2017.06.004
0889-9746/© 2017 Published by Elsevier Ltd.

Please cite this article in press as: Li, D., et al., Structural dynamic reanalysis method for transonic aeroelastic analysis with global structural modifications.
Journal of Fluids and Structures (2017), http://dx.doi.org/10.1016/j.jfluidstructs.2017.06.004.
2 D. Li et al. / Journal of Fluids and Structures ( ) –

aeroelastic analysis, the linear aerodynamic theories formally fail due to the presence of flow nonlinearities (e.g. shocks,
separation, vortices) (Dowell et al., 1983; Bendiksen, 2011). Transonic aerodynamics have to be modeled by nonlinear
equations for satisfactory results (Ashley, 1980; Sahu et al., 2014). Thus, the current treatment in the unsteady transonic
mostly adopts the computational fluid dynamics (CFD) method based on Euler or RANS models, due to its generality in
dealing with nonlinear aerodynamics (Guruswamy, 1990; Förster and Breitsamter, 2013).
The preliminary design play an essential role in an aircraft design process (Oroumieh et al., 2013). The overall outer shape
of aircraft structure is usually fixed in the early stages of modern aircraft preliminary structural design. In order to meet the
design targets, The structural model needs to be modified repeatedly by modifying the distribution of mass and stiffness
(Kamle et al., 2017). The structural mass and stiffness are particularly parameters in identifying structural modeshapes and
frequencies. Marques et al. (2012) have shown that the change of modeshapes and frequencies must be taken into account
for practical aeroelastic design. Hence, for modified structure model using a constant set of normal modes will produce an
inaccurate aeroelastic results. Unfortunately, most of the traditional standard flutter prediction methods are constructed
based on the prescribed modeshapes. In order to achieve an accurate aeroelastic results with a structural modification, both
structure and aerodynamic model have to be recalculated. Furthermore, these reconstruction have to be repeated manual
every time whenever a structure model parameter is modified in the aircraft structure. This approach called direct method
is widely used in practical engineering design.
Within the modern aircraft industry, there are thousands of combinations of possible structural parameters and each
of these calculation cycles take a considerable time. A more rational way to meet the design targets is to develop an
automated procedure (Murthy and Kaza, 1989; Styuart et al., 2011; Schwochow and Jelicic, 2015) that can manage thousands
of calculations within a reasonable cost in time and labor. Picardi et al. (2014) used the substructuring approach to set up
an automated procedure for flutter clearance assessment of aircraft under damage during flight. The main challenge of the
automated procedure is to create an efficient link between the aeroelastic code and the design parameters. To create the
link, a more practical approach is to introduce the structural changes without changing the modal coordinates (Karpel and
Wieseman, 1994a; Pak and Li, 2010) when structural model parameters are changed. Karpel (Karpel and Wieseman 1994b)
applied the fictitious mass method to generate fixed-coordinate time-domain aeroelastic models for dynamic response
involving large local structural changes.
However, most of the above studies focused on the local structural modifications, ignoring the structural modifications
made at global level. Indeed, in the procedure of aircraft structure design, two levels of procedure need to be considered:
global level and local level (Ciampa et al., 2010). As the first step, the global level analysis initializes all the global quantities
and responses, and then provides information to each of the local level subproblem. At global level, global modifications
is taken into account by varying the mass and stiffness of wing to meet design targets (Gasbarri et al., 2010). It is very
useful to find an efficient automated procedure to rapidly obtain the new aeroelastic results with good accuracy and manage
thousands of calculations in transonic aeroelastic analysis, when structural model parameters are changed at global level.
Rather than reconstructing new structure and aerodynamic models as the direct method when structural model
parameters are changed at global level, we propose an new automated approximate aeroelastic method by introducing
the approximate structural dynamic reanalysis method. The present study is not only to set up an automated procedure
to between the aeroelastic codes and design parameters, but also to reduce the problem size and computation time. The
new proposed automated procedure would be expect to have the great potential to support the transonic aeroelastic
design optimization. The paper is constructed as follows: Section 2 give the brief introduction of the standard construction
procedure for transonic aeroelastic systems; Section 3 give a description on the approximate structural dynamic reanalysis
method, and the automated procedure with global structural parameter variation is described. In Section 4 the method is
validated by the AGARD 445.6 aeroelastic wing model with three different structural parameter variation cases, and the
results of the approximate method are compared with the direct method. Section 5 is the conclusion and discussion of the
proposed automated procedure.

2. The aeroelastic problem in transonic regime

2.1. Unsteady aerodynamic model

The aerodynamic model used here is based on a finite volume discretization of the unsteady Euler equations, attention is
limited to inviscid compressible flows. However, this method is also applicable to the Navier–Stokes equations. The flow is
governed by the three-dimensional time dependent Euler equations. The equations in Cartesian coordinates are given by
∂U ∂F ∂G ∂H
+ + + =0 (1)
∂t ∂x ∂y ∂z
where U is the vector of conserved variables, F, G and H are the inviscid flux vectors, which are given by
ρ ρu ρv ρw
⎡ ⎤ ⎡ ⎤ ⎡ ⎤ ⎡ ⎤
⎢ ρu ⎥ ⎢ ρ uu + p ⎥ ⎢ ρ uv ⎥ ⎢ ρ uw ⎥
U = ⎢ ρv ⎥ , F = ⎢ ρ uv ⎥, G = ⎢ ρvv + p ⎥ , H = ⎢ ρvw ⎥.
⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥
⎣ρw⎦ ⎣ ρ uw ⎦ ⎣ ρvw ⎦ ⎣ ρww ⎦
ρE u (ρ E + p) v (ρ E + p) w (ρ E + p)
Please cite this article in press as: Li, D., et al., Structural dynamic reanalysis method for transonic aeroelastic analysis with global structural modifications.
Journal of Fluids and Structures (2017), http://dx.doi.org/10.1016/j.jfluidstructs.2017.06.004.
D. Li et al. / Journal of Fluids and Structures ( ) – 3

Fig. 1. Flow chart of CFD/CSD couple algorithm.

This study is based on ideal gas model with a constant specific heat ratio, and the pressure p is represented by the following
state equation:
[ ]
1( 2
p = ρ (γ − 1) E − u + v 2 + w2
)
2
where γ is the ratio of specific heats. The multi-block structured mesh based unsteady Euler solver uses the cell-centered
finite volume formulation, and the spatial discretization is based on the second-order Van Leer scheme (Van Leer, 1979).
For time marching, dual time-stepping (Pulliam, 1993) and lower–upper symmetric Gauss–Seidel method (Chen and Wang,
2000) are used.

2.2. Structural model

In general, linear structural model is a common practice as the generalized coordinates in transonic aeroelastic analysis
(Platten et al., 2007). The basic assumption of the modal approach is that the physical structural displacement of the aeroelas-
tic system are linear combinations of [ a limited]set of low-frequency modeshapes. The physical structural displacement can
be written as d = Φu, where Φ = ϕ1 , ϕ2 , . . . is the modeshapes calculated from a finite-element model of the structure
with the deflections defined at the structural grid points, and u is the generalized displacements. For the stability analysis
such as the flutter prediction, the structural damping usually can be neglected (Schuster et al., 2003). Thus, the equations of
the motion neglected structural damping are described by a set of modal coordinates:

Mü (t ) + Ku (t ) = fgen (t ) (2)

where M = ΦT MΦ and K = ΦT KΦ are the generalized mass and stiffness matrices, respectively. fgen (t ) is the vector of
generalized aerodynamic forces (GAF) (Winter and Breitsamter, 2014) which can be written as:

fgen (t ) = ΦT f (t ) (3)

f(t) is the vector of external forces acting on the structure. The generalized aerodynamic forces in the relation for the ith
generalized aerodynamic force vector element in the time domain:

fgen,i (t ) = q∞ · cp (t ) · ϕi ◦ dS (4)
S

where cp (t ) is pressure coefficient, q∞ is freestream dynamic pressure.

2.3. Fluid–structure interaction procedure

For the time-domain fluid–structure coupling, a fully implicit partitioned approach is employed. The flow and structure
model are coupling together to exchange aerodynamic loads and elastic deformations on the fluid–structure interface at the
pseudo-time stepping iterations (Jameson, 1991). In Fig. 1, a flow chart of the fluid–structure coupling process is presented.
The transonic aeroelastic calculation was started from a converged steady-flow solution about a rigid structure. First, the
aerodynamic loads are transferred into structural domain through Infinite Plate Spline (IPS) method (Harder and Desmarais,
1972). Then, the structural elastic deformations are computed by the computational structural dynamics (CSD) solver. Next,

Please cite this article in press as: Li, D., et al., Structural dynamic reanalysis method for transonic aeroelastic analysis with global structural modifications.
Journal of Fluids and Structures (2017), http://dx.doi.org/10.1016/j.jfluidstructs.2017.06.004.
4 D. Li et al. / Journal of Fluids and Structures ( ) –

the elastic deformations are mapped to the CFD surface mesh through the IPS method. At this time, update volume mesh
using the Radial Basis Function (RBF) method combined with transfinite interpolation (TFI) algorithm (Tsai et al., 2001).
The aerodynamic loads are then recalculated on the updated CFD volume mesh. This loop is repeated until the variation
of the deflections of the interface nodes convergence. The accuracy of the aeroelastic solver had been validated thoroughly
and applied to several two-dimensional and three-dimensional complex aeroelastic models, where the more details can be
founded in our further publications (Chen et al., 2012; Zhou et al., 2014, 2016).

3. Automated procedure based approximate dynamic reanalysis

3.1. Approximate structural dynamic reanalysis method

It is very important to recompute the modeshapes and frequencies of modified structures in the practice of structural
optimization. For each modification, the multiple repeated modal analyses usually involve extensive computational effort,
especially for large scale problems. Thus, it is necessary to seek an efficient computational method for structural reanalysis.
Many methods have been proposed to reanalysis the modeshapes and frequencies with high efficiency for the modified
structures without performing the full analysis. Generally, these methods can be classified into two categories: direct
method and approximate method (Song et al., 2014). Most direct methods are based on Sherman–Morrison–Woodbury
(SMW) formulas (Akgün et al., 2001), which is usually suitable for large but local (or low-rank) modifications. However
these methods limited by scale of the problem and rank of the modification, and not suitable for global modification. These
methods also limited in practical application because these methods requires a large amount of computation. Recently,
Khodaparast (Khodaparast and Cooper, 2014) presented a linear aeroelastic problem of an aircraft structure under gust
loads which is shown that the SMW formula can be used directly for efficient gust reanalysis of the modified structure for
low-rank structural modifications.
There are various kinds methods had been proposed (Beliveau et al., 1996; Chen and Yang, 2000; Kirsch et al., 2007) for
structural dynamic reanalysis of the modified structure where the modeshapes of the modified structure are approximated
as a linear combination of the modeshapes of the original structure. The most advantage of approximate methods can give
good approximate modeshapes and frequencies for global (or high-rank) modifications of the structural parameters with
higher computation efficiency. On the approximate reanalysis methods, the extended Kirsch combined method (Chen et
al., 2000; Chen and Yang, 2000) is an special efficient approach for global structural modifications. The extended Kirsch
combined method used the second-order perturbation terms of eigenvectors as the basis vectors to obtain the modeshapes
and frequencies of the modified structure with less computational effort (Chen et al., 1994). The extended Kirsch combined
method would be selected as the structural reanalysis module in our proposed automated aeroelastic procedure. More details
about this method are described as the follows.
[ 1Consider ] an original structure with stiffness matrix K0 and mass matrix M0 . The corresponding modeshapes Φ0 =
ϕ0 , ϕ20 , . . . and modal frequencies λi0 are calculated by solving the generalized eigenproblem:

K0 ϕi0 = λi0 M0 ϕi0 (5)


when K0 and M0 are perturbed into the form K0 + 1K and M0 + 1M, in which 1K and 1M are the perturbations in the
stiffness matrix and mass matrix, respectively. The generalized eigenproblem of modified structure can be written as

Kϕi = λi Mϕi (6)


where λ and ϕ are ith eigenvalue and eigenvector of the modified structure, respectively.
i i

The extended Kirsch combined method use the second-order eigenvector terms as the basis vectors in the following
modeshapes reduced basis:

ϕi = ϕiB yi (7)
where
ϕiB = ϕi0 , ϕi1 , ϕi2
[ ]

yi = (y1 , y2 , y3 )T ∈ R3×1
where ϕi1 and ϕi2 are the ith first-order and second-order perturbations of original eigenvector (modeshapes), respectively.
The two perturbations of original eigenvector can be obtained as follow:
n
∑ 1 [( ) (
s T
) ] 1 [( i )T ]
ϕi1 = ϕ0 1K − λi0 1M ϕi0 ϕs0 − ϕ0 1Mϕi0 ϕi0 = ϕ0 Yi1 (8)
s=1,s̸ =i
λ −λ
i
0
s
0
2

n
∑ 1 [( ) (
T ( )T ( )]
ϕi2 = ϕs0 1K − λi0 1M ϕi1 − λi1 ϕs0 M0 ϕi1 + 1Mϕi0 ϕs0
)
s=1,s̸ =i
λ −λ
i
0
s
0 (9)
1 [( i )T ( )T ( )]
− ϕ0 1Mϕi1 + ϕi1 M0 ϕi1 + 1Mϕi0 ϕi0 = ϕ0 Yi2 .
2
Please cite this article in press as: Li, D., et al., Structural dynamic reanalysis method for transonic aeroelastic analysis with global structural modifications.
Journal of Fluids and Structures (2017), http://dx.doi.org/10.1016/j.jfluidstructs.2017.06.004.
D. Li et al. / Journal of Fluids and Structures ( ) – 5

Fig. 2. Flowchart of the automated procedure.

Substituting Eqs. (8) and (9) into Eq. (7), Φ can be written as
⎤T ⎡
Φ10 . . . Φi0 I0 . . . Ii0
⎡ ⎤⎡ 1 ⎤
0 0 0 0 y1 0 0
Φ=⎣ 0 Φ10 . . . Φi0 0 ⎦⎣ 0 Y11 . . . Yi1 0 ⎦ ⎣0 y2 0 ⎦ = Φ0 Y (10)
0 0 Φ10 . . . Φi0 0 0 Y12 . . . Yi2 0 0 y3
where
⎤T ⎡
I0 . . . Ii0
⎡1 ⎤
0 0 y1 0 0
Q= ⎣ 0 Y11 . . . Yi1 0 ⎦ ⎣0 y2 0⎦. (11)
0 0 Y12 . . . Yi2 0 0 y3
( )T
Substituting Eq. (7) into the modified analysis equations Eq. (6), and premultiplying by ϕiB , we can obtain a set of
(3 × 3) matrix equation:

KiR yi = λi MiR yi (12)


where
( )T
KiR = ϕiB (K0 + 1K) ϕiB (13)
( i )T
MiR = ϕB (M0 + 1M) ϕiB . (14)
i
Thus, y can be calculated from Eq. (12) which is then substituted into Eqs. (11) and (10), then we can obtain a coefficient
matrix Y and the approximate eigenvectors Φ. Finally, using the Rayleigh quotient, the ith approximate eigenvalue λiK are
computed from the following equation:
( i )T
ϕ (K0 + 1K) ϕi
λiK = ( )T . (15)
ϕi (M0 + 1M) ϕi
To summarize, the extended Kirsch combined method consists of the following steps:

(1) Run the eigenproblem of original structure, and perturb the structural parameters to form the modified stiffness
matrices K0 + 1K and mass matrices M0 + 1M;
(2) Compute first-order and second-order perturbations of original eigenvector using Eqs. (8) and (9);
(3) Form the reduced stiffness matrix KiR and the reduced mass matrix MiR using Eqs. (13) and (14);
(4) Use Eq. (12) to calculate the coefficients yi ;

Please cite this article in press as: Li, D., et al., Structural dynamic reanalysis method for transonic aeroelastic analysis with global structural modifications.
Journal of Fluids and Structures (2017), http://dx.doi.org/10.1016/j.jfluidstructs.2017.06.004.
6 D. Li et al. / Journal of Fluids and Structures ( ) –

Fig. 3. Structural model (left) and surface CFD grid (right) of AGARD 445.6.

Fig. 4. Flutter boundaries vary with Mach number.

(5) Substitute yi into Eq. (11) to get a constant matrix Y and use Eq. (10) to compute the approximate eigenvectors Φ;
(6) Finally, compute the approximate eigenvalue λiK using Rayleigh quotient by Eq. (14).

3.2. Application of extended Kirsch combined method for aeroelastic analysis

The modeshapes Φ0 of the original structure is taken as the basic modeshapes for basic aeroelastic system. For modified
structure, substituting Eq. (10) into Eq. (3), the generalized aerodynamic force is written as

fgen (t ) = (Φ0 Y)T f0 (t ) = YT fgen,0 (t ) (16)

where fgen,0 (t ) = (t ) is the generalized aerodynamic force of original structure.


ΦT0 f0
The structural dynamic equations of modified structure are:

Mü (t ) + Ku (t ) = fgen (t ) (17)

here, M = YT ΦT0 (M0 + 1M) Φ0 Y, K = YT ΦT0 (K0 + 1K) Φ0 Y. In addition, when perturbations of stiffness matrix and mass
matrix is zero matrix, Y is the unit matrix, the aeroelastic system is equivalent to original aeroelastic system.

Please cite this article in press as: Li, D., et al., Structural dynamic reanalysis method for transonic aeroelastic analysis with global structural modifications.
Journal of Fluids and Structures (2017), http://dx.doi.org/10.1016/j.jfluidstructs.2017.06.004.
D. Li et al. / Journal of Fluids and Structures ( ) – 7

Fig. 5. First four modeshapes of the AGARD 445.6 wing: (a) modeshapes and frequencies of original structure, (b) exact modeshapes and frequencies for
large modification (ε = 1/3) obtained by full modal analysis, (c) approximate modeshapes and frequencies for large modification (ε = 1/3) obtained by
extended Kirsch combined method.

The flowchart of the automated procedure is illustrated in detail in Fig. 2, the automated procedure can be outlined step
by step as follows:

(1) According to the original structure, build basic aeroelastic system based on the basis modeshapes Φ0 ;
(2) Update the structural model, and use extended Kirsch combined method to get a coefficient matrix Y and an updating
modeshapes Φ;
(3) Approximate generalized aerodynamic force fgen (t ) is computed based on fgen,0 (t ) and coefficient matrix Y using
Eq. (16);
(4) Finally, combine with the structural dynamic solver Eq. (17) to compute the responses corresponding to the updating
modeshapes Φ.

The responses corresponding to the updating modeshapes can be computed by repeating steps 1–4, when the global
structural parameters are modified. The automated procedure is fully automated so that it is very convenient for aeroelastic

Please cite this article in press as: Li, D., et al., Structural dynamic reanalysis method for transonic aeroelastic analysis with global structural modifications.
Journal of Fluids and Structures (2017), http://dx.doi.org/10.1016/j.jfluidstructs.2017.06.004.
8 D. Li et al. / Journal of Fluids and Structures ( ) –

(a) Generalized force of first bending mode. (b) Generalized force of first torsion mode.

(c) Generalized force of second bending mode. (d) Generalized force of second torsion mode.

Fig. 6. Comparison of generalized aerodynamic forces obtain by two methods for large modification (ε = 1/3).

Table 1
Error of modal frequencies and MAC for the first four modes.
Parameter Mode Exact (Hz) Approximate (Hz) Error (%) MAC
1 10.208 10.207 −0.005 1.0000
2 41.228 41.232 0.010 0.9999
ε = 1/12
3 50.756 50.792 0.071 0.9998
4 95.962 96.034 0.076 0.9991
1 10.854 10.856 0.021 1.0000
2 42.457 42.470 0.030 0.9998
ε = 1/6
3 51.819 51.930 0.220 0.9993
4 96.633 96.873 0.248 0.9970
1 11.933 11.943 0.082 0.9999
2 44.354 44.398 0.101 0.9994
ε = 1/3
3 53.591 53.894 0.565 0.9991
4 97.712 98.412 0.716 0.9911

analysis with various modified structure. In the procedure, the CFD calculation is the most expensive, however for its
automatic capability the total calculation and tedious manual work time can be obviously cut down.

4. Application and numerical examples

4.1. CFD/CSD coupled solver validation

Before demonstrating the efficiency and accuracy of the automated procedure for transonic aeroelastic analysis with
global structural parameter variations, the present CFD/CSD solver is validated firstly using the AGARD 445.6 aeroelastic

Please cite this article in press as: Li, D., et al., Structural dynamic reanalysis method for transonic aeroelastic analysis with global structural modifications.
Journal of Fluids and Structures (2017), http://dx.doi.org/10.1016/j.jfluidstructs.2017.06.004.
D. Li et al. / Journal of Fluids and Structures ( ) – 9

(a) Generalized displacements of first bending mode.

(b) Generalized displacements of first torsion mode.

Fig. 7. Comparison of time histories of generalized displacements obtain by two methods for ε = 1/12 at various dynamic pressures: V∞ = 274 m/s (left),
V∞ = 334 m/s (right) (Ma = 0.960, AOA = 0◦ ).

wing model (Yates, 1987). The AGARD 445.6 standard aeroelastic configuration with a 45◦ quarter-chord sweep angle, and
a panel aspect ratio of 1.6525, and a taper ratio of 0.6576 with a symmetrical NACA65A004 airfoil section. The wing material
properties were with the density of 381.98 kg/m3 . The elastic modulus in the span-wise direction and the chord-wise
direction is 3.151 and 0.416 GPa, respectively. The shear modulus is 0.4392 GPa. The Poisson’s ratio of 0.31 (Zhong and
Xu, 2016). The structural model, shown in Fig. 3(a), consists of 231 nodes and 200 elements. The thickness distribution of
the wing was governed by airfoil shape.
A multi-block structured mesh was employed for the flow solver, consisting of 61 computational nodes around each
airfoil section and 20 nodes along the wing semispan, as shown in Fig. 3(b). The spatial convergence of the CFD mesh was
analyzed by Zhou et al. (2016), which reported a good agreement of‘ the results from both the medium and fine grids. The
total number of computational nodes for this CFD mesh is 456,365 (119 × 59 × 65). Fig. 4 shows the flutter speed predicted
by the our coupled CFD/CSD solver and others results. For comparison, experimental data (Yates, 1987) and Lee-Rausch’s
numerical results (Lee-Rausch and Batina, 1995) results are also included. The good agreement of our CFD/CSD solver with
the existing experimental and numerical results for different Mach numbers (0.499 to 1.141), indicted the accuracy of our
solver, in which the well-known transonic dip of the flutter speed was also captured (Silva et al., 2014). Indeed, the accuracy
of the CFD/CSD coupled solver had been thoroughly evaluated in our precious aeroelastic investigations (Chen et al., 2012;
Zhou et al., 2014, 2016).

4.2. Accuracy evaluation of the structural dynamic reanalysis method

The numerical example is based on AGARD 445.6 aeroelastic wing model, the structural model of the AGARD 445.6 wing
is divided into four sections along spanwise, as shown in Fig. 3(a). Each section with 50 elements and it has same material
properties. Some assumptions have been made to allow variation of stiffness and mass in the wing. To obtain global variations
of the mass and stiffness distribution, material properties are assumed to vary as
Section 1, E1 = (1 + 3ε) E0 , ρ1 = (1 + 3ε) ρ0
Section 2, E2 = (1 + 2ε) E0 , ρ2 = (1 + 2ε) ρ0

Please cite this article in press as: Li, D., et al., Structural dynamic reanalysis method for transonic aeroelastic analysis with global structural modifications.
Journal of Fluids and Structures (2017), http://dx.doi.org/10.1016/j.jfluidstructs.2017.06.004.
10 D. Li et al. / Journal of Fluids and Structures ( ) –

(c) Generalized displacements of second bending mode.

(d) Generalized displacements of second torsion mode.

Fig. 7. (continued)

Section 3, E3 = (1 + ε) E0 , ρ3 = (1 + ε) ρ0
Section 4, E4 = E0 , ρ 4 = ρ0
where E0 and ρ0 are the Young’s modulus and density of baseline structural model, respectively. It is worth observing that
these assumptions may not be completely accurate in physical terms, but for the demonstration purposes of this research,
a well define and simple model is completely essential.
To assess the accuracy of the approximate modal characteristics obtain by extended Kirsch combined method when
structural parameters changed at global level, two criteria are introduced. The first represents the error of the modal
frequencies:
fAi − fEi
Error =
fEi
where fEi denotes the exact modal frequencies and fAi denotes approximate modal frequencies obtain by full modal
analysis and extended Kirsch combined method, respectively. The second criterion is the Modal Assurance Criterion (MAC)
(Ewins, 1984) of modeshapes, which is written as
2
|ΦT ΦA |
MAC (ΦA , ΦE ) = ( T E) ( T )
ΦA ΦA ΦE ΦE
where ΦE represents the exact modeshapes and ΦA represents the approximate modeshapes obtain by full modal analysis
and extended Kirsch combined method, respectively. For two fully identical modeshapes, the value of MAC is identically 1,
whereas for two unrelated mode shapes, a MAC value close to 0.
Here the first four modes of the AGARD 445.6 aeroelastic wing model are used in the analyses. There are three modified
cases are considered: ε = 1/12, 1/6, 1/3. The modeshapes and associated frequencies for the original and modified AGARD
445.6 wing are shown in Fig. 5. These modeshapes are identified as the first bending, first torsion, second bending, and second
torsion. The frequencies errors and MACs between approximate and exact are listed in Table 1 for the three cases. For all cases,

Please cite this article in press as: Li, D., et al., Structural dynamic reanalysis method for transonic aeroelastic analysis with global structural modifications.
Journal of Fluids and Structures (2017), http://dx.doi.org/10.1016/j.jfluidstructs.2017.06.004.
D. Li et al. / Journal of Fluids and Structures ( ) – 11

(a) Generalized displacements of first bending mode.

(b) Generalized displacements of first torsion mode.

Fig. 8. Comparison of time histories of generalized displacements obtain by two methods for ε = 1/6 at various dynamic pressures: V∞ = 293 m/s (left),
V∞ = 353 m/s (right) (Ma = 0.960, AOA = 0◦ ).

Table 2
Comparison of CPU time (s).
Parameter Full modal analysis Extended Kirsch combined method
ε = 1/12 1.122 0.556
ε = 1/6 1.154 0.586
ε = 1/3 1.106 0.600

the errors of modal frequencies not more than 1%, even for large modification (ε = 1/3). For MACs, the values of the four
modeshapes are all closest to 1, it indicate a good to excellent agreement between the exact and approximate modeshapes.
All of these indicate that the extended Kirsch combined method can give enough accurate approximate modeshapes and
frequencies, without performing a full modal analysis for each modified structural case.
The comparison of CPU time of the full modal analysis and extended Kirsch combined method is listed in Table 2.
Table 2 only collected solution times by the two methods, excluding the costs of assembling stiffness and mass matrices.
All numerical examples are tested on Windows 7 system with Intel⃝ R
Core(TM) i7-2600 CPU (3.40 GHz, 8 cores, but only one
core used) and 16 GB RAM. As can be seen, the computational efficiency of the extended Kirsch combined method is higher
than full modal analysis.

4.3. Accuracy evaluation of the approximate aeroelastic method

After evaluation the accuracy of CFD/CSD coupled solver and the extended Kirsch combined method for AGARD 445.6
wing, which are both the key procedures of new approximate aeroelastic method, the whole automated procedure will be
evaluated in this section. The time histories of the generalized aerodynamic force (GAF) predicted by the direct method and
approximate method are shown in Fig. 6. For brevity, only the responses with the largest structural parameter modification
(ε = 1/3) at the Mach number of 0.960, a critical point of the transonic dip, with zero angle of attack (AOA) is shown
in Fig. 6. As can be seen, the GAF responses predicted by direct method and approximate method agree fairly well. The

Please cite this article in press as: Li, D., et al., Structural dynamic reanalysis method for transonic aeroelastic analysis with global structural modifications.
Journal of Fluids and Structures (2017), http://dx.doi.org/10.1016/j.jfluidstructs.2017.06.004.
12 D. Li et al. / Journal of Fluids and Structures ( ) –

(c) Generalized displacements of second bending mode.

(d) Generalized displacements of second torsion mode.

Fig. 8. (continued)

Table 3
Results of flutter analysis obtained by direct method and approximate method at Ma = 0.960.
Parameter Flutter speed (m/s) Flutter frequency (Hz)
Direct Approximate Error (%) Direct Approximate Error (%)
ε = 1/12 304 306 0.658 14.045 14.085 0.282
ε = 1/6 323 326 0.929 14.903 14.948 0.299
ε = 1/3 349.6 358 2.403 16.129 16.287 0.978

well agreements indicate that the proposed approximate method can produce the aerodynamic responses corresponding
to the mode excitations with high accuracy when global changes of structural parameters are made, even for very large
modification (ε = 1/3).
For further demonstrating the effectiveness and accuracy of the automated procedure, two typical responses (decaying
and diverging) of time histories of the generalize displacement for the three different structural parameter variation cases
under different free stream dynamic pressures are compared as shown in Figs. 7–9. It can be seen that in all the three different
structural modification cases, both the divergent and convergent of time histories of the generalize displacement are well
consistent. It indicates again that the automated procedure has good accuracy for aeroelastic response prediction, even for
very large range (ε = 1/3) of structural parameter variation.
For these low V∞ , Fig. 7 (left), Fig. 8 (left) and Fig. 9 (left) show that all of four modes amplitude of generalized
displacements decay with time marching. At a higher V∞ , the amplitudes of all the four modes would diverge with time
marching are shown in Fig. 7 (right), Fig. 8 (right) and Fig. 9 (right). Between these two V∞ conditions, there must have a
particular point where the system is neutrally stable. In the given flight conditions (Mach and AOA), these particular points
is called flutter speed, are shown in Table 3. Table 3 provides a comparison of the flutter speeds and frequencies obtained
by direct method and approximate method. It shown that the results agree well with each other. These results indicate that
automated approximate method can predict the flutter speeds and frequencies with good accuracy.

Please cite this article in press as: Li, D., et al., Structural dynamic reanalysis method for transonic aeroelastic analysis with global structural modifications.
Journal of Fluids and Structures (2017), http://dx.doi.org/10.1016/j.jfluidstructs.2017.06.004.
D. Li et al. / Journal of Fluids and Structures ( ) – 13

(a) Generalized displacements of first bending mode.

(b) Generalized displacements of first torsion mode.

Fig. 9. Comparison of time histories of generalized displacements obtain by two methods for ε = 1/3 at various dynamic pressures: V∞ = 320 m/s (left),
V∞ = 380 m/s (right) (Ma = 0.960, AOA = 0◦ ).

5. Conclusions

Traditional standard aeroelastic analysis methods are based on the prescribed modeshapes. In order to maintain the
accuracy, both the structure model and aerodynamic model must be reconstructed when a modification is made to the
structure model. Each of these modifications will spend a considerable of tedious manual deal with time. In the practical
aircraft design, the structural model needs to be modified repeatedly to meet the design targets by modifying the distribution
of mass and stiffness, which would be an terrible task for manual direct method. In order to deal with such problem,
an automated approximate aeroelastic reanalysis procedure was proposed based on the approximate dynamic reanalysis
method for transonic aeroelastic analysis with global structural parameters changes.
The automated procedure has been validated by the AGARD 445.6 wing model with global structural parameter modifi-
cations. Firstly, the accuracy and efficiency of the extended Kirsch combined structural reanalysis method was evaluated by
the approximate modal characteristics with global level structural parameters changes. And then the different aeroelastic
time responses in transonic flows with three different structural parameter modification cases were compared with the
direct method in different free steam velocities, including the time responses associated with generalized aerodynamic
force and the generalized displacements. Finally the flutter speeds and frequencies were also compared for the three
different structural parameter modification cases. The good agreements of the simulation show that the proposed automated
aeroelastic prediction procedure can capture the aeroelastic characteristics of the AGARD 445.6 wing with good accuracy
when the global changes of the structural parameters are made.
The main advantage of the proposed automated aeroelastic prediction method is not only be able to provide an accurate
approximate results without changing structural modal coordinates at higher efficiency, but also the whole automated
procedure is efficient in managing the link between aeroelastic codes and structural design parameters, which might cut
down the great time consuming in the manual efforts in the direct method. The automated procedure has the great potential
to support the transonic aeroelastic design optimization where the manual work are not permitted. In next step we will
further extend and apply the proposed automated aeroelastic reanalysis method in the flutter optimization by combined
with the reduced order modeling for the nonlinear aeroelastic problems.

Please cite this article in press as: Li, D., et al., Structural dynamic reanalysis method for transonic aeroelastic analysis with global structural modifications.
Journal of Fluids and Structures (2017), http://dx.doi.org/10.1016/j.jfluidstructs.2017.06.004.
14 D. Li et al. / Journal of Fluids and Structures ( ) –

(c) Generalized displacements of second bending mode.

(d) Generalized displacements of second torsion mode.

Fig. 9. (continued)

Acknowledgments

This work was partially supported by the National Natural Science Foundation of China (11672225, 11272005), National
Key Foundation (MJ-2015-F-010), Shaanxi Natural Science Foundation (2016JM1007) and the Funds for the Central Univer-
sities (xjj2014135).

References

Akgün, M.A., Garcelon, J.H., Haftka, R.T., 2001. Fast exact linear and non-linear structural reanalysis and the sherman–morrison–woodbury formulas.
Internat. J. Numer. Methods Engrg. 50, 1587–1606.
Albano, E., Rodden, W.P., 1969. A doublet-lattice method for calculating lift distributions on oscillating surfaces in subsonic flows. AIAA J. 7, 279–285.
Ashley, H., 1980. Role of shocks in the ‘‘sub-transonic’’ flutter phenomenon. J. Aircr. 17, 187–197.
Ashley, H., Zartarian, G., 1956. Piston theory, a new aerodynamic tool for the aeroelastican. J. Aeronaut. Sci. 23, 1109–1118.
Beliveau, J.-G., Cogan, S., Lallement, G., Ayer, F., 1996. Iterative least-squares calculation for modal eigenvector sensitivity. AIAA J. 34, 385–391.
Bendiksen, O.O., 2011. Review of unsteady transonic aerodynamics: theory and applications. Prog. Aerosp. Sci. 47, 135–167.
Chen, S.H., Song, D.T., Ma, A.J., 1994. Eigensolution reanalysis of modified structures using perturbations and rayleigh quotients. Commun. Numer. Methods.
Eng. 10, 111–119.
Chen, G., Sun, J., Li, Y.-M., 2012. Adaptive reduced-order-model-based control-law design for active flutter suppression. J. Aircr. 49, 973–980.
Chen, R., Wang, Z., 2000. Fast, block lower-upper symmetric Gauss–Seidel scheme for arbitrary grids. AIAA J. 38, 2238–2245.
Chen, S.H., Yang, X.W., 2000. Extended kirsch combined method for eigenvalue reanalysis. AIAA J. 38, 927–930.
Chen, S., Yang, X., Lian, H., 2000. Comparison of several eigenvalue reanalysis methods for modified structures. Struct. Multidiscip. Optim. 20, 253–259.
Ciampa, P.D., Nagel, B., Tooren, M.V., 2010. Global local structural optimization of transportation aircraft wings, In: Aiaa Structures, Structural Dynamics,
and Materials Conference.
Dowell, E.H., Williams, M., Bland, S., 1983. Linear/nonlinear behavior in unsteady transonic aerodynamics. AIAA J. 21, 38–46.
Ewins, D.J, 1984. Modal Testing: Theory and Practice. Research Studies Press Letchworth.
Förster, M., Breitsamter, C., 2013. Computational aeroelastic analysis in time and frequency domain using non-linear and linearized Euler-/Navier–Stokes-
methods.
Gasbarri, P., Chiwiacowsky, L.D., Velho, H.F.D.C., 2010. A hybrid multilevel approach for aeroelastic optimization of composite wing-box. Struct. Multidiscip.
Optim. 39, 607–624.
Guruswamy, G.P., 1990. Unsteady aerodynamic and aeroelastic calculations for wings using Euler equations. AIAA J. 28, 461–469.
Harder, R.L., Desmarais, R.N., 1972. Interpolation using surface splines. J. Aircr. 9, 189–191.

Please cite this article in press as: Li, D., et al., Structural dynamic reanalysis method for transonic aeroelastic analysis with global structural modifications.
Journal of Fluids and Structures (2017), http://dx.doi.org/10.1016/j.jfluidstructs.2017.06.004.
D. Li et al. / Journal of Fluids and Structures ( ) – 15

Jameson, A., 1991. Time dependent calculations using multigrid, with applications to unsteady flows past airfoils and wings. In: 10th Computational Fluid
Dynamics Conference, p. 1596.
Kamle, S., Kitey, R., Mohite, P., Upadhyay, C., Venkatesan, C., Yadav, D., 2017. Design of aircraft structures: an overview. In: Aerospace Materials and Material
Technologies. Springer, pp. 231–250.
Karpel, M., Wieseman, C., 1994a. Modal coordinates for aeroelastic analysis with large local structural variations. J. Aircr. 31, 396–403.
Karpel, M., Wieseman, C.D., 1994b. Time simulation of flutter with large stiffness changes. J. Aircr. 31, 404–410.
Khodaparast, H.H., Cooper, J., 2014. Rapid prediction of worst-case gust loads following structural modification. AIAA J. 52, 242–254.
Kirsch, U., Bogomolni, M., Sheinman, I., 2007. Efficient dynamic reanalysis of structures. J. Struct. Eng. 133, 440–448.
Lee-Rausch, E.M., Batina, J.T., 1995. Wing flutter boundary prediction using unsteady Euler aerodynamic method. J. Aircr. 32, 416–422.
Marques, S., Badcock, K., Khodaparast, H., Mottershead, J., 2012. How structural model variability influences transonic aeroelastic stability. J. Aircr. 49,
1189–1199.
Murthy, D.V., Kaza, K.R.V., 1989. A computational procedure for automated flutter analysis. Int. J. Numer. Methods Biomed. Eng. 5, 29–37.
Oroumieh, M.A.A., Malaek, S.M.B., Ashrafizaadeh, M., Taheri, S.M., 2013. Aircraft design cycle time reduction using artificial intelligence. Aerosp. Sci. Technol.
26, 244–258.
Pak, C.-g., Li, W.W., 2010. Application of approximate unsteady aerodynamics for flutter analysis.
Picardi, F., Paletta, N., Belardo, M., Pecora, M., 2014. Fast automatic procedure for flutter-clearance assessment when in the presence of significant structural
damage. J. Aerosp. Eng. 27, 05014001.
Platten, M.F., Wright, J.R., Worden, K., Dimitriadis, G., Cooper, J.E., 2007. Non-linear identification in modal space using a genetic algorithm approach for
model selection. Int. J. Appl. Math. Mech. 3, 72–89.
Pulliam, T., 1993. Time accuracy and the use of implicit methods. In: 11th Computational Fluid Dynamics Conference, p. 3360.
Rizk, M., 1981. Propeller in slipstream/wing interaction in the transonic regime. J. Aircr. 18, 184–191.
Sahu, J., Fresconi, F., Heavey, K., 2014. Unsteady aerodynamic simulations of a finned projectile at a supersonic speed with jet interaction. In: 32nd AIAA
Applied Aerodynamics Conference, p. 3024.
Schuster, D.M., Liu, D.D., Huttsell, L.J., 2003. Computational aeroelasticity: success, progress, challenge. J. Aircr. 40, 843–856.
Schwochow, J., Jelicic, G., 2015. Automatic operational modal analysis for aeroelastic applications. In: 6th International Operational Modal Analysis
Conference.
Silva, W.A., Chwalowski, P., Perry III, B., 2014. Evaluation of linear, inviscid, viscous, and reduced-order modelling aeroelastic solutions of the AGARD 445.6
wing using root locus analysis. Int. J. Comput. Fluid Dyn. 28, 122–139.
Song, Q., Chen, P., Sun, S., 2014. An exact reanalysis algorithm for local non-topological high-rank structural modifications in finite element analysis. Comput.
Struct. 143, 60–72.
Styuart, A.V., Livne, E., Demasi, L., Mor, M., 2011. Flutter failure risk assessment for damage-tolerant composite aircraft structures. AIAA J. 49, 655–669.
Tsai, H., Wong, F.A., Cai, J., Zhu, Y., Liu, F., 2001. Unsteady flow calculations with a parallel multiblock moving mesh algorithm. AIAA J. 39, 1021–1029.
Van Leer, B., 1979. Towards the ultimate conservative difference scheme. V. A second-order sequel to Godunov’s method. J. Comput. Phys. 32, 101–136.
Winter, M., Breitsamter, C., 2014. Reduced-Order Modeling of Unsteady Aerodynamic Loads Using Radial Basis Function Neural Networks, Deutscher Luft-
Und Raumfahrtkongress.
Yates, E.C., 1987. AGARD standard aeroelastic configuration for dynamic response, candidate configuration I. Wing 445, 103–129.
Zhong, J., Xu, Z., 2016. Coupled fluid structure analysis for wing 445.6 flutter using a fast dynamic mesh technology. Int. J. Comput. Fluid Dyn. 30, 531–542.
Zhou, Q., Chen, G., Li, Y., 2014. A reduced order model based on block arnoldi method for aeroelastic system. Int. J. Appl. Math. Mech. 6, 1450069.
Zhou, Q., Li, D.-f., Da Ronch, A., Chen, G., Li, Y.-m., 2016. Computational fluid dynamics-based transonic flutter suppression with control delay. J. Fluids
Struct. 66, 183–206.

Please cite this article in press as: Li, D., et al., Structural dynamic reanalysis method for transonic aeroelastic analysis with global structural modifications.
Journal of Fluids and Structures (2017), http://dx.doi.org/10.1016/j.jfluidstructs.2017.06.004.

You might also like