Chinese

Journal of
Aeronautics
Chinese Journal of Aeronautics 21(2008) 417-422
www.elsevier.com/locate/cja
A New Approach for Aeroelastic Robust Stability Analysis
Wu Zhigang*, Yang Chao
School of Aeronautic Science and Engineering, Beijing University of Aeronautics and Astronautics, Beijing 100191, China
Received 12 November 2007; accepted 25 March 2008
Abstract
Air vehicles undergo variations in structural mass and stiffness because of fuel consumption and the failure of structural compo-
nents, which might lead to serious influences on the aeroelastic characteristics. An approach for aeroelastic robust stability analysis tak-
ing into account the perturbations of structural mass and stiffness is developed. Applying the perturbation method and harmonic unsteady
aerodynamic forces, the frequency-domain linear fractal transformation (LFT) representation of perturbed aeroelastic system is modeled.
Then, the robust stability is analyzed by using the structured singular value -method. The numerical results of a bi-spar wing show its
effectiveness and low computational time in dealing with the robust problems with mass and stiffness perturbations. In engineering
analysis for solving aeroelastic problems, the robust approach can be applied to flutter analysis for airplane with the fuel load variation
and taking the damage conditions into consideration.
Keywords: aeroelasticity; robustness; flutter; structured singular value; uncertainty; perturbation
1 Introduction
*

Aeroelasticity considers the interaction be-
tween aerodynamic forces, inertial forces, and elas-
tic forces. Since the advent of flight, the study of
aeroelasticity has been an important discipline, and
aeroelastic stability has been a necessary condition
for the safe flight of air vehicles. The relevant speci-
fications require that all the configurations of air-
plane should be free from any aeroelastic instability
for all combinations of altitude and speed within the
limit speed versus altitude envelope enlarged at all
points by the airspeed safety margin
[1-2]
.
There are always some differences between the
nominal system and the actual system, which are
due to modeling errors, manufacturing errors, aging
of hardware, and variation of flight conditions. For
example, fuel consumption and external stores lead

*Corresponding author. Tel.: +86-10-82317510.
E-mail address: wuzhigang@buaa.edu.cn
Foundation item: National Natural Science Foundation of China
(10432040, 90716006)
to variation of inertial characteristics, and the fail-
ures of some components lead to degradation of the
structural stiffness. The variations of structural mass,
stiffness, and aerodynamic forces have serious in-
fluences on the aeroelastic characteristics of air ve-
hicles.
In traditional engineering analysis of aeroelas-
tic stability, the structural dynamics is often based
on the linear natural modes, unsteady aerodynamics
is represented as the form of aerodynamic influence
coefficients, and the k method or p-k method is ap-
plied to solve flutter characteristic equations
[3]
.
To make sure the aeroelastic system is stable
under the influences of modeling errors and varia-
tion of flight conditions, the safety factor of 1.15 is
required. To deal with the flutter problems that arise
with the perturbations of structural mass and stiff-
ness, the parametric analysis is employed. To evalu-
ate the variations of multiple parameters, all the
combined conditions should be enumerated and
418 Wu Zhigang et al. / Chinese Journal of Aeronautics 21(2008) 417-422

analyzed. This approach is relatively simple, but
requires enormous computational time. In addition,
it is not possible to study all the possible combina-
tions between different parameters, and there is a
chance to omit the worst case, which would result in
serious consequences.
With the development of control theory, the
concept of robustness has been introduced into
aeroelastic analysis. Flutter analysis considering
variations of parameters can be seen as the problem
of robust stability under the perturbation of uncer-
tainties. In recent years, robust aeroelasticity is of
immense importance in the field of aeroelasticity. In
robust problems, the parametric uncertainties are
often structured as block diagonal operators. Ne-
glecting the operator structures would lead to con-
servative analysis results. The structured singular
value -method can solve the robust problems with
structured and unstructured uncertainties
[4]
, and it
has proved to be an effective tool for robust aeroe-
lastic studies.
Lind and Brenner
[5-6]
firstly presented a method
for robust aeroelastic analysis by combining -
method and flight test data, and applied it to the
estimation of robust flutter and aeroservoelastic
margins of F-18 research aircraft. In recent years, -
method has paved the way for robust flutter
[7]
and
aeroservoelastic analyses
[8-9]
. Wu, et al.
[10]
extended
- method for analyzing static aeroelastic robustness.
The above mentioned works are all based on the
time-domain linear fractal transform (LFT) model,
which requires the rational function approximation
(RFA) of unsteady aerodynamic forces
[11]
and the
state-space modeling of aeroelastic system. But, the
time-domain LFT method has some shortcomings.
RFA brings aerodynamic augmented states and ap-
proximation errors. Besides, time-domain LFT
modeling for aeroelastic system with physical pa-
rameter perturbation is complex. Borglund
[12]
pre-
sented a robust flutter analysis method based on
frequency-domain. This method uses harmonic un-
steady aerodynamic forces and analyzes robustness
based on frequency responses, which greatly simpli-
fies uncertainty modeling.
The objective of this article is to develop an
approach for aeroelastic robust stability analysis
considering the perturbations of structural mass and
stiffness. The application of this method to engi-
neering analysis is also discussed.
2 Equations of Motion of Perturbed Aero-
elastic System
In the physical coordinates of structure, the
equations of motion of the nominal aeroelastic sys-
tem can be written as
2
S S S
1
2
V + = + M x K x Q x y (1)
where x is the displacement vector of nodes, y ex-
ternal force vectors, M
S
structural mass matrix, K
S

structural stiffness matrix, the density of atmos-
phere, V flight speed, and Q
S
unsteady aerodynamic
influence coefficient matrix.
The normal modes of the nominal system are
denoted as |
1
, |
2
, . On the basis of the principle of
modes truncation, the displacement x can be repre-
sented as the linear combination of the first n modes.
The vector consisting of the first n modes is denoted
as q, the mode matrix is represented as u, and = x
q u is substituted into Eq.(1), then
2
1
2
V + = + Mq Kq Qq f (2)
T
S
T
S
T
S
T
=

`
=

=
)
M M
K K
Q Q
f y
u u
u u
u u
u
(3)
Because of the orthogonality of the normal modes,
the generalized mass M and the generalized stiffness
K of the nominal system are diagonal matrices.
Considering the perturbation of structural mass
and stiffness of the air vehicle, the mass matrix, and
the stiffness matrix of perturbed system can be writ-
ten as
S S M M
1
k
i i
i
o
=
= +

M M W (4)
S S K K
1
l
j j
j
o
=
= +

K K W (5)
where W
Mi
and W
Kj
represent the perturbation
Wu Zhigang et al. / Chinese Journal of Aeronautics 21(2008) 417-422 419

weighted matrices, and o
Mi
and o
Kj
denote the mass
and stiffness uncertainties, respectively. In fact, o
Mi

and o
Kj
are physical parameters representing the
mass and stiffness characteristics of the system,
such as the lumped masses of nodes, the stiffness
coefficient of elements, and the elastic modules of
materials.
Then, the equations of motion of the perturbed
aeroelastic system are
S M M S K K
1 1
2
S
( ) ( )
1
2
k l
i i j j
i j
V
o o

= =
+ + + =
+

M W x K W x
Q x y

(6)
It should be noted that |
1
, |
2
, are not the
normal modes of the perturbed system any more. If
the number of modes n is large enough, the struc-
tural displacement x can be still represented as the
linear combination of the first n modes. Under the
condition of small perturbation, the number of n is
required to be larger by 3-6 than the number of in-
terested modes, in general. Substituting = x q u
into Eq.(6), and considering the orthogonality of the
natural modes u for nominal structural mass M
S
and
stiffness K
S
, we can get
T T
M M K K
1 1
( ) ( )
k l
i i j j
i j = =
+ + + =

M W q K W q o o u u u u
2
1
2
V + Qq f (7)
3 Analysis Method of Robust Aeroelastic
Stability
3.1 Frequency-domain LFT model of per-
turbed system
On the basis of the assumption of harmonic os-
cillation, the frequency-domain equations of motion
of perturbed aeroelastic system can be written as
2 2
2 T T
M M K K
1 1
1
( )
2
k l
i i j j
i j
V
= =
| |
+ = +
|
\ .
| |
|
|
\ .

M K Q q f
W W q
e e
e o o u u u u (8)
where e denotes the cyclic frequency of vibration.
Here,
1 M
2 K
,
( (
= =
( (

R
z q w z
R
A
A
(9)
where
M M1 M
diag( , , )
k
o o = I I " A (10)
K K1 K
diag( , , )
l
o o = I I " A (11)
| |
T
1 n n n n
k

= R I I "

(12)
| |
T
2 n n n n
l

= R I I "

(13)
Incorporating Eq.(9) into Eq.(8), we get
1 1 2
2 2
( ) ( ) e e e

(
+

q = F f F L L w (14)
where
2 2
1
( ) ( )
2
V e e e = + F M K Q (15)
T T
1 M1 Mk
(
=

L W W " u u u u (16)
T T
2 K1 Kl
(
=

L W W " u u u u (17)
Substituting Eq.(14) in Eq.(9), we get
1 1 1 1 2
2 2
2 2
( ) ( ) e e e

( (
(
= +
( (


R R
z F f F L L w
R R

(18)
Combining Eq.(14) with Eq.(18), we obtain the
LFT formulation of the perturbed system (see Fig.1).
The mathematical formulation of the model is rep-
resented as
11 12
21 22
( ( (
=
( ( (

P P q f
P P z w
(19)
where the transfer function matrix for w to z is
1 1 2
22 2 2
2
( ) ( ) e e e

(
(
=
(


R
P F L L
R
(20)

Fig.1 LFT formulation of the perturbed aeroelastic system.
It should be noted that the calculation of the
420 Wu Zhigang et al. / Chinese Journal of Aeronautics 21(2008) 417-422

transfer function matrix is based on the harmonic
aerodynamic influence coefficients in frequency-
domain. The state-space model of the aeroelastic
system is not required, so the additive aerodynamic
lags and approximation due to RFA can be avoided.
3.2 V- method for robust stability analysisG
According to the structured singular value
based robust stability theorem
[4]
, the perturbed
closed-loop system (see Fig.1) is robustly stable
with respect to the set A which is norm bounded
by 1

s A for all e A A if and only if

22
sup ( ) 1
j
e
e
e
( <

R
P (21)
A value of 1 < implies that no perturbation
exists within the set A , which will destabilize the
closed-loop system. Obviously, is related to the
block structure of set A , and it is a challenge to
compute it, which appears to be an NP-hard pro-
blem. Alternatively, the upper and lower bounds of
may be easily computed. The computation of up-
per bound may be posed as an optimization problem
to present a conservative bound on the worst-case
stability properties of the system. The aeroelastic
analysis usually focuses more on identifying worst-
case stability margins, thus the computation of in
this article concentrates on the upper bound of ,
which can be solved by MATLAB -analysis and
synthesis toolbox
[13]
. G
The procedure of aeroelastic robust stability
analysis, which uses the relationship between flight
speed V and structured singular value also named
as V- method, is given as follows.G
Step 1 Compute the flutter speed V
f
of the
nominal system by the traditional methods, such as
p-k method.
Step 2 Determine the levels of the uncertain-
ties of structural mass and stiffness.
Step 3 Select a sequence of flight speed V
i

f 1 2
V V V > > >"
Step 4 Select a sequence of cyclic frequency
e
j

1 2
e e < <"
Step 5 For each value of flight speed V
i
:
(1) For each value of frequency e
j
:
Compute the corresponding generalized
aerodynamic coefficient matrix ( )
j
e Q ;
According to Eq.(20), compute the fre-
quency response matrix
22
( )
j
e P ;
Compute the corresponding structured sin-
gular value
22
[ ( )]
j
e P .
(2) Plot e - curve at the flight speed V
i
.
(3) Find the maximum value sup = within
the range of computed frequencies.
Step 6 Find the flight speed V
rob
for which
satisfies the condition of 1 = .
The computed value V
rob
is the robust flutter
speed considering the perturbations of structural
mass and stiffness. Furthermore, the worst-case can
be obtained which shows the critical levels of per-
turbed mass and stiffness at flight speed V
rob
.
4 Numerical Results
The structure and geometry of a large aspect
ratio wing with two spars is shown in Fig.2. Here,
two kinds of variations of physical characteristics
should be considered: there are eight fuel tanks
located within the wing that are shown as small
panes, and the fuel consumption could bring re-
markable variations of the wing mass and mass dis-
tribution; the elastic module of the material of
the fore-spar decreases, which induces the variation
of structural stiffness.

Fig.2 Sketch of a large aspect ratio wing with two spars.
First, the wing with half amount of fuel is de-
fined as the nominal model, in which the lumped
masses of fuel tanks numbered 1-8 are 2, 3, 2, 3, 2,
3, 2, 3 kg, respectively. Considering the first six
normal modes, the nominal p-k flutter analysis of
the wing with half amount of fuel gives a flutter
speed of V
f
= 93.0 m/s at a frequency of 11.2 Hz.
Then, the three cases of perturbations of struc-
Wu Zhigang et al. / Chinese Journal of Aeronautics 21(2008) 417-422 421

tural mass and stiffness are considered as follows.
Case I Perturbation of the masses of fuel
tanks numbered 1-8. The uncertainty operator of
structural mass is written as
M M1 6 6 M8 6 6
diag( , , ) o o

= I I " A
where
Mi
o represents the uncertainty of lumped
mass of the ith tank, and it is normalized to
M
1
i
o s , +1 represents full amount of fuel, and 1
represents empty.
Case II Perturbation of the elastic module of
fore-spar. The uncertainty operator of structural
stiffness is defined as
K K 6 6
o

= I A
where
K
o represents the uncertainty of the elastic
module, and
K
o is normalized to
K
1 o s , 0 repre-
sents the nominal model, and 1 represents the elas-
tic module being decreased by 10%.
Case III Both perturbations of Case I and
Case II are considered simultaneously. In this case,
the uncertainty operator is written as
M K
diag( , ) = A A A
On the basis of the above mentioned cases, ro-
bust aeroelastic analysis gives the e - curves at
various flight speeds by V- approach as shown in
Figs.3-5.
On the other hand, all the combinatorial condi-
tions of perturbations are enumerated and analyzed.
Comparing these results with the results obtained by
V- approach, it is observed that they show a good
correlation. Hence, the accuracy of the robust analy-
sis approach is validated.

Fig.3 Robust analysis e - curves for Case I.

Fig.4 Robust analysis e - curves for Case II.

Fig.5 Robust analysis e - curves for Case III.
In addition, the worst-case is analyzed for Case
III, and the obtained results are listed in Table 1.
The results show that if the elastic module of
fore-spar is decreased by 10% and the lumped
masses of fuel tanks numbered 1-8, respectively, are
4.0, 1.2, 0, 0, 0, 6.0, 0, 6.0 kg, the flutter speed of
the wing reduces to the minimum value of 78.9 m/s.
Table 1 Worst-case of perturbation for Case III
o
M1
o
M2
o
M3
o
M4
o
M5
o
M6
o
M7
o
M8
o
K

1.0 0.6 1.0 1.0 1.0 1.0 1.0 1.0 1.0
5 Conclusions
The numerical example demonstrates that the
V- approach for robust aeroelastic stability pre-
sented in this article can be applied to analyze the
influences of physical parameters perturbations of
structural mass and stiffness on the aeroelastic
characteristics. In actual calculation, only the har-
monic aerodynamic influence coefficients within
frequency-domain and the structural element mass
matrix and stiffness matrix are required. Compared
422 Wu Zhigang et al. / Chinese Journal of Aeronautics 21(2008) 417-422

with the traditional engineering analysis method, the
V- approach requires less computer time and has
good accuracy.
In the condition of small perturbation, the ro-
bust approach can be applied to flutter analysis for
airplane with the fuel load variation and damaged
conditions taken into account. It should be pointed
out that up to now the V- approach is limited to
dealing with the problems with two bounds pertur-
bation. For some actual problems with single bound
perturbation, this approach may give conservative
results. To overcome the limitation, further research
is required in this aspect in the future.
References
[1] Commission of Science, Technology and Industry for National
Defense. GJB67.7-85 Military airplane strength and rigidity
specification: aeroelastic instabilities. 1985. [in Chinese]
[2] Commission of Science, Technology and Industry for National
Defense. GJB540.8-88 Cruise missile strength and rigidity speci-
fication: aeroelastic instabilities. 1988. [in Chinese]
[3] Rodden W P, Johnson E H. MSC/NASTRAN aeroelastic analysis
users guide. Log Angeles: The MacNeal-Schwendler Corporation,
1994; 97-101.
[4] Wu M, Gui W H. Modern robust control. Changsha: Central South
University Press, 1998; 124-126. [in Chinese]
[5] Lind R, Brenner M. Robust flutter margins of an F/A-18 aircraft
from aeroelastic flight data. Journal of Guidance, Control, and
Dynamics 1997; 20(3): 597-604.
[6] Lind R, Brenner M. Incorporating flight data into a robust aeroe-
lastic model. Journal of Aircraft 1998; 35(3): 470-477.
[7] Lind R. Match-point solutions for robust flutter analysis. Journal
of Aircraft 2002; 39(1): 91-99.
[8] Wu Z G, Yang C. Modeling and robust stability for aeroservoelas-
tic systems with uncertainties. Acta Aeronautica et Astronautica
Sinica 2003; 24(4): 312-316. [in Chinese]
[9] Wu Z G. Aeroservoelastic modeling, analysis and synthesis of
elastic air vehicles. PhD thesis, School of Aeronautic Science and
Engineering, Beijing University of Aeronautics and Astronautics,
2004. [in Chinese]
[10] Wu Z G, Yang C. Robustness analysis of static aeroelastic systems
with physical parameters perturbation. Acta Aeronautica et Astro-
nautica Sinica 2006; 27(4): 565-569. [in Chinese]
[11] Tiffang S H, Karpel M. Aeroservoelastic modeling and applica-
tions using minimum-state approximations of the unsteady aero-
dynamics. AIAA-89-1188, 1989.
[12] Borglund D. The -k method for robust flutter solutions. Journal
of Aircraft 2004; 41(5): 1209-1216.G
[13] Balas G J, Doyle J C, Glover K. MATLAB -anslysis and synthe-
sis toolbox users guide. Natick: The Math Works Inc., 1998;
49-66.
Biographies:
Wu Zhigang Born in 1977, he received the Ph.D. degree
in flight vehicle design at Beijing University of Aeronautics
and Astronautics (BUAA) in 2004. Presently, he is a lecturer
in School of Aeronautic Science and Engineering, BUAA.
His main research interests include structural dynamics,
aeroelasticity, and active control.
E-mail: wuzhigang@buaa.edu.cn

Yang Chao Born in 1966, he received the Ph.D. degree in
BUAA in 1996. He is currently a professor in School of
Aeronautic Science and Engineering, BUAA. His main re-
search interests include aeroelasticity, flight mechanics, and
flight vehicle design.
E-mail: yangchao@buaa.edu.cn