Вы находитесь на странице: 1из 9

A nonlinear stability analysis

of elastic ight vehicle


Saeed Shamaghdari and S.K.Y. Nikravesh
Electrical Engineering Department, Amirkabir University of Technology, Tehran, Iran
Abstract
Purpose The purpose of this paper is to present a nonlinear model along with stability analysis of a exible supersonic ight vehicle system.
Design/methodology/approach The mathematical state space nonlinear model of the system is derived using Lagrangian approach such that the
applied force, moment, and generalized force are all assumed to be nonlinear functions of the systemstates. The condition under which the systemwould
be unstable is derived and when the systemis stable, the region of attraction of the systemequilibriumstate is determined using the Lyapunov theory and
sum of squares optimization method. The method is applied to a slender exible body vehicle, which is referenced by the other researchers in the
literature.
Findings It is demonstrated that neglecting the nonlinearity in external force, moment and generalized force, as it was assumed by other researchers,
can cause signicant variations in stability conditions. Moreover, when the system is stable, it is shown analytically here that a reduction in dynamic
pressure can make a larger region of attraction, and thus instability will occur in a larger angle of attack, greater angular velocity and elastic displacement.
Practical implications In order to carefully study the behavior of aeroelastic ight vehicle, a nonlinear model and analysis is denitely necessary.
Moreover, for the design of the airframe and/or control purposes, it is essential to investigate region of attraction of equilibrium state of the stable ight
vehicle.
Originality/value Current stability analysis methods for nonlinear elastic ight vehicles are unable to determine the state space region where the
system is stable. Nonlinear modeling affects the determination of the stability region and instability condition. This paper presents a new approach to
stability analysis of the nonlinear exible ight vehicle. By determining the region of attraction when the system is stable, it is demonstrated analytically,
in this research, that decreasing the dynamic pressure can produce larger region of attraction.
Keywords Stability (control theory), Flight control, Elastic ight vehicle, Nonlinear modeling, Stability analysis, Sum of squares, Region of attraction
Paper type Research paper
Nomenclature
Symbols
f
z
(x, t) distributed normal force along the ight
vehicles body (N)
c
z
(a, Mach, x) distributed normal force coefcient along
the ight vehicles body
D Rayleigh dissipation function (J)
S
ref
reference cross section (m
2
)
F
z
normal force component along z axis (N)
F
Ax
axial force component along x axis (N)
l moment of inertia around lateral axis
(kg m
2
)
x
D
distance of resultant aerodynamic force
application point fromcentre of mass (m)
x
T
distance of resultant thrust force
applicationpoint fromcentreof mass (m)
x
cg
distance of centre of mass fromthe nose of
the vehicle (m)
L length of ight vehicle (m)
m ight vehicle mass (kg)
Mach Mach number
M
1
rst generalized elastic moment (kg m
2
)
M
y
moment component around y axis (N, m)
q angular velocity component around y axis
(rad/s)
Q dynamic pressure ((1/2)rv
2
) (Pa)
q
i
jth generalized coordinate
Q
j
jth generalized force
Q
z
rst vibration mode generalized force
r distance vector of each deected vehicle
point from body-xed coordinates origin
R position vector of each deected vehicle
point relative toinertial coordinates origin
t time
T kinetic energy (J)
U potential energy (J)
u and w velocity components along x and z axis (m/
s)
x, z body-xed coordinates
X
I
, Z
I
inertial coordinates
a angle of attack (rad)
d
z
normal elastic deection (m)
u pitch angle (rad)
z
i
(t) ith generalized coordinate
m rst bending mode damping
r density of air
w
i
(x) ith bending mode shape (m)
v
1
rst bending mode natural frequency (rad/
s)
Q
div
divergence dynamic pressure (Pa)
The current issue and full text archive of this journal is available at
www.emeraldinsight.com/1748-8842.htm
Aircraft Engineering and Aerospace Technology: An International Journal
84/6 (2012) 404412
q Emerald Group Publishing Limited [ISSN 1748-8842]
[DOI 10.1108/00022661211272954]
404
V Lyapunov function
P rst bending mode natural frequency
c invariance subset
l Lyapunov function level set
Introduction
Stability analysis and investigation of the dynamic behavior of a
exible ight vehicle has been the subject of many research
works in the eld of aeroservoelasticity. In studying a dynamic
system, modeling and analyzing methods are key features which
can greatly affect the results. Meirovitch and Nelson (1966)
investigated stability of spinning elastic ight vehicles by
combining ight and elasticity equations based on the Lagrange
approach and linear analysis. Considering aeroelastic stability
of spinning rockets, Platus (1992) proved, using linear analysis,
that in some special cases, structural damping causes instability
in spinning vehicles. Elyada (1989) studied static aeroelastic
instability of non spinning rockets ying in plane, just by
applying structural effectiveness coefcient in a closed form
without using all necessary equations. Haddadpour used modal
analysis, as well as generalized forces and coordinates, to drive
equations of motion. He derived equations governing rigid and
elastic motions using the Lagrange method and analyzed
aeroservoelastic stability of an elastic ight vehicle by linear
analysis in frequency domain (Haddadpour, 2006). Fathi et al.
(2009a, b) developed an analytical nonlinear formulation to
model the nonlinear aerodynamic force and moment
coefcients using orthogonal modeling polynomials based on
an extracted database from computational uid dynamics
(CFD), and studied the stability of the system in time domain
trajectories. Some other researchers have also studied the
dynamic behavior of a system either by using a linear or
nonlinear model and linear analysis approach, or by a nonlinear
model and time domain plotting of trajectories (Chae and
Hodges, 2003; Fathi et al., 2009a, b; Huang and Zeiler, 2006;
Murphy and Mermagen, 2001; Nesline and Nesline, 1985).
Considering geometrical nonlinearities inaeroelastic analysis of
an aircraft, Patil and Hodges (2001) proposed stability analysis
using the nonlinear nite element method which the beam
dynamics has been considered nonlinear.
In many cases of aeroelastic ight vehicle systems, nonlinear
modeling is necessary. Since the aerodynamic forces and
moments are highly nonlinear as the angle of attack increases, a
linear model is just valid around the equilibrium state, i.e. zero
angle of attack. But, time domain analysis by assessment of
state trajectories is not a secure method to analyze stability of
the system, and a nonlinear analysis method must be used.
Also, in the process of designing and controlling, it is essential
to know how large the state space is for which the system is
stable. This is an important specication of a nonlinear system
and can be used in a robust stability analysis of the system. So,
this region, so-called the region of attraction (ROA) of
equilibrium state should be determined in a stability analysis.
In this research, the mathematical model of a nonlinear
aeroelastic ying vehicle in the pitch plane is derived using
Lagrangian approach, and then the state space model of the
system is obtained. This is followed by the stability analysis
for the nonlinear system, based on the Lyapunov theory and
sum of squares (SOS) optimization method. An elastic
supersonic ight vehicle with small elastic displacement
assumption is used in this study.
In order to validate the nonlinear model and stability analysis
method given in this paper, the procedure is employed for a
special test case which was already used by some other
researchers. Comparing results of this study with others show
that in linear modeling there is acceptable agreement, but in
nonlinear modeling, there are signicant differences in the
stability conditions of this study with others. The study is
followed by obtaining the ROA, and also demonstrating that a
reduction in dynamic pressure, can cause an increase in the
ROA. Finally concluding remarks is presented.
Mathematical model
The model of the system can be obtained based on the
governing equations of motion of a supersonic ight vehicle. In
this research, the stability analysis of a supersonic vehicle ying
in a plane is considered and the mathematical model of the
system is derived. Based on Lagrangian approach, the kinetic
and potential energy of the system are calculated and is used to
derive the equations of motion. Then, the generalized forces
are derived using ight loads to construct the dynamic
aeroelastic model. General form of Lagranges equation can
be expressed as (Platus, 1992):
d
dt
T
_ q
i
_ _
2
T
q
i
_ _

U
q
i
_ _

D
_ q
i
_ _
Q
i
1
where T is the kinetic energy, U is the potential energy, D is the
Rayleigh dissipation damping function, q
i
is the ith generalized
coordinate, and Q
i
is the generalized forces.
Displacement of each element of the elastic vehicle is
equivalent to sum of orthogonal vibration modes. Assuming
that the body is symmetric, then it can be dened as follows:
d
z

n
i1
w
i
xz
i
t 2
where w
i
(x) is the ith normal bending mode shape and z
i
(t) is the
corresponding generalized coordinate. High vibration modes
have high natural frequencies and very low amplitudes and can
have perturbation effects on the system. In order to analyze the
stability of a exible body at supersonic ow, the rst vibration
mode can be considered (Platus, 1992). In this study only the
rst vibration mode is taken into account, so hereinafter w(x)
and z
i
(t) represents the rst bending mode shape w
1
(x) and rst
generalized coordinate z
1
(t), respectively.
Displacement of each element of the body in body
coordinates and inertial coordinates is shown in Figure 1.
The kinetic energy due to rigid and elastic motions is as
follows (Haddadpour, 2006):
T
1
2
mu
2
w
2

1
2
Iq
2

1
2
M
1

_
z
2
q
2
z
2
3
where u and w are linear velocity components in body
coordinates, q is pitch angular velocity, m is the mass of the
vehicle, and I is vehicle moment of inertia about lateral axes. M
1
is the generalized elastic moment of inertia for the rst vibration
mode and can be expressed as:
M
1

_
L
w
2
xdm 4
A nonlinear stability analysis of elastic ight vehicle
Saeed Shamaghdari and S.K.Y. Nikravesh
Aircraft Engineering and Aerospace Technology: An International Journal
Volume 84 Number 6 2012 404412
405
where L is the length of air vehicle. The potential energy of the
vehicle in terms of the modal parameters can be written as
(Meirovitch and Nelson, 1966):
U
1
2
M
1
v
1
z
2
5
The Rayleigh dissipation function can be written in terms of the
generalized forces and coordinates as follows (Meirovitch and
Nelson, 1966):
D m
1
v
1
M
1
_
z
2
6
where m
1
is the bending modal damping, and v
1
is the natural
frequency of the body for the rst vibration mode. After
determinationof the kinetic energy, potential energy andRayleigh
dissipation function, equations of motion can be derived.
Assuming that f
z
(x, t) is the normal aerodynamic load
distribution along the body, the work done due to the virtual
displacement d
z
can be written as follows:
dW
_
L
f
z
x; twxdztdx dzt
_
L
f
z
x; twxdx 7
Based on the virtual work theory, the relation between the
generalized force and elastic displacements can be written as
follows:
Q
z

_
L
f
z
x; twxdx 8
where Q
z
is the generalized force corresponding to the
generalized coordinate z. Using the equations of kinetic and
the potential energy, as well as the virtual work, substituting
them into the Lagrange equation (1), the elastic equation is
obtained as follows:

z 2m
1
v
1
_
z v
2
1
2q
2
_ _
z
1
M
1
_
L
f
z
x; twxdx 9
Using the Lagrange equation again, we constitute the force and
the moment differential equations in body coordinate system.
Considering pitch angle u (such that
_
u q) as generalized
coordinate, results in moment equation as follows:
d
dt
T
q
_ _
M
y
10
where M
y
is the external moment around y axis.
Substituting the kinetic energy into equation (10), yields:
I M
1
z
2
_ q 2M
1
qz
_
z M
y
11
In the left hand side of equation (11), M
1
z
2
term is very small
with respect to I, therefore to preserve polynomial type
functions of a state space model, it will be neglected from the
equation.
Similarly, considering linear displacement z as generalized
coordinate, the force equation is derived in inertial coordinates.
Using Coriolis effect the resultant force equation can be
expressed as follows:
_ z w
d
dt
T
w
_ _
2q
T
u
_ _
F
z
12
where F
z
is the force along the z axis of body coordinates.
Correspondingly, substituting thekinetic energy inequation(12),
yields:
m _ w 2qu F
z
13
The force in the right hand side of equation (13) is derived in the
body-xed coordinates system.
The moment in the right hand side of equation (11) is the
summation of the external moments due to the aerodynamic
and thrust forces and can be computed as follows:
M
y

_
L
x 2x
cg
f
z
x; tdx F
Tx
ztwx
T

2F
Tx
x
T
ztw
0
x
T
F
Ax
ztwx
D

2F
Ax
x
D
ztw
0
x
D
14
where F
A
x
is aerodynamic force along x axis, F
T
x
is thrust force
along x axis, x
cg
is the centre of mass, x
T
is the distance of the
point of application of the thrust force from x
cg
and x
D
is the
distance of the point of application of the aerodynamic force
from x
cg
. w
0
(x) is the derivative of w(x) with respect to x.
The force in the right hand side of equation (13) is the
summation of the aerodynamic and thrust forces and can be
expressed as:
F
z
2
_
L
f
z
x; tdx F
Tx
ztw
0
x
T
F
Ax
ztw
0
x
D
15
The external aerodynamic load f
z
(x, t) is determined as
follows:
f
z
x; t QS
ref
c
z
a; Mach; x 16
Figure 1 Elastic displacement in body coordinates (x, z) and in inertial
coordinates (X
I
, Z
I
)
X
1
Z
1
Z
R
E
r
d
z
C.G
x
A nonlinear stability analysis of elastic ight vehicle
Saeed Shamaghdari and S.K.Y. Nikravesh
Aircraft Engineering and Aerospace Technology: An International Journal
Volume 84 Number 6 2012 404412
406
where, Q 1=2ru
2
w
2
is the dynamic pressure, S
ref
is the
reference area and C
z
is dimensionless normal force coefcient
distribution, which is a function of a (angle of attack), Mach
number of ow and distance from the nose of the vehicle. This
distribution can be obtained from CFD methods (Fathi et al.,
2009a, b; Morelli and DeLoach, 2003).
The angle of attack for rigid bodies is constant at everywhere
on the body. For elastic bodies, angle of attack is a function
of elastic displacement and steady angle of attack of the ow,
a
0
. Expansion of a(x, t) over the length of the vehicle is as
follows:
ax; t a
0
2ztw
0
x
1
u
_
ztwx 2
q
u
x 17
Approximating _ w u _ a, nally, the state space equations for
elastic ying vehicle dynamics based on equations (9), (11) and
(13) is as follows:
_ a q
F
z
mu
_ q
1
I
M
y
22M
1
qz
_
z
_
z
1
z
2
_
z
2
22m
1
v
1
z
2
2 v
2
1
2q
2
_ _
z
1

1
M
1
Q
z
18
Note that in above equations, z
1
z and z
2

_
z.
Stability analysis
Consider the following nonlinear system:
_ x f x; v 19
where x is the state vector and v is the control input. To analyze
the stability of this system, all of the equilibrium states of the
system must be obtained by solving the nonlinear algebraic
equation (20), and each of them is analyzed for stability:
f x; 0 0 20
Assume origin is an equilibrium state of the system,
i.e. f(0, 0) 0, and the linearized system around origin is as
follows:
_ x Ax Bv 21
If the linearized systemis unstable, i.e. at least one eigenvalue of
Ais strictly in the right-half complex plane, then the equilibrium
state of the nonlinear system is unstable (Khalil, 2002). In the
elastic ying vehicle dynamic system (18), that the origin is an
equilibrium state, the state equations are functions of dynamic
pressure parameter Q too. It is desired to analyze the stability of
origin in terms of the variation in the value of Q.
To determine the dynamic pressure Q
div
, at which that an
instability is occurred, it is sufcient to linearize the system
around the origin and check sign of the real part of eigenvalues
of the linear system. In fact Q
div
is the divergent dynamic
pressure which makes the system unstable. For the dynamic
pressures that the linear system is stable, the nonlinear system
is locally asymptotically stable and so it has a ROA. The ROA
of an equilibrium state refers to the set of all points such that
trajectories initiated at these points eventually converge to the
equilibrium state. In the stable condition, we should determine
the ROA, i.e. in each dynamic pressure Q for which the system
is stable, how large is the ROA? For this reason, the ROA of
the origin is calculated using Lyapunov theory and Sum of
Square optimization programming methodology (Tan and
Packard, 2008).
Consider the nonlinear system, equation (18) the right hand
sides of these equations are polynomial functions in state
variables, so they are continuously differentiable. The constant
matrix A (an n n matrix of elements f
i
/x
j
) denotes the
Jacobian matrix of equation (18) with respect to x at x 0,
where x a; q; z;
_
z
T
.
Let us consider a quadratic Lyapunov function candidate
V as:
V x
T
Px 22
where P is a symmetric positive denite matrix, and is the
solution of Lyapunov equator (23):
A
T
P PA 2Q
g
23
For a symmetric positive denite matrix Q
g
, because of system
stability in Q , Q
div
, there is symmetric positive denite
matrix P satisfying equation (23). So
_
V in equation (24) is
negative denite and V satises the conditions of the basic
theorem of Lyapunov stability (Khalil, 2002):
_
V _ x
T
Px x
T
P_ x 2x
T
Qx 24
and the origin of nonlinear system, equation (18), is locally
asymptotic stable. For this locally stable equilibrium state it is
desired to nd the largest region that all initial conditions inside
it reach to origin as time goes to innity, i.e. the ROA.
Let x
0
a
0
; q
0
; z
0
;
_
z
0
be the initial condition of system
(18) and x(t, x
0
) denotes the solution at time t with the initial
condition x
0
. The ROA for the equilibrium state x 0 of the
system is:
ROA x
0
[ R
n
:
t!1
limxt; x
0
0
_ _
25
A set M is called invariant for the system (19), if every system
trajectory which starts froma point in Mremains in Mfor all the
future time, i.e. if x(t, x
0
) [ M for all t . 0 and x
0
[ M.
Computing the ROA precisely and even an estimation of it is
very challenging and difcult, in general. For two or three
dimensional systems, an estimation of ROA can be visualized
by simulation of system for a lot of initial conditions and
plotting their trajectories, such as phase plane for two
dimensional system. But, for four dimensional system, it is
impossible, so an analytical approach should be considered. Of
course this approach is desirable for two and three dimensional
systems too.
To determine the ROA, the Lyapunov function V in
equation (22) is a positive denite quadratic polynomial
function, and so it is continuously differentiable. Concerning
Lyapunov stability theory (Khalil, 2002), an invariant subset c
of the ROA can be dened as:
c {x [ R
4
: Vx # l} 26
Such that:
c {x [ R
4
: 7Vxf x; 0 , 0} 27
A nonlinear stability analysis of elastic ight vehicle
Saeed Shamaghdari and S.K.Y. Nikravesh
Aircraft Engineering and Aerospace Technology: An International Journal
Volume 84 Number 6 2012 404412
407
Then, because of local stability of the system, all x
0
[ c, and
the solution of equation (19) satises x(t, x
0
) [ c, for all t $ 0
and lim
t!1
xt; x
0
0. It is evident that c is an invariant
subset of the ROA for the equilibrium state x 0. To nd
maximum subset of ROA for given V, we have to nd the
maximum of l, therefore we need to dene an optimization
problem in order to nd l
max
. To this end, note that both sides
of equation (27) are dened in terms of inequalities, or set
inclusion, and generalized S-procedure (Boyd et al., 1994) can
be used to verify inclusion. Using S-procedure, if l(x): R
n
!R
is positive denite and s(x): R
n
!R is positive semidenite,
the set inclusion (27) can be written as bellow:
2lx 7Vxf x sxVx 2l $ 0 for all x 28
To verify that equation (28) is a sufcient condition for equation
(27), take an arbitrary nonzero state x such that V(x)(l, then
sxVx 2l $ 0, and it follows from equation (28) that
7Vxf x # 2lx , 0. To determine l
max
, we have to solve an
optimization problemwith constraint (26) and it needs to verify
if 2
_
V is positive denite. Since f(x) and V are nonlinear
polynomial vector functions, the problemcan be solved by SOS
decomposition method (Tan and Packard, 2008).
A polynomial p is an SOS if there exist polynomials p
1
,. . . p
N
such that p

N
i1
p
2
i
. The set of SOS polynomials in the
vector variable x is denoted by S[x ]. Now, the SOS
optimization problem is dened as:
l;s
max l Subject to : sx [ Sx
2lx 7Vxf x sxVx 2l [ Sx
29
where l
max
is obtained by solving this SOS optimization
programming, andsothe maximumof ROAcanbe determined.
Results
In order to validate the proposed approach, we have done
stability analysis for the rst test case introduced by Elyada
(1989), i.e. vehicle A. The geometry of this vehicle and its
distributed parameters are shown in Figure 2.
The mass properties and linear normal force coefcient of
the vehicle have been given. The normal force distribution at
ns has been considered as a pair of concentrated forces by
Elyada (1989). The arrows in the gure are related to this pair
of forces. It should be noted that in this paper the distributed
normal force over the length of the vehicle has been determined
using CFD. Based on the data from CFD runs, a polynomial
function in terms of the angle of attack, Mach number and x
(the distance fromthe nose) has been generated. The nonlinear
analytical function for normal force distribution is obtained as
follows:
c
z
a; Mach; x 35:46ax; t 236:4ax; t*x
11:24ax; t*x
2
2ax; t*x
3
4:45ax; t*x*Mach 27:9ax; t*Mach
20:55ax; t*x
2*
Mach 20:17ax; t*x*Mach
2
0:62ax; t*Mach
2
21:61ax; t
3
30
This equation has been derived by interpolating the CFD data
which has been generated for the air vehicle in the research via
FLUENT software, and minimizing predicted square errors
(PSE) (Fathi et al., 2009a, b). At low angle of attack, linear
assumption of aerodynamic distribution offers negligible errors,
but at high angle of attack, when the states are far from
equilibrium state, the aerodynamic characteristics are going to
have nonlinear behavior and can cause considerable error in the
nal outcomes of the model.
In the work of Elyada (1989) and Haddadpour
(2006), the aerodynamic normal force coefcient is not a
function of Mach number. To compare the results whit
their results, the Mach number in equation (30) is assumed
constant.
Substituting equation (17) in equation (30) and the resultant
into equations (8), (14) and (15), and integrating along the
body length, the nonlinear polynomial form of aerodynamic
force, moment and elastic generalized force can be obtained as
follows:
F
z
2QS
ref
2678:1z
3
108:1a
3
214:33z 10:08a
0:0122q 715:25z
2
a 2266:58za
2
0:0047z
_
za
20:1566zaq 20:0034
_
za
2
0:477a
2
q 0:252z
2
q
20:023z
2
_
z
Figure 2 Distributed properties of the vehicle A
1
0

6

E
I
,

N
.
m
2
m
(
x
)
,

K
g
/
m
c
L

(
x
)
,

m

1
60
40
20
0
0 1 2 3
x.m
4 5
0
2
1.5
0.5
0.0
1.0
Source: Elyada (1989)
A nonlinear stability analysis of elastic ight vehicle
Saeed Shamaghdari and S.K.Y. Nikravesh
Aircraft Engineering and Aerospace Technology: An International Journal
Volume 84 Number 6 2012 404412
408
M
y
2QS
ref
0:3136zaq 1:556z 23:189a 0:0007
_
z
0:011q 15:55z
2
a 76:72 20:013z
_
za 20:0022z
2
_
z
0:464a
2
q 0:046z
2
q 0:0013aq
2
20:0125
_
za
2
2:05z
3
15:13a
3
Q
z
2QS
ref
20:145a 0:106z 21:66z
2
a 241:5a
2
z
20:0013z
_
za 20:024zaq 20:02a
2
q 0:0025
_
za
2
0:000226z
2
_
z 20:0045z
2
q 20:2z
3
21:91a
3

31
The equilibriumstates of the systemare the solution of equation
(20). The system has three equilibrium states which vary with
the variation of dynamic pressure. By linearizing the system
around the origin and calculating eigenvalues, the instability
condition can be derived. Figure 3 shows the real parts of the
eigenvalues with variations of dynamic pressure Q, when the
force, the moment and the generalized force are linear functions
of states. In this case, the instability condition occurs at
Q
div
1.087 10
6
Pa, but inthe case of a nonlinear assumption
for the above functions, at Q
div
1.341 10
6
Pa the system
turns unstable. The real parts of the eigenvalues when the force,
moment and generalized force are considered nonlinear
functions of state variables is shown in Figure 4, and as can be
seen there is signicant difference in the calculated Q
div
.
For validation of our method, the results of some other
researchers where all used linear functions, have been
compared and presented in Table I.
As can be seen in Table I, our results in the case of linear
functions assumption are in good agreement with the results
reported by other researchers who have used linear force,
moment and generalized force. But, when the analysis is done
with nonlinear functions of force, moment and generalized
force, the obtained Q
div
is 1.341 10
6
Pa. As it can be seen in
Table I, there is a signicant difference in the determined value
of Q
div
between linear and nonlinear functions. Therefore, it
seems that it is crucial to use nonlinear functions in analytical
studies.
In the case where Q , Q
div
that the equilibriumstate x 0 is
stable, the other two equilibrium states are symmetric with
respect to the origin and both of them are unstable. The ROA
for the equilibrium state x 0 of nonlinear system, equation
(18), is obtained by Lyapunov function, equation (22), and
solving the optimization problem by SOS optimization
programming. The Lyapunov function, V, for this system is
as follows:
V 3086:57*a
2
5:7642*a*q 21823:188*az
1:4642
*
a
*
_
z 1:647
*
q
2
1:4405
*
q
*
z
20:021463
*
q
*
x4 1161:6703
*
z
2
20:45412
*
z
*
_
z
0:063346*
_
z
2
32
l
max
is obtained by solving the SOS optimization problem,
equation (29), so, in the region c {x [ R
4
: Vx # l
max
},
the
_
V is negative denite. The SOS decomposition of equation
(28) establishes negative deniteness of
_
V and is as follows:
2lx 7Vxf x sxVx 2l
max
Z
T
Q
z
Z 33
Figure 3 Instability conditions with linear functions
Dynamic Pressure (x1e6 pa)
R
e
a
l

P
a
r
t

o
f

E
i
g
e
n

v
a
l
u
e
s
0.8 1 1.2 1.4 1.6
40
20
0
20
40
Eig. Val. 1
Eig. Val. 2
Eig. Val. 3
Eig. Val. 4
Linear Force, Moment and Generalized Force
Qdiv = 1.087 1e6
Figure 4 Instability conditions with nonlinear functions
Dynamic Pressure (1e6 pa)
R
e
a
l

P
a
r
t

o
f

E
i
g
e
n

v
a
l
u
e
s
0.8 1 1.2 1.4 1.6
20
0
20
Nonlinear Force, Moment and Generalized Force
Eig. Val. 1
Eig. Val. 2
Eig. Val. 3
Eig. Val. 4
Qdiv = 1.087x1e6
Table I Comparison table for dynamic pressures equivalent to
instability (Q
div
)
References Q
div
(Pa)
Platus (1992) 1.089 10
6
Elyada (1989) 1.149 10
6
Haddadpour (2006) 1.087 10
6
Fathi et al. (2009a, b) 1.013 10
6
Presented method (linear functions) 1.087 10
6
Presented method (nonlinear functions) 1.341 10
6
A nonlinear stability analysis of elastic ight vehicle
Saeed Shamaghdari and S.K.Y. Nikravesh
Aircraft Engineering and Aerospace Technology: An International Journal
Volume 84 Number 6 2012 404412
409
where Z is a 34-element vector of all monomials of degrees 1 to
3, and Q
z
is a positive denite 34 34 matrix. In this case, Z is
as follows:
Z a; q; z;
_
z; a
2
; aq; az; a
_
z; . . . ; a
_
z
2
; q
_
z
2
; z
_
z
2
;
_
z
3
34
For Q , Q
div
, by solving the SOS optimization, one can nd a
specic l
max
(Q) that specify related ROA. The resulted l
max
for
several dynamic pressures are presented in Table II.
Because these ROAs are in four dimensional space and due to
impossibility of visualizing them, one can show some cross
sections of the ROA in 3D or 2D spaces. For example,
projections of ROAs of equilibrium state x 0 in these
dynamic pressures in a-q plane (z 0,
_
z 0) and a-z plane and
q-z plane are shown in Figures 5-7, respectively. Each closed
curve in the gures is the projection of V level sets in related
plane, i.e. the invariance sets projections. As it can be seen in
these gures, decreasing Q from Q
div
, causes larger ROA,
i.e. lower Q has larger stability margin.
Figure 8 shows projection of ROA of equilibrium state
x 0 for dynamic pressure Q 0.33 10
6
Pa in a-q-z space,
i.e. cross section of the ROA in
_
z 0.
In a qualitative analysis approach, the changing of system
equilibrium states in terms of Q is studied. For Q , Q
div
, the
system has three equilibrium states, one in origin that is stable
and the two others are symmetric with respect to origin that
are unstable. Therefore, the ROA of x 0 in direction of
unstable equilibrium states is limited and cannot include
unstable equilibrium states. Figures 9-11 show changing of
components a, q, z (
_
z 0 in all E.P.) for three equilibrium
states of the system in terms of Q in conditions that Q , Q
div
.
As it can be seen in the gures, increasing in Q, can make the
unstable equilibrium states closer to origin and consequently
constriction of the ROA.
Conclusions
In order to carefully study the behavior of some aeroelastic
ight vehicles, a nonlinear model and analysis should be
considered. Moreover, for the design of the airframe and/or
control purposes, it is essential to investigate the ROA of an
equilibrium state of the stable ight vehicle. This paper
presented a new approach to stability analysis of the nonlinear
exible ight vehicle. The mathematical nonlinear model of the
system has been derived using Lagrangian approach such that
applied force, moment and generalized force are assumed to be
nonlinear functions of the system states. In order to determine
the instability condition, i.e. the dynamic pressure Q
div
causing
instability, a linearized model is used. The method is applied to
a slender exible body test case, which are referred by other
researchers. In this study, it has been shown that neglecting
nonlinear functions in mathematical modeling can greatly
affect the stability analysis and the determined stability
conditions. In the situation that the zero equilibrium state is
stable, the ROA is determined by Lyapunov theory and SOS
Figure 5 Projections of ROAs for several Q in a-q plane
0.03 0.02 0.01 0 0.01 0.02 0.03
1.5
1
0.5
0
0.5
1
1.5
(rad)
q

(
r
a
d
/
s
e
c
)
ROA in -q Plane vs. Dynamic Pressure (10
6
pa)
1.27
1.02
0.75
0.52
0.33
Figure 6 Projections of ROAs for several Q in a-z plane
0.04 0.03 0.02 0.01 0 0.01 0.02 0.03 0.04
0.05
0.04
0.03
0.02
0.01
0
0.01
0.02
0.03
0.04
0.05
ROA in Plane vs.Dynamic Pressure (10
6
pa)
1.27
1.02
0.75 0.52
0.33

(rad)
Table II Maximum level set of V in several Qs
Q (10
6
Pa) l
max
0.33 1.72
0.52 0.7
0.75 0.31
1.02 0.1
1.27 0.012
Figure 7 Projections of ROAs for several Q in q-z plane
1.5 1 0.5 0 0.5 1 1.5
0.04
0.03
0.02
0.01
0
0.01
0.02
0.03
0.04
q (rad/sec)
ROA in q - Plane vs. Dynamic Pressure (*10
6
pa)
0.33
0.52
0.75
1.02
1.27

A nonlinear stability analysis of elastic ight vehicle


Saeed Shamaghdari and S.K.Y. Nikravesh
Aircraft Engineering and Aerospace Technology: An International Journal
Volume 84 Number 6 2012 404412
410
programming. It is interesting to note that in the case of
considering nonlinear functions for applied force, moment and
generalized force, signicant differences in stability conditions
can be observed. By determining the ROA when the system is
stable, it is demonstrated analytically in this research that
decreasing the dynamic pressure can produce a larger ROA,
i.e. instability occurs at a larger angle of attack, angular velocity
and elastic displacement. The approach has been presented in a
plane, but it can be extended to a three dimensional case in the
future. Comparison of this studys results in the linear case with
other researchers results for the same case scenario has shown
an acceptable agreement. But, since there are no published data
for the case of nonlinear functions of applied force, moment
and generalized force, results of this study cannot be logically
veried through comparing with others results.
References
Boyd, S., El Ghaoui, L., Feron, E. and Balakrishnan, V. (1994),
Linear Matrix Inequalities in Systems and Control Theory,
SIAM, Philadelphia, PA.
Figure 8 Projections of ROA for Q 0.33 10
6
in a-q-z space
0.06
4
0.04
Alpha
0.06
0.04
Vlin4 = 1.7000
0.04
4
3
2
q (rad)
Zeta
Figure 9 Variation of rst component (a) of three equilibrium
states vs Q
Dynamic Pressure (1e6 pa)
E
q
u
i
l
i
b
r
i
u
m

P
o
i
n
t
s
:

A
l
p
h
a

(
r
a
d
)
0.4 0.6 0.8 1 1.2
0.2
0.1
0
0.1
0.2
Stable E.P.
Unstable E.P.
Unstable E.P.
Figure 10 Variation of rst component (q) of three equilibrium states
vs Q
E
q
u
i
l
i
b
r
i
u
m

P
o
n
i
t
s
:

q

(
r
a
d
/
s
e
c
)
0.4 0.6 0.8 1 1.2
1.5
1
0.5
0
0.5
1
1.5
Dynamic Pressure (1e6 pa)
Figure 11 Variation of rst component (z) of three equilibrium states
vs Q
Dynamic Pressure (1e6 pa)
E
q
u
i
b
l
i
r
i
u
m

P
o
i
n
t
s
:

z
e
t
a
0.4 0.6 0.8 1 1.2
0.2
0.1
0
0.1
0.2
A nonlinear stability analysis of elastic ight vehicle
Saeed Shamaghdari and S.K.Y. Nikravesh
Aircraft Engineering and Aerospace Technology: An International Journal
Volume 84 Number 6 2012 404412
411
Chae, S. and Hodges, D.H. (2003), Dynamics and
aeroelastic analysis of missiles, Proceedings of the 44th
AIAA/ASME/ASCE/AHS Structures, Structural Dynamics,
and Materials Conference, Norfolk, VA, pp. 1-6.
Elyada, D. (1989), Closed-form approach to rocket vehicles
aeroelastic divergence, Journal of Spacecraft and Rockets,
Vol. 26 No. 2, pp. 95-102.
Fathi, J.M., Nobari, A.S., Sabzehparvar, M. and Haddadpour,
H. (2009a), Aeroelastic stability considerationof supersonic
ight vehicle using nonlinear aerodynamic response
surfaces, Journal of Fluids and Structures, Vol. 25 No. 6,
pp. 1079-101.
Fathi, J.M., Nobari, A.S., Sabzehparvar, M., Haddadpour, H.
and Tavakkolie, F. (2009b), Aeroelasticity consideration of
supersonic vehicle using closed form analytical aerodynamic
model, Aircraft Engineering & Aerospace Technology: An
International Journal, Vol. 81 No. 2, pp. 128-36.
Haddadpour, H. (2006), Aeroservoelastic stability of
supersonic slender-body ight vehicles, Journal of
Guidance, Control, and Dynamics, Vol. 29 No. 6, pp. 1423-6.
Huang, X. and Zeiler, T.A. (2006), Dynamics of exible
launch vehicles with variable mass, Proceedings of the 44th
AIAA Aerospace Sciences Meeting and Exhibit, Reno, NV,
pp. 1-33.
Khalil, H.K. (2002), Nonlinear Systems, Prentice-Hall, Upper
Saddle River, NJ.
Meirovitch, L. and Nelson, H.D. (1966), On the high-spin
motion of a satellite containing elastic parts, Journal of
Spacecraft and Rockets, Vol. 3 No. 11, pp. 1597-602.
Morelli, A. and DeLoach, R. (2003), Wind tunnel database
development using modern experiment design and
multivariate orthogonal functions, Proceedings of the 41th
AIAA Aerospace Sciences Meeting and Exhibitat, Reno, NV,
pp. 1-15.
Murphy, C.H. and Mermagen, W.H. (2001), Flight
mechanics of an elastic symmetric missile, Journal of
Guidance, Control and Dynamics, Vol. 24 No. 6, pp. 1125-32.
Nesline, F.W. and Nesline, M.L. (1985), Phase vs gain
stabilization of structural feedback oscillations in homing
missile autopilots, Proceedings of the American Control
Conference, Boston, MA, pp. 323-9.
Platus, D.H. (1992), Aeroelastic stability of slender, spinning
missiles, Journal of Guidance, Vol. 15 No. 1, pp. 144-51.
Tan, W. and Packard, A. (2008), Stability region analysis using
polynomial and composite polynomial Lyapunov functions
and sum-of-squares programming, IEEE Transactions on
Automatic Control, Vol. 53 No. 2, pp. 565-71.
Further Reading
Chang, C., Hodges, D.H. and Patil, M.J. (2008), Flight
dynamics of highly exible aircraft, Journal of Aircraft,
Vol. 45, pp. 538-45.
Corresponding author
Saeed Shamaghdari can be contacted at: shamaghdari@
aut.ac.ir
A nonlinear stability analysis of elastic ight vehicle
Saeed Shamaghdari and S.K.Y. Nikravesh
Aircraft Engineering and Aerospace Technology: An International Journal
Volume 84 Number 6 2012 404412
412
To purchase reprints of this article please e-mail: reprints@emeraldinsight.com
Or visit our web site for further details: www.emeraldinsight.com/reprints

Вам также может понравиться