Sunteți pe pagina 1din 24

Journal of Sound and Vibration (1996) 193(4), 823–846

AEROELASTIC ANALYSIS OF A FLEXIBLE


AIRFOIL WITH A FREEPLAY NON-LINEARITY
S.-H. K  I. L
Department of Aerospace Engineering, Korea Advanced Institute of Science and
Technology, Taejon, Korea

(Received 10 January 1995, and in final form 10 October 1995)

A two-dimensional flexible airfoil with a freeplay non-linearity in pitch has been analyzed
in the subsonic flow range. Structurally, the airfoil is modelled as finite beam elements and
two spring elemens in pitch and plunge. A doublet lattice method is used for the
two-dimensional unsteady aerodynamics to include the camber deflection effect. The
fictitious mass modal approach is adopted in order to use the consistent modal co-ordinates
for the structures with non-linearity. Non-linear aeroelastic analyses for both the frequency
domain and time domain are performed for rigid and flexible airfoil models to investigate
the flexibility effect. Results are shown for models of different pitch-to-plunge frequency
ratio. Responses involving limit cycle oscillation and chaotic motion are observed and they
are highly influenced by the pitch-to-plunge frequency ratio.
7 1996 Academic Press Limited

1. INTRODUCTION
Recently, many investigations for aeroelastic analysis have been made for systems with
structural non-linearities. Concentrated structural non-linearities such as those of freeplay,
bilinear and cubic types, are known to have significant effects on the aeroelastic responses
of aerosurfaces. These non-linearities have significant effects even for small vibrational
amplitudes, and provide non-linear aeroelastic responses of four categories: damped
stable motion, periodic motion, chaotic motion and divergent flutter. Parameters affecting
these responses have been investigated in previous studies: initial conditions, frequency
ratio and amount of non-linearity, etc. However, more studies are still needed to
understand the non-linear aeroelastic phenomena comprehensively. A model with reduced
degrees of freedom is desirable for non-linear aeroelastic analysis, since the responses are
complicated.
Aeroelastic analyses of a two-dimensional airfoil (or a typical section model) with
concentrated non-linearities have been performed by several investigators, because of its
simplicity and usefulness [1–11]. The typical section model gives a lot of insight and
information about the aeroelastic phenomena. Previous studies on the two-dimensional
model usually used the rigid typical section model with a concentrated non-linearity in
pitch. The aerodynamic loads were calculated by using the Theodorsen function in the
frequency domain or the Wagner function in the time domain; these are based on the
assumption of chordwise rigidity. In the first non-linear aeroelastic analysis for an airfoil
model, Woolston [1] and Shen [2] showed that the limit cycle oscillation may occur below
the linear flutter boundary. McIntosh et al. [3] performed experimental work with a wind
tunnel model of two degrees of freedom. Lee [4] developed an iterative scheme for multiple
non-linearities using the describing function method and the structural dynamics
modification technique. Yang and Zhao [5] studied the limit cycle oscillation of an airfoil
823
0022–460X/96/240823 + 24 $18.00/0 7 1996 Academic Press Limited
824 .-.   . 
with pitch non-linearity subject to incompressible flow using the Theodorsen function.
They made a comprehensive study of limit cycle flutter. Zhao and Yang [6] also studied
the chaotic responses of an airfoil with cubic non-linearity in pitch and subject to steady
incompressible flow. Brase and Eversman [7] developed the transformation method which
converts the harmonic aerodynamic force to a transient one, and found a jump response
in a two-dimensional model with pitch freeplay non-linearity. Hauenstein et al. [8, 9]
made experimental and analytical analyses of an aerosurface with freeplay non-linearities
in pitch and plunge degrees of freedom. They showed qualitatively good results between
the experiment and analysis. They concluded from the results that chaotic motion does
not occur with the presence of a single non-linearity, but Price et al. [10, 11] pointed out
this is not true. They made a comprehensive aeroelastic study of a two-dimensional
airfoil with pitch non-linearities such as those of freeplay, cubic and bilinear stiffness
types.
The plunge-to-pitch frequency ratio (vh /va ) of previous models are usually less than
unity, which implies a large aspect ratio wing. The rigid chord assumption is valid for the
large aspect ratio wing. However, the chordwise bending effect on the aeroelastic response
is significant when the aspect ratio of the wing becomes small, or when the mass ratio is
small. In order to understand the aeroelastic phenomena more comprehensively it is
required to know the linear and non-linear aeroelastic response of a two-dimensional
flexible airfoil model. The chordwise bending effect of a cantilevered plate was investigated
in the supersonic range by Dugundji and Crisp [12]. Crisp [13] mentioned that the elastic
chordwise bending effect in the typical section model may be an important factor in flutter
analysis. However, results of non-linear aeroelastic analysis for two-dimensional flexible
model have not yet been reported.
In this paper, a two-dimensional flexible airfoil with a freeplay non-linearity in pitch
and subject to subsonic flow has been studied. A finite beam element and two spring
elements (one is non-linear) have been used for structural analysis. A doublet lattice
method (DLM) [14] has been used to accommodate the two-dimensional chordwise
bending effect on the aerodynamic forces. The fictitious mass modal approach [15, 16] was
adopted in order to use the consistent modal co-ordinates for structures with non-linearity.
Non-linear aeroelastic analyses for both the frequency domain and time domain have been
performed for rigid and flexible airfoil models to examine the flexibility effect. For the
frequency domain analysis the describing function has been used with the V-g method. The
method of Brase and Eversman [7] was used to obtain the transient aerodynamic forces
in the time domain. Results are shown for two models of different pitch-to-plunge
frequency ratio (va /vh ). In the frequency domain, the limit cycle flutter speed and
frequency were obtained for various amplitude ratios. The time domain analysis was
performed for various initial pitch amplitude ratios and free stream velocities. Responses
including the limit cycle oscillation and chaotic motion are observed and are shown to be
highly affected by the pitch-to-plunge frequency ratio.

2. THEORETICAL ANALYSIS

2.1.   


The structural non-linearity must be transformed to an equivalent linear system to solve
the non-linear aeroelastic problem in the frequency domain. The describing function
method used in this analysis is that which was used by Laurenson and Trn [17].
   - 825

Figure 1. The describing function of a freeplay non-linearity.

The freeplay non-linearity can be described by

8 9
Ka (a − s), aqs
{ f (a)} = 0, −s Q a Q s , (1)
Ka (a + s), a Q −s

where a is the displacement, and s is the magnitude of the freeplay. Freeplay non-linearity
is shown in Figure 1. The describing function for freeplay non-linearity d can be obtained
as

6 7
0, AEs
d= , (2)
(1/p)[p − 2a − 2 sin 2a], Aqs

where A is the vibration amplitude, s is the freeplay gap and a = sin−1 (s/A). Then, the
elastic force becomes f (A) = dKa A. This function is dependent on the amplitude ratio s/A
and the conventional linear flutter analysis algorithm can be applied when using this
function. Here the V-g method was used, which can predict the flutter speed of a
period-one limit cycle oscillation. A time domain approach must be used to obtain the
characteristics of the complicated motion.

2.2.    


Generally, the aeroelastic analysis in the time domain is conducted on the generalized
modal co-ordinate to reduce the computation time and memeory requirement. In
non-linear aeroelastic problems, structural properties vary as the displacement changes.
Hence, using a constant set of normal modes from a fixed structural model gives inaccurate
results. To overcome this problem, Karpel proposed the fictitious mass method [15, 16].
In this method, a large fictitious mass is added to the degree of freedom of mass matrix
where structural change will occur. Then, the normal modes obtained from the free
vibration analysis for the system with fictitious mass are used for the aeroelastic response.
The basic idea of this method is that the local deformation due to a large mass enables
one to account for the structural changes.
The free vibration equation of motion of an n d.o.f. system with fictitious masses is

[M + Mf ]{ü} + [K ]{u} = {0}, (3)


826 .-.   . 

Figure 2. A flexible beam airfoil model.

where [M] is the mass matrix, [K ] is the stiffness matrix and [Mf ] is the fictitious mass added
to the d.o.f. where structural change occurs (a list of symbols is given in the Appendix).
The value of the fictitious mass is large enough not to induce numerical difficulty. This
value is not so sensitive and can be selected from a simple model examination. Normal
mode analysis for equation (3) gives a set of nf low frequency fictitious vibration modes
[ff ]. Then the generalized mass and stiffness matrices are given as

[GMf ] = [ff ]T[M + Mf ][ff ], [GKf ] = [vf ]2[GMf ] = [ff ]T[K ][ff ], (4)

where [vf ] is the diagonal matrix of natural frequencies.


A co-ordinate transformation is then performed to clean out the fictitious masses and
to form an actual basic case the stiffness matrix of which may differ from that of the
nominal case by [DKb ]. The transformation is based on the natural frequencies [vb ] and
eigenvectors [xb ] associated with the equation of free undamped vibration in modal
co-ordinates:

([GMf ] − [ff ]T[Mf ][ff ]){j f } + ([GKf ] + [ff ]T[DKb ][ff ]){jf } = {0}. (5)

The mode shapes calculated for the fictitious mass finite element model are transformed
to the basic case:

[fb ] = [ff ][xb ]. (6)

The basic case mode shapes [fb ] serve as a constant set of structural generalized
co-ordinates throughout the response analysis. More detailed explanation of the fictitious
mass modal approach are given in references [15, 16].

2.3.   


The linear relationship between the aerodynamic force acting on the nodal point and
the vertical displacement of the nodal point is obtained as

{Fs } = q[Q]{us }, (7)

where [Q] is the aerodynamic influence coefficient calculated from the DLM. The structural
displacement is transformed into modal co-ordinate as

{us } = [fb ]{u }, (8)


   - 827
where {u } is the modal displacement. Then the generalized aerodynamic force can be
written as

{F } = [fb ]T{Fg } = q[fb ]T[Q][fb ]{u } = q[Q] {u }, (9)

where the underlined quantities are the modal co-ordinate quantities, and [Q] is the
generalized aerodynamic influence coefficient matrix.
Generally, the unsteady aerodynamic influence coefficient matrix is calculated for a
discrete reduced frequency k rather than calculated as a continuous function of the circular
frequency v. Thus the aerodynamic influence coefficient matrices should be approximated
as rational functions. There are many methods [18] of rational function approximation,
but the Roger and Abel method [19] has been used here for simplicity and fast computation
time. The approximate form is

M+3
[Am ](ik)
[Q(k)] = [A1 ] + [A2 ](ik) + [A3 ](ik)2 + s . (10)
m=4
(ik) + km

Here the Ai ’s are calculated from a least square fit. The km ’s are constants to be determined
for best fit. A simplex direct search method [20] is used to determine km for minimizing
fitting error.

Figure 3. The natural frequencies of the flexible model: (a) the effect of torsional spring stiffness; (b) the effect
of beam thickness. –––, Rigid; ——, flexible.
828 .-.   . 

Figure 4. The effect of the torsion to plunge frequency ratio on the flutter characteristics of a flexible typical
section model (e/c = 0·375, t/c = 0·0128, vb /vh = 3·74): (a) flutter speed; (b) flutter frequency. –––, rigid
(present); ——, flexible; w, rigid (NASTRAN).

The transient aerodynamic force of an arbitrary wing motion can be written as

0 1 0 1 0 1
M+3 2M+3
b b
F (t) = q [A1 ] + s [Am ] u(t) + q[A2 ] u̇(t) + q[A3 ] ü(t) − q s [Am ]pm z m ,
m=4
Ua Ua m=4

(11)

where pm = (b/Ua )km and z m = f0t u (t) e−pm (t − t) dt.


The equation of motion for an aeroelastic system with concentrated structural
non-linearity can be written as

[M]{ü} + [C]{u̇} + {R(u)} = {F}, (12)

where {R(u)} is the elastic restoring force, which is a function of displacement. For
piecewise non-linearity, the restoring force can be written as

{R(u)} = [K]{u} + { f (a)}, (13)

where [K] is the linear stiffness matrix without freeplay, and { f (a)} is the restoring force
vector, the elements of which are zero except for the non-linear element. For freeplay
   - 829
non-linearity, { f (a)} is given in equation (1). The procedure for applying the fictitious mass
method to freeplay is to generate the nominal modes with a nominal pitch stiffness (in this
case zero), and then remove the pitch stiffness from the stiffness matrix and replace it by
the non-linear force vector.
Transformation of equation (12) into the modal co-cordinate ({u } = [fb ]{u}) gives

[GM]{ü } + [GC]u̇ } + {GR(u )} = [F ], (14)

where the generalized mass, damping matrix and restoring force vector are as follows:

[GM] = {fb }T[M]{fb }, [GC] = {fb }T[C]{fb },

[GR(u )} = [GK]{u } + {fb }T{ f (a)}. (15)

Usually, the damping matrix [GC] is not calculated from the finite element formulation.
Here we assumed the modal damping matrix to be given by

[GC] = 2[z][va ], (16)

Figure 5. The effect of thickness on the flutter characteristics of a flexible typical section model (e/c = 0·375,
Ka = 2400Nm/rad, va /vh = 1·412): (a) flutter speed; (b) flutter frequency. –––, rigid; ——, flexible.
830 .-.   . 

Figure 6. The limit cycle flutter characteristics of flexible typical section model with a freeplay
(Ka = 2400Nm/rad, va /vh = 1·412): (a) flutter speed; (b) flutter frequency. –––, D.F. (rigid); w, r, time
simulation (rigid); ——, D.F. (flexible); W, R, time simulation (flexible); — . —, flexible linear flutter; — . . —,
rigid linear flutter.

where [z] is a diagonal damping coefficient matrix, usually taken with values from 0·005
to 0·02, and [va ] is the natural frequency matrix of linear model.
The state variable and matrices are defined by
{n } = {u̇ }, {ṅ } = {ü }, [M
] = [GM] − q(b/Ua )2[A3 ],

[C
] = [GC] − q(b/Ua )[A2 ], 1 ],
[K
] = [GK] − q[A [A
m ] = qpm [Am ],

M+3
1 ] = [A1 ] + s [Am ].
[A (17)
m=4

By rearranging the above formulae, equation (14) can be written as

M+3
[M
]{ṅ } = −[C
]{n } − [K
]{u } − {f}T{ f (a)} + s [A
m ]{z m }, (18)
m=4

żm (t) = u (t) − pm z m (t), z m (0) = 0. (19)


   - 831
Combining equations (17) and (18), yields the final state space equation as

F {n ˙ } J K − [M
]−1 [C
] −[M
]−1 [K
] −[M
]−1 [A
4 ] ··· −[M
]−1 [A
M + 3 ]L
G {u ˙ } G G [I] [0] [0] ··· [0]
G
G G G G
g {z 4˙ } h = G [0] [I] −p4 [I] [0] [0] G
G * G G · G
* * [0] · *
G G G · G
f{z M˙ + 3 }j k [0] [I] [0] ··· −pM + 3 [I]
l
F {n } J F{f}T{ f (a)}J
G G G G
G {u } G G {0} G
× g {z 4 } h − [M
]−1 g {0} h. (20)
G * G G * G
G{z G G G
f M + 3 }j f {0} j

Figure 7. The limit cycle flutter characteristics of flexible typical section model with a freeplay
(Ka = 800Nm/rad, va /vh = 0·815): (a) flutter speed; (b) flutter frequency. –––, D.F. (rigid); ——, time simulation
(flexible); — . —, flexible linear flutter; — . . —, rigid linear flutter.
832 .-.   . 

Figure 8. A comparison of time responses of free oscillation between the fictitious mass model and the full
model: (a) linear case (s = 0); (b) non-linear case (a0 /s = 10). ——, Full simulation; w, fictitious mass model.

The number of equations in the matrix equation (20) is (2 + M) × N (M = number of


simple poles, N = number of normal modes). Integrating equation (20) gives the time
response of a non-linear system. Here, the 5–6th order Runge–Kutta–Verner algorithm
is used for the adaptive integration step.
Integration of equation (20) requires an initial condition. In non-linear analysis, the
response is dependent on the initial conditions, so the condition must be chosen carefully.
Since the computation is performed in the modal co-ordinate of the fictitious mass model,
the initial condition is to be expressed in terms of the modal co-ordinates. In this study,
a rigid body rotation of freeplay angle and the first mode of the linear model are combined
to assign an initial condition. Therefore the initial pitch amplitude is defined as
a0 = s + ae1 , (21)
where s is the freeplay angle, and ae1 is the pitch angle produced by the deformation
proportional to the first linear vibration mode.

3. RESULTS AND DISCUSSION


3.1.   
A linear flutter analysis has been performed prior to a non-linear analysis to determine
the aeroelastic characteristics of the linear mode. The model consists of a uniform
   - 833
aluminum beam, a torsional pitch spring and a plunge spring, as shown in Figure 2, where
a and h are defined as the pitch angle and the plunge displacement at the elastic axis
(pitching axis), respectively. The elastic axis is located at 37·5% chord. The value of the
plunge spring stiffness Kh is 500 000 N/m. The deformation of the beam is assumed to be
small so that the rotatory inertia effect is neglected. The chord length of the beam 15·6 cm
and the thickness of the beam is 2 mm unless otherwise specified. Eight beam finite
elements were used for structural analysis. Also, a reduced modulus of elasticity
(E' = E/(1 − n 2 )) was used for the plane strain condition. Structural damping coefficients
in all modes are assumed to be g = 2z = 0·01.
Linear flutter results have been obtained for both flexible and rigid models to investigate
the flexibility effect. Also, the computation has been performed for a torsional spring with
various stiffnesses and for a beam with various thicknesses. A rectangular wing (aspect
ratio = 20), which consists of 200 DLM elements (ten chord-wise and 20 span-wise), was
used to determine the two-dimensional aerodynamic pressure. In this analysis, a normal
mode approach was used to save computation time. Four vibration modes proved to be
sufficient to obtain a converged flutter speed in this analysis.
In Figure 3 are shown the natural frequencies of the flexible and the rigid models, where
va and vh are the uncoupled natural frequencies of pitch and plunge, respectively, and vb
is the natural frequency for the chordwise bending mode. The natural frequencies are
expressed as follows:
va = zKa /Ia , vh = zKh /m , vb = (1·506p)2zE'I/mL 3 . (22)
Here Ia is the rotational inertia of the beam about the elastic axis, m is the mass of the
beam, and vb is obtained from the free–free beam boundary condition. vn is the natural
frequency of the typical section model. In Figure 3, ‘‘rigid’’ means that the chordwise

Figure 9. The parameters map of a rigid and flexible typical section with a freeplay for air speed versus the
initial condition ratio (Ka = 2400Nm/rad): (a) rigid model; (b) flexible model. F, divergent flutter; C, chaotic
motion; L, limit cycle oscillation; P, periodic motion; ,, damped stable motion.
834 .-.   . 

Figure 10. The parameters map of rigid and flexible typical section with a freeplay for air speed versus the
initial condition ratio (Ka = 800Nm/rad): (a) rigid model; (b) flexible model. F, divergent flutter; C, chaotic
motion; L, limit cycle oscillation; P, periodic motion; ,, damped stable motion.

bending stiffness of the airfoil is infinite and the chordwise bending deflection is zero, and
‘‘flexible’’ means that the airfoil is flexible. The effect of the frequency ration (va /vh ) on
the natural frequency of the typical section is shown in Figure 3(a). Here, the various
frequency ratios have been determined by changing the torsional spring stiffness from 200
to 4000 Nm/rad. As the frequency ratio increases, the flexibility effect on the first and
second mode increases. The effect of the change in beam thickness on the natural
frequencies of the typical section is shown in Figure 3(b). The increase in beam thickness
results in increases in the mass and bending stiffness. The flexibility effect of the typical
section on the natural frequency increases as the thickness decreases.
Flutter analysis was performed by using the V-g method, for sea level conditions. The
divergence speed, the flutter speed and the frequency of the model of Figure 3(a) are shown
in Figure 4. For the rigid model, the present results are very close to those of
MSC/NASTRAN, in which strip theory is used with the Theodorsen function. The
divergence speed is lower than the flutter speed for va /vh Q 0·7. The chordwise bending
effect on the flutter becomes significant for va /vh q 1·0. The flutter speed of the flexible
airfoil is lower than that of the rigid airfoil for va /vh q 1·0. The divergence speed is lower
than that of the rigid airfoil. Also, the flutter frequency of the flexible airfoil is lower than
that of the rigid airfoil.
Thickness effects on the flutter characteristics are shown in Figure 5. As the thickness
of the airfoil decreases, the flutter speeds of the rigid and flexible airfoils decrease. The
   - 835
flutter speed of the flexible airfoil is lower than that of the rigid airfoil. The flexibility effect
of the airfoil on the flutter becomes significant as the thickness decreases.

3.2. -  


Non-linear aeroelastic analysis was performed for the model with a pitch freeplay
non-linearity. Rigid and flexible models with the frequency ratio (va /vh ) smaller than 1
and larger than 1 were selected for the computational model. The plunge spring stiffness
(Kh ) was set to 500 000 Nm/rad and the pitch stiffness (Ka ) was selected to be 800 Nm/rad
(va /vh = 0·815) and 2400 Nm/rad (va /vh = 1·412).
The frequency domain analysis using the describing function was performed first. The
effects of the limit cycle amplitude ratio (A/s) on the flutter speed were examined and are
shown in Figures 6 and 7. The case of Ka = 2400 Nm/rad is shown in Figure 6. The overall
trend is similar to that of Figure 4, since the equivalent linearized pitch stiffness increases
as the amplitude ratio increases as shown in Figure 1. The minimum flutter speed is
observed near A/s = 2·5. For a velocity above 150 m/s there are three types of motion:

Figure 11. The divergent flutter speed of a rigid and flexible typical section with a freeplay for air speed versus
the initial condition ratio: (a) Ka = 2400Nm/rad; (b) Ka = 800Nm/rad. w, rigid divergent flutter; r, flexible
divergent flutter; –––, rigid linear flutter; ——, flexible linear flutter.
836 .-.   . 

Figure 12. The bifurcation diagram of the pitch amplitude (Ka = 2400Nm/rad): (a) rigid model (a0 /s = 1·0);
(b) flexible model (a0 /s = 2·0).

static equilibrium, stable and unstable limit cycle oscillations. Stable limit cycle oscillation
occurs below the linear flutter boundary, and approaches linear flutter speed as the
amplitude ratio increases. The time domain analysis gives either stable limit cycle
oscillation or static equilibrium position according to the initial condition. For the
stable limit cycle oscillation the results of the time domain agree well with those of the
frequency domain analysis. The frequency of the limit cycle oscillation increases as the
amplitude ratio increases. The flexibility effect lowers the limit cycle flutter speed and
frequency. The results for the case of Ka = 800 Nm/rad are shown in Figure 7. In this case,
unstable limit cycle oscillation occurs above the linear flutter boundary. The frequency of
the unstable limit cycle oscillation increases as the amplitude ratio increases. The time
domain analysis shows more complicated responses like chaotic motion, as will be shown
later in this paper. The simple describing function method is not appropriate for this case.
To verify the fictitious mass modal approach in the structures with the freeplay
non-linearity, the free vibration analysis for the non-linear model was performed in the
time domain. Leading edge displacements obtained by the fictitious mass model are
compared with those given by a full d.o.f. model in Figure 8. Good agreement between
the two models is observed. Therefore the fictitious mass modal approach is appropriate
for the modal transformation of a structure with concentrated non-linearity.
   - 837
The time domain analysis was performed for the initial amplitude ratio (a0 /s) and the
free stream velocity. The results are represented in the parameter map as shown in
Figures 9 and 10. Time responses are classified into five categories: damped stable motion,
limit cycle oscillation, complicated periodic motion, chaotic motion and divergent flutter.
Here the periodic motion means the motion of period-n or the quasi-periodic motion,
including the period-1 motion which is not symmetric about the mean position. The
chaotic motion was recognized by phase plane diagrams, FFT analysis, Poincaré maps,
etc. Transient chaos is not considered here. The case of Ka = 2400 Nm/rad is shown in
Figure 9. A broad velocity region of the limit cycle oscillation is observed below the
divergent flutter speed in contrast to the linear case. The characteristics of the responses
are less affected by the initial pitch amplitude ratio. However, chaotic motion and periodic
motion appear for small amplitude ratios of the flexible model. The case of
Ka = 800 Nm/rad is shown in Figure 10. In this case various responses appear as the initial
pitch amplitude ratio and velocity vary. The rigid and flexible models show similar trends.
The effect of the initial pitch amplitude ratio on the divergent flutter speed is shown in
Figure 11. The case of Ka = 2400 Nm/rad is shown in Figure 11(a). The divergent flutter
speed is almost the same as the linear flutter boundary. The effect of initial pitch amplitude
ratio is not significant in the rigid model. When the initial amplitude ratio of the flexible

Figure 13. The bifurcation diagram of the pitch amplitude (Ka = 800Nm/rad, a0 /s = 1·0): (a) rigid model;
(b) flexible model.
838
.-.   . 

Figure 14. The timehistory and phase plane diagram of the rigid model (Ka = 2400Nm/rad, a0 /s = 1·0, Ua = 180m/s): (a), (c) pitching motion a; (b),
(d) plunging motion h.
   -

Figure 15. The time history and phase plane diagram of the flexible model (Ka = 2400Nm/rad, a0 /s = 2·0, Ua = 160m/s): (a), (c) pitching motion
a; (b), (d) plunging motion h.
839
840
.-.   . 

Figure 16. The time history and phase plane diagram of the rigid model (Ka = 800Nm/rad, a0 /s = 1·0, Ua = 200m/s): (a), (c) pitching motion a;
(b), (d) plunging motion h.
   -

Figure 17. The time history and phase plane diagram of the rigid model (Ka = 800Nm/rad, a0 /s = 1·0, Ua = 205m/s): (a), (c) pitching motion a;
(b), (d) plunging motion h.
841
842
.-.   . 

Figure 18. The time history and phase plane diagram of the flexible model (Ka = 800Nm/rad, a0 /s = 1·0, Ua = 190m/s): (a), (c) pitching motion
a; (b), (d) plunging motion h.
   -

Figure 19. The time history and phase plane diagram of the flexible model (Ka = 800Nm/rad, a0 /s = 1·0, Ua = 200m/s): (a), (c) pitching motion
a; (b), (d) plunging motion h.
843
844 .-.   . 
model is small, the divergent flutter speed is higher than the linear flutter boundary. The
case of Ka = 800 Nm/rad is shown in Figure 11(b). In this case the divergent flutter speed
is higher than the linear flutter boundary, and decreases as the initial pitch amplitude ratio
increases. It is observed that the flexibility effect generally lowers the divergent flutter
speed.
The bifurcation diagrams for the pitch angle of the elastic axis with velocity change are
shown in Figures 12 and 13. The displacement of the pitch angle the velocity of which is
zero are dotted when the motion is settled to steady motion. The case of Ka = 2400 Nm/rad
is shown in Figure 12. Both the rigid and flexible model show similar trends. Limit cycle
oscillations are observed after the damped stable motion. The case of Ka = 800 Nm/rad
is shown in Figure 13. More complicated structures are observed than in the previous case.
Hopf bifurcation and period doubling phenomena are observed. A chaotic motion band
is seen between the periodic motion bands.
The detailed time history and phase plane diagrams for the pitch and plunge motions
of the elastic axis are shown in Figures 14–19. The phase plane diagrams were drawn after
the motion settled to steady motion. The limit cycle oscillation of the case of
Ka = 2400 Nm/rad is shown in Figures 14 and 15. Figure 14 is for the rigid model and
Figure 15 is for the flexible model. In the flexible model, fluctuation of the pitch velocity
is observed due to the flexibility effect, whereas the plunge motion shows a more smooth
phase plane curve. The results for the case of Ka = 800 Nm/rad are shown in
Figures 16–19. Figures 16 and 17 are for the rigid model and Figures 18 and 19 are for
the flexible model. Figure 16 represents the periodic motion, and Figure 17 shows the
chaotic motion of the rigid model. The shape of the phase plane diagram resembles that
of buckled plate flutter [21]. The periodic and chaotic motions are shown in Figures 18
and 19. The shapes of the phase planes are similar to those of the rigid model: however,
fluctuations of the pitch motion curve in the phase plane diagram are observed due to the
flexibility effect.

4. CONCLUSIONS
Aeroelastic analysis for a flexible two-dimensional airfoil with a freeplay non-linearity
has been performed. Both frequency domain and time domain analysis were performed
for the rigid and flexible models. Concluding remarks from the analysis are as follows.
1. The flexibility effect in the chordwise bending deflection on the linear flutter becomes
significant when the frequency ratio of the pitch to the plunge (va /vh ) is greater than 1.
The thickness effect on the linear flutter becomes significant as the thickness of the airfoil
decreases.
2. From the comparison with the full d.o.f. model the fictitious mass modal approach
is shown to be an appropriate modal method for the non-linear aeroelastic analysis
including concentrated non-linearity.
3. Complex periodic motion and chaotic motion are observed, and occur more easily
when the frequency ratio is smaller than 1.
4. The flexible model behaviour shows trends similar to that of the rigid model.
However, the flexibility makes the pitch velocity fluctuate.
5. When the frequency ratio is larger than 1, the divergent flutter speed is almost the
same as the linear flutter boundary. However, the flexibility increases the divergent flutter
speed for small initial pitch amplitude ratio. When the frequency ratio is smaller than 1,
the divergent flutter speed is larger than the linear flutter boundary, and approaches it as
the initial pitch amplitude ratio increases.
   - 845
REFERENCES
1. D. S. W, H. W. R and R. E. A 1957 Journal of Aeronautical Sciences
24, 57–63. An investigation of effects of certain type of structural nonlinearities on wing and
control surface flutter.
2. S. F. S 1959 Journal of Aerospace Science 28, 25–32, 45. An approximate analysis of
nonlinear flutter problems.
3. S. C. MI, J., R. E. R, J and W. P. R 1981 Journal of Aircraft 18, 1057–1063.
Experimental and theoretical study of nonlinear flutter.
4. C. L. L 1986 American Institute of Aeronautics and Astronautics Journal 24, 833–840.
An iterative procedure for nonlinear flutter analysis.
5. Z. C. Y and L. C. Z 1990 Journal of Sound and Vibration 123, 1–13. Analysis of limit
cycle flutter of an airfoil in incompressible flow.
6. L. C. Z and Z. C. Y 1990 Journal of Sound and Vibration 138, 245–254. Chaotic motions
of an airfoil with non-linear stiffness in incompressible flow.
7. L. O. B and W. E 1988 Journal of Aircraft 25, 1060–1068. Application of transient
aerodynamics to the structural nonlinear flutter problem.
8. A. J. H, R. M. L, W. E, G. G, I. Q and A. K. A
1990 AIAA-90-1034-cp. Chaotic response of aerosurfaces with structural nonlinearities.
9. A. J. H, J. A. Z, W. E and I. Q 1992 AIAA-92-2547-cp. Chaotic and
nonlinear dynamic response of aerosurfaces with structural nonlinearities.
10. S. J. P, B. H. K. L and H. A 1993 AIAA-93-1474-cp. An analysis of the
post-instability behaviour of a two-dimensional airfoil with a structural nonlinearity.
11. S. J. P, B. H. K. L and H. A 1994 AIAA-94-1546-cp. The aeroelastic response
of a two-dimensional airfoil with bilinear and cubic structural nonlinearities.
12. J. D and J. D. C. C 1959 OSR Technical Note 59–787. On the aeroelastic
characteristics of low aspect ratio wings with chordwise deformation.
13. J. D. C. C 1974 Noise, Shock, and Vibration Conference, Monash University, Melbourne,
Australia. On the hydrodynamic flutter anomaly.
14. E. A and W. P. R 1969 American Institute of Aeronautics and Astronautics Journal
7, 279–285. A doublet-lattice method for calculating lift distributions on oscillating surfaces in
subsonic flows.
15. M. K and C. D. W 1994 Journal of Aircraft 31, 396–403. Modal coordinates for
aeroelastic analysis with large local structural variations.
16. M. K and C. D. W 1994 Journal of Aircraft 31, 404–410. Time simulation of flutter
with large stiffness changes.
17. R. M. L and R. M. T 1980 American Institute of Aeronautics and Astronautics
Journal 18, 1245–1251. Flutter analysis of missile control surface containing structural
nonlinearities.
18. S. H. T and W. M. A, J. 1988 NASA TP-2776. Nonlinear programming extensions
to rotational approximation methods of unsteady aerodynamic forces.
19. I. A 1979 NASA TP 1367. An analytical technique for predicting the characteristics of a
flexible wing equipped with an active flutter suppression system and comparison with wind
tunnel data.
20. J. A. N and R. M 1969 Computer Journal 7, 279–285. A simplex method for function
minimization.
21. E. H. D 1982 Journal of Sound and Vibration 85, 333–344. Flutter of a buckled plate as
an example of chaotic motion of a deterministic autonomous system.

APPENDIX: LIST OF SYMBOLS

[C] structural damping matrix


E Young’s modulus
E' reduced modulus of elasticity
f (a) non-linear restoring force
{F} external force vector
[GM] generalized mass matrix
[GK] generalized stiffness matrix
g modal structural damping coefficient
846 .-.   . 
h plunge displacement of elastic axis
k =vb/Ua, reduced frequency
Ka pitch spring stiffness
[K] stiffness matrix
[DK] change in stiffness matrix
[M] mass matrix
[Mf ] fictitious mass matrix
q dynamic pressure
[Q] aerodynamic influence coefficient matrix
{R} elastic restoring force
s freeplay angle
Ua free stream velocity
z m augmented states
a pitch rotation angle
r density
[ff ] mode vector of the fictitious mass model
[fb ] mode vector of the basic system
[xb ] transformation matrix of the basic system
v natural frequency in radian
d describing function
z =g/2, modal structural damping coefficient
n Poisson ratio

Subscripts
0 initial condition
b basic system
f fictitious mass model

Miscellaneous
(·) =d()/ dt
– (underline) generalized co-ordinate
– (overbar) amplitude of harmonic motion

S-ar putea să vă placă și