Vous êtes sur la page 1sur 8

1

Flutter Suppression by Active Control - Theory and Experiment

1. INTRODUCTION. FLUTTER & FLUTTER SUPPRESSION

Aeroelastic phenomena of aircraft structures appear as a result of interactions


between deformations of the elastic structure and the aerodynamic forces induced by the
structure deformations ([1], [2]). The most important aeroelastic phenomenon is flutter,
i.e. a self-excited oscillation of the structure under the action of the aerodynamic loads.
Flutter often exhibits an explosive behavior with catastrophic consequences; the structure
must thus be definitely flutter-free throughout the flight envelope of the aircraft.
The concept active flutter suppression appears as part of a rather new technology in
aviation, the so-called control configured vehicle ([3], [4]). It means controlling by some
active devices - typically through activated controls - the natural instabilities of the
structure, so as to make this flutter-free beyond its "nominal" flutter boundary. The
validity of this concept has been proven theoretically and experimentally ([5]-[9], etc.).
With all that, it must be said that, Friedmann [5], "…currently there is no vehicle in
production that uses active flutter suppression" and "…much remains to be done before
one can consider the routine incorporation of such systems in production aircraft".
Despite this, the concept still retains its actuality. One of the main reasons is that it is
strongly related to the "much more realistic" problem of controlling the "conventional"
aeroelastic structural vibrations.
Consequently, the active structural control
is basically investigated and designed both
for flutter prevention and structural load
alleviation.
In this paper the results of a
theoretical and experimental research are
presented.

Transducers
M 2. THE EXPERIMENTAL SET-UP
w Plunge A
 An overall view of the
α Pitch Fβ
β Flap
aeroservoelastic system installed in the
w low-speed wind tunnel at the Faculty of
Aerospace Engineering is given in Fig. 1.
α
It consists of a constant-chord wing
R model mounted on a moving platform and
T elastically suspended both in plunge and
pitch motions, to thus reproduce the state
Mechanical Devices

T Translation: Rail - Ball Bearing Block


R Rotation: Rod - Ball Bearings Assembly
M Flap Driving-Motor Housing
A
Fig. 1 The experimental aeroservoelastic model
F installed in the wind tunnel
2
Flutter Suppression by Active Control - Theory and Experiment

of a typical deformable aircraft wing.


The model represents basically a coupled two-degree of freedom dynamical system. A
more detailed drawing of the wing itself is given in Fig. 2.
The system parameters were chosen such that the model would be capable of
exhibiting autooscillations (flutter) at relatively moderate air speeds and reduced
frequencies compatible with the tunnel.
For the control of the model, a simple electrically driven trailing-edge flap was
added over the entire wing span.
 For tracking the plunge/pitch system motion, a couple of linear/angular
transducers have been mounted on wing. Further, a flap deflection angular sensor has
been provided for in order to have a precise control on this latter's movement.
 For sensing the motion,

Z
a couple of linear accelerometers
C= 182
A1 and A2 are provided for on the
X = 77 CG

X
S CG
X AR
model, the first one lying on the
Disc BH
[240883] w AR
O moving platform, the other some
L= 400
distance l2 apart from the axis.
 [200]

A 2
It is clear that this device is
A 1 capable of sensing both the
l 2
translation and rotation
Fig. 2 The wing-model. Geometry. Accelerometers accelerations of the model.
3. SYSTEM DYNAMICS AND
EQUATIONS

z
"e"
 The conventional system is
L = b(½+a)
[L] drawn in Fig. 3. Notations are those in
MCE 
standard textbooks on Aeroelasticity.
MS
The added Table 1 lists all principal
w system parameters.
 bc
bx ba  The governing equations of this
CE two-degree of freedom system (including
S CG M [F ] U[]
x
K K the effect of a flap deflection) can be
b b written ([1], etc.) in matrix form as - recall
Kw
that both w(t ) and  (t ) denote small
CV CA
perturbations from an initial equilibrium
Fig. 3 Aeroelastic system - Notations
state -:
3
Flutter Suppression by Active Control - Theory and Experiment

M  S w C 0 w K w  L 


0
  w   w      (1)
 S I   0 C   0 K    M CE 
Here L (lift, positive up) and MCE (aerodynamic moment in CE, positive nose-up)
represent the destabilizing aerodynamic forces; their definition contain the air velocity U.
The second entry above is a linear viscous-damping term that models friction effects.
 The aerodynamic forces. In a pure-stationary approximation, the aerodynamic
forces are functions of the model-incidence  (t ) and flap-deflection  (t ) only and
in phase with them. The formulas are then (with "L" designating the wing span)

 L[N]  0 CL  w(t )  CL


  " L " U b    " L " U b   (t ) (2)
2 2
  
 CE
M [N  m]  0 2bC M CE  ( t )  2bC M CE

 L[N]  not! w


or, simply    Fwα     Fβ   (3)
 M CE [N  m]  
 State-space representation. The system of two second-order differential
equations (1) or (4) can be cast into a state-space form. We choose the "natural" states
def !
T
x   w(t )  (t ) w(t )  (t )  (4)
With the matrices in (1) denoted M (mass), Z (damping) and K (stiffness), one
readily builds the canonic system as ("0" and "I" stand for zero and unity matrices)
0 2 2 I 2 2  021 
x(t )   x ( t )   1 2    (t ) (5)
M 1  K  Fwα (U 2 )  M 1Z  M Fβ (U ) 

or, in standard notations x  A wα  x  Bβ   (6)


Table 1 - System parameters
Symbol Unit Value Designation
C = 2b [m] 0.182 wing chord (b = semichord)
L [m] 0.4 wing span
F 25%C aerodynamic center [ ¼C ]
CE(AR) ~ 28%C "elastic center" (rotation axis)
CG ~ 42%C mass center of the model
S ~ 70%C flap hinge line
M [kg] 1.855 model mass
Sα [kgm] 0.01729 model mass static moment about CE (static unbalance)
Iα [kgm2] 2.745*10-3 model mass moment of inertia about CE
w [m] "plunge"-displacement (positive up)
α [rad] "pitch"-rotation (positive nose-up)
β [rad] flap deflection (positive down)
Kw [N/m]=[kg/s2] 321.6 "plunge"-suspension stiffness constant
Kα [Nm/rad]=[kgm2/s2] 1.25 "pitch"-suspension stiffness constant
4
Flutter Suppression by Active Control - Theory and Experiment

 Flutter. The conventional flutter problem is formulated by simply discarding


the inhomogeneous part in (5) or (6), that is through building the system "free dynamics"
x  A wα (U 2 )  x (7)
 Actuator dynamics - the extended system. For practical applications on flutter
suppressors, one must take into account the dynamics of the command line (actuator).
For the present study, as in [7], a simple first-order time-delay model will be considered
 (t )     (t )    u(t ) (8)
The nominal system (6) and the actuator equation (8) are now combined to build the
extended system (in this representation the flap deflection becomes a state variable!)
not!  x (t ) 
 x  A wα Bβ  x  041  
        u or x     x  A  x  B  u (9)
  01 2        (t ) 

3. THE SYNTHESIS OF CONTROLLER FOR FLUTTER SUPPRESSION

The objective of this section is to formulate the problem of the active flutter
suppression and to build a methodology for the synthesis of an ad-hoc controller.
 The LQR regulator. Consider the linear system (generic notations!)
x  Ax  Bu ; y  Cx (10)
and a full-state feedback connection
u( t )   K  x ( t )  x  ( A  B  K )  x  B  u ; y  C  x (11)
The optimal regulator results through the minimization of the quadratic index

1  
I  u( t )    xT (t )  Q  x(t )  uT (t )  R  u(t )  dt  min (12)
2 0  
With Q and R prescribed, the solution to this is the well known Riccati regulator.
 The "direct connection" scheme. Implicit in the full-state feedback concept is
the condition of full access to all states. Since in the proposed set-up accelerometers are
used as sensors, the state vector (displacements and velocities) must be built from
measured accelerations. In this respect, an attempt has been made to "construct" the
states by simply successively integrating the accelerations - Fig. 4. The integration
operation implies some initial conditions which are meaningless in an "indefinite"
process as flutter is. Despite that, the proposed scheme has been proven effective in
suppressing the flutter. The explanation comes from the fact that the action of a feed-back

Initial
Command OUT Arithmetic block
conditions

u(t)[]
+ Servo  + w w (t)
-  
(t)[] w
β-sensor +  (t)
w+ 

U[m/s]
 

Regulator
u = – Kx

Fig. 4 System with controller - The "direct connection" sqeme


5
Flutter Suppression by Active Control - Theory and Experiment

regulator is to bring the system to the equilibrium point, that is to x  0 , anyway!

 The state estimator. The alternative to the former is the observer (Fig. 5), a "pure
mathematical" device through which the states are being reconstructed from the output.
With x̂ denoting the state estimate, the dynamics of the system plus observer is [10]
x   A BK  x  B 
u(t )   K  xˆ (t ) and       u (13)
xˆ   LC A  BK  LC xˆ  B 
The asymptotic estimator results from the condition ( x is the error of estimation)
-K
x̂ x  x  xˆ  x  ( A  L  C)  x ; t    x  0
Further, one shows that this last can be cast into
OBSERVER
[ A, B, C, L ] the problem of minimizing a quadratic index similar to
u y (12) - see for instance [10]. This again leads to a
SYSTEM
+
u  K  xˆ
[ A, B, C ] Riccatti equation which finally gives the L matrix:
Fig. 5 System plus observer AP  PAT  PCT R-1CP  Q  0  L  PCT R-1 (14)
The regulator and estimator can be designed independently (separation principle!).
The C matrix above must be defined. In the proposed layout - Fig. 2 - what the two
accelerometers measure would be, in a linear approximation, two linear accelerations
 a1  w ,  a2  w  l2   (15)
One interprets these accelerations as output of the system in a generic sense, that is
def !  a 
1  w  ! 1 0   w  not! w
y          F  (16)
a2   w  l2  1 l2     
With the derivatives in RHS extracted from the initial system (1), one gets finally
 w(t ) 
 
 1  2   1   (t )   1
   F  M Fβ (U )    (t )
2
y (t )  F   M  K  Fwα (U )   F  M Z  
 w(t ) 
 (t ) 
or, in standard notation y (t )  C  x ( t )  D   ( t ) (17)

4. FLUTTER SUPPESSOR DESIGN AND DEMONSTRATION

 System identification. Tests have shown the model to enter a violent flutter at
exactly 17.8 m/s and ~17 rad/s frequency. These correspond to a reduced frequency of
about 0.09; it follows that the stationary assumption is fully justified for this study.
Now, the system free dynamics (7) has been simulated through the Laplace
transform and the root-locus (RL) procedure - Fig. 6 - with the basic mechanical
parameters determined experimentally. In addition, values in Table 2 have been entered.
6
Flutter Suppression by Active Control - Theory and Experiment

Table 2. System data


Symbol Unit Value Description Observation
a - -0.44 "EA" location aft of midchord (in semichords) standard notation
c - 0.4 "S" location aft of midchord (in semichords) standard notation
l1 [m] 0 position of the A1 accelerometer from "EA" -
l2 [m] 0.180 position of the A2 accelerometer from "EA" -
 [rad/s] 200 lag constant for the actuator dynamics
"plunge" damping coefficient def !
ζw - 0.1 Cw   w  2 K w / M 
(fraction of critical damping)
"pitch" damping coefficient def !
ζα - 0.1 Ca   w  2 K / I 
(fraction of critical damping)
 The aerodynamic coefficients are by far the most decisive model parameters.
– In a first attempt (as many articles on the subject do [7], [8]), we have used the
conventional 2D-values (Appendix). Figure 6, a reveals flutter at a speed of about 9 m/s,
which differs drastically from the actual value; this approach fails as expected.
– In a second attempt, one has tried to determine those parameters experimentally.
Table 3 lists the coefficients used in the analysis compared to some theoretical
counterparts – the differences are evident! Figure 6, b shows flutter at a speed of about
15.5 m/s, still different from the measured one. Despite that, one has concluded that this
mathematical model was "the best that could be obtained" within a simple approach and
has decided to validate the model as it is and further try to use it for flutter suppression
studies. We retain the experimental value as actual flutter speed, that is UF = 17.8 [m/s].
Table 3 . Aerodynamic coefficients of the model
Lifting-line [1] Coefficients used in the present analysis
Symbol 2D-theory   L / C  2.2
Value Observation
(no end-plates effect)
C L 6.2832 (2) 3.1778 2.17 measured in tunnel

CL 4.1516 – 0.96 measured in tunnel

CMCE 0.1885 – 0.0651 calculated with the measured C L

CM CE
adopted from 2-D theory
– 0.5170 – – 0.5170
(apparently overestimated!)
7
Flutter Suppression by Active Control - Theory and Experiment

 Flutter suppression system (FSS) design


– We set as design point the actual flutter speed of the model U des  U F  17.8[m/s] .
– For this condition, the controller (namely the minimum energy one) has been
calculated as well as the state observer. The entries and the solutions are given in Table 4.
Table 4. Controller and estimator design
Q = 0 | R=I K des  1.7892 0.0825 0.0496 0.0146 0.0327

25
15
0
a b c
U=0
jω U=0 U=0

10 10
6
20 12
15 12
0 14 26 DESIGN
8
FLUTTER 14 FLUTTER POINT
16 22
U = 17.8 U = 17.8
10 10 16
U des  U F  17.8 8
14 16 18
18
16
15
0 12
6 10 10
U=0 U=0
U=0

σ σ σ
10
–4 –2 0 2 –4 –2 0 2 –4 –2 0 2
Fig. 6 Numeric simulation
a) 2D Aerodynamics; b) Measured coefficients; c) The controlled system

U des  U F  17.8  0.9761 0.1472 0.1980 5.8821 0.0075 


Q = I | R=I LTdes   
 0.1089 0.9575 0.7515 2.3578 0.1527 

 Numerical simulation. By definition, the flutter suppressor must stabilize the


system for all speeds beneath the design point. Due to its intrinsic properties however, the
Riccati-regulator would be effective even far beyond that point - see RL on Figure 6, c.
 Tunnel tests. Decisive in a project of this art is, naturally, the experiment itself.
An extensive test program has been conducted on the system (out of which a few results
obtained with the system plus observer layout are shown - Fig. 7). The conclusions are:
1) The efficiency of the flutter suppressor has been demonstrated. The proposed
control laws are effective not only at the design speed but even above that; this fact is
otherwise in accord with the engineering objective of setting-up "constant" controllers.
8
Flutter Suppression by Active Control - Theory and Experiment

30
U = UF = 17.8 U = 19.5 >UF
20
w [ cm ] w [ cm ]
α [deg] α [deg]
10 β [deg] β [deg]

-10

-20
FSS ON t [sec] FSS ON t [sec]
-30
0 1 2 3 4 5 6 0 1 2 3 4 5 6
Fig. 7 Flutter Suppression System Demonstration
2) Despite the mentioned inconsistency of the mathematical model with the physical
system, the control laws based on the former are still operative. To an extent, this fact can
be theoretically explained through the intrinsic robustness of the Riccati-controller itself.
3) Similar conclusions may be drawn regarding the dynamic response (not shown).
Although strongly influenced by the controller, the system wing plus flap preserves its
full functionality as a "lift producing device" even in the initially unstable domain.

Appendix. 2D - aerodynamic influence coefficients ([1])


C L  2 C L  2T10 
2CM
CE
 CL 
e e
; 
b b
 12  a  
2CM
CE
 T4  2aT10

T4   cos 1 c  c 1  c 2 ; T10  1  c 2  cos 1 c

REFERENCES
1. Bisplinghoff, R.L., Ashley, H., Halfman, R.L., Aeroelasticity, Addison-Wesley, Cambridge, Mass, 1955.
2. Petre, A., Teoria aeroelasticităţii – Fenomene dinamice periodice, E.A., Bucureşti, 1973
3. Thompson, G.O., Kass, G.J., Active Flutter Suppression – An Emerging Technology, Journal of
Aircraft, Vol.9, No. 3, March 1972, pp. 230-235.
4. Garrik, I.E., Aeroelasticity – Frontiers and Beyond, Journal of Aircraft, Vol.13, No. 9,
Sept. 1976, pp. 641-657.
5. Friedmann, Peretz P., Renaissance of Aeroelasticity and Its Future, Journal of Aircraft,
Vol. 36, No. 1, Jan.-Feb. 1999, pp. 105-121.
6. Horikawa, H., Dowell, E.H., An Elementary Explanation of the Flutter Mechanism with Active Feedback
Controls", Journal of Aircraft, Vol. 16, No. 4, April 1979.
7. Liebst, Bradley S., Accelerometer Placement in Active Flutter Suppression Systems, Journal of
Guidance and Control, Vol. 10, No. 5, Sept.-Oct. 1987, pp. 441-446.
8. Jeonghwan, K.O., Kurdilla, A.J., Strganac, T.W., Nonlinear Control of a Prototypical Section with
Torsional Nonlinearity, Journal of Guidance and Control, Vol. 20, No. 6, Nov-Dec 1997.
9. Lind, R., Brenner, M., Robust Aeroservoelastic Stability Analysis, Springer, London, Berlin, 1999.
10. Unbehauen, R., Systemtheorie 1, R. Oldenbourg Verlag, München, Wien, 1997.

Vous aimerez peut-être aussi