Aeroelasticity Proc - Gojny PDF

Download as pdf or txt
Download as pdf or txt
You are on page 1of 17

DAMPING AUGMENTATION OF FLEXIBLE STRUCTURES

— A ROBUST STATE SPACE APPROACH —

Udo B. CARL
Marcus H. GOJNY
Technical University of Hamburg–Harburg
Institute of Aircraft Systems Engineering (2.08)
21071 Hamburg
Germany
Email: gojny@tu–harburg.de

The present paper addresses the regulation of weakly damped aeroelastic wing structures by means of the primary
flight controls (PFC). Synthesizing a sufficient low–order linear multi model system of the aggregate aeroservoe-
lastic plant requires the design of a robust state feedback. This ensures a remarkably augmented damping ratio in
comparison with the original aeroelastic system. Moreover, the influence of the actuation system performance on
the aeroservoelastic damping augmentation feedback is investigated. The suitability of the resulting controller is
verified by simulation as well as validated by real tests.

Keywords: aeroservoelasticity, flexible aircraft, analytical modelling, structural vibrations, typical


section, primary flight control actuation system, hydraulic actuator, uncertain paramters, robust
control, parameter space design

1 INTRODUCTION

Aeroelastic control can be more challenging than conventional controlled structures prob-
lem, in that the dynamics of the system change dramatically with the flight conditions. It
holds the promise of significant improvements in performance: reducing the ambient vibration
level, increasing the maneuver responsiveness and stabilizing an otherwise unstable system [4].

Considering the usual mission of a large commercial aircraft, aerodynamic forces and
moments entail a substantial deformation of the elastic structures of fuselage, tailplane and
wing. Each incremental change of the structural shape yields a new aerodynamic state which
causes aeroelastic interaction. Due to the increasing sizes of transport aircrafts, the spectral
gap between flightmechanical motion and internal structural modes decreases continuously.
Hence, dynamically coupled modes of the flexible aircraft arise, which can be excited by
gust loads as well as manoeuvres and flight mechanical stability augmentation functions.
These oscillatory modes are usually characterized by weak damping without supplementary
measures. The resultant vibrations reduce the fatigue life of the structure and may lead, under
worst conditions, to a complete loss of the aircraft’s controllability. In order to solve this
problem for light weight structures, new perspectives aim at an additional functionality of
the primary flight control surfaces, to ensure an active modal damping augmentation. Due to
changing flight operation conditions, as e.g. speed, and masses of the dynamic system by e.g.
fuel in the wing, the physical parameters of the present plant vary considerably. Moreover,
the required control circuit consists of a substantial nonlinear electrohydraulic servo actuation
system. Although the complete system is characterized by high order, nonlinearities and
significant parameter uncertainties, a linear controller of low order is aspired. In order to
outline a general approach, the basic plant consists of a quasi–stationary formulation of the
aerodynamics linked together with a mechanical airfoil model, which is connected with the
model of the electrohydraulic actuation system. Adapting its essential overall properties to a
seventh order linear multiple model system, allows the design of a robust static state feedback
which provides the desired additional active damping function.

The paper is organized as follows: The basic aeroservoelastic plant is presented in section 2.
The performance specifications and principals of the applied controller design procedure are
outlined in section 3. Section 4 presents the practical validation of the closed loop system as
well as current PFC actuation system aspects. Concluding remarks finalize this paper.

2 MODEL AGGREGATION

2.1 Aeroelastic model


Although, in practice the dynamic instability of an elastic body in an airstream usually de-
scribed as flutter, is a complicated phenomenon, this approach is based on a quite simple
aeroelastic system. It exhibits some of the dominat properties, which can be met at a current
flexible wing configuration in more formidable guise. With regard to an experimental variable
camber wing of large aspect ratio, the incisive reduction leads to a two–dimensional structure,
which is characterized by the center of gravity C, the elastic axis E, the aerodynamic center A
and the hinge line G (figure 1). Immediately, this structure can be adapted to the representative

k
k q
h

(a) Swept, twisted wing (b) Uniform cantilever wing

Figure 1: Mechanical structure


h
k h .
q (y ) a = a 0 + q + h / U
q k q
z M
L
a M H
A E C G

(a) Representative section [6] (b) Quasistationary aerodynamics

Figure 2: 3DOF aeroelastic model


section [2], a rigid airfoil section, suspended in an airstream and with degrees of freedom in
bending and torsion by suitable suspension from two sets of springs. Assigning the geomet-
ric and inertial properties of this system to a wing cross section at three–quarters of the wing
span yields a basic mechanical representative dynamic model [6]. Hence, the set of generalized
coordinates is denoted by a downward vertical displacement h of the line of attachment (E),
a leading edge up angular rotation θ about this line and the trailing edge downward aileron
deflection ζ
h iT
q̃ = h θ ζ : (1)

In further discussion the latter will be ommitted by inserting a suitable actuator representative.
With the total stiffness kh ; kθ , the total mass m, static unbalance SθP and mass moment of inertia
JθP about the generalized reference point P 2 f A; E ; C; G g, the sum of potential energy of
strain V and the kinetic energy T can be expressed as function of the generalized coordinates
and its derivatives. By assumption that small deflections are to be expected, the application of
Lagrange’s Equations
   
∂ (T ; V ) ∂ (T ; V )
; dtd ∂ q̃˙
+
∂ q̃
+Q = 0 (2)

on this lumped parameter system in the presence of airstream [4] results in a non conservative
linear system. The mechanical equation of motion reads:
2 3
m SθE SG
6 f 7
6 E
6 Sθ
4
JθE (xθ + aζ ) S G G
f + Jf
7 ¨
7 q̃ +
5

SG
f (xθ + aζ ) S G G
f + Jf J fG
| {z }
M̃S
2 3 2 3 2 3
d
6 h
0 0
7
k
6 h
0 0
7 6
;L 7
 6
4
7 ˙ 6
5 4
7
5
6
+ 6 0 dθ 0 7 q̃ + 6 0 kθ 0 7 q̃ = 6 M + e L
4
7
7
5
: (3)
0 0  0 0  MH
| {z } | {z }
B̃S K̃S

In this case the () elements in (3) will be determined by the electrohydraulic actuation system
model. The right hand side of (3) denotes the lift L, the summarised aerodynamic momentum
M and the hinge momentum MH .

The essential features of the aerodynamic physics are described by a quasi–stationary


model which reflects incompressible wing–aileron lift characteristics [4]. The effective angle
of attack α(t ) then results as sum of a steady state initial value and additional terms from wing
torsion and bending
ḣ(t )
α(t ) = α0 + θ(t ) + : (4)
U
The coupling between the aerodynamic forces and the generalized mechanical motion leads to
a set of static gain matrices
2 3 2 3
6
;L 7 6
; ρ
2 S cLα 0 0
7
6
6 M +eL
4
7 6
7 = U 6
5 4
ρ
2 S (c cMα + e cLα )
7
0 0 7 q̃˙ +
5
 (5)
ρ
MH 2 Sf c f cHα 0 0
| {z }
B̃A
2 3 2 3
6
0 ; ρ2 S cL ;ρ S cL ; ρ2 S cL
 7 6 7
α ζ α

 62
+U 6 0
4
ρ
2 S (c cMα + e cLα ) ρ S c cMζ + e cLζ
7 6
7 q̃ + U 6
5 4
2 ρ
2 S (c cMα + e cLα )
7
7 α0 :
5
ρ ρ
0 2 Sf c f cHα ρ S f c f cHζ 2 Sf c f cHα
| {z } | {z }
K̃A F̃A

Assembling (3) and (5), results in the aeroelastic equation where the homogeneous aerody-
namic terms have been moved to the left hand side
 ;  ; 
M̃S s 2 + B̃S ; U B̃A s + K̃S ; U 2 K̃A q̃(s) = U 2 F̃A α0 : (6)

Bearing in mind that the aileron deflection is due to the control law of the electrohydraulic
PFC actuation system, ζ truely does not represent a third degree of freedom (DOF) within the
present system. Thus, the 3DOF system (3) must be reduced to a 2DOF plant (q̃ ! q). The
third column elements of (6), which describe the inertia coupling between aileron motion and
wing, then must be considered as additional input. Whereas the equation represented by the
third row elements of (6) describe the load input of the PFC actuation system, the third column

+
m

Q
[k g ]

-
m
m

-
U U +

U [m s -1 ]

Figure 3: General operating domain


contains the inertia effects of a control surface deflection on the wing. Therefore, regarding
the aeroelastic subsystem, the number of generalised coordinates (1) decrease. Defining the
structural state and subsystem input vector
h i h iT h iT
xS = qT q̇ T = h θ ḣ θ̇ ; uS = ζ ζ̈ α0 (7)

and considering the resulting 2–by–2 mass, damping and stiffness matrices yield the equivalent
state–space realisation
2 3


0 I
ẋS = 4 5 xS + (8)
;MS ;1 ; KS ; U KA

;M ;1 (B S ; U BA )
2
S
2 3
0 0 0
6 7
6 7
 6 7
0 0 0
+6 7 uS
6
4 ;ρU 2 S cL ζ
SG
f ; ρ2 U 2 S cL
α
7
5
  ;  ρ 2
ρU 2 S c cMζ + e cLζ xθ + aζ S G G
f + Jf 2 U S (c cMα + e cLα )

yS = C xS + D uS ; xS (0) = 0 :

U ∈ [0, 330] m s−1


25
bending
torsion
U→

20

15
Im {s}

←U

10

5
−5 −4 −3 −2 −1 0 1 2
Re {s} →

(a) Pole zero map, root loci

0.06 25

20

0.04

15
ω [rad s−1]
σ [−]

0.02 10

5
0
0
0 100 200 300 0 100 200 300
U [m s−1] → U [m s−1] →

(b) Damping and eigenfrequency

Figure 4: Structural eigenmodes vs. true airspeed


Scrutinizing the parameter dependencies of (8) leads to a significant insight. Obviously, the
eigenvalues of (8) are severely effected by a few uncertain parameters:
 airstream velocity U 2 fU ; U g and squared value U 2, respectively
;
+

 mass m 2 fm ; ; m + g, mainly effected by the varying fuel mass.


Figure 3 outlines the corresponding operating domain. For reasons of simplicity, the latter un-
certain parameter m is initially assumed to be fixed, because it varies quite slowly due to fuel
consumption. This leads to two dominant parameter dependent root locus branches within the
complex plane (figure 4(a)). When the torsion branch crosses Refsg = 0, the allocated air-
speed reads U = Ucrit . The damping of this eigenmode shifts its sign and the system becomes
instable. Simultaneously, the absolute values of the poles tend to converge (figure 4(b)). As
damping decreases rapidly with speed near the critical speed Ucrit, the certification regulations
require a significant margin between Ucrit and the maximum dive speed UD of the aircraft, say
Ucrit  1:15UD.

2.2 Primary flight control actuation system


Exhaustive investigations have been made to improve the performance of the electrohydraulic
actuation system and realization aspects [11, 12, 13]. Based on robust state control and esti-
mation techniques the bandwidth of a current system could be trebled with simultaneous im-
provements in damping characteristic and minimal phase lag. Additionally, the resulting closed
actuator control loop ensures robust properties regarding uncertain system parameters caused
by e.g. fluid temperature. Further details outline [10, 12] et al.. Assuming low–level signal range
operation, a linear third order model represents the PFC actuation system. With the demanded
control surface deflection ζc(t ), the external load F (t ) as input and the actual deflection ζ(t ) or
its derivatives as output, respectively, it leads to a multiple input single output description
2 3
h i Zc(s)
Z (s) = Gc (s) Zc (s) + Gd (s) F (s) = Gc (s) Gd (s) 4 5 : (9)
| {z } F (s)
GA (s)

This characteristic yields a 1–by–2 transfer function matrix GA (s), where Gc(s) describes the
command transfer characteristic and Gd (s) is synonymous to a complex stiffness [19]. Figure
in e r tia (c o n tr o l s u r fa c e ) lo a d s im u la tio n

a c tu a to r (a c tiv e )

s e r v o v a lv e

s h a ft
p r e s s u r e
s u p p ly a c tu a to r (s ta n d b y )

Figure 5: Electrohydraulic PFC actuation system (test rig)


2 −180

magnitude [dB]
0 −135

phase [deg]
−2
measurement −90
−4
simulation −45
−6
−8 0
−10 45
0.1 0.2 0.3 1 2 3 10 13 20
frequency [Hz]

(a) command transfer function

−160 −135
magnitude [dB m/N]

measurement −90
−170

phase [deg]
simulation
−45
−180
0
−190 45
−200 90
0.1 0.2 0.3 1 2 3 10 13 20
frequency [Hz]

(b) disturbance transfer function

Figure 6: Measured PFC actuation system frequency responses


6 shows the Bode plots of both transfer functions comparing analysis and experimental results.
Measurement results are based on a conventional inboard aileron actuation system of the Air-
bus A330/340 installed on a test rig at the TUHH Institute of Aircraft Systems Engineering
(figure 5). The inertia of control surfaces can be imitated by the exchangeable disk mass and
actuator loads are controlled by a dynamic load simulation. Torsion stiffness of the aileron spar
and flexible adjustment of the actuators to the wing box can be simulated by shaft flexibility
and proper adjustment. The test rig equipment comprises a real–time hardware–in–the–loop
simulation software which allows coupling to large–scale simulation models of flexible aircraft
structures. The transfer function matrix of the real system reads approximately
h i
GA (s)  1
a3 s 3 + a2 s 2 + a1 s + a0
b0 f 1 s + f0 : (10)

The coupling of aeroelastic and actuation system can be deduced from (3). Therefore, the actu-
ator load leads to
;  G 
F (t ) =
MH (t )
r?
; SG
f ḧ ; xθ + a ζ S f + J f θ̈ ;
G
(11)

with r? representing the control surface lever.

2.3 Aggregate aeroservoelastic model


Finally, the entire aeroservoelastic model comprises a fourth order wing model, aerodynamics
with a constant feedthrough characteristic and a reduced, third order closed loop actuation
h
L
q
.
h
M w in g .
q
.. s tr u c tu r e ..
z h..
q
U
.
L h
M a e r o -
q
z
M d y n a m ic s
H

a 0

z z
C .
.. c o n tr o lle d z
h. . ..
q a c tu a to r z
M H

Figure 7: Blockdiagramm of the assembled model


system model. The resulting state, input and output vector reads
h iT h iT h iT
x = h θ ζ ḣ θ̇ ζ̇ ζ̈ ; u = ζc α0 ; y = h θ ḣ θ̇ : (12)

Assuming the sensors to be ideal grants the availability of necessary state information: vertical
displacement and rate, torsion angle and rate respectively. Thus, the aeroservoelastic plant can
be summarized in a seventh order linear multi model system

ẋ(t ) = A(p) x(t ) + B(p) u(t )


y(t ) = C(p) x(t ) + D(p) u(t ) ; x(0) = x0 ; (13)
 
where p = [ p1; p2 ] = U ; U 2 denotes the present uncertain parameter vector [1]. The interact-
ing quantities within the model (13) and the interdependencies of the submodels are illustrated
by the block diagram shown in figure 7. Assuming the angle of attack α0 = constant the fre-
quency domain input–output modelling yields a 4–by–1 transfer function matrix
h iT
Y(s) = Gh (p; s) Gθ (p; s) Gḣ(p; s) Gθ̇(p; s) Zc (s) ; (14)
| {z }
G(p;s)

where G(p; s) denotes the ”polynomial plant family” [1]. Analysing the eigenvalues of the ag-
gregate model yields the aeroelastic modes maintaining their dominant character. As the actu-
ation system mainly operates as low–pass filter embedded within the forward path of the entire
aeroservoelastic system and has significant faster eigenvalues compared to the aeroelastic struc-
ture, the objective of the subsequent section focusses on a significant damping augmentation of
the aeroelastic modes: bending and torsion (figure 4).
h
U q (y )
y
a 0
z
q

z c

K (s )

Figure 8: Damping augmentation control loop

3 VIBRATION CONTROL

In order to accomplish the outlined goals, the uncertain plant family requires a parametric
robust control law. Keeping pragmatic realisation aspects in mind, a low order compensator is
preferred, because an increase in order results in additional closed–loop eigenvalues that must
be robustly stabilised, too [1]. The preceding analytical modelling of the aeroservoelastic plant
favours the application of the Parameter Space Design method [1]. Aiming at a minimum order
controller the basic approach consists of a static state feedback (figure 8)
h i
K(s) ;! k= kh kθ kḣ kθ̇ : (15)

3.1 Assumptions
With regard to the aeroelastic eigenmodes (figure 4), the dynamics of the state controlled ac-
tuation system reveal a very high bandwidth fB  12 Hz. Considering the third basic rule of
robust control [1]:
Be a pessimist in analysis, then you can afford to be an optimist in design,
an incisive simplification only concerning the controller design process can be postulated:
h i
GA (s) ;! G̃A := 1 0 : (16)
If any robust stabilising k exists, subsequent examination must focus on the effect of the ac-
tuation systems command and disturbance performance. Moreover it is assumed, that the state
quantities of the simplified aeroelastic plant are completely measurable. This implies no re-
striction, because the corresponding states can be derived from integration of the accelerometer
signals. The principle structure of the damping augmentation feedback control loop displays
figure 8. The input of the actuation system is composed of the basic aileron deflection com-
mand signal ζc and the weighed state feedback ;ky (12), which acts as a compensator relating
to the critical eigenmodes of the wing structure (figure 4).

3.2 Parameter boundaries



The uncertainty domain is comprised by a lower and upper boundary pi; ; pi+ ; i = 1; 2.
Even though the entire operating domain of the flexible aircraft covers U 2 f0; UDg, the
p 2
+

p 2

p 1 U c r it
p 1
+

s ta b le u n s ta b le

Figure 9: Modified operating domain of dependent parameters p


main effect of the feedback controller is desired to augment the damping at the interval
U ; < UD < Ucrit < U + ; U ; > 0. As the open loop system indicates a stable characteristic
for all values U < Ucrit, the lower boundary is raised from U ; = 0 m/s to some higher value,
e.g. U ; = 200 m/s, whereas the upper boundary must cover a safety margin with respect to
the critical speed: U + = 330 m/s. It should be emphasized, that the resulting operating domain
contains stable and unstable representatives (figure 4). The general representation in figure 3
degrades to an unequivocal assignment p2 = f ( p1 ), because of the declaration of p. This results
in an one–dimensional parameter space Q. Due to practical considerations the p–dependency
will be included by a finite number of operating points n distributed between f p1;; p2; g and
f p1+; p2+g along the outlined parameter trace (figure 9). Hence the problem reads: simultane-
ous stabilizing of a finite plant family [1]. The parameter effects on the open loop eigenvalue
characteristic are illustrated in figure 4(b).

3.3 Performance specifications


Generally, the demands on the closed loop time response of a linear time–invariant system are
formulated indirectly by eigenvalue specifications. With regard to the root locus design of a
single–input system pole placement yields a unique solution for state feedback. Unfortuately,
this unique solution refers to a single nominal plant. Assuming an uncertain plant family it
is a reasonable requirement that all eigenvalues are located within a specified region Γ. The
pole region in figure 10(a) is bounded by a circular arc and its damping by a hyperbola that
guarantees damping according to the asymptotes and negative real part [1].

Bearing in mind that the real system also obtains an electrohydraulic actuation system
within the control loop, the requirements are shaped according to the second basic rule of
robust control [1]:

When you close a loop with actuator constraints, leave a slow system slow and
leave a fast system fast.
Moreover, the quantitative description of the desired Γ–region is dominated essentially by the
demand on the augmented damping characteristic. For the analytical modelling a structural
damping factor DS  0:02 in the absence of any airstream is assumed [6]. Hence, a moderate
minimal damping heuristically reads D ; = 0:1 covering the entire operating domain (figure
10(a)). The maximum permissible dynamic of this loop is prescribed by the bandwidth of the
PFC actuation system. In order to follow the statement of separated design, the latter should
be three to four times higher, than the maximum bandwidth of this loop with fmax  3:5 Hz.
Preserving unnecessary high loop gain the negative real part denotes

2π fS = min (jsi j) D ;  0:18 Hz ; n = 10 :


i=1:::n

U = 200 300
20
200
10 U = 200
100

0
0 Γ
κA [−]
Im{s}

−100
−10 −200

−300
−20
−400
−25 −20 −15 −10 −5 0 10 20 30 40
Re{s} → κB [−] →

(a) Pole zero map, open loop (b) Invariance plane, complex (—) and real (;;) mar-
gins

U = 200
20

10 U = 200

0 Γ
Im{s}

−10

−20

−25 −20 −15 −10 −5 0


Re{s} →

(c) Pole zero map, closed loop

Figure 10: Simultaneous Γ stabilisation


Finally, the complete Γ–pole region is outlined in figure 10(a), it is bounded by the edge ∂Γ and
overlapped by the open loop aeroelastic root loci.

3.4 Simultaneous Γ–stabilisation


The core of the Parameter Space Design Method represents the derivation of a controller, which
shifts the entire set of open loop eigenvalues into the desired Γ–region. Assuming any static
state feedback k ((15) and figure 8 respectively) the objective is to determine the set k 2 KΓ for
which the polynomial family in closed loop configuration

G(s; p ( j); k) ; j 2 [1; n] is Γ–stable :

Considering the Boundary Crossing Theorem it suffices to map ∂Γ of each representative p( j)


into the new parameter space, the k–space [1]. The set of robust Γ–stabilising controllers re-
( j)
sult from a superposition of all related maps KΓ . Unfortunately, the Boundary Representation
Theorem accomplishes this mapping only for two controller coefficients ki ; i 2 [1; 4] simultane-
ously. Since a reduction of free controller parameters is not possible, a two–dimensional cross
section plane is selected in the state feedback gain space, where the boundaries are displayed.
This invariance plane approach becomes a sequential procedure, identifying the most critical
eigenvalues, here the torsional motion λ( p1+ ; p2+) and successively shifting two eigenvalues
into the desired region. Decomposing the pre–image of the Γ–region boundary yields four sub-
sets (figure 10(b)): Two dashed straight lines, the real margins, which represent Imf∂Γg = 0
and two solid graphs, the complex margins, which relate to the hyperbola and circle branches
Imf∂Γg 6= 0. Thus, selecting [κB ; κA ] from the intersection of all ∂Γ images effects a shift of
these eigenvalues, while the rest remain fixed [1]. Finally, the simultaneous Γ–stabilising con-
troller is determined indirectly by choosing an operating point from the admissable set within
the invariance plane. The total state feedback gain vector therefore results from the summation
of N iterative design steps
N
(
2 KΓ



n
)

∑ km
( j)
k := ; where km = km Γ ; n = 10 : (17)
m=1

j=1

3.5 Verification
Bearing in mind that the outlined design procedure only represents a necessary condition
requires an adjacent verification of the closed loop performance. Selecting k = k  ((+) shown
in figure 10(b)) as representative candidate yields the closed loop eigenvalues in figure 10(c).
Although the pole zero map reveals that G(s; p ( j); k ) ; j 2 [1; n] meets the requirements, a
nonlinear simulation is essential for verification, including the actuation system characteristics.
Figure 11 shows the system response following a step input α0 . For U = Ucrit the open loop
system executes coupled bending/torsion oscillations as expected. By means of the robust state
feedback controller a well damped time response is achieved.
U = Ucrit U = Ucrit
3 1
bending


torsion 0.5
2 aileron
θ, ζ [ o ], h [10−1 m]

θ, ζ [ o ], h [10−1 m]
deflection 0

1 −0.5
bending
−1
torsion
0
−1.5 aileron
deflection
−1 −2
0 1 2 3 4 5 0 1 2 3 4 5
t [s] → t [s] →

(a) open loop (b) closed loop

Figure 11: Simulated α0 step response


Returning to the preliminary assumption (16) raises the question to what extent the ro-
bust aeroservoelastic control loop characteristic depends on the PFC actuation system
performance. While keeping the outer loop gain vector constant k = k , the controller of the
actuation system depends on the adjusted bandwidth fB 2 [3; 5; 8; 12; 15] Hz. To demonstrate
the effect of actuator bandwidth regulated by a robust state controller and for comparison with
a proportional controller of f B = 3 Hz, figure 12 shows the root loci of the closed loop system.
Although the low–pass characteristic of GA (s) increases with decreasing bandwidth, a robust
state feedback within the inner actuator control loop allows to decrease fB half to one of todays
robust state space controllers for adequate performance. Certainly, the prompter the actuation
system response the smaller is the phase lag of the compensating aileron deflection. Thus,
a high actuation system bandwidth is advanageous by increasing the damping of the torsion

U Î [2 0 0 , 3 3 0 ] m s -1

2 0 f B = 3 H z (P )
f B = 5 H z to r s io n
f B = 8 H z
f = 1 2 H z
1 5 f
B
= 1 5 H z
B
®

U
I m {s }

1 0
b e n d in g

P F C a c t. s y s .
0
-2 0 -1 5 -1 0 -5 0 5
R e {s } ®

Figure 12: Influence of PFC actuation system performance


trajectory with increasing actuator performance. For the proportional controlled actuation
system the actuator bandwidth is similar to the torsional eigenfrequency of the aeroelastic
structure. In this case an aeroelastic feedback controller in unable to stabilize the system for
the whole flight regime considered.

4 VALIDATION

To demonstrate the feasibility of the previously developped damping augmentation control


loop, an implementation of the outlined control law has been performed on the test rig. This in-
cluded the wing structure, sensors and aerodynamics. The PFC actuation system test rig (figure
5) has been upgraded to a real–time system. Two usual personal computers with Intel Pentium
III processors at fT = 450 MHz, which are linked together in target–host configuration, provide
the universal real–time simulation platform based on Matlab/Simulink. The robust state con-
trolled actuation system operates hardware–in–the–loop within a virtual dynamic wing struc-
ture simulation in the presence of quasistationary aerodynamics. A modular architecture of the
simulation model simplifies the transfer to the test set–up, as only the actuator model has to be
replaced by the test rig I/O module. The actuator control law as well as the damping augmenta-

1
to r s io n
0 .5
b e n d in g
0
-0 .5
-1 a ile r o n h [m ]
d e fle c t io n q [d e g ]
-1 .5 z [d e g ]
-2
0 1 2 3 4 5 6
t [s ]

(a) robust state controlled PFC actuation system

0 .5
0
-0 .5
-1 h [m ] b e n d in g
q [d e g ] to r s io n
-1 .5 z [d e g ] a ile r o n d e fle c tio n
-2
0 1 2 3 4 5 6
t [s ]

(b) current PFC actuation system, proportional SISO controller

Figure 13: Measured α0 step response at U  Ucrit


.
tion feedback is realized within the simulation environment. The validation focusses especially
on two aspects: practical functionality, that means confirmation of the simulation results with
physical actuators in the loop. Beyond that, further examination of the interaction of a propor-
tional SISO controlled actuator and the presented damping augmentation control loop should
be done. In order to ensure comparability to the simulation results (figure 11), the measured
time response applies to the initial step in angle of attack at t = 1 s. Figure 13 illustrates the
superior effect of damping augmentation by the robust state controlled actuation system. By
fast actuation system dynamics the compensation signal ζ(t ) effectively operates as an active
oscillation damper, whereas the significant phase lag of a conventional controlled PFC actua-
tion system causes further exitation of the coupled oscillatory motion h; θ. These experimental
results confirm the conclusions already gained from linear analysis (figure 12).

5 CONCLUDING REMARKS

It has been demonstrated that active damping augmentation by primary flight control surfaces
could efficiently effect the weakly damped aeroelastic wing motions. The deduction of the
aeroservoelastic synthesis model from principles of physics led to an analytical plant descrip-
tion. The resulting uncertain seventh order multi–model–system comprised originally stable
and unstable representatives. As physically motivated boundaries could be determined for the
uncertain parameters, Ackermann ’s Parameter Space Design method yielded a set of robust sta-
bilizing state feedback vectors. A significant damping augmentation was verified by simulation
with the actuation system hardware–in–the–loop. Practical investigations revealed the consider-
able influence of the actuation systems performance. Especially, the common SISO controller
of current electrohydraulic actuation systems caused an unsufficient phase lag, which led to
instability of the entire system. The aeroelastic compensation feedback grants a fast rejection
of exogeneous perturbations in connection with a double bandwidth than applied in actual
actuation systems. Finally, it has been demonstrated that the implemented controller design
method was capable of shaping a suitable actuator controller as well as an aeroelastic feedback
which simultaneously provides damping augmentation throughout the entire uncertain parame-
ter set. The ongoing investigations focus on elaborating enhanced models of current flexible
wing structures and examinating optimal sensor locations. Future work will also aime at ap-
plying suitable order reduction methods to preserve a state controller which mainly effects the
critical modes of the aircraft and can be transferred to an equivalent output controller.

ACKNOWLEDGEMENT

The author thanks DaimlerChrysler Aerospace Airbus GmbH for promoting and supporting the
project Aktuatorregelung in aeroelastischer Umgebung.
NOTATIONS AND ABBREVIATIONS

Notations
α [ ] dynamic angle of attack
α0 [ ] initial angle of attack
aζ [m] distance E G
A [;] system matrix — state space representation
B [;] input matrix — state space representation
B [;] damping matrix
c [m] distance from leading to trailing edge, hinge line to trailing edge
cL ; cM ; cH [ ;] aerodynamic coefficients
C [ ;] output matrix — state space representation
e [m] distance A E
D [;] damping coefficient
D [;] feedthrough matrix — state space representation
∂Γ [;] stability boundary
f [Hz] frequency
Γ [;] stability region
G [;] SISO transfer function
G [;] transfer function matrix
θ [ ] leading edge up angular rotation
h [ ] ; downward vertical displacement
J [kg m 2 ] moment of inertia
κA;B [ ;] basis of the invariance plane
k [kg/s 2 ] mechanical stiffness
k [ ;] vector of feedback gains
K [ ;] stiffness matrix
L [N] aerodynamic lift
M [N m] aerodynamic moment
MH [N m] hinge moment
M [ ;] mass matrix
m [m] total mass
p ;
[ ] uncertain parameter representative
p [;] vector of uncertain parameters
q [;] vector of generalized coordinates (2DOF)
q̃ [;] vector of generalized coordinates (3DOF)
Q [;] vector of unconservative forces
ρ [kg/m 3 ] air density
r? [m] equivalent control lever
s [rad/s] complex frequency s = σ  jω
S [kg m] static unbalance
S [m 2 ] equivalent wing or aileron aera (vertical projection)
T [Nm] kinetic energy
U [m/s] equivalent airspeed
UD [m/s] maximum dive speed
u [ ;] input vector
V [Nm] potential energy of strain
xθ [m] distance E C
x [ ] ; state vector
y [ ] ; output vector
ζ [ ] trailing edge downward aileron deflection
Z [ ;] Laplacian of ζ
Indices
(Note: The listed symbols may be used as lower as well as upper index.)
; [;] lower boundary
+ [;] upper boundary
α [;] angle of attack
A [;] aerodynamic
B [;] bandwidth
c [;] command
crit [;] critical
d [;] disturbance
f [;] flap, aileron
θ [;] angular rotation
h [;] vertical displacement
max [;] maximum
min [;] minimum
P [;] reference point P 2 fA; E ; C; Gg
S [;] structural
ζ [;] aileron deflection

Abbreviations
A [;] aerodynamic center
C [;] center of gravity
E [;] elastic axis
G [;] hinge line
DOF [;] degrees of freedom
I/O [;] input–output
PFC [;] primary flight control
SISO [;] single input single output
TUHH [;] Technical University of Hamburg–Harburg

REFERENCES
[1] Ackermann, J. et. al.: Robust Control — Systems with Uncertain Physical Parameters. London: Springer, 1993.
[2] Bisplinghoff, R.L.; Ashley, H.; Halfman, R.L.: Aeroelasticity. Mineola: Dover Publications, 1996.
[3] Bossche, D.; El Sherif, F. et. al.: Airbus A330/340 Primary Flight Control System — Aileron Actuation. SAE Committee A6. At-
lanta/USA, 1991.
[4] Dowell, E.H.(Ed.): A Modern Course in Aeroelasticity. 3 rd edition. Dordrecht: Kluwer, 1995.
[5] Foersching, H.W.: Einfluss servomechanischer Steuerungs– und Stabilitaetssysteme auf das Flatterverhalten von Flugzeugen. Z. Flug-
wiss. 21 (1973), S.22–31, 1973.
[6] Foersching, H.W.: Grundlagen der Aeroelastik. Berlin: Springer, 1974.
[7] Gelb, A.; Vander Velde, W.E.: Multiple–Input Describing Functions and Nonlinear system Design. New York: McGraw Hill, 1968.
[8] Green, W.L.: Aircraft Hydraulic Systems. Chichester: J. Wiley & Sons, 1985.
[9] Guillon, M.: Hydraulic Servo systems — Analysis and Design. London: Butterworths, 1968.
[10] Kliffken, M.G.; Kruse, U.: Robust Control of Electro Hydraulic Actuators in Primary Flight Controls. at—Automatisierungstechnik
11/45, 1997.
[11] Kiffken, M.G.: Nichtlineare strukturelle Regelung, angewandt auf Stellsysteme der Flugsteuerung. PhD thesis, Technical University of
Hamburg–Harburg. Fortschritt–Rep. No.737, Series 8. Duesseldorf: VDI-Verlag, 1999.
[12] Kliffken, M.G.; Gojny, M.H.: A Unified Control Strategy for Flight Control Actuators. Recent Advances in Aerospace Hydraulics.
Toulouse/France, 1998.
[13] Kliffken, M.G.; Gojny, M.H.: Numerical Robust Implementation of Sampled–Data Controllers for Flight Control Actuators. at—
Automatisierungstechnik 9/47, 1999.
[14] MacKinnon, A; Stollery, J.L..: Variable Camber Wing. COA Rep.9304, Cranfield University, 1993.
[15] Lind, R; Brenner M.: Robust Aeroservoelastic Stability Analysis. London: Springer, 1999.
[16] Middleton, R.H.; Goodwin G.C.: Digital Control and Estimation – A Unified Approach. Englewood Cliffs: Prentice–Hall, 1990.
[17] N.N.: Donnees Aerodynamiques — Aileron Hinge Moments. Aerospatiale, Toulouse, 1991. Note Aerodynamique 445.343.
[18] Raymond, E.T.; Chenoweth, C.C.: Aircraft Flight Control Actuation System Design. Warrendale: SAE, 1993.
[19] Vaughan, N.D.; Raval, R.: The Use of an Impedance Analysis of an Hydraulic Servo to Investigate Control Compensation for Flutter
Suppression. Recent Advances in Aerospace Hydraulics. Toulouse/France, 1998.

You might also like