N A S A TECHNICAL NOTE

FLUTTER SUPPRESSION USING

ACTIVE CONTROLS BASED ON THE
CONCEPT OF AERODYNAMIC ENERGY

by E. Nissim
Langley Research Center
Hampton, Va. 23365 ;' '
% .

N A T I O N A L AERONAUTICS A N D SPACE A D M I N I S T R A T I O N W A S H I N G T O N , D. C. MARCH 1971

it I
r
i
I' TECH LIBRARY KAFB, NM

~­ 01133267 ­
1-
~

1. Report No.
NASA TN D-6199
4. Title and Subtitle
I 2. Government Accession No. 3. Recipient's Catalog No.

5. Report Date

FLUTTER SUPPRESSION USING ACTIVE CONTROLS BASED March 1971

6. Performing Organization Code
ON THE CONCEPT O F AERODYNAMIC ENERGY
7. Author(s) 8. Performing Organization Report No.
L-7525
E. Nissim
10. Work Unit No.
9. Performing Organization Name and Address 126-14-14-02
NASA Langley R e s e a r c h Center 11. Contract or Grant No.

Hampton, Va. 23365

13. Type of Report and Period Covered
112. Sponsoring Agency Name and Address
Technical Note
National Aeronautics and Space Administration 14. Sponsoring Agency Code
Washington, D.C. 20546
15. Supplementary Notes
T h i s r e s e a r c h was accomplished while the author held a National R e s e a r c h Council
Postdoctoral Resident R e s e a r c h Associateship at NASA Langley R e s e a r c h Center. Author is
now Associate P r o f e s s o r , Technion, Israel Institute of Technology, Haifa, Israel.
~

__ - .
16. Abstract

T h e problem of flutter suppression is treated from a n energy point of view, whereby

the energy dissipated p e r cycle of oscillations is reduced to a quadratic form involving a
diagonal m a t r i x of energy eigenvalues which are of aerodynamic origin only. A simplified
binary bending-torsion flutter of a wing s t r i p is investigated by s e v e r a l systems: leading-
edge (L.E.) control s u r f a c e , trailing-edge (T.E.) control surface, and combined leading­
edge-trailing-edge (L.E.-T.E.) control surfaces. I n each case the control s u r f a c e s are
allowed to be driven by a linear s e n s o r , a rotational s e n s o r , and a combined linear-
rotational s e n s o r system. T h e r e s u l t s show that the flutter of the wing s t r i p can be sup­
pressed, a single control law being used, over a wide range of reduced frequencies, irre­
spective of the m a s s of the s y s t e m , its stiffness, center-of-gravity location, elastic axis,
the mode of vibration, and the Mach number (in the subsonic range). P a r a m e t e r s affecting
the implementation of the control law, such as power requirements, phase lags, and ampli­
tude gains, are investigated. T h e extension of the r e s u l t s to three-dimensional flow cases
is discussed and the u s e of the flutter s u p p r e s s o r as a gust alleviator is a l s o considered.

18. Distribution Statement

Unclassified - Unlimited

. - .
19. Security Uanif. (of this report) 20. Security Classif. (of this page) 21. NO. of Pages 22. Price'
Unclassified Unclassified 114 $3.00

For sale by the National Technical Information Service, Springfield, Virginia 22151
. ~ . __ .
.

CONTENTS
Page
SUMMARY ....................................... 1
INTRODUCTION .................................... 1
SYMBOLS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
THE ENERGY APPROACH TO FLUTTER SUPPRESSION . . . . . . . . . . . . . . 8
Basic Requirements for a Flutter-Suppression System .............. 8
The Energy Analysis . . . . . . . . . . . . . . . . . . .
............. 9
...............
Control System Law for the Simplified Flutter Model 11
......................
Analysis of the Simplified Flutter Model 13
...............................
Optimization Procedure 14
...........................
Data and Scope of Optimization 16
OPTIMIZATION AT ZERO MACH NUMBER . . . . . . . . . . . . . . . . . . . . . 17
Ranges of Optimization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
Results for Trailing-Edge Control Only . . . . . . . . . . . . . . . . . . . . . . 18
Results for Leading- and Trailing-Edge Control Surfaces . . . . . . . . . . . . . 20
Leading-Edge Control Surface Only . . . . . . . . . . . . . . . . . . . . . . . . 21
Subsequent Considerations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
Remarks on the Comparison Between the Two-Controls System and the
Single-Control System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
POWER REQUIREMENT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
Preliminary Considerations . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
Data for Power Estimate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
Effect of Power Requirement . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
Reoptimization and Reevaluation of the Power Requirement With
Unimportant Control P a r a m e t e r s Set to Zero . . . . . . . . . . . . . . . . . . 26
COMPRESSIBILITY EFFECTS ............................ 26
SOME PRACTICAL ASPECTS OF THE OPTIMIZED CONTROL LAWS . . . . . . . 27
Preliminary Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
Amplitude Effects.................................. 28
Sensitivity of Optimized System to the Determination of the Frequency ..... 28
.........
Some Further Consideration of the Control Law G P a r a m e t e r s 29
GENERAL DISCUSSION AND RESULTS . . . . . . . . . . . . . . . . . . . . . . . 31
The C r o s s Effects Between Active and Nonactive Strips on a Wing . . . . . . . . 31
Suppression of Flutter in Three-Dimensional Wings . . . . . . . . . . . . . . . 33
Suppression of Rigid-Body Modes . . . . . . . . . . . . . . . . . . . . . . . . . 35

iii

..
 Page

..............................
Gust-Response Problems 37
.........................
Effect of Chordwise Deformations 37
V-g Plots of a Two-Dimensional Flutter System ................. 39
Some Further Use of the Energy Approach .................... 40
Determination of the Energy Absorbing Modes .................. 40
................................
Mechanism of Flutter 41
CONCLUDING REMARKS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
APPENDIX A .THE ENERGY ANALYSIS . . . . . . . . . . . . . . . . . . . . . . 44
APPENDIX B . DERIVATION OF THE GENERALIZED AERODYNAMIC
FORCES OF A L.E.-T.E. SYSTEM ......................... 48
APPENDIX C - ESTIMATION O F THE POWER REQUIRED TO ACTIVATE THE
CONTROL SURFACES ............................... 53
APPENDIX D .TRANSFORMATION O F THE FLUTTER EQUATIONS USING
THE ENERGY MODE SHAPES MATRIX ...................... 56
REFERENCES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58

FIGURES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60

iv
 FLUTTER SUPPRESSION USING ACTIVE CONTROLS BASED ON
THE CONCEPT OF AERODYNAMIC ENERGY

By E. Nissim"
Langley Research Center

SUMMARY

The problem of flutter suppression is treated from an energy point of view, whereby
the energy dissipated by the system per cycle is reduced to a quadratic form involving a
diagonal matrix of energy eigenvalues which a r e of aerodynamic origin only. A simplified
binary bending-torsion flutter of a wing s t r i p is investigated by s e v e r a l systems: leading-
edge (L.E.) control surface, trailing-edge (T.E.) control surface, and combined leading­
edge-trailing-edge (L.E. -T.E.) control surfaces. In each case the control surfaces a r e
allowed to be driven by a linear s e n s o r , a rotational s e n s o r , and a combined linear-
rotational s e n s o r system. The results show that the flutter of the wing s t r i p can be sup­
pressed, a single control law being used, over a wide range of reduced frequencies i r r e ­
spective of the m a s s of the system, its stiffness, center-of-gravity location, elastic axis,
the mode of vibration, and the Mach number (in the subsonic range). The leading-edge­
trailing-edge control system driven by linear-rotational s e n s o r s system is shown to be
the most efficient system. Estimates are made for the power required to drive the pro­
posed leading-edge-trailing-edge system and the p a r a m e t e r s affecting the implementa­
tion of the control law (such as phase lags and amplitude gains) a r e investigated. A
broad discussion is also included regarding application of the flutter-suppressing system
to three-dimensional wings, the p a r a m e t e r s affecting the reduction of the number of con­
trol s y s t e m s required to suppress flutter, and the use of the flutter s u p p r e s s o r s as gust
alleviators. The analytical developments a r e given i n s e v e r a l appendixes and the results
obtained a r e illustrated in figures and charts.

INTRODUCTION

The recent technological advances made in the field of control s y s t e m s and the
increased reliability of control system components offer a new way of treating the prob­
lem of flutter instability. T h i s approach consists of a rapidly responding control system
__ - - ~ __ ~

* NRC-NASA Resident Research Associate, now Associate P r o f e s s o r , Technion,

I s r a e l Institute of Technology, Haifa, Israel.
which is actuated by the motion of the main surface and which leads to an appropriate
deflection of the control surface. In this way, u s e is made of control-surface aerody­
namic forces to combat flutter instability. Such a system is r e f e r r e d to as an active
control system.
Attempts to u s e active controls were made i n recent y e a r s mainly i n conjunction
with problems associated with gust alleviation (the objective being to reduce the rigid-
body response to turbulence) and with mode stabilization (the objective being to reduce
structural dynamic response to gusts).
The study, design, and test of gust-alleviation systems have been explored over a
number of decades and although they have been shown t o be effective and technically
feasible (refs. 1 to 4) they have lagged i n practical application. Mode stabilization, to
reduce structural response, is akin to the problem of flutter suppression. It has been
seriously considered only during the past decade, and has culminated with the installa­
tion of active systems on the B-52 aircraft to control the response of the rigid-body mode
and one elastic mode (first aft body bending) to gust inputs. (See refs. 5 and 6.) No
attempt will be made to review the extensive literature in the field of gust alleviation
with mode stabilization (for excellent reviews, see refs. 7 and 8) and mention will only
be made of some of the contributions and results which have direct bearing on flutter
suppression.
Flutter suppression implies, essentially, the control of structural modes only.
Therefore, mode stabilization methods are of prime importance in flutter suppression.
Associated problems can be classified within a few categories as follows:
(a) The determination of the contribution of the different structural modes to a
given a r b i t r a r y deformation of the aircraft
(b) The determination of a control law which will actuate the control surfaces in a
certain fashion, proportional to the participation of the structural mode in
the given deformation, and give r i s e to a stable control system. An unstable
control system can arise, for example, when a signal sensed from one struc­
tural mode gives rise to a control force which actuates a second structural
mode and which i n turn gives r i s e to a force which drives the f i r s t structural
mode. This condition is known as control system induced instability and is
caused by the sensor-force couplings.
(c) The determination that the control forces arising from the determined control
l a w a r e of the type to stabilize the motion sensed
(d) The determination that the stabilized system is insensitive to changes in flight
configuration, altitude , Mach number and mode shapes

2
 Different approaches and methods were proposed and studied for the solution of
the problems mentioned in (a), (b), and (c). These methods include multisensor systems
for modal observation, force-sensor modal decoupling networks (very sensitive to
changes in flight configuration), or mode discriminating s e n s o r s to avoid control system
induced instability, and aerodynamic damping forces for mode stabilization. Linear
optimal control theory has recently been used (ref. 9) to provide a systematic way of
treating the problem of control system instability and to increase the control-force sta­
bilization effectiveness. However, the control law which is derived by this method, and
which depends on the location of the sensors, cannot often be attained by real systems.
Furthermore, it requires as many control surfaces as modes considered. Recently,
Wykes (ref. 10) suggested the identically located accelerometer and force (ILAF) scheme
to insure control system stability and made use of control "aerodynamic damping" forces
(that is, control deflection proportional to linear velocity) for modal stabilization.
These procedures permit some measure of optimization of a specific system at a
specific flight condition (that is, flight configuration, altitude, and Mach number) and for
specifically defined mode shapes. However, the problem common to all mode stabiliza­
tion control systems is a tendency to be very sensitive to changing flight conditions.
Thus, an optimized control system at one flight condition may either show degradation o r
even give rise to adverse effects at another flight condition. (See ref. 10.) Furthermore,
even the optimization procedures used for a specific flight condition may often lead to
control systems which are difficult to realize in practice. Wykes (ref. 10) has used only
aerodynamic "damping" forces with the ILAF concept and in s o doing he imposed a
s e v e r e limitation on the optimization technique. It should also be remembered that
extensive literature exists (refs. 11 to 13) showing that addition of positive damping can
turn a stable elastic system into an unstable one. It is therefore concluded that the
approaches to mode stabilization do not appear to offer an appropriate basis for flutter-
suppression investigations and that a new approach is desirable.
Mention should be made here, prior to any elaboration on a new approach, of two
experimental works c a r r i e d out with the objective of increasing the flutter speed by the
use of active control systems. (See refs. 14 and 15.) It is not believed, however, that
these experiments were methodically planned, and they therefore resulted i n very minor
changes in flutter speed that were often accompanied by poor controllability. It s e e m s
that extensive analytical work which will lead to deeper understanding of the basic param­
e t e r s affecting the suppression of flutter by active controls is required before new exper­
iments are made.
The approach taken in this paper is based on aerodynamic energy considerations
and is formulated in a manner insuring lack of sensitivity to changing flight conditions.
Conceptual studies of some basic control systems are made and their relative m e r i t

3
determined. Figures are furnished partly to illustrate the results and partly to supply
information for general preliminary design considerations.

SYMBOLS

b semichord length

Cij
element i,j, of control law matrix [c]

ck s c a l a r multiplier of [G] matrix

d representative oscillatory amplitude

fpfs generalized f o r c e s i n 0 and 6 coordinates and required to establish a

desired control law

Gij
element i , j , of the control law matrix G
[I
h,E bending displacement, positive in down direction

hl,h2 bending amplitudes defined by sketch (0)

hS h at x = x s

k reduced frequency, ob
-
v

distance defined i n sketch (B2)

M Mach number

m l = npb4 s

n number of degrees of freedom of elastic system

-
P energy dissipated by system per cycle

4
- power required to drive control surfaces
P1

generalized aerodynamic force along h, CY,p , and 6, respectively

smax maximum s t r e s s at a specified location

S reference length or wing semispan

t time

V flight speed
7

virtual work along coordinates h, CY,p, and 6, respectively

X downstream coordinate measured from midchord point

XS value of x at which the sensor is located divided by b

--
Y ,= leading-edge displacements of the control surfaces defined in equations (B9)
and (B10)

CY, z oscillatory angle of attack of wing, o r oscillatory amplitude, positive in

nose-up direction

% value of CY at x = xs

PY6 leading-edge and trailing-edge control surface deflections , respectively,

positive directions defined i n sketch (Bl)

x eigenvalue of [u]
xi ith eigenvalue of [U]

P fluid density

W oscillatory frequency

5
Matrices:

real and imaginary part, respectively, of aerodynamic matrix A

[AR],[AI]
inertia matrix

[B11’[B2]j
PI
,PI
[B3 submatrices of defined by equations (C2)

PI control law matrix

[Co] ,[C 11,[C2] control law matrices defined by equation (9)

E structural stiffness matrix

structural stiffness submatrices defined by equations (C2)

GI column matrix of forces

force matrix defined by equation (C4)

[Fll
LG1 control law matrix

[Go] ,[%I @2] control law matrices defined by equation (10)

complex aerodynamic matrix defined by equation (B 16)

r i t - i I

[I
submatrices of H defined by equations (C2)

matrices defined in equations (D9)

r e a l and imaginary part, respectively, of energy eigenvector modal matrix

[Q~],[Q]
@I complex response vector

6
 value of {q} at flutter speed
C9F1
Cg3 complex amplitudes of response vector

{qR} ,cqI) real and imaginary part of

c 01
q

aerodynamic energy matrix defined by equation (A9)

[UI

PJ diagonal matrix of eigenvalues of [u]

real and imaginary parts, respectively, of generalized energy coordinates
((R)

@I eigenvector of [u]
Other notations :

11 absolute value

T
transposed matrix

* complex conjugate

li row matrix

0 column matrix

Subscripts:

min minimum

max maximum

opt optimum

r reference

Dots over symbols denote derivatives with respect to time.

I
 . ... .
..__

THE ENERGY APPROACH TO FLUTTER SUPPRESSION

Basic Requirements for a Flutter-Suppression System

In determining a new approach to the problem of flutter suppression, the following
goals were set:
(1)To provide some analytical means which will insure, during the process of
optimization, minimum sensitivity of control system effectiveness to flight
conditions
(2) To provide a simple analytical model which will yield some insight into the
mechanism of flutter suppression by active controls and still retain the
essence of the problem
The simplest possible flutter model is the bending-torsion flutter of a rigid two-
dimensional strip. The idea of using such a model, although very attractive, s e e m s to
present some inherent difficulties. F o r example, what should the inertia matrix o r the
elastic matrix values b e ? What values should be assigned to p a r a m e t e r s such as m a s s
ratio o r elastic-axis location and how could these results be correlated, even within some
approximation such as aerodynamic s t r i p theory, to a three-dimensional wing ? These
points should be carefully considered before discarding such a simple model. The natural
question which follows is how do the values of inertia o r elastic t e r m s , for example, affect
the flutter instability ? Clearly, these t e r m s affect the natural frequency and the natural
mode shapes of the oscillating s t r i p and the variation of these p a r a m e t e r s over a wide
range of values will yield, for each frequency of oscillation, a wide range of mode shapes
which represent, in essence, different combinations of bending and torsional displace­
ments. Since the frequency is itself a parameter, this statement implies infinite com­
binations of frequency and mode shapes, It may therefore be possible to accomplish this
result by a direct variation rather than indirectly (through inertia o r elastic t e r m s ) .
Indeed, since a wing is a continuous system, one isolated s t r i p on the wing can vibrate at
many frequencies and many mode shapes and therefore, this variation of frequencies and
mode-shape combinations is also essential for the extension of the results to a three-
dimensional wing. Similarly, the sensitivity to changes i n flight conditions reduces in
essence to the effect of changing the mode-shape-frequency combinations. Hence, the
requirement that the binary system be insensitive to flight conditions implies that the
increase i n stability of the binary system be independent of the mode shapes over a wide
range of frequencies. T h e r e remains now to determine an appropriate approach which
will allow an easy variation of frequencies and mode shapes and indicate their effect on
the stability of the binary flutter system. At this stage, an energy approach s e e m s to be
the natural outgrowth of these requirements. T h i s is t r u e since one can expect that the
work done per cycle by the system will depend on the aerodynamic forces (which a r e

8
functions of the reduced frequency for any given Mach number) and mode shapes only,
whereas the magnitude and sign of the work done per cycle indicate the state of stability
of the system.

The Energy Analysis

In this section the energy concept is developed (the details are presented i n appen­
dix A) without imposing any restrictions on the s i z e of the system and the nature of the
aerodynamic forces. The r e s u l t s obtained apply to three-dimensional wings as well as
to the two-dimensional model described.
Let

{F} = -w2[B + npb4S(A~+ iA~)]{q} + [E] {q}

where, at flutter,

{F) =O

and w represents the frequency of oscillation; [B3 , the m a s s matrix; [AR] and [AI], the

real and imaginary parts of the aerodynamic matrix, respectively; [E], the stiffness
matrix; p, the density of the fluid; s, a reference length; b y a reference semichord
length; and {q}, the response vector.
As shown i n appendix A, the work done by the system on its surrounding per
cycle can be written as (eq. (A7))

where (from eq. (A2))

The sign of is of importance i n the determination of stability, and therefore, it

will be advantageous to convert equation (2) to a more convenient form. As shown i n
appendix A (eq. (A15)), i? can be reduced to the form

9
or

where [A] is a diagonal matrix of the eigenvalues X i , necessarily real, of the Hermitian
matrix (eq. (A9))

[U] = [-(AI + AT) + i(AR - Ai]

and where the vectors (SR> and {tI} are defined by the transformation (eq. (A10))

The matrix QR
[I l
+ i&r is a square modal matrix of the principal eigenvectors.
tion (5) contains some very interesting features:
Equa­

(1)The work done per cycle has been brought to a principal quadratic form
(which involves a diagonal matrix) in t e r m s of the system responses. Hence, the sign of
-
P will be independent of the values of the responses provided it is possible to render all
the X t e r m s positive.
(2) The vaiues of X a r e determined from the aerodynamic matrices only. They
depend on the assumed modes but a r e independent of the response of the system. The
values of X will, however, depend on the reduced frequency of oscillation and Mach
number. In this way, complete decoupling has been achieved between the aerodynamic
inputs and the response of the system.
It can now be seen that a necessary and sufficient condition for the suppression of

i
flutter for arbitrary values of responses (R and (1 is that all the X t e r m s be posi­
)
tive. This requirement which stresses the arbitrary nature of the response of the system
is a s e v e r e one in most cases. This statement is t r u e since the values of the ( t e r m s
-
of a particular system (in a specified flight configuration) may be such as to render P
i n equation (5) positive despite the existence of some negative values of A. It is essen­
tial, however, to introduce this requirement to insure the dissipation of energy under any
conceivable flight condition. The objective of the following work will be to find a relation
between the main surface oscillatory displacements t o the control surface deflections to
insure that the smallest value of X (that is, Xmin) of the simplified flutter model
becomes and remains positive and assumes adequately large values over a wide range of
reduced frequencies and Mach numbers.

10
 It is interesting to note here that several investigators have, in the past, used an
energy approach to treat various aeroelastic problems (refs. 13, 16, and 17) while they
restricted the energy expressions to real variables only. It now appears that the use of
complex notation throughout the analysis, as shown in the appendix, is advantageous in
that it leads to a diagonalized quadratic expression.

Control System Law for the Simplified Flutter Model

The two-dimensional rigid s t r i p shown in sketch (a) has two degrees of freedom,
translation (bending) and rotation (torsion). Therefore, two control surfaces a r e required
i n order to maintain precise control of the two degrees of freedom of the main surface.
The main surface is thus allowed to have both leading-edge (L.E.) and trailing-edge (T.E.)
control surfaces (the choice of a leading-edge-trailing-edge control system is further

Sketch (a)
discussed in a subsequent section of this paper). It is possible to see, at this stage, that
eight parameters (four �or each control surface) exist which permit a general deflection
of the control surfaces with respect to the motion of the main surface. These deflections
are:
(1) Deflection of the T.E. control surface actuated by the in-phase displacement of
the main surface
(2) Deflection of the T.E. control surface actuated by the out-of-phase displacement
of the main surface
(3) Deflection of the T.E. control surface actuated by the in-phase twist of the main
surf ace
(4) Deflection of the T.E. control surface actuated by the out-of-phase twist of the
main surface
The additional four parameters similarly relate to the actuation of the L.E. control
surface by the in-phase and out-of-phase (90° phase lead) movements of the main surface.
This relation can be put in matrix form by using the notation illustrated in sketch (b), that
is,
11
 -
G12

G22
-
Equation (8) represents a control law, where the C and G t e r m s are essentially the
gearing ratios between the motion of the main surface and the deflection of the control
surfaces.

__- -- - I- - - - -
U n d i s t u r b e d p o s i t ion
- - -- - ­

1
Sketch (b)
No attempt is made at this stage to consider the practical feasibility of mechanizing the
control law. Such considerations are left f o r a later stage of this work, after gaining
some insight into the relative importance of the different control parameters.
It is possible to increase the number of parameters by letting the C and G
t e r m s in equation (8) be functions of the reduced frequency. An expansion of the C and
G values in t e r m s of reduced frequency will be of the form

[C] = [CO] + k[C1] + k2[C2] + . . . (9)

[GI = [Go] + $[GI] + k2[G2] + . . .

It should, however, be noted that the aerodynamic coefficients of both the main surface
and the control surfaces are quadratic i n k" and the coefficients vary only slowly with
the reduced frequency k. Therefore, a quadratic representation of C and G should
be sufficient to represent all the coefficients of the main surface for all practical pur­
poses. Hence, equation (8) can be looked upon as a special form of equations (9) and (10)
and can be expected to yield good results at low reduced frequencies. The matrices [Ca]
and [Gz] will affect the aerodynamic derivatives of the main surface mainly through the
control surface aerodynamic inertia t e r m s , and these values are known to be very small

12
I

i n wing flutter at the normal range of frequencies. It can therefore be concluded that
although equation (8) will be used throughout the present work, one might consider the
u s e of the following equations i f a high range of reduced frequencies needs to be
considered:

[GI = [Go] + "[q

In the following discussion, it will be assumed that [C] = [CO] and [GI = [GO],

Analysis of the Simplified Flutter Model

The first step in the analysis is to determine the aerodynamic matrices which lead
to the formation of the aerodynamic energy matrix [U] defined by equation (6). The treat­
ment of the L.E. control surface presents no special problems since, effectively, it is
equivalent to a deflection of a large T.E. control surface. A detailed derivation of the
two-dimensional generalized aerodynamic forces per unit span is presented in appendix B.
The aerodynamic forces per unit span have the form (eq. (B17))

[I
where H represents the matrix of aerodynamic coefficients, Q represents generalized
forces, and the subscript refers to the degree of freedom upon which these for c e s act.
Substitution of the control law defined by equation (8) into equation (13) yields

where

[HI e

13
I

Hence HI, Ha, HQ, and H4 are 2 X 2 submatrices of [HI. The generalized forces
Q p and Q6 need to be known only when evaluating the power required by the control
system to establish the control law defined by equation (8). The bending-torsion aero­
dynamic forces can therefore be written as

where

[AR
IC
+ iAI = H1 + H2C + iHzG]

1 [ I
Equation (16) shows how [C] and [G] affect the value of the m a t r i c e s AR and AI
[ .
Finally, substitution of the values of
[ I [I
AR and AI into equation (6) yields the required
matrix [U] which, in turn, yields the energy eigenvalues.
Mention should be made h e r e that the derivation of [U] for the simplified model
implies that the control surfaces represent irreversible systems, that is, they can be
deflected by the control system only and are not affected by other forces.
As a first step, [U] was calculated herein by using aerodynamic derivatives perti­
nent to Mach number M = 0, derived from potential theory solutions of Theodorsen,
Garrick, KGssner, and Schwartz, and summarized in references 18 to 21. In a later
stage, Mach number effects were considered i n calculating [U] by solving, numerically,
Possio's integral equation (ref. 22) with the two-dimensional compressible kernel. An
extensive comparison with the numerical results appearing in references 23 and 24 has
shown that, in all cases, the accuracy of the numerical results was within 3 percent.

Optimization Procedure
The eigenvalue solution of the simplified flutter model yields two values of A, the
s m a l l e r of which is designated by Amin. Imagine now that Amin is plotted against k'l
for different values of a particular control parameter, for example, Cij. The optimiza­
tion procedure was initially programed to determine the smallest value of Amin a t each
value of a specific Cij (or Gij) as shown in sketch (c). Comparison was then made
between points such as 1, 2, and 3 and the optimized value of Cij was defined as that
value which produced the largest minimum value of Xmin (that is, Cij = 1-3 in
sketch (e)). The deficiencies inherent in this definition of optimum values made it
impossible to implement in practice.

14
 M = Constant

Xmin

Ilk
Sketch (c)

Sketch (d) illustrates how one may be misled to choose the value of Cij = 1-2 as opti­
mum, on the basis of this definition, whereas the overall performance of Cij = r1 is
superior.

Sketch (d)

The optimization with respect t o the control parameters w a s therefore redefined to

yield the values of the control parameters which give r i s e to maximum area under the
curve of hmin against k-1. This definition has possible theoretical deficiencies, such
as shown in sketch ( e ) . Practice has shown that this definition is workable and that,
unlike the curve i n the sketch, the curves are well behaved. The optimization procedure

1/k
Sketch (e)

15
consists of the variation of a single control parameter at a time, all the other control
p a r a m e t e r s being kept constant. The value of the control parameter which yields the
largest area under the curve of X against k'l is then assigned to the parameter and
another control parameter varied i n a s i m i l a r fashion. This whole procedure is repeated
until convergence of the values of the control p a r a m e t e r s is reached. Experience has
shown that convergence is reached after 3 to 4 parameter sweeps. It is, however, advis­
able to accompany each optimization with a plot of the optimized hmin against k'l
curve, together with sensitivity curves which indicate the effect of the off-design values
of [C] and [q.
Data and Scope of Optimization
The binary flutter model was allowed 20 percent chord L.E. and T.E. control s u r ­
faces. No variation was allowed i n these values and they were kept constant throughout
the investigation. Similarly, a reference point for sensing the motion of the main s u r ­
face was kept constant at 30 percent of the chord (measured from the leading edge). This
point (x = -0.4b) simulates the location of the s e n s o r s on the wing. If it should be desired
to relate the results to another sensor location (x = xsb), then the following transforma­
tion may be used to modify the optimized values of

a! = CY,

h = hs - ~ X +S 0.4)b
or, i n matrix form

The optimum control law will then assume t..e form:

or

16
where

k+iG3,;6.idopt~-xS - 0.4 ]
The flutter model w a s optimized over a wide range of reduced frequencies
(0.0128 S k S 19.5).
The numerical work which follows was c a r r i e d out i n four stages:
(1)Stage 1, optimization at M = 0: Optimization of Xmin with respect to [C] and
[GI at M = 0. Comparative studies of optimum conditions were made for T.E. control
system, L.E. and T.E. controls, and L.E. control only (optimization c a r r i e d out, in all
cases, for both single-sensor control law and a double-sensor control law).
(2)Stage 2, power requirement: Rough assessment of the power required to drive
the optimized control system and its sensitivity to the control parameters.
(3) Stage 3 , compressibility effects: Mach number sensitivity of Amin, [ClOpt
r i

(4) Stage 4, some practical aspects of the optimized control laws: The practical
achievement of the control law from both a block-diagram point of view and performance
point of view, together with off-design sensitivities.
The work w a s organized to determine the possibilities offered by the analysis and
to look for any deficiencies which might make the results impractical. Allowance was
also made for possible interactions between the different stages requiring some com­
promising results. Where interactions were light, no reoptimization followed since it
was felt that slight refinements should be left to a later stage.

OPTIMlZATION AT ZERO MACH NUMBER

Ranges of Optimization
The variations of the control p a r a m e t e r s were confined, i n this stage, within the
following ranges, to i n s u r e s m a l l control deflections:

-0.5 ZC11 50.5

-1.0 5 c12 9 1.0

-0.5 2 C21 50.5

17
 -1.0 S G12 5 1.0

-0.5 2 G21 5 0.5

-2.5 5 G22 5 2.5

As can be seen, the L.E. control surface was allowed a somewhat s m a l l e r range of
variations than the T.E. control surface. Some of these ranges were changed at subse­
quent stages, on the basis of the results already obtained. Figure 1 shows the variations
of Amin and Am= with k'l for a wing s t r i p having no control surfaces (or
[C] = [GI = 0). It can be seen that Xmin is negative throughout the range of ' k
(0.0128 5 k 5 19.5) and that Amax is positive throughout the s a m e range of k. Fig­
u r e 1 represents the basic values of X and any improvements due to control deflections
should be determined by comparison with this figure.

Results for Trailing-Edge Control Only

The optimized values of [C] and [G] for the c a s e where only a T.E. control surface
w a s allowed a r e

The graphical descriptions of these [C] and [GI values a r e illustrated i n sketches (f)

Sketch (f) Sketch (g)

18
Figure 2 shows the variations of Xmin with k" around these optimum values of [C]
and [GI. Figure 3 shows a s i m i l a r variation of Am= with k-1 around the optimum
[
values of [C] and G] .
The following points emerging from these figures are worth noting:
(1)The value of (Amin) is only marginally positive (except at high k values)
opt
and is highly sensitive to off-design values
(2) The most sensitive control p a r a m e t e r s are C22 and G22
(3) The values of C22 which improve Xmin cause Am, to deteriorate
(4)There is an optimum value of the parameter G a l which, essentially, gives
rise to aerodynamic damping. (See fig. 2(c).) This result is consistent with
the observations of Wykes (ref. 10) concerning the effects of damping
Figures 4 and 5 are basically identical to figures 2 and 3 except for the magnified
k'l scale.
By r e f e r r i n g back to sketches (f) and (g), it is interesting to note that the main
effect of the in-phase deflection of the control surface is to counteract any l i f t buildup;
that is, the lift increase due to the angle of attack a is opposed by the f o r c e s created by
the deflection of the T.E. control surface. Furthermore, the out-of-phase deflection of
the T.E. control surface i n c r e a s e s the damping i n bending, on one hand, while the tor­
sional damping is reduced on the other. It can therefore be seen that flutter suppression
is achieved by both reducing the energy input into the system and increasing the dissipa­
tion of energy.
Comparison can also be made between this T . E . control system, which r e q u i r e s
two s e n s o r s (to determine a and h/b), and a T.E. control system which makes use of
one sensor only. The control law of such a single-sensor T.E. control system may
assume either of the following f o r m s .

Figure 2 shows the reduction i n Xmin caused by either letting C21 and
approach zero. It can thus be seen that the system having the two s e n s o r s is superior to
any single-sensor system.

19
 Results for Leading- and Trailing-Edge Control Surfaces
The optimized values of [C] and [q f o r this case, where the system was allowed to
have both L.E. and T.E. control surfaces and was activated by two s e n s o r s , are

0.5

-0.5
= L0.45

Illustration of the physical meaning of these [C]and [GI values is shown in

sketches (h) and (i). Figure 6 shows the variation of Xmin with k - l around these
optimum values of [C] and [G] whereas figure 7 shows a similar variation of Am=. An
identical representation but with a magnified k - l scale is shown in figure 8.

.45 i / b
.5 6lb .05 h / b

Sketch (h) Sketch (i)

A study of figures 6 to 8 yields the following points which are worth summarizing:
(1) The optimum value of Xmin is large and positive over the whole range of k-l.
(2) The off-design sensitivity of the optimized values of [C] and [GI is greatly
reduced as compared with the T.E. control surface only.
(3) The most sensitive control parameters which affect both Amin and Am, (in
practically opposite fashions) are, as before, C22 and G22.
(4)The parameter G12 mainly affects Amin whereas the parameter G21
mainly affects Am,,.
(5) The parameters (211, (212, C21, and G11 have only very s m a l l effect on
both Xmin and Am=.

20
 (6) A variation of G21 can reduce Amax to the point where Am, = hmin. F u r ­
ther variation of G21 may cause the interchange between Amax and Amin. Hence,
t h e r e must be a critical value of -1 below which, the variation of G21 will greatly
affect Amin
(7)The p a r a m e t e r s C11, C12, G11, and G12 have optimum values coinciding
with the limits of their range of variation; thus, an expanded range might improve the
results.
The preceding sketch of the activated (optimized) oscillating s t r i p shows again that
the in-phase control deflections counteract any lift buildup whereas the out-of-phase
control deflection provides both bending and torsional damping forces.
Once again, the effectiveness of a single activating s e n s o r w a s tested. Figures 9
and 10 show typical results of a variation around such an optimized system. Note that
the rotational sensor activates all the important control p a r a m e t e r s except Gal. A
rotational sensor only reduces effectively to letting C11 = C21 = G11 = G21 = 0 and thus
causes through Gzl) the interchange between Amin and .A" Clearly, the two-
(
s e n s o r system is the m o r e effective sensing system and shows a very large effect on
Amin.

Leading-Edge Control Surface Only

An investigation into the effectiveness of an optimized L.E. control surface activated
only by two s e n s o r s w a s made. A typical variation of Amin around the optimum values

[I [I
of C and G , shown in figure 11, indicates that a L.E. control surface, on i t s own, is
effectively of very little use.

Subsequent Considerations
To cut down the amount of work in the following investigation, and on the basis of
the results obtained s o far, it was decided to consider only the two most promising s y s ­
tems: the L.E.-T.E. system and the system comprising the T.E. control only. Further­
m o r e , each of these s y s t e m s w i l l always be considered as being activated by two s e n s o r s .

Remarks on the Comparison Between the Two-Controls System

and the Single-Control System
It is felt, at this stage, that the use of both L.E. and T.E. control systems requires
an additional justification. This is t r u e since i t may be argued that an unfair comparison
has been made i n the simplified flutter model, where an attempt was made to control a
two-degrees-of-freedom system by a single control surface (T.E. only) and thus justify
the L.E.-T.E. control system. It may further be argued that the introduction of two T.E.

21
control surfaces, suitably controlled, may produce some improvements similar to those
of the L.E.-T.E. control system. Sketch (j) shows two such T.E. control surfaces con­
trolling the bending-torsion degrees of freedom of the strip.
On the basis of the aerodynamic s t r i p theory, one can divide the s t r i p shown i n
sketch (j) into two s t r i p s each having a single T.E. control surface. Since these two
s t r i p s perform an identical oscillation, any optimization technique, including the energy
approach, will yield

Sketch (j)
identical control laws for these two control surfaces. Hence, the two control surfaces
effectively oscillate as a single T.E. control surface, the performance of which has
already been investigated. It can therefore be stated, on the basis of this argument and
the results presented i n this work, that the L.E.-T.E. control system is superior to a T.E.
control system.* The physical reason for this superiority lies i n the fact that the T.E.
control surface has combined beneficial-detrimental effects on the system. On one hand,
the T.E. control surface is very powerful in decoupling the aerodynamic nonsymmetrical
c r o s s t e r m s , but on the other hand, it reduces both the aerodynamic damping in torsion
and the torsional aerodynamic stiffness t e r m (this latter t e r m w i l l be later shown to affect
the nonactivated s t r i p s on an elastic wing). The optimization results reflect the com­
promise obtained between these conflicting effects. The L.E. control surface, although
very ineffective in decoupling the nonsymmetrical aerodynamic t e r m s , provides a means
of adding aerodynamic damping in pitch and positive torsional aerodynamic stiffness
t e r m s . Hence, these two control surfaces supplement each other and can yield good
results if properly used.
-- __ -_ ~ .

*This result appears to be self-explanatory. It has, however, an additional impor­

tance in that it points out that two T.E. control systems, when located on a wing, will be
m o r e efficient if the elastic deformations at the controls locations a r e widely different.
Two adjoining T.E. controls can therefore be expected to be relatively inefficient at the
low frequencies (leading to almost identical elastic deformation a t the two adjoining sec­
tions) whereas at high frequencies they may be expected t o be more efficient.

22
 POWER REQUIREMENT

Preliminary Considerations
In evaluating the power required when activating the control surfaces, some further
definition of the fluttering system has to be made. T o avoid analyzing very specific sys­
tems where force-response amplitude and phase characteristics a r e of prime importance,
it was decided to generalize somewhat the problem by attempting to assess the upper
bound of the power requirement. This assessment can be made by assuming that the
force-response-phase relationship is such as t o require maximum power. Furthermore,
only the peak power requirement will be evaluated as compared with average power
required per cycle. The detailed analysis of the power requirement, presented in appen­
dix C, shows that the upper power bound can be written (eq. (C11))

The notation I I means the modulus of the complex number, and [I

F 1 denotes the sum of
the 2 X 2 inertia, elastic, and aerodynamic matrices which appear in the equations of
motion relating to the control surfaces. (See eq. (C4).) The pi]matrix has inertia
t e r m s (both structural and aerodynamic) proportional to w2, structural elastic constant
t e r m s , aerodynamic stiffness t e r m s proportional to velocity square V2 and aerodynamic
damping t e r m s proportioned to wV. Initial results have shown that the aerodynamic
stiffness t e r m s (proportional to V2) contribute most to the power requirement. Hence,
-
the value of P 1 , m S will, approximately, be proportional to wV2 and to the response
squared.
Gust power-spectral input considerations indicate that low frequencies, where both
the input power and the response are large, should be critical from the point of view of
the power requirement. Similar conclusions can be reached i f a uniform cantilevered
beam is considered to oscillate to produce a specified maximum s t r e s s Sm, at a speci­
fied location. T h e inertia loading is proportional to w2d, where d is a representative
oscillatory amplitude. Hence,

d a - Smax
w2
23
Thus, since

and using this relation yields

n

or

Thus, it is again shown that low frequencies will, in general, be critical when the power
requirement is considered.

Data for Power Estimate

The flutter model was assigned the following values to estimate PI,":

Wing chord, 3.048 m e t e r s (10 feet)

Velocity, 243.84 meters/sec (800 ft/sec)
Altitude, sea level
L.E. control, 20 percent chord
T.E. control, 20 percent chord

Iho/bl = 0.2

ho = 0.3048 meter = 1 foot

= 0.044 radian = 2.5O

w = 12 radians/second

[CI =

Mass ratio, 4.

24
 0.04

[B33= [0.04
-0.018666

0.050666 1
"1
0.010666

and where [B3], [B4], b3], and [Eq] a r e defined i n appendix C. This flutter example was
extracted from reference 25. The control surfaces were allowed inertia values appropri­
ate to flat-plate-type controls.

Effect of Power Requirement

Figure 12 shows the variation of the power requirement with the variation of the
control surface p a r a m e t e r s around their optimum value. The upper curve represents the
total upper bound for the power required to actuate a unit span of the wing. The other two
curves represent the decomposition of this power into its L.E. control and T.E. control
constituents. The following conclusions can be determined from figure 12:
(1) The value of Pl,max p e r unit span is very sensitive to the values of C11,
C21, and G11. If i t is remembered t h a t t h e s e parameters have negligible effect on
Amin, one can s e e that the values of 32.3 kW/m (13.2 hp/ft) can be reduced by about
21.6 kW/m (8.84 hp/ft) i f C11 = C21 = G11 = 0.

(2)All the other p a r a m e t e r s , except G21, which mainly affect Amax contribute
little to the power requirement. The parameter C12 is later shovn to be of great
importance and it is worth noting that it has a very s m a l l effect on F l,".
It s e e m s therefore, rather unexpectedly, that all the important p a r a m e t e r s (C12
and those affecting Amin) contribute little to the power requirement whereas the parasitic
t e r m s which do not affect Amin and G21 and which affect Am, have a l a r g e effect
on the power requirement.

25
 Reoptimization and Reevaluation of the Power Requirement With
Unimportant Control P a r a m e t e r s Set to Zero
On the basis of the forementioned results, i t was decided to reoptimize [C] and [G]
by constraining C11 = C12 = C 2 1 = G11 = 0 and'by extending the range of G12 to
-1.5 5 G12 51.5.
The optimized values obtained were

Figure 13 shows the large reduction i n power requirement due to these changes.
The value of 32.3 kW/m (13.2 hp/ft) was thus reduced to 10.7 kW/m (4.36 hp/ft).
Some additional 25 percent reduction in power can be obtained by using L.E. aerody­
namic balance obtained by shifting the hinge by 5 percent chord. Figures 14 and 15 show
the improvement obtained i n both Xmin and Am=. This improvement followed the
increased range of G12, which, i n turn, w a s increased because of the elimination of C 1 1
and G11 and, for the time being, C12. Sensitivity tests around those new values of [C]

[I
and G (the figures are not presented here, except for fig. 16) did not show any new
features which have not already been seen in previous figures.
It can therefore be concluded that the results obtained are very efficient in t e r m s of
power requirement and that these upper bounds are small i n value and therefore cannot
present major problems in practical applications.

COMPRESSIBILITY EFFECTS

[ CI
The simplified flutter model w a s optimized with respect to C] and G under the
constraints that C 1 1 = C12 = C 2 1 = G11 = 0 for M = 0.5, M = 0.7, M = 0.8, M = 0.85,
and M = 0.9. For the case of T.E. control only, no constraints were imposed. In both
[I [I
cases negligible changes in C and G were obtained for all Mach numbers; hence, i t can
be concluded that the optimized control law remains invariant with Mach number.
The curves of Xmin a t the optimized values of the control parameters show
changes with Mach numbers. Figure 16 shows such a variation of a T.E. control system
whereas figure 17 shows similar variation of the L.E.-T.E. control system. It can be

26
seen that Mach number effect is beneficial for the whole k range of the L.E.-T.E. sys­
tem, whereas the T.E. system shows some improvement, at the high k range, with
Mach number increase, and a deterioration at the lower range of reduced frequencies.
It may be noted that the curves of h against k-1 for the M = 0 case were com­
puted by using a numerical aerodynamic solution. It can be seen that the agreement with
the closed-form solutions shown earlier is generally good, except for the very high val­
ues of k. (See fig. 17.) At the very high values of k, the closed-form solution for
M = 0 yields positive A values throughout the whole k range whereas the numerical
solution yields negative hmin values for k > 4.6. In the case of the T.E. system, both
the numerical and analytical solutions yield negative Xmin values for the high ranges of
k. (See fig. 16.) Another point of interest lies in the effect of compressibility on Amin
at high k values. Both figures 16 and 17 show that at high k, all the M f 0 curves
coincide. It can thus be concluded that the compressibility effect is beneficial throughout
the whole range of k values.

SOME PRACTICAL ASPECTS OF THE OPTIMIZED CONTROL LAWS

Preliminary Discussion
In forming the block diagrams representing how the optimized control l a w s can be
achieved, it was assumed that the control surfaces' hinge moments are parameters which
vary rapidly with flight conditions and which may show substantial deviations from theo­
retical values. It was therefore decided to form a control system which does not require
any knowledge of the hinge moments, provided the power requirement is adequate. Fig­
u r e s 18 and 19 show the block diagram of such T.E. and L.E.-T.E. control systems,
respectively. It can be seen that the feedback loop e r r o r function is based on the differ­
ence between the actual control deflection and the desired ueflection as determined through
the optimized control l a w s . Two main problems present themwlves at this stage:
(1) How can the frequency w be determined in o r d e r to perform the required
division shown in the block diagrams and how sensitive will the system be to e r r o r s in i t s
determination ?
(2) The feedback control loop is, essentially, a second-order system having its
resonance frequency and damping coefficients. It can therefore be expected that phase
lags and amplitude changes will be incurred between the desired and actual control deflec­
tions and that these phase lags and amplitudes will be a function of the frequency Of
oscillation.
It is therefore necessary to investigate how sensitive the system is to these varia­
tions and whether practical systems o r system components can meet any ensuillg require­
ments. Figure 20 shows the effect of phase lags for the T.E. control systeln whereas
27
figure 21 shows similar variation of the L.E.-T.E. control system at M = 0.9. How
sensitive the system is to phase lags at the low range of reduced frequencies can clearly
be seen. Figure 21, f o r example, shows that a large positive value of Xmin can be
changed into a negative value by a m e r e 20° of phase lag. On the other hand, at high
reduced frequencies, the phase lag effect becomes much s m a l l e r (see fig. 20) and becomes
negligible at values of k > 1. This result indicates that the feedback control loop needs
to be designed to have a high resonance frequency, relative to the lowest structural fre­
quency, when flutter is expected to occur a t the low range of reduced frequencies. Lower
relative feedback loop resonance frequencies are required when flutter occurs at the high
ranges of reduced frequencies. The effects of the phase lags were investigated at various
Mach numbers; in every c a s e a similar pattern of variation as i n these figures was shown.

Amplitude Effects
Figure 22 shows a typical variation of the amplitude effects due to the feedback loop
for a L.E.-T.E. control system at M = 0.9. It can be seen, once again, that the low
reduced frequency range shows a very great sensitivity to amplitude gains whereas some
beneficial effects can be seen a t the high range of reduced frequencies for moderate val­
u e s of amplitude gains. These results correlate with the phase-lag results for the reso­
nance frequency of the feedback loop. Furthermore, it shows that an intermediate value
of damping is required. Too high a damping value will lead to both a rapid variation of
phase lags and reduction in amplitude with frequency increase, whereas too light a
damping, although yielding small variations in phase lags, will cause large amplitude
variations.
The T.E. control system behaves i n a similar manner as can be seen from figure 23.

Sensitivity of Optimized System to the Determination of the Frequency

The determination of the frequency, through the determination of the period of oscil­
lation, i n a manner shown in figure 24, was suggested by Harley Brixey of the Boeing Co.,
Wichita, Kansas. This method yields directly the l / w t e r m which s e r v e s as a multi­
plier in the controls block diagrams. F o r a multifrequency signal, it has a tendency to
yield an average frequency. The problem therefore reduces to the effect on hmin of
multiplying the whole [GI matrix by a constant different from 1. Figures 25 and 26 show

[I
the results of a constant multiplier of G on the optimized Xmin curve for the T.E. and
L.E.-T.E. control systems, respectively, both a t M = 0. The T.E. control system shows
such a low sensitivity that an e r r o r of 200 percent is hardly noticeable in figure 25. At
very high values of multipliers, a deterioration in Amin, which starts at the high k val­
ues, creeps well into the low k values. Furthermore, the off-design sensitivity, shown
in figure 29, becomes very much amplified (the multiplier Ck = 18). For multipliers

28
smaller than 1, the values of Xmin become proportionally smaller but remain positive.
(See fig. 25.) F o r the L.E.-T.E. control system, improvement i n hmin can be seen in
figure 26 for the lower range of k with a deterioration, which becomes relatively large,
for large values of Ck and high ranges of reduced frequencies. (See fig. 26.) It can
therefore be seen, i n all cases, that inaccuracies in the determination of l / w of the
order of 200 percent to 300 percent (that is, Ck = 2 o r Ck = 3) have little effect on the
results. For Ck < 1, a reduction in Xmin is shown in figure 26 but i n all cases, Xmin
remains positive and still assumes relatively large values. Sensitivity t e s t s for the
L.E.-T.E. control system around the optimum values of [C] and [GI for a large multiplier
( c k = 18) does show a decrease in sensitivity (see fig. 29 as a representative example)
rather than a n increase in sensitivity. Similar results pertinent to M = 0.9 are shown
in figures 27 and 28 for the T.E. and the L.E.-T.E. control systems. These results of
the l / w effect on Amin shed a new light on the damping t e r m s G and the form of
the control law. This effect is discussed in the following section. At this stage, however,
it appears that the optimized control law can be achieved in practice by a careful design.

Some Further Consideration of the Control Law G P a r a m e t e r s

The effect of the accuracy of the l / w t e r m , discussed in the previous section, has
a direct bearing on both the control law and the mechanism of increasing the damping of
the motion of the fluttering system. These points are discussed herein.
A different control law.- It has been shown that the division by w gives rise to
some problems, which although they can be brought to a relatively simple solution, war­
rant some investigation into the possibility of eliminating the need to determine w.
Consider the simplified control law

and modify it to

where the dots represent differentiation with respect to time. Equation (25) can be easily
implemented and does not require the determination of w . Since h/b and 6, are
dimensional, equation (25) should be reduced to a nondimensional form. This nondimen­
sionalizing is done by dividing the [q t e r m s by a reference frequency w r which lies

29
within the range of the expected frequencies of the system. Thus, equation (25) can be
written, for sinusoidal oscillation, as

o r , in a different form, as

Equation (27) is identical to equation (8) except for the factor w/Or. Therefore the effect
of the variation of Ck, as discussed in a previous section, is identical to the effect of
applying the control law defined by equation (26) and moving to a different Ck curve for
different frequencies. (See, for example, fig. 26.) It should be noted that under this
definition, W / w r can assume values l a r g e r or s m a l l e r than 1. Figures 25 and 29(a)
which relate to the T.E. control system show that the deterioration of Xmin coupled
with the high sensitivity to off-design values does not offer any attraction to such a sys­
tem. On the other hand, the large improvement in Xmin at low k values for the L.E.-
T.E. control system for > 1 and the decrease in the off-design sensitivity makes this
wr
simplified control law a very attractive proposition. It may be possible to allow the val­
ues of C and G to be dependent on k through the introduction of filters; thus, one
could choose a small value of w r and yet maintain positive values of Xmin at the very
high values of W/wr. Some initial work done in this direction yielded some results which
proved to be very sensitive to Mach number a t high k and high w/wr values. Some
further work using some more sophisticated programing which will allow one to cope with
the large number of parameters involved in such an optimization is required.
The damping increase.- A second point emerging from the l/w analysis indicates
that although a maximum value of hmin is obtained by varying a single G i j t e r m , a
monotonically increasing Amin can be obtained by varying all the Gij t e r m s by a con­
stant factor (as i n the case of the L.E.-T.E. control system) or a shallow decrease in
hmin may result (as i n the case of the T.E. control system). This behavior was not
detected by the optimization procedure since relatively large parametric increments were
used there. For k values smaller than 1, the G values contribute mainly to the aero­
dynamic damping. From existing literature (refs. 11, 12, and 13) an increase in damping,
in a single degree of freedom only, of a fluttering system may show some improvement
accompanied by subsequent degradation in stability as damping is further increased. It is
also known that an increase in damping by a constant factor in all the degrees of freedom

30
always leads to an i n c r e a s e in stability. On the basis of these results, it appears that
for a T.E. control system, a single s e n s o r like Wykes' (ref. 10) linear accelerometer,
for example, essentially introduces damping i n bending, whereas a two-sensor system is
m o r e efficient in that it enables one to provide some small amounts of damping i n pitch.
The L.E.-T.E. control system allows the control of damping i n both degrees of freedom
and therefore leads to substantial improvements.
This correlation of results with known effects of linear damping suggests that the
effect of linear damping on flutter speed may be explained i n t e r m s of the X t e r m s and
that this energy approach may have a wider application than originally was anticipated.
Some further possible u s e of the energy approach will be mentioned in a subsequent
section.

GENERAL DISCUSSION AND RESULTS

The C r o s s Effects Between Active and Nonactive Strips on a Wing

An isolated s t r i p has so far been considered with the objective of transforming it
from an energy-absorbing element into an energy-dissipating element. If the intention is
to span the wing with such activated s t r i p s , then the problem is solved since the whole
wing turns into a dissipative system. If the intention, however, is to put a very s m a l l
number of such s t r i p s i n some isolated locations along the span, then c a r e must be taken
to insure that the nonactivated s t r i p s a r e not adversely affected by the activated ones. In
other words, the "neighboring" nonactive s t r i p s should not absorb more energy and offset
the beneficial effects arising f r o m the dissipative active strips. With this effect in mind,
possible sources for such adverse interactions a r e sought. However, any such s o u r c e
must be of aerodynamic nature since these forces a r e the only additional forces that the
active s t r i p s introduce. F r o m these aerodynamic forces, the aerodynamic damping t e r m s
and the aerodynamic cross-coupling t e r m s a r e of prime importance i n turning the acti­
vated s t r i p into a dissipative element. Therefore, the possible effects of the direct aero­
dynamic stiffness t e r m s on the nonactive s t r i p s of the wing remain to be investigated.
[I1
These t e r m s do not appear i n the matrix U since they a r e eliminated while it is being
formed. It should therefore be made clear that any changes made to those direct aerody­
namic stiffness t e r m s cannot possibly have any effect on the values of Amin.
Inspection of the optimized oscillation of both the L.E.-T.E. and the T.E. control
s y s t e m s show an upward deflection of the T.E. control with positive values of a (arising
because of the negative values of (222) as shown i n sketches (k) and (1). This upward
deflection gives r i s e to a large nose-up pitching moment that leads to a negative torsional
aerodynamic stiffness t e r m . T h i s negative stiffness t e r m leads to a reduction of the f r e ­
quencies of the whole wing as the airspeed is increased and thus affects the neighboring

31
 L.E.-T.E. c o n t r o l system T.E. c o n t r o l system

Sketch (k) Sketch (2)

s t r i p s i n an invariably unfavorable manner. In addition, this condition may result in very

low reduced frequencies which could not have been anticipated on the basis of the natural
frequencies and velocity of the system; thus, the system is "pushed" outside the k range
of optimization.
In the case of the T.E. system, nothing can be done to produce positive aerodynamic
torsional stiffness t e r m s . The L.E.-T.E. system, however, offers a simple means of
counteracting the nose-up moment by an appropriate deflection of the L.E. control. It is
found that a value of C12 =: 5.6 eliminates the nose-up moment about the quarter-chord
point (aerodynamic center). This value of C12 5.6 i n s u r e s that the aerodynamic tor­
sional stiffness of the oscillation s t r i p can never be negative. For wings having elastic
axes location behind the quarter-chord point, the counteraction of the T.E. control to the
l i f t buildup cancels some existing negative aerodynamic torsional stiffness t e r m s . Swept-
back wings, however, may show a reduction in their positive aerodynamic torsional stiff­
n e s s because of this reduction i n lift.
In order to minimize any possible c r o s s effects between the activated s t r i p s and the
nonactivated strips, C12 should be assigned a value which will offset the s t r i p nose-up
pitching moment about the aerodynamic center caused by the negative values of C22
(around k = 0.5). It is therefore hoped that by not allowing the system to assume nega­
tive aerodynamic stiffness t e r m s and thus maintaining the full structural stiffnesses at
all speeds, only a small number of s t r i p s w i l l be required to eliminate flutter. The
rigid-body modes form an exception to this rule since, i n the absence of any structural
stiffness t e r m s in these modes, the aerodynamic t e r m s are the sole contributors to the
stiffness. The special case of rigid-body modes is discussed in some detail i n a subse­
quent section.
In regard to the aerodynamic bending stiffness, it should be noted that a positive
value of C21 gives r i s e to positive bending stiffness and vice v e r s a f o r negative (221.
The L.E.-T.E. optimized control law s e t s C21 = 0 and thus leaves the bending stiffness
unchanged. The T.E. control law where C21 = -0.35 incorporates some reduction in the
bending stiffness of the system. It can however be modified, if necessary, since its effect
on Xmin is small.

32
 Suppression of Flutter i n Three-Dimensional Wings
The suppression of flutter of a three-dimensional wing can theoretically be treated
as a direct extension of the two-dimensional flutter suppression case. This treatment
can be done at various levels of sophistication ranging from the simplified s t r i p theory
approach which requires some "engineering judgment" to some more elaborate lifting
surface theory used i n conjunction with a modal approach. These approaches a r e
described i n this section.
Consider a wing which flutters at a speed VF and that it is to be increased to let
VF lie outside the range of speeds 0 to VR. The first step involves the inspection of
the deformation of the wing at VF with the objective of determining the spanwise loca­
tion which exhibits the largest product of b4q2. (See eq. (A7).) For constant-chord
wings, this location may be around the wing tip whereas for tapered wings it will be fur­
ther inboard. This latter location will form the center line of a L.E.-T.E. s t r i p of length
L, governed by the optimized control law. This s t r i p will not only stop absorbing energy
but will also dissipate a considerable amount of energy. It is further hoped on the basis
of the precautions taken while the direct aerodynamic stiffness t e r m s were considered
that the neighboring s t r i p s will not extract more energy from the surroundings and thus
offset the net energy input into the system, The flutter equations a r e then solved and the
new flutter speed VF1 is determined. If V F ~still lies within the range 0 to VR, a
second strip is added. This procedure is repeated until the required speed range is
cleared. It can be s e e n that this procedure is possible to accomplish since by spanning
the wing with such s t r i p s , the whole wing is turned into a dissipative system. The
engineering judgment is required to determine the best locations of the s t r i p s to keep
their number to a minimum.
A second possible approach, still based on aerodynamic s t r i p theory, involves the
use of the structural mode shapes to compute the energy matrix [U]. Assume that the
system is allowed to have m mode shapes and therefore m values of X will be
obtained. Determine the span 2 of the L.E.-T.E. control surfaces and optimize Xmin
[ I [I1
with respect to both the control p a r a m e t e r s C and G (in much the s a m e way as in the
case of the two-dimensional strip) and the spanwise location. Once the optimum control
parameters and optimum spanwise location a r e determined, vary the L.E. control param­
eter C12 to offset the negative torsional aerodynamic stiffness effects due to the opti­
mized value of C22. Determine the important p a r a m e t e r s affecting Xmin and discard
the parasitic t e r m s i n the same way as w a s done for the two-dimensional strip. Reopti­
mize Xmin with r e s p e c t to the control parameters constraining the value of C12 to
the specified value and the parasitic control t e r m s to zero, the spanwise location
remaining unchanged. At this stage, the flutter equations are solved again and the new
value of VF is checked against VR. If insufficient i n c r e a s e is obtained, an additional

33
 s t r i p is added in an identical way. Note that a maximum number of n/2 s t r i p s are
required to control the n modes and that if n is sufficiently large, the optimized con­
t r o l law of the s t r i p must be identical to that f o r the two-dimensional s t r i p results pre­
sented. This statement is t r u e since if n is sufficiently large, the mode shapes can be
combined i n such a way to produce z e r o deflections at all the n/2 s t r i p s (which span the
wing) except f o r the active s t r i p treated. Since only a single s t r i p is oscillating under
such modal conditions, the optimum control laws must be identical. In this fashion con­
siderable labor can be saved since the optimized results of the two-dimensional s t r i p
(with or without the t e r m O / W r ) can be used right at the start, and the flutter equations
solved for different spanwise control locations to determine the best location. Alterna­
tively, the spanwise station which produces the maximum value of Xmin is chosen to be
the best location.
A third approach is i n essence identical to the modal approach which does not make
u s e of the preoptimized s t r i p results. T h e only difference lies i n the fact that lifting s u r ­
face aerodynamics, incorporating control forces, are used i n evaluating the matrix U]. [
The s t r i p ' s values for the control parameters can be used, i n this case, to provide the
range of values around which the optimum is expected t o lie.
It should be s t r e s s e d h e r e that the number of mode shapes that have to be considered
in the analysis depends on the design of the control system. The rule should be that all
the modes that actuate the control surfaces have t o be considered. It has already been
shown (when the W / W r effect was discussed) that a control law which proves to be bene­
ficial at the low range of reduced frequencies may become detrimental at the high ranges
of k values. It will, therefore, be incorrect to pursue a flutter analysis involving four
modes, for example, on the basis that these are the modes which yield flutter within the
range 0 to VR. It may well be that the control law may deteriorate the higher frequency
modes to the extent of producing a flutter speed, at a high k value, within the range 0
to VR. It is therefore extremely important to make s u r e that all the modes which lie
within the frequency band that actuates the controls are considered. This requirement
gives rise to a new problem which relates to the practical implementation of the frequency
bounds of the control law. The lower frequency bound can be established by means of a
"washout" filter, and no special difficulties arise. The upper bound does give rise to
some problems since the introduction of a filter for this purpose is accompanied by large
phase lags and amplitude changes, the effect of which has already been shown to be detri­
mental. It is therefore felt that a very good way of filtering out the high frequencies is
through integrations of signals; i t should be remembered that the high frequency signals
a r e of less importance because of the integrations of the accelerometer signals. If fur­
ther reduction of the high frequency signals is required, additional integrations can be
performed on the two signals which form the feedback loop e r r o r function. T h i s proce­
dure is permissible since the control law (eq. (8)) will still be valid i f both sides of the

34
equations are integrated an equal number of times. In this way the proper bounds are
s e t to embrace the f u l l range of the desired frequencies, and to reduce the high frequency
signals to the level of the actuator's friction.
The optimized control law for the L.E.-T.E. control system was applied, analyti­
cally, on a large delta-wing-type transport airplane at M = 0.9, under NASA contract
NAS1-9808. The results showed that a single L.E.-T.E. control system covering 12.5 per­
cent of the wing semispan s and located about 0.1s inboard of the wing tip gave rise to
28 percent increase i n flutter speed. An additional adjoining L.E.-T.E. s t r i p covering
12.5 percent span and located further inboard led to a complete suppression of flutter i n
the range of speeds that were considered. (That is, the maximum speed tried was 41 per­
cent greater than the flutter speed of the nonactivated wing. The flutter speed with two
active s t r i p s appears to be much above this speed, i f at all existent.) This result is very
encouraging i n that it shows that two L.E.-T.E. s t r i p s a r e sufficient to suppress flutter in
a r e a l system.

Suppression of Rigid-Body Modes

The suppression of rigid-body modes does not warrant a special treatment from the
point of view of Amin, provided the reduced frequency lies within the range of optimiza­
tion and the phugoid motion is not considered. The reservation concerning the phugoid
originates from the energy approach used in this work. It should be noted that in devel­
oping the work done per cycle by the system, it has been assumed that the aerodynamic
forces a r e linearly dependent on the attitude of the components of the system. This state­
ment is true for all elastic vibrating modes and it presents a good approximation for all
the oscillating rigid-body modes with the exception of the phugoid. The phugoid will not
be treated h e r e and the p a r a m e t e r s controlling its motion can be readily identified i n the
control law. With these reservations understood, any s t r i p fitted with the L.E.-T.E. con-
t r o l system will dissipate energy when the airplane moves in a rigid-body mode. Special
consideration, however, may have to be made regarding the contribution of the neighboring
s t r i p s to the energy equation (the assumption being made that the activated s t r i p s do not
span the wing). For illustrative purposes, consider two types of airplanes: one with
sweptback wings and one with "straight" wings.
A wing s t r i p with a L.E.-T.E. control system located around the tip of a sweptback
wing is shown in sketch (m). A positive increase in the angle of attack a,which is nor­
mally accompanied by an upward increment in lifting force, activates the T.E. control
upward to counteract the lift buildup. T h i s reduction in l i f t due to the activated control
deflection introduces a destabilizing moment about the center of gravity of the airplane,
or effectively, a relatively l a r g e and negative contribution to the rigid-body aerodynamic
pitching stiffness (due to the large value of Ax). A s already mentioned, since the

35
 Sketch (m)
aerodynamic stiffness is the only stiffness i n pitch, changes leading to its deterioration
may affect the response of the rigid airplane i n a way which causes the "neighboring"
s t r i p s t o absorb m o r e energy and thus, lead to a net destabilizing effect. It can t h e r e f o r e
be seen that a small reduction i n lift at a section located far from the center-of-gravity
location may cause a l a r g e reduction in pitching stiffness. This effect can be readily
overcome by the introduction of a linear stiffness t e r m due to C21 (which was shown to
have a negligible effect on the suppression of flutter). It can thus be seen that a s m a l l
bending stiffness t e r m that is, positive C21) introduced at an activated s t r i p located at a
(
l a r g e distance behind the center of gravity compensates for this deterioration i n stiffness.
In nonswept wings, this reduction i n lift leads to a negligible change in the rigid-body
aerodynamic pitching stiffness. A review of these considerations which essentially deal
with avoiding the degradation of the rigid-body modes due to the activation of an isolated
s t r i p indicates that this problem is of minor importance (since the washout filter required
to suppress the very low frequency range and allow control of the airplane can be designed
to filter out the rigid-body frequencies) as compared with the problem which may a r i s e if
an attempt is made to control the rigid-body modes to improve their stability.
Assume that i t is desired to control, for example, the pitching mode of an elastic
aircraft. On the b a s i s of the foregoing results, a L.E.-T.E. control s t r i p at the root of
the horizontal tail (for minimum effect on its neighboring strips) will i n s u r e the increase
i n dissipation of the elastic modes of the horizontal tail. Adjustment* of C21 will then
provide the stiffness control of the rigid-body stiffness in pitch (together with fuselage
bending stiffness). Any possible adverse effects on the flutter speed due to possible inter­
action with the wing may be treated by fitting activated s t r i p s on the wing at some appro­
priate spanwise location in a manner identical to that already described. A canard s u r ­
face can be treated s i m i l a r to the horizontal surface. The yaw control also reduces to a
s i m i l a r procedure applied to the vertical-tail stabilizer. Finally, it should be noted that
._ ~

*When C21 is given a value different from zero, C11 may be introduced. The
reason for this effect is identical to the one which led to the introduction of C12 due to
the existence of C22. The ratio of C11/C21 should therefore be approximately the
s a m e as ~ 1 2 / ~ 2 2 .

36
such active strips, located at the tail surfaces, can be used to compensate any rigid-mode
degradation due to the wing's active s t r i p s , fitted with the objective of suppressing flutter,
such as the c a s e of the sweptback wing previously mentioned.

Gust-Response Problems
The control of the rigid-body modes, as discussed i n the preceding section, does
not give r i s e to requirements which are contradictory to those necessary to control the
elastic modes. Hence, no difficulties should a r i s e when attempting to control, simulta­
neously, both the rigid-body and the elastic modes. Since this type of problem falls
within the province of gust-response problems, it is evident that the L.E.-T.E. control
system, properly adjusted f o r rigid-body control, should also prove to be a very efficient
gust alleviator.
As an example, consider the response of an airplane to a vertical gust. It is known
that the gust problems a r e usually associated with the rigid-body translation of the air­
plane, whereas others a r e associated with its elastic response. A L.E.-T.E. control s y s ­
tem located near the root of the wing should lead to the dissipation of l a r g e amounts of
energy due to rigid-body movements. Additional control of rigid-body translation move­
ment can be achieved, i f necessary, by introducing aerodynamic translational stiffness
through positive values of C21. As noted i n the preceding section, the existence of
c 2 1 f 0 may lead to the introduction of (211. The problems arising from the elasticity
of the airplane a r e best tackled by fitting a L.E.-T.E. control system at a location near
the tip of the wing C11 and C21 may again be introduced if necessary . It should,
(
however, be remembered that although C11 and C21 have negligible effects on the
)
flutter-suppression characteristics of the system, the penalty they introduce i n t e r m s of
power requirements is by no means negligible.

Effect of Chordwi s e Deformations

The results obtained s o far relate to wings with rigid chords. The extension of the
analysis to a chordwise flexible s t r i p can be made at various degrees of sophistication.
In the following discussion, only the simplest possible extension is indicated. It is
assumed that the chordwise deformation can be approximated by t h r e e straight lines
representing the L.E. control surface, the niain wing surface, and the T.E. control s u r ­
face. T h i s approximation is illustrated i n sketch (n). The problem thus reduces to
determining the control deflection due to the deformation and comparing that deflection,
i n a feedback loop, with the desired deflection determined through the control law. T h i s
result can be achieved i n a simple manner by employing four s e n s o r s : two Linear accel­
e r o m e t e r s and two rotational accelerometers. The two linear accelerometers were
located near the control s u r f a c e hinges at points 1 and 2 indicated i n sketch (0). The

37
 Actual deformation

Sketch (n)

1
4 1

Sketch (0)

output of these two accelerometers can be processed to give a and h/b of the main
p a r t of the surface. The two rotational accelerometers, are located on the control s u r ­
faces, somewhere near the hinge line, at points 3 and 4. (See sketch ( o ) . ) Now

hl - h2
c%= Ax12

where the subscript e refers to the equivalent deformed chord deflections. (See
sketch (p).) F r o m the knowledge of + and h/b, the signals are processed as before to
implement the control law, and then the values of the existing control deflections & and
b e are subtracted while the e r r o r function is formed in the feedback loop.

38
 V-g Plots of a Two-Dimensional Flutter System
The results obtained were applied to a s t r i p flutter problem, the objective being to
check the control-law effectiveness i n t e r m s of a V-g plot. Figure 30(a) shows a V-g plot
of the binary system previously used for the estimation of the power requirements (taken
from ref. 25). The additional inertia and elastic data are

B2 = [ 0'04
-0.018666
-0.04
0.050666 1
0.0625
El=[ 0 o.02J

the m a s s ratio is 4 and B1, B2, E l , and E2 are the submatrices used i n appendix C.
The system has a flutter speed VF/w& of 1.54 and a divergence speed VD/U& of 2.2.
Figure 30(b) shows the s a m e system activated by the suggested control law; that is,

39
It can be seen that flutter has completely been suppressed and that the g values are in
the o r d e r of hundreds.
Figure 30(c) shows the s a m e system but with a very small activation that amounts
to 10 percent of the suggested value. It is interesting to note the resulting major change
i n the V-g plot and the fairly substantial change in flutter speed. This particular result
indicates that wing designs having large m a s s e s , like engines, attached near the wing
trailing edge may introduce camber changes equivalent to control deflections and may
lead to flutter results highly sensitive to the mode-shape definition.

Some Further Use of the Energy Approach

It has already been mentioned that the energy approach may be applied to investigate
the effect of linear damping (structural, viscous o r aerodynamic) on the flutter stability in
t e r m s of energy eigenvalues and energy mode shapes. It can also be used in either pre­
liminary o r advanced design to determine those mode shapes which absorb energy from
the surroundings.

Determination of the Energy Absorbing Modes

For preliminary design, it is possible to represent the mode shapes of the structure
in t e r m s of either synthetic modes or some other known modes taken from a similar air­
craft. Determine the matrix [U] (eq. (A9)) and solve for the eigenvalues (by using

smallest value of X
ill
eq. (A8)) to form the matrix A and the modal matrices QR
[ + iQ]. Determine next, the
(generally negative) and isolate i t s corresponding modal columns
(qR + isI}. Inspection of the elements of both this qR and qI will yield the physical
modes which participate most i n this energy mode shape. If this t e r m is the only nega­
tive X t e r m , design parameters like engine location, external s t o r e s locations, and so
forth, can be used in a manner which will suppress those modes which contribute most to
the energy mode. If there are few negative X values this procedure can be repeated and
preference can be given to those modes which correspond to the lower frequencies
(smaller number of nodes) and higher negative values of A.
F o r an advanced design stage where the flutter mode of vibration {qF} is known,
equation (A10) can be used to determine the energy vector {tR + itI} at flutter. Substi­
tution into the energy equation (eq. (A15)) yields the largest negative product of

. This ith energy mode can then be identified as outlined previously.

40
 In a similar manner, the reduction of the l a r g e r flutter system into a smaller sys­
tem can be achieved by ignoring all those modes which either give rise to very small V a l ­

ues of X (preliminary design) o r very small values of X i (advanced design).

Mechanism of Flutter
An additional use of the energy equation is to further the insight into the mechanism
of flutter. It is shown i n appendix D that the flutter equation

can be reduced to the following equation (eq. (D10)) by the u s e of the energy modal matrix
[QR + '&I] as a transformation matrix; that is,

where the matrices NS1 and Ns2 are symmetric whereas N A ~and N A ~ are anti-
symmetric. All the N matrices and m l are defined in appendix D. The form of equa­
tion (D10) is very interesting since it shows that the A matrix (which is diagonal) is the
equivalent damping matrix. The matrices N A ~and NA2 are nondissipative matrices
since they a r e antisymmetric and their effect is to provide out-of-phase coupling t e r m s
only in the equations of motion. Hence, any nonconservative linear problem can be
reduced in a similar fashion, to a problem of considering on one hand the existence o r
nonexistence of some basically unstable mode shapes (negative X t e r m s ) and on the
other hand, the amount of coupling between those unstable modes and the other stable
ones. It can therefore be seen that the key point for flutter to be at all possible is the
existence of a mode shape with negative A.
The relative importance of the low frequency modes over the high frequency modes,
when flutter problems are considered, is widely accepted in aeroelasticity (excluding panel
flutter). The analysis of an elastic system is invariably performed by coi-sidering its
first few elastic modes. The normal justification, based on the responses of the system,
is not convincing especially when dynamic stability problems are considered. It is, how­
ever, well known that practice provides the best justification for this approach. Inspec­
tion of figures 1 and 2 s e e m s to provide some explanation and justification. At a certain
value of flight speed, the low frequency modes will give rise to low reduced frequencies
whereas the high frequency modes give rise to high values of reduced frequency. This
statement means that the low frequency modes will lie on the far right-hand side of fig­
u r e s 1 and 2, whereas the high frequency modes will lie on the near left-hand side of

41
 these figures. Since Xmin decreases i n absolute value continuously and rapidly as k
increases, i t reaches a value of k at which Xmin is completely canceled by the struc­
tural damping of the system and thus makes it impossible f o r flutter t o exist.

CONCLUDING REMARKS

The advantages of using a leading-edge-trailing-edge control system, driven by

two sensors, and the usefulness of the energy approach have been demonstrated in the
present work. The r e s u l t s obtained show that flutter can be suppressed but that great
c a r e must be exercised. It is appropriate t o stress here, once again, some of the salient
points which were brought to light i n this work:
(1) Oscillatory aerodynamic derivatives are essential i n this type of work. F u r ­
thermore, they must have good accuracy up to high values of the reduced frequency k.
The accuracy of the aerodynamic derivatives used i n the present work was not sufficient
at the very high k values, as shown by comparison with computed exact results. Com­
parisons made with tabulated results, and which showed good agreement, were for values
of k < 1 (since no higher values of k were tabulated). The results presented i n this
work, using numerical aerodynamic solutions, lead to somewhat s m a l l e r values of the
energy eigenvalue X i n the high k range than the exact solution.
(2) Mach number effects are of a complex nature and no overall correction, like the
Prandtl-Glauert correction, can be considered satisfactory. An example which showed
excellent stability from the Amin point of view, over a n extremely large range of k, at
a z e r o Mach number showed some very adverse effects at a Mach number of 0.9 and
k > 0.5. Hence, the oscillatory Mach number effects must be considered.
(3) Control surface and actuator dynamics are of prime importance. Phase lags
and amplitude changes can easily deprive the system of its effectiveness, especially at
the low k range.
(4)All the structural modes that give rise to signals that actuate the control s u r ­
faces must be considered in the flutter analysis with active control. This statement is
true irrespective of the number of structural modes required to represent the important
flutter modes of the unactivated system.
It is believed that the control system which consists of the leading-edge-trailing­
edge controls and which utilizes the frequency ratio O / W y t e r m (and obviates the need
to determine the frequencies) is the best from the point of view of practical implementa­
tion and performance. This condition is particularly t r u e since the very low reduced fre­
quencies lying around the phugoid or short-period modes do not need such large positive
values of Xmin as compared with those at the higher values of k. Further

42
improvements i n the control law can be obtained i f the k range of the system is more
limited than the one treated here.
The implications of a successful flutter suppressor on design philosophy are many.
They span a very wide range of problems: elastic stability, structural efficiency, struc­
tural fatigue life, and riding qualities of the aircraft. Additional work is, however,
required before such a system can be incorporated into a design. Analytical work on a
typical airplane can provide answers to questions relating to the number and s i z e of s u r ­
faces required to s u p p r e s s flutter for different types of airplanes and different categories
of planforms. As a direct extension of this work, systematic analyses can be done on
simple types of airplanes where some form of s t r i p theory can be used. The extension
using lifting surface theories r e q u i r e s the refinement of present numerical schemes to
i n s u r e sufficient accuracy to allow partial span control at high Mach numbers and high k
values. Finally, some carefully designed experimental work is required to provide sup­
port to these analytical results.

Langley Research Center,

National Aeronautics and Space Administration,
Hampton, Va., January 27, 1971.

43
 APPENDIX A

THE ENERGY ANALYSIS

Let

(F} = (-w2 [.+ n p b 4 s ( A ~+ iAI] + [El) {q>

where, at flutter

{F} = 0

and w represents the frequency of oscillation; [B], the m a s s

r e a l and imaginary parts of the aerodynamic matrix,
matrix; p , the density of the surrounding fluid; b, the half of a reference chord; {q},
the response vector; and s, the semispan of the wing.

The vector {q) can be written as

where {qo} is a complex vector of amplitudes. The real p a r t of equation ( A l ) is essen­

tially the s a m e as the imaginary part, except for the initial conditions at t = 0. In a
harmonic motion, where no transients a r e treated, the r e a l part only of equation (Al) can
be considered without losing the generality of the equations. Hence, the forces (F> which
the system exerts on its surroundings a r e given by

where the asterisk denotes the conjugate vector. The vector {F} is clearly a r e a l
vector.
The velocity vector {q) can be obtained by differentiating equation (A2) to obtain

44
 APPENDIX A - Continued
The real part of this velocity vector R{q} is given by

Hence, the rate a t which the system does work on its surroundings is given by R q
and using equations (A3) and (A5) yields
1’KF1

Hence, the work done per cycle by the system on its surroundings can be found by
integrating equation (A6) between t = 0 to t = 2.
Thus,

+ n P b 4 s ( A ~+ iA1)] + [“]){qo]

This equation is a s c a l a r equation and therefore the first expression on the right-hand
([I [I
side of this equation can be transposed to obtain B and E being assumed to be
symmetric
)

Note that the matrices within the square brackets f o r m a Hermitian matrix.
Now determine the eigenvalues and eigenvectors of the following Hermitian matrix
b] extracted from equation (A7):

45
 APPENDIX A - Continued
where

PI =

and represent the vector {qo} i n t e r m s of the eigenvectors of equation (A8), that is,

are square matrices whose columns are the real and imaginary
parts of the eigenvectors of equation (A8), and
c (R + i[I} are generalized modal coordi­
nates (defined by eq. (A10)) associated with the aerodynamic energy.
All the eigenvalue and eigenvector solutions of equation (A8) can be expressed i n
the classical f o r m (ref. 26, p. 6) [QR + iqJ[AJ = [U][QR + iq]. Postmultiplying this
equation by {tR + itr} and premultiplying i t by q"
14 1 I'
o r tR - itIJ QR - iq yields
['

wherePA1 is the eigenvalues matrix, that is,

L J

�4=

The right-hand side of equation ( A l l ) is identical to the right-hand side of equation (A7)
except f o r the ?r2pb4u2s/2 factor. Hence, the left-hand sides of these two equations
can be equated to obtain:

46

I
I

APPENDIX A - Concluded
The matrix IQ: - iQTI represents the modal matrix of the complex left-hand
L -I
(normalized) eigenvectors and therefore (ref. 26, P. 26)

[I
where I is the unit matrix. Hence i? reduces to the form

or

The energy input per cycle into the surroundings has thus been reduced to a quadratic
form i n t e r m s of the generalized energy modal coordinates.

47
 APPENDIX B

DERIVATION OF THE GENERALIZED AERODYNAMIC FORCES

O F A L.E.-T.E. SYSTEM

Consider the determination of the aerodynamic forces acting on the system shown
in sketch (Bl)
Undisturbed p o s i t i o n
- - - -
0

Sketch (Bl)

and assume that the forces acting on a somewhat similar system (as used in ref. 16)
described i n sketch (B2)are known. Note the difference between the h,a,P,6 coor­
-_-­
dinates and the h,qP,6 coordinates

-----
Undisturbed p o s i t i o n

- 4 I I ,Quarter-chord point

W ' O VQ
I -
% *
Sketch (B2)

The arrows indicate the direction of either positive displacements, distances, or forces
The symbols L and M denote the total lift and pitching moment and L is assumed to
act through the quarter-chord point. The force acting on the aileron-tab combination

48
 APPENDIX B - Continued

through the aileron L.E. point is denoted by P and the aileron-tab moment is denoted
by T. The force acting on the tab through its L.E. point is R and the tab moment is
Q. The forces i n sketch (B2), their direction, and points of application a r e identical with
those of Smilg and Wasserman (ref. 18) and can be determined directly from expressions
appearing i n the reference for M = 0. These expressions summarize the works of
Theodorsen, Garrick, Kfissner, and Schwartz (refs. 20 and 21).
The object of the following analysis is to determine the generalized forces acting on
the h, a, p, and 6 coordinates. T h i s object will be accomplished by the application
of the principle of virtual work. The virtual work Wh i n the h coordinate is given by

W h = b -h L
b

and therefore, the generalized force Qh will be given by

Similarly, the virtual work W, i n the CY coordinate is given by

Hence, the generalized force QCYis

The virtual work W p in the p coordinate is given by

W p = -(M - T - P+)p + (L - P)Z4p

and hence

Qp = 3=
60
P(Z2 - 14) + LZ4 + T - M

Similarly ,

W g = Qb - R136

49
 APPENDIX B - Continued
and hence,

T h e s e equations can be condensed into the following matrix equation:

b 0 0

-11 1 0

14 -1 12 - 14
0 0 0

but the forces a r e of the form

where
I{ L

= TPW 2

E = (12

= -138
- 1q)p
33

[&] is a 6 x 6 matrix and z' and

b4

define the hinge locations and are given by

039)

(B10)

Therefore, one can write the following equation based on equations (B9) and (B10)

1 0 0 0
0 1 0 0
0 0 1 0
0 0 (12 - Zq)/b 0
0 0 0 1
0 43JF
-0 0
50
 APPENDIX B - Continued
Substituting equations (B8) and (B11) into equation (B7) yields

where

c ­
1 0 0 0
0 0 0 0 0 1 0 0
1 0 0 0 0 0 1 0
-1 1 (12 - Z4)/b 0 0 0 0 (22 -z4p 0
0 0 0 0 1

0 0 0 -h/b
----
Comparison between the coordinates h,a,u,P,6, and h,a,,P,6 yields the following
relations:

P=P

These equations can be expressed as

-2lb
1
0
0
z4/b
-1
1
0
][I 01

Substituting equation (B14a) into equation (B12) and rearranging finally yields
(B 14a)

51
 APPENDIX B - Concluded

(B14b)

where

1 -21/b
0 1
0 0
[Dl =
0 0

0 0

0 0

and [qT is the transpose of p].

If

then equation (B14b) can be written in the m o r e concise form:

52
 APPENDIX C

ESTIMATION OF THE POWER REQUIRED TO ACTIVATE

THE CONTROL SURFACES

Consider the equations of motion of a L.E.-T.E. wing s t r i p of unit span which is

allowed to oscillate i n bending and torsion i n addition to its control-surface deflections;
that is,

where f p and fs are external forces applied by the control system on the control s u r ­
faces and the matrices a r e of o r d e r 4 X 4. The first algebraic equations in the matrix
equations (Cl) have already been used in the energy analysis in conjunction with the con­
trol law defined by equation (2). The last two equations, when used with the control law,
yield the forces required for the implementation of the control law.
Denote

PI =

r-

H1 ;
[H] = --f
H4

where all the submatrices are of o r d e r 2 X 2. Substitute the control law in the last two
equations of equation (Cl) to obtain

53

L
 APPENDIXC - Continued

( - d p 3 + B4C + iB4G + npb4(H3 + H4C + iH4Gj] + E3 + E4C + i E 4 9 { (:$

h/b
a} = (C3)

By denoting

+ B4C + iB4G + rpb + E4C + i E 4 q

equation (C3) can be written in the form

The matrix (:$ is complex. Its real part will be given by

where the asterisks denote complex conjugates. The angular velocities of the control
surfaces are obtained by differentiating equation (8); that is,

-
The maximum instantaneous power requirement Plyma is given by

max

or

54
 APPENDIX C - Concluded

Equation (C6) indicates that the maximum possible value of R L ~ , ~ J will be given by
(C8)

where the notation I [ means the modulus of the complex number written within the lines,
and

Similarly, using equation (C5) yields

Substitution of equations (C8) and (C 10) into equation (C7) finally yields

Equation (C11) formed the basis upon which the power estimates were made.

55
 APPENDIX D

TRANSFORMATION OF THE FLUTTER EQUATIONS USING

THE ENERGY MODE SHAPES MATRIX

The flutter equation

where

can be transformed to an interesting special form by the u s e of the energy modal matrix
[QR + iq] as a transformation matrix, that is,

-
Substitute equation (D3) into equation (Dl) and premultiply the resulting equation by
bE -
-

iQT] to obtain

Denote
\

where the first matrix on the right-hand side of equations (D5) is symmetric whereas the
second is antisymmetric. Substituting equations (05) into equation (D4) and making use of
equations ( A l l ) and (A13) yields

56
 APPENDED - Concluded

Denote
I

where TS1 and TS2 are symmetric matrices and T A l is an antisymmetric matrix.
Substituting equations (D7) into equation (D6) yields

+ kz - iQT]Eu2(Tsl + iTA1) + Tsq[QR + i%])([R + i<I} =0 (D8)

Denote

i
PSl] = [QETSIQR + qTTSIQJ [QiTA1% - QTTAIQd
[
ps21 = QRTS2QR+%TT S2QI]

F A l ] = [QzTSIQI - QTTSIQR] [QzTAIQR qTTAlQJ

F A 4 = CQzTSZQI - qTT,ZQ.3
Ns2] are symmetric and are antisymmetric. Rear­
ranging equation (D8) and making use of the notations defined in equation (D9) yields
(w2[Nsl] + [NS2] + i
[ 2A -
y w2NA1
{tR + itI } = (D 10)

It should be noted that all the real square matrices i n equation (D10) are symmetric
whereas all the imaginary square matrices are antisymmetric, except for �A] which is
diagonal.

57
 REFERENCES

1. Connor, Roger J.; Hawk, John; and Levy, Charles: Dynamic Analyses for the C-47
Airplane Gust Load Alleviation System. Rep. No. SM-14456, Douglas Aircraft
Co., Inc., July 29, 1952.
2. Knowles, J. A.; and Shockley, E. J.: Flight Tests of the C-47 Airplane Equipped With
a Gust Alleviation Device. Rep. No. Dev-947 (Contract AF-33(038)-6655),
Douglas Aircraft Co., Inc., Nov. 25, 1952.
3. Kraft, Christopher C., Jr.: Initial Results of a Flight Investigation of a Gust-
Alleviation System. NACA TN 3612, 1956.
4. Hunter, Paul A., Kraft, Christopher C., Jr.; and Alford, William L.: A Flight Inves­
tigation of an Automatic Gust-Alleviation System in a Transport Airplane. NASA
T N D-532, 1961.
5. Dempster, John B.; and Roger, Kenneth L.: Evaluation of B-52 Structural Response
to Random Turbulence With Various Stability Augmentation Systems. AIAA Pap.
No. 66-998, Nov.-Dec. 1966.
6. Dempster, Johr. B.; and Arnold, J a m e s I.: Flight T e s t Evaluation of an Advanced
Stability Augmentation System for the B-52 Aircraft. AIM Pap. No. 68-1068,
Oct. 1968.
7. Davis, H. Max; and Swaim, Robert L.: Controlling Dynamic Response i n Rough Air.
AIAA Pap. No. 66-997, Nov.-Dec. 1966.
8. Wykes, John H. : Structural Dynamic Stability Augmentation and Gust Alleviation of
Flexible Aircraft. AIAA Pap. No. 68-1067, Oct. 1968.
'9. Smith, Ralph E.; and Lum, Evan L. S.: Linear Optimal Theory Applied to Active
Structural Bending Control. AIAA Pap. No. 66-970, Nov.-Dec. 1966.
10. Wykes, John H.; and Mori, Alva S.: Techniques and Results of an Analytical Investi­
gation Into Controlling the Structural Modes of Flexible Aircraft. AIAA Symposium
of Structural Dynamics and Aeroelasticity, Aug. -Sept. 1965, pp. 419-433.
11. Broadbent, E. G.; and Williams, Margaret: The Effect of Structural Damping on
Binary Flutter. R.&M. 3169, British A.R.C., 1960.
12. Done, G. T. S.: The Effect of Linear Damping on Flutter Speed. R.&M. No. 3396,
British A.R.C. , 1965.
13. Nissim, E.: Effect of Linear Damping on Flutter Speed. Part I: Binary Systems.
Aeronaut. Quart., vol. XVI, Pt. 2, May 1965, pp. 159-178.

58
14. Theissen, J. G.; and Robinette, W. C.: Servo Control of Flutter. AIAA Structural
Dynamics and Aeroelasticity Specialist Conference and ASME/AIAA 10th Struc­
tures, Structural Dynamics and Materials Conference (New Orleans, La.),
Apr. 1969, pp. 228-240.
15. Moon, F. C.; and Dowell, E. H.: The Control of Flutter Instability in a Continuous
Elastic System Using Feedback. AIAA/ASME 1l t h Structures, Structural Dynam­
ics, and Materials Conference (Denver, Colo.), Apr. 1970, pp. 48-65.
16. Garrick, I. E.: Propulsion of a Flapping and Oscillating Airfoil. NACA Rep. 567,
1936.
17. Duncan, W. J.: Introductory Survey. Vol. I of AGARD Manual on Aeroelasticity,
W. P. Jones, ed., 1959.
18. Smilg, Benjamin; and Wasserman, Lee S. : Application of Three-Dimensional Flutter
Theory to Aircraft Structures. ACTR No. 4798, Material Div., Army Air Corps,
July 9, 1942.
19. Wasserman, Lee S. ; Mykytow, Walter J. ; and Spielberg, Irvin: Tab Flutter Theory
and Applications. AAF TR No. 5153, Air Technical Service Command, Army Air
Forces, Sept. 1, 1944.
20. K k s n e r , H. G.; and Schwartz, L.: The Oscillating Wing With Aerodynamically
Balanced Elevator. NACA TM 991, 1941.
21. Theodorsen, Theodore; and Garrick, I. E.: Nonstationary Flow About a Wing-Aileron-
Tab Combination Including Aerodynamic Balance. NACA Rep. 736, 1942.
22. Bisplinghoff, Raymond L.; Ashley, Holt; and Halfman, Robert L.: Aeroelasticity.
Addison-Wesley Pub. Co., Inc., c.1955.
23. Turner, M. J.; and Rabinowitz, S.: Aerodynamic Coefficients for a n Oscillating Air­
foil With Hinged Flap, With Tables for a Mach Number oi 0.7. NACA TN 2213,
1950.
24. Van d e r Vooren, A. I.: The Theodorsen Circulation Function Aerodynamic Coeffi­
cients. Collected Tables and Graphs, Pt. VI of AGARD Manual on Aeroelasticity,
W. P. Jones, ed., 1964.
25. Theodorsen, T . ; and Garrick, I. E.: Mechanism of Flutter - A Theoretical and
Experimental Investigation of the Flutter Problem. NACA Rep. 685, 1940.
26. Wilkinson, J. H.: The Algebraic Eigenvalue Problem. Clarendon Press (Oxford),
1965.

59
 0­
14000

-2000 ­ 12000

-4000 -
low0 c
8000 ­-i
A m i n -6000­
-6000- max
\

6000 ­
-
-8000 ­
-
4000 ­-
-10000 -
2000 ­
-
I
-12000 ; I I
10 J
-12000 ;00 10 20
20 30
30
, l 40
40 l 50
50 \ 60
60J 70
70 80
80 0 10 20 30 40 50 60 70 80I
l/k
Ilk

(a) hmin. (b) Amax.

Figure 1.- Variation of h with l/k of a wing strip with no control surfaces.
 2- 500 ­
C2l
c22
-L 9 (optimum)
-.25
0 -2 0
-1.8
15 ~ -2 1
-4 - -1.7
-500 - -22
-6 ­
-1.6
-8 ­
A min
. -1000- -2 3
A . -1. 5
min - l o t

-12 c -1500 - -24

-14 c

--18
l6I

-20 1 I I I I I I I I
-2000

-2500
t
0 10 20 30 40 50 60 70 80 0 10 20 30 40 50 60 70 80
Ilk I/ k

(a> c21. (b) c22.

Figure 2.- The effects on hmin of variation of T.E. control system parameters around their optimum values.
 G21
.35 (optimum1 2wr G22

.4 0
.3
.45
0, 2 .
.5
-10- .25 -200 -
.3
-.1
-400 -
.4

- hmin -6M)L \ \

\
-30

-800 ­
-40 - \
-1w ­

-50 ­ .15
-1200 ­
-60 I I I I I I I I
0 10 20 30 40 50 60 70 80 -1400 I I I I I I 1 I
10 20 30 40 50 60 70 80
Ilk Ilk
(c) (321. (4 (322.

Figure 2.- Concluded.

 Id c22
65 85 Id
-I. 3
/-14

5­

4­

?nax 3­ max

2­

0 10 20 30 40 50 60 70 80

Figure 3 . - The effects on h m u of variation of the T.E. control system parameters around their optimum values.
 8 X lo3

6­
t m)

.5 5­

.25
max
h max 4­
.2

. 15 3­

2-
2­

Ilk

(4 G22.
Figure 3.- Concluded.
 .I- c21 2­
c22
/-_ 5
-.
,
.-'
-/ 4
-.35(optimum)
45 -2 ­ -
1 -17

-4 ,­ \
' -21

-6 ­
\
, -16
\ \

'-22
-8 ­
hmin ' -L 5
-10­

-12 ­

-.4L \ -18

-.5 0 I I
1 . 2 3
I I
4
I
5
I
6
I
7
I
8
I
9 10
-.15 -L 3
-"O 1 2 3 4 5 6 7 8 9 1 0
I/k I/k
(a) c21. (b) c22.

Figure 4.- The effects on hmin of variation of the T.E. control system parameters around their optimum values.
 L1

-
.3
.45

-* O5
-.10 ­
t

E:
-,25
\ \.2
h .
min

-8 ­

-.40
-10 ­
-.45
-.50 10I I
1
I
2
I
3
I
4
I
5
I
6
I
7
I
8
I
9
.15
10 -12 I I I I I I I I
1 2 3 4 5 6 7 8 9 10
I/ k 11k
( 4 G21. (4 (322.

Figure 4.- Concluded.

 140 - 160 ­
-1.3
c21
140 ­
120 -

120 ­
loo ­

100 ­
80 ­

h max 80­

60 ­
60 ­

40 ­
40 ­

20 20

0 1 2 3 4 5 6 7 8 9 1 0 0
Ilk Ilk

(a) (31. (b) c22.

Figure 5.- The effects on xmax of variation of the T.E. control system parameters about their optimum values.
 G21 G22
4
/'

140 1 / .3

lop tim um)

max 80 ­ max

60 ­
40 ­

0 1 2 3 4 5 6 7 8 9 10
Ilk
(c) G21.
Figure 5. - Concluded.
h .
1800 ­

16M) ­

1400 -

1200 ­

min 1M30­

800 ­
32 C 1 5.65

600 -
600 - /

I I I I I I
1 60 70 EO
10 20 30 40 50 60 70 80 0 10 20 30 40 50 60 70 EO
0 40
llk
1 41, 50
11k

Figure 6.- The effects on Amin of variation of the L.E.-T.E. control system parameters around
their optimum values.
1800 r c22

-L7 (oDtimum)

1400 - 1500
I- //
-1. 8
-1. 6
-L 9
-1. 5
1200 ­ -2 0

1000 ­
1000 - -14

-2 1
800 - -1.3
500 ­
600 ­
-2 2
400 ­
-L2

-23
0 10 20 30 40 50 60 70 J
80 -500 : I I I I I I 1 1
0 10 20 30 40 50 60 70 80
Ilk

(4 c22.
Figure 6.- Continued.
 1800 - 2500 -
G 12
1600 ­
y\-. 45

1400 -
-.4 2000 ­
..35
3 ,1. 0
1200-
, .9
1500 ­
1000 ­
hinin hmin
800 ­
loo0 ­
600'­
v
400 ­
500 -

I I I I I I
0 10 200 30 40 50 60 70 80 0
Ilk I/k
600'­

4m!­
!
200 -

I
0 10 20 30 40 50 60 70 80
I/ k
 3
8x10

7- 7­

6- 6­

5- 5-

Amax 4 - Amax 4­

3­ -

2- 2 ­>

1­

1 - I
0 10 20 30 40 50 60 70 80 0 10 20 30 40 50 60 70 80
11k I/k
(a) q 1 . (b) q 2 .

Figure 7.- The effects on X m u of variation of the L.E.-T.E. control system parameters around
their optimum values.

4
W
 10 x Id
r

6­

5­

max 4 ­ max

3­

2­

I
0 10 20 30 40 50 60 70 80

Figure 7.- Continued.

 7­ 7­

6­ 6­

5­ 5­

max 4 ­ hmax 4­

3­ 3­

2­ 2­

1-

I
0 10 20 30 40 50 60 70 80 0 10 20 30 40 50 60 70 SO
Ilk

(e) q p
Figure 7.- Continued.
 G22
10 x Id G21 9 x 103 .5
r
.6 I / .4

7 '­

6­

hmax 5- max

4­

3­

2­

0 10 20 30 40 50 60 70 EO
I/ k

(g) (321.

Figure 7.- Concluded.

 30 - 30,r

25- 25 ­

20 - 20 ­

15- hmin 15 ­

10- 10 ­

5- 5-

I
0 1 2 3 4 5 6 7 8 9 1 0 0 1 2 3 4 5 6 7 8 9 1 0

Figure 8.- The effects on hmin of variation of the L.E.-T.E. control system parameters
around their optimum values.

4
4
 C22
-17 loptimum)

25 ­ 25 -

20 ­

h m i n 15­

10 ­

Figure 8.- Continued.

 G12

30 - G.
i.
i
35 ­
L3
1.2
1. 1

25 ­ 30 ­ 1.0

.9
25 - .a
20 ­
.7

20 - .6
hmin
15 ­
15 ­

10 ­
10 ­

5­

0 1 2 3 4 5 6 7 8 9 10 0 1 2 3 4 5 6 7 8 9 10

Figure 8.- Continued.

 G22
.2loptimum)
30r .I
.3
0
.4

-.1
.5

-.2
15 ­

10 -

0 1 2 3 4 5 6 7 8 9 i b
Ilk Ilk

(g) G21. (h) (322.

Figure 8.- Concluded.
 -8000 I I I I I
0 10 20 30 40 50 60 70 80
Ilk
Figure 9.- The effects on Xmin of variation of C11 around the optimum
of the L.E.-T.E. system (C12 = C22 = G12 = G22 = 0) driven by a linear
sensor only.

180

hmin
- //-----­ 1.0

.9

.I L.u
~- --
0 10 20 30 40 50 60 70 80
Ilk

Figure 10.- The effects on Amin of variation of C12 around the optimum
of the L.E.-T.E. system ( C l 1 = C21 = G1l = G21 = 0) driven by a rota­
tional s e n s o r only.

81
 0

-20w

-4000

. -6000
min

-8000

10.000

1 12,000 ' I I I I
0 10 20 33 40 50 60 70 80
l / k

Figure 11.- The effects on A d n of variation of C11

around the optimum of the L.E. system.

82
 Power W
m

: Total Power

Power for L. E. control

optim m

@tiTm

am
Power for T. E. control

I
-.2 I
I
.2
I
.4
I
.6 .8
uLO
c12

(b) c12.
Figure 12.- The effects on the power requirement of variation of L.E.-T.E.
control system parameters around their optimum values. k = 0.075;
V = 243.84 m/sec.
83
 .
1

3l
Total power
n

\
Power for L. E. control

20

Optim

lo0&-.5 -.4 -.3 -.2 -.1 1 1 1 l 1

.1 .2 .3 .4 .5
c21

( c ) c21.

c
45 X Id

40 60/

30

25

I 1 I

a4
 45- Optimum

\
30­
4obtimum

-
- 25
35
30
Power + 2025
15- 20­
Power
15 ­
10 -

10 ­

, 5-
5 < Optimum _____.
Power )r
- T . E. control

0- 1 I 1-A.
-.1
O b

-.4 -.3 -.2

~

-. 5

--:-==
40

35 Optimum

25 Power for L. E. control

Power _W
20

10

Power for T. E. control C)ptiVm

V

1 1 I I I I I
.4 .6 .8 1.0 12 14 1!6

G12

(f) (312-
Figure 12.- Continued.

85
 45 x 103
r 6or

10 ­

I 1 I I I I I I
0 .1 .2 .3 .4 .5 .6 .7 .8
G21

( g ) G21.

3i ; 103
___.
Optipm
­
Total power
v
3(
Opti um Powerfor L. E. c o n t c
8
25

2C
Power

15

10

5 Power f o r T. E. control
. ~­

0 I I I I I I
.1 .2 .3 .4 .5 .6
n

b22

(h) G22.
Figure 12.- Concluded.
86
 25 x 103
r

Power

Total power

10 ­

12 ­

8-
Power for L. E. control
+ - .
Power ! 6 - 9 8 ­
m m

6­
4-
OptiRum Power for T. E. control

2­

0-
4l I
2

0-.4
I
-.2 0
I
.2
c12
I
.4
I
.6
I
.8
I
LO

Figure 13.- The effects on the power requirements of variation of the

L.E. -T.E. control system parameters around their optimum values
k = 0.075; V = 243.84 m/sec.
87
 I

Optimum
Power
Y

0pt)plum

I I I I

Power

2L :t
0
-2 5
1
-2 0
..

(4
Figure 13 - Continued.

88
 25

2(

1:

Power 1
m

1c

16 -

14 ­
10­

12 ­

8­
10 ­
hp
Power W
6- 8- Power for L. E. control

4-
%timum
Power foYT. E. control
2­

I I I I I I I
O- 0 .2 .4 .6 .8 LO 1.2 1.4 L6 1.8
G12

(f) G12.
-
Figure 13. Continued.

89
 zox 103
r
25 r

Total power
2timum
v

Optimum Power for L. E. control

___.
e
Power

@timum Power for 5. E. control

- . A ­
v

1 I I I I I
.1 .2 .3 .4 .5 6
G22

(h) G22.
Figure 13.- Concluded.

90
 Ilk
Figure 14.- The effects on hmin of variation of C11 around optimum
of L.E.-T.E. system optimum with C11 = C12 = C21 = G11 = 0,
G21 = 0.4).
(

9 x 103
r

Ilk

Figure 15.- The effects on Am, of variation of Cll around optimum

of L.E.-T.E. system optimum with C11 = C12 = C21 = G11 = 0,
(
G21 = 0.4).
91
 30r
20 -
M
0
10 ­
-. 5

0
.7

.a

.85

L 1 I 1 I I
30 40 50 60 70 80
11k

M
a
. I

14

42F
021
Ilk

Figure 16.- Compressibility effects of the T.E. system.

92
 M

Ilk

Figure 16. - Concluded.

93
 M
..9
4500

4000 ­
/ .85

.8
3500 ­
.l
3000 -
.5
h min 2500 -
0
2000 ­

1500 -

1000 -

I
0 10 20 30 40 50 60 70 EO
Ilk

hmin

Ilk

Figure 17.- Compressibility effects of the optimized L.E. -T E. system.

94
 M

A .
min

Ilk
Figure 17.- Concluded.

95
 -
s c2 1
-
Amplifier-- Motor
surface

Figure 18.- Block diagram of the T.E. control system.

T.E.
Amplifier-
T.E.
lfotor
- control 'out
surface

1
Amp li fie r
-
ut

Figure 19.- Block diagram of the L.E.-T.-E. control system.

96
 Phase lag, deg

-14 -

-16 -
\\\ 60

-18 ~~ I I I I I I
10 20 30 40 50 60 70 80
Ilk

(a) Wide k-1 range.

Phase lag, deg

I
1 1 5 I
4
I
5 k \ L
y:
; o:
70

Ilk

(b) Reduced k-l range.

Figure 20.- Variation of Xmin with phase lag by use of
optimum values of C and G from M = 0.9 results
(T.E. system).

97
 Phase lag, deg

0
-5
5
10
-10
15
20
25
30

35
40
45

50

55
60
\\ 65
1 ,70
I I I
-”0 10 20 30 40 50 60 70 80
Ilk

(a) Wide k-1 range.

Phase lag, deg

80 r ,-10

h
-lz0O 1 2
l
3
1
4
I
5
1
6
1
7
1 b>070

Ilk

(b) Reduced k’l range.

Figure 21.- Variation of hmin with phase lag by u s e of
optimum values of C and G from M = 0.9 results
(L.E.-T.E. system).

98
 Feedback
amplitude gain
Feedback L
amplitude gain 25
1. 5
3
3. 5
1
4
-1 ­
4.5
-2 ­
5
hmin
-3 -

-A ­
.5

-5 -
\\
-7 I I I I 1 I I \,5 -1
0 10 20 30 40 50 60 70 80 0 .5 LO 15 20 25 3.0
Ilk 11k
(a) Wide k - l range. (b) Reduced k-1 range.
Figure 22.- Variation of Xmin with feedback amplitude gain using the optimum values of
C and G. M = 0.9; L.E.-T.E. system.
 Feedback
amplitude gain
1
l.5
.5
2

2. 5

3. 5

\\ 4.5

I/k
Figure 2 3 . - Variation of Amin with feedback amplitude gain
using the optimum values of C and G. M = 0.9;
T.E. system.

9.1v
Steady-state s i g n a l
VlL p
h

+I0 4).
J- ,
4
_ _ - A
I

Figure 24. - Period-measuring system analog diagram

and signal sketches.

100
 -'I
0
1
10
I
20
I
30
1
40
Ilk
I
50
1
60
I
70
1
80

W
(a) - 2 1; wide k'l range.
W r

l2 r 4

I
10 I
20 I
30 d0 0 -
Ilk
(b) d 1; wide k-1 range.
Wr
Figure 25.- Variation of hmin with W / W r using the opti­
mum values of C and G from the M = 0 results
(T.E. system).
101
 .4

._
.2

.I

0
hmin
-. 1

.-.2

;. 3
-.4
-.5 I I 1 I
I .5 LO 15 20 25 3. 0
Ilk
W
(c) - 2 1; reduced k'l range.
Wr

.35

.3c

.2:

.20

Amin .15

. 1c
.05

-.05 I .5 1. 0
1
L5
I
20 25
I I
3. 0
Il k
W
(d) -5 1; reduced k'l range.
Wr
Figure 25.- Concluded.

102
 45 lo3 K
wr

2500 ­
40 ­

2000 ­
/ .15

1500 ­
10
9 hmin

1000 -

O
0
b
1

(a) -
or
11k
w 2 1; wide k'l range.
500 10

(b)
20

or
30 40

Ilk
50 60

5 1; wide k - l range.
70 80

Figure 26.- Variation of hmin with w/wr using the optimum values of C and G from the
M = 0 results (L.E.-T.E. system).

c.r
0
W
 -
W
Wr
1

2.51 .15

2. 0 ­
.5

1. 5 r
hmin
A .
min
1.0 ­ .25

.5­

-.51 I I I I I I
0 .5 1. 0 L5 2.0 25 3.0
1Ik
Ilk
w
(c) -2 1; reduced k'l range. (d) 5 1; reduced k'l range.
wr wr
Figure 26.- Concluded.
 50 r

-50 -

-100 -

-150 -

hmin -200 -

-250 -

-300­

-350
-

-400 I I 1 I 1
0 10 20 30 40 50 10 0-
Ilk
W
(a) -
W r
2 1; wide k'l range.

20

\
I I -1 I I
10 20 30 40 50 b - H O
ltk

(b) -
W
W r
5 1;wide k-1 range.

Figure 27.- Variation of Amin with W / W r using the opti­

mum values of C and G from the M = 0.9 results
(T.E. system).
105
 hmin

Ilk

(c) -
W
Wr
2 1; reduced k-1 range.

hmin

Ilk

(d) 5 1; reduced k-1 range.

Wr

Figure 27.- Concluded.

106
 9 X Id

I/ k

(a) w
W r
2 1; wide k-l range.

w
w3

hmin

Ilk
w
(b) - I1; wide k'l range.
Wr

Figure 28.- Variation of Xmin with W / W r using the opti­

mum values of C and G from the M = 0.9 results
(L.E. -T .E. system).
107
 w
-
wr
45 r 18 171615

hmin

-5 I I I I I I I
0 .5 LO 1. 5 20 2. 5 3. 0
I/k
(c) -
w 2 1; reduced k-1 range.
Or

hmin

Ilk
w d 1; reduced k-1 range.
(d) -
wr
Figure 28.- Concluded.

108
 5 x 103

G22

I
-250
I
10
I
20
I
30
I
40
I
50
I
60
I
70
\ 1-.3
80
l/k

(a) Variation of G22; T.E. system.

l/k

(b) Variation of (222; L.E.-T.E. system.

Figure 29.- The effects on Xmin of variation of the control system p a r a m e t e r s

around their optimum values. C11 = C12 = G l l = G12 = 0; Ck = 18; M = 0.

109
 5r$
4t
4-
G22
.1

3­

2-

Amin
1­

0 -.3

-1 -

0
I
10
I
20
I
30
I
40
l/k
I
50
-­
60 70 80

( c ) Variation of G22; L.E.-T.E. system.

Figure 29.- Concluded.

110
 ‘r

(a) Basic binary system; no active controls.

(b) Binary system with optimized control

law; L.E. -T.E.system.
Figure 30.- V-g plots of a two-dimensional,
two-degrees-of -freedom (bending­
torsion) system.

111
 (c) Binary system with 0.1 of optimized control law values;
L.E. -T.E. system.
Figure 30.- Concluded.

112 NASA-Langley, 1971 - 1 L-7525

NATIONAL
AERONAUTICS
AND SPACE ADMINISTRAT
ION
WASHINGTON,D. C. 20546
OFFICIAL BUSINESS FIRST CLASS MAIL
PENALTY F O R PRIVATE U S E $300
POSTAGE AND FEES PAID
NATIONAL AERONAUTICS AN
SPACE ADhfINISTRATION

0 5 U 001 26 51 30s 71088 00903

A I R F O R C E WEAPCNS L A B O R A T O R Y /bJLOL/
K I R T L A N D A F B I NEW M E X I C O 87117

ATT E - LOU BOWMAN, C H I E F p T E C H . L I B R A R Y

rVJs,.Lx.JTER: If Undeliverable (Section 158
Postal Manual ) Do Not Retun

‘The aeronautical and space activities of the United States shall be

.
conducted so as to contribute . . t o the expansion of human knowl­
edge of phtW07JZena in the atmosphere and space. The Administration
shall provide for the widest practicable and appropriate dissemination
of inforiiiation concerning its activities and rhe results thereof.”
-NATIONAL AND SPACE ACTOF 1958
AERONAUTICS

NASA SCIENTIFIC A N D TECHNICAL PUBLICATIONS

TECHNICAL REPORTS: Scientific and TECHNICAL TRANSLATIONS: Information
technical information considered important, published in a foreign language considered
complete, and a lasting contribution to existing to merit NASA distribution in English.
knowledge.
SPECIAL PUBLICATIONS : Information
TECHNICAL NOTES: Information less broad derived from or of value to NASA activities.
in scope but nevertheless of importance as a Publications include conference proceedings,
contribution to existing knowledge. monographs, data compilations, handbooks,
sourcebooks, and special bibliographies.
TECHNICAL MEMORANDUMS :
Information receiving limited distribution TECHNOLOGY UTILIZATION
because of preliminary data, security classifica­ PUBLICATIONS: Information on technology
tion, or other reasons. used by NASA that may be of particular
interest in commercial and other non-aerospace
CONTRACTOR REPORTS: Scientific and applications. Publications include Tech Briefs,
technical information generated under a NASA Utilization Reports and
contract or grant and considered an important
Technology Surveys.
contribution to existing knowledge. i

Details on the availability of these publications may be obtained from:

SCIENTIFIC AND TECHNICAL INFORMATION OFFICE

R
f=
NATIONAL AERONAUTICS AND SPACE ADMINISTRATION
Washington, D.C. PO546