Ra - Ja. W?a: Finite State Modeling OF Aeroelastic Systems

NASA CONTRACTOR NASA CR-2779
REPORT
O_
r,,,.
r_
FINITE STATE MODELING

OF AEROELASTIC SYSTEMS
Ra.ja. W?a
Prepared by
STANFORD UNIVERSITY
Stanford, Calif. 94305
for Langley Research Center
NATIONAL AERONAUTICSAND SPACE ADMINISTRATION • WASHINGTON, D. C. • FEBRUARY 1977

1. Report No. 2. Government Accession No. 3, Recipient's Catalog No,
NASA CR-2779
4. Title and Subtitle 5. RePort Date
February 1977
FINITE STATE MODELING OF AEROELASTIC SYSTEMS
6. Performing Organization Code
7. Authorls) 8. Performing Organtzation Report No,
Ran jan Vepa

10. Work Unit No,
9, Performing Organization Name and Address
Stanford University '11. Contract or Grant No.
Stanford, California 94305 NGL 05-020-243

13. Type of Report and Period Covered
12. S_nsoring Agency Name and Address

Contractor Report
National Aeronautics and Space Administration 14, Sponsoring Agency Code
Washington, DO 20546
15. Supplementary Notes
Adapted from Ph.D. Dissertation, May 1975

Langley technical monitor: Robert V. Doggett, Jr. Topical report.
i 16. Abstract
A general theory of finite state modeling of aerodynamic loads

on thin airfoils and lifting surfaces performing completely
arbitrary, small, time-dependent motions in an airstream is
systematically developed and presented. In particular, the nature
of the behavior of the unsteady airloads in the frequency domain is
explained. This scheme employs as raw materials any of the
unsteady linearized theories that have been mechanized for simple
harmonic oscillations. Each desired aerodynamic transfer function
is approximated by means of an appropriate Pad_ approximant, that
is, a rational function of finite degree polynomials in the Laplace
transform variable.
The modeling technique is applied to several two-dimensional

and three-dimensional airfoils. Circular, elliptic, rectangular
and tapered planforms are considered as examples. Identical
functions are also obtained for control surfaces for two- and
three-dimensional airfoils,
117. Key Words (Suggested by Authorls)) 18. Distribution Statement
Unsteady aerodynamics
Aeroelasticity Unclass i f i ed-Unl im i ted
Active controls
Subject Cateqory 39
19. _urity Classif. (of this re_rtl 20. Security Classif. (of this page) 21, No. of Pages 22. Price"
Unclassified Unclass i f ied 188 $7.00
* For sale by the National Technical InformationService, Springfield, Virginia 22161

INTRODUCTI ON ............................................. 1
General l)c's_-r-il_t io_. ,_! !i_. l'r_,t)tem .................. 2
Rev icw _) Pc_-ctincnt i[ _:]',][ill t . ..................... 3
,qllnl!)lnF,/ i (kn_,t C[_lt[ i_.i S ............................ 5
SYNBOIff-] ................................................ 6
FINLTE SfA'[F, W)l)l'!,!h',; c_i ; l!:i:_ t[ _2h;....................... 10
Finite }<]em_mt bI,)de!>; ............................... 10
[dealizati(m u_; a th':_m K,M .......................... ll
FINITE S'I'AI'I] blI)l)['_l_I2q(: OV :'_t'R_'_IP,'N:\H[(- l,(-!Al)_q ............... 17
'['wo I)imells[ulllll A 1 l(_i [:; ............................ 17
[n(t_mprt'_s[l_ e t tow ............................ 17
S!t})_:()Ili, _ I(_.,.: .................................. 34
,qlltS(__V-,,)[i]C _d¢t kt'_llta;_,llJ(" f](IW .................. 40
Tllree 1)_mensioilttl i,ii k i'n,4 Snvfcwes .................. 44
SUGGESTIONS FOR F[:'I'irRt" k'l<<I,?ARr:ll .......................... 59
CONCLIlI) [NG Rt:tMAI{KS ....................................... 64
APPENt)]X A - [Ni,'i l_t,iP,l_:i] (i<_i'i,'!_:lt.r_'i's !-oI_ A BEAH-ROI) WITH A

FLEXIBLE ROOT . ......................................... 65
APPENDEX I{ I[!TE(;b:Ai, F.t_t'A*I'I{3NS 1;'{)I< IINSgTIEAI)Y At:.R()I)YNAM[CS 66
APPENDIX (1 - t:Ai(;[ I._'T[')!; ('_ \! !<',!ii'/kA!11('. f,()AI)S FOR SIMPLE

HAfCHONICAId){ O}{(]ll,!Fd]'iT;(; ',l}¢1'(lll,q ANI) ].[FTTNG SURFACES .. 83
APPENt)tX b - PAl}l{ ..\i'}'}<tJ;tl_!,,'_;i':¢ ANI_ i5!II2(;RAI, EQUATIONS ... 98
R[-FERI<NCES .............................................. 106

TABLES
Number Page
i. Coefficients of Rational Function Approximations to

Non-dimensional Aerodynamic Load Coefficients for a
Two-dimensiona] Airfoil,: M = 0.3 ...................... 117
, Coefficients of Rational Function Approximations to

Two-dimensional Airfoil, M = 0.4 ...................... 118
o Coefficients of Rational Function Approximations to

o Coefficients of Rational Function Approximations to

5_ Coefficients of Rational Function Approximations to


Non-dimensionalAerodynamic Load Coefficients for a

Two-dimensional Airfoil, M = 1.75 ..................... 123
8_ Coefficients of Rational Function Approximations to

Two-dimensional Airfoil, M : 2.0 ...................... 124
9_ Rational function approximations of generalized aero-

dynamic loads in p_unging (1), pitching (2), chord-
wise _ending (x/r) , (3) and, spanwise bending
(y/r) (4) for a circular wing in incompressible
flow ................................................... 125
lO. Indicial lift in plunging for circular and elliptic

126
wings ,,.,,,,,,,°.o,6o.o,,,..,.,4.°°.oo.,e..,,.o,°o,.,,,
Indicial lift in plunging .............................. 127
Other indicial functions in incompressible flow ........ 128
iv
Number Page
15. Indicial functions for generalized forces for impulsive

plunging (1) and pitching (2) displacements M = 0.7,
AR = 6, Rectangular Planform ref. length = semi-span .... 129
14. Coefficients of Polynomial Approximations to Mode Shapes

in an Axis System along the Free Stream (x) and Normal
to it (y). (Origin at 40% Root Chord from the Leading Edge) 130
15. 0 ql
Quasi-steady generalized loads qij= qij + ik ij by the
Doublet-Lattice method ................................. 131
16. Generalized piston theory loads using lattice

integration scheme at M = 0.6 .......................... 132
17. Rational function approximations to generalized loads

on a swept wing, AR = 4.17_ with a control surface
in first bending (1) first torsion (2) and
control surface modes (3) ................................ 133
V
Figure Pa_t_
i. Swept wing with two f]ap_: _L<i _ir _ i p_Jnts ........ 134
2. Theodorsen's function C(k) _ i _ i ik) ............. I {5
3. _(s) for rcal va]ue_ _,,f_ ........................... l_
4. Error in F(k) i'Tru_ - Ap_,r<_-:j_:-_J : ..................... I _i
5. Error in _'(Ii) ';_ ,'u, - App!:'o:<i:r" : ................... 1 _'S
6. Error in _('s) .................................... 139
7. Wagner' := ffunc. Lion k[ (t) ,:ob+ ai m-_ :_rd_:__x:i'mat, ely
by Kussne,'h 14(I
_° Error in K(t) ....................................... t4!
9. Indicia] _"esponsc to;" m_',,,_.:k !r_:_,,, _:e _u2",_ . ....... 142
iO. Indicial response for m<,v] ...... ..... " _usts 14%
ii. Indicial lift for an impuA:<],.':_ i'! ::,_-_'_
displacement_ AR ............................... 144
12. Indicial lif't for an :im_,uHs],',: P_ ' ! _C
displacement .................................... I ' _:
13. Indicial !ilt due to i::@ul_:iv: ti_[ ._tation ......... 1_{)
14. Indicia! moment in plun_<i_4_ ......................... 14
15. Indicia! moment in p i<ching ......................... 14S
16. Indicia! moment due _o N;g) _,: ,! . ................ [:*'._
17. Indicia! partial mome"n_L. JL, p,,_:_,:, :: ................. 1 _{
16. Indicia.]_ pa_.tial momsnt in -._ ,_ _, 191
19- Indie{a _ _a_'kial r_b,metK _k_ ! , _,, l)i!

Figure Page
Influence of compressibility on Wagner's function ...... 153
Indicial lift due on impulsive plunging displacement,
M _ O • e , • • • • • ,, • • , • • • • • • • • • • • ,, , • • • • • * o, • , • • • • , • e , ,, , ,
154
22. Indicial lift due to impulsive plunging, circular
planform, M = 0 ....................................... 155
23. Indicial moment due to the impulsive plunging,
156
circular planform, M = 0 .............................
24. Indicial lift due to an impulsive parabolic span-
157
wise bending mode, circular planform_ M = 0 ...........
25. Indicial moment due to an impulsive parabolic
158
spanwise bending mode, circular plan_form, M = 0 .......
26. Variation of with aspect ratio for rectangular

C_ b
159
and elliptic wings ...................................
27. Indicial lift due to impulsive plunging obtained
from exact Pads approximants, elliptic planform,
160
M _ 0 e,e,,,e.,,,6_.,eo,..eee,.e.. °o,.o,...o.°e,,,,,ol
28. Indicial lift due to impulsive plunging obtained
161
numerically, elliptic planform, M = 0 ................
29. Indicial lift due to impulsive plunging,
162
rectangular planform,AR = 6.0 ........................
30. Indicial lift due to impulsive plunging,
163
rectangular planform, AR = 4.0 .......................
31. Influence of AR on the indicial lift in plunging,
164
rectangular planform, M = 0 ..........................
vii
Page
Figure
32. Influence of finite aspect ratio and sweep on
the indicial function for impulsive plunging
165
at M= 0.7 ...........................................
33. Indicial generalized forces in plunging (i)
and pitching (2) displacements rectangular
wingAR = 6.0, M = 0.7 ............................... 166
34. Influence of compressibility on the indicial
function for impulsive plunging, rectangular
planform, AR = 6.0 ................................... 167
35. Rolling moment on an impulsively rolling wing,
AR = 6.0, M = 0.0 ...................................... 168
36. Indicial lift for impulsive plunging, tapered
planform, AR = 5.84, M = 0.0, TR = 0.524 ............... 169
Swept wing, AR = 4.16, A = 45 ° , M = 0.6 .............. 170
Indicial function for qll' due to impulsive
displacement in the first bending mode, swept wing_
AR = 4.16, A = 45 ° , M = 0.6 .......................... 171
39. Indicial function for q12 due to impulsive displace-
ment in the first torsion mode_ swept wing,
AIR = 4.16, A = 45° ,M=0.6 " .... 172
Indicial function for q13' due to impulsive flap

173
rotation .............................................
174
Indicial function for q21 .............................
175
42. Indicial function for q22 .............................
176
viii
Figure Page
177
45. Indicial function for q32 .............................

178
46. 179
Indicial function for q33 ............................
4?. Block diagram for control law synthesis
procedure ............................................. 180
ix
FINITE STATE MOPEI,IN(; OF AFROEI,ASTIC SYSTEMS
Ran jan Vepa

Stanford IJnivers_ ty
SFM_b\RY
A general theory of finite state modeling of aerodynamic loads on
thin airfoils and lifting surfaces performing completely arbitrary, small,
time-dependent motions in an airstream is systematically developed and
presented. In particular, the nature of the behavior of the unsteady
airloads in the frequency domai> is explained. This scheme employs as
raw materials any of the unsteady l inearized theories that have been
mechanized for simple harmonic oscillations. Each desired aerodynamic
transfer function is approximated by means of an appropriate Pad6
approximant, that is, a rational function of finite degree polynomials
in the Laplace transform variable.
The modeling technique is applied to several two-dimensional and
three-dimensional airfoils. Circular, elliptic, rectangular and tapered
planforms are considered as examples. Identical functions are also ob-
tained for contro] surfaces for two- and three-dimensional airfoils.
] N TR(H)UCTI (?N
In the last decade r,tpid advances have taken place in the area of
automatic control of practicg_l engfneering systems. The vast tech-
nologicai developments in autopilot: design and in the design of aircraft
take-off ,and landing systems hn_ led to the possibility of applying
this technology to control the vibration modes of aircraft wing structures
and the elimination of aer(,clastic instnbitities in the flight envelope

of the aircraft. Although the mathematical theory of distributed parameter
systems has made rapid advances recently, it seems more expedient, from a
practical point of view, to approximate aeroelastic systems, mainly aircraft
wings and control surfaces, by finite state models. The techniques of
approximating aircraft wing structures by finite state models which make use
of finite elements and other structural idealizations are well known. No
systematic techniques exist, however, for approximating the aerodynamic loads
on these structures by compatible finite state models for aeroelastic purposes.
Thus a systematic theory for approximating aerodynamic loads on aircraft wings
by finite state models used along with well-known techniques of structural
idealization could be tremendously useful not only for understanding aero-
elastic instabilities but also in the development of control systems for
suppressing aeroelastic instabilities. Such theories can also prove helpful
in the minimum weight design of aircraft structures.
This paper is concerned with the finite state modeling of aeroelastic
systems. The well-known theories of modeling of aircraft wing structures are
briefly presented. A general theory is then developed for the modeling of
unsteady aerodynamic loads on wings and airfoils. These aerodynamic models
may be used in conjunction with structural models for aeroelastic purposes.
General Description of the Problem
The general techniques of calculating unsteady aerodynamic loads for
simple harmonically oscillating airfoils and lifting surfaces are out-
lined in [_ , [2J and [3]. With these techniques it is possible to
calculate the unsteady aerodynamic loads for different modes of oscilla-
tion at a given frequency of oscillation. Little is known about che
analytical behavior of these loads in the frequency domain. Thus it is
2
first essential to identify the behavior in the frequency domain of the
aerodynamic loads on airfoils and lifting surfaces.
The next step is to approximate this behavior in a manner that will
permit the construction of aerodynamic models, which may be used along
with structural models for aeroelastic purposes. Also, for the work
described in this paper existing techniques of calculating aerodynamic
loads for airfoils and lifting surfaces were utilized whenever possible.
Review of Pertinent Literature
The theory of finite state modeling of structures for dynamic

analysis is well known. The various methods of weighted residuals [4],
finite element techniques [5] and variational techniques [4] have all
proved extremely useful for analytical modeling purposes. Recently
system identification techniques [6] have been formulated for approxi-
mating structures by finite state models from experimental data.
The author is not aware of any systematic modeling procedures in
unsteady aerodynamics. However, there were several related developments
in the past fifty years. In 1925, Wagner[1,3] first studied the

growth of lift on a two dimensional airfoil in incompressible flow due
to an impulsive change in the vertical velocity of the airfoil. Garrick

[7] , later showedthe relationship between Wagner's solution and Theo-
dorset' s solution [8] for the lift on an oscillating airfoil. Sears [9]
showedthe relationship between solutions for a sharp edge gust and a
sinusodial gust. R. T. Jones [I0] first considered the aerodynamic forces
on finite wings of elliptic platform in non-uniform motion in incompressible

flow. W. P. Jones [11] calculated the lift on rectangular and tapered
wings for impulsive motion in incompressible flow. Lomax, et al., [12]
solved the problem of obtaining the lift and momentfor impulsive motion,
3
exactly, at the starting instant< _ _ _ ,:_,i _ i <r., ,.,J_th_,_itb _ipersonic
steady flow over three dimensio_la] l_f:t _ _ - _-ze_:_,,_, and brisc]_ler
[13] and Drischler [cl4qj used the Fo1_ri_ _ t _-_....i, _i.... l_tJonship for osci]]a-
tory and impulsive motions deve]oppd })_ _:' : _ ,,:, {:,!tafne_i approximations
to the indicial lift an<] moment in pl<u_:,_il_ :{i_ {_ :i_it_ in compressible
flow, over two-dimens_onal airfoi]s. "_i i, _ ., _ <:.m_idc, red t},__ trans:ient
loading on wide delta w_nzs and recta.n£_];i_ _' _;,, _ _ul)ersonic speeds.
Recently Djojodihnrdjo and Widnail i ]):] _r, }i,.:; [18-] developed a
numerical technique for arbitrarily mov[_1_ ,_[_i, _ : m<i lifting surfaces
based on Green's representation theu,,_.__; i:_ ._,t_, i '_ <he,__ry.
While the theo<{ to be prese_ted i_ ri, s.... _, _{_ so[le ,7_iill_onfeatures
with each one of the above techniques, one ,],._s;_,,_i _,_i_ <_om them any
insight into the general behavior of ti_e _ ,,i,_,_,_. i_ads and pressure
distributions in the frequency dol_

_,a]n.
" F]_
. {i__
"" "_{ r<_<es in the beha¢ior of
incompressible and compressible fLow.,_ }_._.v,.: _t i,.:,_,_]u]L_: exp]ained.
Several techniques have been presented [r_ _, _ ,'_-'t!<r the calculation
of aerodynamic loads for simple ha_-m,.u_.i-,I I._.... <. , .: _,_{ airfoils and
lifting surfaces. _ese techniqu_'s h_,, !)_'_'_ t_ _._ <_ by ,<sh!ey, Widna[l
and Landahl _9], Landahl and Stark i2_Ii. :r..........

r ,iij and Ashley _22j.
Recent methods now widely in use are th_ !.,._,q_, :_._ t lon techn_quer_ of
Cunningham _3j, Rowe [24j and Laschk_ ;_ . .,- r, doublet-lattice
I .,.- ]
technique of Albano and Rodden Lzoj. i _,_
• _, _< Iio ]-e :{n extremely
useful and practical tool for aer<_e]asti _,_,_, _ ,. [J(,me t_f t rlese
methods and their extensions are b_iel ]v h, 4, _ } :, :i Ai_pendix C.
In this paper _h:: th_-ory of rinit{ _ _.{i_ _. of stm,ctures is

first outlined. Then a systematic procedure is presented for the modeling
of unsteady aerodynamic loads. Some contributions are also made towards
extending and modifying methods for calculating aerodynamic loads for
oscillating airfoils and lifting surfaces.
Summary of Contributions
(I) A general theory is presented to describe the behavior of
the aerodynamic pressure and loads in the frequency domain.
(2) A numerical technique has been developed for modeling the
aerodynamic loads in the frequency domain.
(3) A numerical technique is developed for calculating the aero-
dynamic loads on two dimensional airfoils with trailing and leading
edge flaps, for a fairly large frequency bandwidth. For three
dimensional lifting surfaces the bandwidth, for which the doublet lattice
method leads to reasonable results, is improved.
(4) Indicial functions have been obtained for various types of
impulsive motions for wings of different planform.

SYMBOLS
distance of elastic axis from the center of gravity.
b reference semi-chord of wing.
speed of sound in the free street.
Co distance of flap hingeline from the center of gravity.

l
C(k) Theodorsen' s flmetion.
non-dimensional pressure coefficient _ Ap/q

Cp
C (x,_) bending influence function.

zz
Cee(x,_) torsional influence function.
E1 flexural rigidity.
GJ torsional rigidity.
vertical displacement of wing elastic axis.
Hankel functions.
I unit matrix.
Ie moment of inertia of wing.
moments of inertia of flaps.
modified Bessel functions of the first kind.
_b
k reduced frequency -9- "
6
Kh wing root stiffness in shear
wing root stiffness in lateral rotation.
wing root stiffness in torsion.
K(t) Wagner's indicial function.
stiffness matrices.
Ko,K1
K2(X,s ) Kernel function for two dimensional aerofoils.
K20(X,s) Kernel function for two dimensional aerofoils in incompress-

ible flow.
K3(X,Y,s)Kernel function for three dimensional lifting surfaces.
Control surface stiffness constant.

Ksi
defined in figure i.
I
semi-span of wing.
_w
Lh lift in the plunging mode.
re(x) mass per unit length.
Mo, Mass matricies
M Mach number
Me aerodynamic pitching moment.
aerodynamic flap moment.

M_ i
NB number of collocation points.
NF number of flaps
p Laplace transform variable.
p(x,s),
p (x,y,s) pressure distribution.
_p(x,_),
pressure jump across the wing.
_p(x,y,s)
q dynamic pressure = 1/2 pV 2 •
nodal displacement vector.

qo I
ql modal displacement vector.
Laplace transform of qo' q1"
.th
i, j element of Q(s,M).
qij (s ,M)
Q(s,M) aerodynamic load matrix.
pb
S -
V
se(x) unbalance of entire wing.
unbalance of i th flap.
S_ i
t time
.th
Tc_i control torque on i flap.
U gust velocity.
U(n,a,t) Kummer functions.
V free stream velocity.
w(x,y,t) downwash velocity.

w(x_y,p) Laplace transform of w(x_ y, t).
x ,y _z _!artesian coordinates.
Y= y- _
Z = z -
th
zj (x,y) displacement of the j mode.
angle of attack
_ for M < i_ _-i for M > i.
8i (t) flap rotation of i th flap
o(_,t) torsional rotation
Laplace transform of Wagner s function.
_(s) : sm(_)
Vt Vt
T or --
b
w
Cartesia_ coordinate system

FINITE STATE MODELING OF STRUCTURES
Finite Element Models
The finite element method is often regarded as generating a dis-
crete model of a physical system. The method may also be viewed as a
variational technique, where an attempt is made to minimize the demand
for ingenuity in the construction of trial functions. This discrete
matrix technique for the formulation and solution of linear dynamic
problems in engineering mechanics has been widely discussed in the
literature [5]. This approach has proved useful in obtaining approxi-
mate analyses of complex structural configurations that are difficult
to handle by exact mathematical formulations. The bookkeeping required
for solving the large number of linear simultaneous equations involved
is readily handled by matrix algebra techniques and the resulting
analytical formulations are tremendously simplified. Thus concurrent
development or revision of different sections of large digital computer
programs to perform the analysis is feasible. This feature has led to
its use and acceptance as a basic tool for dynamic structural response
calculations. The aim of this section is to briefly describe how the
technique can be useful for finite state modeling of wing structures.
The efficient utilization of the discrete element influence-
coefficient approach in a digital computer would provide for con-
struction of a stiffness matrix and a mass matrix for the entire wing
I0
structure by simple superposition of each matrix, from corresponding
matricies for the discrete elements that model the wing in detail. A stiffness
and mass matrix generator program (SAMGEN)

was written for this pur-
pose. Several different types of elements were incorporated in the
program. This program was found to be extremely useful in structural
optimization work [27].
Idealization as a Beam-Rod
The swept wing structure shown in Figure i is considered. The
wing is connected structurally to the fuselage, so that the root is not
completely rigid. Further it is assumedto have a finite number of trail-
ing edge flaps, NF in number, which provide for the control torques.
The flaps are assumedto function as rigid bodies which are good approx-
imations for most control flaps. The wing is assumedto have a large
aspect ratio, at least, for structural purposes. This permits the
wing to be modeled as a beam-rod. The actual root is replaced by an

effective root normal to the elastic axis of the wing. Rotational
inertia, shear deformations and sectional bending are neglected.
Let h(x, t), e(x, t), _i(t) i = l, 2, 3, .-- NF, be the vertical
displacement of the wing along the elastic axis, the torsional dis-
placement of the wing along the elastic axis and the angular displace-
ments of the flaps respectively. The kinetic energy of the wing/flap
system may be written as
11
i=l i
_-- I' .i )
th
where _i(t) is as_-mr_!_,:i
to be equal _,{__:<,_'c :_ "egJons where the i
flap does not cx_,_!:,.

• _,_,_
fr,_ c poL_tieN ._ , n, _l, i . _y:-_t.em may be written
as
-,7% fx t_
1 _
---- _
(EI (
\X/'
4- _j (ix ( -- ..... .> )dx
C"
, j
1 O, 'p .-]Z.W :
r_ . _:( t "1
A prime (') w_ll b,' use_i to rqu'_--':en_ ..... '-._, ! . . ' to repr_sent
37
The aerodynamic fc_ ,,_:: and control L ':.<.-.. :to <, _,_,..._,<] a,s non-
conservative for<u_: .
Using the [_ra:'_<i'>.u tv_,h;;iauc_. _,. _,_,._ ,.I _owin_ equations
of motion and boLLn_!ari{ ,'malt +.i<:ns .
12
NF
m(x)}_(x,t)- Se (x)e'(x,t) - (i_lSSi_i(t))+ (EI(x)h"(x,t))" = Lh(X) (2.3a)
NF
-S8(x)_(x,t)+Ie(x)_'(x,t)+ (i=l(ISi
+7 b(c i- a)Ssi)fli'"(t))-(GJ(x)e'(x,t))'
= Me (x) (2.3b)
K8i8 i (t)
-Ssi_ (x,t)+ (Isi+ b (ci- a)Ssi)_" (x,t)+ Isi_i (t)+ _i- _i-i
= Msi(x)+ Tcsi , i = 1,2,_,...NF

(2.3c)
where Lh(X), M0(x ) and Msi(x), i = i, 2, 3, -.. NF, are the aero-
dynamic forces and moments and Tcsi , i = i, 2, 3_ ... NF, the control
torques on the flaps.
The boundary conditions at x = O are given by,
(EI(x)h"(x,t))'- _h(x t) : 0 (2.4a)
EI(x)h"(x,t)+ KRh'(x,t ) : 0 (2,15b)
GJ(x)e'(x,t)+ _@(x,t) = 0 (2._c)
The boundary conditions at the wing tip x = _w are
El(x)h"(x,t) = 0 (2.5a)
(EI(x)h"(x,t))' = 0 (2 .Sb)
13
GJ(x)@'(x,t) = 0 (2.5c)
If El(x) and GJ(x), the flexural and torsional rigidities are
zero at the wing tip, the boundary conditions are,
EI'(x)h"(x,t)
= o (2.6a)
2El'(x)h'''(x,t) + El''(x)h''(x,t) = 0 (2.6b)
- GJ'(x)e'(x,t) = 0 (2.6c)
In the above equations the inertia and aerodynamic loads are also
assumed to be zero at the wing tip •
Since the properties of a practical wing are not uniform it is
difficult to solve the above boundary value problem even in the
absence of external loads. Hence these differential equations are
reduced to integral equations and then discretized.
The influence functions Czz(X_ _ ), the static deflections at
station x due to a unit load at station _ along the elastic axis, and
Cee(x , _ ), the state torsional deflection at station x due to a unit
moment at station _ along the elastic axis _ for a wing with a flexible
root_ are given in Appendix A.
The inertial properties are assumed to be lumped at NB collocation
points, xi, along the elastic axis. Then following standard techniques
it is possible to rewrite the equations of motion, in the absence of
14
external loads, in the form,
[I ] [qo } = "[Ko ]-I [M0 ] [qo ] (2.7)
where
[I] is the unit matrix,
{qo _ is the vector of nodal displacements, torsional
rotations and flap rotations,
[Ko ]-l is the influence coefficient matrix_ and [MO] is
the inertia matrix.
(2.8)
Defining [Uc]T : [TCSl _ Tc82 _ ... TCSNF]
NF x NB • NF x NB . NF xNF
and [BO]T = [ 0 " 0 : I ] (2.9)
we may rewrite the entire equations of motion as
[Kl]ql + [MI ] ql : FO (t) + BU c (2.1o)
where Fo(t ) is the aerodynamic load vector, U c is the vector of con-
trol torques, B the control torque distribution matrix, and ql the
vector of modal displacements. It is also assumed it is possible to
measure modal displacements and rates and these measurements can be
written as,
: Eql + 241
15
If a linear aerodynamic theory is assumed, the Laplace transform
of this load vector, Fo(P), where p is the Laplace transform variable,

is specified as,
where Q(s,M) = [qij (s,M)] is an aerodynamic load matrix, defined by
the relations,
Zi(x'Y) ApJ(x'y's'M) dxdy

qij(s'M) =_wA1 _A b 1 v2
w
where A is the area of the lifting surface for a three dimensional

w
lifting surface and equal to the chord for two dimensional aerofoils,
i
z (x,y) is the modal deflection surface in the ith mode and Ap] (x,y3
s,M), is the Laplace transform of the pressure difference in the ]th
mode. The modal deflection Surfaces may be obtained by solving
the free vibration problem given by equation 2.7
A general theory for approximating the aerodynamic loads qij(s,M)
in the frequency domain, will be presented in the next chapter. This
theory will permit the entire aeroelastic system to be replaced by a
finite state modal.
16
i'[N['l_:: %'IAiE _!i',l)<l.lN!; C)F .,\F,Ri_I)YNA_I[('. I_OA1)S
In thi_ _ecti,_n, a gen,_'al method o£ approximating the aero-
dynmr_ic load_ in the frequency _}omain is presented for wings in two
and three dimensional 11ow.
'£'h_ Pade _._[,_i,

io..:Jmant r_,:
_!_,! ic app]iud t,:->
(:b_,ain !,he so].u%ion
to the. pce_u_'u _li_<trJbuti<,n on wing.- perfc_r';_:lng arbitrary oscillations
in two dimens:ional inc<mlg_-_:_':_ibl__ Ylow. _-'he m<_thod is then generalized
and applJ_.d to two {[imun,_[ real comp.-o:::_[ble flow p_'ob]ems.
[j_cO].pres_ih]e !l_.'. -- Aerody,lamic l_nds _m nrbitrarily
os<i]lating rigid wiI_Hn in tw_--:iimeI_i{,l_a] incc,uq_'<'_sib]<' flow can be syn-
thesized f_'or,_Wagne_' '_ ;:olut]on loz" bhe lJ ft o_l a wJn_ d__tc i,oa sudden step-
wise change in the do_,m_,,a_h as _qeserJbed in Her. i. R. T. Jones [I0]
has obtained an app_o×Jmate Laplace 1:ransfoz_r_ of Wagner's indicial
function. It can b,_ shown that eeuations [or the pres,oure distribution
and aerodyns.n_-i{', loadf' may b,_

_ obtaJrl{= ! for eonw_rging or diverging
oscillat]om: by _._:placJng C(k) by 6(s _6(s) where 9(s) is the
Laplace transform of W_q._n,_r':', indicia! _imct:ion. Thus equations of
motion of a wJng_ with a tr.a.!]ing edge flap [:' ] or a wing with an
a:,leron and _ Ilap _: , ; '] _ay b_ obtained by using the correspond-
ing equ_t:iom_ for o_<c[ilatot.g moi, ior_. ,_"or converging oscillations
the wake is asstm_e.d to b_: finit_ an,J dupendunt on initial conditions.
In this s<:etion_ Theodor_mn'._ uircu!ation Izg function is
an'_lytical.ly e.oniJnu,--ci [7_[- (:orl_,,.!_'iy_!_T '_'_[ ["_, L_t ;_:_ ;iil [o[] w_th no
17
oscillating components in two dimensional incompressible flow, and
the physical significance of the results explained. By representing
Theodorsen's function as a series of Kummer functions, an asymptotic
expression for Wagner's indicial function is obtained. The first two
terms of this approximation are identical to the approximation for
Wagner's function obtained by Garrick. By applying the Pade
approximant theory, it is shown that Theodorsen's function may be
represented by a sequence of rational functions which converge
uniformly. This method of approximating Theodorsen's function is
generalized so one can construct rational function approximations not
only to Theodorsen's function, but to gust response problems also.
These rational function approximations may be easily inverted, using
Laplace inversion_ to obtain good approximations to various indicial
functions •
The lift and moment about the elastic axis on a rigid thin air-
foil_ performing vertical translational or torsional oscillations can
be easily obtained by integrating the pressure and its moments (given
in Appendix B).
These expressions are,
, 2 • • i
Lb = 2wqb 2 --5 ...... f ....... , ..... l---f---5--Z ..... f ........... f .....

[ M ] as s2+ 2s_(s) ' s-as +2$(s)(l+(_-a)s)
+2(a+_)@(s).sls(a-_)-(7_+a )s +(a+_)2._5(s)(l+(_-a)s
×
h/b
C_
18
These equations may be written also as,
E] Lb
Ms
= 2wqb 2 s
as2
s
s(a-i/2)-(1/8
- as
+ a2)s2
+
¢(s)[sl(1+(
_ -
I 2a+ 2 1 1 a)s)] l
Thus the circulatory part of the aerodynamic matrix can be decomposed
into the product of two vectors. As a consequence we have,
2 h
S - as
= 2Trqb2
i a 2 )s2
M_ as 2 s(a - _)-(_+
z _ a)s)]
y(s) = $(s)[s,(Z+( _ [hi
Thus it is sufficient to construct a finite state model of _(s) (shown
in Figures 2 and 3 ) in order to obtain finite state models of the
generalized aerodynamic forces. _n application of this approach to
flutter suppression studies is presented in Ref. 30.
19
In general_ if the influence of comprc_;_ibility and finite span
are included_ it is impossible to deeompo_:e!i_e matrix in the above
manner. In two dimensional ineomp_s:_:ibl _ J<l,_w_

the circulatory aero-
dynamic forces are dependent only on tLe ! _w_m_ash

velocity at one
point (the _,/4 chord point for a :,dgi_I ,_iri_r_l ). In compressible
flow or for three dimensional lifting _ar:iac_.__the aerodynamic forces
are dependent on the downwashat a finit-= _lumberof collocation

stations. Further the contribution of t_he iownwashat each of the
collocation point_ to different generali;,_ I _oYces_is different. Thus
one cannot assumeapriori that a decompo_:iti<_ of the form discussed
above is possible. Thus, the above _Pl)_:,_:'I_ is not useful in com-
pressible flow or for three dimensional lif{,:ing surfaces.

Garrick has shown that Wagner's inflicial function and
Theodorsen's function are related by the Fou,;er transform relation-
ships. Thus,
Kl(t): i [ C(k)-i eil_t ....

,_- for t _> 0
-_ 0 for t < O.
This may be written in the form
Wg(t ,_ Kl(t + T)- i df
C(k)-1
where Cf ( 2_ e
ik
2O
This equation suggests that numerical quadrature may be used to
compute K(t) for T _ t _ 2T for any given T. An equivalent possibility
is the use of discrete Fourier transforms like the fast Fourier trans-
forms [31]. In order to show the disadvantages of this technique a
brief discussion of the method is presented.
To observe the effect of sampling both in the frequency and time
domain, at finite but equal intervals, the above equation is evaluated
at the points,
tj = j.At, j = O, +_ i, +_ 2, ..., +_ (N-I) and
f = n'Af, n = O, + i, + 2, ..., + (N-l)

n _ -
where N = TF, with F = 1/At and T = i/Af. At. is chosen small enough
J
for the highest frequency present to be sampled at least twice during
each cycle. With the assumption that the indicial function and the
spectrum Cf(f) are periodic with period T we have,
N-I )e+2Wi n.j/N

1 _ Cf(n. At
Wg(j.At) : _ n=O
N-I
T E W_(j.At)e -27ri n.j/N
Cf(n'Af) = _ n=O _
- 2_i/N
Let A
n A Cf(n,Af), Xj A TWg(j ,At) and w e
We then have the discrete transform pair,
= N-I Awn_+n-J {xj ] = [fl]{An }

E
Xj n=O
N-I -nj
I Z X.w,
{A n ] = [_]-I{xj}
An= N j= O 0
21
From the orthogonality properties and the properties of the exponential
function it is possible to factor

[_] in such a manner that the number
N
of computations involved for inversions are reduced from N2 to _ log2N
_rovided N = 2m). This decomposition technique is the Fast Fourier trans-
form. Thus it is possible to obtain numerical values for Wg(tj) for
various values of tj by using different sampling periods T to any desired

degree of accuracy. Also errors due to the assumption of periodicity
can be corrected for_using WindowTechniques. Howeverthis does not lead

to a finite state model. From the numerical data generated for the
indicial functions_ Kl(t), it is essential to fit the data with a finite

number of exponentials (2 or 3). This is usually a non-linear technique
[32]. The Laplace transform of such an approximation would then lead to
a finite state model in the frequency domain. This whole procedure is
undoubtedly extremely tedious and involves several computational steps.
Conceptually the procedure involves approximating a distributed parameter
system by a discrete time system which is then approximated again by a
continuous system in the time domain. Thus a direct procedure of finite
state modelling from knownvalues of the response at f = fn' with the

correct initial and steady state behavior is desirable. In order to develop
such a technique_ Theodorsen's function is analyzed in detail.
Analytic continuation of Theo_orsen's function: The function _(s)

will nowbe evaluated in the left half of the 's' plane along the real
axis and the physical significance of the results will be explained. In

this case, _(s) may be written as
22
¢(s)= 1 -
_e-S cosh v cosh v dv
I+Lt
0
e-S cosh v dv
0
The integrands of the two integrals are monotonically increasing in the
domain of v. Further they are both divergent. Hence the limit may be
evaluated by using L'Hospitals rule_ which immediately leads to
#(s) : 1 for Re (s) < O, Im(s) : 0 .
This result is of fundamental importance. It was assumed that at time
t = O, the wing was stationary and unperturbed. Clearly, converging
motion with no oscillatory components from an unperturbed state is not
possible. This implies that on the negative real axis of the 's'
plane the 'motion' remains unperturbed. As _(s) = l, is the solution
for the case of steady translation of the wing, the quasi-steady solution
for the pressure distribution is the exact solution on the negative
real axis of the 's' plane.
Asymptotic expansions of Wagner's indicial function: In terms of
Krummer functions we have
1
_(s) = Re(s) > O
i+½ l, 2s)
s (1½,3, 2s)
The Kummer functions U(a, n, z) are a class of confluent hypergeo-
metric functions that are multiple valued with the principal branch
defined for -v < arg(z) < v. The order of the singularities for large
23
and small values of z are determined by a and n respectively.
As a consequenceof the properties of Kummerfunctions $(s) maybe
represented as an asymptotic series for large s as_
N+4
1
: 1 - (4sU(1,1,4s)
&.
+ E Bi (Cis)U (]. ,1 ,Cis )+ 0 (s)
i=l
N : 0, l, 2, ...,
where B i and C i are constants. Apart from being asymptotic for large
s the expression also satisfies the condition,
T,t : l
s-*0
Further_ the inverse Laplace transform of the asymptotic series can be
easily found. Hence it is also possible to find an asympZotic expression
for Wagner's indicial function. In particular_ for N = 0,
3
2 T
K(t) = 1 T + I_ + 768 'P(_)
4
where P(T) = i + 0.875T + 1.28435T 2 + 1.8428_ 3 + 4.09134T
The first two terms are identical to the indicial function given by
Garrick. In general the correction to Garrick's approximation may be
written as,
2 3 (Polynomial of degree N in r)
K(t) - 1 + _ = _ (Polynomial of degree N+4 in _)
The rational function in the brackets is usually referred to as a
[N + 4, N] Pad_ approximant. Baker [33] has discussed the methods of
24
construction and the properties of Pad_ approximants.
Pad@approximants of Theodorsen's function: For small values of s,

W. P. Jones [34] gives the following series for _(s):
2
_f(s) z i + ys + y2s + s
where y = log(s/2) + _. The correct asymptotic series for small values

of s is given by,
2
Y
_(s) = 1 + ys + y2s2 + s3[ z_l 2 + y3+_ ]+ s4<y4
y4 y2 _)
+ sS(y 5-_y3 4
+_ _ y3
- 19
i---g_2 +liy-_
ll ll
)
+ 4 19.2 77
of:l,
R_(s) >0
Obviously it is quite difficult to construct a Pad_ approximant from
this series. However for large values of s, we have the asymptotic
expansion,
l(i + i
_(s) : _ 1 + 7 19 143 629 8273081
8s 2 64s 3 1282 + 512s 5 +
+ o(-_))
i024s _ 4194304s 7 S
Re (S) > 0
The fact that the lag function has a value of 0.5 for s = _ and i for
s = 0, suggests that it may be approximated by an [N_N] Pad_
25
approximant quite accurately. To demonstrate this, the first four
Pade approximants are constructed from the asymtotic series for large
s. They are asymptotic to the lag function uptothe first, third,

i
fifth and seventh orders in --S respectively. They are given by
s+o.5 s2 + 1.5 s ÷ 0.375

2s ÷ 0.5 '
2s2 + 2.5s + 0.375
s3 ÷ 3.5 s2 ÷ 2.7125s ÷ 0.46875 and

2s 3 + 6.5s 2 ÷ 4.25 s ÷ 0.46875
4
s 4 4.64696s 3 + 9.33371s 2 + 5.51735 s + 0.49334
, Re(s) > 0 ,
2s 4 ÷ 8.79392s 3 + 16.71894s 2 ÷ 7.67296s ÷ 0.49334
These expressions are found to converge rapidly over the entire right
half of the 's' plane. However the above approximations are not very
good for small values of s as they are obtained from the asymptotic
series for large s.
This method of approximating Theodoresen's function and evaluating
the Wagner indicial function approximately is similar in principle to
the method suggested by Luke [35] for evaluating the Randall function
in the theory for the unsteady aerodyn_nics of oscillating cylindrical
shells in supersonic flow.
The method of construction of the Pad_ approximants may be
generalized. Since the lag function is known exactly along the
imaginary axis s = ik, the Pad_ approximant may be constructed by a
suitable least squares technique . If C(k) = N(ik)(D(ik)) -1 , where
N(s) and D(s) are polynomials in s, the coefficients of tne polynomials
may be determined by minimizing
26
m
z (D(ikj)C( j)- 2
j=l
where kj, j = i, 2, 3, ..., m are the values of k at n different points.
This technique needs only the values of C(k) at a finite number of
points in the frequency domain_ and hence may be generalized to other
cases, where the aerodynamic loads can be calculated numerically only.
This problem may then be reduced to the solution of an over determined
linear system for the coefficients of the polynomials N(s) and D(s).
This procedure results in
s4÷ 0.761036s3+ 0.i02058s2+ 0.00255067s + 9.55732 × 10 -6

"
2s4÷ 1.063996s3+ 0.i13928s2+ 0.00261680s + 9.55732 × 10 -6
Re(S) > O.
Clearly Wagner's indicial function may be approximated using the
formula
4 -BiT
K(t) : Z -I _(s)__= 1 - _ A.e
s 1
i=l
where _i are the roots of the denominator polynomial where
Ai _i
0.011285763 0.0044482234
0.043280564 0.027697193
0.21639860 0.096054968
O.229O35O8 0.40379780
27
By taking the Laplace transform of the approximation for Wagner's
function obtained by R. T. Jones [i0], we have
@(s) = _+ 0.5615s + 0.0273 Re(s) > 0

2s2+ 0.6910s + 0.0279 _
The [4,4] Pad_ approximant obtained by least squares and the corres-
ponding indicial function are compared to the exact values and the
results of R. T. Jones (Figures 4 - 8).
This approximation is very good both in the frequency and the time
domain. The simplicity of this method compared to the one given by
R. T. Jones is quite obvious. The main advantage of this method is that
it can be generalized to three dimensional lifting surfaces.
It is also interesting to note that the poles of the six approxi-
mations for the lag function are negative real values. This indicates
that the transient pressure response to a downwash input is
asymptotically stable. Further since the negative real axis is a
branch cut of the exact expression for the lag function it represents
a continuous distribution of poles.
Application to gust response problems: In applying the above
theory to gust response problems, a certain amount of caution is essential.
It must be emphasized that only the Laplace transform of the temporal
portion of the circulatory pressure distribution can be approximated by
a sequence of [N,N] Pad@ approximants. This does not hold for sinu-
sodially convected gusts.
Sears [9] considers a gust velocity of the form,
28
WG = Woeie(t-x/V)
and showsthat lift to be equal to
Wo ic_c
L(k) = 4_ q'b. -_- e S(k)
where S(k) = (Jo(k)- iJl(k))C(k)+ iJl(k ).
Kussner's indicial function is related to S(k)e -ik in the same manner as
Wagner's indicial function is related to c(k). However to obtain an
exponential approximation of Kussner's function one should construct
a [N,N-I ] sequence of Pad6 approximants of S(k)e -ik In fact, using the
least squares smoothing technique developed above and taking the inverse
Laplace transform we obtain the following expression for Kussner's
indicial function K2(t ),
4 -_.T
1
K 2(t) = 1 - Z Aie
i=l
where
Ai Bi
0.012994467 0.0049896174
0.062319920 0.0349314o0
0.40920539 0.13796713
0.51548O22 1.1645811
29
In order to showthe relationship between the temporal and spatial
portions of the gust response, it is essential to consider a gust of
for_n_
i_t - iFx/V
WG = WOe
Kemp [36] shows the lift to be of the form
: Vo b ei_t
where K_,k) = ((Jo(k)- iJl(k))C(_+ _ Jl(k))
and k = _- .
The first term in the function K(k_X) is the circulatory part while
the second is the virtual inertia effect. Clearly it is now sufficient
to approximate C(k) by an [N,N] Pad_ approximant to obtain the indicial
function corresponding to a downwash of the form
W G -- WO e'iFx/V H(t)
where H(t) is a step'function.
A sinusoidally convected chordwise gust past an airfoil at a
steady angle of attack G, of the form, U G = UO elf(t-x/v), is known to
produce a lift equal to [37],
Uo
L(k) = 4_q-_- _be ic_t T(_
where T_) = C_)(J0 _)-iJl(k))+ J0(k)+ i2J l(k).
3O
From this result we may obtain the indicial function for a sharp edged
chordinse gust striking the leading edge of the aerofoil at t = 0
Us _
U X >Vt -b
U0 X <Vt - b
The indicial lift is given by
Uo
L(t) : 4vq -_- _bK3(t )
In
where K3(t ) may be found in the same manner as Kussner's function.
fact K3(t ) is approximately given by
4 -8iw
K3(t ) = 2 - _ A. e
i=l I
i
Ai 8i
1 0.026965564 o.002369o134
0.044268810 0.o4o6473o2
0.45282297 0.14614369
1.4759427 1.73o58o2
Separating the spatial and temporal portions of the gusts into
iwt - iyx/V
Ug= Uoe
31
we may show the lift to be
U0
L : 4vq _- _be i_t V(k,k)
where V(k,k) = C(k)(Jo(k )- iJl(k))+ jo(k)+ iJl(X)+
Sharp edge gusts moving with respect to inertial observers are
consideredj i.e., gusts of the form
wa= 1 0 X > Ut -b
! Wo X < Ut - b
X >Ut -b
and
°
Uo X <Ut -b
where U is different from the free stream velocity V. U0 is assumed to
be very small compared to U so only first order effects need be con-
sldered. The corresponding indicial functions are related to K(k,_k)
e -i_k and -i_k

V(k,_k)e _ where _ = V/U, in the same manner as Kussner's
function is related to S(k)e -ik and hence may be written as,
4 -Bj%
KT = i - jTIA/= e
KC = 2 - FACe J
J=l J
These indicial functions are shown graphically in Figure 9 and Figure
lO.
Before going into the nature of the solution of the pressure distri-
bution in two dimensional compressible flow it may be instructive to
32
review the effects of thickness and viscosity on _(s). Hewson-Browne
[38] has investigated the oscillation of a thick airfoil in an incom-

pressible flow and obtain a general expression for _(s) (which he
evaluates only for simple harmonic motionS) including thickness effects.
The generalization is different in the expressions for lift and moment.
This would destroy the decomposition property of the circulatory portion
of the aerodynamic matrix discussed earlier. The singularity in the

derivative of _(s) continues to exist. Woods[39] has presented a new
approach for the calculation of the unsteady two dimensional flow about
aerofoils performing arbitrary motion in an incompressible fluid. While

the flow is assumedinviscid_ the potential flow boundary conditions
are modified semi-empirically to make someallowance for viscous
effects. The method is applicable to thick airfoils, the only limita-

tion being that the velocities and displacments of the unsteady perturba-
tion about the meanmotion be small. Results are obtained for the lift
and momentby an application of Blasius theorem. Howeverthese results,
while being very impressive mathematically, seemto be practically useless
as most of the integrals cannot be evaluated for general motion. Pro-
vided the Reynolds number is sufficiently large for boundary layer

theory to be applicable, viscosity has three main effects on the
theoretical potential flow. These are_ I) the Kutta condition is
modified in that the position of the rear stagnation point is indepen-
dent of incidence 2) the velocity distribution of the meansteady flow
particularly near the rear trailing edge is modified 9) viscosity

contributes to the damping that is independent of the velocity but does
not effect the inertial loading. Woodsaccounts for i) and 2) but not
3)- The results indicate that the reduction in the wakevelocity due
33
to the thickness and viscous effects, changes the flat plate deriva-
tives by an amountwhich is quite large for k large but neglegible

for small values of k. Chen and Wirtz [40] have included second order
wake effects, but it appears that the singularities they obtain are
no different from those of the first order solution.
It seemsprobable that the nonlinear and viscous effects in the
wake may reduce the wake velocity and distort the wake which may be
significant especially for low values of k and large values of Isl.
So far there is no theoritical justification of this. The above consid-
erations are important if experimental verifications of Wagner's solu-
tion are sought.
Subsonic flow. - In Appendix C, the behavior of the pressure
differential across the airfoil in subsonic flow was shown to be
essentially like incompressible flow for low values of reduced frequency
and mach number. However for large values of reduced frequency, unlike
incompressible flow, the force required to generate impulsive motion may
be expected to remain bounded for all time and finally reach a steady
state value. In particular, the work required to generate impulsive
motion is not entirely recoverable, so that the process is irreversible_
and some part of the starting force must be regarded as "damping". It
is well known, that for large values of reduced frequency the pressure
is given by Piston theory [i] for all mach numbers, M, greater than
ze ro •
_p 4 _z
i.e.,
q MTY
34
where z is the displacement of the airfoil.
Another curious fact, that has not received the attention of any
previous investigators in unsteady aerodynamics_ is that the steady state

aerodynamic load on an airfoil in steady subsonic flow or supersonic
flow is always finite as long as the mode shapes are bounded. This
indicates that the aerodynamic system by itself is a completely stable
system_ in subsonic or supersonic flow_ in spite of being a circulatory
system. This physical fact fully explains the reason for all the poles,
of each one of the PEui6 approximants of $(s)_ being in the left half of
the 's' plane. The stability problem is especially significant in
transonic shock free flows and hence is discussed in greater detail in
the next section.
In Appendix B_ it was shown that the integral equation for the
pressure distribution can be reduced to a Fredholm integral equation of
the second kind with a completely continuous kernel. The significance
of complete continuity will be discussed in 3.2. In Appendix D_ it was
shown that for an equation of this type _ the solution has the form of
a [N_ N+J] Pade approximant. Clearly_ comparing this result, with the
Piston theory result for large s _ we conclude that the loads must be of
the form,
aosml+ al ""
qij (s,M) = it N-I
N-._ MsN+ b2s ... bN+ I
where the coefficients a i and b i are functions of Mach number alone.
Clearly as s _ % qij = a_/M and a0 can be obtained from piston
0 (M) and
theory. In the limit as s -_ Oj qij (O,M) = aN+l/bN+ 1 = qij
equals the steady state value. As the mach number approaches zero,
35
one of the poles of the denominator moves to negative infinity in the
's' plane. This clearly indicates that in incompressible flow an impul-
sive force is needed to generate impulsive motion. Hence for large values
of s_ qij behaves like an inertial force in incompressible flow. In

compressible flow, unlike incompressible flow, the coefficients bi are
different for different types of loads and modeshapes.
This difference in behavior may be easily explained physically.
In incompressible flow there is a lag in the development of the circulation
around the aerofoil, while the pressure variations are propogated with an
infinite speed. In compressible flow in addition to this lag the
pressure variations caused by the motion of the aerofoil are propogated
by a finite speed. This introduces an additional lag in the loads.
On the other hand as the speed of the airfoil increases, it is clear
that the pressure variations move upstream more and more slowly in
relation to the aerofoil. If the airfoil movesfaster than the speed

of sound_ it cannot make its presence known to the fluid ahead of it and
the mechanismfor bending the streamlines ahead of the airfoil no longer
exists. Eventually for very high speeds, the influence of the pressure
variations is only local. At this stage the a_rfoil behaves like a one
dimensional piston and piston theory leads to the exact pressure on the
aerofoil for all values of s.
Thus
aOs
qij(s,M)=.. _ for M _ _
It was also shown in Appendix D that the sequence of rational
functions for increasing values of N converge and hence can be
truncated and assumed to be of the form
36
aoml + alsN + %{ + qO (M)bml
"'" iJ (3 .i)
qij (s ,M) = N N-I
Ms + b2s ... + bN+ I
for some fixed N.
Further, in view of the aerodynamic system being completely stable,
we may conclude that all the poles of qij (s,M) lie in the left half
of the 's' plane. This is an important and useful property, as it can
check on the validity of the rational function approximants after they
have been constructed.
The exact solution for subsonic flows, can be obtained by Mathieu
functions [1]. However the actual computation of the solutions is ex-
tremely tedious. Furthermore, the solutions are not closed form
solutions and hence are only approximate. Thus it is difficult to
obtain series expansions in frequency of the aerodynamic loads, in order
to analytically construct the Pad_ approximants.
Hence a numerical technique used often in the area of system iden-
tification and for digital filter synthesis [41] is proposed and
applied to various cases. The problem of computing the approximants
is reduced to a least squares minimization problem. This is then
transformed to a overdetermined linear system for the coefficients of
the Pad_ approximants and the solution obtained by QR decomposition.
Now qij (s ,M) may be easily calculated by the numerical method out-
lined in Appendix C for a given mach number, and for various real values
of reduced frequency.
k = k. = -is m = i, 2, 3, ..., m 0
m m
37
Also aO can be obtained from :piston theory and qijO (M) from steady state
calculations. Thus the above equation may be written as_
-2 + ''"
alSmI+ Sma2 -N - qij(Sm'M)(b2Sm
SmaN _o
_-+ "'" bNSmN)
- (N+I)
+ sm (qOj(M)- qij (Sm'M))bN+l = smlMqij (Sm'M)" ao
m = i, 2, 5, ".. m0 (3.2)
Hence these equations are a system of m0 complex equations in the 2N real
unknowns al_ a2, ... a N and b2, b3 ... bN+ I. If mo, is chosen larger
than N_ this system of equations may be solved by the linear least
squares technique.
Numerical experience indicates that a large number of values must
be chosen fork, in the low and high frequency regions. In the low
frequency regions the coefficient of bN+ 1 could be a small complex
quantity while at high frequencies the right side of equation (3.2)
would be a small complex quantity. Hence a large number of values
must be chosen for km, in the low and high frequency regions, in order
that the approximation be sufficiently accurate in these frequency
regions.
These approximants for nondimensional generalized forces in (1)
plunging, (2) pitching and (3) flapping mode and tabulated in Tables
1 - 5. The flap hinge is assumed to be at the rear quarter chord point.
In most of these cases N was chosen equal to 2. In these cases it was
38
found that more accurate results could be obtained with N = 4. However
these results are not tabulated.
It was assumedthat it is possible to calculate the aerodynamic
loads qij (s ,M) accurately for all values of k and M in the subsonic
regime. This assumption is generally true in two dimensional subsonic
flow, provided a fairly large number of collocation and integration
points are used in the method described in Appendix C. The method is
very accurate even in the presence of control surfaces, provided
collocation points are not chosen too close to the hinge line.
The influence of compressibility which is well known in steady
flow is different for the steady and unsteady components of the loads.
For low mach numbers (0.i < M < 0.3) it was found that higher order
m
Pad_ approximants were required to model the pitching moments, where
the poles of the approximants have a tendency to move towards the
imaginary axis of the s plane. This could be due to the higher order
of the approximants without any physical significance.
For high subsonic mach numbers (0.7 _ M _ 0.9) the loads were less
stable even with lower order Pade approximants. On the other hand it
is well known [42, 43 and Section 9.2, l] that oscillations of a two
dimensional aerofoil about a pitching axis may become unstable for low
values of reduced frequency and high mach numbers_ if the axis lies in
a certain region, ahead of the quarter chord axis and if the moment
of inertia about this axis is sufficiently large. This has been
attributed to the fact that the logarithmic term in frequency in the
kernel function is stronger at high mach numbers for two-dimensional
aerofoils than for three dimensional lifting surfaces.
By taking the Lapalce inverse transform of qij (s,M)/s_ it is
possible to obtain the indicial response in the i th mode due to a tin_e-
39
.th
wise step change in displacement of the 3 modeat time t = O. These
results are shown in Figures ii to 19, and have been obtained from
[4, 9] Pad_ approximants in all cases except in the case of qll at M = 0.7.
From these figures it is quite clear the loads corresponding to the same
pressure mode have the same general behavior in the time domain. Also
the loads corresponding to the flap mode take much longer to reach the
steady state value_ than the loads due to plunging and pitching.
The influence of compressibility on Wagner's indicial function may
be studied. In this case, the lift qll' due to a rigid plunging mode is
calculated at several Mach numbers and the corresponding indicial
function is obtained by taking the Laplace inverse of qll/S 2. This
result is shown in Figure 20. These curves are compared with that of
Mazelsky and Drischler [15] for Mach number = 0.5, in which case the
error was found to be the highest.
Supersonic and transonic flow. - In supersonic flow the
expressions for the lift, moment and partial moment for an aerofoil
with a flap are given in [44] and for an aerofoil with both trailing
and leading edge flaps in [45]. The Pads approximants for the exact
expressions for the lift and moment of a rigid aerofoil in plunging and
pitching may be obtained. The lift coefficient for an airfoil in pure
plunging motion is,
Lh . 2+ h
--= 4( s as+b )V or
qb Ms2+ cs + 8b
s+ 2 h
+ 83 ) V
4O
whe re
a- 3MB
[SMB + 3 + 3M3(_-M)- M2_J_ A = M2(TM 2- 3)
b : -_3_ (_2!B2(__M)+ 3_2M) 31B2M2 ((3M2+ i)(_-M)+ 4M)

c= 7
The moment coefficient in pure plunging motion is,
Mh als h -5_
_ 4( ) V ' where aI =
qb 2 Ms2+ ClS + bI 3M (IIM 2- 3)
bI - 564 , and cI - IOM62

M (IIM2-3) IIM 2- 3
The lift coefficient in pitching is given by,
L a2s + b2 M 3M 2- 2 ), b2 _ 3M_
qb - 4( Ms + 6b 2 )G ' where a2 = 5( 3M2+ I 3M2+ i
The moment coefficient in pitching is given by,
M s2+ a.Ss M2 2 3(2 - M2)M_2+ 3_ 5
qbC_2- 4( 3Ms + b3 )C_, a_._ - _2- b3' ]03 = 2M2
All four approximants have the correct asymptotic behavior in the
limit s _ 0 and the piston theory limit was enforced for s = _. Also
the poles of all four of the approximants satisfy the stability cri-
terion discussed in the last section. The approximants are good
41
approximations for the loads for M > 2.0. Though not extremely accur-
ate these approximations are certainly better than piston theory or
quasi-steady theory.
It was observed that for higher order Pade approximants, the poles
behaved in an erratic manner especially for low supersonic mach numbers,
moving often into the right half of the 's' plane. In transonic flow,
it is well known from the exact solution for the pressure [46], that
the linear aerodynamic theory leads to infinite aerodynamic loads in
the steady state as the loads have a pole of the form _. This was
probably the reason for one of the poles of the approximants moving
towards the origin for low mach numbers in the supersonic regime and
high machnumbers ( > 0.8) in the subsonic regime. For the reason
indicated above it seemsappropriate that in transonic flow_ the
aerodynamic loads may be written as a [N,N+2] Pade approximant in
with a pole at v_ = O. However this result has no practical value, as the
linear theory is known to be invalid in steady transonic flow.
The physical interpretation of this result is that two dimensional
shock-free transonic flows are unstable on the basis of linear theory.
One of the arguments against the physical existence of shock-free
transonic flows was based on the supposed instability of these flows
with respect to upstream moving disturbance waves [47]. The argument
was that these waves when superposed on a steady shock-free basic flow
could move upstream as long as the local steady flow speed was subsonic,
but as they entered a region of supersonic flow they would come to a
standstill and coalesce to form a steady shock wave.
42
HoweverPearcey [48] showedusing laborious experimental tech-
niques, that the disturbance wavesmoving upstream can penetrate the
supersonic region, and that in fact, shock-free transonic flow
could be experimentally realized. This is attributed to the non-

linear turning effect of the wavesmoving upstream into the supersonic
region, which was not theoretically accounted for earlier [49]. Then
based on geometrical acoustics one may show that disturbance waves
moving upstream can penetrate the supersonic region. The problem of

shock-free transonic flow is too complex at present for a strict
mathematical treatment.
The numerical technique developed in the last section is also

applicable to supersonic flow. The results are tabulated as in the
subsonic case for N = 2 in Tables 6 - 8. It may be noted that the
derivatives of the indicial response functions to time-wise step
changes in the downwashin plunging and pitching [12] are not continuous,
since the period of influence of an impulsive source is finite. The
above technique approximates these derivatives by continuous exponential

functions (Figure 21). Thus higher order Pad6 approximants are
essential to approximate the derivatives of the above indicial

functions.
It was found that for Machnumbers less than 1.5 second order Pad_
approximants were insufficient for modelling the loads. The possible
reason for this has already been discussed. On the other hand for
Mach numbers greater than 2, a first order Pad6 approximant was sufficient
43
for the lift and momentin plunging and pitching in the frequency domain.
This is to be expected as for a wide range of frequencies. In fact
Piston theory may be considered as a zeroth order Pad_ approximant
for high machnumbers.
Three Dimensional Lifting Surfaces
In this section a systematic method of obtaining Pade approximants
of the aerodynamic pressure loads is presented. Though the method is
actually quite difficult to apply in practice_ it will prove the
existence of Pade approximants and the form of the solution. The method
is also applicable for two dimensional airfoi]s.
First it must be noted that Piston theory is also valid for three
dimensional lifting surfaces for high frequencies and that any solution
for the pressure distribution must converge to this limit for high
frequencies.
In subsonic flow for two-dimensional airfoils, it is well known
that the non-dimensional pressure _(_,_,ik) may be written as_
For three-dimensional lifting surfaces it is given by
_p = E p}on (ik)n+ (ik)21og ik p_(ik)n (3- 4)

n=O
n n
where PSO and Pgl are independent of f_equency. Assuming the downwash
to be WOO + ik wOl and using the kernel function expansion in Appendix
44
B, for low frequency (equation B18), we have
8 (wlO÷
ikWol) co
V Z an(ik)n+(ik)21og ik bn(ik)n]_(_,_,ik)d_dw_
n=O
Hence we have the infinite system of equations,
Woo 1 j,j. o
-9- = -_ AaOP30
wOl 1
-_- = - _-_ _;(alp300 + aOp310)d$ dq
0 : (a2P + alP30 + aOP )d_ dq
The first equation in this infinite set, involves the calculation of the
pressure distribution by application of steady state lifting surface
theory. After having determined the steady state pressure the second
term is also determined from steady state lifting surface theory since
O 1
P30 is known and P30 occurs in combination with the steady state kernel
a O. Thus the coefficients may be obtained numerically by a procedure
llke the kernel function technique. In this manner it is possible to
i i
obtain all the coefficients P30 and P31" One special case of importance
0 0
is when WOO = O. In this case P30 = 0 and P31 = O. Hence in this case,
1 2 k2 O(k3i + O(k31og ik)

_P = P30 ik - P30 +
45
Thus the existence of a solution of the form given by (3.2) is estab-
lished.
Next we note that
QO
s log(so(S/So))
:s (-i)
i+l + s log sO O < s < 2s O-
i=l
From (3.2) we have
J
(k _ i)
ko
_p= Z psn0(ik)n+ [ik 2 (-1) j+l P31 (ik)n+(ik)21°g iko ]
n=O J n=l
This can be rearranged in the form,
2N-2
_pp = E C (ik) n (3 5)
n=o n
The coefficients Cn are functions of the number of terms included
in the series for the logarithm. However the coefficients converge.
Further the first two coefficients are unaffected. We may now construct
a [N, N+l] Pad_ approximant such that Piston theory is satisfied in
the limit s _. The coefficients of [N, N+l] Pade approximants may be
obtained in terms of C n by solving a linear system. Thus one may con-
struct a sequence of [N, N+l] Pade approximants. These would have the
same asymptotic behavior as equation (3.5) for small values of frequency
and the correct limit for high frequency.
It is now essential to establish the convergence of this sequence of

p
Pade approximants.
46
In Appendix D, a similar problem was considered. First the Neumann
series solution of the problem was established. It was then shown that
convergence of the Pade approximants is assured if the kernel function is
a completely continuous kernel.
Symbollically we may denote the lifting surface problem as_
w = Lp = [L0 + ikL 1 + (ik)2L 2 ... + (ik)21og ik(M0+ ik M1 ...)]p
where Li and Mi are integral operators. The solution to the steady state
problem is
Psteady = LO1 w
The iterative method described can be symbolically written as_
p : T,olW
- 'ol[ikLl+(ik)%,2+
...+(ik)21og(ik)(MO
+ ik ...)
]p
The above is a Fredholm integral equation of the second kind of the type
considered in Appendix D. Complete continuity of the kernel function in
the above equation can be assumed in view of the uniform numerical con-
vergence of the kernel function method. As explained in Appendix D the
condition of boundedness of the kernel function is not related to the
condition of boundedness of the operator or complete continuity.
Complete continuity ensures that the integral equation may be reduced
to an algebraic system and that the solution converges uniformly. If
47
it is so, L may be thought of as a matrix and
[ADJ(L)1
p = L-lw_
det (L)- w
If the elements of L are rational functions or polynomials of a parameter
[A_J (L) ]
s, then elements of Det(L) are ratios of polynomials. It is from this
fact that we are able to conclude that there exists a sequence of Pade
approximants if the kernel function is completely continuous. Hence
the theory developed in Appendix D may be applied to the above system
and it could be concluded that there exists a sequence of [N,N+I] Pad6
approximants that converge to the exact pressure distribution as N _ _.
The same conclusions hold for the aerodynamic loads.
Thus we may write qij(s,M) approximately as,
N+I N
aoS + als ... aN+ 1
qij (s'M) "
MsN+ b2 sN-I ... bN+ I
In compressible flow a0 may be calculated exactly from piston theory.
In the case of M = O, we can easily verify from the equation that the
loads are given by a sequence of IN,N+2] Pads approximants. In the
limit M 40 and s _ _ we have the integral equation for virtual inertia
given by
s
2 z(x,y)
b - _
l J'_A d_ 'rl
,s)(X2 +
l
y2)9/2 )
48
This is a self adjoint integral equation and can be solved by a kernel
function technique. Unlike compressible flow the pressure distribution
has no singularities along the boundaries. It is important to realize
that virtual inertia and Piston theory loads are independent of the
direction of flight. However as circulation begins to develop the Kutta
condition is of importance and the classical theory of motion of a body
through an ideal fluid is not valid. W. P. Jones [50] has proposed a
simple method for solving the above equation for rectangular wings. Two
dimensional broadside configurations such as circles and ellipse with
symmetrically placed fins have been treated by Bryson [51, 52, 55] with a
view to analyzing the stabilizing effect of control surfaces on aircraft.
For a circular wing of radius ro, in incompressible flow it could

easily be shown (Appendix C) using the sametechnique as van Spiegel's
[54] solution for low frequencies, that for large frequencies
Cp(r,O) = O(k2)
The actual computation of the Pads approximants, though possible
in principle, was not carried out. Exact Pads approximants could also
be computed in compressible flow. However the amount of computations
involved to get meaningful results is enormous_ as it is essential to
expand several special functions and their products in power series.
Also circular and elliptic wings are not used often in practice. The
numerical technique outlined in the previous sections with some modifi-
cations for three dimensional lifting surfaces is more useful.
For three dimensional lifting surfaces the aerodynamic loads may
be written as
49
0 1
qij(s_M) = qij(M) + sqij(M)
_i N 2
CoS ÷ ClS + ... CN_ 1 s
÷ N N-I
s ÷ b2s ''" bN+ 1
0 (M) ÷ I (M) are the exact loads in the limit s -_0 and
where qij sqij
0
Co = ao - qijl(M)'M. qijO(M) and qijl(M) may be calculated from P30
i
and P30 "
If [A0 + ikA I ] is the influence coefficient matrix for small values
of k and [wj} = {wj0 i ikWjl}, the downwash vector, the pressure
coefficients for small k are given by
PJ = [AO]-l[{wj}_ ik AI[A 0]-lwj0 ]
[AO] may be calculated exactly as in steady flow. [AI] is calculated
from the coefficient of ik in the expansion for the kernel function
given by equation BI$, using either the doublet-lattice or kernel function
technique. Integrating the pressure with appropriate displacement mode
o(M) and I(M).

shapes leads to qij qij
The numerical evaluation of the coefficients b i and c i in compres-
sible flow is similar to the two dimensional case. In incompressible
flow since the numerical technique for the calculation of the virtual
inertia loads for arbitrarily shaped wings was not completed, a 0 was
assumed to be unknown and b 2 euqal to unity.
5O
After some numerical experiment N was chosen to be equal to 2.
This leads to a first order rational function in incompressible flow.
A simple lag is sufficient to model the circulatory effects in this
case. In compressible flow two lags are essential_ for large aspect
ratio wings at low m_ch numbers, one to account for the lag due to
compressibility effects at high frequencies and the other to model the
lag due to circulatory effects at low frequencies. For moderately low
aspect ratios at high subsonic mach numbers the lag due to compressibility
effects may be modelled by a single lag. For low aspect ratio wings
(AR < i) there exists no physical reason for aerodynamic lag and hence
no lags are required at all Mach numbers. Some examples are presented.
(I) Circular Planforms: A circular wing is considered in
incompressible flow. For circular wings the exact results of Van Spiegel
and Benthem and Wouters [55] for simple harmonic motion in plunging_
pitching_ chordwise bending and spanwise bending modes are used.
Rational function approximations to these loads are given in Table (9).
The error in the virtual inertia lift in plunging is about 30% • This
high value is due to the lack of any data points for k > I. The growth
of lift and moment for a timewise step change in the downwash in the
plunging and the spanwise bending modes are shown in Figures 22 to 25.
No other results were available for comparisons. The corresponding indicial
functions and the indicial functions for the lift and moment due to a time
wise step change in the displacement in different modes are tabulated in
Tables i0 and 12. The lift and moment in the spanwise bending mode
reached the steady state value much faster than the chordwise bending
mode. This is to be expected as in the first case_ the major contribution
51
to the load is from the pressure at the wing tips where it does not change
aprreciably with time.
(2) Elliptic Planform: An elliptic wing, AR = 6, is considered
in incompressible flow. R. T. Jones considered elliptic wings of aspect
ratio 6 and 3 and computed the indicial response for the lift due to
plunging. In the limiting case for high frequencies, he obtains approx-
imately,
qll(S) = _( l _ )[ (st + s(l+O( ))+o(sO))

S(71-(_) _)
where E(') is the complete elliptic function of the second kind and
represents the ratio of the semi-perimenter to the span. The first
term represents virtual inertia. It may be noted that it is independent
of the direction of flight and hence is minimum for circular planforms.
For low reduced frequencies, we may compute the asymptotic value
of qll by the iterative technique described, using the kernel function
method to solve the integral equations. Fifteen spanwise integration
points, three chordwise collocation points and ten chordwise integra-
tion points were used for the calculation. Thus for small s we have,
2 2 ._)
11 (s) : sq11
i + s ql I + O(s + O(sPlogs)
qll
1 is identically equal to the stability derivative C_ may be
approximated by the formula,
C_ :: _:_( (13AR)2 + 1.7D72(13AR) )

c_ (BAR)2 + _.7680(15AR) + 6.94_ii
The corresponding formula for rectangular wings is,
52
2_ (BAR) 2 + 0.7881 (BAR) )
C_c : -_- ( (BAR)2 + 3"5760(_ AR) + 3.1526
Both the above formulae have the correct limiting solutions for AR -_
and O.
C_.
2(: v .
qll 2_b ) is also computed for elliptic and rectangular wings
for increasing values of AR and shown in Figure 26. The curves indi-
cate a maxima for circular and square lifting surfaces. For lower
aspect ratios, low aspect ratio wing theory is a good approximation.
i 2
For a circular wing q II and q Ii may be obtained from the exact cal-
culations of Van Spiegel. The errors in the numerically calculated
values, using the method described above, were found to be 0.16_ and
0.9% respectively.
From these asymptotic values it is possible to construct exact
[i, 3] Pad_ approximants. Taking the Lapalce inverse transform we
obtain the indicial response to a impulsive displacement. These results
are compared to those R. T. Jones in Figure (27). Though the difference
in the two results is not too large for high frequencies (') it is con-
siderable in steady flow.
Next the lift due to an elliptic wing_ performing simple harmonic
oscillations in the plunging mode was calculated at ten frequency points
using Laschka's collocation technique with fifteen spanwise integration
points, eleven chordwise integration points and three chordwise colloca-
tion points. The least squares technique for computing Pad_ approximants
described earlier was used and the indicial function was computed for an
A2 = 6 lifting surface. This is compared to the results of R. T. Jones
53
in Figure 28. This result comparesvery well with the corresponding
indicial function, obtained by using only th_ asymptotic values. This
is quite significant for as it clearly indicates that with only the
calculation of the asymptotes and the aerodyn_nic loads for one
frequency of oscillation, it is possible to predict aerodynamic loads
for all types of arbitrary motion.
(3) Rectangular Planform: Rectangular planforms of aspect ratio
4 and 6, in incompressible as well as compressible flow, are considered.
The aerodynamic loads for simple harmonic motion are obtained using
Laschka's method with fifteen spanwise collocation points, three chord-
wise collocation points and eleven chordwise integration points. The

optimumnumber of chordwise integration points is the integer value of
[n(m + 0.5)], n = i, 3, 5, 7, ..., m = numberof chordwise collocation
points. It has been shownby Rowe [24] that not using the optimum
number of integration points could lead to considerable amount of
oscillations in the predicted values of the loads as comparedto the exact

values. The numerical technique is used to calculate Pade approximants
and indicial functions. The indicial function_ for plunging are compared
with those of W. P. Jones [Ii] for AR = 6.0 and 4.0 in incompressible

flow (figures 29, 90, and 31, Table ii). In [56] the result for AR = 6.0
is comparedto the one obtained by a potential flow technique. The fact
that the virtual inertia for an AR = 6 wing is slightly lower than the
case AR = 4.0 cannot be attributed to any physical reason. In fact it
can be argued that the virtual inertia must be a minimumfor a square

lifting surface (AR= i) and have the samevalue as a two dimensional
airfoil AR = 0.
54
Whencompressibility effects are included (Figure 32) the influ-
ence of finite aspect ratio is to reduce the steady state lift while
the starting value given by Piston theory is unaffected. Sweeping the
lifting surface backwards, with the plauform area a constant, further
reduces the steady state lift without effecting the starting value.
In Figure (33) and Table 13 is shown a comparison of the
generalized aerodynamic loads due to plunging (1) and pitching (2) for an
aspect ratio 6 wing. This indicates that the basic behavior of the loads
due to the pressure in any one of the modes is about the same. Thus
this raises the possibility of approximating all the loads due to the
pressure in any one mode, by the same combination of exponentials
with different magnitudes. Also one may conclude that the exponential
rise times in all the loads due to the pressure in the same mode are
interelated. This fact led to a simplified scheme of approximating
the aerodynamic load matrices for purposes of flutter analysis. This
scheme is briefly discussed in the next chapter.
In Figure (34) the influence of compressibility on the indicial
response in the plunging mode for a finite aspect ratio wing (AR = 6.0)
is shown. This again is similar to its two dimensional counterpart
shown earlier. The steady state lift is considerably reduced due to
induction effects.
The indicial rolling moment for an impulsively rolling wing is
shown (Figure 35) for a straight and swept wing of AR = 6. Here the
influence of sweep is to reduce the lift uniformly for all value of
time •
(4) Tapered Planform: A tapered planform of AR = 5.84 and a
"taper ratio of 0.524 was considered. The indicial lift for plunging
55
was comparedwith that of W. P. Jones in Figure 36, and indicates
good agreement. The influence of taper is to raise the magnitude of
the indicial lift quite uniformly for all time. This is to be expected
as the pressure falls to zero, near the wing tips and hence does not
contribute much to the total lift. Hence the influence of tapering
with AR and planform area fixed would be to increase the total lift,
both in steady and unsteady flow.
(5) Swept Wing (A = 45 ° ) with a control surface (AR = 4.16):
This wing (Figure 37) was studied extensively using the improved
version of the doublet-lattice method described in the appendix to
calculate the aerodynamic loads on the wing for simple harmonic
oscillations.
The wing was modelled as a beam-rod and its mode shapes (first
bending and first torsion) were calculated by curve fitting discrete
mode shapes (Table l_). These two polynomial mode shapes and a rigid
control surface mode were used to calculate the aerodynamic loads
on the wing.
First the convergence of the doublet lattice method in the limits
of small and large reduced frequencies was investigated. For low
reduced frequencies, the method was applied with l) 6 chordwise boxes
(NCB) and _ spanwise boxes (NSB) 2) NCB = 6_ NSB = lO and 3) NCB = 9,
NSB = 5 (Table 19).
This indicated that increasing the number of spanwise boxes
(from 9 to 7, 8, and I0) resulted in a converging result. However
changing the number of chordwise boxes did not alter the results signi-
ficantly.
56
To veri_y the usefulness of the box integration schemeat high
frequencies it was assumedthat the pressure at the sending point
(quarter chord point) of each box is given by 4/M times the downwash
at the receiving point (three quarter chord point) of each box. In

this manner the generalized forces are computedin each of the three
cases given above and comparedwith the exact piston theory results.
This did indicate convergence with increasing number of chordwise
boxes. These results are shownin Table 16.
Finally the method was used to calculate the loads at a number of
frequencies starting from small values of frequencies to high reduced
frequencies (up to 5 based on semi-chord).
The generalized forces due to the pressure in the bending and
control surface modes did converge slowly to the piston theory result.
However. the generalized forces _ associated with the torsion mode did
not converge to the results of piston theory. Hence using the results
of the first case and the numerical method for calculating Pade
approximants, it was found that the poles of the approximants were
unstable in the loads associated with the torsion mode. In the sec-
Ond case (Table 17) the results did improve considerably at low fre-
quencies hut not at high frequencies. It was not possible to con-
struct a stable first order Pade approximant for the generalized
force in the second mode due to the pressure in the first. Also the
rise times for the loads (Figures 38 - 46) associated with the torsion
mode were small when compared to the first bending and control surface
modes. The partial moment on the partial span flap has the same
behavior as the corresponding indicial function for two dimensional
57
aerofoils. Also since the results for simple harmonic motion did
converge to the piston theory limit in this case the indicial function
seems quite accurate. However there is no check on the accuracy of
the results as no comparisons are available.
From squations (3.3) and (3.4) we may conclude that the influence
of finite aspect ratio would be to push the poles of the approximants
further into the left half of the 's' plane. For this reason also,
it is important to calculate the aerodynamic loads for higher reduced
frequencies in order to get accurate representations of the loads
than would be necessary in the two dimensional case. An efficieint
and fast method for calculating unsteady aerodynamic loads at all
frequencies is therefore essential.
On the other hand for aeroelastic purposes _ it may not be necessary
to enforce the piston theory limit for low aspect ratio wings. How-
ever extensive flutter calculations are essential for any conclusive
results.
The doublet lattice method does not converge uniformly for all
frequencies. The reason for this appears to be the lattice integra-
tion scheme. Hence the doublet lattice method is not too useful for
finite state modelling of aeroelastic systems.
58
SUGGESTIONS FOR FUTURE RESEARCH
Generally for purposes of flutter analysis, it is sufficient
to calculate aerodynamic loads accurately at the frequency and
flight velocity at which flutter begins to occur. However for
designing control systems for the suppression of flutter, it is
not only essential to predict the root locus in the neighborhood
of the flutter frequency, with respect to the flight velocity,
but also some of the stable roots of the transcendental characteristic
equation which influence the unstable locus. Thus aerodynamic
models used for flutter suppression studies must be fairly accurate
over a range of frequencies. Numerical techniques must converge
uniformly over a large bandwidth. Collocation techniques discussed
earlier for the calculation of aerodynamic forces generally satisfy
these convergence requirements.
The aerodynamic modeling theory presented in this report
suggests the possibility of representing the aerodynamic load matrices,
Qx' in a certain modal coordinate system X, as
-I
Qx = {qxij } = {Pxij(s)/rxj (s)} = PX RX
where R is diagonal. Transforming to the actual modal coordinate

X
system Z = TX
Q = T-IQx T = T-IPxT [T-IRx T ]-I = P(s)R(s)-!
59
where R(s) is not diagonal. Hence in general
N
n N+l
P(s) = Z Pn s + PN+I s (4.1)
n_O
and
N-1
R(s)= z R n + N
n
n=O
The interpretation of this approximation is based on the fact that the
loads due to the pressure in a_mode may be approximated by the same
combination of exponentials with different magnitudes. If M = O,
this reduces to a [N + l_ N - l] matrix Pade approximant. The
coefficients Pn' Rn may be obtained numerically using a combination of
Lagrangian interpolation and least squares technique. This has already
been verified for two-dimensional aerofoils and in certain s_ecial cases in
three dimensional wings • In these cases _ however, the Piston theory limit was
not enforced and N = 0 and N = i were sufficient in the range of frequencies for
which the phenomenon of flutter becomes a possibility.
The above approximation of the aerodynamic load matrix has
several advantages over the method of approximation already discussed.
The matrlx approximants lead to only Mn'N (Mn(N-I) in compressible
flow) states, in addition to states required to define the structural
system, where M n is the number of vibration modes considered for
modelling the structure. Approximating every individual element of
the aerodynamic matrix by a Pad_ approximant leads to M n 2"N (Mn2(N-I) in
incompressible flow) additional states.
The equations of motion of the entire system can be written as
6O
_il + 5ql - _vP(S'M)R-l(s'M)ql
: BUc (_..2)
8v = q Awb
The equation reduces to,
[MIR
(s,M)--_
s% 2 + 5R(s,M)-BP(s,M)]R-_s,M)q (_.3)
I : BUc
The flutter speed is determined by
det[MiR(s,M ) _2 2 + KIR(S,M )- 8vP(S,M)] = 0

(_.4)
The equations of motion can be reduced to the form
X= FX+GU
Zm=HX
where X is the vector of states {y]{#] etc. and U are the control
torques on the aerodynamic control surfaces. There are several
criteriawhen such a realization would lead to a system which is con-
trollable and observable [57]. These are generalizations of the
criteria for scalar transfer functions. The survey article by
Silverman [58] presents an excellent description of the realization
problem,
61
The model is useful for synthesizing control laws for active
flutter suppression. Control laws maybe written in terms of the
state vector [X] as
U= -CX
may be obtained by arbitrary pole placement. The characteris-
tic polynomial of the open loop system is assumed to be
N
n n-i
X(F) = s + Z ais
i=l
and that of the desired closed loop system,
N
A
n-i
(F - Gc) : sn + E pi s
i=l
The coefficients a. are related to F by the relations,

I
k
1 r a. Sk+l_j , k = i, 2, 3_ ... N
ak-- _ j:o J
where S$ = trace (F_) and a0 = i. Similar relationships hold for

#%
(F- GC) and Pi" Using the notation,
T A
P = [PI' P2' "'" Pn ]
T A
a = [al, a2, ... an]
and
62
i 0 0 ... 0 0
0 ... 0 0
h 1
A
I 0
an_ 1 an_ 2
an an_ 1
it is possible to obtain a relationship between the control gains C and
[p]-[a]. It may be tinted, that A -1 always exists and can be evaluated
easily• In [99] it is shown that provided F is cyclic (if not an

,% *k
initial gain Ci may be chosen such that [F - GC i] is cyclic)
trace GC
*%
trace FGC -i
:A (p-a)
• •,
trace Fn-lGc
This is a linear system of equations in terms of the unknowns in C and
may be easily solved. Preliminary results are extremely encouraging.
A detailed discussion of the modeling procedure and numerical results
is beyond the scope of this paper. A simple scheme for control
law synthesis is shown in Figure _7. This technique is outlined by
Lyons, et al. [60]. Optimal laws based on quadratic synthesis may be
obtained by theaigenvector decomposition technique as applied in
Ref. 61. An offshoot of this technique would be a new method
of flutter analysis which would be useful in structural optimization [62].
63
CONCLUDING REMARKS
A general theory of finite state modeling of the aerodynamic
loads on oscillating airfoils and lifting surfaces has been systema-
tically developed and presented. In particular, the nature of the
behavior of the unsteady aerodynamic loads in the frequency domain
is now well understood. The analytical reasons for the difference in
the behavior of these loads at high frequencies in incompressible and
compressible flows are explained. Extensive studies need to be
carried out for supersonic lifting surfaces, where it is expected that
higher order Pad_ approximants would be necessary for an accurate
prediction of the aerodynamic loads for arbitrary motion. The theory
presented has several useful applications, especially in the control of
the vibration modes of aircraft wing structures and in the active
suppression of aeroelastic instabilities.
This study suggests the possibility of approximating aerodynamic
loads as Pad6 approximants in Mach number as well. However, in the
transonic regime nonlinearities need to be considered to account for
shock waves.
Several other suggestions for improving the results and for
applying the theory have also been presented.
64
APPENDIX
A
INFLUENCE
COEFFICIENTS
FORA BEAM-ROD
WITHA FLEXIBLEROOT
The influence coefficients for bending of a beamof length _ are
given by (the deflection at point x due to a unit load at point _)
Czz(X,_)
_-]
0 (_-_)(x-_)
EI(X) dk + _ + _ 1 , for g > x _> 0
_-_ (_-X)(x-_)
d_ + _x 1
o E_(_) _ + _ ' for _ _>x _>
where EI(_) is the flexural rigidity distribution , _ is the root restraint
stiffness constant for rotation in the plane of bending, and Kh is
the root stiffness constant in shear. The influence coefficients for
torsion of a rod of length _ are given by (torsional deflection at
point x due to unit moment at point $)
cee(x,_) = O_ Gj-rO_7+_
dk i , fo_ __>x_>O
dk i
=o
S
where GJ(k) is the torsional rigidity distribution and K e is the
torsional spring constant at the root. If the root is rigidly fixed_
the influence coefficients reduce to those given in [1].
65
APPENDIX
B
INTEGRAL
EQUATIONS
FORUNSTEADY
AERODYNAMICS
Three Dimensional Lifting Surface
The source solution of the linearized equation for the velocity

potential is given by
47r@= -H(T - r/c) (B.I)

r
where H(-) is the Heaviside step function, r is the distance between
the observation and the source point, T is the time measured from the
instant of the disturbance and c is the speed of sound. The observa-
tion point is affected at time T if the disturbance moving with a
velocity c can trayerse the distance r in that time; the strength of
the disturbance is -1/4wr.
A coordinate system (g,_ _ _ r) moving with respect to a fixed
coordinate system (Xo,Yo,Z0,to) , along the negative x 0 axis with a
velocity v, fixed in the wing_ is considered. The fixed and moving
coordinate systems are assumed to coincide at time t = t. The

o
coordinates of the point (_, _, _, T) referred to the fixed axes are
(_ - v(t 0- t), _, C , to). A unit acceleration potential source at the
point (_, _, C ) in the moving reference frame will produce an
acceleration disturbance at the point, (x, y, z), after some time delay;
the magnitude of the disturbance is
66
= [HC'v'-r/c)
i _ ]=-_' l
2
2
r = (x - _ + v(t o- t))2+(y _ _)2 + (z - C), (B.2)
In fixed coordinates, the velocity potential is related to the accelera-
tion potential as
@(x,y,z,t) = _t2_(x,y,z,_)dt
tI
where [tl,t 2] is the time domain of influence of _, which affect_ the
velocity potential at time t in fixed coordinates, t 2 and tI are the
last and first instants of time at which the moving source can affect
the observation point.
Therefore the velocity potential at (x,y,z,t) due to a moving
doublet at (_, _, _, to) of strength A_(_, _, _, to) is given by,
t2
i _ i
¢(x,y,z,t) = - I_ _ _ A_(_, _, _, to) _ dt 0 (B.3)
tI
The time t 2 satisfies the equation,
((x - _) ÷ v(t 2- t))2+(y-_)2+(z-_)2 = c2(t 2- t) 2 (B.4)
This is the equation for en ellipse and determines the region of
integration for equation (B.3). If X = x- _, Y = y - _ and Z = z -
and t2 - t = T, we have
67
MX+R
cT - or + R -MX i
_B.5)
I -M2 M2 - i
where R =_2 + (i - M2)(Y 2 + Z2)
For M < i, only one of the solutions is negative. Hence
IMX -R
t2= t+ (B.6)
c i - M2
If the perturbations in the flow field existed for infinite time in
the past then tI = _ • 0n the other hand if the flow field is
unperturbed for time t < 0, then clearly
i MX -R
t2 = tO +
c i - M2
and (B.7)
IMX -R
tI = 0 provided t >
c i -M 2
IMX -R
= t2 = 0, as the disturbance has not reached
If t < c i M2 ' tl
the observation point. For M > i_ both solutions for T are negative
and the period of influence is the time an advancing spherical wave
would take to pass the point of observation. Therefore,
IR -MX
t2= t+
c M2 -i
(B.
IR+MX
tI = t
c M2 -i
68
provided, of course, both t I and t 2 are positive. For M = i, (transonic
limit), we may let M = I + Eor i - e and take the limit as z _0. If
we denote U i A= X + V(t i - t), i = 0, i, 2, then in subsonic flow,
X -MR
U2 - i = x - Vt (B. 9)
-M 2 ' UI
provided the flow field is unperturbed for time t < 0.
For supersonic flow, t sufficiently large,
- (MR + X) (B.IO)
U2 - MR
M2 - X1 ,UI= M2 - 1
Hence the velocity potential at (x, y, z, t) due to a moving
doublet at (_,_,_,to) is equal to
U2 U O- X
_(x,y,z,t) = 1 _ A_(_,I]_ ,t + --V--- )
-_ J" dU o (B.II)
u1 _/Uoe+y2 + z2
The total velocity potential at (x, y,z,t) due to a distribution of
moving doublets on the lifting surface S, at _ = 0, is given by
U 0- X
)
4vr@(x,y,z,t) = - V1 j'j'd_ d_ _z jU2 A_(_,_,O,t + T
dUo (B.12)
s u1 _/Uo2+y2 + z2
The downwash on the surface is _iven by
A¢(_,n, 0,t + U -X)dU

o o
1Lt _ If d_ dy _ fU2
4_ w(x,y,t) = - V z-_0 S--z _z
S UI
/ 2
U + y2 + Z2
o (B. 13)
69
It can be shown that a discontinuity across a doublet layer exists
such that [I]
A$(x,y,O,t) = _(x,y,O,t) - _(x,y,-0,t) = Ap(x,y,t)

0 =J
(B.14)
where Ap is the pressure differential across the surface.
Substituting (B.14) in (B.13) and performing the differentiation
using Liebnitz rule, we have the following integral equation for the
downwash in subsonic flow.
M MX+R 1
w(x,y,t) = 1 ff dEdn [Ap(_,_ t + U_X)
V 8_q S ' R I X2+y 2
+ I U Ap _,n,t +--Uo-X i I dU
3 o
O V I(/U 2+y2 )
O
+ I tAp(_,q,(t-T)) VdT ] (B.15)

_(_-x/v)
0 ((X-VT)2+Y 2)
X-MR I
where Ap(_,n,t) = 0 for t<O, U -
I-M 2 '
1 RI-MX
q = ½p®V2,Rl =/X 2 + (I-M2)y 2 and t >
c I-M 2
In the limit as t + 0, U + X, MX + R 1 and X 2 + y2 + 0, and only the
first of the 3 terms remain in the equation B.15. Integrating over a
small region around the receiving point (x,y), we have
w(x,y,O) = _ _M Ap(x,y,O)
V 4 q
70
which is the familiar piston theory result. At the initial instant, the
lifting surface behaves like a piston, and the flow over it is
similar to one dimensional compressible flow. It may be noted that a
similar equation obtained by Drischler[63] may be reduced to the
sameform as (B.15) after somealgebraic manipulations.
Taking the Laplace transforms of Equation B.15, we may obtain an
integral equation relating the Laplace transform of the downwash

w(x,y,. pb
_-) and the pressure Ap(x,y, pb
-_-), where p is the Laplace trans-
form variable.
..... pb pb
81rw(x,y_(pb/V))v ql _ de d_ _(¢,_, _-)K3(X,Y, -_ )
where K3(x,Y, _ )=
re
p(X - MR1 -x>
[ R 1 X 2 + y2
X -MR 1
l_M2 _(Uo-X)
+ _ eV H(t - X/V) ] ] (B.16)
0 (U02+ y2)3/2 dUo + V ,_ { ((x - vt)2+ y2) 3/2
£ (m) represents the Laplace transform of m. The third term converges
only for real (s) _> O, and - X

0 - p-!_
V
e
R1 + j, dT
K3(X,Y Pb_ _ [e vP-(_l_-M2) MR I MX+
_V) = e X 2 + y2
m__-x(_2+y2)3/2
1 -M 2
ye ) IHI-z(v V
71
where IHI_I(') is a Struve function and
Y(-) is a Bessel function of the second kind.
For simple harmonic motion this reduces to the kernel function given
by Watkins, et al. (64). For purposes of calculations however it is

pb
convenient to write, K3(X, Y_ -_ ) as,
pX P MRI-X
pb - V [ -_[ _ ] M MX+ R1
K5(x,Y,V ) = e [ e R1 X2 + y2
j,
o
+ MI:{I_ X
e
v
(1.2+
dT
y2)5/2
] (B.17)
As shown by Landahl [65], the integral is also equal to,
Ii [ ]
-- Exp -PY_ dT
X 2 + y2
However equation B.17 is most convenient for calculation purposes. The
second term in equation B.17 is a singular integral at Y = O. Hence
the integration over the lifting surface must be done in the sense
of Hadamard, as expalined by Mangler [66]. This is indicated in
equation B.16 by the symbol _.
72
For small values of the reduced Laplace transform variable s = pblV ,
we may write the Kernel function as [67].
0O
sn s2
Ks(X,Y,s ) = E (an + _n(s).bn sn) (B.18)
nr-O
i
(1+ x )
where a 0 - y2
_1/X2÷ 132y2
X2 y2
l 1 (X+ + )
al = b y2 _X2+8% 2
b 0 = 1/2
bl : - X
2b
The coefficients b
n are functions of the spanwise variable Y only (and
not of X]
In steady flow, for s : O,
X
1 (i+ (B.19)
<x,Y
,o):
and in the limit as s _ _, from piston theory
K3(X,Y,S)Is_ : 2WMS(X,Y) (B.20)
In the limit as M -_ O, we have,
73
(T+x)
b
K 3 (X,Y,s)IM:O = J' e
d"[ (B.21)
-X (T2+ y2).5/2
In two dimensional incompressible flow we have,
-s (_+x)
d_" dY
K2(X's) IM:O a
= K20 (x,s)= J' J" e
-_ -x (T2+y2)3/2
2 + 2s -s X/b
: 2 ye si(sxfo)
2
X _7(i,i _- -b-
sX ) (B.22)
E (sX]
where i'-b-" is the exponential integral and U(l,l,-sX/b) is a Kummer
function.
The above Kernel function indicates a distribution of time delays
over an infinite time domain. Further they are not valid for Real(s)
< O. However if a solution to the integral equation can be obtained
for Real (s) > O, the convolution principle may be used to find the
solution for any s.
For purposes of calculations, we note that the function
]HI-I _'s--i_b)- YI ( _b ) in equation B.17 may be approximated by a
convergent sequence of rational functions in b--

_ or Pads approxi-
i
mants. It follows therefore that can be approximated by a
(2+z)_/2
finite sum of exponential functions. Two such approximations are
known, one due to Laschka [25] and the other due to Runyan and Watkins
74
[68] • The errors in these approximations are comparedin
aT
[69].
b
Laplace transforms of exponentials of the type e are given by
rational functions. Hence, in principle, it is possible to approximate
the entire Kernel function by a rational function in s. Further in
order that it converge to the piston theory kernel for large S_ the
approximation must be [N, N] Pade approximant. This point of view
has useful consequences, regarding the behavior of the aerodynamic
loads in the frequency domain.
In supersonic flow the kernel function may be easily shown to be

x+ mi su
sX s X+MR1
b
_-i e b e
K3(x,Y,s)= e
+ R1 X2+y 2
M MX+RI b M2_l for X > _- 1 IYI

+ R1 X2÷y 2 e s X'MRI 1
(B.23)
= 0 for X < 2_-i IYI
In this case the kernel function is valid for all s, including the
left half of the 's' plane.
In transonic flow though the linea_ized theory cannot predict
shock waves We may obtain the kernel function by proceeding to the
limit as M _ 1 from the supersonic case. We have,
75
sU y2
sX m b s (x- )
b
K3(X,Y,s ) = e j. e dU + e
-X y2 (U2+y2)3/2 X2+ y2
X>O
= o eor x < o (B.24)
In this case again the kernel function is only valid for real (s) _> O.
Two Dimensional Airfoils
In two dimensional flow, the integral equation relating the down-
wash and the pressure distribution may be obtained by integrating the
corresponding three dimensional kernel function with respect to the
spanwise coordinate from - _ to + ®. For the case of arbitrary motion
of an airfoil , one can show after some computation, that
Vt -X
aT
87rw(x,t) _ _I _ X2 ea_ J' _p (e,t- X÷T]
--g--.
-X
V q Xl
I+M
T_I M2T2
(_X) 2
-X
I+M
(l-_(X))d_ X+_
(B.25)
+ J' Ap(g, (t - -V-))
-X___ 2_I M2T 2

1-M (T+x)2
76
where xI = x Vt(1 M + M) ' if x - Vt (l+M)
M > XL
= xL , if x - Vt (I+M)<
M x_
Vt(1 - M)
x2= x+ M if x + V (t_M M) < xT
= xT if x + V (lt_MM >x T
Eqtmtion B.25 is valid for any type of airfoil motion. In incompressible
flow we have,
XT Vt -X
8Ww(x ,t)
i _ 2 d_ _ Ap(_ t (X+_)) d_ (B.26)
V
qk -x v
At time t = O, we have
w(x,O) M @(x,O) (B.27)

V - _ q
Non-dimensionalizing the time variable t to T = Vt/b and taking the
Laplace transforms of the integral equations, we have
8_(x,s)
V
1 # K2(x - _, s) _(g,s)d_ (B.28)
q_
where w(x,s) is the Laplace transform of the downwash and _(x,s),
77
the Laplace transform of the pressure
K2(x,s)
= 2s*S£e+S*_[ Ixl)+%(M/Ixll
(_l(Ms*
+ e-S x_ _n( l+_+ 1 s x(x-l)

M "- sx _e K0(Ms * lXlX)dX]] (B.29),
0
where s = s/_ 2, K0(. ) and Kl(. ) being modified Bessel functions of the
second kind.
Also K2(X,s)Is- o0: + 2M_6(x) (B.5o)
26 (B.31)
and K2(X's) Is- 0 - X
In incompressible flow, equation B.29 reduces to equation B.22. Equation
B.22 has a well known inverse given by, (after non-dimensionalizing
all distances by the semi-chord).
i
_o(X,S) = _ a(x,z,_)_(z,s)dz (B.32)
-1
where
G(x,z,s) = - 2_ [sA(.,z) _ 1-_
-x i:_ (¢(s)- _ + _---__x
)] (B.55)
1 i -xz+ -_xe./_ - z2
^(x,z) = _ log (B.3_)
i _ xz _ __--_x2 __z 2
Kl(S)
real (s) > 0 (B.35)
$(s) : Kl(S)+ K0(s ) , continued
78
2su(1_,3, 2s)
, real (s) > 0 (B.35)
2so(1½,
3, 2s)+ l, 2s) continued
2 -i -xt ax]
= S "_[i - [(K0(x)_ KI(X))2+ 2(I0(x) + ii(x)) ] e 2
0 x
Ki(" ) and li(- ) are modified Bessel functions and U(a,b,z) is a
Kummer function.
Equation (B.28) may be reduced to several alternate forms. One
method of doing this reduction was given by Fettis [70]. The integral
equation may be reduced to a non-singular Fredholm integral equation
with a completely continuous kernel function (defined in Appendix D)
given by,
Ap(x,s) =
_Po(x,,)
._ - s
.M2jf Kcc(X,_,s)_p(_,s)d_ (B.56)
-i
KNc(X,_,s) = H(_ - x)- cos

7T
x _ (_(s)-l)
IT
l-x
- I_ _ _-IG(X'Z's)K(s(z-_))dz"
H(_ - x) is the unit step function and
K20(z- _,s) 2 M2
_(s(z-_) = K2(z - _,s)- B + _
Another important representation of the integral equation may be
obtained. The kernel function K2(X,s) may be written as,
79
K 2 (X,s) : e+M2Xs* _{K2o(X,s*)-
T[K2o(X,s
M 2 * )-K20
s (X ,s*) ]
+ _-_-.s*.[ -- in 2+B 2 in s - 4 (l-B)in M _ 2¥]

___2 _ * M2
M2 .2 4 2
2 s [_anl+-- _ - 1 -_-_in - ---
+ o(M4).o(x2+ax log x)} (B.37)
+ s *XM 2 * M2
= e B{K20(X,s )- _[K20(X,s*)-K20S (X,s*)]
+ M2s* (l-2 [in (_Ms )+ _])+ M m V(X,s )] (B.38)

2
where s
K20(X,s *) = 2( ½ + s*In X) (B.59)
Kv(X,s ) : O(M°+ alnM) (B._)
Using equation B.38 we may invert equation B.28 using the well known
solution for incompressible flow. Thus we have
a_(x,s) _Pov(x's*) _s*x M2 _s*x[ 1

q - V e - 2"_ e _ A._(s,*x)ji(s,-I)
*
i:o
Bi(s,. x)Ji(s*,x)]- M
_ 4 eM2s*x 1 _ 1 G(x,z,s . )Kv(Z-g,s )e_M2s*_.
+ J" _(_,s)
-i-i q
• d_dz (B. 41)
where
8O
APov(x's*) i i .
q = _ j"c(_,_,s)_(_,s)e
-M2szdz (B.4_)
-I
i _M2s* _
Ji (s*'x) : 7 _ Ap(_,s) e
q
X
Ao(S* x ) = ln-lx(2s*_s*2x)+ l-_x_x

l+x (s*@(s*)in 2-(@(s*)-l)
S _ * *
+ -_-(_(s )+s (l+x))+s*2(l+ in 2)(1 + x)).
Bo(S ,x) = -(2 - s _)Trs
Al(S ,x) = s*21n-lx + s*(_(s*)-l) 1-x
* *2
B l(s ,x) = -TTs
= [i - 2(in( 1 Ms*)+ V)]
Thus for sufficiently small M we have for Ap,
_p(x,s) _%v (x'_*)M2s*x

= e
q q
M2 M2s*x[ _ A i(s* 0 * * 0 *
- --_ e i:O ,x)J (s,-l) + Bi(s ,x)Ji(s ,x)]
+ O(M4s*4(a + b in M)) (B._3)
81
where jO (s*_x) = ]l_i APov(_s) d_.
I q
x
This solution is exact to O(M2s*). The first term is identical
to Osborne's thin airfoil theory [71]. The usefulness of this
representation will be briefly discussed in Appendix C.
82
APPENDIX C
CALCULATION OF AERODYNAMIC LOADS
FOR SIMPLE HARMONICALLY OSCILLATING AIRFOILS AND LIFTING SURFACES
In this appendix some of the existing methods and their extensions
for the calculation of aerodynamic loads on wings with and without
control surfaces will be briefly discussed.
Two Dimensional Airfoils
Analytical methods. - In incompressible flow Theodorsen's
solution is well known for calculating pressure distributions and air-
w n
loads on two dimensional aerofoils. For a downwash V = WnX ,
we have
ik(l+x)
Ap =
-_- Cp = _ 4_ Wn[xn+( n-i
_ x n-_ -12-_ _ : )+ n+l
_=0 I(_)
n
( z xn-_ 2-_ _ : )+(C(k)-l)
2-I]i]: (C .i)
_:o,2,_,... T(_) _ ]
2
n , for n even
where I(n) = (5 ")
n+l , n-i ,
= -_ • V . for n odd.
For a down%ash given by w _-_(x - xc)n for x > x

C
= 0 for x < x
c
83
4 n k-i n-r-i C kij +j
_P--_]l_ _n [nl
z r=O
cr { nrl
z (-i)
k=O
kx t n-r-1-]k n-r-k )]"
• (x - Xc)r
nZ rrblc ( n-rZ (-1 )k-lxk n-r_ Ck j n_r.k). )r

- (x-x c )ncoslX o+ _.__(I+x)
_. _-y r
(x-x c
r:O }_0
+(1 -C(k)) Z
r=O
n r nCr(-Xc) (Jn-r + Jn-r+l ) l
4 ik(x-xo I l-XcX
+ i-x2
---W [(x--Xc]T_n_ n + 1 }log[
n
I (_ - xe) t (c.2)
vZi_x 2 _-i
(2e-1) (2_-3)... (2e-2R+l) x 2 _,-2k-i )
c (x2_-i + Z
where J2_ - 2g k=l 2_(_-l) (_,-2)... (,_-R) C
26 _ -i
+ cos x
C
22_ :£ :
and
c x2_+ _ -i k+l
- Z #2,g - " ('_ -1) (,g -2). • • (,_ -k) c2'_ -2k-2
J2,_+l 2Z + 1 c k=O (26-3)... (2_-2k+1) x
The expressions for the loads may be easily obtained by integrating
the corresponding moments of the pressure distributions.
Equation B.43 extends Theodorsen's solution to subsonic flow for
low mach numbers and frequency. The first term of equation B.43 is
identical to Osborne's theory• Kemp [72] has provided the closed form
84
expressions for the lift and moment based on Osbourne's theory. How-
ever it was found that this approach is not too attractive for purposes
of numerically computing pressure distributions or aerodynamic loads.
Further these solutions are not valid for either high subsonic mach
numbers or high reduced frequencies, when compressibility effects are
not small relative to circulatory effects.
For high frequencies, and Mach numbers M _ O, the method of
acoustic planforms of Lomax, et al. [12] is useful. The two dimen-
sional unsteady problem for compressible flow can be formulated in a
coordinate system x'_y' _z',t' fixed in the undisturbed fluid_ by the
equation,
1
q°x'x' + _z'z'- 2 _t't'
C
with the appropriate boundary conditions. For the three dimensional
steady supersonic case, in body fixed coordinates, it is
Hence it is seen that the solution of the unsteady problem can be
obtained from the solution for the steady flow about a swept forward
wing tip placed in a supersonic flow with Mach number M = 4. The
angle of sweep is determined by the Mach number of the two dimensional
aerofoil from than relation A = ta_ll/M. The wake vortices in the
unsteady two dimensional problem correspond to the tip vortices in the
steady three dimensional problem. The three dimensional steady flow
85
problem may be solved by Evvard's method [i]. For example the pressure
distribution obtained this way, corresponding to an indicial down-
wash w0, for t _ O_ is given by_
(_,x)- 4Wo A
M = _pl
for T < min( M M (i -x))

- I-_ ' Y_
A C1p(_,x) = A Cpl + _8 -_-

Wo [ tan -I _ - 1 + I-_VM--_-_
M I/(I+M)T - 1 ]
for l-_M(l+x) _<T -<i--_(i-x)
8 Wo , I-M)T
A Cp(_,x) : ACpl + _-_ tan "I VM-_-_ - 1
for _M (l-x) _< T _< I_M(I+x)
vt
where • - and 0 <M < i
b
Expanding in the variables (l-x) and (l+x) indicates that for t > O,
at the leading edge (x = -i)
1
_p(_,x)* ® as
and at the trailing edge (x = l)
_Cp(_,x)÷ o as
86
The indicial aerodynamic load coefficients may be obtained by
integrating the pressures over the entire chord of the aerofoil.
Taking the Laplace transforms, and using the asymptotic expansions

of the resulting Frenel integrals
R -ih -iR
Fr(R) : _ e du - l-i + ie (i i 3__)
0 _ 2 _ SiR 4R2
we may obtain the expressions for the non-dimensional lift and moment
coefficients in plunging and pitching for large values of k.
These are
Lh
2b'q - 4ik
M + _(I_M) _1-i _ t e-ikl_M + (I-M) 2 e -i_-/_
"k2-M 1
+ o(i/k)
-ikO_M 2_M
Mh _
2b----2q 1___2 _
M_-7_ Ie I+M - (I_M)2 e -i i-_ I + 0(i_ )
.k_M k_4
2bqL : _ M_ + _l'i _k e-i--

I+M-(I-M)2e I-M
-i-- I + 0(i_)
M_
2b2q _ _4 M-
ik + --2(l'M)
MS + _----_-/_V_
l-i ./_{e -dkl-_-M + (I-M) 2 e-ik_-M" 1+ 0(i/k).
It is interesting to note that it is not possible to expand these
coefficients in a power series in (l/k). Also they are not valid for
M = 0. Thus it is possible to analytically obtain the behavior of the
87
aerodynamic loads for low and high frequencies. While they contribute
to a better understanding of the behavior of the pressure and aero-
dynamic loads at high and low frequencies, they do not serve ar_v other
useful purpose. For this reason and for computational efficiency, it
is convenient to resort to numerical techniques which are valid for
all values of frequency and mach number.
Numerical methods. - One of the earliest methods to calculate
aerodynamic loads on two dimensional aerofoils is due to Dietz
(described in Refs. I_ 2_ and 3). For wings with trailing or leading
edge flaps_ Fettis suggests a method that is fairly accurate. From
a computational point of view Hsu's method [73] for wings without
flaps seems to be simplest. However Hsu's method can be extended for
wings with control surfaces. This modification is briefly described
below. A pressure loading function of the form,
M N
ACp = E Pmhm(x) 2 6 _. an(X. xc)n
m=O 7r 8 n=O
log
_XXc+ 2 2
_
is assumed, where hm(X ) are identical to Hsu's chordwise loading
functions in compressible flow and are the exact loading functions
n
corresponding to a downwash of x in incompressible flow, given by
equation C.1. 5 is the angular deflection of the trailing edge flap.
The above loading function is substituted in the integral equation and
integrations of those terms that contribute to discontinuities in the
88
downwashand its chordwise derivatives are performed analytically. The
coefficients an can then be found by comparing like terms on both
sides of the integral equations. The kernel functions K2(x,s ) may be

written as
K2(x,s
) ~ 2B
x 2s lnlxl-(
-8- s
-_ )2(_-384)x in I_I
+ r_ _xn+n inlxl Z 8nxn

n=O r_2
aO, a1 and a 2 depend only on the terms outside the summation signs and
are given by,
2
S
: = 2(i + 8 2), a2
a0 i, a1
and
s )3.o(M
)
a 3 = o( _-_
In incompressible flow a3, a4, ..., etc., are equal to zero (from
equation C.2).
The singular part of the above kernel function and the pressure
loading functions are integrated exactly. A computer program is
developed that evaluated the Possio kernel by Gaussian and Berthod
Zabrowski quadrature, integrates the assumed pressure loading functions
and the non-singular portion of the kernel function numerically and
then solves for Pi" The program then obtains the corresponding
generalized aerodynamic forces for any mode shapes. Results indicate
that for M = 0.O and 0.7 , the error in using a ten point collocation
89
schemeand not including the flap singularity is 20 - 25%of the four
point collocation solution with the singularity included. In incom-
pressible flow using the four point collocation schemeresults in an
error of less than 1%, in the momentabout the flap hinge line. The
program may be used for values of 0 < k < lO and can include trailing
and leading edge flaps. For high frequencies a large nu_nberof colloca-
tion and integration points are required.

In supersonic flow, the expression for the pressure and airloads
given in Ref. 1 may be integrated using Gaussian integration with no
difficulty as control surfaces do not contribute to any singularities.

Results for a rigid airfoil with trailing and leading edge flaps are
given in [44] and [45].
Three Dimensional Lifting Surfaces
Analytical techniques. - There are no known analytical tech-
niques available for arbitrarily shaped three dimensional lifting
surfaces. However, as pointed out by Miles [74], the solution of the
partial differential equations for the velocity potential by separating
the variables is possible only in eleven Euclidean coordinate systems.
Of these only three define lifting surfaces of practical interest.
These are (i) elliptic cylindrical (two dimensional aerofoil) (2)
oblate spheroidal (circular planform) and (3) ellipsoidal (elliptic
planform).
The circular oscillating lifting surface in incompressible flow
was considered by van Spiegel [54]. It is interesting to note that from
van Spiegel's general solution, we may extract the solution for the
9O
virtual inertia loading. The orthogonal coordinate system for the
circular wing is given by
x= + 2 -_ cos V y= + -u sin v j
z = u_ where 0 < _, -i < _ < i, 0 < _ < 27[ so that the entire space
is covered just once. The wing itself is given by _ = O, while the
part of the x_ y - plane outside the surface is given by _ = O.
The problem is treated by van Spiegel by the method of the
acceleration potential. The regular solution is equal to
_l = _ r_ Pi J(u)Qi j(i_){cijcos i v + S iJsin i v]

i=l j=O
where Pi j and Qi j are Legendre's associated functions of the first and
second kinds. Since @l should be an odd function of z and u, the
summation sign is restricted to i + j = odd. This is indicated by
the prime added to the summation over j in the equation for @. The
constants C. j and S. j are determined by the prescribed normal derivatives

1 1
at the airfoil. By expanding _iz = _i_/_ in a series of surface
harmonics the constants are determined• It is found that_
i 2_
_l(n"_' ) = _ _ _0 _lz(_l'_l)G(n'_'V'O"_l'Vll_ldY
-i
where the Green's function of the second kind is equal to,
W m
E E 1 2n+l(n-m) : m m
v;o, k' vl) -- n=l m=O _m 27[ _ Pn (U)Pn (_l)
m
Qn (i_)
• • Cos m( V-V i)
_Qj(iO)
91
where T0 : 2, E1 = _2 = _3 ..... i.
The above solution is one that has the exact prescribed acceleration
on the aerofoil. Further the solution is not singular at the leading

and trailing edges and behaves as V_ where r is the distance of the
leading edge or trailing edge. The solution that has the exact normal
velocity w(u,V), prescribed on the airfoil is
_t 2_ i _
_*(_,_,v): ( + '__) o -1 'Vl
"G (q,m ,V ;O,!_l,Vl)dV, d'a
This solution leads to apressure singularity at the trailing and lead-
ing edges due to the differentation in x.
Consider now
o, o, vl) _- 1+ 1__ 2
This is a solution of Laplace's equation having zero normal velocity
on the aerofoil and a singularity at the point G : O, _ = O, V = V1
which lies either at leading or at the trailing edge. The singularity
is of the type _(cos G)/r 3 , where r is the distance from a point on
the aerofoil to the singular point and G the angle between the vector v
and the radius of the circle. Thus the singular solution is,
92
3_/2
/2
where the integration is along the leading edge only, since no singu-
larities are admitted at the trailing edge. As in the two dimensional
case the singularity in _2 is of the type V_-/_, where r is the dis-
tance from the leading edge. The function a (rl) should be deter-
mined from the condition that the normal velocity at all points on
the aerofoil due to the total acceleration potential _i + @2 agree
with the prescribed normal velocity. Hence the normal velocity at
all points due to the acceleration potential _i + @2 " @* should be
equal to zero. This leads to an integral equation which may be
solved by Fourier series expansions.
On examining this solution carefully, we find that very large
frequencies of oscillation or at the starting instant (_i + _2 - _*)
is independent of frequency. Hence we conclude that the solution
corresponding to the virtual inertia loads is,
2 1 _2
_I*
= oSo__i(--_t
2 Z(_l,
v l))a(_,_,v_°,_l,vl)_
v,d'_
Like the regular solution this is not singular at both the lead-
ing and trailing edges.
The effect of compressibility may be taken into account by means
of spheroidal wave functions. Also the oscillating elliptic wing can
be treated when Lame functions are used. However, these analytical
93
solutions are not really exact and the degree of computations is
enormouswhen comparedto numerical methods. Hence these techniques

are not too useful.
Numerical techniques. - Numerical techniques may be broadly
classified into two groups (a) finite element methods and (b) kernel
function techniques. The doublet lattice method of Albano and Rodden
[26] and potential flow methods [16] in subsonic flow and the Mach
box method in supersonic flow [7_ 76] are examples of the first
type. The kernel function methods of Laschka [25], Cunningham [23]
and Rowe [24] in subsonic flow and Muller's method [67] in supersonic
flow are examples of the latter type. Two of the methods that were
actually used in a slightly modified form during the course of this
work are briefly described.
The doublet lattice method: This method for lifting sur-
faces is based on the developments of Albano and Rodden. The surfaces
are divided into trapezoidal panels, the downwash being taken at the
midspan of the 5/4 chord. A distribution of acceleration potential
doublets whose strength is to be determined, is assumed along the
quarter chord of each panel. The assumed pressure loading is
then integrated along the quarter chord line. In the original method,
the integrals were obtained by approximating the kernel function in
each box with a parabola in the direction of the quarter chord of
each panel ' divided by rl 2 ' where rl 2 = YO 2 + z0 2 ' Y0 and z0 being
relative y and z distances of the midpoints of the quarter chords
(sending points) and the three quarter chord (receiving points).
94
The force on the panel due to the pressure distribution is assumedto
be equal to the force on the doublet line segment. The accuracy of
the method was remarkable for aeroelastic purposes. However for
large frequencies (reduced frequencies k > l) it was found that the
method was not only not accurate, but also did not converge to the
results obtained by piston theory. It was found that by approximating
the kernel in each box with a quartic in the spanwise direction of the
2
quarter chord point of each panel divided by rI , the results were
tremendously improved for planar lifting surfaces for high fre-
quencle s •
Using the same notation as Rodden, Giesing and Kalman [77] we
may rewrite equation 27b of Ref. 77 as,
P (,i) [Klexp{-i_o(x-n tan k s)/V} - KIO]
+ + Cl
: AO _2 + BO_ +
Assuming that P (_) is evaluated at _ = +_ ae, +_ be and O, where e is
the semi-span of each panel we have,
1
1 (P(ae) -P (-ae) )- I(P (be)-P (-be) ) ]
2 (a 2-b 2 )e 3
95
= +
2 (b2-a2)e a "j
cI : P(O)
'a' and 'b' were chosen to be 1.O and 0.5 respectively. For
A 0 = B0 = O, the formulas for A1, B 1 and C1 reduce to those given
by Albano and Rodden. The above approximation is very good for high
frequencies, especially for swept wings (ks _ 0). At high frequencies,
for swept wings_ P (_) varies rapidly for different values of _. How-
ever extensive numerical experiments have to be carried out determine
the optimum values of a, b and the receiving and sending points in
various frequency domains.
In spite of the improvement in the results for moderately high
frequencies, the results did not converge to the piston theory limit
for certain mode shapes. For convergence the distance between the
sending and receiving point on each panel must vanish as this condition
is known to hold in the case of piston theory. This would require,
theoritically, am infinite number of boxes in the chordwise direction.
Also for a proper spanwise pressure distribution, which critically
effects the aerodynamic loads a large number of spanwise boxes are
required at all frequencies. This leads to an excessively large
computer storage requirements and thus the method is not too attrac-
tive for high frequencies.
96
Laschka's kernel function method: In this method the pressure
distribution is taken in the form
N
z a.(y)h
(x,y)= i=l i(x)
I
Following Laschka, the terms in the kernel function which are singular
are separated from the part that is regular. The singular terms are
integrated analytically while the regular terms are integrated
numerically. The integral equation is reduced to a linear system and
the values of the unknown coefficients ai(Y ) are obtained at different
spanwise and chordwise stations. For the chordwise integration, the
integrations points and the collocation points may be chosen inde-
pendentally. For high frequencies one may choose fewer collocation
points and a large number of integration points, in order to converge
to piston theory. Thus the kernel function method is more efficient
than the doublet lattice method.
Control surfaces may be handled by both methods, without taking
into account the singularities at their leading edges. However, the
computational time for high frequencies is quite high when compared
to the corresponding time at low frequencies, in both methods.
97
APPENDIX
D
PADEAPPROXIMANTS
ANDINTEGRAL
EQI_TIONS
Pad@Approximants
The [N_M] Pad_ approximant of a matrix function F(z) is defined by
F(z) = P(z)R-l(z)
where
P(z) is a matrix polynomial of degree M and
R(z) is a matrix polynomial of degree N
If in particular F(z) is a power series defined by,
L
i
F(z) : E Fiz and if
i=O
(D .2)
M zi M .z i A
P(z) = Pi , R(Z) = _ R , RN = I
i=O i=0
• and R.
we have the following equations for Pl i
N-I
k = 0, I, _,
_i 3, ° • • M
Fk_ j Rj + Fk_ N = Pk'
J=0
(D .3)
N-I
Fk_jR j = -Fk_N, k = M i i_ M ÷ 2, M ÷ 3, ... L

J=O
98
For L _> M + N_ equations D.3 may be solved for the matrices Rj and Pk
successively.
On the other hand, if F(z) is known for z = zk, k = 0,1,2,3 ... L
we have
M N-I
F(Zk)Z kN = _. PiZk i - F (Zk) r. R.z i, k = 0,I,2,3,... L

i=O i:O i
(D .4)
It is now assumed that PO and PM are known apriori. (If PM is not
known, we may increase value of M by 1 and assume that PM+l = [O]).
From the L known values of Zk, we pick M-1 values, _, k I = 1,2,3,
... M-l, and rewrite equations D.4 in part as
M-I N-I
i=l i:O
: i_ 2, 3, ... M-I
We now have M-1 unknowns P. and M-I equations and hence we may solve
1
for the unknowns P. in terms of R. as

1 1
N-1
Pi = J=O
?' BijR'J + BO (D.6)
Hence equations D.4 reduce to a system of L-M+1 equations for the N
unknowns Rj_ j = 0,1,2, ... N-l, which may determine by least squares
provided the system of equations are solvable. On. the other hand it is
also possible to solve equation D.4 directly by least squares. These
equations are also applicable for scalar quantities. Properties of
Pad_ approximants are discussed further in [78].
99
Properties of Integral Equations Related to Pad_Approximants
([79] and [80])
In this section someof the basic definitions and properties of

I
integral equations of the form
b
_(x) = f(x,s) + s J' dy K(x,y,s)_(y) (D.7)
a
where f(x_s) and K(x,y,s) are defined on the closed internal [a_b],
are discussed. We can regard_
b
K -- j' dy K(x,y,s)(-) (D.8)
a
in equation D-7 as an operator acting on the function $. In operator
form the integral equation is
= f + sK_ (D.9)
The function K(x,y,s) is the kernel of the equation and in order
to ensure a valid definition and a unique solution of D.7, various
restrictions can be imposed on _, f, and K. It is usual to assume
and f are functions of bounded norm II "If in the Hilbert space L2
of functions defined on the inter_al [a,bl; for example,
b
II 112=J' < (D.ZO)
a
A sequence of functions [fn} is bounded if
i00
IIfnll _< B < - (D.II)
Bounded operators. - The operator defined by D.8 is bounded
if there is a finite constant M, such that
II II
_<Mllhll (D •12 )
for all functions h of finite norm. This is a different condition from
the boundedness of the function K (x,y,s) given by IK(x,y_s)l j c < = •
The norm of a bounded operator K is the least value of M satisfying
D.12. The operator is unbounded if no finite M can be found such that
(D.12) holds for all h.
Kernels of finite rank.- A kernel K is of finite rank n

n
if it is expressible in the form
Kn(x,y,s ) = E _i(x)Bi(y) (D.13)

i=l
where [Gi ] and [8i } are linearly independent sets of functions in L 2.
If the kernel K(x,y,s) is a kernel of finite rank and _i and 8i are
rational functions or polynomials of s equation (D.7) may be solved
exactly. For simplicity it is assumed that f(x,s) is independent of
S.
The kernel function method_ - The scalar product of functions
f and g in L 2 is
101
b
(f,g) = j' _ f(Y)g(Y) (D .14)
a
Substituting (D-13) in (D.7) gives,
n
_(x) = f(x)+ s Z ($,_i)(_i(x) (D.15)
j---1
The coefficients ($,_i) are complex numbers and ar_ solution of (D.7)
may be of the form,
_ = f(x) + E _j _.
j=l 0
where the coefficients _j and to be determined.
Substituting (D.16) in (D.19), _j are determined by
i=l[6ij - s(°ti'_J)SEi = s(fi,6j).
Let A(s) be the determinant of the coefficients of _j and the

th
minor associated with i,j element be Aij.
n
Then
_i = A-1 j=EIAijs(f'_j)
and
n n
q9 = f(x)+ Z A"I Z Ajis(f,Bi)
J=l i=l
n n
Hence,_ = a-i[_ (x)+ Z s (f(x),Si (x) ] (D.16)
As a function of the parameter s _ this a rational function, the numera-
tor degree and the denominator degree being dependent on oi and Bj.
102
Compact operators. - Let T be a linear operator that trans-
forms any function h in L 2 into another function Th of L2; in parti-
cular T could be of the form given by equation D.8. Every linear
operator of this type is continuous in the sense that it transfors a
sequence [h n] converging in the mean to h into a sequence [Thn]
converging to Th.
Compactness or complete continuity of a linear operator is a
stronger condition than continuity. Let [fn } be any bounded infinite
sequence in L 2, so that [Tf n} is another infinite sequence in L 2.
Then if for all bounded [fn},{Tfn] contains a subsequence converging
to some function h as n-_ _, so that
llTf n - hll = O
then T is completely continuous. Reiz and St.-Nagy [80] give necessary
and sufficient conditions for complete continuity.
Completely continuous operators can be uniformly approximated by
operators with kernels of finite rank. Also, if a kernel function
can be approximated with kernels of finite rank, then the operator
is completely continuous. In unsteady aerodynmmics the kernels are
implicitly approximated by kernels of finite rank and resulting solu-
tions are known to converge to the exact pressure distributions uni-
formly, for all frequencies.
Thus there exists an infinite sequence of kernels K1,K2,K 3 ...
of rank i, 2, 3, --. such that
103
with ¢ -, 0 as n-_ _ for all h ¢ L2_ uniformly for all s.
n
m
Let K n be of the form E Gi (s ,x)Bi (s ,y) where G. (s,x) and _i(s,x)

i=l l
are rational functions or polynomials in the variables s. Hence it
may be assumed that as s _ = K n is of O(s E) for all n. Then from
equations D.16 we obtain a sequence of solutions, which are rational
functions of s. The degree of the numerator would be N-E and that of
the denominator N where N _ _ as n _ m •
Neumann series solutions, -
Another method of solving (D-7) is to substitute the series
_(x) = f(x)+ E snfn(X,S) (D.17)

r_-l
and equate the coefficients of s on each side. This gives
fn+l(X'S) : S dy K(x,y,s)fn(y,s)
a
n : 1,2,9, ...
The series (D.17) is known as a series and may be written in
operator notation as
_(x) = f(x) + Kf(x) + K2f(x) ...
When K is uniformly bounded in the norm, then the series converges in
the norm if,
104
-1
IsI_<11KII
When s does not satisfy this condition Chisholm [79] has shown that
Pad6 approximants of the series may be formed which will give approxi-
mations to the solution for all s.
From Chisholm's theorem it follows that: If the kernel
function K(x, y, s)of the integral equation D.7 can be approximated
by a sequence of ksrnels of finite rank 3 then there exists a sequence of
solutions which are rational functions in the 's' domain, the degree of the
numerator being N + J and the degree of the denominator being N, which
converges uniformly to the exact solution as N -_ _ , where J depends
on the limiting behavior of K(x_y,s) and f(x_s) as s ÷ _ •
Thus if the behavior of the pressure distributions in unsteady
aerodynamics are known for high frequencies and in steady flow and if
it is possible to show that the associated kernel functions are
completely continuous for all s_ then it is possible, in principle,
to construct Pad_ approximants for the aerodynamic loads from the
numerically obtained solutions.
105
6. REFERENCES
i. Bisplinghoff_ R. L., et al., "Aeroelasticity," ist ed., Addison-
Wesley Publish Company, Inc., 1955.
o Bisplinghoff, R. L. and Ashley, H., "Principles of Aeroelasticity,"
John Wiley and Sons, Inc., 1962.
9° Fung, Y. C., '_n Introduction to the Theory of Aeroelasticity,"
Dover Publications, 1969.
4. Leipholz, H., "Stability Theory, '_ _.cdemic Press, 1970.
, Przemieniecki, J. S., "Theory of Matrix Structural Analysis,"
McGraw-Hill Book Company, 1968.
.
Flannelly, W. G._ et al., "Research on Structural Dynmnic Listing
by Impedance Methods," USAAMRDL-TR-72-65, Nov., 1972.
. Garrick, I. E., "On Some Fourier Transforms in the Theory of Non-
Stationary Flows," Proc. Fifth Int. Cong. Appl. Mech. (Cambridge,
Mass., 1955). AIAA Selected Reprint Series on "Aerodynamic
Flutter," Vol. V, March, 1969, pp D2-D5.
. Theodorsen, Theodore, "General Theory of Aerodynamic Instability
and the Mechanism of Flutter," NACA Rept. 496, 1955.
Q Sears, W. R., "Some Aspects of Non-Stationary Airfoil Theory and
Its Practical Application," Jour. Aero Sci., Vol. 8, No. 3,
Jan. 1941, pp 104-\08.
106
i0. Jones, R. T., "The Unsteady Lift of a Wing of Finite Aspect Ratio,"
NACA Rept. 681, 1940.
ii. Jones, W. P., "Aerodynamic Forces on Wings in Non-Uniform Motion,"
R. & M. No. 2117, British A.R.C., Aug. 1945.
12. Lomax, H., et al., "Two-and Three-Dimensional Unsteady Lift
Problems in High Speed Flight," NACA Rept. 1077, 1952.
13. Mazelsky, B. and Drischler, J. A., "Numerical Determination of
Indicial Lift and Moment Functions for a Two-Dimensional Sinking
and Pitching Airfoil at Mach Numbers 0.5 and 0.6," NACA TN 2739,
1952.
14. Drischler, J. A., "Calculation and Compilation of the Unsteady
Lift Functions for a Rigid Wing Subjected to Sinusoidal Gusts and
to Sinusoidal Sinking Oscillations," NACA TN 3748, Oct. 1956.
15. Miles, J. W., "Transient Loading of Wide Delta Airfoils at
Supersonic Speeds," Jour. Aero. Sci., Vol. 18, No. 8, Aug. 1951,
PP 543-554.
16. Miles, J. W., "Transient Loading of Supersonic Rectangular Air-
foils," Jour. Aero. Sci., Vol. 17, No. i0, Oct. 1950, pp 647-652.
17. Djojodihardjo, R. H. and Widnall, S. E., '_ Numerical Method for
the Calculation of Nonlinear Unsteady Lifting Potential Flow
Problems," AIAA Journal, Vol. 7, No. i0, Oct. 1969, pp 2001-2009.
107
18. Hess, J. L., "Calculation of Potential Flow About Arbitrary
Three-D_lensional Lifting Bodies," Douglas Aircraft Company
Rept. No. MDC J5679-01, Oct. 1972.
19. Ashley, H., et al., "New Direction in Lifting Surface Theory,"
AIAA Journal, Vol. 3, No. i, Jan. 1965, pp 3-16.
20. Landahl, M. T. and Stark, V. J. E., "Numerical Lifting Surface
Theory-Problems and Progress, _'AIAA Journal, Vol. 6, No. ii,
Nov. 1968, pp 2049 - 2058.
21. Woodcock, D. L., '_ Comparison of Methods used in Lifting Sur-
face Theorys" AGARD Rept. No. 583.
22. Ashley, H., "Some considerations relative to the predictions of
unsteady airloads on lifting configurations," Journal of Air-
craft, Vol. 8, No. i0, Oct. 1971, pp 747-756.
23. Cunningham, A. M., Jr., "Unsteady Subsonic Collocation Method for
Wings with and without Control Surfaces," Jour. of Aircraft,
Vol. 9, No. 6, June 1972, pp 413-419.
24. Rowe, W. S., "Collocation Method for Calculating the Aerodynamic
Pressure Distributions on Lifting Surfaces Oscillating in Sub-
sonic Compressible Flow," AIAA Symposium on Structural Dynamics
and Aeroelasticity, Boston, Mass., Aug. _O-Sept. i, 1965, pp 31-45.
25. Laschka, B., "Zur Theorie der Harmonisch Schwingenden tragenden
Fl_che bei Unterschallstromung," Zeitschrift fiir Flugwissen-
schaften II, July 1963, pp 265-291.
108
26. Albano3 E. and Rodden, W. P., '_ Doublet-Lattice Method for Cal-
culating Lift Distributions on Oscillating Surfaces _n Subsonic
Flows," AIAA Journal, Vol. 7, No. 2, Feb. 1969, pp 279-285.
27. Gwin, L. B. and Mclntosh, S. C. Jr., "Large Scale Flutter
Optimization of Lifting Surfaces,"AFFDL-TR-73-91, January 1974.
28. Theodorsen, T. and Garrick, I.E., "Non-Stationary flow about a
wing-aileron-tab combination including aerodynamic balance,"
Rept. No. 736, NACA, 1942.
29. Jones, W. P., '_erodynamic Forces on an Oscillating Aerofoil-
Aileron-Tab Combination" R.&M.1948, British A.R.C., Sept. 1941.
30. Buchek, P. M., et al., "Modern Control Techniques in Active Flutter
Suppression Using a Control Moment Gyro," (Stanford University;
NASA Grant NGL-05-020-498.) NASA CR-138494, 1973.
31. Cooley, J. W. and Tukey, J. W., '_n Algorithm for the Machine
Computation of Complex Fourier series," Digital Signal Processing,
Edited by Rabiner, L. R. and Rader_ C. M., IEEE Press, New York
1972, pp 223-227.
32. Golub, G. H. and Pereyra, V., "The Differentation of Pseudo-
Inverses and Nonlinear least squares problems whose variables
separate," SIAM J. Numer. Anal., Vol. lO, No. 2, April 1973.
33. Baker, G. A.,Jr., "The Pade Approximant Method and Some Related
Generalizations," The Pad6 Approximant in Theoritical Physics,
109
Edited by Baker, G. A., Jr. and Gammel, J. V., ist Ed., Academic
Press, New York, 1970.
34. Jones, W. P., "Stm_nary of Formulae and Notations Used in Two
Dimensional Derivative Theory," British A.R.C., R.&M. No. 1958,
April 1942.
35. Luke, Y. L., "Approximate Inversion of a class of Laplace Trans-
forms Applicable to Supersonic Flow Problems," Quart. J. of
Mech. and Appl. Math., Vol. 17, 1964, pp 97-103.
36. Kemp, N. H., "On the Lift and Circulation of Airfoils in Some
Unsteady-Flow Problems_" Journ. Aero. Sci., Vol. 19, No. lO,
Oct. 1952, pp 713-714.
37. Horlock, J. H., "Fluctuating Lift Forces on Aerofoils Moving
Through Transverse and Chordwise Gusts ," ASME Jour. Basic. Engg.,
Dec. 1968, pp 494-500.
38. Hewson-Browne, R. C., "The Oscillations of a Thick Aerofoil in
an Incompressible Flow," Quarterly J. of Mechanics and Applied
Mathematics, 16, 1963, pp 79-92.
39. Wood, L. C., "The Lift and Moment Acting on a Thick Aerofoil in
Unsteady Motion," Phil. Trans. A., Vol. 267, 1994, pp 131.
40. Chen, C. F. and Wirtz, R. A., "Second order theory for flow past
oscillating airfoils, " AIAA J ., Vol. 6 , No • 8 , 1968 , pp 1556-1562/
110
Dat, R. and Meurzec, J. L., "Exploitation par lissage mathe-
matique des mesures d'admittance d'um syste"me lin_are, " Extract
de la Recherche A_rospatiale, No. 1972-4, 1972, pP 209-215.
42. Miles, J. W., "Quasi-Stationary Thin Airfoil Theory," Jour. of
Aero. Sci., Vol. 16, 1949, pp 440.
43. Smilg, B., "The Instability of Pitching Oscillations of an Air-
foil in Subsonic Incompressible Potential Flow," Jour. Aero. Sci.,
Vol. 16, 1949, PP 691-696.
44. Garrick, I. E. and Rubi"ow, S. I., "Flutter and Oscillating Air-
Force Calculations for an Airfoil in Two Dimensional Supersonic
Flow," NACA Rept. No. 846, 1946.
Hassig, H. J., '_erodynamic Flutter Coefficients for an Airfoil
with Leading- and Trailing-Edge flaps in Two-Dimensional Super-
sonic Flow," Journ. Aero. Sci., Vol. 21, No. 2, Feb. 1954,
pp 131-132 •
46. Landahl, M. T., "Unsteady Transonic Flow," Pergamon Press, N.Y.,
1961.
47. Kuo, Y. H., "On the stability of two-dimensional smooth transonic
flows," Jour. Aero. Sci., vol. 18, 1951.
Pearcey, H. H., "The Aerodynamic Design of Section Shapes for
Swept Wings," Advances in Aeronautical Sciences, Vol. 3, London
1962.
111
49. Nieuwland, G. Y. and Spee, B. M._ "Transonic Airfoils: Recent
Developments in Theory, Experiment, and Design," Annual Review
of Fluid Mechanics, Vol. 5_ 1973, pp 119-150.
50. Jones_ W. P., "Calculation of Additional Mass and Inertia
Coefficients for Rectangular Plates in Still Air," R.&M. No.
1947, British A.R.C., Nov. 1941.
51. Bryson, A. E.,Jr., "Stability Derivatives for a slender missile
with application to a Wing-Body Vertical Tail Configurations,"
Journal of the Aeronautical Sciences, Vol. 20, No. 5, PP 297-308,
May 1953.
52. Bryson, A. E.,Jr., "Comment on the stability derivatives of a
wing-body-vertical tail cross section," Journal of the Aero-
nautical Sciences, Reader's Forum, Vol. 21, No. l, pp 59,
January 1954.
53. Bryson, A. E. ,Jr., "Evaluation of the Inertia Coefficients of the
Cross Section of a Slender Body," Readers Forum, Journal of
Aeronautical Sciences, Vol. 21, No. 6_ pp 424-427_ June 1954.
54. Van Spiegel, E., "Boundary value problems in lifting surface
theory," Technical Rept. W1, National Luchtvaartlaboratorium,
Amsterdam_ March 1958.
55. Benthem_ J. P. and Wouters, J. G._ "The calculation of aerody-
namic forces on the circular wing in unsteady incompressible
flow/' Nationaal Lucht-En Ruimtevaartlaboratorium, NLR-TN-W25,
1963.
112
56. Summa, M. J., "Potential Flow About Three Dimensional Stream-
lined Lifting Configurations, with Applications to Wings and
Rotors," AIAA, 13th Aerospace Sciences Meeting, Pasadena,
Calif., Jan., 1975-
57. Wolvovich, W. A., "The determination of state space representa-
tions of linear multivariable systems," Automatica, Vol. 9,
No. i, 1973.
58. Silverman, L., "Realizations of linear dynamical systems,"
IEEE Transactions on Automatic Control, Vol. AC-16, No. 6, 1971.
59. Gopinath, B., "On the control of linear multiple input-output
systems," The Bell System Technical Journal, Vol. 50, No. 3,
March 1971, PP 1063-1081.
60. Lyons, M. G., et al., "Control Law Synthesis and Sensor Design
for Active Flutter Suppression," AIAA paper No. 73-832, 1973.
1 . Hall, W. E., Jr. and Bryson, A. E.,Jr., "Synthesis of Hover
Autopolits for Rotary-Wing VTOL Aircraft," (Stanford University;
NASA Grant NAS2-5143.) NASA CR-132053, 1973.
62. Segenriech, S., 'Weight Optimization under Flutter Constraint,"
Doctoral Dissertation, Dept. of Aeronautics and Astronautics,
Stanford University, Jan._ 1975.
63. Drischler, J., "An Integral Equation Relating the General Time-
Dependent Lift and Downwash Distributions on Finite Wings in
Subsonic Flow," NASA, TN D-1521, 1963.
I13
64. Watkins, C. E., et al., "On the Kernel function of the Integral
Equations relating the Lift and Downwash distribution of
Oscillating Finite Wings in Subsonic Flow," NACA TN 3171,
January 1954.
6_ • Landahl, M. T., "Kernel Function for Nonplanar Oscillating
Surfaces in Subsonic Flow," AIAA Journal, Vol. 5, No. 5,
May 1967, pp 1045-1046.
66. Mangler, K. W., "Improper Integrals in Theoretical Aerodynamics,"
RAE Report Aero 2424, ARC 14394, June _ 1951.
67. Milne, R. D., '_symptotic Solution of Linear Stationary Integro-
differential Equations," British ARC, R.& M. Rept. No. 3548,
July 1966.
68. Runyan, H. L. and Watkins, C. E., "A systematic Kernel function
procedure for determining aerodynamic forces on oscillating or
steady finite wings at subsonic speeds," NASA Tech. Rept. No.
R-48, pp 8.
69. Muller, A., "Berechrung Der Druckverteilung an Schwingenden
Tragflugeln in uberschallstromung mittels eines kollokations-
verfahrens #ur die Kussnersche Integra!gleichung," Doktor-
Ingenieurs Dissertation, Der Technischen Universitat Munchen_
July, 1973.
70. Fettis, H. E., "An approximate method for the calculation of
non-stationary airforces at subsonic speeds," WADC Technical
Rept. No. 52-56, Wright Air Development Center, 1952.
114
71. Osborne, C., "Unsteady Thin-Airfoil Theory for Subsonic Flow,"
AIAA Journal, Vol. ii, No. 2, Feb. 1973, PP 205-209.
72. Kemp, N. H., "Lift and Moment for Arbitrary Power-Law Upwash
in Oscillating Subsonic Unsteady Thin-Airfoil Theory," AIAA
Journal, Vol. 12, No. 3, March 1974, PP 413-415.
73. Hsu, P. T., "Some Recent Developments in the Flutter Analysis
of Low-Aspect-Ratio Wings," Proceedings of the National
Specialists Meeting on Dynamics and Aeroelasticity at Fort
Worth, Texas, Nov. 6-7, 1958, pP 7-26.
74. Miles, J. W., "On Solving Subsonic Unsteady Flow Lifting Surface
Problems by Separating Variables," Jour. Aero. Sci., Vol. 21,
No. 6, June,1954, pp 427-428.
75. Zartarian, G. and Hsu, P. T., "Theoretical Studies on the Pre-
diction of Unsteady Supersonic Airloads on Elastic Wings,"
WADC Technical Report 56-97, Dec. 1955.
76. Moore, M. and Andrew, L., "Unsteady Aerodynamics for Advanced
Configuration, Part IV, Application of the Supersonic Mach
Box Method to Intersecting Planar Surfaces," FDL-TDR-64-152,
May 1965.
77. Rodden, W. P., Giesing, J. P. and Kalman, T. P., "New Develop-
ment and Applications of the Subsonic Doublet-Lattice Method
for Nonplanar Configurations," Symposium in Unsteady Aerodynamics
for Aeroelastic Analysis of Interfering Surface_ AGARD Conference
Proceedings No. 80-71, Part II, 1971.
115
78. Baker, G. A. ,Jr., "Essentials of Pads Approximants," Academic
Press_ New York, 1974.
79. Chisholm, J. S. R., "Pad6Approximants and Linear Integral
Equations," The Pad_ Approximant in Theoritical Physics,
Edited by Baker, G. A. Jr. and Gammel, J. V., ist Ed.,
Academic Press, New York, 1970.
80. Riesz, F. and B. St.-Nagy, "Functional Analysis/' Ungar Publishing
Co., New York, 1955.
116
.H
O C)b
.H cO 0
0_ C_
oo _
O U%
og
0
0 0
O
o
cO co
t,_
®o
:O O CH
0
O O
(]J
o-, _ _ _' O_1o _ o _,'_o o-,

o,,
.H
o.
CU 0,1 _ _ cO
ooo_0_88o_
_
_ 0 qD Ckl _
8°
oo8o
N_ N_
_
w'-,
li"
• • •
O O O O O O O O !_1 O O O O O O O O O
I I I
%
? 0
0
% CO CO
% ,D
CO .-d- NA m-I C_ N_ _O CO
+
_o _
_, _ o o o o °°, _ _' _' °, o 9 _ ° ° °,
.4-
N -H
% O
4_
kO 0 Oh Oh °H
, -_-
rt cO 0 _b-- ib-'-
r--I 8 cO
03 0 ,-I 0 0,I 0 OJ 0 0 O,J '0 r--I 0 ,--4 0
o
•H
N
O -H
co . _ _ o _ _ _ 0_ _ o . o,. o ._ o
I I I I I I
O
4_
O_
°H
O O cO O _ _0 co _£) 0J r-t
oH
o o o _" o o o o
_-I I I I I O
I
,Q
ID
%
(D
°H ,-4
U
(i)
O
117
r_
I::I
Q)
Q,I
O
_D
II)
_o _ O
f-I
Ib--
'I'¢X
_
--,',_
CO
',D
C_
GI
["-.-
_lP
CO
0d (_ OJ _ I._1'_ _0 :1_
0 II "I-
,H
oo,_ i_' o_'0, ,:_ o°

o 0_ o ,-, o
o_g_8
o o to o o, o o o,o o o, o o 6 o o o, o
-_c.+
,._°N
OJ
____ _ o_I o _._ _ -__
___
S'
_-
_,_8
_ao
°H °H .._'t-
,o
I
o o o o
I
o o
I
H to
I
o o
I
o o
I
t_ o o o
I
o
._._ _
!0",, b'- -_ Ch 03 O_
.-.-I" CO I_ I'G J _ '1_ r_

,--I iO _ O _ O .--if O GI O 0_ O 0,1 O _1 O _1 O
|t
O % °r'-0 / I I i I
•H O
4-) _--t
iO _ cO O _G::) i'_
?O
!o o _
o ._°' o ,-,
_
OJ
o
_ -_
_
_OJ
o
_
0
_0
0_
-el o o o o /_ o o !0 o
4-_
!
I °1
°.°
%-I
O
I:I
e
e_
II)
O
118
°_
t_
0
0
,-cl
o
t',-- ,-t co
o _'_, _ _. _0 _ _ 0-, o_ °
o U 4-
.,-I 008 o800 o_ 8
0 0 0 '0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
! I I
I 0 _
°_ _ _ _ ° o _ o. _ o
_o o, o 8 o o o o o o o o o, 0_ o o o o
O.r_ .._'F
o % .r-o
.r..I
_D ,o .O
0 ,-I
0 0 LC_ 0 .D
o cO 0 0 0 OJ 0 0 0 C_
._ I I I I
4_
.el OJ _ ,-4 Od C_ _
o
.el
ID
o
119
I:I
.,-I
o
.M
if,
0
0
0
i-1
o
0
%
O/ _0
---_ _ _0 -.-t cu P-- _D t'--
Oh Oh _d- OJ -_ -:_ _-- I) _ °h _ 0
_° +
-t- tc_ _o _ ___ _o_oo
oo8888oooo
o _oo88 o_o_o8:88
%
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 .0 0
I I I
A g_ + < _ o 0_
__ o_ --k
____o_
oo o_,_o_ _9_ _o___ __o
% ? oooo oo o o o oo, oo _ oo ?
4_
J
, _ %
,_ _o o_ :_ _ _ _ _o
0 _ 0 0 _ 0 i_ -_ 0 _'_ 0 _ iO 0 0 _ 0
_ i_ _ c_ _ _ _ _ _ o _ o _ o _ o
I I I I I
0 0
4-_
o@ o _W o _ _ _ k
0 oI o 6I o o_ ,S oI S 6I
I
=1
ul I
IL
(1)
.H r'l OJ i_ ,-'I 0.I _ ,--I OJ I_%
o
.v.t
120
Q)
.H
,D
.rl ,-I
cH 0
0
0 0
0
_.) 16 o
0
.el
,r-t LI%
-.=t"
O
,O O
O I
%
[_- _
O_ cO O'x r-I 00 0 0 OJ _ O_
cO _ _ 0J
qD :0 GI _ !O ::0
---# 0"_ _C) -if"
o m
,--I O 0 O O
.--t _ _ + '_ OJ 0
0 g 8 g :0 0 0 0 0 0 0 0 0 0
• • i •
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
I I I
0 'b- _ OJ
OJ !GI
:0 CU
-.-I- qD
,.D 0 _
LG ,-I
,--I ,b--
:0 _ .:.-.if _--I 0 ,--I 0 I'G 0 0 -.d-
,,__ d o d o o o o o o o o _ _o o o o
•.O % % I I I I I I
i-g : /
m oJ B_ O :r_ o-x _ ,o co
['.-
co .o
[_- !_::_ (M
BI -.-t- {_ [,-- o
[,--
,-4
_-
O ,-I O r._ O 'el O LG 0 p__O 20
O O co
oJ
O O CLI O
II
,_ ,_ m, ,_ aJ H ,H H H H o H o H o H o H
.r'_ I I I I I
O o .H
•_ _
4-_
o
?O
O ,-I_
#,
O
o
_
c;
0
o
_,
_
0_
0d
o
_
c;
0_.
CLI
o
,--I
,-t
o
or.t I I I I I
4.)
I
It_ I:::ll,.al:::l I:I , I:::I! I:I _ I:I I::I I:I I::I, I:I I_
o
_-I C_I _ ,--I CU _ r-I CU
.H
O
.r-I
g--t
I1)
O
121
.H
0
r+-i
0
0
II _0 12-- qD 0 oO CO O0
O +
+ _
o _ 0_ _',_ _ o _ o
_ $_,oO_O _ cO¢-,
_ _ o ,_ o ,o_ t¢_
",
% o o ,-) o 0_ 0_ !S o o o ,--, 0_ o o o o o 6
I I
<
o m
-4-
•H ._e+
%
4_
_, ua+
¢+
% %
L(_ -.-I- rt_ _ _ -1- --I" LCx Ckl
Lr_ 0 ckl b--- _Q _ ',.D
0 sO 0 r-.--t 0 _ 0 cO 0 <.) o 0 0 _o.I 0 o..I 0
II
_11" ,-.-.t od ,.-I _11 ,-I 0I H 0I r.-I 0 ,,.-t Oi ,....I 0 ,_t 0I i_..I
O O .H
•H r_
,.0
b- b-- 0_ c_ _-- O_
".0 _O cO sO _ aD
0 ',.O _O cO _ ',.O _'h
'.-0 0 _-I 0 cO ,-I ,-I ¢-I 0
o
.el _, o o, ,o oi o 6I o o
4_
O _ _ _ _ _ _ _ _i_ _ _ _ _
i
¢-I Od _ t-I CkJ _ ¢-I Od
¢-I ,-4 t-I O.I 0.1 Od _ _
,el
o
122
123
I:1
R-I
i
O
NA b-'- _ b'-"
0 II -I- ICX _-I OJ 04 ,,0 0_ _ OJ .-.-i- 30 C_ t--
o _ a_ ,_ _ _ o I_ ._- c_ _a
+ _a o _ ,_ _ _ o_ o _ o _ o _ o ,o_ o _ _ o
% o o _ o o o o 6 ,o 6 o o o o o o o o
!
Z Oh
,-I 0 _D
C_ _ '_D_ °O _O _ O _._ _ o _ _ o_ _ L_ !CO OJ

-i- 0,t r-I OJ _1 O LCX _ O _IEX O '.4D _1 CO Od cO
_ o _ _1 r-I .Od _1 _1 _ O O O © ,--I _OJ O O O O C_ ,-1
I I I I I I
O*rl ..._.-I-
o, _+
g
%
:o
.-1"
'_D _ ._ _ L_ OJ !_55. O4
cO C_ _ _ !Ckl O
_ _ o o o c0 o o o _io
o_ _o o o o !o_ a_ o
_ _ _ o _ o . o _ o _ o _ _o . o
I I I I ] I
0
*_I 0 .H
.OeH
D-
O O D.- OJ O Od x4D
Od _ O Od O -_'-_
O O r-I O _D r-I _1 ,-I O
O
.el
0_
I
o oI Io 6I o 6I o oI
4_
o_
O
124
..=I-
°H 0J
_ _ o
l,n O
0d I._
O O
0J .._
O _ _GI _
O O
GI
O O O O O C) O O
.H
.,-I
_ _ _
t.rX _, _O _
O o
0 A o, M- o_
o 8 o 8 o
O Oo o
O o
O o
O
o m r--I
O O O O O O O O
I I !
r-4
.H
E GI
O
0d
O
t'-- O
MS)
GI
O_
00
cO O
cO
cO O
o _ o o! _ o o o
H _
KD
% m O cO _ O O'x _
cO Ox
cO I.f'X _ ..-a _ O
0
• . . _ . . _ .
r"l OJ O O O O O O
I I
•H _ m
o__ o
•t-I _ 0 N_
4-_ I_ _ II O_ GI
',.0 ["- _ O'x _ O --_'"
o _ _ -_ _ o_'
•N _ _ dr_ _1 KO ,-I O _-I Lf'X N'X O
,--I O O O O O O O
!
_o 0 q-i
o_
0 0J
.r-_ r--I OJ _1 0J N'_ _ tr_
*el
0 •_ _--I r-I OJ 0.1 r-I ,--I GI GI

0 -H
•H 1:_
125
cO
cO C_ co _ Lr_ cO
0 cO
0
,-I oJ Lr_
0 0 6 o o 6 d d
°,...I
_L 0J cO
O'x
,-I
0,1 co OX
0 r-i uO 0J b- 0 o ,a ,_
,r"l 0 0 ,-I ,-I 04 0J ,-I (_l
Q
I:_ t" 0 0 0 0 0 0 0 0
.1-1
¢'-I rn
_D cO
I
,-I O_
OJ
0 (5 0 0 0 6 o 6
S "6 +
Lr_ cO
r._ o i- °_ _- o_ ....-t
Lr_
0J ct/
o 0 0 0 o 0 0 0
o_
r-4 oJ
ii o
• r'l _
o
_ M _o _
0 0
M
_ _ _0
_ , ,,
.el _o -P II
_ N
•r_ ,_ .,-I
.p 0
.,-I _" • II
126
0"_
Lr_
oooo_ _° 0
0 0 0 0 .-_ 0,1 ;-.-I 0
Bi
b-
kid
cU 0,1 O4 _ _1 r-I _1 OJ O_
o6o666o o 0
I-
I
0 _ _0 0 9
0 _ _ 0
0 0 0 0 _ _ _ 0
0 0 0 0 _ 0 0 0
,_ ¢--
I
r-t G) 0 _ 0 _ _ _
OJ
al • • •
0 0 0 0 0 0 0 0
• t'4 _
0 0 0 0 0 0 0 0 0
_ _ _ 0 0 0 0
6o_oooSo 0
II 0
M
r.-,
_ _ _ u _ _ ",
_o
III _,_ o_,i _ ,r,-P 5"
.I_ II _ II _ U • II
127
oo -_ _ o _
,--4
en oo
0.1 0,1 _
C_I o
OJ _
O0 0 0 0
co
o_ t"- tc_ ,-t co r4
8 0
(S)
! • Q
0 0 0 0 0
• G:)L
O2 I
m
O_
',D
OJ
!
,-.t 0 0 _1
0 0 0 0 c;
r-i
',,D
_ + ',.D
0 r-t
4_ t- 8 _ 8 o 0
0 0 0 0 0
I I
o
! .p
o
II • -_
,.Q "_ .rl 0
0
P_
• _,_- _,._
4._ .,-4
°_ •r--I °r-t .!
128
oJ 0 o_
C_
r,rl 0 D.--
,-4 o
r-I
i'n
O'_ ff_ b-" ,-4
.H la_ o S (5 6
hO
t-
OJ
I
•,-4 4-> oJ Oh OJ
m O N-h 0 .-.M-
4-- S _ S
t"-
.-d-
! <O
O h'h OJ
I t_ _ _q o_-
_1 .r-I ,-I Oh -ul- 0
"b I¢_ .-d'- 0 0
Ob--O
_ ° _ ,-I 0 0 0
_q OO t- I I I I
v
,--I
--I-
0
co ko
N @ t.- 0 ,-I 0
v
0 L?X 0 0
rl
,-4
N _ II
O m
.r-_ ,--I
.M
OJ
,--I
O 0J b-- 0 0 0
4._ L_X 0 0 0
I I
-el
_.r-t
•r-_ r-I 0.1 ,'-t 0.1
O
0_ O_
t-t
129
_D
U_ O
u_ .-.'t
t¢'3
u'_ 0
0 0
•_I 0 0
0 0
< o 0 O 0 O 0 O O O O
0
0 c_ c; O 6 6 o c_ o
.,l_,l
0
O_
"4 0 o_ 0
•_ 0 0 0
0 0
O 0 O 0 0 0 O
< S S o S S
c_ 0 I I
0
tD
r_
.-.1" o',
m. O_
c-,I *-.1"
7 °_ _0 O
0
,_--_ 0 0 0 0 0 0 0 0 0 O
< c; S o o S o c;
O_ I I
0 I
_J O_
< 0 P_ °_
•H _--_ ,.D -..1" O
c_
v c¢3 ',.D P,_ 0
<
O O O 0 0 0
X 0 < 0 0 0 0 O O 0 0 0 O
0 _J
U_ Dq II c; c; c; c;
,r-_ ! t I I
0
Z II
0
0 c'.l _ _ .....1-
o.,I _ _ u_ ...1- ,DO u_
,--, O0 00 '...O rm 0 00 uh 0 -,...1"
O0 eq _ r-,. 0 e_ O_ ,,.-1" 00 o
0 •_ 0 _ c-.I 0 0 om r.-- o", -.:1- 0
0 0 0 0 0 cq o_ r'_ 0 0
•_ 0 0 0 0 0 0 0 0 0 0
0
I I 1 I I I
q;
c; q_ L_
"r_
O_
E.) _J
o,i
0
130
0
,._
4._
0
,-I
t_
.r-I II 0 0 _ .-.-t- u_
4.a ,-I _ kD C_ 00 o'_ 0
c_ C_ ,-I o0,-_ ,--_ Lm,

r-'t .r'_ • + • • •
.M
-.4 o _, _ _ c_o
4._ I ,-I I I _ c_J I c',_ r--
i-4 r-'-I I
,'-I !
II
0
0 r-t L_ _ cO _-- _.-0 u"_ u'_
r_ L_ O_ cO _ C_'_ o"_ r_
Q)
I CLI ,_ I cO r-I I
I I I
t_
II 0 OJ _ cO ',.0
,--4
,
M _m, 0", e--I r-I _I oo oo

c0 i_, cA oI ,.6
I -4
_ _S&_:
_ I. r,'_ r.-.
(Yx ,--I ,--I I
+ I
,-I II t_
0 CO oD c,_
,-I
,r,_ cO O_ _ _ r.-le,,i c_ c,'_
0 .r-o _0 03 L"-- _ _ ,-lu'_ o4.,_
II .M • + . • •
c0 _ & _ _ oooo
°r.._ I L'%] f'('% I CO OJ l, ,--_ u_
I I I
d_
N_
0
0 -_" "_ _J_
'L'_- _ r--I _ D-- C_.. o.] ,-_
o II ,--1 OJ _-I ,--I rc_ 0 ,-_ _'_' ,----+
r--I "r-'o • . . . +
.r-t
CO
N_ I+ c"_ t"_
r--t rt I
N !
.el
r-t
+o
II ,.-t
t_ U'_
b-- tO',
0.1 D_ _ 0N_ 0_',£ _,--+
.._ ..,_-
0 .r-o _:) 0J r_ Lr_ Lr'X 0 u'_ 0 00
.H
_, o& &
_
_I _
CO
_oo_
,,--"1 I, '-_ u"_
I I I
<1)
•ca _-I CLI _ ,-4 Od l,r_,-_ c,i o_
I
.el
•M _I ,.-I r---I OJ OJ OJ rm _ <"_
131
o
kD ,-4
O 4
II _0 O L_ k.C) OJ O_
I'_ ..-., ;.--I ',..O ;-.-...I-4- C_ &
..t-0
4 I ,-.4 I I o..I I>-
I I I
,.q
o
.M
Lr'x B_ I1_ Lr'x CU B'_ _ O_ r-_
(:_ rE", O.I O i,_ kO (O OJ LG
I I I
°r-_
LFX
CQ ',D ',.O _'_ b-

O
;--I
or.-I -:1- CU _ cO _ H _ {I3
'_'0_ B_ C_ b-- _-_ H L_ r--I
I ,,'-I I b A I (lJ ['--
,d ! I I
,._
o c_ _ o ,-4 oO C) o O cO
..p -._ _ ',43 '.,D O s_ _..D h-",
O_ i'G O _ Lr_ LF', O LP_ O'_

,--I _ _'_ ,--4 _ h-
I 1 I I
"d
ID
t_
,H *r_ ,--I OJ _ ,--4 (21 :,,'_ ,--I oJ _
H
(1)
.H ,--I H i--I O.J OJ OJ [4% _,
(1)
o
132
r-I
cO C_ r_ _0
M_ I_ O_ 0 CLI I_ cO C_
I1
M? o
0 0 _
0 0 kO ,-I _
o 0 0 0 0 ,-I 0 0
.,-I
o ..:I-
+_.M C_
0 ,-.I
O_ 0
o Ckl 0 ,--I ,-I _ _1
t'- 0 ,-I 0 Ifx ID- 0.1 ou
rm _ Od
0 0 0 _ 0 0 ,--I 0
0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0
0°_ I I I
0 "_J m_l ..1_ 0J

Lr'X cO 0 If_
0 co Lr_ 0 C_ _'X CO
r.-I _ _U'N _ o_ _-
g'_ O _ o_
0"_
sN_
.M 0 OJ i'_ 0 cO i-I 0 0 .-d-
f1_ 0 r-.t 0 0 O 0 0 0 0 0 0
F- 0 0 0 0 0 0 0
,-I 0 0
! I I I I I
CD'_ 0
0 0
CO
b--
-_" OJ 0 CO 0
t"-
00c.O -r-_ 0 g'X _-I ,-I 0 O_
,-4 ,-4 O.I 0 ,-I _ ,-I _ ,-t
0 0 0 0 ,-I 0 0 0 ,-I
0
I
0 o o o o o o ?
N
o,-4
% o
o
_t Od
4_
.el
H C_ CLI C'kl I_ _
o
.r..i
4-_
133
Stream Velocity
V = Free
I
I
I
I
I
7
9
I
1
I
I
I
Fig. i.
Swept wing with two flaps and nine node points_
134
0
F_
II
m °
0 o
.r-I
c_
o
o
o
C_
O0 _C> _d
q
0 d c_ c_
135
t
o
,-4
(:5
136
c5
N
0
C_ !
-r-t
o
,%
%
r-_
.el
r_
o
I
d
CD
CD 0
c5 0 c5
I
137
0
i
.H
._
0
0
d 0
d
!
_OUU3
C)
c5
139
0
_4
4_
N
.,-4
N
O
%
I1)
4_
O
.el
4-_
ej
z
I1)
.r't
u_C:3
q_-q-b:l_L g
l
140
CD
CD
CD
c_
O3
"ID
I.--
Or)
LLI_ (D
Z
0 (D
>
0
I,_.
°_
05
_0
,,-4
c:_
C_ I_0 C_
I
141
{D
¢1
rD
I=
o
r_2
r-I
-r-I
0
d,
CD
0 LC_ 0
CD
n aja-q.b-.l _
"_ _,,t!1
142
(-_
,cl
t_
o
,._
_D
_D
o'_
0
(D
_D
c;
DaJD-q-b :EL
143
! LC)
I
I
I .4--" 8
0
II
I E
i 0 2
U'}
O3 E_
I
I o
I 0
I t'"
0 -r-I
I I '1o
I
I QJ _0
O3
,-4
;>
.r-I
_-_
._
.r-I
0
r.-.t
.el
CD
., LLb
144
t t l
I tit
I lit
t Ill
I Ill
L'--
0 000
If II It II l|
,-..I
°r-f
.r.-i
._
c_
g
°,-I
%
0
4_
o_
,--t
.el
°_
o;
0
0
-.,4
145
C)
i I
I I
I I I
I I III
I I III
l I
I III
II
(i)
0 d d o_o
n II II II II II
/
/
/
-CD or)
-0
i,m
0 0
-,-4
f- .._
(._
o °_
!
.._
0
E
(1) ,-q
CH
of)
02
g
(1) .r-i
0
a_ 4._
>
0
I..-
4_
q_
(1) .r-t
,--t
("xl C.)
c" ,-4
0
.r-I
"0
r-'t
.r"t
I
(Z3
I
!
146
I
O3
0
r
0
I
c-
°_
"U
,--t
> .H
0
c51 o c5_ c5_ c5_
_%---.-
C) r-4
C
o
ID
_4
r-t
.r-t
i i
I I !
147
0
,@
,_
*--4
.el
148
d
00
!-,,-
-cO _
"ID
I,,-,.
0
r-
!
E
rd")
F: 0
Cl r.t-i
I 0
..p
I
I >
I U
I.
I
I 0
-0,1
.r.I
0
-H
U'I
H
,.6
.el
_,lqp---__ __
149
f...
(,/')
C_
..........Lcb
150
CD
! I
!
151
0
b..i i i'
.I
II I
i-i +-.
0 iW._
i-- C_iO
i.--
"0 I__ o
0 o r-
E- v _r_
0
r
,0. C_
C_. IX') _ "<I" CO E

G_
Ci _ 0 0 0 CZ}
II II II II II
r 0
._
\ (lJ r-4
>
\\
\\\,
C
0 0
\ m
°_
.M
-0
0
e,4
0
_A
.r,l
...q.
"0
t'N ll==.-- 0
c5 0
I
IN3NON
152
I f I t I w C_
["-4 _Ot L..-"_i-. r_lC _
ul ' I
I I .! I
11
o
.,.-]
4._
0
%
q)
,,-t
.rt
u?
0]
I1)
%
g.-t
o
i1)
o
,--t
d d
il
I
I
i
!
i l !
153
0
II
c_
A
I
o
r-I
o
C_)
!
,m
\
\ E I-I
\
\
\
\ C
\ 9
\ "0
\
0
\
>
0
.r-I
,4
c ,H
0
0 .H
7O
I
I ed
I
.H
r_
c:_ q_ "b'q c_
_!7
154
C3
I
Ol
4_
O3
ol
II
,.-4
j--
u3
"I0
L..
0
0
(-
(3
I
.r--I
or)
r-t
r-
._
0)
.r't
Q_ _'J
Q)
>
0
._
o
4-_
C
,o
0
,--t
"o
I,-I
&
G.I
| i
I i , , CD
O_ co
c,4 C,4
c5 d
DajD.q:bz_
155
C)
o
H
©
rt--_
g
.r-I
(U
,r..i
c_
,,--_
L_
__-T)
156
C_
g
-00
t
r
t-t
E
(IJ
L. II
.rt
_.4
e-I
o
r
o_
"0
0
CO
0
c_ c_
157
C3
14
L_
*r.'l
g
CO
0
i--
U3
0 ¢j
o_
1:3.. r-t
U3 o
!
°r,I
l"-
°_
",4"
(D
0
1:3
4._
0 llJ II
C
0
,el o
,_ e_
"0
r-I
I-4
Ol
o | 0
O0
CD 0
c_ 0
158
,r-I
(.) 4a
E oE ,r-I
i-4
f..4
CL
_c_ 4a
0
Om
0
r._
IZ)
Pc 0
0
II (.)
43
If}
&
II
(-
U')
r,..)
\
\ o
cl
0 o
,r.i
,5
0.1
0
Q
0
q'll_
_19 "A
159
-'-II I
r_ t
-p
I!:I
II '
I
o
I
I
I p_
0
r- I
E
o-- o
X
0
I-.
r'_
CI
Q_ .el
.el
0
r- --)
I--'
•el II
rY
E o :E
.p
Q) 0
r- -_
0
u'1
I.. • ff'l •el r---t
CL
I i r-I 0
•rt 4-_
i ' I
t
0
*el
_
.el
i ' I
t
i ' I
Cd
CD
160
6
(223
,k--"- U
_2
o
£.C, ,-4
II
q_ O
.,-I
4-_
p._
r..4
r-I
0..)
.rt
©
O_
U')
O "O
E ._.__.__-
._.-_
___o
or_E__'-_
° I I / O
r"
Q.)
rj .H
I a3
or)
E ,Q
o
I O3 t_0
E -tO
I C2
,,i _ _ , \ °_
I ,.-4
I
I .r-t
m
r_
.el
¢.)
.el
I-'4
_d
Od
'.,'O
i
I,O
O
D ajD.q-b-ll _; c5
_,;t'1
161
c5 o c;
DaJ D-q-b-_
_t!-1
162
0
C'J
I!
II
II
6
t
"U
II
0
r- P_
U'}l
G.)I o
9-4
f=%i
_=.Jl
r- I r-_
Q21 if)
L..
cl o
\
/ E S-.
0
r
o
e
if)
121
I
I
e-g ._
1 ,--4
I
r- ?-
.el
\ / '
I
\ \ la)
_ o
\
\
(1,) "_
0
mt
•4"-" .el
.m_ _
_ H
._
!
cO
0 c_ cJ
DaJD'q'b'l I _
_!7
163
(2)
0,1
C)
r-4
hi?
c_
0
bJ5
r-t
-r-I
4D
.r--t
,.--t
,--t
°r-t
.,-t
4_
ps
6,
qd
_-_
r-4
.r-I
164
U_
I
I
I
I
I
I
&
I I ' I
Da.ID" _-b-J.L _
165
OJ
Or)
r-_
n_
tr) b_
c-
-cq _)_
U')
!
E
(b
bO
C
Q_
>
--,r- CI_
Q_
(.)
c-
*_
U
S
-G _0
hi?
M3
II
r-t
C)
co (Z)
d_
*el
166
_4
CD
¢J
_4
>
E 0
>-
| °_
e _ S
\ r-
0
o
0 II
°_ o
_ o
_4
0 0
Q
c5
167
E
o
c_
O
If
4O
DI
\ ,r,P
,el
o_
0
4._
r4
r-I
o
c_
168
!
_4
oJ
I
I o
I
H
I
I:z;
E-_
I
I 0
II
.t"
(.h oD
(D
H
0
,<
E
0
"d
¢.
Q)
h0
.,'-I
(b
;>
.rq
_YJ
.r-t
%
©
,r'-t
,-t
,---t
°r-t
¢.)
\ .,-t
"u
,.d
\
"ELastic
Axis
2O 2O
5O
Ref. Semi-chord: 11.985
Fig. 37. Swept wing, AR = 4.16, A = 45 ° , M = 0.6
170
tO "o
aJ Or) o
X
0 X
..Q 0
'0
r_
,.Q
aJ -p
CO
.,-.I
-0
L_ C
0 C]
r c__
o Or)
to "0
f.--
4-_ o o 0
o ¢... .o
0
I
CO
,cl
(L)
or)
,S
o
I--.
.el II
0
r- 4_
°_
o
"0
II
a) .x H
,....-..e
(L)
%
0 o
L.-
J
cO
o II
II II °_
m o
C) C) t-
Z O
.4--* .r-.t
(/1
,r-I
"0 e)
u?
ci)
CD .el
CD
0 CD
o C_
171
",4"
O
.H
0_
4_
-rH
O
.H
F:
r,.n C)
c_ u3
t
I,.,,.,-
! O nd
I
I I
._
._
I
I
-c,q E '-do
O
.m @ LF_
ng
C_J
C_
I,,,-..
O :
O _ bid
t-
_-t -H
CJ
,r-i
ug o 4_
H _
uo
CD
C,O
O O
_ ........
2Lb
t72
-'..1"
E I.-
0
0
0
c-
c.)
I
Qm
-e,,Il:z
l,/'l
>
I,..
o
"qt---"
o
or.I
r-
D
Q_ ,'-t
.el
e,..l
.el
C_
0 °_
c_
0
Q
etb q
173
0
E
Cl
I..
O
om
4_
Q.I
0
0
.rl
,m .a-i
i.--i
C_
i.1"
CD 0
CD C_
174
"<1"
o CO
E
I,/'1
I
I
I 0
I r"
I
I
I
CIJ
ON
"U
O4
O
i.- o
,,-I
.p
(J
q.-i
r..- ,--4
°_
ed
°_
grl
om
°_
m C_
0 O_ GO
0 CD
0 0 CD
t
175
,,,1-
CO
0
.0--"
CL
E
>,
1
I--
U_
I
I O
I
I o4rJ
I
I
I
r-
om
OJ
G.I i,4
0
_-_
>
El 0
i. 4_
G.I
E
El
._
I--t
,A
tO
41 ,
!
O
bO CD
E---- C)
CD CD C_
Io i• i °
176
o 1
O_
E
"tD
I,--,,
0
o
I c-
I o
I
I
I
I or)
i-t
"0
O_
(ll
(_ 0
0
I.- ._
(_
0
i--I
C
°_
cJ
._
"0
Im
H
"o
.r-t
I CD
177
I
*-'I
"0
L.
0
c"
(_)
I
©
q-t
o
°r"'l
..t.n
r...4
-_I
°M
°_
l
I
CO
0 C_
c5 0
c5
178
"-.I"
|1
I
I
I
I
I
co
1
0
'k'_l m
¢-.
_-_
(1;
0
>
0
0 °_
4J
_J
U
C: .r-I
0
ID .r-I
_3
:-6
I 1 , -'0
I_0 o
0
c5 oaJD-_q-b_:uZ=_:_: c:;
I
luawolg
179
rree Vibration Continuous
i
Mode. Loads For I......dl',,todeL For 1
Anatysis i
I/ i
Shapes s.n.t,,_. I IA_o. L_,:,od_
1
i
't
IL-
'
I
) I
'
l
FLutter-
I
Anatysis _J
Dynamic
Modet
Synthesis
{
Controt
Law
Synthesis
©
Contro[ Laws For Flutter
Suppression
Fig. 47. Block diagram for control law synt_-esis procedure.
180
_U.S. GOVERNMENT PRINTING OFFICE: 1977 - 73b-004/31

Ra - Ja. W?a: Finite State Modeling OF Aeroelastic Systems

Uploaded by

Copyright:

Available Formats

Ra - Ja. W?a: Finite State Modeling OF Aeroelastic Systems

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Ra - Ja. W?a: Finite State Modeling OF Aeroelastic Systems

Uploaded by

Copyright:

Available Formats

NASA CONTRACTOR NASA CR-2779

FINITE STATE MODELING

Stanford, Calif. 94305

for Langley Research Center

NATIONAL AERONAUTICSAND SPACE ADMINISTRATION • WASHINGTON, D. C. • FEBRUARY 1977

7. Authorls) 8. Performing Organtzation Report No,

Ran jan Vepa

9, Performing Organization Name and Address

Stanford University '11. Contract or Grant No.

Stanford, California 94305 NGL 05-020-243

12. S_nsoring Agency Name and Address

Adapted from Ph.D. Dissertation, May 1975

A general theory of finite state modeling of aerodynamic loads

The modeling technique is applied to several two-dimensional

117. Key Words (Suggested by Authorls)) 18. Distribution Statement

Unclassified Unclass i f ied 188 $7.00

* For sale by the National Technical InformationService, Springfield, Virginia 22161

General l)c's_-r-il_t io_. ,_! !i_. l'r_,t)tem .................. 2

Rev icw _) Pc_-ctincnt i[ _:]',][ill t . ..................... 3

,qllnl!)lnF,/ i (kn_,t C[_lt[ i_.i S ............................ 5

FINLTE SfA'[F, W)l)l'!,!h',; c_i ; l!:i:_ t[ _2h;....................... 10

Finite }<]em_mt bI,)de!>; ............................... 10

[dealizati(m u_; a th':_m K,M .......................... ll

FINITE S'I'AI'I] blI)l)['_l_I2q(: OV :'_t'R_'_IP,'N:\H[(- l,(-!Al)_q ............... 17

'['wo I)imells[ulllll A 1 l(_i [:; ............................ 17

[n(t_mprt'_s[l_ e t tow ............................ 17

S!t})_:()Ili, _ I(_.,.: .................................. 34

,qlltS(__V-,,)[i]C _d¢t kt'_llta;_,llJ(" f](IW .................. 40

Tllree 1)_mensioilttl i,ii k i'n,4 Snvfcwes .................. 44

SUGGESTIONS FOR F[:'I'irRt" k'l<<I,?ARr:ll .......................... 59

CONCLIlI) [NG Rt:tMAI{KS ....................................... 64

APPENt)]X A - [Ni,'i l_t,iP,l_:i] (i<_i'i,'!_:lt.r_'i's !-oI_ A BEAH-ROI) WITH A

APPENDEX I{ I[!TE(;b:Ai, F.t_t'A*I'I{3NS 1;'{)I< IINSgTIEAI)Y At:.R()I)YNAM[CS 66

APPENDIX (1 - t:Ai(;[ I._'T[')!; ('_ \! !<',!ii'/kA!11('. f,()AI)S FOR SIMPLE

APPENt)tX b - PAl}l{ ..\i'}'}<tJ;tl_!,,'_;i':¢ ANI_ i5!II2(;RAI, EQUATIONS ... 98

R[-FERI<NCES .............................................. 106

i. Coefficients of Rational Function Approximations to

, Coefficients of Rational Function Approximations to

o Coefficients of Rational Function Approximations to

o Coefficients of Rational Function Approximations to

5_ Coefficients of Rational Function Approximations to

, Coefficients of Rational Function Approximations to

, Coefficients of Rational Function Approximations to

8_ Coefficients of Rational Function Approximations to

9_ Rational function approximations of generalized aero-

lO. Indicial lift in plunging for circular and elliptic

Indicial lift in plunging .............................. 127

Other indicial functions in incompressible flow ........ 128

15. Indicial functions for generalized forces for impulsive

14. Coefficients of Polynomial Approximations to Mode Shapes

16. Generalized piston theory loads using lattice

17. Rational function approximations to generalized loads

2. Theodorsen's function C(k) _ i _ i ik) ............. I {5

3. _(s) for rcal va]ue_ _,,f_ ........................... l_

4. Error in F(k) i'Tru_ - Ap_,r<_-:j_:-_J : ..................... I _i

5. Error in _'(Ii) ';_ ,'u, - App!:'o:<i:r" : ................... 1 _'S

6. Error in _('s) .................................... 139

7. Wagner' := ffunc. Lion k[ (t) ,:ob+ ai m-_ :_rd_:__x:i'mat, ely

_° Error in K(t) ....................................... t4!