Vol. 49, No. 4, December2003 O Vol. 49.

no 4, <lecembre2003

Methodfor FlutterAero-servoelastic
Open
LoopAnalysis

Ruxandra Mihaela Botez * Alexandre Doin * Diallel Eddine Biskri x lulian Cotoi * Dina Hamza * Petrisor Parvu *

NonTEncLATURE
Abstract
Aero-servoelasticity(ASE) is a multi-disciplinarystudy of 0) naturalfiequcncy
interactions among structural dynamics, unsteady n generalizedcoordinatesas functions of frequency
aerodynamics,andcontrol systems. In this paper, the n(co): [n.(co)n, (ro)l1.1(o)]
Aircraft Test Model (ATM) developed by the NASA generalizedcoordinatesfor elasticmodes
l.
Dryden Flight ResearchCenteris used,and the velocitiesat
I' gcneralizedcoordinatesfor rigid rnodes
which flutter occurs are calculatedby use of the Structural
Analysis Routines(STARS) aero-servoelastic software. la generalizcdcoordinatesfor control modes
For the validation of our aero-servoelasticstudy, the C] non-di mensi onal general i zed coord inat es, wit h
STARS aero-servoelasticsoftware, also developed by respectto tirle
NASA, is used.We developeda new aero-servoelastic tool non-dirncnsionalseneralizcdcoordinatesfbr elastic
4c
in Matlab to consider these interactions,and the results rnodes
obtainedthrough our method are comparedwith the ones
4r non-dirnensionalgeneralizedcoordinatesfbr rigid
obtainedthrough STARS.
modes
Key words: aerodynamics. aeroelasticity,
servo-controls,aero-servoelasticity. Qa seneralizedcoordinatesfbr control
non-di rnensi onal
modes
R6sum6 M modal incrtia or massmatrix
L'a6roservo6lasticite(ASE) est I'objet d'une etude D m o d a ld a r n p i n gm a t r i x
multidisciplinaire des interactions entre la dynamique 6r modal dampingcoefficients
structurale,l'adrodynamiqueinstable et les systemesde
K rnodalelasticstiffnessmatrix
commande.Dans le prdsentdocument,le matdriel utilizd
est la maquette d'adronef mise au point par le Dryden a modal generalizedaerodynarnicfbrccs
Flight Research Center de la NASA. Les vitesses Qr imaginary part of modal generalizcdaerodynamic
auxquellesle flottementseproduit sontcalculdesau moyen forccs matrix
du logiciel d'adroservoelasticitdSTARS. real part of rnodal gencralizedaerodynamic forces
Qx
Nous avons egalement utilize le logiciel rnatrix
d'adroservodlasticite STARS mis au point par la NASA
pour validernotre6tudesur I'adroservodlasticitd,
puis nous o modal sensormatrix
avonselabordun nouvel outil d'aeroservodlasticite dansle -.1s sensorlocations
< Matlab )) pour analyserces interactions.Les r6sultats p true air density
obtenuspar le biais de notre mdthodesontcompar6sir ceux rcl -crcncc
ai r densi ty
Po
qui ont 6td obtenusir I'aide du logiciel STARS.
o airdensityratioo:o= P (l)
Mots clds .' Aerodynamique, adrodlasticite,
Po
servocommande,a6roservoelasticite. V truc airspeed
vo referencetrue airspeed
tr/E equivalentairspeedVr,:.[oV (2)
* Ecole de technologicsupcrieurc
Departcmcntdc genie dc la production ai rspccdrati ov:Lt (3)
vo
automatizeie
1 1 0 0 ,r u c N o t r c D a m e o u c s t ,M o n t r e a l ,Q C
H3C lK3. Canada.
Qa dynamicprcssureqd :1,,, (4)
wing chord lcngth
E-rnail: rr"rxandra(z)gpa.etsmtl.ca h semi-chord,b : cl2

Rcccived7 August 2003

i!r)2003 CASI
 CanadianAeronauticsand SpaceJournal Joumal aeronautique
et spatialdu Canada

k f r equ e n c yk : Y -' b
r edr - r c ed (5) wind tunnel test program; generic hypersonic vehicles;
2VV benchmarkactive controls testing project; high-speedcivil
Mach Mach number transport;etc.
Anothcr software fbr developing a statc space model
rcprcscntation of a flexible aircraft for usc in an
INrnooucrtoN aero-servoelasticanalysis has also becn presented.This
techni quei s basedon determi ni ngan equi valcntsyst cmt o
In this paper, a rnethod for open-loop fluttcr
matchthe transf-er-function lrequcncyresponse. The theoryhas
acro-servoelasticanalysis is presented. This rnethod is
been irnplernented in a computercode called FAMUSS at the
validatedon the Aircraft TestModel (ATM) with the aid of the
McDonnell Douglas Aircrafl Cornpany.FAMUSS has been
Structural Analysis Routincs (STARS) cornpLlterprogram
used internallyat McDonnell Douglason its aircraft.
developedat NASA Drydcn Flight ResearchCcnterby Gupta
Recently,an acroelasticcode,ZAERO, hasbeendeveloped
(teet).
atZona Technologyby Chenet al. (2002),which couldbc used
Two main bibliographicalresearchthcrncsarc considered
also for aero-servoelastic analyscs.The influence of thc
here.Thc first one givesa shortreviewof the aero-servoelastic
acrodynamicstores on the aeroelasticinstability has bcen
analysissoftwarc in the literature.and the second gives a
studiedusing a number of aerodynamicmodels for the F-16
reviewof unsteadyaerodynamicforcesapprorirnationrnethods
aircraftconfigurations. includingthe isolatcdwing-tip launchcr
fiom tlie reduced frequencyfr dornain into the Laplacc .r
rnodcl. and the rvhole aircraft with and without stores.The
dornain.
resultsshou,good agreementbetweenthe prcscntnumcrical
A numberof aero-servoclastic analysissoftwaretools exist
predictionsand the flight-flr-rttcr tcst data.
in the aerospacc industry,rnainlyin the U.S.A.Thesetools arc
A short rcvioi' of eristing aero-servoelastictools is
STARSdevelopedby Cupta(1991)at the NASA DrydenFlight
presentedhere. Nert. a rcvic$, of the aerodynamicforces
ResearchCentcr;the Analog and Digital Aero-servoelasticity
approxirnation fl'orn the fi'equency dornain to the Laplace
Me th od( A DA M ) dev c l o p c db y N o l l e t a l . (1 9 8 6 )a t th c Fl i ght
dornai ni s gi vcn. Thc fol l ol r' i ngrnethods:l eastsquar es( LS) ,
Dynarnics Laboratory Air Force Wright Aeronautical
rratrix Pade(MP). and rninin-rurn state(MS) arc ncccssaryfor
Laboratories (AFWAL); Intcraction of Structures.
aero-servoelastic analyses.
Acrodynamics,and Controls(ISAC) dcvelopedby Adarnsand
The conventional LS method has been used in the
Ho a d l c y ( 1993) at N A SA L a n g l e y ; a n d F l e x i b l e A i rcrafi
acro-servoelastic computerprogramcalled ADAM. where its
Mo d e l ingUs ingS t at eSp a c e(F A M U S S)d e v c l o p e d b y P i tt and
capability is to determinePad6approximationsof any order
Goodman(1992) at the McDonncll DoLrglasCiornpany.
suchthat the sun-rof the numeratorand denominatortermsdocs
ln this paper,we usc STARS among thc cxisting software
n o t e x c c c d1 5 .
tools for aero-servoelastic analysis.The STARS prograrnis
Thc state spacc equatior-rs includc augmentedstatesthat
designedas an efficienttool for analyzingpractical cngincering
representthc acrodynamiclagsl their nutnberis dependenton
problerns and (or) supporting relevant research and
the nurnberof denominatorrootsin the rationalapproximation.
developrncnt activitics,and it hasan interfacewith NASTRAN,
The aerodynamicentire matrix has been approximatedby a
which is still very much used in the aerospaceindustry.
ratio of matrix polynornials.In tlie MP approximationrnethod
STARS has been applied to various projccts such as the
by Roger et al. ( 1975),each tcrm of the aerodynamicrnatrix
X-29A, F-18 High Alpha RcsearchVehicle/ThrustVectoring
rnay be approxirnated by a polynornialratio in s. Howevcr,it
Control System,B-52lPcgasus, GenericHypersonics,National
has also bcen found that comfflondenominatorroots are also
Acro Spac eP lane( NA S P ),S R -7l /H y p e rs o n i cL a u n c hV ehi cl e,
eflective in defining the correspondingpolynomials.Other
and High SpeedCivil Transport.
rnodificationsof the MP rnethodwcre suggcstedby Karpel
Another softwareused for aero-servoelasticity analysesis
(1982) and D unn (1980).A hi ghcr number of denom inat or
thc Analog and Digital Aero-scrvoclasticity Mcthod (ADAM)
roots is requiredin the MS approxirnationmethodby Karpel
computer program, which was developed at The Flight
(1990),whcrc the nurnberof augmentedstatesis equalto the
Dynamics Laboratory. ADAM has bcen applied on the
nurnberof denominatorroots.
unaugmentedX-29 A and the following two wind-tunnel
The capabilities for enforcing or relaxing equality
m o d e l s :( l ) t h e F D L r n o d e l ( Y F - 1 7 ) t e s t e di n t h e N A S A
constraintswere includcdin thc LS, MP, and MS rncthodsby
Langley l6 ft transonicdynamicstunncl and (2) the Forward
Ti fl bny and A darns (19U 4, 1987).These cap abilit icswer e
Swept Wing FSW model mountedin the AFIT 5 ft subsonic
abbreviatcdELS, EMP, and EMS, and they were introducedin
w i n d tunnc l ( l f t - 30 .4 8c m).
thc aero-servoel aslctl i c ntputerprograrnIS A C . Tlic r ninir nur n
Thc softwarecalled ISACI(The Interactionof Structures,
stateapproach(MIST) was selectedrecentlyin the ASTROS
Aerodynatnics,and Controls) was developedat the NASA
computerprogramby Chen et al. (1917).This rnethodoffcrs
LangleyRcsearchCcntcr.ISAC hasbeenusedon variousflight
savingsin the numberof addedstateswith little or no penaltyin
rn o d e l ss uc has DA S T A R W -1 a n dA R W -2 ;D C -1 0w i n d-tunnel
thc accuracyof rnodellingthc acrodyuamicforces.Howcver,its
flutter rnodel;genericX-wing feasibilitystudics;analysesof
e l a sti c ,oblique- winga i rc ra ft;Ap ' W (A c ti v e F l c x i b l e W i ng)

180 r) 2003CASI
 Vol. 49, No. 4, December2003 I Vol. 49, no 4, decembre2003

applicabilityto the unsteadyaerodynamics in the transonicand wherc n(0) is the initial value of a generalizedcoordinates
hypersonicregirnesremainsto be established. vector;),and Vrare,respectively, the vectorof eigenvalues and
As all thesesoftwaresand theorieswcre mainly developed the matrix of cigcnvcctors associatedwith the system
i n th e U. S . A . .t he nc c d fo r a th e o re ti c aal e ro -s e rv o el asti c
tool represented by Equation(6); and r represcntsthc timc.
also existsin the aeronauticalindustryin Canada.The prcscnt If the general i zcd
coordi natesvectori s of dim cnsionn, as
tool is developedin Matlab on the basis o1' thc existing thc cquationof acroelasticdynarnicsis of seconddegrec,then
thcoreticaltools and expertisein the litcrature. the vector of cigenvalucsis of dirnension2n
By usc of STARS,the lateraldynarnicsof a half aircrafttest
modcl is stLrdicd.Thc ATM is modelcd by finitc-elernents
rnethods,and the detailsof its modeling arc given by Gupta tr,lt (8)
(1997). Following the free vibration arralysis of the
flnite-clernentmodcl of thc ATM. tlrrce perf'ectrigid-body where each cigenr.,alueis written as follows:
modes,two rigid-controlrnodes,eight elasticrnodes.and thrce
rigid-bodytnodesare gencratcd. ) , i- d i + . j ( \ where 1<i<2n (e)
Thc aerodynarnicunsteadyfbrces were getlcratedin the
reducedfrcquencyfr dornainwith the Doublet Latticc Method whcre q is the irlaginary part of thc eigenvalue. represcnting
(DLM) rnethodin STARS. Then, for flutter calculations.wc the liequency, and r/, is thc real part representingthc damping.
used the p and P/r linear and nonlinearflutter rnethodshere The rnatrix ei' is defined as lbllows:
describedand programmedin Matlab.
Two main comparisonshavc been performcdbctweenthe e i r _ d i a g ( e 7 . rsr7 2 r. . . J i t ) (10)
resultsobtaincd( l) by introducingthe LS rnethodin the P
flutter methodand comparingthe resultsobtainedby our own The rnatrix of eigenvectors Vn contains in each of its
P-LS rnethodversusthe onesobtainedin STARS.(2) rvith the colurnns, thc cigenvectors associated with cach eigenvalue.
thrce approxirnationrnethodsLS, MS. and MP. Then. thc Thcn. \\re express the systern dcscribcd by Equation (6) in thc
advantages of using thc MS methodarc explained. followins r-natrixfonn:

Pk MsrHoD - LtnsAR SoLUrroN

The formulationfor linearaeroelastic
analvsisin the caseof
:t '(K:- q,tQ) Ml 'D_lln_l
[n] ltll-,hl (,,)
Ln-l l-M Ln_l
the Pk flutter r-nethodis
andwe calcr,rlatethc cigenvalucsandcigenvectors of rnatrixl.
Mi + Dn + Kn + 4aQ(k,Mac'h)1= Q (6)
The solution (expresscdas the aircraft rnotion) bccomcs
r"rnstablcwhcn the real part of the systcrn eigenvalues
where p, thc n-rodalgeneralizedaerodynamiclorcestnatrix,is (expressedin tcnrs of darnping)becomespositive.Once the
usuallycornplex.The realpart of p denotedby gn, is calledthe
dynamic prcssLlre(expressedin terms of spced,altitude.or
"aerodynamicstiffness",and is in phasc with the vibration
both) and the Mach numbervary,wc calculatethe parameters
displacement; the irnaginarypart of p denotcdby Qt is called
valucs(aircraftfluttervelocities)wherethe flutterphenolnenon
the "aerodynarnicdamping"and is in phasewith the vibration
takespl ace.
velocity.
This dynamics equation is a sccond-degrccnon-linear
cquationwith rcspcctto thc generalized coordinatcsvariable11. Pk NlErHoD- Nox-LrlrnanSolurrox
The non-linearitycolncs from the fact that the acrodynarnic
Now, in the non-stationarycasc, where the aerodynamic
generalizedforcesrnatrixQ is a functionof reducedfreqr,rency
generalizedforcesrnatrixp dependson the reducedfrequency
/r, dependingof the naturalfiequency0)as shown in E,quation
(s) t, Equation(6) becomesnon-linear.
Many algorithnrslnay give a good approxirnationof the
If we considerthe problcrnto be quasi-stationary,
wherethe
systerncigcnvalues, without giving a solutionto the problern.
aerodynamicgeneralized forccsmatrix Q is indcpendentof the
The first objectiveof the aeroclasticity is to analyzcthe stability
reduced frequency k, then Equation (6) becomes a linear
of the solutionandnot the solutionitself,so thatthe knowledge
equation,parameterized by the dynarnicpressureq,1and the
Mac'hnumbcr. of the ei-eenvalues obtainedis enoughto judge the stabilityof
the sol uti on.
The solutionof the problcm becornes
The Pfr methodis one of thc rncthodsthat allows accessto
the systern eigenvalues. This rnethod gives a good
n(1)= v,,e)"'vrtn(o) (t)
approxirnationof eigenvalues.Its algorithm, presentedin
Fi gure l . consi stsi n fi ri ng a Machnumbcrandcalculat ing t he
eigenvalucsfbr a sivcn nunrberof speedsthroughan iterativc

(er)
2003CASI Iul
 CanadianAeronauticsand SpaceJournal Journalaeronautique
et spatialdu Canada

the form in which the equationsare presentedis different.

Actually, the P method is a non-dimensionalform of the Pft
Fix small V
method,wherethe generalized coordinates,speed,and time are
nonnalized.
As Q may be decomposcdinto two parts pp and Qv we
Approximation
associatepo with 11and Q' with 11.Thercfore,as the Q matrix is
k=abN alreadya factorofq, we divide Qrby cqto expressp1 as a factor
of q. Thus, Equation(6) becomes

I
onstructQ@ matrix M i + ( D + - qa Q ) n + ( K + q , r p n ) n: 0 (t2)
0)

g and cofrom Equations(4) and (5) into

B y s u b s t i t u t i nqo
C a l c u l a t e l v=l d 1 + j a 1 E q u a t i o n( l 2 ) . w e obtai nthe fol l ow i ngequati o n:

\/r.)
Mi+(o. vcQtln*lK*.pV'QBln=0 (t3)
*, /\z)

Thc valueof o is givenby Equations( I ) and (2), from where

we obtain the following equation:
For each7, calculate
@;bN =k? p V 2= p o V J (l4)

Then,both sidesof Equation(1a) are dividedby l and the

definition of ofrom Equation(2) is used,therefore,we obtain

Vr2 V' -
Pv:or'i:Prve;:Prvylo (15)

In the aeroelasticEquation(13), Equation(15) is substitued

Yes
into the factor of Qr and Equation( 14) into the factor of pu, and
For V calculate we obtain
7"i=di+ jlJ.i
Then -> Next 1", (t\(r^)
M i + lo * ] , p"u , . G -r/r r e , l rnc* l' - p o v l el*n : 0 r l o l
No l"i -> IncreaseV \ 4k \ 2 )

A variable change technique is further introduced, this

F i g u r e l . A l g o r i t h m o f t h e P / <M e t h o d .
introduces the reference airspeed V0 through the normal
referencefrequencycr5definedas ob : Vslc.The conversionof
processon the reducedfiequencyfr, so that the eigenvaluesof the generalizedcoordinatesq in the codomaininto r1Pin the coP
the system for a given Mach numbcr and speedinterval are domain, is realizedwith the following technique:
obtained.Of course,as we hypothesizedthatthe Mach number
was constant,the result is invalid only if the speedintervalis 0
nP(oP) = n(o) where 0)P - (17)
centeredand sufficiently close to the speedcorrespondingto ('D6

the Mach number. If it is not close enough,we should restart

the algorithmby choosinga closer Mach number value. from where

ro : coPob (18)
P MnrHoD - Nox-DrnnpusloNAl
REpnESENTATIoN and, by deriving Equations (ll) and taking into account
E quati on(18), w e obtai n
The linear and non-linear solutionsof the Pft method have
alreadybeen describedin the previoustwo sections.In this
section,the P method is described.This method comes from
I=ololP and n=tino (1e)
thePk method.The processingof iterationsis the samebut only

182 O 2 O O 3C A S I
 Vo l . 4 9 , N o . 4 , D e c c n rb cr2003 0 V ol . 49, no 4, di cembre2003

To cxpress thc reduced fiequency k as a function of where Ai are the coetflcients rnatrices and Fi arc lug
parameters o)P,v, and o. we substitutecogivenby Equation( l8), coeft-rcients
of the approxinrationrnodel.Further.the objcctivc
tr'g i venby E quat ion( 2 ), l t g i v c nb y E q u a ti o n(3 ), a n d ofi orn functiortto be rninirnizedis deflnedas follows:
Eq u a tions( l) and ( 2) i n to Eq u a ti o n(5 ). a n d w c o b tai n
r :LI 1
-e,,{}kp
(D(' (' n V6
I w,i,1g,,tjk,t (24)
, .' p I,, Vo iil
k=-:-0)'0),-L=-(0'
2V 2 Vr, , ,, VF
rvlrere14',,7
i s gi vcn by
(2 0 )
I
-
Wilt = (2s)
k={o,on
,'0*(t.
lo,tlr ,)
2v
and klis thc /th normalizcdficquency.After minirnization,the
In the end, the aeroelasticnonnalizcdequationis obtained
coeftlcientmatricesare lirrthercalculated
by the P tnethod, after introducing thc ncw generalized
v c c t orql' b y E q u a ti o n (s1 7 )a n d (1 9 ),a n dby usi ng
co o rd ir lat es
the airspeedratio v from Equation(3) |0",,1
:| wj'1Yr'Yo'
+ nfin,,1 (26)
M P i t ' + ( D P + u . u 6 l u o ) n n+ ( K P + v 2 K ' P ) n P: 0 ( 2 1) f'l"l i?
where Lf , DP,DgI', KP,and Kul' matricesarc
" w,r^['0n'ir
1 B[ra''i]
f;
MP =M, D t ' _! O = D t
cQr
In Equation (26), we havc cxpresscd Bot1,811.Qni1r.and Qtilr
as follows:
Kt'= Ln x, Dr' - Much\.
oo- |orL'|et(k^ - ^;2t ^t
l- , / l

B^u/, - l I o k. lt
't- t t i * F i k.i 1+ $ . l
1 etl
,.P |
Ka' = ; P,,c'Qp.(k^Much)
L
I
-kh. I
g,, : [o -kt -k'h'
" L- I o :

r i * P i ki +lt,
I

In thc following scctions,thc three approximationsfor thc l

aerodynamicsconversionfionl the frequencydomain into the
Laplacedomainare described.They are the leastsquares(LS), and
the matrix Pade (MP), and thc minimum state (MS)
approxin-rations. Qniir: Reloi(lfrr)l and Qriir:lmlQ,,Q k)l (2tl)

As can be seen in Equations(26) to (28), the coefficient

L Elsr-SeulnEs AppnoxrMATToN
matricessolutionsdepcndon the lag coefficientsF7 variablc
The aerodynamic forces approxirnation for Q(k,M values.The quasi-Newtonrnethodis further usedto minin-rize
calculatedby the DLM is writtcn in the Laplacedornainas the obj ccti vefuncti on.
fbllows:

Mnrntx-Pnne, AppnoxrMATToN
J
Q ( r ): A , ,+ A ' . s+ A . . s *l i O , , (22)
l-l
t+F, This method is similar to the LS approximation.however
different[3; are calculatedfor each colurnn,so that the next
rnodel is further used.see Tiffany and Adarns (1987)
or, as the Laplace variable is ^s:.7fr, then

S o ' - ,- J = k Q r ( s:) A q r1 A r r s+ A , r r ' * t A , r * ; , r a # (2e)

Qffr) : A o * A , jk A,k 2 + (23) '\ r
7 ik+Fi | P,rl

and in the reduccdfiequencyft dornain.it becornes

io 2003 CASI
 CanadianAeronauticsand SpaccJournal Journaladronautique
et spatialdu Canada

- l

Q f i n - A 0 r* A r ri k - A . i k '+) l A 1 , , r ),
The objectiveis now non-linearwith respectto the E and D
(30)
,ih matri ces.The sol uti oni s obtai nedusi ng the i t er at ivelincar
quadrati csol uti on. Matri x D i s now fi xed a nd t he lincar
quadraticproblernis solvcd accordingto thc E matrix. Next,
whcre A,, are the coeft-icicnts matriccsof 7th column of the
the E rnatrixis hxed and thc lincarquadraticproblcrnis solved
matrix A, and F,,r orc lag cocfficicntsof thc 7th colurnn of
accordi ngto the D rratri x. Thi s schernei s r epcat edunt il
approxin-ration rnodel. In this case. the lbllowing objcctive
convergence is attained.To solvc the linearquadraticproblcrn
fu n cti o nis r ninir niz c d :
accordingto thc E rlatrix with thc D matrix fixcd. thc objcctive
lirnction is rewrittenbv colurnnas fbllows:
t, =4ZtrilQil(ik)-Q,,t
ir,)l' ( 3 1)
t=\t, (36)
In the MP rlcthod with respectto the LS rnethod,thc
u'herethe colurnnerroris
matricesBp7and 8,, arc diffcrcnt for cach column.

t, -4.ltr,lwi e/k) (37)

vtunn-SrnrnAppnoxl MATIoN
M rr,,{r
where
The MS approximationis written in thc Laplacedornain

I ) Q ; ( l / ' 7 )- Q i ( 7 1 1 )
e i ( f t r= (38)
Q ( s ): A o + A , s + A , 3 ' + D l s l - R l E . r (3 2 )
Furtherrnore.the objectivefunction (36) is rninimizedfcrr
In the frcqucncydotnain,this approxirnationbecornes each col urnr.t.A ccordi ng to E quati on (35). solving t he
on cnr (37; i s equi val cntto solving t he
r-r-ri ni rni zati probl
Qfjtl = Ao * A,.ik+ A,k2 +olikl Rl tr.jt (33) following cquationfbr eachcolumn of the E rnatrixas shown
b y G o l u ba n d V a n L o a n ( 1 9 8 9 ) :
where the D and E matrices are explained in thc next
paragraphs and arc relatedto the convergencc
of thc solution.R A"Ei - B" (3e)
is thc aerodynarniclags diagonal rnatrix. To rninirnize the
objcctivefunction.wc apply thc next constraints,thus Karpel whcrc A,. and 8,, are expressedas follows:
( ree0)
A" : I clDtw,,, DCor+ cilurwr?7
DCy
ReQ(0)l= ReQ(g)1, R c O ( / / ' r ) l= R e Q ( j k / ) i I
- Br.,r)
8,,: Icl,DtwT?r(Qni(frr) -B,,r)
+c/,nfwi(Qr7(frr)
I r n Q ltf r * ) l= l m p l r k . ) i

We know that The error is writtcn as a row crror and the linear quadratic
problem,accordingto the D matrix,is equivalentto solvingthe
(:jkr -R) I : (k2r+ R2)r(-/frI-R) (34) following equationfbr each row of D:

theseconstraint
Following thernodelbecomes:
applications. DiA,t : B,/ (40)

Qffr r ) = B R1+ DCp T u* j ,B,, + D C ,7 E) (35) wherc

i n wh i c h A,t: ICo,Ewit,EtCL +CyEwit,EtCI

I
t.)
BR/: Qn(o)- * tq*to)-Qn(tr)l B,/: (k)- BB,,)wi?rEtcil, -Br/)wiEtcil
+(eri(rtr)
ftl It0,.i
In conclusion.the MS approxirnation is generallybctterthan
cnr= nfttir + R2)t tt it+ Rr) ' )r? the LS and MP approrirnations.When the nr,rmber of lags is
greaterthan thc nurnberof lrequcncies,the problembecomcs
i l l condi ti oned.
= L,
Br1 To solve this problern, two adclitional features are
frQ,tL"t consi dcrcdat cachi terati on:( l ) the l i nearquad r at icpr oblem s
accordingto thc E rnatrix and after that accordtngto thc D
c , : o [ f t 3 , + R r ) - r- ( k i t + R r ) | matri xaresol vcdusi ngsi ngul ar-val ue decornpo sit ion( SVD) as
-lo*, show n by Gol ub and V an Loan ( 1989); (2) t he opt im al

lu4 r , 2 0 0 3C ' A S I
 Vol. 49, No. 4, Dccember2003 Vol. 49, no 4. decernbre2003
0

comprornise betweenthe prescntand the lastitcrationis choscn Firstly, we have to cxpress tlic aerodynamic forces Q under the
tcl cnsurethc algorithmcouvergcncc.Next, w e w o u l d l i k e t o Padi polynomial fbnn
cxplain thesctwcl features.
Equat ions( 39) an d (4 0 ) n ra y b e w ri tte n irr the fclllowing
fbrrn: Qk)- A,,+ ltiA,+ ( iD A. *'f ,'-ik 1.*,
- (46)
7 ik+flt
Ax-B 1al)
rvlrereA, arethc coclficicntsof a dirnensioncqr-ral to the rnatrix
i n whic h x is an un k n o w n v c c to r.W h e n th i s s y stcrni s i l l O and obtained fl'orn thc LS algorithm. and Fi are the
co n d i t ioned. t hc LS s o l u l i o nc a n b c o b ta i n c cul s i n gSV D of A . acrocl ynanrilcags. Thc opti mal val ucs of the se lags. which
In l a c t , A c an bc dc c o rn p o s eads l b l l o w s : mi ni rl i zc the appro.ri rnati on
crror bctw eenthe aer cldynar nic
(.)(/r)nratri x and i ts approxi nrati onby P adepolynom ials.is
A - USVT (42) calcLrlatedf urthcr. Equations (5). anc'l the n,cll-knor,vn
crpression.\' : /(Dgivcs ili - hslV trnd by placing this ncw
i n w h i c h U and V ar e o rth o g o n anl te ttri c c a s .n d S i s a di a_qonal cxprcssi oni n E quati on(6) w e obtai u
rn a trix .T he diagona lo 1 ' th e S r-n a tri rc o n ta i n s i. tr ordcr.the
si rrg ularv aluc of A s o th a t. i f ra n k (A) - r { n . th c ri I' l ast M i + D r l+ K \ + q , r (4t)
e l crn c ntofs t hc diagon aal rcn u l l . U s i n -eth i sd c c o n -rp osi ti orr.
thc
L S solLr t ion of E qr , r a ti o(4
n 1 ) c a n b c o b ta i n c da s fo ll ow s: tJ , / | \l
-,,
, l ]

+t,rr
"1r,, .li,)0.* tL l.v - ",q.
+ t, In
I
i - t . s+ - ; F r
( 4 3) L t)l
I

The statcvectorof acrodynamicrnodesis firrthcrintroduccd

whcrc u; and vr arc,respectively, thc coluntnsof tnatriccsU and
V and ^s,is thc ith elernentof diagonaln-ratrixS. \
L ct E lprand E , n r ; b c th c o p ti rra ls o l u ti o nto th c p r esentand ' -\
I
:- -ll
l'
'l (48)
last iteration so that the optirnal cornplonrisebetrveenthc .i+l-0,
p re se ntand las t it c ra ti o n i s o b ta i n c d b y rn i n i rn i zi ngthe
o b i e c t iv ef unc t ion( 37 ) a c c o rd i n gto u l v i th and by repl aci ngE quati on(48) i n E quati on(4 7) . r , vcobt ain

E - c x E , nr ) + ( l cr)E1py (41)
:u
ui*on * rn nn,,l'f,'A:*,x,-] (4el
Sirnilarly,fbr the D rnatrix. the ob.jectivefurrctioncan be Li-.r l
rninimizcdaccordinsto cr with
whcre
D : c r D , n, , * ( l - c r ) D , n ) (45)

Fo r t hes c s c ala r rn i n i rni z a ti o n s " w e u s c d a scal ar tI:ivt+ 0., D=D. r,,(l)^,

,,,11)'
rninimizationalgorithrn.

and
LS MsrHoD lNrnonucEDrNruE P METHoD
- /A\
Equat ion( 21) dc s c ri b i n gth e d y n a rn i c so f a e ro -s c rvoel asti c K:K*U,r[ (s0)
systcmsofl-ersonly serni-lincarreprescntations. In lbct, all ;1,
terms related to the aerodynarnicfbrces present strong
non-lincariticswith respectto the rcdr-rced lrequcncyk. Then, Wc substituteli givenby Equation(3) into Equation(2) and
th e mu lt it udcof t he a n a l y s i sa n d rn o d c l i n ga l g o ri th r ns appl i ed v, and o. Then,Equat ion( l8)
w e obtai n V as a functi onof V 1,.
to the linear systemsrequiresthc rnotivatiorrto obtaina linear gives cq)as a function of crlPand rr) and by equatingthe two
aero-servoelastic system. Various mcthods arc available to E quati ons(20), gi vi ng the val ue of /t. w hi le t aking int o
applythis linearrzatron, seePitt and Goodrnan(1992),Rogeret consideration that b : c'12"
we firrtherobtainthc valucof theblV
a l . ( 1 9 7 5 ) ,K a r p c l( 1 9 8 2 ) ,D u n n ( 1 9 8 0 ) .T i f l a n y a n d A d a m s ratio as function of o, v, and 06, as follows:
( 1 9 8 4 .1 9 8 7 ) P, o i r i o n( 1 9 9 5 ) a
, n dC o t o ia n d B o r e z( 2 0 0 1 ) . W e
h a vechos ent he s im p l e s o t n e .th e L S mc th o d b e causei t i s h_ ^[6
( sl )
r,rsed in STARS- and in this wav we can comparethe rcsults. L' 2r,cqy

i c r2 0 0 3 C A S I 185
 CanadianAeronauticsand SpaceJournal Journaladronautique
et spatialdu Canada

The variablechangetechniqucintroducesZtr: c(erand by

.p
T
I
00 0 n'
replacingthis expressionforVs into Zp given by Equation(3). ..P

we obtain Vy:.: (:(\1y.Furthcrmorc,wc substitute V, into qr1

0 ilrP o n'
g i vcn b y E quat ions( 4) a n d (1 4 ), a n d w e o b ta i n 0 OI x,o (s7)
0
I , . , = ' Ir r r t ' r t ' - ' i v r 1,. I-
q,t : (s2) 0 OI /l
Ltl
;P,,Vt:
p
Thcn,we substituteq6 givenby Equation(52) andblZ gir,'en OI 0 0 n'
by Equation(51) into Equation(49) and into the coefficients -fre -fie -rt A! -v2 A! .p
n'
1,+ nl,
givenbyEquation(50) and we divide the resultingequationby xl
cofrand we obtain
OI +l'1 0 0
0 0
"l -01 1
f l-,,, OI 0 ;, t1$

t4oi' + Dtnt + RPqP+ vrl F Al- T t ,x,t I : 0 (53)

| /-/ I
Lr-r l
REsulrs ANDDrscussroN
where
To validate our rncthod.u'e uscd the STARS Aircraft Test
Modcl devel opedby Gupta (1997) at N A S A Dr yden Flight
ilf' : MP + oAl, DP - DI' + r^[" ,q|
ResearchCenter.This rnodclincludesaero-structural elements
(flexible aircraft)and control surf-aces (aircraftcommands).
To : KP+v2AP
Firstly, a frcc r,'ibrationanalysis was performed in the
absenceof aerodynamics to obtainthe free modesof vibration.
with the coefficients
We obtainedthe samefrequenciesand modesof vibrationby
our method as well as by the one used in STARS.
'^p,,r'2
^ll
A P c= , A r , . A l = L pot'bA1, Al = Secondly, to calculate the aerodynamic forces in the
I lo,,b''t, frequencydomain by the doublct lattice method (DLM), the
^l
simulationparametersare consideredin the STARS computer
A'i : ^prt'-A for 3 <j <2 + n, program:the referencescmi-chordlength h - 81.50 cm, the
refcrenceair densityat scalcvcl ps : 1.225kg/mt. the altitude
L

at the sea level Z - 0 km" the ref-erence Mach number Mach :

The first equationof the systemof Equations(19) and also
0.9,thereferencesoundairspccdat sealevela6: 340.3rn/s.
the ratio Vlb given by Equation (5 I ) arc substitutedinto In Table l, the resultsof an aeroelasticnon-linearanalysis
Equation(48), and wc obtain
arc shown"and velocitiesat which flutteroccursare calculated.
In Table l, the resultsgiven by our Pfr methodare cotnparcd
*,*4roo$iX, -,on,to (s4) with the results givcn by STARS through the three flutter
!o open-looprnethods(k method,Pk rnethod,and ASE method).
Actually,we shouldonly comparcthe resultsobtainedby our
Next, salre variablc transfotmationis applied ro X, as the P/r method(row 4) with the resultsobtaincdby the P/r method
o n e a p pliedt o 11giv en b y E q u a ti o n(1 9 ) in the STARSprogram(row 2). We canconcludcthatwe obtain
a goodcomparisonin tennsof fluttervelociticsandfiequencies
f i' :0fh *, (5s) for both fl utterpoi nts.
In Table 2, we prescnta comparisonbetweenthe results
obtainedon the linearizedATM with our P rnethodmodifiedto
givenby Equatton
With this lastncw variabletransformatiot-t include the LS fbrmulation shown on row 4 with the ones
(55), and by dividing Equation(5a) by ob, we obtain obtainedby useof the STARSsoftware,calledthe ASE method
(row 3).
open-loop analysis is further
X l * P i X i = r l P , w h e r eV' i : ? F , (56) A lirrear aero-servoelastic
r/o perfbnnedon the ATM. This analysisusesthe P-flutterrnethod
where the aerodynamicunsteadyfbrccs werc approximatcdby
Then, we obtain the final matrix equation the LS rnethod.Resultsare shownin Figures2 to 4. As shown
in Figure 4, the unstablemodeswhereflutter occursare elastic
mode2 (flutterpoint I ) and mode7 (flutterpoint 2). Mode 2 is
the fuselagefirst bendingand mode 7 is the fin first torsion
mode.

t86 ic,r2003 CASI

 Vol. 49,No. 4, December2003 I Vol. 49, no 4, ddcembre
2003

Table l. Aeroelastic Non-linear Analvsis.

Flutter I Flutter 2
Equiv. airspeed(km/h) Frequency(radls) Equiv. airspeed(km/h) Frequency(rad/s)
STARS
/c method 821.10 77.4 1596.5 t47.3
Pk method 818.74 17.4 1600.4 147.l
l,SE method 878.79 77.3 l35t.l 136.2
Our Pk method 817.44 77.4 1404.7 145. 6

Table 2. Linear Aeroelastic Analvsis.

Flutter I Flutter 2
Equiv. airspeed(km/h) Frequency(rad/s) Equiv. airspeed(km/h) Frequency(rad/s)
STARS
k method 82r.70 77.4 1596.s0 14 7. 3
Pft method 8 18 . 7 4 71.4 1600.40 147.1
,4SE method 878.79 77.3 t3sl.10 136.2
Our P-LS method 878.79 77.5 I s 0 8 .01 144.2

35il
i-i Eler:_;tii-
t-r-rrtr-le
1
.Ir ElAStir_t-il'tde?
+ Elastirr-nide3
3fiil + Elastii:tr-rtrje
4
n Elastirmo'Je5
{;} Elastictrr'trjEfi
F ElasticrxildeT
:5il I

a Elastirrlo,le I
U]
E
{U
t*

]*t
TJ
C ?il0
fi]
=
{F
il)
!*
LL
150

1ilil

5il
-3ttil -:[0 - 1ilil 1ilil
t)amping {g.20il}
Figure 2. P-LS Method - F r e q u e n c y ( r a d / s ) v e r s u s D a m p i n g ( 9 * 2 0 0 ) .

i.) 2003cASI 187

 Canadian Aeronautics and Space Journal 0 Journal a6ronautiqueet spatial du Canada

o Elasticmode1
x ElaEtic mode2
+ Elasticmode3
* Elasticmode4
tr Elasticmode5
+ Elasticmodefi
v E l a s t im
c o d e7
; a ElasticmodeB
{n
'j(]

s
lF*
L}
C
0)
:5
(r
(I}
: l t t-

5fi0 1il00 15*il 2fi*il

Equivalent
airspeed{krn/h}
Figure 3. P-LS Method - Frequency (rad/s) versus Equivalent Airspeed (km/h).

The lineanzation of aerodynamic forces by the LS method An optimrzatron algorithm minimizes the quadratic error
offers many advantages over the comprehension of the denotedby -/ between the exact aerodynamicforce matrix and
dynamics of aeroelasticsystems as well as over controller the approximation of the aerodynamic force matrix
conception. In our study, we have chosen a linearization approximatedby Paddpolynomials.This algorithm varies the
techniquethat is relativelysimple,basedon the optimizationof valuesof aerodynamiclags, where the approximationof Pade
aerodynamic lags By and on the LS minimization of Pad6 polynomials is realized with the LS algorithm.
coefficients4,. Resultsof the LS method applied on the ATM for a Mach
When we observethe equivalentairspeedsfor the flutter numberof referenceM:0.9 are shownin Figure 5. A total of
phenomenon,the resultsof the combinedP andLS methodare six aerodynamiclags have been used to obtain a satisfactory
closer to the results of the Pft method than those of the ASE aerodynamicforcesapproximation.The left graph in Figure 5
method programmedin STARS. This conclusiondependson showsthe evolutionof the norm of the quadraticerror between
the quality of the hnearrzattonapplied on the ATM, becausethe the aerodynamic force matrix and the approximated
results of the non-linear model (Pfr method) and of the aerodynamicforce matrix as a function of iterationnumber of
linearizedmodel (LS and P method) are close.Our linearized the optimrzatron algorithm.
rnodel divergeswith respectto the ltneartzedmodel conceived There is also compromise between the number of
with STARS. This divergence may be explained by the aerodynamiclagsand the quality of approximation.The greater
precision of each hnearrzation,and on the other hand by the the number of aerodynamic lags, the better is the
differenceof the hneartzatronmethodsand more particularly of approximation. At the same time, the aeroelastic model
the optimrzatron techniques. becomesmore cumbersomefrom a computing time point of
view as the number of aerodynamiclags is increased.

188 ( O 2 O O 3C A S I
 Vol. 49,No. 4, December
2003 0 Vol.49, no 4,ddcembre2003

o E l a s t im
c o d e1
H E l a s t im
c o d +2
+ Elasticmode3
# E l a s t im
c o d e4
n E l a s t im
c tde5
+ E l a s t im
c o d eS
-sfi v E l a s t im
c o d e7
a ElasticmodeB
{}
{}
f-'l
lc
- 1ilfi
#J

{}r
'a
r

f; - 15il
ffi
n
-TilS

-?5fi

-300
t^l
LJ silfi 1ff0fi 150il ?il*il
Equivalent
airsp*ed
{km/h}
Figure 4. P-LS Method - Damping (9*200) versus Equivalent Airspeed (km/h).

rigid, and control modes,for which the aerodynamicforcesare

defined.
o 250 Concerning the existence of the other methods of
o
200 optimization,a comparisonis realisedon the useof the LS, MP,
o
c and MS algorithmsfor the aerodynamicforcesapproximation.
o 150 Figure 5 showsthe total error for LS, MP, and MS modelsis of
IE
the same order (160). These results shows us clearly the
E
x
100
o superiorityof the MS model.
CL 50
a,
0 CoxclusroNs
LS wit h 16 0 M P w i th 1 6 0 MS w i th 1 6 0
v ar iables v a ri a b l e s v a ri a b l e s In this paper, firstly, a comparison was reahzedat the level
( 4 lags ) (4 l a g s ) (1 6 0 l a g s ) of an aero-servoelasticnon-linear analysis. This analysis
concernedour modified Pk methodvalidatedagainstthe flutter
Figure 5. Errors for the Three Methods LS, NIB and NIS.
methods programmed by using STARS. The Aircraft Test
Model existing in STARS at NASA Dryden Flight Research
The reason is that with each ncw aerodynamrclag, we Centerhas been used.A good coherenceat the level of flutter
introduce new aerodynamic l a g s i n th e d e s c ri p ti onof the predictionin open loop, appearedat flutter velocitiesof 834.12
aeroelasticmodel dependingon the total number of el asti c. and 1482.90km/h.

((,r2003CASI
 CanadianAeronauticsand SpaceJournal 0 Journal aeronautiqueet spatial du Canada

Secondly, the influence of the aerodynamic forces K a r p e l , M . ( 1 9 9 0 ) . " T i m e - D o m a i n A e r o - s e r v o c l a s t i cM o d e l i n g U s i n g

WeightedUnsteadyAerodynamic Forces".J. Guid. Control, Vol. 13, No. I I,
lineanzation LS studies on the flutter prediction, firstly in an
pp. 30 37.
aeroelasticcontext, and later, in an aero-servoclastic context.
N o l l , T . , B l a i r , M . , a n d C e r r a ,J . ( 1 9 8 6 ) . ' A D A M , a n A e r o - s e r v o e l a s t i c
The comparison was good from flutter (aeroelastic) and
A n a l y s i sM e t h o df o r A n a l o go r D i g i t a l S y s t e m s "..1 .A i r c r .V o l . 2 3 , N o . I 1 .
aero-servoelasticpoints of view.
Pitt. D.M.. and Goodman, C.E. (1992). "FAMUSS A New
The aerodynamicforces approximationprecision may be
c o d e l i n g T o o l " . A I A A P a p .9 2 - 2 3 9 5 .
A e r o - s e r v o e l a s t iM
increasedby augmentingthe number of aerodynamiclags on
Poirion. F. ( 1995). "Modelisation Tcmporclle des Systdmes
one hand,or by useof the most powerful linearizationmethods
. p p l i c a t i o n A I ' E t u d e d e s E f f e t s d e s R e t a r d s " .L a
A d r o s c r v o i l a s t i q u c sA
that guaranteea good precision,by decreasingthe number of RechercheAerospaliole.No. 2. pp. 103 I 1,1.
aerodynamicmodes.
R o g e r ,K . L . , H o d g e s .G . E . . a n d F e l t . L . ( 1 9 7 5 ) ." A c t i v c F l u t t c rS u p p r c s s i o r r
The comparisonof three hneanzationmethods shows the A F l i g h t T e s t D e m o n s t r a t i o n "J. . A i r c r . V o l . 1 2 , p p . 5 5 1 - 5 5 6 .
superiorityof the MS approximations.In the MS method,the
T i f T a n yS , . i l . , a n d A d a m s .W M . , J r .( 1 9 8 4 ) ." F i t t i n g A e r o d y n a m i cF o r c e si n
systemis ill-conditionedwhcn the number of approxirnation t h e L a p l a c eD o m a i n : A n A p p l i c a t i o no f a N o n l i n e a rN o n - g r a d i e nT t echniquc
lags is larger than the number of reduced frequencics.An t o M r - r l t i l e v eCl o n s t r a i n c dO p t i r - n i z a t i o n "N. A S A T M - 8 6 3 I 7 .
additionalfeatureis presentedto solvethis problem.We should Tiftany. S.H.. and Adarns,W.M.. Jr. (1987). "Nonlinear Prograrnrling
choose,at each iteration,an optimal compromiscbetweenthe E,xtcnsions to Rational Approxirr-ration Methods of UnsteadyAerodynarrics".
presentand the last iteration.[n the end, the presentpaper Proc'cetlirtg.so/ thc )8th AIAA/ASME/ASC'l',/AHS Structure.s, Struc'tural
showsa goodvalidationof our aero-servoelastic methodon the Dt'nomit'.s,ttnd ,)lutet'iul.sConfarenc'c,Montcrey, Cialifornia,6-8 April 1987.
A m e r i c a n I n s t i t u t eo f A e r o n a r - r t i casn d A s t r o n a u t i c sN
, ew York, New York.
ATM model in STARS.
pp. 406 .120.

AcTxowLEDGEMENTS
The authorswould like to thank Dr. Kajal Gupta at NASA
Dryden ResearchFlight Centerfor allowing us to usethe ATM
in STARS. Many thanks are due to the other members of
STARS Engineeringgroup for their continuousassistanceand
collaboration:Tim Doyle, Can Bach, and Shun Lung.

RETEnnNCES
Adams, W.M., and Hoadley, S.T. ( 1993). "ISAC: A Tool fbr
i co d e l i n ga n d A n a l y s i s " .N A S A T M - 1 0 9 0 3 1 ,p p . I 1 0 .
A e r o - s e r v o e l a s tM

C h e n . P . C . ,S a r h a d d i .D . , L i u , D . D . . a n d K a r p e l , M ( 1 9 7 7 ) . " A U n i f l e d
Aerodynamic lnfluencc Cocfllcient Approach fbr
A e r o e l a s t i c / A e r o - s c r v o e l a sat ni cd M D O A p p l i c a t i o t t s " A
. I A A P a p .9 7 - l l t 3 l ,
pp. l27l 1271.

C l h e n , P . C . , S u l a e m a n .E . , L i u , D . D . . a r r d D e n e g r i , C . M . . J r . ( 2 0 0 2 ) .
" l n f l u e n c e o f E x t c r n a l S t o r e A e r o d y n a m i c so n F l u t t c r / L C I Oo f a F i g h t e r
Aircraft". Proc'eetlingso/ tha 1-1rdAIAA/ASME/ASCE/AHS/ASC Stru<'tures,
Slntc'lurtrl D.\'ttctntit:.s and Materiuls Conferenc'e.Denver, Colorado, 22 25
April 2002. American InstitLrte o f A c r o n a u t i c sa n d A s t r o n a u t i c sA
, merican
S o c i c t y o f M e c h a n i c a lE n g i n e e r s R. e s t o n ,V i r g i n i a . A I A A P a p . 2 0 0 2 - 1 4 1 0 ,
pp.l ll.

C o t o i , I . , a n c lB o t e z .R . M . ( 2 0 0 1) . " O p t i m i z a t i o no f U n s t e a d yA e r o d y n a m i c
Forccs for Acro-scrvoclastic Analysis". Pntt'eedings fi the L4STED
Internolionol Cctnferanca on Conlrutland Applicaliorr.r,BanfI. Alberta. 27 29
. l u n e 2 0 0I . E d i t e d b y M . l l . H a m z a . I A S T E D /A C T A P r e s s , A n a h e i m ,
C a l i f b r n i a p. p . 1 0 5 1 0 8 .
D u n n . I I . J . ( I 9 8 0 ) . " A n A n a l y t i c a lT e c h n i q L rfcb r A p p r o x i m a t i n gU n s t e a d y
A e r o d y n a r n i c si n t h e T i m e D o r n a i n " .N A S A T P - 1 7 3 8 .

G o l u b .G . H . , a n d V a n L o a n ,C . F .( 1 9 8 9 ) ." M a t r i r C o m p u t a t i o n s "T, h e . l o h n s
H o p k i n s U r r i v e r s i t yP r e s s ,B a l t i r n o r e ,M a r y l a n d .

Gupta, K.K. ( 1997). "STARS An Integrated Multidisciplinary.

Finitc-Elemcnt, Structural, Fluids, Aeroelastic, and Aero-servoelastic
A n a l y s i s C ' o m p u t e rP r o g r a m " .N A S A T M - 1 0 1 7 0 9 .

K a r p c l . M . ( 1 9 8 2 ) . " D e s i g n f b r A c t i v e F l u t t e r S u p p r e s s i o na n d G u s t
A l l e v i a t i c r nU s i n g S t a t eS p a c eM o d c l i n g " . J . A i r c ' r .Y o l . 1 9 , p p . 2 2 1 2 2 7 .

190 is.r2003CIASI