Você está na página 1de 13

System Mode Shapes in the Flutter of

Compressor Blade Rows'


FRANK LANEf
New York University

ABSTRACT INTRODUCTION

The problems associated with the prohibitive number of pos-


sible system modes for a fluttering compressor or turbine blade
row are eliminated by the development that comprises the present
T H E C H I E F CONTRIBUTION of t h e research reported
herein lies in t h e establishment of a m e t h o d t h a t
permits t h e linear flutter analysis of configurations
report. T h e existence and uniqueness of extremely simple sys-
tem flutter modes are proved for blade rows consisting of identical characterized b y a large n u m b e r of identical fluttering
Downloaded by OLD DOMINION UNIVERSITY on June 1, 2014 | http://arc.aiaa.org | DOI: 10.2514/8.3502

blades equally spaced about a common rotor. These simple members and, accordingly, encumbered b y a prohibi-
system modes, if properly interpreted, have the effect of reducing tively large number of degrees of freedom. I n order
by a factor of n t h e number of degrees of freedom necessary to
to place this development in t r u e perspective it is
analyze an w-bladed configuration. Stated differently, the system
of n blades m a y be considered, with no loss of generality whatso- necessary to outine t h e methods b y which m a n y -
ever, in terms of a single "equivalent blade." T h e proof holds bladed configurations have been flutter analyzed in
under any type of flow and any and all types of interblade the past a n d to indicate wherein t h e present t r e a t m e n t
coupling, so long as a linear analysis is permissible. Moreover, extends or supersedes this previous research.
since it is the flutter-inception point t h a t is of interest in pre-
dicting critical velocity or rotational speed, it m a y well be t h a t W i t h t h e availability of well-established methods for
the conclusions developed apply even to the onset of stall flutter. flutter analysis of single, isolated wings, t h e first a n d
Practical application of the method to stall-flutter calculations most obvious method for treating multibladed systems
would, of course, require the availability of aerodynamic stall-
is to ignore all forms of interblade coupling a n d t o
flutter coefficients.
proceed under t h e assumption t h a t each blade flutters
The development is carried out first under the assumption of
infinite rotor inertia or, in other words, constant rotor velocity. just as it would in t h e absence of all other blades.
This restriction is then relaxed, and the treatment is expanded to This approach can b e considered as zeroth order a p -
permit torsional oscillations of the rotor itself. I t is proved that proximation a n d m a y lead to unconservative estimates
under certain conditions the assumption of infinite rotor inertia of critical flutter velocity.
introduces no error whatsoever.
T u r n e r a n d Duke, 9 in an investigation of propeller
SYMBOLS flutter, neglect aerodynamic interaction b u t dispose
of t h e interblade coupling t h a t arises from t h e physical
A = flutter matrix with infinite rotor inertia assumed
connection of t h e blades of a three-bladed propeller
A = flutter matrix with finite rotor inertia
to a common h u b b y resolving t h e propeller system
V = relative flow velocity at some reference radius
mode into " r e s o n a n t " and " a n t i r e s o n a n t " components
k = reduced frequency of flutter ( = co&/ V) at reference radius
t h a t are t h e n shown t o be uncoupled. This, in t u r n ,
b semichord length at reference radius
allows t r e a t m e n t of a single blade in formulating t h e
n = number of blades in row
flutter problem, all under t h e assumption of zero aero-
/ = time
dynamic interaction between blades. T h e cumber-
p = number of degrees of freedom alloted to each blade
some n a t u r e of t h e analysis b y resolution into resonant
g = generalized-coordinate vector excluding rotor deflection
and antiresonant modes (it appears awkward even for
q = generalized-coordinate vector including rotor deflection
as few as three blades) indicates t h e unsuitability of
qtt = generalized coordinate associated with rotor deflection
this approach to systems such as compressors a n d t u r -
Qa = generalized coordinates associated with blade twisting
bines with m a n y blades. Lilley 2 noted, from experi-
qu = generalized coordinates associated with blade bending
mental observations of t h e flutter of a series of elastic
12 = rotor angular velocity, rad. per sec.
wings arranged in cascade, t h a t adjacent blades were
o) = flutter (circular) frequency
approximately 180 o u t of phase when fluttering.
ov = argument (radians) of vth of the n nth. roots of unity
This phenomenon was termed "antiphase m o t i o n "
a = phase-lag angle between adjacent blades at flutter
by Lilley, who recognized immediately t h a t , for t h e
Received September 20, 1954. case of zero stagger, a cascade fluttering in antiphase
* The investigation reported in this paper was sponsored by the motion is aerodynamically equivalent to a single wing
Wright Air Development Center, Air Research and Development oscillating between parallel walls with its a t t e n d a n t
Command, under Contract A F 33(616)-25 with the Research image system. T h e aerodynamics of t h e single wing
Division, College of Engineering, New York University. Project
oscillating under t h e interference of tunnel walls h a d
Director, Chi-Teh Wang, Professor of Aeronautical Engineering.
f Research Associate Professor of Aeronautical Engineering,
been studied previously b y Reissner 4 ' 5 a n d others
Daniel Guggenheim School of Aeronautics. and is presented in somewhat different form b y Lilley 2
54
with a view toward obtaining oscillatory air forces or even stalling (separating and reattaching boundary
for the unstaggered cascade in antiphase motion. layer)so long as a linear analysis is permissible a t
Lilley then applied these air forces to t h e unstaggered the flutter inception point. Moreover, the develop-
cascade flutter problem for the particular values of m e n t permits of all forms of couplingaerodynamic,
elastic and inertial parameters corresponding to his elastic, or inertialbetween blades. Incidentally, as
experimental system, under the assumed condition a consequence of this proof of existence of a simple
t h a t the system flutter was in antiphase motion, and system flutter mode, it is shown t h a t the antiphase
the agreement between theory and experiment proved motion observed b y Lilley is to be expected for t h e un-
excellent. I t is to be emphasized t h a t Lilley's observa- staggered cascade a t zero incidence b u t not necessarily
tions indicated t h a t , for b o t h staggered and unstaggered for other configurations. T h e degree of departure from
cascades, adjacent blades a t flutter were approximately exact antiphase motion in the case of nonzero stagger
out of phase and t h a t he derived air forces and solved or nonzero incidence, or both, m u s t be determined as a
the cascade flutter problem assuming antiphase motion result of t h e flutter calculation for the particular con-
for only the unstaggered cascade. Previous results figuration in question.
for tunnel interference effect on a single oscillating
wing also, for obvious reasons, apply only to the un-
Downloaded by OLD DOMINION UNIVERSITY on June 1, 2014 | http://arc.aiaa.org | DOI: 10.2514/8.3502

ANALYTICAL D E V E L O P M E N T
staggered case. T h e importance of Lilley's work lies
in the fact t h a t he was able to assume in advance t h e (1) Consequences of the Cyclic Characteristics of the
form of the system mode, based on experimental evi- Configuration
dence, and hence to determine the oscillatory air
I n order to keep the explanation from becoming too
forces (including interaction between blades) and
cumbersome, the details will be developed for a system
proceed with a solution of the flutter problem subject
in which just two degrees of freedom are alio ted to
to the restrictions imposed b y the antiphase assumption.
each blade (one in torsion qa and one in bending qu).
Chang and C h u 1 present a derivation of the oscilla- T h e method is, however, equally applicable when any
tory lift and m o m e n t acting upon the blades of an un- number, p, of degrees is allowed per blade, and, a t
staggered two-dimensional cascade a t zero or very small appropriate points in the development, t h e method of
incidence, under the restrictive assumption t h a t all extending to more coordinates per blade will be in-
blades oscillate in phase with one another. This as- dicated. T h u s we consider a cyclic system of n blades
sumed condition permits a solution of the aerodynamics all identical and identically supported and all equally
problem by mapping techniques similar to those used spaced. B y "cyclic" arrangement we mean only to
by Theodorsen 8 for the fluttering single wing. As signify the usual propeller, turbine, or compressor
will appear evident from calculations to be reported configuration wherein the last or nth. blade is adjacent
in a subsequent paper, the in-phase assumption is the to t h e first blade. We denote b y qj, qj} the generalized
least conservative one possible with respect to the coordinates corresponding to the torsion and bending
prediction of critical flutter velocity. In fact, it will modes, respectively, of t h e jth blade, with (harmonic)
appear t h a t critical flutter velocities predicted on t h e time dependence removed, and we arrange the vector
basis of an assumed in-phase motion of all blades will of generalized coordinates in the form
exceed the critical velocity of one of the blades alone,
whereas actual blade-row critical flutter velocities are
2a l
considerably less t h a n t h a t of an isolated blade.
A study of Lilley's paper suggested to the present
2z
author the possibility of performing a general investiga-
qj
tion of system flutter modes for multibladed systems
(i)
with the objective of predicting, or a t least restricting
in advance the form of t h e system mode which exists a t
the fluttering condition. W e content ourselves here
q"
with a somewhat oversimplified s t a t e m e n t of t h e con-
.qu"
clusions reached as a result of t h e investigation. I t is
proved t h a t a system of identical blades identically where OJ is the flutter frequency. We assume initially
supported and equally spaced a b o u t a common rotor, infinite rotor inertia so t h a t no coordinate is necessary
whether this rotor be rigid or elastic and of infinite or for the rotor itself. Now, under the most general con-
finite inertia, will flutter in a system mode or a t worst
ditions of coupling, the final system of (2) flutter
in a linear combination of modes, in which all blades
equations will take the form
oscillate with identical blade modes and in which each
blade leads or lags its neighbor b y a common phase Aq = 0 (2)
angle, the value of this angle constituting one portion
of the solution of the flutter problem for a particular wherein the matrix A m u s t be of cyclic form in its
configuration. T h e conditions under which this state- second-order submatrices as a consequence of t h e cyclic
m e n t is proved are sufficiently general to encompass geometry of the configuration. T h e flutter equation is
any type of flowincompressible, subsonic, supersonic, then of the following form, in expanded notation:
55
56 JOURNAL OF T H E AERONAUTICAL S C I E N C E S JANUARY, 1956

d\ h G2 #3 h dn bn Hz*1]
Ci & C2 d2 Cz d% dn On
an bn CLl h a2 b2 dn-\ bn-l qJ
Cn dn Ci d1 C2 d2 Cn-1 dn-i q_u \
dn-i bn-l dn bn Q>\ h dn-2 bn-2 qj
Cn-l dn-i Cn dn Ci dt c
n-2 dn-2 qj = 0 (3)

a2 b2 a b di h qan
c2 d2 Cz d, 4 di Ci di J .qunA

where elastic, dynamic, and aerodynamic coupling we m a y write Aj} Bjy Cj} Dj equally well in the following
m a y exist between blades, as well as within a single form:
blade.
Here at represents the total effect of the torsion mode Aj = (ai + d2a>j + anu>j~l + an-iUj ~ +
Downloaded by OLD DOMINION UNIVERSITY on June 1, 2014 | http://arc.aiaa.org | DOI: 10.2514/8.3502

in the sth blade upon the torsion mode of the (s + 1 Bj = (61 + b2ccj + bno)j-1 + W bn_!o>j-2 +
i)\h blade, bt represents the effect of the bending mode etc.
in the 5th blade upon the torsion mode of the (s + 1 (5a)
i)th blade, ct is t h e effect of the torsion mode in the sth Now since the determinant | A | is unaltered in magni-
blade upon t h e bending mode of t h e (s + 1 i)th tude b y adding to any row multiples of t h e remaining
blade, and dt is the total effect of the bending mode in rows, we form new first and second rows p / and p2 in
the sth blade upon the bending mode of the (s + 1 the following manner (p7 denotes the original j t h row
f)th blade. T h e torsion mode of t h e nth blade affects and p / the n e w j t h row).
the bending mode of the second blade in exactly the
same way t h a t the torsion mode of the first blade af- Pi' = Pi + w / _ 1 P 3 + o)jn~2pb + ... + Wjp2n-l
fects t h e bending mode of t h e third blade, and so on. (6)
P2 = P2 + 0)jn-lpi + Ujn~2pe + . . . + 0)jP2n
I t is to be noted t h a t the degrees of freedom (charac-
terized b y au in bending and qa in torsion) are taken as This leaves | A \ unaltered in magnitude and in the
the same spanwise or radialwise functions for each following form (recalling t h a t (a>j)" = 1):
blade. In other words, some single bending mode is
Aj BJ AJ B, B,
taken as the bending mode for each blade and some AJ Bj
single torsion mode is t a k e n for the torsion mode of all
blades. No assumption is m a d e beforehand, however, LJ -Oj Cj Dj DJ
CJ D,
with regard to the relative magnitude or phasing of the Uj COj CO/ CO/ Uj
occurrence of the bending and torsion modes in any one \A\ = dn bn di bi a2 b2 . Q>n1 bn-l
or in different bladesi.e., the vector solution a is ^n dn di d! c2 d2 Cn-1 dn-l
completely unrestricted.
We now proceed to show, b y a s t u d y of the deter- (Remaining rows unaltered)
m i n a n t | A | of matrix A, t h a t a great deal can be said c2 d2 Cz dz 4 d\ ... C\ d\
in advance a b o u t the resulting form of q and t h a t , in (7)
fact, t h e problem m a y be reduced from one requiring
2n degrees of freedom to one with only two degrees (in Suppose now t h a t we expand the determinant \A\ by
general the reduction is from pn to p degrees). T h e the Laplace expansion with respect to the first two rows,
s t u d y of | A | stems from an extension of the special using I A I in the form (7). We see t h a t all minors of
circulant determinant (to which | A | would reduce if the form
we allowed initially only one degree of freedom per
Aj Aj Bj Bj
blade).
8
Let o)p be the pth of the n nth roots of unityi.e., (COj) (C0,)*|
and
(wy) s
(,)*
Cj
c7 D^ Dj
(4) (C0,-) S (Uj)'1 (COJ)8 {UjY

define the scalar quantities


vanish since they are expressible as determinants with
A1 = (ai + a2ccj + azaij2 + . . . + anoijn~l) identical columns. Moreover, all nonzero minors
Bj = (6x + b2a>j + hail + . . + bnU*-') formed from the first two rows contain
(5)
Cj = (Ci + C200j + C30)/ + . . + CnCo/"1)
Dj = (di + d2ccj + dsooj2 + . . + dnCOj"-1) \Aj Bj
(8)
Cj Dj\
I t should be noted t h a t , in view of the fact t h a t

CO/ = 1 for all j as a factor. T h u s , since Aj is a factor for each of the n


SYSTEM MODE SHAPES 57

different values of j (corresponding to t h e n different M, for the present development, assuming t h a t if nec-
nth roots of unity), we m a y express | A \ in the product essary, t h e entire process presented herein will be re-
form peated with new trial values of M until a trial value of
M and the resulting critical flutter velocity V are con-
L4, 5,1 sistent with the free-stream sound velocity. More-
K n A, =K n (9)
5 = i over, we mean by b (semichord) V, k, or any other pa-
with i as yet undetermined; in fact K m a y depend rameter, the value of the parameter a t some reference
upon the various a's, &'s, c's, and d's until we prove station along the blade span, since in rotating blade
otherwise. Now the product (9) is already of order n systems with radially varying chord t h e usual defini-
in any ajy bj} cjf or dj,' hence K cannot depend upon tions of k, V, and b actually vary along a blade. Further,
these elements, b u t m u s t in fact be simply a numerical we assume t h a t the velocity V is interchangeable with
constant since | A | itself is of order n in each of the %, the rotor velocity 0, since in general either V, 0, or
bjf Cj, dj. Moreover, since the leading diagonal term of the ratio of the two will be fixed in advance.
IA | contributes (ai)n-(di)n we see immediately t h a t K Returning for convenience, b u t with no loss of gener-
must be unity. T h u s we obtain the expansion of | A | ality, to the case wherein two degrees of freedom are al-
in the particularly simple form lowed to each blade, we continue the s t u d y of the system
Downloaded by OLD DOMINION UNIVERSITY on June 1, 2014 | http://arc.aiaa.org | DOI: 10.2514/8.3502

flutter problem through an examination of the factored


A, Bj
I A I = n A, = n (10) form (10) of the flutter determinant, \A\ . Obviously,
j = i j = i Cj Dj\ if IAI is to vanish, as it m u s t for existence of a non-
where the product extends over the n nth roots of trivial flutter mode, a t least one of t h e n factors A,
unity. m u s t vanish. Hence the scheme which suggests it-
At this point we m a y remark t h a t the initial use of self a t first is the following. Annihilate, in turn, each
more t h a n two degrees of freedom per blade would of the n minors A, by a proper choice of k and subse-
have led, through similar arguments, to t h e same form quent solution for V (call it Vj). This m a y be per-
of expansion for \A\ except t h a t the factors A, would formed, for each A7, by the method of successive trial
be of order p instead of order twoi.e., values of k, known as Theodorsen's method 6 or b y t h e
method of including an undetermined parameter.
A
Ahp Having fixed upon a trial value of k, frequency m a y be
A, replaced by velocity through the relation
\A\ = n A, = n (11)
j = i j = i kV/b (12)
Ajpl A,
Then the elements of any A, become complex first de-
the elements of these A7 being defined in the same man- gree polynomials in V2, and the expansion of A, becomes
ner [see E q . (5)] as the Ajt Bj} Cj, Dj were defined in a second (in general pth) degree polynomial in V2, with
the two degree of freedom per blade situation. This is complex numerical coefficients. Separating the coef-
easily seen b y remarking t h a t the original p X n order ficients into real and imaginary parts, for each A;-, we
determinant would be expanded, after Laplace's get a pair of simultaneous algebraic equations in V2.
method, by the first p rows following a process such as These are then solved respectively for the roots VR2,
Eq. (6) in which, to each of the first p rows were added VJ2 which, in general, will not be the same. New
linear combinations of every pth following row multi- trial values of k are inserted and the roots V2 m a y be
plied by a suitable power of co,-. plotted versus k, until an intersection is obtained^
i.e., until some choice of k leads to a common solution.
(2) Determination of Critical Flutter Velocity VR2 = Vi2, for both the real and imaginary equations.
Now the flutter problem consists in finding the critical I n case several intersections occur, the solution with
velocity V and corresponding flutter frequency co such the minimum value of V is used, since this will provide
t h a t the determinant | A | vanishes and a nontrivial the most critical condition for the particular minor A^
system mode q is possible as a solution of Eq. (2). T h e under consideration. T h e final critical velocity corre-
lowest velocity V which causes [ A | to vanish and sponding to the jth minor A?, will be called Vj and the
which, with the corresponding value of co, is consistent associated k and oo, kj and ooj. T h e process is repeated
with an initially chosen reduced frequency k ( = oob/ V) for each A^, until values of Vj} kj and co;- are available
is the critical flutter velocity. T h e nontrivial vector for each of the n minors. T h e minimum V} then gives
or vectors q which then satisfy Eq. (2) constitute the the flutter velocity, and the associated kj provides the
system flutter mode or modes. I t is to be recalled flutter frequency co; through E q . (12). If one or more
t h a t the elements aJ} bjt cjf dj, of A involve k and, in of the A^ fails to lead to a physically realizable (real)
compressible flow, the M a c h N u m b e r M, in such a pair, Vj, CCJ then this index is merely neglected in the
complicated transcendental form t h a t it is necessary to comparison to determine the minimum Vj. Summariz-
insert numerical trial values of these parameters, k and ing, we have found the velocity V and the associated k
M, into A before proceeding with the flutter problem and o) which annihilate one (or possibly more t h a n one)
or, alternatively, to include an undetermined parameter of the minors A; and which achieve this annihilation
in the analysis. We will forget a b o u t M a c h Number, with minimum velocity. This group, k, F, oo would
58 J O U R N A L OF THE A E R O N A U T I C A L S C I E N C E S JANUARY, 1956

have been t h e one arrived a t b y t h e far more cumber- rapidly with increasing distance, we see t h a t Aa, Ba,
some procedure of annihilating t h e original flutter Ca, Da and hence also PR and Pi, take the form of finite
determinant \A\ directly and choosing t h e annihilat- complex Fourier series in t h e angle cr, with rapidly de-
ing group with minimum velocity. creasing coefficients for the higher harmonics. T h u s
A t this point a tremendous simplification suggests it is to be expected t h a t V2(a) and k{a) will v a r y
itself for turbine and compressor blade systems with smoothly in a despite the presence of high harmonics
large numbers of blades. For large n, the nth roots of (for large n). Indeed, as will be seen subsequently, if
u n i t y approach a condition of density a b o u t the unit the coupling between blades is of an aerodynamic na-
circle. T h u s we replace the preceding process of pick- ture only, then for incompressible flow t h e infinite
ing a minimum Vj from among a large set of discrete Fourier series corresponding to E q . (14) are expressible
values b y a minimization process with respect t o t h e in closed form as smooth (actually analytic) functions
root angle, a. I n other words, we set equal to zero the of a for 0 < a < 2ir. Hence the plotting process to
general minor locate a minimum V and the replacement of t h e mini-
mizing angle a b y the nearest admissable value av
A9 Bt
(13) should be a valid means of obtaining the critical flutter
velocity and associated frequency. If t h e high har-
Downloaded by OLD DOMINION UNIVERSITY on June 1, 2014 | http://arc.aiaa.org | DOI: 10.2514/8.3502

wherein the n discrete values of t h e nth roots of u n i t y monics occurred with large coefficients then t h e plot-
are replaced b y t h e angle a which is temporarily as- ting process and the replacement of the minimizing value
sumed to be continuously variable between 0 and 2-ir, of the angle a b y an adjacent admissable value av would
and wherein we use an extension of E q . (5a) to define be open to question, b u t the physical n a t u r e of the prob-
Aa, -Dai Ca, Da\ lem eliminates this difficulty.
r Looking a t this point in a slightly different light, it
Aa = (ai + a2eta + ane la + aze2t<r + an + ) seems reasonable to believe t h a t a blade system with n
B9 = (ftx + b2ei<T + bne~iv + he2i + br -ie + ) blades, where n is large, will exhibit a critical flutter
(14)
velocity which differs only slightly from a system with
Now t h e annihilation of the general minor A,,, leads to identical blades, identically supported, b u t (n + 1) in
the two equations, number. This reasoning provides additional support
for our contention t h a t V(<r) will v a r y "smoothly"
PB(V\ k, a) = 0
(15) with cr, thereby justifying the minimization process
Pi(V\ k, a) = 0
and the replacement of the minimizing angle a, b y its
in which PR, PI are polynomials in V2 with coefficients nearest admissable value, <rv.
depending transcend en tally upon k and upon a. After
insertion of a trial value of k, Eqs. (15) m a y be solved, (3) Determination of the Flutter Mode
respectively, for VR2(<J) and Vi2(a). W e shall show Having determined the critical flutter velocity and
subsequently t h a t velocity (and frequency) m a y be the associated values of k, it remains to compute t h e
expected to v a r y smoothly with t h e angle <J. T h u s , corresponding flutter mode q. This mode t u r n s out to
to each of several (say ten) values of a between 0 and be very easily described, and in t h e case where \A\
2w we find t h e value (or, possibly, values) of k which vanishes simply (or doubly in the case of a symmetric
gives consistent roots, blade configuration) to have a very interesting physical
relation to t h e angle a or <sv. Consider t h e minor A
TV(cr) = VT2(a) (16)
which is annihilated b y t h e critical velocity Vv and t h e
2
T h e consistent roots' V (o) and t h e corresponding associated kvi forgetting, for the present, t h e possibility
values of k = k(<r) are then plotted against a. T h e of multiple degeneracyi.e., assume for t h e present
flutter condition is found simply b y picking the mini- t h a t A vanishes simply and t h a t no other Ay vanishes
m u m value of V2(ap) where <rv is one of the admissable upon insertion of V = Vv, k = kv. T h e n
discrete values of a.
\AV Bv
0 (18)
<TV = 2-Ku/n (17) \CV Dv
T h a t is, on t h e curve (or curves, in case more t h a n one and there exists a nontrivial vector (qa, qu) such t h a t
consistent condition exists for some range of values of YAV / j Vqa'
a) of V2 versus a we find the minimizing value of a and = 0 (19)
t a k e t h e nearest admissable <rv with its corresponding
Expanding the elements AVBV etc., and t h e Eqs. (19),
Vv, & as t h e flutter condition.
we find
T h e replacement of t h e problem of annihilation of n
determinants Av b y t h e annihilation of the general (ai + aw,, + a3co,2 + . . . + anuvn-l)qa +
2 n-l\n =
minor Aa a t a m u c h smaller (than n) number of points (6i + &2w + huv +...+& o
and subsequent plotting of annihilating combinations (20)
2 n l
(ci + c2co + czu + . . . + cno)v )qa +
V, k, versus a to determine t h e critical flutter condition
(di + d2oi, + dzu2 + . . . + dnuvn~l)qu = Oj
is justified as follows: referring to E q . (14) and remark-
ing t h a t the influence of distant blades becomes small B u t these are equivalent to
SYSTEM MODE SHAPES 59

ai bi a2 b2 az h \~q
C\ d\ c2 d2 c% fl>3 cn an qu
uvqa
uvqu
uv2qa
= 0 (21)
u2qu

*vn-lqa
l
U/- q j u

Or, in other words, the vector under the condition t h a t the minor A is annihilated b y
insertion of the critical flutter velocity. Now multiply
qa both Eqs. (20) through by co obtaining
<lu
uvqa (an + diuv + a2uu2 + . . + an--i(avn-l)q_a + J
Downloaded by OLD DOMINION UNIVERSITY on June 1, 2014 | http://arc.aiaa.org | DOI: 10.2514/8.3502

uvqu (bn + &ico + b2u2 + . . + bn-iwvn-l)q_u = Of


(22)
(cn + cxuv + c2u2 + . cn-navn-l)qa +
{dn + dio)p + d2uv2 + . . + dn^iuvn-l)qu = 0/
l
0)p q* (23)
l
J*v qu^
upon recalling t h a t {uv)n = 1.
is a solution of the first two of the original 2n E q s . (3) Eqs. (23) m a y be rewritten in the form

dn bn Q>\ h a2 b2 . an-i bn-l ~q


&n dn Ci dx c2 d2 . Cn-1 dn-l qu
^vq<x
uuqu
u2qa
= 0 (24)
uv2qu

uvn- lq<\
l
\j*vn~ qu]

T h u s the vector q of Eq. (22) is also a solution of the N e x t we consider the possibility of multiple-degener-
third and fourth equations of system (3). B y multi- acy. Multiple-degeneracy of A, it should be noted,
plying Eq. (20) by successively higher powers of cop we implies t h a t upon insertion of the critical flutter veloc-
see t h a t Eq. (22) is indeed a nontrivial solution of t h e ity V = Vvi either one of the minors vanishes multiply
complete original flutter equation. T h e nontriviality or several of the minors vanish simply or multiply for
follows from the existence of t h e nontrivial pair qa, a single value of k. If only one minor, A^ is annihilated
qu. Now in the case where insertion of V = Vv causes b y the flutter velocity Vv and if this minor is doubly-
only simple degeneracy of the matrix A, our solution degenerate (or possibly multiply-degenerate in the case
(22) is unique to within an arbitrary multiplicative of ^>-order minors) then there exist a t most two (or
constant. Moreover t h e form of the solution (flutter possibly more in the ^>-order case) linearly independent
mode) is amazingly simple. qa, qu is simply the vector solution vectors {q a , qu}. T o each of the linearly inde-
solution to the homogeneous second-degree system (19) pendent blade modes there corresponds a unique
which is easily determined once the elements of A are (within a multiplicative constant) system mode of pre-
available, and this represents t h e fundamental blade cisely t h e same form as Eq. (22). Hence, in this case,
mode. All other blades flutter with exactly the same the final system flutter mode is expressible a t worst as a
blade mode, with same magnitude of occurrence of this linear combination of solutions of the form (22) with
mode (all the powers of coy are of unit magnitude) and different blade modes b u t equal frequencies and phase
with a phase shift equal to the angle av occurring be- shift between blades. If Vv, kv cause the simultaneous
tween successive blades. T h u s , as soon as we find the vanishing of more t h a n one minor A, then to each
angle, <rvf if A is simply degenerate upon insertion of vanishing minor there corresponds one or more solu-
Vv, we can solve a second (or, in general a pth) order tions of the form (22). T h e final system flutter mode
system for the blade mode and immediately write down will then once again be expressible as a linear combina-
the system mode solution with no further effort. tion of solutions of the form (22) with equal frequency
60 J O U R N A L OF T H E A E R O N A U T I C A L S C I E N C E S J A N U A R Y , 1956

b u t with t h e possibility of n o t only different blade quasi-multiple-degeneracy occurring simultaneously,


modes b u t also different phase shift between blades for with still greater complications.
each of t h e component vectors. Finally, t h e entire preceding discussion with regard
Of special interest is t h e symmetrical blade and flow to flutter mode is easily generalized t o t h e case in which
configuration (zero stagger, zero incidence, a n d sym- p degrees of freedom are allowed to each blade. Here
metrical elastic a n d dynamic influence) in which a t the equation corresponding to (19) will possess a non-
least double-degeneracy will always occur. F o r in this trivial >-order blade mode, and t h e process which led
case, up to t h e system mode (22) is duplicated for t h e higher
order system. Hence again in t h e simply degenerate
a2 = any a 3 = an-h a4 = aw_2, etc.
case, a single blade mode of p components characterizes
h = bn, h = & w _i, b4 = bn^2, etc.
(25) all blades, a n d all blades experience this mode with
2 = cn> 3 cn\, c = cn2, etc.
equal magnitude, equal frequency, and with a common
d2 = dn, d3 = dn-i, d = dn^2, etc.
phase-shift angle a between adjacent blades. Multi-
Hence, t o a n y minor A a corresponding minor AM_ ply-degenerate cases lead to flutter modes which are
will vanish with t h e insertion of t h e velocity Vv. This expressible as linear combinations of solutions of this
is easily seen b y introducing E q . (25) into t h e defining type. I t is t o be emphasized t h a t multiple degeneracy
is of no consequence in t h e determination of t h e flutter
Downloaded by OLD DOMINION UNIVERSITY on June 1, 2014 | http://arc.aiaa.org | DOI: 10.2514/8.3502

relations (5) for Av, By, Cp, and Dv. Under symmetry,
velocity Vv. Once t h e minimum Vv V(av) is found,
Av = {a,i + 2a 2 cos <rv + 2a 3 cos 2uv + this is t h e critical flutter velocity, and t h e possibility
2a{yi + i)/ 2 cos [(n 1)/2](TV\ fornoddf t h a t this V m a y cause | A \ to vanish multiply does n o t
(26)
Bv = {bi + 2b2 cos <JV + 2&3 cos 2<JV + .. alter this fact.
+ 26(M+i)/2 cos [(n l)/2]ap\ for n odd J
etc. (4) Significance of the Flutter Mode
and Having completed the analysis of t h e multi-bladed
system in very general terms, we m a y now give an in-
Av = {<2i + 2a 2 cos o- + . . 2an//2
2 cos [(w 2)/2R + terpretation to the results which simplifies tremendously
ina /2 ) for w even
tt(n + 2)/2 & v the linear flutter analysis of a n y blade row. Examina-
etc.
tion of E q . (19) or t h e extension of E q . (19) to t h e p-
and obviously, if Vv causes A to vanish it will also cause order case, shows t h a t this corresponds actually to a
AM_ to vanish, since situation in which we solve for t h e flutter of only one
blade when this blade is under a very special form of
cos 2-K(m/n)(n v) = cos 2ir(m/n)v\
influence from all other blades; namely, all other blades
and (27) oscillate with the same blade mode, t h e same magnitude
" / ( 17 It* TT-t It f* I of occurrence of this blade mode, and with a phase
6'( - ) = ^ f()r ^ eyen
shift of (as y e t undetermined) angle a between adjacent
for all integers m. Hence, if only t h e two minors A blades. Since under multiple-degeneracy t h e flutter
and AM_ vanish (simply) upon insertion of F, &, then mode is still composed of a linear combination of modes
the solution (flutter mode) will be expressible as a linear of this type, t h e present considerations cover t h e mul-
combination of t w o modes of t h e form (22) with identi- tiply-degenerate case as well.
cal blade modes and with equal b u t opposite phase Keeping t h e foregoing remarks well in mind, t h e solu-
shift between blades. This satisfies our intuitive feel- tion of t h e flutter problem for a cyclic arrangement of
ing a b o u t t h e symmetric case, since, with s y m m e t r y of m a n y blades m a y be performed as follows. Allow only
flow and configuration, there should be no "preferred enough degrees, p, of freedom to describe satisfactorily
direction"i.e., there is no more reason w h y the jth. the motion of a single blade in flutter. Derive for one
blade should lead t h e (j l ) t h blade b y angle uv than blade t h e air forces experienced b y this blade when it
lag t h e ( j l ) t h blade b y this angle. oscillates in bending and twisting a n d all other blades
I t should be noted t h a t there is a possibility of t h e oscillate in t h e same blade mode with equal magnitude
occurrence of another type of degeneracy which we and with a phase shift of some undetermined angle a
shall call "quasi-multiple-degeneracy." This situation between adjacent blades. Do t h e same for t h e kinetic
occurs whenever | A | vanishes with two or more dif- and potential energy matrices. I t is to be emphasized
ferent values of k, each being associated with t h e same t h a t no restriction is involved in this assumption and
value of V, a n d when this V is t h e minimum of all an- t h a t all interblade interference is correctly accounted
nihilating Vs. Physically, this implies t h a t as the for. T h e n solve t h e flutter problem for this single
critical velocity V is approached, flutter m a y occur in "equivalent b l a d e " leaving t h e phase angle a unde-
either of t w o different modes, each having a different termined. Finally, minimize t h e flutter velocity with
frequency, or in some undetermined linear combination respect t o a t o obtain t h e critical flutter velocity a n d
of these two modes. Since the possibility of occurrence replace a b y t h e closest admissable value av. If one
of this quasi-multiple-degeneracy is rather remote, we value of <TV minimizes V, a n d if t h e p order determinant
shall n o t dwell further upon t h e matter, except to note vanishes simply upon insertion of t h e critical Vv and av
t h a t there is also t h e possibility of both multiple- and then t h e system flutter mode m a y be written directly
SYSTEM MODE SHAPES 61

as given by Eq. (22). If the ^-order determinant ordinates (with harmonic time dependence, elw\ re-
vanishes multiply or if several distinct values of <JV moved) corresponding to the torsional and flexural
give the same minimum value of F f o r the same k, then, degrees of freedom in the jth blade. Now we arrange
while the system flutter mode cannot be written down the complete flutter mode vector (call it q) in the form
expressly, still it is known to be expressible as a linear
combination of a finite number of easily determined
linearly independent vectors of the form (22). T h e qal
occurrence of quasi-multiple-degeneracy does not affect
the critical velocity V, b u t does again complicate the q*
flutter mode prediction. qu2
(28)
Finally, it should be remarked t h a t a phase shift per
blade of angle <rv, where (et<rv) is one of the nth roots
of unity, is consistent with t h e physical n a t u r e of an n-
bladed system, for it is easily seen t h a t any blade is
"in phase with itself" as it certainly m u s t be in actual- qun
ity. This suggests a further simplification in the prob-
Downloaded by OLD DOMINION UNIVERSITY on June 1, 2014 | http://arc.aiaa.org | DOI: 10.2514/8.3502

lem of obtaining the interference air forces for a blade where, once again, harmonic time dependence has been
row with a large number of blades. We merely assume removed and t h e q's are simply complex amplitudes.
t h a t any blade has an "interference effect on itself" as T h e complete, final system of flutter equations will
though it were located n blades away from itself, and then be expressible in the simple form
this allows us to express the entire interference for all
Aq = 0 (29)
blades in the form of a complex infinite Fourier series
in the undetermined angle a, with coefficients depend- w h e r e in A h a s t h e f o l l o w i n g f o r m :
ing upon the blade number multiplied b y t h e gap. In
7 8 V 8 rj . . 8 V
effect we replace the finite number of blades n by an
a ai 6i a2 b2 .. . an bn
infinite number, and hope t h a t the resulting Fourier
series will be recognizable in closed form. In the case 0 C\ dx c2 d2 Cn dn
a dn bn ai 6i .. 0"n-\ bn-l
of a cascade with or without stagger and with zero or
very small flow deflection in incompressible flow, it is A = na Cn dn C\ d! .. cn-\ dn-l
(30)
an-i bn-l dn bn an-i bn-2
indeed the case t h a t the Fourier series m a y be replaced
j8 Cn-1 dn-i Cn d n Cn-2 dn-2
b y a comparatively simple single function of a, valid
for 0 < a < 2T. This will be described in detail in a
subsequent report.
Oi d2 b2 d\ bi
(5) Extension to the Case Where the Rotor May Oscillate $ C2 d2 C\ di
in the Rotary Direction
T h e justification for writing A in the form (30) follows
We now extend our considerations of blade system immediately from the physical arrangement of the
flutter modes to the situation in which the blade row is blades upon a rotor. T h e form of the submatrix ob-
mounted upon a rotor of finite m o m e n t of inertia. We tained b y deleting the first row and first column is a
assume t h a t the rotor is rigidly constrained to remain direct consequence of t h e fact t h a t the blades are ar-
in its plane of rotation (no axial deflection) b u t t h a t it is ranged cyclically about the rotor and the nth blade
capable of rotary deflection and t h a t , through the effect affects the first blade in precisely the same manner as
of the rotor shaft and of other components mounted on the first blade affects the second, and so on. I t is to be
this shaft, we can a t t r i b u t e to the rotor a certain spring noted t h a t A allows for all types of coupling, aerody-
constant and a damping constant which govern re- namic, elastic, and dynamic, between the blades and
spectively the resistance to angular deflection and rate for t h e influence of the blades upon the rotor and of the
of deflection. This makes it necessary to introduce an rotor upon the blades. T h e first element y in t h e
additional degree of freedom q1 corresponding to the principal diagonal of A includes the rotor inertia, the
rotor deflection angle. Let effective rotor spring constant and the effective rotor
damping constant. T h e alternating form (6, 77, 8,
<Z = feiat
rj, . . .) of the remainder of the first row follows from the
and then drop the bar (-) in further considerations as fact t h a t the bending mode in all blades affects the rotor
we have done tacitly in E q . (1). qaettat is the instan- equally, and t h e torsion mode in all blades affects the
taneous angle through which the rotor is deflected away rotor equally. Likewise, the first column { a, /3, a, /3
from its equilibrium rotating position tit and is con- . . . } reflects t h e fact t h a t t h e rotor deflection affects
sidered positive in the same sense as 12. all blade bending modes identically and affects all
Once again we restrict our detailed considerations to blade torsion modes identically. N o t e t h a t no assump-
the case where two degrees of freedom are allowed to tions are implied in the form of A other t h a n the re-
each blade, and again the extension to more degrees is striction to identical blades, identically supported and
trivial. Again, let qj, qj denote the generalized co- equally spaced.
62 JOURNAL OF T H E AERONAUTICAL S C I E N C E S JANUARY, 1956

W e now proceed in a m a n n e r similar, b u t not identi- p2 = P2 + P4C0J,-1 + P6CO-2 + pgCOy~ + P2nUv


cal, to t h a t pursued in the case where t h e rotor inertia Pz' = P3 + P5C0,,-1 + P7CO~2 + p9C0
3
" + + P2n+IMV
was assumed infinite. Letting co denote the z^th of v = 1,2,3, ... (n - 1 )
the n n t h roots of unity, (33)
U/j/ K/ (31) (The value of \A\ is unchanged b y substituting the
new rows p 2 ' and p 3 ' for the original rows p2 and p3,
and defining Av, Bv, Cv, Dv as beforei.e.,
since we have merely added to the original rows, linear
Av = (ai + a2co + a^co,- 1 + a%wv2 + aw_ico-2 + . . .) ] combinations of succeeding rows). Now making use of
Bv = (bi + b2o)v + bnuv 1 + &3co2 + &_:iw ! the fact t h a t co/ = 1 for all v and
etc. +
v = 0,1,2, . . . . (n - 1) = 0
YJ tO*
(32) p = 0 P = i
for v j * 0
we leave the value of the determinant | A | of matrix A (34)
for v T^ n
unchanged b y forming new second and third rows as
follows (pj denotes an original row and p / , the altered we m a y write the determinant \A\ of matrix A in the
Downloaded by OLD DOMINION UNIVERSITY on June 1, 2014 | http://arc.aiaa.org | DOI: 10.2514/8.3502

row): form

7 8 V 8 V 8 V . 8 V
Av Bv A Bv A, B
0 A, Bv co C0 co,2 co/ co/"1 cV^1

0 cv Dv c, cv D.
00v 00v co/ co,2 ' ' ' CO/"1 CO j ,

a an bn dl h a2 h . aM_i bn-l
13 ^n dn C\ di Ci di Cn-l dn-i (35)
a an~i bn-l
P Cn1 dn-i

[Remaining rows unchanged from original


form (30)]
valid for v 1,2, . . . (n 1)

T h e zero elements in the first column of each of the and third rows of E q . (35) contain
second and third rows follow directly from relation (34)
Bt
and from the m a n n e r (33) in which the new second and
third rows were formed. N o t e t h a t E q . (35) is not
valid for v = 0 or v = n since Eq. (34) does not hold as a common factor, for any fixed value of v. Since
in these cases. Now we consider t h e Laplace expansion this is true for v = 1,2,3, . . (n 1) we see t h a t \A\
of | A | b y its second and third rows, using the form (35) contains the (n 1) factors
for \A\. W e see immediately t h a t all minors of the
v = 1,2.3 . . {n - 1)
form A, = (36)
A
Av Av B. B
N o w we perform operation (33), only with v = 0
CO/ co/ co/ co
i.e., we form new second and third rows as follows (re-
and
calling t h a t coo = oo n = 1)
cv c, D, D
co/ co/ co/ oov P2 = P2 + PA + P6 + + Pin
(37)
PZ = PZ + P5 + Pi + ... + P2n + 1
vanish due to t h e fact t h a t they are expressible in terms
of determinants with identical columns. T h e only T h e resulting expression for \A\ [which again is un-
nonzero minors which m a y be formed from the second changed b y the process (37) ] is

7 8 V 8 V 8 rj 8 7}
na Ao Bo Ao Bo Ao Bo .. . Ao Bo
n(3 Co Do Co Do Co Do ... Co Do
a an bn ai h a>2 b2 an-i bn-i
A = (38)
13 ^n dn C\ dl 2 di Cnl U"nl

Oi
rerr Lainitlg rows unaltered from original form
(30)
SYSTEM MODE SHAPES 63

where A0 = ai + a2 + a3 + . . + an velocity and frequency we proceed exactly as for the


n n n infinite-inertia-rotor case with t h e reservation t h a t the
Bo = E K Co = Cr, 2 ^ 0 = <*r (39) root angle, av for v 0 becomes exceptional. We m a y
r = 1 r = 1 r = 1 elaborate somewhat on this point. In order for flutter
This time we expand \A\, in the form (38), b y its first (a nontrivial solution vector q) to occur, the determi-
three rows. Noting t h a t a n y third order minors formed n a n t \A\ m u s t be annihilated b y some combination of
from the first three rows vanish if they have two identi- velocity V and flutter frequency co (or, alternatively, V
cal columns, we see t h a t the only nonzero minors of and k). This is completely equivalent to the annihila-
order three contain tion of one or more of the factors

7 8 V Ao, Ai, A2, Aw_i


na A0 Bo (40)
T h u s we determine, for each factor AoAiA2 . . ., the V
n/3 Co Do
and k which annihilate this factor (picking the combina-
as a factor. tion with minimum V in case more t h a n one pair, V,
T h u s we have shown t h a t A I is expressible in the k, annihilates any factor) and remark t h a t the flutter
factored form condition is given b y the pair, V, k, with minimum V.
Downloaded by OLD DOMINION UNIVERSITY on June 1, 2014 | http://arc.aiaa.org | DOI: 10.2514/8.3502

Obviously, if the pair with minimum V corresponds to


\A\ = KAo n A, (41) one of the minors A [v = 1,2,3, . . (n 1)] and if the
= 1 pair k, V which annihilates
where K m a y contain any of the elements (aif bi} cit dif
A0 Bo\
7, a, /3, 8, 7]) of \A\ until we prove otherwisei.e., Ao =
Co Do\
all we have shown thus far is t h a t each of the factors in
the product (41) is indeed a factor of \A\. B u t every exhibits larger velocity V t h a n this minimum Vv, then
term in Eq. (41) is of (total combined) order (2n + 1 ) the flutter frequency and critical velocity for the finite-
in the elements (aif bif ch dif 7, 8, a, @, rj) of A and, inertia-rotor are identical to the frequency and velocity
likewise each term in the complete expansion of | A | is for t h e flutter condition of the infinite-inertia-rotor.
of order (2n + 1) in these elements. T h u s K m u s t Stated differently, if for a given rotor and blade system
be a constant, independent of these elements. Now we solve for t h e flutter velocity and frequency in two
we note t h a t the leading diagonal of | A | contributes ways, once assuming infinite rotor inertia and once
the term taking proper account of rotor inertia, spring constant,
damping and rotor-blade coupling, then we will find
7(ai)W
exactly the same critical flutter conditions (frequency
to the expansion of \A\, and the coefficient of y(ai)n and velocity) for the two cases, provided t h a t A0 is not
{dxY in Eq. (41) is simply K. Thus the factor with minimum annihilating velocity in t h e
infinite inertia case and A0 is not the factor with mini-
K = 1 m u m annihilating velocity in the finite inertia case. If
and we have finally A0 were the factor (under the infinite-inertia-rotor as-
(n - 1
sumption) with minimum annihilating velocity a n d / o r
Ao the factor with minimum annihilating velocity in t h e
\A\ =i n A,
more exact case of finite rotor inertia, then obviously
we can no longer state t h a t the critical flutter velocities
or
and frequencies would be equivalent in t h e two cases.
y/n 8 rj Now for t h e systems with large numbers of blades
L4 = n a Ao B0\ (42)
P Co Do
n (and finite rotor inertia) we m a y proceed in a manner
similar (but not identical) to t h a t followed in the in-
From t h e factored form (42) of | A | we m a y draw a finite-inertia-rotor t r e a t m e n t . Recognizing t h a t the
host of conclusions concerning t h e solution of the flutter ^ t h roots of unity become dense about the unit circle
problem for a blade row m o u n t e d on a rotor of finite as n becomes large, once again we annihilate the general
mass. First we note t h a t as the rotor inertia becomes minor
infinitely large, | 71 > 00} and we have simply
Av Ba
A, =
B, Ca Da
(43)
N- = 7
V =
n 0|
D,
where
Hence as | 71 - > o the problem reduces to the form
A, = (01 + a^a + ane-' + a^1" +
treated earlier. Atcr 1 \

Next we see t h a t the product (42) contains ' 'rotor- % Zta


Ba = (&x + b2e" + bne- + he +
t e r m s " only in the first factor, all other factors de-
pending only upon the blades. Moreover, these latter bn^e-^+ ..)
etc.
factors are identical to those of the infinite-inertia-
rotor case. In the determination of critical flutter (Again, recalling t h e smoothness of V(<r), k(a), we
64 J O U R N A L OF T H E A E R O N A U T I C A L S C I E N C E S J A N U A R Y, 1 9 5 6

need annihilate A^ a t only a relatively small number of satisfies t h e second and third equations of t h e original
values of a\) T h e annihilating values of V{a) and flutter equation system (29). Moreover, we see from
k(a) are plotted against a in t h e range 0 < a < 2T and Eq. (34) t h a t q, as given b y E q . (46), satisfies t h e first
the minimum of t h e discrete set of velocities V(<TV) = of Eqs. (29) as welli.e.,
Vvf is recorded. N o t e t h a t we do not a d m i t a0 (or an)
as a possible minimizing value of a nor do we state a t qj J2 to/ + quV J2 0 (47)
this point t h a t t h e above-mentioned minimum is t h e P = 0 p = 0
flutter velocity. T h e reason for these statements is, of Multiplying Eqs. (45) through b y co a n d recalling t h a t
course, t h e presence of t h e factor A0 in t h e expansion
of | A | and t h e fact t h a t A0 does n o t fall into t h e cate- (oj.r = 1
gory of Aa as a -> 0. Thus, to complete t h e determina- we see t h a t
tion of flutter velocity we find t h e velocity, V, a n d re-
(an + accp + a2o)u2 + a^vz + . .)qa +
duced frequency k, which annihilate Ao (picking t h e
pair V, k with minimum V if more t h a n one annihilat- (bn + hwv + b2u2 + bdcopd + . .)qu 0
(48)
ing pair exists) a n d compare with the minimum Vv. (cn + CiO)v + c2u2 + czuvz + . .)qa +

If t h e velocity which annihilates Ao exceeds t h e mini- (dn + dio)p + d2uv2 + dzuvz + . .)qu = 0]
m u m Vv then Vv is indeed t h e critical flutter velocity, Hence E q . (46) is a solution of t h e fourth a n d fifth of
Downloaded by OLD DOMINION UNIVERSITY on June 1, 2014 | http://arc.aiaa.org | DOI: 10.2514/8.3502

and t h e flutter frequency follows from t h e reduced fre- Eqs. (29). Continuing t h e process-(i.e., multiplying
quency k corresponding to t h e minimizing av. If t h e (45) b y successively higher powers of o)p}we see t h a t
velocity Vo which annihilates A0 is less t h a n t h e mini- Eq. (46) is indeed a solution of t h e entire system of
m u m of t h e velocities Vv then V0 is t h e flutter velocity flutter Eqs. (29) under t h e condition t h a t Av(v 9^ 0) is
and t h e corresponding k = ko gives t h e flutter fre- annihilated b y t h e critical flutter velocity a n d fre-
quency. quency.
U p t o this point we have said nothing about flutter Moreover, since we have stipulated simple degener-
modes q which correspond to t h e flutter conditions acy, E q . (46) is unique within a multiplicative con-
noted above. W e proceed now to show t h a t these stant.
modes bear a striking resemblance t o t h e modes ob- Now t h e solution (46) is of t h e same form as t h a t ob-
tained under t h e restriction of infinite rotor inertia. tained under t h e assumption of infinite rotor inertia
Suppose first t h a t t h e vanishing of one of t h e Av[v = with t h e exception of t h e additional rotor deflection
1,2, . . (n 1)] leads to t h e critical flutter condition component which is zero. T h a t is, if t h e critical flutter
and suppose t h a t this A vanishes simply upon insertion condition for t h e system with rotor of finite inertia
of t h e critical k, V ( = kv) Vv) a n d t h a t no other factor corresponds to t h e vanishing of one of t h e minors
of | A | vanishes with these k, V. Then, with t h e Av(v 9^ 0) then t h e flutter mode is one in which all
critical Vv, kv inserted blades oscillate with identical blade mode {qa, qu} a t
equal magnitudes a n d with a (nonzero) constant phase
0 lag angle o\,(= 2-irv/n) between successive blades.
Cv
Moreover, t h e rotor does n o t oscillate b u t remains un-
and there exists a nontrivial vector \qa, qu\ such t h a t affected. W e can give extra evidence in support of
Av B this conclusion, based on E q . (34). T h e rotor m a y be
(44)
DJ \su\ influenced b y t h e blades through some linear combina-
tion of blade torsional a n d flexural displacements,
Expanding E q . (44), we m a y write, velocities, a n d accelerations, a n d hence through some
(ai + a2co + azu2 + . . . + anwvn-l)qa + linear combinations of qa} qu to which velocity a n d ac-
(&i + b2wv + 630?/ + + bnuvn-l)qu = Of celeration m a y be reduced under harmonic motion.
2 n l (45) Now, since all blades affect t h e rotor in the same w a y
(d + C203v + CZ0)y + . . " cnwv - )q_a +
2 and since t h e blades display a constant nonzero phase
(di + d2co + d%uv + . . . + dnuvn-l)qu = 0)
lag for t h e condition where t h e rotor deflection van-
Now we see from E q . (45) t h a t t h e vector ishes, we see t h a t t h e net torque exerted upon t h e rotor
by t h e blades a t a n y instant m u s t vanishi.e.,
r
q<*
qu Torque = eiat E (TaqJ + TuqJ)
= 1
uvqa n 1 n 1
uvqu = T&Jat E W + Tuqueiat E (*>)>
3 = 0 j = 0
u2qa (46)
= 0 for v ^ 0, v ^ n
2
uv <Lu
(49)
This adds support to our somewhat surprising discov-
ery t h a t q1 = 0 in t h e case where critical flutter corre-
l sponds to t h e vanishing of some A with v ^ 0. N o w
i*
l
1uJ if t h e critical V and k cause t h e simultaneous annihila-
SYSTEM MODE SHAPES 65

tion of two or more Av(v ^ 0) or cause one Au(v ^ 0) zero-phase mode. If t h e zero-phase mode is not criti-
to vanish multiply, then we know t h e flutter mode cal in either case, then the results in t h e two cases are
m u s t be expressible as a linear combination of vectors identical. If t h e zero-phase mode is critical in the in-
of the form (46) and hence t h e rotor still exhibits no finite-inertia-rotor case then the zero-phase mode oc-
deflection. curs with no rotor oscillation. I t is difficult to gener-
We next discuss t h e flutter mode for t h e condition alize in advance, b u t if it should t u r n out, in practice,
where Ao = 0 leads to the critical flutter velocityi.e., t h a t t h e zero-phase mode is seldom or never critical in
where the V which annihilates Ao is t h e minimum anni- either t h e infinite-rotor-inertia or the finite-rotor-
hilating velocity. Now if t h e critical velocity and re- inertia case, then there would be little or no need to
duced frequency are inserted, then Ao = 0 and there deal with t h e rotor a t alli.e., infinite inertia could be
exists a nontrivial three-component vector \qQ, qa, qu\ assumed a t t h e s t a r t and no need would exist for the
such t h a t determination of the elements (7, 8, rj, a, f3). This,
however, is merely conjecture and, it m u s t be empha-
y/n 8 V -qal sized, is not to be assumed until definite evidence is
a Ao Bo Ice = 0 (50)
available to the effect t h a t Ao = 0 and A0 = 0 are
Co Do_ Au_ seldom or never t h e critical conditions.
Downloaded by OLD DOMINION UNIVERSITY on June 1, 2014 | http://arc.aiaa.org | DOI: 10.2514/8.3502

Expanding Eq. (50) we see t h a t t h e components g", T h e occurrence of multiple degeneracy will be men-
Qm u, which we have shown m u s t exist, satisfy t h e rela- tioned only briefly since the considerations involved
tions are similar to those for t h e infinite-rotor-inertia condi-
tion. Again we define t r u e multiple degeneracy as the
yqn + n8qa + nr)qu = 0
condition where one pair V, k causes several (Ao, A) to
aqQ + (ai + a2 + a 3 + . . + an)qa +
vanish or one of these minors to vanish multiply.
(6i + h + h + + bn)qt 0 (51) Under these conditions t h e flutter mode will be expres-
0(f + (ci + c2 + cz + . . + cn)qa + sible a t worst as some (in general undeterminable)
(dt + d2 + dz + . . + dn)qu = 0 linear combination of modes of the form (46) or (52).
Hence the vector If none of the vanishing minors is Ao, then the flutter
mode will exhibit no rotor oscillation. Quasi-multiple
Q degeneracy, the existence of two equal, minimum anni-
hilating velocities V corresponding to two different re-
q_u duced frequencies, leads to the possibility of flutter in
either of two different frequencies a t the single critical
(52) flutter velocity.
Now it is i m p o r t a n t to note t h a t all steps performed
up to this point m a y be generalized to the case where we
allow p r a t h e r than two degrees of freedom per blade.
Once again \A\ is of circulant-type form with respect
to its ^ t h order minors except for t h e first row and first
where {qn, qa) qu} is t h e solution of E q . (50), satisfies column which t a k e the respective forms
all (2n + 1) of the original flutter Eqs. (29) and, in
(7, &u 82, . 8P, di, . . 8P> 8h . .
case of simple degeneracy, is the unique solution of
[7, ah a2, . Ctp, Oil, . . aP) a\,
Eq. (29) to within an undetermined multiplicative
constant. Now the solution vector (52) displays in Again t h e replacement of rows 2,3 . . (p + 1) b y rows
general a nonzero rotor oscillation qQ and a blade sys- composed of t h e original row plus multiples of oov times
tem mode with zero phase lag. T h a t is, all blades os- each pth following row leads to a form analogous to
cillate with identical blade mode qa, qu a t equal magni- form (35) while leaving t h e value of \A\ unaltered.
tude (with equal frequency) and all blades are exactly Again we expand \A\ b y factors consisting of {n 1)
in phase. Moreover, the rotor is no longer unaffected t h order minors and one (p + l ) t h order minor. T h e
by the blades, b u t undergoes an oscillation of amplitude solution for the critical flutter conditions and modes fol-
\(t\ a t the same frequency as t h a t of the blade oscilla- lows in exactly the same way as for the case with p 2,
tion. T h e fact t h a t a nonzero rotor oscillation ac- and t h e general conclusions with respect to the resulting
companies the in-phase blade mode is consistent with flutter mode are unchanged.
other considerations, since with a = 0, sums of the t y p e I n summarizing our findings for t h e case of a blade
(49) are no longer zero, b u t r a t h e r add u p to n, the num- row mounted on a rotor of finite inertia we m a y state
ber of blades. T h u s all blades exert equal torque upon the following surprising conclusion. If t h e velocity
the rotor with equal frequency and in phase with one Vo which annihilates A0 exceeds the smallest of the
another. Obviously under these conditions, t h e rotor velocities Vv which annihilate the A(v ^ 0) and if the
will in general experience a nonzero oscillation. T h u s velocity Vo which annihilates Ao also exceeds t h e smal-
we see t h a t the only difference between the flutter lest Vvi then (1) t h e assumption of infinite rotor inertia
problem for a system with rotor of finite inertia and causes no error whatsoever, (2) flutter frequency and
t h a t for a system with infinite-inertia rotor lies in the velocity are t h e same for t h e infinite-inertia and finite-
66 JOURNAL OF THE A E R O N A U T I C A L S C I E N C E S JANUARY, 1956

inertia rotor conditions, (3) t h e blade mode and the Vo. T h u s , there exists some finite value 70 such t h a t
phase lag between adjacent blades are t h e same for the for | 71 > | 7o|, Vo > Vv and Vv is critical.
two conditions, and (4) t h e rotor undergoes no oscilla- A final remark m a y be helpful in the actual process of
tory deflection (g a = 0). On the other hand, if either plotting V(o) or V2(a) for the case of finite rotor iner-
Vo or Vo is less t h a n the minimum Vvi then t h e infinite- tia. While Ao = 0 does not actually fall into the scheme
inertia assumption leads to an error. If Vo is less t h a n of t h e finite-rotor-inertia case, still, the evaluation of
the minimum Vv b u t Vo is not less t h a n the minimum the velocity and reduced frequency, Vo and ko, which
Vv then the actual flutter mode (for finite rotor inertia) annihilate A0, provides useful end-points on t h e curves
corresponds to the combination Vy ku (which was not V{<r) and k(a) for plotting purposes.
critical in the infinite-rotor-inertia case) and flutter
occurs with zero rotor oscillation, blade mode corre- REFERENCES

sponding to A, and phase lag <rv between blades. If 1


Chang, Chieh-Chien, and Chu, Wen-Hwa, Aerodynamic
Vo is less t h a n the minimum Vv then flutter (for the Characteristics of an Oscillating Cascade, Air Research and De-
finite-rotor-inertia system) occurs with a nonzero rotor velopment Command, USAF, (AF 33(038) 12919), September 1,
deflection (q ^ 0) and with a blade mode correspond- 1952.
2
Lilley, G. M., An Investigation of the Flexure-Torsion Flutter
ing to the matrix for which Ao is the determinant.
Characteristics of Airfoils in Cascade, Report No. 60, The College
Downloaded by OLD DOMINION UNIVERSITY on June 1, 2014 | http://arc.aiaa.org | DOI: 10.2514/8.3502

Moreover flutter occurs with all blades in phase (a- = of Aeronautics, Cranfield, May, 1952.
3
0). Mendelson, A., Effect of Centrifugal Force on Flutter of Uni-
I t is interesting to state one of our results in a slightly form Cantilever Beams at Subsonic Speeds with Application to Com-
different way: if the system leads to a critical flutter pressor and Turbine Blades, NACA T N 1893, June, 1949.
4
Reissner, E., Wind Tunnel Corrections for the Two-Dimensional
mode with nonzero lag between blades under the as-
Theory of Oscillating Airfoils, Cornell Aeronautical Laboratory
sumption of infinite rotor inertiai.e., V0 > Vv for Report No. SB-318-S-3, April 22, 1947.
some vthen there exists a finite value of rotor inertia 5
Reissner, E., Boundary Value Problems in Aerodynamics of
such t h a t for any value of inertia larger t h a n this finite Lifting Surfaces in Non-Uniform Motion, Bulletin of the Ameri-
value the assumption of infinite inertia is exact. We can Mathematical Society, Vol. 55, No. 9, pp. 825-850, Septem-
ber, 1949.
can show this as follows: 6
Scanlan, R. H., and Rosenbaum, R., Introduction to the Study
r i. T/ ^ T/ u i ^o annihilates A0 of Aircraft Vibration and Flutter, First Edition, p. 266, The
Let Vo > Vv where < T7 ., M ^ MacMillan Company, New York, 1951.
( Vv annihilates A 7
Sisto, F., Flutter of Airfoils in Cascade, Doctoral Dissertation,
i.e., Vv is critical under the assumption of infinite rotor Massachusetts Institute of Technology, 1952.
8
inertia. Now as | y | becomes large Theodorsen, T., General Theory of Aerodynamic Instability
and the Mechanism of Flutter, NACA T R 496, 1934.
9
Ao - * 7A0 Turner, M. J., and Duke, J. B., Propeller Flutter, Journal of
the Aeronautical Sciences, Vol. 16, No. 6, pp. 323-336, June,
Hence Vo (the velocity which annihilates Ao) approaches 1949.

I n t e r a c t i o n of a S h o c k W a v e w i t h a T h e r m a l B o u n d a r y L a y e r
(Continued from page 22)

make it desirable to operate the test section a t reduced velocity and sulfur hexaflouride the lowest. T h e corre-
pressure. Since an alternative approach to the prob- sponding extremes of surface temperature would then
lem appears quite promising no serious effort has been be 4400K. and 54K. An experimental model has
made to resolve these difficulties. been constructed and we hope to report on the results
The principal mechanism b y which the thermal layer in a subsequent article.
disturbs a shock is a change in t h e local speed of sound.
REFERENCES
In a sense the change in density which accompanies 1
Lighthill, M. J., Reflection at a Laminar Boundary Layer of a
heating a t constant pressure is extraneous to the prob-
Weak Steady Disturbance to a Supersonic Stream, Neglecting
lem. Prof. Walker Bleakney has suggested t h a t the Viscosity and Heat Conduction, Quart. Jour, of Mech. and Appl.
change in sound speed be produced by allowing gases Math., Vol. 3, P a r t 3, p. 303, September, 1950.
2
of different molecular weight to diffuse through one Adams, M. C , On Shock Waves in Inhomogeneous Flow,
another. I n fact, the differential equations for dif- Journal of the Aeronautical Sciences, Vol. 16, No. 11, pp. 685-
690, November, 1949.
fusion and h e a t conduction are the same. A zone of 3
Ribner, H. S., and Moore, F. K., Unsteady Interaction of Dis-
nonuniform gas concentration suitable for experiments turbances with a Shock Wave, 1953 Heat Trans, and Fluid Mech.
analogous to those with a hot or cold plate could be Inst., also, NACA T N 2864 and 2879.
4
generated b y allowing one gas to diffuse slowly through Chang, C. T., On the Interaction Between Weak Fluctuating
the air above a porous surface for a short time before Fields of Sound, Vorticity, and Temperature with an Oscillating
Shock, The Johns Hopkins University, Department of Aero-
the shock wave arrives. A light gas corresponds to a
nautics, Preliminary Report, 1953.
hot region and a heavy gas to a cold one. Of t h e 5
Hess, R. V., The Interaction of a Shock Wave with a Thermal
gases readily available hydrogen has the greatest sound Boundary Layer, Unpublished.

Você também pode gostar