Professional Documents
Culture Documents
Acoustoelasticity: General Theory, Acoustic Natural. Modes and Forced Response To Sinusoidal Excitation, Including Comparisons With Experiment
Acoustoelasticity: General Theory, Acoustic Natural. Modes and Forced Response To Sinusoidal Excitation, Including Comparisons With Experiment
A comprehensive theoretical model has been developed for interior sound fields which
are created by flexible wall motion resulting from exterior sound fields. Full coupling
between the wall and interior acousticcavity is permitted. Anefficient computational method
is used to determine acoustic natural frequencies of multiply connected cavities. Simplified
formulae are developed for interior sound levels in terms of cavity, wall and external
acoustic field parameters. Comparisons of theory and experiment show generally good
agreement.
I. INTRODUCTION
The interaction between the internal (acoustic) sound pressure field and the (elastic) flexible
wall of an enclosure will be considered. One might think of this as the field of “acousto-
elasticity”. A good introduction to this genera1 subject is given in the book Sound, Structures
and Their Interaction by Junger and Feit [l]. Also the first author has briefly discussed this
subject in his book Aeroelasticity ofPlates and Shells [2].
From the point of view of applications, such problems frequently arise when the vibrating
walls of a transportation vehicle induce a significant internal sound field. The walls themselves
may be excited by external fluid flows. Cabin noise in various flight vehicles and the internal
sound field in an automobile are representative examples.
A more detailed discussion of the present work may be found in reference [3].
A direct and concise derivation of the acoustic pressure in a closed cavity due to arbitrary
motion of the flexible cavity walls will be given. The pressure is expanded in terms of the
normal modes of the rigid-walled cavity. The result, which is valid for any cavity geometry,
is given in the form of a set of linear ordinary differential equations for the response of each
normal acoustical mode. The equations of motion of the flexible wall are also derived in terms
of in vacua structural normal modes. Finally, the complete coupled fluid-structural equations
of motion are obtained.
Most studies of the acoustic pressure within a cavity due to motion of a flexible wall have
considered only simple harmonic motion, [4-71 although implicitly the more general case of
arbitrary motion can be handled via the Fourier integral theorem. However, a very simple
result for arbitrary wall motions can be obtained directly from the acoustic equations by
using Green’s Theorem without resorting to transform methods. The end result of the present
method is an expansion for the acoustic pressure within the cavity due to motion of the walls,
in terms of the normal modes of the cavity with its walls assumed to be rigid. The derivation
519
520 E. H. DOWELL, G. F. GORMAN, 111AND D. A. SMITH
applies equally to all cavity geometries, and as such is a generalization of the normal mode
expansion (or “guided wave” expansion, in the terminology used there) derived less directly,
and for rectangular cavities only, in reference [8].
Let the cavity occupy a volume Y, and be surrounded by a wall surface A, of which the
portion A, is flexible, while the remainder A, is rigid. If the fluid within the cavity is at rest
prior to motion of the wall, the fluid pressure pt satisfies the familiar wave equation, and
associated boundary condition:
Pp - (l/c;) Pp/W = 0, (1)
aplan = -p. a2 wlat2 on AF,
=o on A R. (2)
In these equations p,, and cc, are the equilibrium fluid density and acoustic velocity within the
cavity, and w is the displacement of the flexible portion of the wall in the normal direction n
(positive outward). (A list of symbols is given in Appendix 2.)
Equation (1) has rigidwall normal mode solutions Fneiwr, n = 0, 1,2, . . ., with the following
properties :
Vz F, = -(w:/c,)~F,, (3)
aF,lan = 0 on A=AF+AR, (4)
; F,F,,dc = 0, r # n,
i’
Y
= Iv;, r = n. (5)
cot is the nth acoustical natural frequency and F” its related natural mode. Superscript A is
to distinguish of and M,A from their structural counterparts to be discussed in a later section.
Note that equation (3) has the solution o,A = 0, F, = 1. All other frequencies wt, n = 1,2, . . .,
are positive and non-zero, however.
The wave equation (1) can be transformed into a set of ordinary differential equations in
time by using Green’s Theorem in the form
(pV’F,-F,V’p)dt:= (6)
*i
”
By defining
1
P,E- PF, dr, W,=; wF,,dA, (7)
POCl!lV J’ F s
V AF
and making use of the fact thatp and F, satisfy equations (1) and (3), and boundary conditions
(2) and (4) the following ordinary differential equations for the acoustic modes are obtained
from equations (6) :
~n+o;:lP,=-Cii,AF/V. (8):
A dot ( ) denotes differentiation with respect to time. The quantities P,(t) are the coefficients
in an acoustical normal mode expansion for the pressure :
P-PIP~~~=I,(P,IM,A)F,. (9X
n
t This is the perturbation pressure j?, of course. The ^ is dropped subsequently for convenience.
$ Note that for n = 0, one has the Helmholtz resonator result: i.e., F0 = 1, W, = (l wdA)/Ap, f’~ =
-W,A,/V, M; = 1; :. p=p,,c;Po=-yoc;(J wdA)/V.
ACOUSTOELASTICITY 521
Since the normal modes F, satisfy the homogeneous boundary condition (4) on the entire wall
surface A, the normal derivative of expression (9) does not converge uniformly on the flexible
portion Af of the wall surface. Expression (9) is suitable, however, for calculating the pressure
itself throughout the cavity and everywhere on the wall surface, including the flexible portion.
This specific result, although broadly indicated in reference [7], is due to Ventres (see
reference [2]).
When one of the walls of the cavity is highly absorbent it is often characterized by a simple
point-impedance model which states that
where the A subscript is used to refer to the absorbent wall characteristics: e.g., wA is the
absorbent wall displacement and zA is the absorbent wall impedance. The boundary condition
aplan = -pofc, on A, (2)
still applies, of course. By using expression ( IO), equation (2) may be written for the absorbing
wall as
+/an = -pop/za on A,. (2,‘)
Upon using expression (2J in equation (6) along with equation (9), equations (8) become
where
C,, Z (1/A,) { [ Vi FJ?,) dA
ove;A,
The effect of an absorbent wall is to couple all of the (rigid wall) acoustic modes. Of course,
as will be shown explicitly in what follows, the flexible wall also couples all of the acoustic
modes as well.
In many technical applications the flexible portion. A,, of the cavity wall may be a structural
element, such as a plate or shell. In such cases the wall deflection, w. is often expressed as a
series of the form
H’= 2 qm yy,, (II)
m
in which the modal functions Ym are defined over the region AF, their properties being deter-
mined by structural considerations. By using equation ( 1I), a set of structural modal equations
describing the wall motion can be derived. As will be shown subsequently, in these equations
the cavity acoustic pressure, p, appears in the form of generalized forces, Q,, :
First, however, the quantities W,, appearing in equations (8) and (8,) can be expressed. by
using equations (7) and (1 I), as
W” = 2 L”!nYnI. (1-J)
m
where
I-+ F,Y,dA. (14)
F J‘
*F
522 E. H. DOWELL, G. F. GORMAN, III AND D. A. SMITH
When the structure may be represented by a linear structural model the total fluid-structure
interaction may be treated in a simple way. Let the structure be represented by a linear
(partial, differential) equation:
sw+ma2w/at2=p=-pE. (16)
S is a linear differential operator, representing structural stiffness. For example, for an iso-
tropic, flat plate, Sr DV4, where V4 is the biharmonic operator. The second term on the
left-hand side is the structural inertia, m being structural mass per unit area. On the right-
hand side one has two pressure loadings, the first due to the cavity acoustics, and the second
due to some specified “external” agent.
For simplicity, we assume that the structural modes, $,,,, (cf. equation (11)) are normal
in vacua modes satisfying an eigenvalue equation,
Multiplying through by rl/. and integrating over AF gives, after using condition (18),
Id2qmldr
’+ &, anI M, = P,t, + Q:, (20)
where (recall equation (12))
Q; = P’I(/mdA, (21)
I
(22)
Hence the structural modal equations, (20), may be rewritten by using equations (23) as
(24)
In summary, the key relations are equations (15) and (24), with Q5 determined from its
definition, equation (22). Equations (15) and (24) are coupled acoustical-structural ordinary
differential equations which may be solved by any standard method. Since one normally is
ACOUSTOELASTI(.‘ITY 523
dealing with systems under externally forced motion, caused by the generalized forces Q&
the initial conditions on q,,, and P,, will usually be the trivial ones
qm = dq,ldt = P, = dP,Jdt = 0 at t = 0.
If 4 is expanded in its natural modes (they are obviously the same modes as for p, as can be
be seen from equation (25))
4 = Z a, F,. (26)
one can, from equations (9), (25) and (26) determine that
t15>*t
The underlined terms are the structural-acoustic coupling terms. The coefficients are anti-
symmetric (to within a multiplicative constant) and this characteristic, combined with the
first time derivatives of q,,, and a,, means that the equations possess gyroscopic coupling.
Thus it is known that (if C,, E 0, i.e., there is no absorbing wall) the coupled structural-
acoustic natural frequencies are real. Hence the system without absorbing walls is an un-
damped resonator. Moreover Meirovitch [9] (in a different physical context) has shown how
one may use conventional matrix eigenvalue methods to determine the natural frequencies
of gyroscopically coupled systems.
The same results may be obtained through a variational formulation. Previously Gladwell
[IO-I 21 and later Craggs [13-141 have considered variational statements in the context of
finite element representations for the coupled structural (wall)-acoustic (cavity) problem.
Craggs’ work has been carried furthest and we use it as a point of departureforourdiscussion.
Although as shown in reference [3] the same results are obtained as before, the variational
statement in its present form is somewhat awkward. This is because one must consider both
direct and complementary energies and work [13, 141. Hence, the authors’ preference is for
the previous formulation in which Green’s theorem is used for the cavity. Others may feel
differently, however, or perhaps be inspired to develop a more elegant and/or simpler
variational statement.
t Here, and in what follows, asterisks are used to denote equations which are key results.
524 E. H. DOWELL, G. F. GORMAN, III AND D. A. SMITH
For definiteness we consider two rectangular cavities although it will be clear that the basic
method is applicable to fairly general geometries (see Figure 1).
Ym
I I
i-l
Cavity / Cawty
1.
a Fi _._~ 1
b ~I-x
Figure 1. Sketch of coupled cavities.
Our interest here is in the acoustic natural modes of the two cavities connected by an opening
of area AF. In this opening there is, in general, a flexible structural member which deforms
(as indicated by the dashed line in Figure 1) and thus permits sound transmission from one
cavity to the next. All other portions of the cavity walls are rigid. As before the structural
member has a displacement, W, with a modal expansion
w= Z4m(r)$m(Y,Z)~
Nt
(28)
and the acoustial pressures in cavities a and b have modal expansions
The modal equations of motion for the acoustic pressures are then (recall section 2)
The + sign in the second of equations (30) and the minus sign in the right-hand side of
equation (3 1) result because the normal to the opening is outward from cavity a and inward
to cavity b.
As in section 2 one may introduce a velocity potential through Bernoulli’s equation
II.b = _po$a. b,
P (32)
with a modal expansion
(33)
V” M;“[g + w;“‘~] = +c; AF 1 L&,&, Vb M,A*[iif, + co;** + a”,]= -c; AF 2 L:,cj,, (34)
m In
Equations (34) and (35) are in the standard form permitting use of Meirovitch’s algorithm
[9] to determine the natural frequencies (eigenvalues) and modes (eigenvectors) of the two
coupled cavities.
Of special interest is the case with a pure opening between the two cavities: i.e., one with
zero mass and stiffness. In this case substantial further simplifications are possible. Thus
ACOUSTOELASTICITY 525
M, = u, = 0 and equation (31) becomes (one could deal alternatively with equations (34)
and (35))
Pa**
n
= T (A,/Ph) 2 {-w2q,L~m&JJ’ + wpb2)}. (3X)
m
Substitution of equations (37’) and (38) into equation (36) gives (before substitution one
should replace the summation index, nr, in equation (38j by I’to avoid confusion with the
m in equation (36))
where
Equation (40) is a non-standard eigenvalue problem because of the form that o2 takes in
Qrm (see equation (39)). However, it has one overwhelming advantage as compared to the
standard (modified for gyroscopic coupling [9]) eigenvalue approach embodied in equations
(34) and (35): The size of the matrix in equation (40) will be much smaller than that in
equations (34) and (35). This is because the number of two-dimensional opening modes, G,,
may be much smaller than the number of three-dimensionalcavity modes, F,"and F,9required
to achieve a given level of accuracy. This advantage will persist even when the opening is a
structural member of finite stiffness and mass, but the mathematics are a bit more cumber-
some.
In closing this section we quote, without detailed derivations. two generalizations of the
above results.
For an external wall of cavity a whose motion. w“, is prescribed (sinusoidal motion is
assumed for simplicity)
wEa = ; q: l+bm. (41)
The equation of motion is
z &, Pm + 0: = 0. (42)
where
where
Lb(r) Lb(a)
nr “In
M;b(_w2 + ofb’)’
(43)
where Agb is the opening area common to cavities a and 6, etc. The generalization to any
number of interconnected cavities should now be clear.
The relationship of this analysis to other methods also deserves brief mention. Morse and
lngard [7] have considered two coupled cavities employing a Green function approach. The
final result is the same as that obtained here: i.e., equation (39) (see Appendix B of reference
[3] for details). Alternatively, in the context of the variational formulation, as far as computing
acoustical natural frequencies of multiply connected cavities with pure openings and other-
wise rigid walls is concerned, the q,,, may be considered as Lagrange multipliers which enforce
the constraint that the pressures in cavities u and b must be equal at their common opening.
Hence the present analysis may be thought of as a component mode synthesis where the
components are the individual cavities. The results are entirely analogous to those previously
obtained in structural vibrations, except the roles of force and deflection are interchanged
[ 151. In structural applications, the Lagrange multipliers are forces of constraint enforcing
common deflections where two structural components are connected; in acoustical applica-
tions, the Lagrange multipliers are deflections of constraint which enforce common forces
(acoustical pressures) where two acoustical components (cavities) are connected.
The principal result of this analysis then is the determination of equivalent single cavity
modes where the single cavity is composed of several component cavities. Thus it is sufficient
to consider a single cavity in the subsequent discussion of forced response in section 5.
For a single cavity with a flexible wall and an external sound source, the theroetical model
has been verified experimentally by several authors [4, 5, 16-I 91.
Here we assess the capability of the model to describe accurately the acoustic natural modes
in multiply connected cavities. Once the combined natural modes of the multiply connected
cavities are determined and verified experimentally, they may be treated as one single cavity.
Hence the earlier work for a single cavity [4, 5, 16-191 then may be taken as experimental
verification for the forced excitation of multiply connected cavities as well.
The experimental studies with which numerical results are compared here were conducted
by Smith [20], who considered several geometrical arrangements for two acoustic cavities
with rigid walls and a partial opening between them (see Figure 2). In Figure 3 the lowest
longitudinal resonant frequencies found from theory (section 3) and experiment are shown
vs. partition size, y,, normalized by cavity height, d. As may be seen the agreement is good
with the exception of the lot mode for Q z 0.4 where a mechanical resonance of the loud-
0
B&K
Type 2603
microphone
ompllfier
Loudspeaker
/T MIcrophone
/
J _l==>: 1 l-l
_il 1
B8 KTyped002 standing wove apparatus
_I
[COb!!YIS 20 I” I 10nx4 In]
Figure 2. Experimental set-up.
Loudspeaker
30 mode
z~__a__d_ilr ____-
11 I.@
IO mode
00 mode (Helmholtr)
Figure 3. Cavity natural frequencies us. partition size (two-dimensional case, longitudinal (i.e., x-direction)
mode). Q = wujcn = 2af/c. -, Theoretical results; C, experimental results.
speaker used to excite the acoustic modes contaminates the data. In practical terms this is
unavoidable as both theory and experiment show that Q,, = O-5 + 0 as the partition is closed
(yJd= 0 -+ 0.5). In these experiments, c0 = 1117 ft/s, a = d= 10 inches and the width
dimension was 4 inches to provide two-dimensional conditions in the frequency range of
interest. The thickness of the partition (assumed zero in the theoretical calculations) is 0.5
inches as is the thickness of all external walls. The cavity is constructed from plexiglass.
In Figure 4, for a constant aperture (opening) size (y,/d = 0.5) the effect of aperture location
is studied. Both theory and experiment show a small effect.
528 E. H. DOWELL, G. F. GORMAN, III AND D. A. SMITH
Loudspeaker
0
1-1 0.1 0.2
yl,/d
o-3 0.4 0.5
Figure 4. Natural cavity frequency vs. aperture location (two-dimensional case, longitudinal (i.e., x.
direction) mode). B = C&X = 2&c; ye/d= 0.5. -, Theoretical results; A, experimental results.
Loudspeaker
4
20‘ A r: A A n ’
30 mode 7
00mode(Helmholtz)
yc ‘d
Figure 5. Cavity natural frequency us. partition size (two-dimensional case, longitudinal modes). Q =
wa/cn = 2Qf/C. -, Theoretical results; A, experimental results.
ACOUSTOELASTICITY 529
0. THEO
=2.00
R rXP=2.04
Figure 6. Longitudinal pressure distributions and resonant frequencies for the first seven (symmetrical
with respect to height) acoustic modes with a full opening between the two cavities. R = wajm. -,
Theoretical results; o, experimental results.
A second configuration studied by Smith consists of two acoustic cavities, one twice the
dimensions of the other, with rigid walls and a partial opening between them.
In Figure 5 the
lowest four longitudinal resonant frequencies from theory and experiment are shown L’S.
normalized partition size, y,/d. Again the agreement is good. In Figure 6, the longitudinal
pressure distributions (along with their resonant frequencies) are shown for the first seven
(symmetric with respect to height) acoustical modes with a full opening between cavities.
The agreement between theory and experiment is generally very good.
An exception is the agreement on the 12 and 30 modes. Theoretically the 12 and 30 modes
have similar resonant frequencies and experimentally they are very close. Therefore, it is
very difficult to excite cleanly each mode experimentally and thus the poorer agreement
between theory and experiment for these modes.
Altogether one concludes the theoretical results are a faithful description of the physical
model. For more detail the reader should consult reference [20].
The total pressure loading on the structural wall is now the sum of the internal acoustic
cavity pressure and an external (prescribed) pressure loading. Hence, the structural modal
equations are, with inclusion of a conventional model of structural damping,
MA& + 21, mm& + ~;q,l = Pm, (44
where
Qm=j$$mdxd~~ p=pC+pE, P’ = PO c; 2 Pn FnIM,A. (45)
n
AF
pE, the prescribed external pressure, is the new element in the theoretical model. Q,,, may be
written as
Qm= PO 4 A, ; U’. LnIM;) + Q:, (46)
where
Qm”
= - Il pE Il/,dxdy. (47)
P = -Pod. (50)
With I$Jrepresented by the expansion
$J=;U”F,, (51)
(52)
where
1’[K]
G[Mmom”
’1
0
AA PO C; cm VM; of .
These equations can be solved numerically by standard methods [9]. In the remainder of
this section, simplifying assumptions are considered which allow useful but approximate
analytical solutions to be obtained. These should suffice for a rough estimate and also serve
as a guide for more accurate and more elaborate numerical solutions. In all cases sinusoidal
excitation is assumed: i.e., pE =IjEeiwt ‘, where aE is the excitation frequency.
Because our model is linear any other time history can be considered through superposition
of sinusoidal excitation. Random excitation may be considered by using the sinusoidal
excitation results and power spectral methods [2 I].
ACOUSTOELASTICITY 531
One may anticipate the largest response will occur when the excitation frequency is near a
structural or cavity resonant frequency.
5.2.1. Exciting frequency = structural wall resonant frequency
If ‘I&% o,, where s denotes a structural resonant (natural) mode, then only qs will be
important and all other qn may be neglected. Then, from equation (49)
P” = -(AJ V) L”,CjS/(-ol + U$>, (53)
and, from equations (53) and (48),
(54)
From the right-hand side of equation (54) it is seen that if o;i2 > I$, then the nth cavity mode
gives rise to equivalent mass, while if 01’ < wf, then the nth cavity mode contributes an
equivalent stiffness. Note that the n = 0 cavity mode always contributes equivalent stiffness.
In many practical examples the structural wall resonant frequencies will be unchanged by
the cavity per se: e.g., the sum on the right-hand side of equation (54) can be neglected. The
circumstances under which this is not true will be considered in section 5.2.2. Making this
assumption for now, from equation (54) one computes, as wE-- wS,
From equations (55), (53) and (45), the cavity pressure may be computed as
p’ =
PO@,
F, L,,, $ @!,/A,)[-wf + co;‘] . (56)*
2(,iV c/%f F
If w: is much smaller than the lowest non-zero acoustical resonant frequency, then only n = 0
need be considered in equation (56). Hence equation (56) simplifies to
(57)*
dA = (2/77)‘, WWj$:dA
(l/A,) j- ICI, = (1/212.
Thus p= z 3p,: i.e., typically the soundpressure in the cavity will be larger than the external
sound pressure, if o, < mA.For w, > w*, one must computepC from equation (56) and typically
pc < pE.
In most cases where there is a significant change in a structural resonant frequency due to
the cavity, only the n = 0 cavity mode will be important [46, 16, 181. Hence equation (54)
may be simplified to
M,[& + 21, u, 4, + oz q.] = -(p. ci A fl V) L& qs + of eioHr. (58)
The effect of the cavity (within our approximations) is to modify the structural stiffness
term: i.e., the total term is now
M,w: + (P&-,~AJV)L%.
532 E. H. DOWELL, G. F. GORMAN, III AND D. A. SMITH
Equation (60) generalizes an earlier result by Dowel1 and Voss [5] for sfructural-cavity
coupling. By using equation (60) the results in section 5.2.1, e.g., equation (56), can now be
generalized to include structural-cavity coupling. By examining the ratio of the two terms in
equation (60), one can assess when cavity effects on structural resonances are important:
Upon using our previous estimates for the integrals, the ratio becomes
(pOcf A,/m VW<)(256/n4).
Clearly the lowest frequency structural mode will be most affected by the cavity. In extreme
cases more than one structural mode may be significantly affected and the cavity-induced
coupling between structural modes must be considered [6, 181.
Another less frequent case where the structural frequency may be changed due to acoustic
action is when a structural mode frequency and acoustic mode frequency are close together.
Clearly this is an undesirable situation, since large sound levels would be anticipated at this
double resonance. This special case is considered in section 5.2.5 below.
;L&=O. (61)
Q,“Jk L2“In
pn.=--.-...- M,A
c
PO&$& m M,,,[--w~~+ 21;, o, iot + o$] i cm M,,,[-of* + 25, w, o$ i + w,$
. (65)
For the special case where o: = w, for some m, then the sums can be well approximated by
a single term? and the above relationship simplifies. Even when o: # o, for any m, there may
t Apart from the pathological case where L,, z O!
ACOUSTOELASTICITY 533
be a predominant structural mode in which case the sum could be approximated by a single
term. First, however, we digress to consider another important limiting case.
[f F, andpE are constants orer AF, then a very interesting result obtains: namely,
the ratio of sums then simplifies and using equations (45) and (65) gives
(66)
SiO”A _=2$i0”AP = 2,
b
Clearly, ifp’),,,, =pE then w = O! To see this in detail recall that
P=/POc”,= P, F,IM,A.
PC 1 jpE$,dAj$,dA $,dA ’
Ir I
-2 c”
and w # 0.
POG o”A =+LW:,
F ” MIA. . .I MrrJ...l
Note that p’ is constant over A,. Only if there is a single dominant structural mode will this
simplify further. This concludes the digression and we now return to the discussion of equation
(65) in the case when there is a single dominant structural mode.
For a single dominant resonant structural mode, m, equation (65) becomes
(67)
and
IfpE = WI,,.+ thenp’ =pE on A, and w = 0 as before. Note that, within reason, rl/, may
be selected as an appropriate linear combination of natural modes. Hence a dominant struc-
tural mode assumption is quite useful.
The central conclusion to be drawn from the above is that, in simple terms, one may think
of the cavity as acting as a vibration absorber for the structural wall when the external
exciting frequency is equal to the cavity resonance frequency. Hence, usually the internal
cavity pressure will be equal to the external pressure at most and indeed, if w, $ w: for all m,
and/orpE varies significantly over the flexible wall, thenp’ < pE. The exceptional case is when
pE is constant over AF but F, is not. Then p’ > pE (see equation (68)).
534 E. H. DOWELL, G. F. GORMAN, 111AND D. A. SMITH
or
L2nm
:2[,ip”=p,,cAA, -pn
F cm M,[-a<* + 21, w, 0: i + of] I MR
Pm”
+
c
m M,[-~0: * + 2i&o,
L”fn
iwt + wf]
(70)
p, = -
0: L”, 1
(72)
POC$AFL&IM,A
I[ 1 + 0;;?4[, i: VM, M .“lpo 4, A; Lk,,) I ’
Hence, the effect of cavity damping is to decrease the internal pressure field. However, the
numerical effect is typically small for the lower acoustical modes.
If one starts again from equation (70) an alternative but more elaborate calculation can be
carried through for ]--w~’ + oi 1% 21;, w, ~0:.
5.2.5. Exciting frequency = structural wail and cavity resonant frequencies in very close
proximity
With, for simplicity, any absorbing wall ignored, consider now a single dominant structural
wall mode and also a single dominant cavity mode whose natural frequencies are very near
each other. From equation (52) the equations of motion are
P
-L-
iRE2F
. m n
(76)
( &/Lnm AF) = { [-QE,’ + 2[, Q,” i + I] [-Q”,’ + f2A,‘]- j.QE,‘]’
where
Note that
i h (po VImA,) [( AFI I’) (~o/aJ12.
From expression (76) the coupled structural-acoustic resonant frequencies may be deter-
mined and the associated cavity pressure. For the special case of Q”, = 1 and /1 G 1 (typically
i. < I ), particularly simple results may be obtained. The resonant frequencies are then
Itjc/(~E/Ln,,,
A,)1 5 (j.“*iX,,,, Fn. (78)*
The left-hand side of expression (78) is approximately p’/p” for the lower acoustic and
structural modes: i.e., L,, - O( 1). Hence for typical [,,, and 1, not too small, p’/p” > 1 when a
cavity and structural wall mode coincide in natural frequencies and L,, _ O( 1).
5.2.7. Summary of key relations from simplijed models to determine caoity pressure
When the exciting frequency is equal to the structural wall resonant frequency, then for
wt: + w,, equation (56)* or (57)* applies, modified by equation (60)* as necessary.
When the exciting frequency is equal to the cavity resonant frequency, then for ut ---+0::
(single dominant structural mode) equation (68)* applies.
When the exciting frequency is equal to the structural wall and cavity resonant frequencies
in close proximity, then for WE --f w, = wt, equation (78)* applies.
When the external excitation frequency is well separated from all structural or cavity
resonant frequencies, there may not be a single dominant structural and/or cavity mode.
However, some simplification may still be possible by neglecting the interaction between
structural wall and cavity: i.e., one may first determine the structural wall motion and then
use that result to determine the internal cavity acoustic pressure. In this approximation the
effect of the cavity sound pressure on the structural wall motion is omitted.
It is also worth emphasizing that for off-resonant exciting frequencies, although the
numerics may be more elaborate due to the necessity of accounting for multiple modes, the
basic theoretical model may be more accurate because the uncertainty with respect to
structural and/or acoustic damping values will be less important.
536 E. H. DOWELL, G. F. GORMAN, III AND D. A. SMITH
I
IO0
Frequency (Hz)
panel. The first two resonances correspond to the first and third panel modes,? and thus
indicate that the panel is driving the cavity at these frequencies. The resonance at 5 I8 Hz is
the fundamental cavity depth mode. Theoretically, if the external pressure were uniform over
the flexible panel, it should be motionless, and the pressure level difference between the
external and internal measurements should be zero when the external frequency equals the
cavity resonant frequency. This is nearly the case, as is shown in Figure 8.
In Figure 9, a comparison between theory and experiment is made. The ratio of panel
frequency (modified by coupling with the cavity) to in LUCUOpanel frequency is plotted against
panel length to cavity depth ratio, a/d. The in zwzuo panel frequencies were computed from
Warburton’s theory [24], and the panel frequencies’ variation with cavity depth were com-
puted from Dowel1 and Voss’ theory [18], which is an earlier version of the present analysis.
There is excellent agreement between theory and experiment at the large cavity depths, with
some variation from theory occurring at shallow cavity depths. Again it should be emphasized
that this interesting change in panel frequency occurs only for flexible panels and stiff
(shallow) cavities.
a/a
Figure 9. Cavity effect on panel natural frequencies. --, One term theory; ------, two term theory;
9. i: , experiment.
Three types of damping will be referred to in this discussion: constant damping, frequency
damping, and experimental damping. Constant damping is the value measured for a 12 in
cavity depth, on the assumption that there is no variation of panel modal damping ratio with
cavity depth. Frequency damping allows for variation of damping ratio with frequency and
employs the data measured at a 12 in cavity depth for various panel resonances. Thus, the
only effect changing this type of damping is the variation of panel modal frequency with
cavity depth (see Figure 9 and reference [16]). Experimental damping is that measured for
the exact conditions under study.
Figure 10 shows plots of the variation of cavity pressure with cavity depth for the eee
different theoretical damping models: i.e., constant damping, frequency damping, and
experimental damping. The exciting external frequency is equal to the fundamental panel
t The second panel mode is antisymmetric and hence is not excited by the symmetrical spatial distribution
of the external sound field.
538 E. H. DOWELL, G. F. GORMAN, III AND D. A. SMITH
IO i
t
0
I I
5.03
1
IO.16
I
15.24 20.36
I
25.40
I
30.4
Cavity depth (cm)
Figure 10. Cavity pressure us. cavity depth. ------, Frequency damping; ----, constant damping; -,
experimental damping; C, experiment.
resonant frequency.? Recall that the damping ratios used in these calculations are those of
the panel and not of the cavity; the latter were neglected. Even though cavity damping has
not been considered, there is excellent agreement between experiment and the theoretical
model when experimental damping is used.
Similar results have been obtained for random external pressure excitation [ 171. Pretlove
[6] has made measurements of panel natural frequency variation with cavity depth; Guy and
Bhattacharya [ 191have measured cavity pressures and panel natural frequencies. Generally
good agreement with theory has also been shown in references [6, 17 and 191.
7. CONCLUDING REMARKS
A comprehensive theoretical model has been developed for interior sound fields which are
created by flexible wall motion resulting from exterior sound fields. Included in the model are
the mass, stiffness and damping characteristics of the flexible wall and of the acoustic cavity.
Full coupling between the wall and cavity is permitted although detailed analysis, numerical
results and experiment suggest that it is the exceptional case when the structural wall dynamic
characteristics are significantly modified by the cavity.
Based upon the general theory, an efficient computational method is proposed and used
to determine acoustic natural frequencies of multiply connected cavities. Simplified formulae
are developed for interior sound levels in terms of in z’c~cuostructural wall and (rigid wall)
acoustic cavity natural modes.
Comparisons of theory with experiment show generally good agreement. The principal
uncertainty remains the structural and/or cavity damping mechanisms. For external sound
excitation and the lower acoustic modes, cavity damping is demonstrated to be generally
unimportant; however, it may be of importance for interior sound sources or the higher
acoustic modes. The results of Wolf, Nefske and Howell [25] and Petyt, Lea and Koopmann
[2@using finite element techniques and Howlett and Morales [27] using modal analysis
similar to that of Cockburn and Jolly [28] also suggest that effective analytical models are
available.
t As cavity depth was changed, the external frequency was continually retuned to match the (coupled)
panel frequency.
ACOUSTOELASTICITY 539
REFERENCES
1. M. JUNGER and D. FEIT 1972 Sound, Structures and Their Interaction. M.I.T. Press.
2. E. H. DOWELL 1974 Aeroelusticity of Plates and Shells. Leyden: Noordhoff International
Publishing.
3. E. H. DOWELL 1976 Princeton University AMS Report No. 1280. Acoustoelasticity.
4. R. H. LYON 1963 Journal of the Acoustical Society of America 35, 1791-l 797. Noise reduction
of rectangular enclosures with one flexible wall.
5. E. H. DOWELL and H. M. Voss 1963 American Institute of Aeronautics and Astronautics Journal
1, 476-477. The effect of a cavity on panel vibrations.
6. A. J. PRETLOVE 1965 Journal of Sound and Vibration 2, 197-209. Free vibrations of a rectangular
panel backed by a closed rectangular cavity.
7. P. M. MORSE and K. U. INGARD 1968 Theoretical Acoustics. New York: McGraw-Hill Book
Company, Inc.
8. E. H. DOWELL 1968 Journal of the Acoustical Society of America 47, 238-252. Transmission of
noise from a turbulent boundary layer through a flexible plate into a closed cavity.
9. L. MEIROVITCH 1974 American Institute of Aeronautics and Astronautics Journal 12, 1337-l 342.
A new method of solution of the eigenvalue problem for gyroscopic systems.
10. G. M. L. GLADWELL and G. ZIMMERMANN 1966 Journal of Sound and Vibration 3, 233-241.
On energy and complementary energy formulations of acoustic and structural vibration
problems.
11. G. M. L. GLADWELL 1966 Journal of Sound and Vibration 4, 172-l 86. A variational formulation
of damped acoustostructural vibration problems.
12. G. M. L. GLADWELL and V. MASON 1971 Journal of Soundand Vibration 14,115-135. Variational
finite element calculation of the acoustic response of a rectangular plane.
13. A. CRAGGS 197 1 Journal of Sound and Vibration 15.509-528.The transient response of a coupled
plate-acoustic system using plate and acoustic finite elements.
14. A. CRAGGS 1973 JournalofSoundand Vibration 30,343-357. An acoustic finite element approach
for studying boundary flexibility and sound transmission between irregular enclosures.
15. E. H. DOWELL 1972 Journal of Applied Mechanics 39, 727-732. Free vibrations of an arbitrary
structure in terms of component modes.
16. G. F. GORMAN, III 1970Princeton University AMS Report No. 925. An experimental investigation
of sound transmission through a flexible panel into a closed cavity.
17. G. F. GORMAN, III 1971 Princeton University AMS Report No. 1009. Random excitation of a
panel-cavity system.
18. E. H. DOWELL and H. M. Voss 1965 American Institute of Aeronautics and Astronautics Journal
3, 2292-2304. Experimental and theoretical panel flutter studies in the Mach number range of
1.o to 5.0.
19. R. W. GUY and M. C. BHATTACHARYA 1973 Journal of Sound and Vibration 27, 207-223. The
transmission of sound through a cavity-backed finite plate.
20. D. A. SMITH 1976 Princeton University AMS Report No. 1284. An experimental study of acoustic
natural modes of interconnected cavities.
21. Y. K. LIN 1967 Probabilistic Theory of Structural Dynamics. New York: McGraw-Hill Book
Company, Inc.
22. L. L. BERANEK 1960 Noise Reduction. New York: McGraw-Hill Book Company, Inc.
23. L. MAE~TRELLO 1965 Journal of Sound and Vibration 2, 270-292. Measurement and analysis of
the response fields of turbulent boundary layer excited panels.
24. G. B. WARBURTON 1954 Proceedings of the Institution of Mechanical Engineers 168, 391-394.
The vibration of rectangular plates.
25. J. A. WOLF JR., D. J. NEFSKE and L. J. HOWELL 1976 Presented at the Automotive Engineering
Congress and Exposition, Detroit, Michigan, February 1976, SAE Paper 760184. Structural-
acoustic finite element analysis of the automobile passenger compartment.
26. M. PETYT, J. LEA and G. H. KOOPMANN 1976 Journal of Sound and Vibration 45, 495-502. A
finite element method for determining the acoustic modes of irregular shaped cavities.
27. J. T. HOWLETT~~~ D. A. MORALES 1976 NASA TM X-72838. Prediction of light aircraft interior
noise.
28. J. A. COCKBURN and A. C. JOLLY 1968 Air Force Flight Dynamics Laboratory Technical Report,
il FFDL-TR-68-2. Structural acoustic response, noise transmission losses and interior noise
levels of an aircraft fuselage excited by random pressure fields.
540 E. H. DOWELL, G. F. GORMAN, III AND D. A. SMITH
Here some useful background information is given on (1) the plate in LYZCUO modes, (2)
the cavity (rigid wall) modes and (3) the interesting change in wall mode shapes as the cavity
depth is reduced.
Normollzed length
Figure 11. Cavity effect on panel normal mode shapes (fundamental mode). 0 ---C --- $3, d= 2 in:
O-U-O, d= 12 in.
ACOUSTOELASTICITY 541
0.50 --d4
Normal~red lwgth
Figure 12. Cavity effect on panel normal mode shapes (third mode). I_--- _, d= 2 in; S---U, d= I2 in.
the fundamental mode shape evolves into that resembling the third mode and the third mode
into that resembling the first. Presumably at even smaller cavity depths (or for more flexible
wall plates), the first mode (which now resembles the third) evolves into the fifth (or next
volume displacing mode) and so forth. In the present experimental apparatus, only a partial
evolution occurs as the cavity depth is reduced from the maximum (12 in) to minimum (2 in)
cavity depth studied. In Figure I 1, the fundamental mode is shown for these two cavity
depths and in Figure 12 the third mode is shown. One can see clearly the tendency toward the
mode shape changes described above. For a more flexible wall plate and a consequent more
dramatic change in mode shape, the reader may consult reference [ 181, which includes a
comparison between theory and experiment.
p density
4k mth structural mode (critical) damping ratio
i” nth acoustic cavity mode (critical) damping ratio
w frequency
Q,,, GA see equation (76)
Subscripts
0 equilibrium
A absorbent wall; acoustic
E external
F flexible wall
m, n, r, s modal numbers
R rigid wall
Superscripts
A acoustic
a, 6, c cavity a, b, c, respectively
c acoustic cavity
- perturbation