See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/243716469

Structural Acoustics Tutorial—Part 2: Sound—Structure Interaction

Article in Acoustics Today · January 2007

DOI: 10.1121/1.2961152

CITATIONS READS
21 5,031

2 authors:

Stephen Hambric John B. Fahnline

Pennsylvania State University Pennsylvania State University
98 PUBLICATIONS 1,085 CITATIONS 72 PUBLICATIONS 1,192 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Stephen Hambric on 03 February 2015.

The user has requested enhancement of the downloaded file.

 STRUCTURAL ACOUSTICS TUTORIAL—PART 2:
SOUND—STRUCTURE INTERACTION
Stephen A. Hambric
and
John B. Fahnline
Applied Research Laboratory, The Pennsylvania State University
State College, Pennsylvania 16804

Introduction tural surface must be identical to the corresponding particle

his is the second of a two-part tutorial on structural velocity in the neighboring fluid. This simple equality allows

T acoustics written for Acoustics Today. The first

appeared in the October 2006 issue, and focused on
vibrations in structures. In that article, I explained
us to couple the equations that define structural and fluid
motion at the fluid–structure interface and solve for the total
sound–structure behavior. While the normal particle velocity
• the various waves that can propagate through structures, is identical in the structure and fluid, the in-plane, or tangent
and how bending waves are dispersive (their wavespeeds particle velocity is not. In fact, we allow a ‘slip condition’
increase with frequency); between the structure and fluid, so that a structure can slide
• the modes of vibration of finite structures; along a neighboring fluid without inducing any sound.
• mobility and impedance, and how they are simply sum- Of course, with any simple concept there are inevitably
mations of individual modal responses; several assumptions. Here are ours:
• structural damping; • homogeneity (the fluid properties are the same every-
• how the mobility of an infinite structure is the mean where),
mobility of a finite one; and • isotropy (the fluid properties are
• modeling vibrations with finite ele- the same no matter what direction the
ment (FE) analysis.
“In this article we will wave propagates), and
Some of the feedback I received • linearity (the fluid properties do
from readers of the first article pointed
explain: what structural not depend on the fluctuating pressure
out that it was more of a ‘cliffs notes’
summary of structural acoustics than a
vibrations do to neighboring amplitude or phase).
With these simplifying assumptions, it
tutorial, but found the summary quite is straightforward to couple the vibra-
useful nevertheless. After thinking about
acoustic fluids, and what tions of structures with those in
it, I suppose that is true, and I thought acoustic fluids. Incidentally, all of the
about renaming the second part of this
sound fields do to information in Part 1 of this article
article ‘A Structural Acoustics made the same assumptions, but for the
Cookbook,’ rather than a tutorial. For
neighboring structures.” structural materials!
continuity, though, I have retained the We will start by explaining what a
original title, but hope the two articles will be a useful short structure’s vibrations do to a fluid—they compress and
reference on the subject, where handy formulas and concepts expand it. The spatial pattern of structural vibrations and
can be easily found. their frequencies determines how much sound is radiated,
In this article, I have added a co-author—Dr. John and in what directions. It may be helpful to think of the
Fahnline—who specializes in analyzing sound–structure acoustic fluid as an elastic blob surrounding the structure,
interaction using boundary element (BE) modeling tech- being pushed and pulled over time by the motion of the
niques. John has already written one book1 on acoustic BE structural boundary (in the normal direction only, of
analysis (with Gary Koopmann), and is working on a second. course). When the fluid’s mass density is comparable to the
In this article we will explain: structure’s, the fluid not only absorbs sound, but also mass-
• what structural vibrations do to neighboring acoustic flu- loads the structure. We will explain how two important
ids, and structures—a circular baffled piston, and a flat rectangular
• what sound fields do to neighboring structures. flexible finite plate—radiate sound and are fluid-loaded by
These problems are complementary (and reciprocal), and we the impedance of the surrounding acoustic fluid.
will use analytic, numerical, and experimental data to Next, we will consider the complementary problem—
demonstrate their basic concepts. As with Part 1 of the arti- how acoustic waves induce vibration in a structure. The
cle, we will supply plenty of useful terms and equations. same physics are at work in this reciprocal problem, as we
The overarching concept of linear sound–structure inter- shall see later. We will conclude by discussing how to make
action is simple: the normal particle velocity in the structure measurements of sound–structure interaction, and how to
and fluid along the fluid–structure interaction boundary must model it using boundary elements.
be the same. This means that when a structure vibrates against A clarifying note: in this article we consider
a fluid, the component of the vibration normal to the struc- sound–structure interaction, not fluid–structure interaction

Structural Acoustics Tutorial 2 9

 (FSI). FSI usually refers to how moving fluids interact with
solid objects, such as the turbulent flow over a structure. To
learn more about FSI, we recommend the references by
Blevins2 and Naudascher and Rockwell.3

Piston vibrating against an acoustic fluid

Perhaps no other structural-acoustic system has been
studied more than a circular baffled piston vibrating against
an acoustic fluid (see, for example, Fahy,4 pages 58–60 and
118–121; Pierce,5 pages 220–225; and Junger and Feit,6 pages
95–100 and 105–109). We will describe how a piston interacts
Fig. 1. Low frequency vibration and acoustic fluid motion of a piston. Left – baf-
with a surrounding acoustic fluid, and how the acoustic fluid fled piston; Right – unbaffled piston.
affects the piston’s vibrations. This simple problem allows us
to introduce many structural-acoustic quantities of interest. vibrations are time harmonic (eiωt), and consider pressures in
Let us start by considering a very slowly oscillating pis- the far-field, or far away from the structure. The far-field
ton in a rigid baffle, as shown in Fig. 1. As the piston moves pressure radiated by a baffled circular piston as a function of
outward, fluid flows from the area of high pressure near the angle θ and distance r is:
center of the piston to its outer edge, which is nearly at ambi-
ent pressure. Conversely, as the piston moves inward, fluid , ( 1)
flows toward the center of the piston, where the pressure is
less than the ambient pressure. The level of sound radiation
is a function of frequency because as the speed of the oscilla- where ρ0 is the fluid density, a is the piston radius, vn is the
tion of the piston increases, there is less and less time for the piston velocity (assumed to be constant over the surface of
pressure to equalize, and eventually the surface vibrations the piston), and ko is the radial frequency ω divided by the
become much more efficient in compressing the adjacent acoustic sound speed co. We do not include the time-har-
fluid and causing sound radiation. Since the piston’s size monic dependence in this, or in any future equations.
controls how far the pressure pulses have to travel before Figure 2 shows the far-field pressure at two frequencies,
they can equalize with the surrounding fluid, it is important normalized to koa, where koa=π and 3π. For koa=π (a full
in determining when it begins to radiate sound efficiently. acoustic wavelength spanning the piston diameter), the pres-
Thus, we can conclude purely from physical arguments sure is in phase at all directivity angles (note that the direc-
that the main parameters which control the acoustic radiation tivity angle is taken from a vector pointing normal to the pis-
from a vibrating structure are its speed of vibration (or fre-
quency) and its size. It is common in acoustics to nondimen-
sionalize the problem in terms of the quantity ka, where k is the
acoustic wavenumber and a is the characteristic dimension of
the vibrating structure. For a piston source, the transition from
an inefficient, low frequency oscillation to an efficient, high fre-
quency oscillation occurs at approximately ka = 1, where a is
the radius of the piston. At this frequency, the acoustic wave-
length is of the same order as the size of the structure.
A source with zero average displacement is an even
more inefficient radiator of sound at low frequencies, like an
unbaffled piston, i.e., a loudspeaker without a cabinet. When
the diaphragm of the speaker moves, the fluid on one side of
the speaker is compressed while the fluid on the other side is
expanded (see Fig. 1). There is a natural flow of fluid from
high pressure to low pressure, and thus the fluid flows
around the edges of the speaker. The pressure equalization is
very nearly complete because the average displacement of the
surface is zero when both sides are taken into account, such
that there is very little residual compression of the acoustic
medium and sound radiation. As the frequency of vibration
increases, the sound radiation also increases because there is
not as much time for the fluid to flow around the edges of the
speaker before it is compressed. At approximately ka = 1, a
source with zero average displacement becomes as efficient a
radiator of sound as a baffled piston.
Let us consider the pressure field radiated by an oscillat- Fig. 2. Pressure distribution in the far-field of a radiating baffled piston. Top –
ing piston at discrete frequencies. We assume the piston’s ka=π; Bottom – ka=3π.

10 Acoustics Today, April 2007

ton surface). However, at higher frequencies, like koa=3π, the
pressure alternates its phase at varying angles. The directivi-
ty of the sound radiated by piston sources is usually plotted
as a pressure magnitude to remove the effects of alternating
phase.
It is hard to generalize how much sound a source makes
by considering pressure at a specific location. Therefore, we
usually measure or compute the total radiated sound power
instead. The total radiated sound power is computed by inte-
grating the active acoustic intensity over space.
The active, or propagating acoustic intensity in a fluctu-
ating acoustic pressure field is:
Fig. 3. Sound intensity distribution in the far-field of a radiating baffled piston,
(2) ka=3π.

where v씮* is the conjugate of the acoustic fluctuating particle Figure 3 shows the far-field intensity for ka=3π on a dB
velocity. Notice that both the particle velocity and intensity scale. Notice how all the peak locations are positive, since
are vector quantities, and point in specific directions. When intensity squares the pressure magnitude. The highest pres-
in the far-field, we consider only the radially propagating sure is normal to the piston, with lower amplitude side lobes
component of intensity. Also when in the far-field, the pres- at various angles. As ka increases, more side lobes will appear.
sure and particle velocity are in phase with each other (this is The total radiated power is computed by integrating the
not the case close to the vibrating surface, or in the near- intensity (which is just the localized power/area) over a far-
field) and v=p/ρoco. So, in the far-field, the radial component field half-spherical surface surrounding the piston. Spherical
of intensity simplifies to: surfaces are used to make the integration simple; the total
sound power may be integrated over any shaped surface,
(3) though.
The radiated sound power is related directly to the radi-
ation resistance of the fluid, which acts over the surface of the

ACOUSTIC MEASUREMENT PRODUCTS

Reliable Acoustic
Measurement Products
A Complete Line of Microphones, Preamplifiers and Accessories
I Prepolarized ICP ™ and Externally Polarized (200 V)
I Type 1 Condenser and Cost-effective Array Microphones
I Linear Measurements to 192 dB
© 2007 PCB Group, Inc. ICP and PCB are registered trademarks of PCB Group, Inc.

I Low-profile Surface Microphones

I Sizes from 1/8” to 1” Diameter
I Full Complement of Low-Noise Preamplifiers
I Your Front-End Solution for Sensors, Power Supplies,
Cables, Calibration, and Accessories

Toll Free in USA 888-684-0013 • 24-hour SensorLineSM 716-684-0001

E-mail [email protected] • www.pcb.com
ISO 9001:2000 Certified • A2LA Accredited to ISO 17025 • AS9100:2004 Certified

Structural Acoustics Tutorial 2 11

 piston. In fact, power can be calculated knowing a structure’s Popular approximations for the low-frequency resistance
radiation resistance and spatially- and time-averaged normal and reactance are:
velocity:

. (4) (7)

The radiation resistance and reactance are the real and

imaginary parts of the fluid loading on the structure’s sur-
face. Imagine once again the fluid as an encompassing elastic
blob surrounding the structure. As the structure pushes At high frequencies, the resistance asymptotes to a con-
against this blob, it encounters an impedance, which resists stant value, while the reactance decreases inversely propor-
the structure’s motion. The impedance, Zo, is complex, and tionally to frequency:
equal to Ro + iXo, where Xo is the reactance. The resistance
and reactance are fluid-loading properties of the structure, . (8)
and do not depend on how the structure vibrates. Equations
for the resistance and reactance (shown without derivation)
of a baffled circular piston are: Now, notice how Ro and Xo in the example are plotted in
absolute terms, but also plotted normalized against ρocoA (see
the axis on right side of the plot) and against dimensionless
, and (5) frequency ka (see the axis on the bottom of the plot). Recall
that you can use ka to visualize how many acoustic wave-
lengths fit over the characteristic dimension a. In our plot, the
. (6) maximum ka is π, which corresponds to a half wavelength
over the piston radius, or a full wavelength across the piston
diameter. Ro /ρocoA converges to a value of 1 at high frequen-
Note once again that no piston structural properties appear in cies. In fact, Ro /ρocoA is called the normalized radiation resist-
the above equations; they are solely determined by the geome- ance, or more commonly, the radiation efficiency:
try of the piston and the acoustic properties of the fluid.
We plot the resistance and reactance for a 16.6 mm
radius piston in water at the top of Fig. 4. At low frequencies, . (9)
the resistance (in red) looks like a parabola, while the reac-
tance (in green) looks like a line with a constant slope.
We will learn more about radiation efficiency when we dis-
cuss how sound is radiated by plate modes.
At the bottom of Fig. 4, examples of how the fluid load-
ing varies over the surface of a piston are shown at low, mid,
and high frequencies. Below a ka of π/2, the fluid loading is
primarily reactive, or mass-like, weighing down the piston.
Above ka of π/2, the fluid loading becomes more resistive,
absorbing energy in the form of sound from the vibrating
structure.
Now, let us suppose that the piston is the mass element
of a simple harmonic oscillator, where the mass (m) rests on
a grounded spring (k) and dashpot (b). We now consider the
effect of the complex fluid loading on the piston resonance,
where the piston mobility in-vacuo is:

. (10)

The fluid loading (resistance Ro and reactance Xo) may be

added to the mobility equation to produce:

(11)

Fig. 4. Radiation impedance of baffled circular piston. Top–resistance and reac-

For a 1 gram piston of 16.6 mm radius, with spring con-
tance as a function of frequency (and ka); Bottom–spatial variability of resistance stant k=1x105 N/m, and a damping constant b of 1, we com-
and reactance at three discrete frequencies. pute the drive point mobility v/F in air (ignoring fluid load-

12 Acoustics Today, April 2007

 The radiated sound power curve in water illustrates the
basic principle of loudspeaker design: adjust the piston prop-
erties to set the fundamental resonance as low in frequency as
possible, so that the piston response is controlled by the mass
term in Eq. 11. Recall that the radiated sound power is the
product of the radiation resistance Ro and the square of the
piston normal velocity. Since Ro at low ka is proportional to
(ka)2, and the piston mobility above resonance is proportional
to 1/ω2, the frequency dependencies cancel, leaving a nearly
frequency independent radiated sound power/F2 response.
In our example in Fig. 5, the in-air radiated sound power
transfer function is not flat above the piston resonance fre-
quency because the radiation resistance is above the frequen-
cy range where it is proportional to (ka)2. To resolve this, a
speaker designer would simply reduce the radius of the pis-
ton, shifting the radiation resistance curve further out in fre-
quency. This solution comes with a cost though—the radia-
tion resistance amplitude reduces with surface area, which
will in turn reduce the radiated sound power.

Structural waves vibrating against an acoustic fluid

How well do structural waves (rather than rigid oscilla-
tors) radiate sound? Since only structural motion normal to
an object’s surface induces an equal motion in a neighboring
fluid, we consider transversely vibrating, or flexural waves
(we acknowledge, however, that longitudinal waves deform a
structure transversely due to an elastic material’s Poisson
effect, but do not focus on the sound radiated by longitudi-
nal waves here).
How well flexural waves in a structure radiate sound
depends on whether the waves, which essentially act as a
source against the fluid, are subsonic (slower than the
wavespeed in the fluid) or supersonic (faster than the
wavespeed in the fluid). Supersonic waves radiate sound, and
subsonic waves do not.
Many structural acousticians like to consider the sound
radiated by structural waves in wavenumber space, and
examine wavetypes on frequency-wavenumber plots.
Consider the traveling flexural waves in an infinite plate
shown in Fig. 6. The flexural and acoustic wavenumbers
(computed by dividing radial frequency by the flexural and
acoustic wavespeeds) are plotted against frequency in the top
of the figure. Since the acoustic waves are non-dispersive, the
Fig. 5. Mobility magnitudes (top), radiation resistances (middle), and radiated
sound power transfer functions (bottom) of a 16.6 mm radius circular baffled pis-
wavenumber curve has constant slope. The flexural waves,
ton in air (black) and in water (blue). however, are dispersive, causing a varying slope in the
wavenumber curve.
ing, using Eq. 10), and in water (including fluid loading using At low frequencies, the structural wavenumbers are
Eq. 11). Next, we multiply Ro by the square of mobility to com- higher than those in the acoustic fluid, corresponding to sub-
pute radiated sound power. Plots of the mobility magnitude, sonic structural waves. These flexural waves radiate no
radiation resistance (in air and in water), and the radiated sound at all (this is only true for infinite plates—we will dis-
sound power for a unit force input are shown in Fig. 5. cuss the sound radiated by finite plates soon). This is because
The effect of mass loading on the piston in water is pro- the particle velocity in the fluid normal to the structure’s sur-
nounced, shifting the piston resonance frequency down- face must match that of the structure. At low frequencies,
ward, and the overall mobility amplitude downward. The acoustic waves are faster than structural ones, so their wave-
radiation resistance (and reactance) of water is much higher lengths are longer. This means the structure simply cannot
than that of air (be sure to note the multiple scales used on induce a propagating wave in the fluid.
the resistance comparison plot). Therefore, radiated sound The frequency at which the flexural and acoustic waves
power is quite different in air and in water, and the piston have the same wavenumber (and wavespeed, and wave-
resonance peaks occur at different frequencies. length) is called the coincidence frequency, and the flexural

Structural Acoustics Tutorial 2 13

 elastic moduli (Young’s Modulus E and Shear Modulus G)
speeds up flexural waves, and lowers the plate’s coincidence
frequency. Increasing a plate’s density increases its mass,
slowing down flexural waves, and raises the plate’s coinci-
dence frequency. Increasing thickness increases both stiff-
ness and mass, but increases the stiffness at a greater rate, so
thickening a plate will lower its coincidence frequency.
Therefore, stiffening a plate lowers its coincidence fre-
quency, allowing it to radiate sound at lower frequencies.
Conversely, mass-loading a plate raises its coincidence fre-
quency, so that the plate does not radiate sound at low fre-
quencies. It would seem that the answer to most noise con-
trol problems would be to simply add mass to a plate while
reducing its stiffness! While this is a good way of reducing
sound radiation, we have never had any of our sponsors
accept it. As most of us have experienced, nearly all new
structures are lightweight and stiff, like carbon-fiber com-
posites reinforced with ribbing. These sorts of structures typ-
ically have low coincidence frequencies, and therefore radi-
ate sound very well. Also, as we will see later, lightweight stiff
structures are very easy to excite by sound waves.
Since most practical structures are finite, we will now
explain how well the mode shapes of a structure radiate (see
Part 1 of this article for a discussion of structural reso-
nances). The classic example studied by early structural
acousticians is our old friend—the simply supported rectan-
gular plate.
Before we proceed with modal sound radiation, remem-
Fig. 6. Bending and acoustic wavenumber-frequency plot, with trace diagrams of ber a key concept: the standing waves in a mode shape are
bending and acoustic waves at and above coincidence. Below coincidence, an infi- comprised of multiple left and right (and forward and back)
nite bending wave radiates no sound!
traveling waves, which propagate at the structure’s wave
speed. It is these traveling waves that radiate, or do not radi-
waves now radiate sound, as shown in the image in the middle ate sound. If our plate were infinite, such that there were no
of Fig. 6. They do so in the plane of the plate, or grazing the reflections from any of the plate boundaries, subsonic flexur-
plate. At frequencies above coincidence, the flexural waves al waves would radiate no sound. However, the discontinu-
continue to speed up—eventually becoming pure shear waves ities at the boundaries ‘scatter’ the energy in subsonic flexur-
at very high frequencies (see Part 1 of this article). At these al waves into many wavenumbers, some of them supersonic,
high frequencies, the sound radiated by the flexural waves so that a finite plate radiates sound below its coincidence fre-
propagates in a preferred direction, which is computed by quency. The amount of sound radiation depends on the radi-
trace matching the flexural wave to the shorter acoustic wave, ation efficiency of each of the plate’s modes.
as shown in the image on the bottom right of Fig. 6. Maidanik7 and Wallace8 computed how much sound a
The angle of dominant radiation (taken from a vector finite rectangular simply supported plate’s modes radiate. In
normal to the structure’s surface) is computed as: particular, Wallace provides formulas for the frequency-

, (12)

where cB is the flexural wavespeed. You can compute a plate’s

coincidence frequency by setting the flexural and acoustic
wavespeeds (or wavenumbers) equal to each other. For
bending waves in thin plates at coincidence:

, so (13)

. (14)

How does coincidence frequency vary with plate parame- Fig. 7. Effects of stiffening and mass-loading on plate coincidence frequencies.
ters? To find out, let us examine Fig. 7. Increasing a plate’s Stiffening a plate reduces the coincidence frequency, and adding mass increases it.

14 Acoustics Today, April 2007

dependent far-field pressure and acoustic intensity fields radi-
ated by a given plate mode, along with an integral for comput-
ing the sound power radiation efficiency for each mode. We
show the far-field intensity below, near, and above the plate
coincidence frequency for the first three low-order modes of a
square 1m x 1m plate in Fig. 8. Recall from the first article that
we define mode orders as (m, n) pairs, where m and n corre-
spond to the number of antinodes (regions of maximum defor-
mation) in the plate’s x and y directions, respectively.
In the figure, the mode shapes of the plate are shown,
along with the corresponding far-field intensity patterns. At
low frequencies, the fundamental (1,1) mode radiates sound
omnidirectionally, like a baffled circular piston. Also at low
frequencies, the (1,2) mode radiates sound like a dipole, and
the (2,2) mode radiates like a quadrupole.
The figure also shows radiation efficiencies as a function
of acoustic wavenumber (frequency/acoustic wavespeed),
with a line shown to indicate the frequencies of the directiv-
ity plots. Notice how they resemble the normalized radiation
resistance of the baffled circular piston (Fig. 4). Below coin-
cidence, the efficiencies increase rapidly with increasing fre-
quency. The efficiencies peak at coincidence (exceeding 1,
showing that they are not true efficiencies!), and then asymp-
tote to a value of one above coincidence.
Above coincidence, the far-field sound directivity
changes, with ‘lobes’ of sound radiated from the structure at
critical angles. These critical angles may be computed using
the trace matching procedure described above for the infinite
plate. The critical angles exist for all plate modes except the
fundamental (1,1) mode, which radiates sound normal to the
plate at all frequencies, but with a spatial ‘beamwidth’ that
narrows with increasing frequency.
Let us re-examine the radiation efficiencies below coin-
cidence for the different mode orders. We see that the (1,1)
mode radiates sound most efficiently, followed by the (1,2)
mode, and then the (2,2) mode, which radiates sound least
efficiently. This trend shows an important result: modes with
odd m and n orders radiate sound much better than those
with mixed orders (odd-even or even-odd), which radiate
better than those with purely even orders (even-even).
Using the analytical radiation efficiencies, we can com-
pute how much sound a rectangular plate radiates when
driven by a point force. To do so, we combine the mobility
formula from Part 1 of this article with radiation efficiencies
computed using the formulas in Wallace.8 Ignoring any fluid
loading effects on the structure (which we know will mass-
load and radiation damp the structure from our exercises
with the baffled circular piston), we show the mobility, radi-
ation efficiency, and radiated sound power for a unit force
drive in Fig. 9. Fig. 8. Far-field sound intensity in air for various low-order simply supported 1 m
square flat plate modes. Top–below coincidence, Middle–near coincidence,
Starting with the mobility (top of the figure), we see how Bottom–above coincidence.
the contributions from the individual modes compare to the
total mobility. The mobility is dominated by resonant mode radiates sound very much like a baffled piston, and the
response at the resonance frequencies, and a mix of non-res- radiated sound power transfer function (power normalized
onant responses away from resonance. by the square of the input force) is dominated by the sound
Next, examining the radiation efficiencies of the modes radiated by that mode. Peaks in the sound power occur for
shows the same trends we observed in Fig. 8—that odd-odd the other modes, primarily the other odd-odd (1,3) mode,
modes radiate sound very well, but modes with mixed and but the non-resonant sound radiated by the (1,1) mode is
even orders radiate sound poorly. The fundamental (1,1) nearly flat with increasing frequency.

Structural Acoustics Tutorial 2 15

 This example brings up one of the key points of this arti-
cle—non-resonant sound radiation can be higher than resonant
sound radiation. The sound radiated by the (2,2) mode at its res-
onance frequency just below 100 Hz is a good example—it is
lower than the non-resonant sound radiated by the (1,1) mode!
When we coupled the radiation resistance of a baffled
piston to a simple harmonic oscillator (where the piston
head was the mass), we added it to the resistance of the oscil-
lator. For finite structures, most analysts use mechanical loss
factors to model structural damping. We can compute a radi-
ation loss factor which can be used in a structural analysis to
represent the radiation damping of an acoustic fluid. A use-
ful equation for the radiation loss factor of a uniform thick-
ness, homogenous plate or shell structure (flat or curved) is

.
(15a)

The equation shows that the radiation loss damping is

directly proportional to fluid density, meaning that the heav-
ier the fluid, the more sound power is radiated, and the more
heavily damped the structure. An example would be a vibrat-
ing bell dunked in water. While in the air, the bell’s vibrations
would ‘ring’ for a long time. When immersed in water, the
vibrations decay quickly. An extension of the above equation,
well known in the Statistical Energy Analysis (SEA) commu-
nity, relates the radiation loss factor to the radiation resistance:

, (15b)

where M is the mass of the plate or shell.

Table 1 provides a useful list of the various sound radia-
tion quantities and how they inter-relate.

The complementary problem—Structural vibrations

induced by acoustic pressure waves
Whereas the sound radiated by vibrating objects is often
just an annoyance, the vibrations induced in structures by
impinging acoustic waves can be so high that the structures Fig. 9. Mobility magnitude (top), radiation efficiency (middle), and radiated sound
power transfer function (bottom) of a simply supported 1m square 5 mm thick flat
Table 1. List of sound power radiation quantities and their interrelationships. steel plate in water driven at its quarter point (x=0.25 m, y=0.25 m).

16 Acoustics Today, April 2007

crack and fail. This is clearly a more serious problem, and has angle of incidence and the fluid’s characteristic impedance:
been studied intently by those in the aerospace and nuclear ,
communities, particularly in the 1960’s and 1970’s. (17)
Try holding your hand or fingers lightly on any struc-
tural surface, like a window when there is a lot of noise out- and the infinite plate’s impedance is
side. You will feel the window vibrating. The physics that
explain this phenomenon are the same as those that explain (18)
,
how a vibrating structure radiates sound. As we will see,
there are several quantities in sound-induced vibration that
are complementary to those in vibration-induced sound. where η is the structural loss factor and h is the plate thickness.
Let us start by considering the acoustic field adjacent to Examining the equation, we can see that the plate will
a flat infinite plate that is struck by an incoming acoustic vibrate most when its structural impedance is minimized.
wave. There are many pressure waves next to the structural This will occur when the stiffness and mass terms in the
surface—the incident wave, a reflected wave, and a wave re- plate’s impedance cancel each other, or when D(ko sinφincident)4 =
radiated by the structure, which has been forced into vibra- ρhω2 Let us find this frequency. To do so, we replace ko with
tion by the incident and reflected waves. The sum of the inci- ω/co, and we find that:
dent and reflected waves forms a ‘blocked’ pressure on the
. (19)
surface, and if the surface is rigid, the blocked pressure field
is the total pressure. If the structure is flexible and vibrates, it
radiates a third pressure wave, which sums with the blocked
pair to form the overall pressure field. ωco looks similar to a plate’s critical frequency, and to its crit-
Figure 10 shows the incident and reflected waves for a 30 ical radiation angle for frequencies above coincidence. In
degree angle of incidence (the angle is taken from the direc- fact, when φincident is 90 degrees (the acoustic waves propagate
tion normal to the plate), along with the total blocked pres- in the plane of the plate, or ‘graze’ the plate), the coincidence
sure field acting on a rigid surface. Notice how the two waves frequency is the critical frequency. The critical frequency, in
combine to form a standing wave pattern in the direction fact, is sometimes called the lowest coincidence frequency.
normal to the surface. The standing wave pattern propagates This means that there is no single frequency where the plate
in the direction parallel to the surface at a speed cosin(φ), vibrates most—there are several of them, each of which cor-
where φ is the angle of incidence. respond to a different angle of incidence.
Infinite plate theory (discussed in Part 1 of this article) Now let us consider how finite structures respond to
can be used to estimate the plate vibration caused by the incident acoustic waves. However, we will do so not for
blocked pressure field: waves that are incident from specific angles, but for groups
of waves that are statistically incident from all angles, more
, (16) commonly known as a diffuse acoustic field. Smith9 consid-
ered how a simple harmonic oscillator responds to diffuse
acoustic fields in 1962, where the mean square velocity
where zfluid and zplate are the impedance (pressure/velocity in response of the oscillator is
this case) of the acoustic space and the plate.
The fluid impedance depends on the pressure wave’s ,
(20)

where M is the oscillator mass, Rradiation and Rstructure are the

radiation and structural resistances, and Gpp is the autospec-
trum of the incident pressure.
The randomly excited simple harmonic oscillator prob-
lem is one of the origins of Statistical Energy Analysis (SEA),
a popular structural-acoustic analysis technique. For multi-
resonant structures, like plates, the plate energy is computed
rather than velocity:

(21)

Here, the radiation and structural resistances are averaged

over a frequency band spanned by ωmin and ωmax. Also, we
Fig. 10. Blocked pressure field acting on a plate (right side of images) at 30 degree include the modal density nplate, which defines the number of
angle of incidence. ‘+’ and ‘–’ signs indicate phase variations in the waves. plate modes in the frequency band. As we can see, SEA is sta-

Structural Acoustics Tutorial 2 17

 tistical, averaging over bands of frequencies. SEA also com-
putes averages of response over space, since <E>plate is aver-
aged over the plate.
A full discussion of SEA is outside the scope of this tuto-
rial, but you can find more information on SEA in the review
paper by Burroughs10 and a more thorough treatment of how
structures respond to diffuse fields in the article by Shorter
and Langley.11
You can use Eq. 21 without using SEA, though. You just
need to know the incident pressure spectrum of the diffuse
acoustic field, and the radiation and structural resistance. You
also need an estimate of the plate’s modal density, which is

,
(22)
Fig. 11. Typical transmission loss plot for variable angles of incidence. Grazing inci-
dence refers to acoustic waves that are nearly in the plane of the plate.
Note that the plate’s modal density depends on frequency (as
do all modal densities). cy). At high frequencies, above the coincidence dip, plate
Examining the equation, we see that the higher the radi- stiffness is dominant, and the transmission loss increases
ation resistance, the higher the plate energy (and therefore with the 6th power of frequency, or 18 dB/octave. Figure 11
the higher the plate’s vibration). So, the better a plate radi- shows what most people already know from experience—it is
ates, the easier it is to excite with incident pressure fields! hard to keep low-frequency sounds from propagating
Now, what happens when there is also fluid on the other through barriers. Consider this the next time you close a
side of the plate? How much of the incident sound gets door to block out sound from a hallway or another room.
through the plate to the other side? This is the classic sound You stop hearing mid-high frequency sounds, but still hear
transmission loss problem, and may be solved easily for an ‘muffled’ low-frequency noise.
infinite plate, and not so easily for a finite one. To visualize the sound field incident on and transmitted
Fahy4 provides a derivation of the sound power by an infinite plate, Fig. 12 compares pressure and displace-
transmission coefficient through an infinite plate in his text- ment of a plate at two conditions: well below, and near coin-
book, and we repeat it here (assuming the fluids on both cidence. In the example, we have set the plate loss factor
sides of the plate are the same): equal to 0. Try setting loss factor to zero and computing the
transmission coefficient in Eq. 23 at coincidence (remember,
this is where the acoustic wavenumber in the plane of the
. (23) plate matches the free bending wavenumber in the plate, or
kosinφ = kb). You should compute a transmission coefficient
of 1, which is perfect sound transmission!
The strength, or depth of the coincidence dip depends
strongly on the plate’s loss factor η. Designers of noise barri-
The red, green, and blue terms in the equation represent the ers (windows, doors) try hard to minimize the depth and
damping, mass, and stiffness of the plate, respectively. The breadth of the coincidence dips. The most common
amount of sound transmitted depends on the fluid proper- approach for mitigating coincidence dips is using con-
ties, the structural properties, frequency, and the angle of the strained layer damping, or CLD (we learned about this in
incident pressure wave with respect to the plate. Part 1 of this article). Automotive glass in luxury vehicles,
Some typical transmission loss plots, computed as and glass in high-end office buildings usually have a thin
10log10(1/τ) are shown in Fig. 11. For acoustic waves not nor- layer of clear vinyl sandwiched between two panes of glass.
mally incident to the plate, sharp dips appear in the transmis- For zero, or normal angle of incidence (sound waves
sion loss. These dips correspond to sharp peaks in the trans- normal to the plate’s surface), the transmission coefficient is
mission coefficient (transmission loss is the inverse of the not indeterminant (you might think it would be, since there
transmission coefficient), and act as strong pass-bands of inci- are several terms in Eq. 23 that divide by sin(φ)). The trans-
dent sound. The dips are at the coincidence frequencies of the mission coefficient actually simplifies to:
plate. Recall that the coincidence frequencies depend not only
on the plate, but on the angle of incidence of the sound waves.
As the angle of incidence changes, the coincidence frequency , (24)
and the frequency of the transmission loss dip changes as well.
At low frequencies, the mass term in Eq. 23 determines
the transmission loss, which increases with the square of fre- which corresponds to the well-known ‘mass law.’ The mass
quency (6 dB/octave, or 6 dB for each doubling of frequen- law transmission loss is shown in green in Fig. 11, and

18 Acoustics Today, April 2007

increases with the square of frequency over all frequencies,
never showing a coincidence dip.
To improve transmission loss, many noise control engi-
neers use two panels, particularly in windows. Called ‘double
glazing,’ the panels are separated by an air (or other gas) gap.
Since transmission loss is additive, there is a substantial
improvement, much more so than simply increasing the
thickness of a single panel. However, most double panel sys-
tems have different panel thicknesses so that the coincident
dips of the two panels occur at different frequencies. While
this leads to two pass-bands for incident sound, it is general-
ly preferable to have two weak pass bands rather than one
very strong one.
In most practical situations, acoustic waves impinging
on a panel do not arrive from the same angle. Consider a
window in an office building or a hotel. Sound waves arrive
from all angles, with a random distribution of the angles over
time (imagine incident sound from passing airplanes and
automobiles, reflecting off of adjacent buildings). Therefore,
statistical methods have been used to estimate a random
angle of incidence transmission loss for low frequencies:

(25)

where TL0 is 10log10(1/τφ=0), or the mass-law normal inci-

dence transmission loss. Since transmission loss is maxi-
mized (transmission coefficient is minimized) at normal Fig. 12. Incident and transmitted sound fields around an infinite 25 mm thick steel
incidence, the random angle of incidence transmission loss is plate, 30 degree angle of incidence. Top–at 50% of coincidence, Bottom–at coinci-
a fraction of that at normal incidence. dence. Dark blue and red indicate high pressures, and green indicates low pressures.

G.R.A.S. Sound & Vibration cost-efficient systems for beam-forming and acoustic holography.

MICROPHONE ARRAY MODULES

Depending on the application and the nature of
the emitted sound field, the microphones are available
as pressure- or free-field types, all with a built-in
Transducer-Electronic-Data-Sheet (TEDS).

For multi-channel systems without TEDS options,

we offer uni-gain versions with pre-adjusted gain.

The array modules for mounting the microphones are

available in various spacing configurations depending
on frequency requirements.
The amount of microphones and modules in the array
is only limited by the channels of the analysis system.

As always, we are ready to act on request, if our standard

solutions do not fit in with your requirements.

G.R.A.S. Sound & Vibration A/S · Skovlytoften 33 · 2840 Holte · Denmark

Tel.: +45 4566 4046 · Fax: +45 4566 4047 · E-mail: [email protected] · www.gras.dk

Structural Acoustics Tutorial 2 19

 Be careful using the above expression at high frequen-
cies! It does not include the coincidence dip, or the high-fre-
quency stiffness effects. To consider these effects, and those
of finite panel boundaries, other techniques are used, like
SEA. In fact, one of the early SEA applications was for single
and double panel transmission loss calculations. Price and
Crocker’s famous papers12, 13 clearly show the coincidence
dips in transmission loss, and the importance of varying
panel thicknesses in double panel systems.

Measurements of sound—structure interaction

The sound radiated by vibrating structures, and conversely,
the vibrations induced in structures by sound fields, are com-
monly measured in two types of chambers: reverberation rooms,
and anechoic rooms. Interconnected rooms are used to measure
the transmission loss of barriers, like windows and doors.
Accelerometers mounted to structures can provide a
spatially averaged normal velocity. Also, non-contact veloci-
ty measurements are often made using laser vibrometry,
eliminating the mass-loading effects of the accelerometers on
the test structure, and reducing test times.
Measuring sound power is more difficult than measur-
ing vibration. In air, arrays of microphones may be used to
measure the spatial variability of pressure. In water,
hydrophones are used. Sound pressures are measured quite
differently in reverberant and anechoic rooms, though.
In a reverberant room, the acoustic modes of the air in
the room are excited by a vibrating structure. When the
acoustic cavity modal density is high, the room’s pressure
field becomes nearly diffuse, such that pressures measured at
just a few locations randomly spaced throughout the room
may be used to estimate the total sound power radiated by a
vibrating object. The sound power is computed using the
measured pressures, along with a few of the reverberation Fig. 13. Impact hammer measurements on cylindrical shell structure submerged in
ARL/Penn State Reverberant Water Tank. Top–shell submerged in tank with diver;
room’s parameters, such as the volume and the reverberation
bottom, shell suspended over tank.
time. The reverberation time is inversely proportional to the
room’s loss factor—the longer a pressure impulse takes to struck by an impact hammer at several points, and the pres-
decay, the lower the room’s loss factor. sures measured at five hydrophone locations are used to esti-
The accuracy of a reverberation chamber sound power mate the sound power transfer function (Prad/F2).
measurement depends on the room’s modal density and rever- The sound power transfer functions for several drive loca-
beration time (among other things). The sound power is actual- tions on an elbowed pipe (the same pipe described in Part 1 of
ly a statistical mean, which is bracketed by a standard deviation this article) are shown in Fig. 14, adapted from another
at each measurement frequency. In general, the higher a room’s NoiseCon article,17 this one from 2005. The measurements
modal density and the wider the frequency bandwidth, the were made in one-third octave bands, and show that the
smaller the standard deviation. Most reverberation room sound sound power transfer functions vary with drive location (just
power measurements are therefore made over wide frequency as mobility functions do). Annotations on the graph show
bands that contain several acoustic modes. One-third octave where various shell modes cut on (again, see Part 1 of this arti-
frequency bands are commonly used. cle to refresh your memory on what shell modes are).
There are standards14, 15 available to help an experimen- It is often useful to compare sound power transfer func-
talist quantify a reverberation chamber’s characteristics, and tions to that of an ideal dipole source:
conduct a sound power measurement. The standards, how-
ever, are specific to rooms filled with air. To make measure- , (26)
ments in a reverberant water tank, some modifications to the
approaches in the standards are necessary. Conlon describes where |F|2 represents the square of the r.m.s. force amplitude.
these modifications, and applies them to ARL/Penn State’s This would be the sound made by an oscillating point drive
reverberant water tank in his NoiseCon 2004 article.16 in space. At some frequencies, most notably those of struc-
Photographs of a metal pressure vessel being tested in tural resonance, the sound power transfer function for the
ARL/Penn State’s tank are shown in Fig. 13. The vessel is elbowed pipe is higher than that of a point dipole, showing

20 Acoustics Today, April 2007

 ed. However, whereas only a few pressure measurements are
required in reverberant rooms to compute radiated sound
power, many more pressure measurements are required to
do so in an anechoic room.
Total sound power must be integrated over a closed sur-
face around the sound source. Also, acoustic intensity must
be measured, sometimes in the near-field of the structure.
This means many intensity measurements are made over
several points around the structure. The points may be
defined over a hemisphere or a box-like shape. Standards are
available18 to guide your measurements, and acoustic intensi-
ty probes, like the one shown in Fig. 16, are widely available.

Fig. 14. Sound power radiated by Schedule 10 Steel 3 inch pipe with elbow, pipe
length/diameter~12.

that the structure amplifies the force drive at those frequen-

cies. At other frequencies, usually between structural reso-
nances, the structure attenuates the drive.
The sound power transfer functions can also be com-
bined with a surface-averaged mobility measurement to
compute radiation efficiency:

(27)
.

Pressure, intensity, and sound power are measured very dif-

ferently in anechoic rooms. The walls of a typical anechoic
chamber are coated with sound absorption materials. Foam
wedges are common, as shown in Fig. 15. The absorbing
material nearly eliminates the reflection of sound by the Fig. 16. Hand-held acoustic intensity probe. Two microphones are separated by a
walls. The pressure measured around a sound source placed known distance ∆x so that particle velocity may be computed using a finite differ-
ence approximation. The mean pressure and estimated particle velocity are com-
in an anechoic room is due only to acoustic waves propagat- bined to compute the active, or real part of acoustic intensity in the direction along
ing away from the source. the x-axis.
Anechoic environments allow for more refined meas-
urements of sound fields, both over space and frequency. Two measurement chambers are required to determine
Directivity plots like those in Fig. 3 and Fig. 8 may be meas- a barrier’s transmission loss. Often, a thick, nearly rigid wall
ured. Also, narrow-band frequency spectra may be comput- is built between the two rooms. The wall includes an open-
ing, into which a barrier is mounted. One of the rooms is
considered the ‘source’ room, and is ensonified with sound
waves, which strike the barrier. The sound power in the
‘receiver’ room is measured, and a transmission coefficient is
calculated. Usually, the source room is reverberant, and the
receiver room is anechoic, with measurements made accord-
ing to the procedures we described above.

Boundary element modeling of sound fields

Boundary element methods have been used to predict
acoustic fields for about 45 years, and there is a vast amount
of literature on the subject. Although these methods can be
used to compute both interior and exterior acoustic fields,
the focus here is on computing exterior acoustic fields where
boundary elements have distinct advantages. Since this arti-
cle is at the level of a tutorial, our main goal will be to give the
reader a brief synopsis of the basic analysis techniques and
the current state of the art. For the sake of simplicity, the
specified boundary condition and radiated acoustic field are
Fig. 15. ARL/Penn State's Anechoic Chamber. assumed to be time-harmonic.

Structural Acoustics Tutorial 2 21

 A completely general solution for the pressure field of a tions should yield finite values. However, we would not
vibrating structure can be derived in the form of an integral expect a hyper-singular function integrated over a surface to
equation, which is an equation with an unknown function yield a finite value. Indeed, the only reason the velocity field
under an integral sign. Even though we cannot actually solve of a dipole source yields a finite value is because the surface
the integral equation for the unknown function explicitly for pressure and normal velocity, which are weighting functions
any but the simplest problems, it is still useful to know the for simple and dipole sources, are related through a gradient
form of the solution in terms of physical variables. Also, the operation. To illustrate, consider a right-angled corner,
solution provides great insight into how the problem might where the surface normal is discontinuous. As we travel
be solved numerically, and, as such, the Kirchhoff- along the surface towards the corner, the pressure field must
Helmholtz integral equation (KHIE) commonly provides the change such that the gradient operation produces the correct
starting point for many numerical formulations: velocity on the other side of the corner.19 The level of conti-
nuity in the pressure field is also important. In the exact solu-
tion, the surface pressure is an absolutely continuous func-
(28) tion, such that the function itself, as well as all of its deriva-
tives, are continuous. This level of continuity is impossible to
duplicate with simple polynomial approximations. However,
In the KHIE, r represents a point in the acoustic fluid the pressure and ∆surface velocity, taken together, must
and q is a point on the boundary of the fluid S, or boundary enforce - iωρ v = - p as we travel along the boundary. Thus,
surface. In the context of sound radiation from structural it is impossible to duplicate the level of continuity in the
vibrations, the boundary surface is the outer surface of the actual surface pressure distribution with simple interpola-
structure in contact with the acoustic medium. The function tion functions and it is also impossible to exactly enforce the
G stands for Green’s function, and G and dG/dn in the KHIE specified boundary conditions!
represent the acoustic fields of simple (monopole) and dipole From a practical point of view, how is all this relevant?
sources, respectively, distributed along the boundary surface. First, many of the research papers written about boundary
The dipole sources are aligned in the direction perpendicular element methods in the last ten years are concerned with
to the boundary surface. The monopole and dipole sources sorting out the mathematical details of the integrals. These
are functions of the point q where they are located and the papers are written by, and for, people writing their own
field point r. The KHIE shows that the acoustic field depends boundary element codes. A novice would find it very difficult
only on what happens at the boundaries since the surface to understand all the mathematical complexities.
vibrations cause the radiated acoustic field. (Admittedly, even after years of dedicated effort, we find
By studying the solution for the acoustic field given by many of the papers incomprehensible.) A reader interested
the KHIE, much can be learned about how to derive a in a simple explanation of boundary element methods will
numerical solution. Unfortunately, its simplicity is deceiv- find the earliest papers on the subject, written in the 60’s,
ing. The acoustic fields of simple and dipole sources are sin- much easier to understand.20, 21, 22 Also, for the casual users
gular functions of the source and field point locations, such who are not trying to write their own boundary element
that they become infinite when the two points coincide code, it is only important to understand that current bound-
(when r 씮 q ). However, in the KHIE, these singular func- ary element codes are not perfect. They probably do not take
tions add together to yield nonsingular pressure and velocity care of the singularities in the KHIE such that the acoustic
fields. Taking R to be the distance between r and q, the sin- field is strictly non-singular. Nonetheless, it has been well
gularities for the different sources can be categorized as: shown over the years that even simple approximations for
(1) pressure of a simple source G ~1/R, the pressure and normal surface velocity weighting functions
(2) velocity of a simple source dG/dn0 ~ 1/R2, yield adequate numerical solutions for many problems.22
(3) pressure of a dipole source dG/dn ~ 1/R2, After constructing and solving the matrix system, the
(4) velocity of a dipole source d(dG/dn0)/dn ~1/R3. pressures and normal velocities are known for each element
The process of taking a derivative makes a singular func- of the boundary surface. The KHIE can then be used to
tion more singular. For example, taking the derivative of r-1 directly compute the pressure at each desired field point
with respect to r gives -r-2. When r is less than unity, the func- location. The overall sound power output can be computed
tion r-1 is smaller than r-2, so that r-2 goes to infinity faster as by setting up a grid of field point locations on a surface
r 씮 0. A function that depends on 1/R is called weakly sin- enclosing the structure and numerically integrating the
gular, one that depends on 1/R2 is called strongly singular, acoustic intensity over the surface. In theory, the power out-
and one that depends on 1/R3 is called hyper-singular. put can also be computed directly using the pressure and
How is it possible, then, for singular functions to add up normal velocity on the boundary surface, but this is prob-
to give finite pressure and velocity fields? First, integration is lematic because this is where the largest errors tend to occur
a “smoothing” operation. Thus, it tends to reduce the level of in a boundary element solution.
singularity, the same way differentiation increases it. Because Aside from the mathematical details in evaluating the
the functions are integrated over a surface, this process integrals, most recent innovations in boundary element
essentially reduces the level of singularity by two orders. methods have concerned ways to increase the speed of the
Thus, it makes sense that weakly and strongly singular func- computations and reduce storage requirements. As with all

22 Acoustics Today, April 2007

numerical computations, boundary element methods have boundary and the finite element displacement vector.
benefited enormously from the rapid increase in computer Including the result for the pressure field on the boundary
processor speed and memory. Since boundary element com- surface in the finite element equations of motion yields
putations are repeated many times at different frequencies, it
is easy to split jobs up and assign them to different comput-
ers or processors. . (30)
In practice, boundary element methods are very com-
petitive for all but the very largest engineering problems, Given an input force vector, it is theoretically possible to
where the matrix inversion and multiplication dominate the solve this equation for the displacement vector, which now
solution times. This is especially true if the time required to includes fluid coupling. In a finite element analysis, the equa-
create models is factored into the analysis. In this respect, tion system is highly-banded because each element only
boundary element methods are relatively quick and easy. interacts with other elements through the nodes. However,
Simple rules of thumb can be used to determine if boundary in a boundary element analysis, every node interacts with
element methods will be applicable for a specific problem. every other node, so that the acoustic impedance matrix is
Generally, six elements are required per acoustic wavelength generally fully-populated. It then becomes very time-con-
to accurately compute an acoustic field. Knowing the surface suming to solve the matrix system in its present form.
area S of the boundary and the maximum frequency of inter- Various alternative strategies are possible where the matrices
est, the required number of acoustic elements is given as are subdivided into degrees of freedom with and without
S/(π/3fmax)2. Three matrices of this size will be required to fluid coupling. It is also possible to treat the acoustic pressure
compute the matrix multiplication, and it is usually best to field on the boundary surface as the primary variable.21
do the computations in double-precision, thus requiring 16 Ultimately, many researchers instead reformulate the prob-
bytes for each element of the matrices. lem in terms of a modal frequency response analysis (which
is the approach we use at Penn State). In a modal frequency
Modeling sound—structure coupling response, the mass, damping, and stiffness matrices are pre-
Recall that a moving structure compresses and contracts and post-multiplied by mode shapes, and the applied force
neighboring fluids, which act as a continuous elastic blob vector is pre-multiplied by the mode shapes. The resulting
around the structure. When the fluid is heavy, its pressure system, which is usually much smaller in size than the system
loading affects the structural vibrations. To compute these defined in physical coordinates, is solved to compute modal
effects, the boundary element model must be coupled to a coefficients. The modal coefficients tell us how much each
representation of the structure. mode contributes to the overall response, and are multiplied
The previous analysis has shown that an acoustic imped- by the mode shapes to compute that physical response.
ance matrix, relating the pressures to the normal component Using the coupled FE/BE formulation, the time required
of velocity on the boundary surface, can be computed as a to compute and store the acoustic impedance matrix as a
function of frequency using boundary element methods. We function of frequency typically dominates the overall solu-
now want to combine this result with a finite element analy- tion times. To speed up this part of the process, it is common
sis of the structural vibrations to include the effects of the to condense the structural displacement variables into a
pressure field. In a finite element analysis, the displacements coarser set of acoustic variables. To illustrate, Fig. 17 shows
are written in the form two structural and acoustic meshes for a loudspeaker.
Since both the structural and acoustic fields require
, (29) approximately six elements per wavelength, this process
assumes that the structural waves vary more rapidly than the
where M, B, and K are the mass, damping, and stiffness
acoustic waves. This is true below the coincidence frequency,
matrices, d is the displacement field of the structure, and F is
and is valid up to a relatively high frequency for structures
the vector of forces applied to the structure.
submerged in water.
To include the pressure field in the finite element analy-
To demonstrate the typical results of a coupled FE/BE
sis, the impedance matrix must be transformed from its
analysis, two example problems will be considered. The first
dependence on normal surface velocity to nodal displace-
ments. For time-harmonic problems, the displacement field
on the outer surface of the structure can be used to compute
the normal component of the surface velocity as a simple
dot-product: vn = v . n = -iω d . n. Thus, a matrix relationship
can be derived to convert the complete displacement field
into a normal surface velocity vector. This matrix will have
zeros for all interior nodes not in contact with the surround-
ing fluid. Similarly, the matrix will also be zero for displace-
ment degrees of freedom on the boundary surface tangential
to the outward surface normal. Post-multiplying the acoustic
impedance matrix by the transformation matrix then yields
a matrix relationship between the pressure field on the Fig. 17. A structural mesh for a speaker with two different acoustic meshes.

Structural Acoustics Tutorial 2 23

 is the familiar example of a thin circular cylindrical shell in
an infinite circular baffle. For simply-supported boundary
conditions, the structural displacements can be determined
analytically. Similarly, the acoustic Green’s function is
known analytically for a simple source outside of a rigid cir-
cular cylinder, and thus for this problem the acoustic field
can also be determined analytically, although it can only be Fig. 18. Finite element predictions for the first two mode shapes and resonance fre-
evaluated numerically. Solutions for the in vacuo and fluid- quencies of a thin shell. The rigid ends on both sides of the shell simulate semi-infi-
coupled resonance frequencies are given by Berot and nite rigid baffles.
Peseux.23 For this problem, the cylinder has radius 0.4 m,
length 1.2 m, and is 3 mm thick. Its material properties are elements in the axial direction, and 16, 24, and 48 elements
given as E = 200 GPa, ρ = 7850 kg/m3, v =0.3, and the sur- around the circumference. Table 2 lists the resonance fre-
rounding water has properties co = 1500 m/s, and ρo = 1000 quency ratios given by Berot and Peseux along with numer-
kg/m3. We note that the thickness–to-radius ratio is less ical predictions for the various acoustic element meshes.
than 0.01 for this example, and thus thin shell theory is In the table, NA is the number of acoustic elements for
applicable and we expect a relatively large shift in the reso- each of the boundary element meshes. Clearly, the numeri-
nance frequencies due to fluid-loading. Our goal will be to cal and analytical predictions match each other closely,
reproduce Berot and Peseux’s results. In the numerical although a relatively fine acoustic mesh is necessary in the
computations, the structural analysis proceeds as usual, but circumferential direction to achieve convergence for the
we cannot simply model the boundary surface in the higher order modes, as was expected. Overall, for structures
acoustic analysis because the circular baffle extends to submerged in water, the fluid-loading analyses has been
infinity in both directions. shown to yield excellent predictions for the added mass.
Following the procedure used by Berot and Peseux in We consider a cavity-backed plate as a second example
their numerical computations, we will truncate the baffle problem with structural-acoustic coupling. In the 1970’s,
on both sides of the cylinder approximately half its length Guy and Bhattacharya24 studied transmission loss through a
beyond the vibrating portion and close off the ends with flat cavity-backed finite plate and their results have been used
endcaps. Figure 18 shows the first two computed in vacuo subsequently by several authors to validate numerical pre-
mode shapes and resonance frequencies for the shell. dictions. We will similarly use the problem to illustrate how
In the finite element analysis, the displacement is set to dipole sources can be used to model interior and exterior
zero at the ends of the cylinder in the radial and torsional acoustic fields simultaneously in a scattering problem.
directions, but the axial displacement is unconstrained. As Figure 19 shows the geometry of the cavity and plate.
we will show, a fairly fine mesh is required in the circum- The plate is 0.914 mm thick and is made of brass with
ferential direction to properly represent the acoustic properties E = 106 GPa, ρ = 8500 kg/m3, v = 0.3, and the
impedance of the higher order modes. For example, to cor- surrounding air has properties co = 340 m/s, and ρo = 1.2
rectly represent a mode with 8 circumferential wavelengths, kg/m3. The plate is simply-supported and the backing cavi-
we need (8)(6)=48 acoustic elements around the circumfer- ty has rigid walls. We want to compute the transmission
ence. The acoustic mesh does not have to be nearly as loss through the plate to a field point location at the center
refined in the axial direction because the structural stiffness of the back wall of the cavity. Guy and Bhattacharya’s trans-
is much higher. With this in mind, the structural mesh has mission loss is actually just a ratio of the incident pressure
24 elements in the axial direction and 48 elements around and the pressure near the back wall of the box:
the circumference. The acoustic meshes have 8, 12, and 12

. (31)

For the numerical analysis, we can perform the compu-

tations in one of two ways. We could simply apply mechan-
ical forces to produce a uniform pressure to the top surface
of the plate, as in the studies by S. Suzuki, et al.25 and M.
Guerich and M. A. Hamdi,26 or we could simulate the
experiment using an acoustic source as the excitation. We
will use the latter method. The surface of the cube is divid-
ed evenly into 144 quadrilateral elements, yielding a
boundary surface mesh with 864 structural elements.
The incident pressure is computed knowing the source
location and the distance to the center of the plate. The
Table 2. Analytical and numerical predictions for the resonance frequency ratios
(in fluid/in vacuo) of the circular shell. Mode orders are indicated as (m,n) pairs,
results in Fig. 20 show very good agreement between our
where m is the order along the axis, and n is the circumferential harmonic. numerical predictions and Guy and Bhattacharya’s experi-

24 Acoustics Today, April 2007

 Fig. 20. Transmission loss as a function of frequency for a cavity-backed plate.

Acknowledgments
We are grateful for our daily interaction with the
ARL/Penn State Structural Acoustics Department (Andrew
Fig. 19. Dimensions for the cavity-backed plate example problem. Barnard, Robert Campbell, Stephen Conlon, David Jenkins,
and Tim McDevitt), along with some of our students in Penn
State’s Graduate Program in Acoustics, in particular, Ben
Doty, who measured the response of the elbowed pipe.AT
mental measurements, thus demonstrating that dipole
sources can be used to model both the interior and exterior
References for further reading
acoustic fields simultaneously. 1 G. H. Koopmann and J. B. Fahnline, Designing Quiet Structures
(Academic Press, San Diego, 1997).
Summary 2 R. D. Blevins, Flow-Induced Vibration, 2nd Edition (Krieger
In Part 2 of this tutorial on structural acoustics, we Publishing Company, Malbar, FL, 2001).
have learned about how acoustic fluids interact with struc- 3 E. Naudascher and D. Rockwell, Flow-Induced Vibrations–An
tures, both as an acceptor and a cause of vibrational energy. Engineering Guide (Dover Publications, Mineola, NY 2005).
We have presented some simple and difficult concepts in a
relatively short article (entire textbooks are devoted to the
subjects we have discussed), and hope the information is
useful as a handy reference. For those of you who are inter-
ested in learning more about these topics, please look
through the references we have provided. You are also wel-
come to enroll in the Sound—Structure Interaction course
offered by the Penn State Graduate Program in Acoustics
for a more thorough treatment of this subject.
Of course, there is much we have not explained, but we
can refer you to other strong references on those subjects.
For example, we have focused almost entirely on the inter-
action of structures with exterior fluids. Acoustic cavities
contain resonances which can interact with the walls that
bound the interior space, particularly at low frequencies.
Some classic papers which introduce this topic are those by
Pretlove27 and Dowell.28 During the Active Noise Control
(ANC) boom of the 1980’s and 1990’s, many people inves-
tigated how to control the sound within acoustic cavity
modes by driving the enclosure boundaries with tuned
forces. Nelson and Elliott’s textbook29 is a good reference on
ANC.
Another relatively modern structural–acoustic topic is
Nearfield Acoustic Holography (NAH), which is an inverse
technique for inferring a structure’s surface vibrations from
a complex pressure hologram measured near the surface.
Once the surface velocities are known, numerical boundary
value techniques can be used to compute the far-field
sound radiation. We recommend the textbook by
Williams30 to those interested in NAH.

Structural Acoustics Tutorial 2 25

 4 F. J. Fahy, Sound and Structural Vibration: Radiation, Degree Elbows, Under Water Loading,” Proceedings of NoiseCon
Transmission, and Response (Academic Press, San Diego, 2005, Minneapolis, MN, 17–19 October 2005 (Institute of Noise
1987). Control Engineering, Ames, IA, 2005).
5 A. D. Pierce, Acoustics–An Introduction to its Physical Principles 18 ANSI S12.35–1990 (R1996), Determination of Sound Power
and Applications (Acoustical Society of America, Melville, NY, Levels of Noise Sources in Anechoic and Hemi-Anechoic Rooms,
1991). (Acoustical Society of America, Melville, NY, 1996).
6 M. S. Junger and D. Feit, Sound, Structures, and Their Interaction 19 L. J. Gray and E. Lutz, “On the Treatment of Corners in the
(Acoustical Society of America, Melville, NY, 1993). Boundary Element Method,” J. Comp. Appl. Math. 32, 369–386
7 G. Maidanik, “Response of Ribbed Panels to Reverberant (1990).
Acoustic Fields,” J. Acoust. Soc. Am. 34, 809–826 (1962). 20 L. G. Copley, “Integral Equation Method for Radiation from
8 C. E. Wallace, “Radiation Resistance of a Rectangular Panel,” J. Vibrating Bodies,” J. Acoust. Soc. Am. 41, 807–816 (1967).
Acoust. Soc. Am. 51, 946–952 (1972). 21 H. A. Schenck, “Improved Integral Formulation for Acoustic
9 P. W. Smith, Jr., “Response and Radiation of Structural Modes Radiation Problems,” J. Acoust. Soc. Am. 44, 41–58 (1968).
Excited by Sound,” J. Acoust. Soc. Am. 34, 640–647 (1962). 22 L. H. Chen and D. G. Schweikert, “Sound Radiation from an
10 C. B. Burroughs, R. W. Fischer, and F. R. Kern, “An Arbitrary Body,” J. Acoust. Soc. Am. 35, 1626–1632 (1963).
Introduction to Statistical Energy Analysis,” J. Acoust. Soc. Am. 23 F. Berot and B. Peseux, “Vibro-acoustic behavior of submerged
101, 1779–1789 (1997). cylindrical shells: Analytical formulation and numerical model,”
11 P. J. Shorter and R. S. Langley, “On the Reciprocity Relationship J. Fluids and Struct. 12(8), 959–1004 (1998).
Between Direct field Radiation and Diffuse Reverberant 24 R. W. Guy and M. C. Bhattacharya,“The transmission of sound
Loading,” J. Acoust. Soc. Am. 117, 85–95 (2005). through a cavity-backed finite plate,” J. Sound Vib. 27(2),
12 M. J. Crocker and A. J. Price, “Sound Transmission Using 207–223 (1973).
Statistical Energy Analysis,” J. Sound Vibration 9, 469–486 25 S. Suzuki, S. Maruyama, and H. Ido, “Boundary element analy-
(1969). sis of cavity noise problems with complicated boundary condi-
13 A. J. Price and M. J. Crocker, “Sound Transmission Through tions,” J. Sound and Vib. 130(1), 79–91 (1989).
Double Panels Using Statistical Energy Analysis,” J. Acoust. Soc. 26 M. Guerich and H. A. Hamdi, “A numerical method for vibro-
Am. 47, 683– 693 (1970). acoustic problems with incompatible finite element meshes
14 ISO 3741/ANSI S12.51, Determination of Sound Power Levels of using B-spline functions,” J. Acoust. Soc. Am. 105(3),
Noise Sources using Sound Pressure – Precision Methods for 1682–1694 (1999).
Reverberation Rooms (Acoustical Society of America, Melville, 27 A. J. Pretlove, “Forced Vibrations of a Rectangular Panel Backed
NY, 1999). by a Closed Rectangular Cavity,” J. Sound Vib. 3, 252–261
15 ISO 354, Measurement of Sound Absorption in a Reverberation (1966).
Room (International Organization for Standardization, 28 E. H. Dowell, G. F. Gorman, and D. A. Smith,
Geneva, 2003). “Acoustoelasticity: General Theory, Acoustic Natural Modes
16 S. C. Conlon, S. A. Hambric, and W. K. Bonness, “Evaluation and Forced Response to Sinusoidal Excitation, Including
of a reverberant water tank for radiated power measurements,” Comparisons with Experiment,” J. Sound Vib. 52, 519–542
Proceedings of NoiseCon 2004, Baltimore, MD, 12–14 July (1977).
2004. (Institute of Noise Control Engineering, Ames, IA, 29 P. A. Nelson and S. J. Elliott, Active Control of Sound (Academic
2004). Press, San Diego, 1993).
17 B. J. Doty, S. A. Hambric, S. C. Conlon, and J. B. Fahnline, 30 E. G. Williams, Fourier Acoustics–Sound Radiation and Nearfield
“Structural-Acoustic Measurements of Pipes with Ninety- Acoustic Holography (Academic Press, San Diego, 1999).

26 Acoustics Today, April 2007

 Stephen A. Hambric Acoustics Division, and as an associate editor of ASME’s
is head of the Journal of Vibration and Acoustics.
Structural Acous-
tics Department at John B. Fahnline is
the Applied Research a research associ-
Lab at Penn State ate at the Applied
and Associate Prof- Research Labora-
essor in the Grad- tory and an assis-
uate Program in tant professor of
Acoustics. Prior to acoustics at Penn
joining Penn State State. He holds a
in 1996, Dr. Hambric Ph.D. in Acoustics
worked for nine from the Pennsylv-
Steve Hambric and his daughter Lily. years in the Comp- ania State University.
u t a t i o n a l As a post-doctoral
Mechanics Office at the Naval Surface Warfare Center, researcher, Dr.
Carderock Division. Dr. Hambric has directed many numeri- John Fahnline (in background) and his niece Mariel. Fahnline designed
cal and experimental flow and structural acoustics research and and implemented
development programs for the Navy, U.S. industry, and the a practical, effi-
U.S. Nuclear Regulatory Commission. He has authored over cient, and accurate method for numerically solving general,
60 conference and journal articles and advised many graduate three dimensional acoustic radiation problems which can be
students at Penn State. He teaches courses in Structural applied to problems in design optimization and active noise
Acoustics, and Writing for Acousticians on campus at Penn control. The code, named POWER, is currently being used
State, and also to off-campus students working in industry and at the ARL Penn State to predict noise radiation from the
government. He currently serves on the board of directors of structural resonances of compound propulsors. Dr.
the Institute for Noise Control Engineering (INCE), on the Fahnline coauthored a book on design optimization with Dr.
Executive Committee of the ASME’s Noise Control and Gary Koopmann.

Certified to meet ASTM E90,

Doors up to STC 55 E336, E1408 and E413

Hundreds of configurations

Unlimited design possibilities

See what’s possible, visit our project gallery at:

www.KriegerProducts.com/gallery

Visit: KriegerProducts.com
Call: 1-800-251-5175

Structural Acoustics Tutorial 2 27

View publication stats