Journal of Sound and Vibration (1998) 211(3), 435448

MULTI- PASS PER FOR ATED TUBE SILENCER S:

A COMPUTATIONAL APPR OACH
N. S. Dicir.
Department of Mechanical Engineering and Applied Mechanics,
The University of Michigan, Ann Arbor, Michigan 48109, U.S.A.
A. Sri:xr1
Department of Mechanical Engineering, The Ohio State University, 206 West 18th Avenue,
Columbus, Ohio 43210-1107, U.S.A.
:Nb
J. M. Nov:i
Engine Product and Manufacturing Engineering, Ford Motor Company, Dearborn,
Michigan 48121, U.S.A.
(Received 13 April 1996, and in nal form 4 March 1997)
A time domain computational approach is applied to predict the acoustic performance
of multiple pass silencers with perforated tube sections. This non-linear, one-dimensional
method may readily include temporal and spatial variations in sound pressure level, orice
ow velocities, and mean duct ow, all of which aect the local orice behaviour of
perforated tube elements, and therefore the overall noise reduction characteristics. The
transmission loss of two anechoically terminated multiple pass muer congurations is
determined computationally and experimentally for the limiting case of low sound pressure
levels and zero mean ow. Comparisons between the numerical results and experimental
data are shown to correlate well for frequencies where the one-dimensional assumption is
justied.
7 1998 Academic Press Limited
1. INTRODUCTION
The silencing components used to reduce the sound radiated from automotive intake and
exhaust systems often employ one or more perforated tube segments. For some elements,
the overall structure may be complex enough to require a detailed, three-dimensional
model to determine its general noise attenuation. However, for many practical
congurations the relatively long wavelengths of the noise generated by the engine intake
and exhaust processes allow simplied analyses. If the wavelength is much longer than any
characteristic silencer dimension, a zero-dimensional, or lumped analysis, is valid and
presents the least amount of diculty. For an automotive silencer, zero-dimensional
models may provide some insight at very low frequencies, but accuracy decreases rapidly
with increasing frequency. Signicant improvement from a lumped approach may be
obtained by considering one-dimensional ow in the various silencer duct passages that
communicate through the perforated interfaces. This coupled one-dimensional method is
useful for low frequency acoustic analysis of many realistic silencer designs characterized
An earlier and abbreviated version of this work was presented at the 1996 International Congress and
Exposition of the Society of Automotive Engineers, February 2629, 1996, Detroit, Michigan.
0022460X/98/130435 +14 $25.00/0/sv971318 7 1998 Academic Press Limited
N. s. bicir. r1 :i. 436
by perforated tubes that are axially aligned within an external duct having a length that
is substantially longer than any cross dimension. At higher frequencies, where the
wavelength is not much longer than any cross dimension, or for very complex geometries,
a coupled one-dimensional approach is no longer justied and multidimensional
techniques need to be used.
The response of a perforated tube silencer is aected by the overall geometric structure,
including duct porosities and orice locations, and the perforate dynamics. The behavior
of a perforated interface is usually presented in terms of a perforate impedance, Z
p,f
, a
frequency domain parameter given by
Z
p,f
=Dp
p,f
/U
p,f
, (1)
where Dp is the acoustic pressure dierence, U is the area averaged velocity and the
subscripts p and f refer to the perforate and fundamental frequency, respectively. It is well
known that the dynamics of unsteady ow through an orice are dependent on a number
of local uid-dynamic parameters in addition to the specic orice geometry. Clearly the
physical uid properties (density, viscosity, etc.) are signicant. Furthermore, the
magnitude of the throughow (mean plus oscillating), which is strongly coupled with the
average pressure dierence and sound pressure level across the perforate, causes drastic
changes in the orice behavior. Finally, the amount of grazing (axial duct) ow on either
side of the perforate alters the local orice dynamics.
For the conditions present in an automotive silencer, it is most likely that the perforated
sections will exhibit non-linear behavior. For example, with zero mean ow, orice
non-linearity is observed at sound pressure levels on the order of 120130 dB
re 20 mPa
[1], well
below the levels accepted to be linear in one-dimensional duct acoustics and much lower
than the magnitudes experienced in typical automotive intake and exhaust systems.
Moreover, the values of uid properties, acoustic pressure and bulk uid ow can vary
with both time and location. Since the perforate behavior depends on these parameters,
it is reasonable to expect signicant non-linearity at the perforate interfaces.
A number of studies have employed a coupled one-dimensional approach to study
perforated tube geometries by using a linear wave equation to model the duct ows. The
majority of these methods have been frequency domain approaches. Sullivan and Crocker
[1] examined the acoustic performance of perforated concentric tube resonators under
linear orice conditions. Later Sullivan [2] analyzed one and two pass geometries using
an iterative segmentation approach which has a much broader applicability, since it allows
for spatial changes in the perforate impedance. However, the mean values of orice and
duct ow, need to be specied a priori. Analytical [3, 4] and numerical [5] decoupling
techniques have also been applied to perforated tube geometries. In the analytical
decoupling techniques, the perforate impedance and mean duct ow are assumed to be
constant along a particular duct. The numerical decoupling approach of Peat [5] allows
orice impedances and mean ow to vary along a duct, although the mean ow must
change linearly. Results for these decoupling analyses were presented for one and two pass
silencers. Correlation between the foregoing frequency domain techniques and
experimental data is generally quite good, provided that the geometry and ow conditions
are consistent with the underlying assumptions of the particular method. As an alternative
to frequency domain approaches, Chang and Cummings [6] have applied a numerical
time-domain technique that uses a nite-dierence approximation of the linearized
one-dimensional wave equation in the main ducts, which may communicate via a
non-linear orice model. Since the solution proceeds in the time domain, temporal and
spatial changes in orice behavior may be accounted for. They obtained good correlation
17
180
120
1,6
Exhaust locat ion
S
o
u
n
d

p
r
e
s
s
u
r
e

l
e
v
e
l

(
d
B
)
12
160
140
7,8 7a,8a 9,10 11 15 16
17
1200
400
1,6
Exhaust locat ion
M
e
a
n

g
a
s

t
e
m
p
e
r
a
t
u
r
e

(
K
)
12
800
7,8 9,10 11 15 16
irrror:1rb 1inr siirNcrrs 437
Figure 1. Measured sound pressure levels at different engine speeds for the exhaust system locations in Table 1:
, 1000 r.p.m; -----, 3000 r.p.m.; , 5000 r.p.m.; W, right bank; w, left bank.
between computational and experimental results for high amplitude, non-harmonic wave
propagation through a single pass resonator.
It is apparent that the analysis of a perforated tube silencer by itself can be a dicult
task. For silencers placed in an internal combustion engine intake or exhaust system, this
diculty is compounded by a number of other factors. To determine accurately the sound
radiated from any system, the source must be modelled accurately. For an automotive
engine, this includes the highly non-linear processes taking place in the vicinity of the
valves. In addition, ow losses created by the duct systems and silencers aect the
backpressure (or inlet pressure) seen at the valve, and therefore modify the source
characteristics. The experimental data shown in Figures 1 and 2 illustrate additional
problems encountered when attempting to determine radiated sound levels of automotive
engines. The variations in sound pressure level and mean temperature at wide open throttle
conditions are shown for a Ford 1992 3.0L V-6 (Taurus) engine and production exhaust
system. The data presented here are part of the experimental study by Selamet et al. [7].
In this system, the exhaust ows from the right and left banks are initially separate as
the combustion products from three cylinders are combined in the manifold and sent
through a short passage to a catalytic converter. Shortly after the catalytic converters, the
Figure 2. Measured mean temperatures at different engine speeds for the exhaust system locations in Table 1.
Key as for Figure 1.
N. s. bicir. r1 :i. 438
T:nir 1
Measurement location descriptions
Location Description Location Description
1 Cylinder 1 primary runner 10 L.B. Catalytic converter exit
6 Cylinder 6 primary runner 11 Resonator inlet
7 R.B. manifold exit 12 Resonator exit
7a R.B. catalytic converter inlet 15 Mufer inlet
8 L.B. manifold exit 16 Mufer exit
8a L.B. catalytic converter inlet 17 Surge tank
9 R.B. catalytic converter exit
right and left exhaust pipes merge into a single duct. After the exhaust ows merge, the
uid is passed through a small resonator, followed by another section of tubing and then
the muer. Beyond the muer is a short tailpipe section connected to a large surge tank,
implemented to dampen pressure oscillations in place of an expansion to ambient.
Measurement locations are described in Table 1, where R.B. and L.B. denote the right and
left engine banks, respectively.
The overall sound pressure levels for the locations described in Table 1 are depicted in
Figure 1. These values are obtained from the measured time-domain data as
SPL(dB) =10 log
10
$
1
Dt
g
t +Dt
t
0
p p
p
ref 1
2
dt
%
, (2)
where Dt is the time period for one complete engine cycle (two revolutions), p is the average
pressure at the sampling point, and p
ref
=20 mPa. At locations upstream of the catalytic
converters, SPL values range from approximately 167 dB to well over 180 dB. Oscillations
of this magnitude would indicate that non-linear eects may also be signicant in the
interactions between the main runners in addition to the processes occurring at the valve.
Inlet sound pressure levels of approximately 164166 dB and 154161 dB for the resonator
and muer, respectively, show that the perforated components in these elements are
almost certainly acting in a non-linear fashion, although grazing ow and perforate
throughow will aect their response and require additional consideration.
In the exhaust system, the speed of sound (or, equivalently, the wavelength of a given
frequency) at a point will mainly depend on the local temperature. Figure 2 shows the
variation in mean temperature with both distance from the exhaust port and engine speed.
For the 1000 r.p.m. case, the mean temperature varies from approximately 825 to 500 K
between locations 1 and 16. With respect to the primary runner, the relative change in the
mean speed of sound between these locations is approximately
Dc
16 1
/c
1
2zT
16
/T
1
1 =022, (3)
resulting in a 22% shift in the wavelength of a specic frequency component. The large
dierences seen in the temperature at a xed location for dierent engine speeds also
require consideration.
For approaches using linearized equations to analyze unsteady duct ows, the
applicability to an overall intake or exhaust system becomes questionable. Moreover, these
approaches require a xed source boundary condition, such as an input waveform or
source impedance to represent the engine (see, for example, Gupta and Munjal [8]).
Therefore changes in the system design which aect the ow through the valves (by
irrror:1rb 1inr siirNcrrs 439
changing backpressure or system dynamics) may only be included externally and require
separate analytical techniques. The complexity and numerous non-linearities observed in
automotive intake and exhaust systems suggests the need for a method that can accurately
model these phenomena. Non-linear one-dimensional techniques have proven useful for
predicting intake and exhaust system ows in the ducts of automotive engines [912]. These
techniques have mainly been used to study the eects of intake and exhaust system
geometry on engine performance parameters (volumetric eciency, power, etc.), although
some simple silencers have been considered. More recently, analysis of single pass
perforated tube geometries for low sound pressure levels and zero mean ow has been
performed [13, 14].
The objective of the current study is to apply a non-linear uid dynamic model to predict
the transmission loss of multiple pass perforated tube silencers a typical example of which
is shown in Figure 3. The multiple pass geometries, selected to be more representative of
realistic muers, are modelled under low sound pressure level and zero mean ow
conditions. The numerical method employed is an extension of the non-linear,
one-dimensional nite dierence scheme used by Chapman et al. [11]. In an earlier work
[14] this technique was applied to model single pass perforated tube silencers under zero
mean ow and low sound pressure level conditions. Good agreement was obtained with
experimental results for both short and long concentric tube resonators. Since the
technique is applicable to practically any geometry, provided that one-dimensional ow
is justied, it can be used to model the entire intake and exhaust system, accounting for
variations in ow dynamics that are dependent on the engine, manifold and silencer
congurations. In addition, the computational program connects the ducts, perforated
tubes and volumes which comprise the silencer in a modular fashion, so that multiple pass
element models are readily described in a program input le.
Though the numerical technique is applicable for the numerous non-linearities observed
in automotive intake and exhaust systems, the present study is restricted to the low sound
pressure level, zero mean ow regime. This limit is chosen to allow comparisons between
the computational results and the experimental data for multiple pass perforated tube
elements in an anechoically terminated impedance tube test facility.
2. DUCT NUMERICS
The governing balance equations of mass, momentum and internal energy for an
unsteady, compressible ow may be expressed as
1r/1t +9 (rU ) =0, (4)
Figure 3. Geometry of the test mufer.
Cont r ol volume (CV)
Cont r ol sur face (CS )
y
p
o
p
i
U
p
N. s. bicir. r1 :i. 440
Figure 4. Nomenclature and control volume used for application of momentum equation to a single orice.
(1/1t)(rU ) +9 (rU U ) +9p 9 t =0, (5)
(1/1t)(re) +9 (rU e) +p9 U t : 9U +9 q =0, (6)
where r is density, U is velocity, p is pressure, t is the shear stress tensor, e is the specic
internal energy and q is the heat ux. The ideal gas equation of state,
p =(g 1)re, (7)
where g is the ratio of specic heats, is used to relate the thermodynamic variables and
close the system of equations. Equations (4)(7) are discretized for one-dimensional ow
in ducts of variable cross-section by employing the explicit nite dierence method of
Chapman et al. [11] as discussed by Selamet et al. [15]. Since the numerical treatment of
the duct ows is essentially unchanged by the incorporation of perforate communication,
the following description will focus only on the orice ow treatment in the model. Details
of the one-dimensional duct numerics may be found in references [11] and [15].
For the perforate interface model, it is assumed that the locally incompressible orice
ows act independently. Application of the y-direction momentum equation to a single
orice, as shown in Figure 4 yields, in integral form,
d
dt
g
CV
rU
y
dV=(p
i
p
0
)A
p

g
CS
t
y
dA
g
CS
rU
y
(U n ) dA, (8)
where the subscripts CV and CS denote the control volume and control surface,
respectively, A
p
is the orice area and n is the unit outward normal vector. For low sound
pressure levels and zero mean ow, the convective term on the right side of equation (8)
may be neglected, resulting in a balance between the inertial, pressure and viscous forces
as
d
dt
g
CV
rU
y
dV=(p
i
p
0
)A
p

g
CS
t
y
dA. (9)
Equation (9) is further simplied to
M
eq
dU
p
/dt =(p
i
p
0
)A
p
RU
p
A
p
, (10)
Per for at e
int er face
Out er
cont r ol
volume
Inner
cont r ol
volume
U
U
U
U
j
U
U
U
U
j + 1
, e, P
, e, P
A
p, Up, Meq, R
irrror:1rb 1inr siirNcrrs 441
where the coecients M
eq
and R represent an equivalent mass and equivalent resistance,
respectively. Division by A
p
yields the expression employed in the current model as
rl
eq
dU
p
/dt =p
i
p
0
RU
p
, (11)
where l
eq
=M
eq
/A
p
is an equivalent length for the orice. This treatment is equivalent to
the work of Sullivan and Crocker [1], although the solution proceeds in the time domain
rather than being determined in the frequency domain. In contrast to the frequency domain
approaches, the values for the eective length l
eq
and ow resistance R may be determined
from the instantaneous local uid dynamics, and therefore may vary with position and
time.
Due to the complexity of the unsteady ow through an orice, analytical
approximations for l
eq
and R only exist for the simplest of conditions. Consequently,
empirical data is generally required. Although a number of experimental studies have been
performed to analyze the eects of high sound pressure levels, mean throughow and
grazing ow [1620], the ndings are almost exclusively presented in terms of a non-linear
orice impedance referenced to a single excitement frequency. Therefore these results may
only be used approximately to determine the variable coecients R and l
eq
for arbitrary
input waveforms. Cummings [16] has presented an essentially equivalent orice ow model
incorporating variable coecients for the case of zero mean ow and no grazing ow. In
addition, Chang and Cummings [6] employed frequency domain experimental data in a
time domain approach to model a single pass concentric tube resonator under multiple
frequency excitation. For the limiting case considered here, l
eq
and R are assumed constant
and are taken from reference [1].
The numerical model employs a staggered mesh that divides a duct into cells with vector
quantities located at node points and scalar quantities at cell midpoints. For a given
computational cell containing perforations, the orice area and perforate ow are assumed
to be evenly distributed over the wall area. Figure 5 depicts the reduction of a single pass
perforated tube silencer into inner and outer control volumes and shows the variable
centering used. In the numerical model, equation (11) is approximated as
r
n
l
eq
(U
n +1
p
U
n
p
)/Dt =p
n
i
p
n
0
RU
n
p
, (12)
Figure 5. Physical element reduction and variable centering for the numerical technique.
3000
20
Fr equency (Hz)
T
r
a
n
s
m
i
s
s
i
o
n

l
o
s
s

(
d
B
)
1200
10
600 2400 0 1800
d
1
d
2
l
e
N. s. bicir. r1 :i. 442
Figure 6. Transmission loss for a short concentric resonator (d1 =508 cm; d2 =762 cm; le =667 cm;
s =0037): , experiment; W, model.
where the superscript n denotes the timestep. Once the updated perforated velocity is
calculated from equation (12), the mass and internal energy uxes through the orice are
computed using upwind dierencing.
3. RESULTS
The computational model has been applied to a number of perforated tube silencers,
consisting of both single and multiple pass geometries. One particular conguration, the
single pass concentric tube resonator, has received much attention due to its practical
signicance and relatively simple geometry [14, 6, 13, 14, 21]. For comparative purposes,
the experimental and computational results for two such congurations are included here.
The geometries considered are evenly perforated over the central tube (see inset Figure 6),
and have dimensions consistent with the short and long resonators considered in more
detail by Sullivan and Crocker [1], as shown in Table 2. Figures 6 and 7 compare the results
of the numerical model with experimental data for the short and long resonators,
respectively. For both geometries, the predictions from the numerical model are in close
agreement with experimental data.
The geometry of the rst multi-pass test muer, a three pass conguration, is shown
in Figure 3. In this silencer, the uid travels along the entrance duct to the rst end
chamber, reverses direction and travels to the second end chamber through the pass tube,
and reverses direction once more prior to leaving the muer in the exit tube. Travelling
oscillations in all three ducts communicate with the outer cavity over the central section
of the muer. The three interior ducts have an outer diameter of 508 cm and a 008 cm
wall thickness, while the external casing is fabricated from a 1651 cm inner diameter
T:nir 2
Geometry of short and long resonators
Resonator le (cm) d1 (cm) d2 (cm) twall (cm) Porosity (s) dorice (cm)
Short 667 508 762 0079 0037 0249
Long 2572 508 1015 0079 0020 0249
3000
30
Fr equency (Hz)
T
r
a
n
s
m
i
s
s
i
o
n

l
o
s
s

(
d
B
)
1200
5
25
20
15
10
600 2400 0 1800
irrror:1rb 1inr siirNcrrs 443
Figure 7. Transmission loss for a long concentric resonator (d1 =508 cm; d2 =1015 cm; le =2572 cm;
s =0020): , experiment; W, model.
polycarbonate tube. The end chambers are separated from the 255 cm long central cavity
by 127 cm thick baes, and have volumes of 831 cm
3
and 892 cm
3
. Interior duct porosities
of 48% were created by drilling 400 holes of 0249 cm diameter in each duct over the
central region. To account for the local multidimensional eects at the expansion from the
central tubes to the end chambers, a small end correction is necessary. For simplicity, the
end correction is determined from the work of Ingard [22] for concentric Helmholtz
resonators as
d =0425d
p
0
1 125
d
p
d
V1
, (13)
where d
p
and d
V
are the pipe and cavity diameters, respectively.
Computational results for the three pass silencer are compared to experimental data in
Figure 8. Correlation between the model and experiment is satisfactory up to
approximately 1000 Hz, where multidimensional eects become signicant. Note that this
frequency is slightly lower than the limit for continuous propagation of the rst diametral
mode in a circular duct, given by f
rst diametral mode
=184c
0
/pd
max
=1220 Hz, for a speed of
sound of 344 m/s in air at atmospheric conditions. Some discrepancy between the model
and experiment is also seen at frequencies below 150 Hz, which may be attributed to slight
imperfections in the experimental facility anechoic termination. The overall behavior is
similar to an expansion chamber with a superimposed resonance near 450 Hz. Since the
central section is eectively an expansion chamber, it appears that the end chambers are
behaving as Helmholtz resonators (the end chamber volumes are very similar, so both
should resonate at approximately the same frequency). Assuming the two pipes attached
to an end chamber act synchronously, and using an average of the two end volumes, the
eective neck length necessary for a resonance frequency of 450 Hz is approximately
64 cm. If the perforations are assumed to be transparent, and equation (13) is applied at
both ends of the bae, the eective neck length is
l
0
+2d =444 cm, (14)
1000
50
Fr equency (Hz)
T
r
a
n
s
m
i
s
s
i
o
n

l
o
s
s

(
d
B
) 40
30
20
10
200 400 600 800 0
1000
50
Fr equency (Hz)
T
r
a
n
s
m
i
s
s
i
o
n

l
o
s
s

(
d
B
) 40
30
20
10
200 400 600 800 0
45
35
25
15
5
N. s. bicir. r1 :i. 444
Figure 8. Transmission loss of three pass mufer: , experiment; --W--, model.
where l
0
is the axial distance between the volume side of the bae and the rst row of
orices. Therefore, the perforated sections do not repress the interaction between the end
volumes and central chamber substantially, but show their presence by augmenting the
eective neck length.
The sensitivity of perforated tube silencers to variations in porosity is demonstrated in
Figure 9, which shows the computed transmission loss of the three pass muer for three
dierent duct porosities. The highest porosity shown (5%) is very close to the value for
the fabricated geometry, and therefore has essentially the same behavior seen in Figure
8. As the porosity is reduced, the inhibited communication between the end chambers and
the central cavity becomes more noticeable. This eect causes a reduction of the resonance
frequency (or, equivalently, an increase in the eective neck length) accompanied by a
decrease in the transmission loss magnitude near the resonance frequency. As expected,
the deviations from the expansion chamber behavior increase as the porosity is reduced.
The departure from an overall dome-like behavior is most prevalent for the 1% porosity
Figure 9. Predicted transmission loss of three pass mufer for different porosities: , s=001; ,
s =003; , s =005.
1000
50
Fr equency (Hz)
T
r
a
n
s
m
i
s
s
i
o
n

l
o
s
s

(
d
B
)
40
30
20
10
200 400 600 800 0
45
35
25
15
5
irrror:1rb 1inr siirNcrrs 445
Figure 10. Transmission loss of two pass cross-ow mufer: , experiment; --W-, model.
case, which shows a noticeable shift of the transmission loss minimum near 825 Hz,
followed by a secondary peak much sharper and of greater magnitude than the higher
porosity cases.
The second multi-pass geometry considered in the study is a two pass cross ow element.
To create this conguration, the three pass muer was modied by removing the pass tube,
sealing the holes in the baes where the pass tube was removed, and sealing the ends of
the inlet and exit ducts. These changes allow the inlet and exit ducts to communicate over
the central cavity, and remove the end chambers entirely. Though this created a simpler
silencer, asymmetry of the structure actually increased somewhat, so multidimensional
eects should be expected to occur at or slightly below the limiting frequency of the rst
muer. Figure 10 compares the model predictions with experimental results for this
conguration. Multidimensional eects are seen to become signicant near 900 Hz, below
which the behaviour shows the repeating broadband domes of a simple expansion
chamber. Assuming the perforated tubes to be transparent, the chamber has an expansion
ratio of m=1127 and length l
e
=255 cm. Therefore the domes should have a maximum
value of 10 log
10
(1 +025(m1/m)
2
) =151 dB, and repeat at intervals of Df =c
0
/
2l
e
=674 Hz, which are very close to the data for the rst dome in Figure 10. Since the
model assumes one-dimensional uid motion, it continues to predict an expansion
chamber behavior beyond 900 Hz, as expected.
4. CONCLUDING REMARKS
A one-dimensional, nite dierence numerical solution of the fundamental balance
equations for unsteady, compressible ow has been shown to accurately predict the
attenuation of sound in multiple pass perforated tube silencer elements. This method
avoids many of the restrictions that are inherent in frequency domain techniques. Although
the present study considers acoustic disturbances in a quiescent medium only, the method
is capable of treating high sound pressures and complex ow elds, including shock waves,
provided that the one-dimensional assumption is justied and a suitable model for
perforate ow is incorporated. Although diametral modes will not continuously propagate
along a duct at frequencies below pd/l =1841, and this expression is commonly used to
determine the upper frequency limit of one-dimensional approaches, it has been shown that
N. s. bicir. r1 :i. 446
multidimensional behavior may become signicant at slightly lower frequencies for fairly
complex perforated tube systems. Since the geometries considered are more representative
of exhaust system silencers, a 900 Hz upper frequency limit determined at atmospheric
conditions is rather conservative. For example, scaling the upper limit with respect to a
speed of sound of approximately 450 m/s in the exhaust gas (based on a gas temperature
of 500 K, roughly corresponding to the muer conditions at 1000 r.p.m. in Figure 2),
shows that the one-dimensional assumption is valid for modelling the above muers to
frequencies of about 1180 Hz. This limit is high enough to include the ring fundamental
and rst few harmonics of most engines even at relatively high speeds (for example, a
six-cylinder engine running at 5000 r.p.m. has a ring fundamental of 250 Hz). A number
of studies are available on the analysis of ow through orices under various conditions,
including high sound pressure levels, through-ow and grazing ow [18, 23]. Work is
currently in progress to incorporate these factors in the present perforate model. The
non-linear, time-domain approach is expected to be useful for a wide range of applications,
including complex geometries, non-linear ow phenomena and entire engine intake and
exhaust systems.
rrrrrrNcrs
1. J. W. Siiiiv:N and M. J. Crocirr 1978 Journal of the Acoustical Society of America 64,
207215. Analysis of concentric-tube resonators having unpartitioned cavities.
2. J. W. Siiiiv:N 1979 Journal of the Acoustical Society of America 66, 772788. A method for
modeling perforated tube muer components. I. Theory. II. Applications.
3. K. J:.:r:x:N and K. Y:x 1981 Journal of the Acoustical Society of America 69, 390396.
Decoupling approach to modeling perforated tube muer components.
4. M. L. MiN:i, K. N. R:o and A. D. S:n:sr:nibnr 1987 Journal of Sound and Vibration 114,
173188. Aeroacoustic analysis of perforated muer components.
5. K. S. Pr:1 1988 Journal of Sound and Vibration 127, 123132. The acoustical impedance at
discontinuities of ducts in the presence of a mean ow.
6. I. J. Cn:Nc and A. CixxiNcs 1988 Journal of Sound and Vibration 122, 243259. A time domain
solution for the attenuation, at high amplitudes, of perforated tube silencers and comparison
with experiment.
7. A. Sri:xr1, S. H. YoN:i, J. M. Nov:i and M. Kn:N 1995 SAE 951260. The eect of vehicle
exhaust system components on ow losses and noise in ring spark-ignition engines.
8. V. H. Gii1: and M. L. MiN:i 1992 Journal of the Acoustical Society of America 92,
27162725. On numerical prediction of the acoustic source characteristics of an engine exhaust
system.
9. G. P. Bi:ir and S. W. Co:1rs 1973 SAE 730160. Noise produced by the unsteady exhaust eux
from an internal combustion engine.
10. G. P. Bi:ir and J. H. McCoNNrii 1974 SAE 740736. Unsteady gas ow through
high-specic-output 4-stroke cycle engines.
11. M. Cn:ix:N, J. M. Nov:i and R. A. S1riN 1982 In Flows in Internal Combustion Engines.
(editor, Teoman Uzkan). Numerical modeling of inlet and exhaust ows in multi-cylinder
internal combustion engines. ASME WAM, Austin, TX.
12. T. Morri, M. F. FirxxiNc and L. A. L:PoiN1r 1990 SAE 900679. Characterization of manifold
dynamics in the Chrysler 2.2 s.i. engine by measurements and simulation.
13. T. Morri, J. Morri and D. A. Bi:srr 1991 SAE 910072. Fluid-dynamic and acoustic modelling
of concetric-tube resonators/silencers.
14. A. Sri:xr1, N. S. Dicir. and J. M. Nov:i 1993 In Proceedings of NOISE-CON 93 (editor,
H. H. Hubbard) pp 291296. A time domain based computational approach for perforated tube
silencers.
15. A. Sri:xr1, N. S. Dicir. and J. M. Nov:i 1994 Journal of the Acoustical Society of America
96, 31773185. The HerschelQuincke tube: A theoretical, computational, and experimental
investigation.
16. A. CixxiNcs 1986 Journal of the Acoustical Society of America 79, 942951. Transient and
multiple frequency sound transmission through perforated plates at high amplitude.
irrror:1rb 1inr siirNcrrs 447
17. A. L. Goibx:N and R. L. P:N1oN 1976 Journal of the Acoustical Society of America 60,
13971404. Measurement of the acoustic impedance of an orice under a turbulent boundary
layer.
18. T. H. MriiiNc 1973 Journal of Sound and Vibration 29, 165. The acoustic impedance of
perforates at medium and high sound pressure levels.
19. U. INc:rb and H. IsiNc 1967 Journal of the Acoustical Society of America 42, 617. Acoustic
nonlinearity of an orice.
20. E. Frbrr and L. W. Dr:N III 1969. NASA CR-1373. Analytical and experimental studies for
predicting noise attenuation in acoustically treated ducts for turbofan engines.
21. K. S. Pr:1 1988 Journal of Sound and Vibration 123, 199212. A numerical decoupling analysis
of perforated pipe silencer elements.
22. U. INc:rb 1953 Journal of the Acoustical Society of America 25, 10371061. On the theory and
design of acoustic resonators.
23. D. RoNNrnrrcrr 1972 Journal of Sound and Vibration 24, 133150. The acoustical impedance
of holes in the wall of ow ducts.
NOMENCLATURE
A area
c speed of sound
CS control surface
CV control volume
d diameter
e specic internal energy
f frequency
k = 2pf/c0, wavenumber
l length
m expansion ratio
M mass
p pressure
q heat transfer rate
R perforate ow resistance
t time
T temperature
U velocity
V volume
x rectangular co-ordinate
y rectangular co-ordinate
Z impedance
Greek Symbols
g ratio of specic heats
d end correction
D difference
l wavelength
r density
s porosity
t shear stress tensor
Subscripts
e expansion chamber
eq equivalent
i inner
f frequency
max maximum
N. s. bicir. r1 :i. 448
o outer
0 mean value
p perforate
ref reference
V volume
y y-direction
Superscripts
n timestep