Applied Acoustics 105 (2016) 156–163

Contents lists available at ScienceDirect

Applied Acoustics
journal homepage: www.elsevier.com/locate/apacoust

Influence of boundary restraint on sound attenuation performance of a

duct-membrane silencer
Jingtao Du ⇑, Yang Liu, Yufei Zhang
College of Power and Energy Engineering, Harbin Engineering University, Harbin 150001, PR China

a r t i c l e i n f o a b s t r a c t

Article history: Low frequency noise in duct is a challenge for the traditional passive noise control techniques. Recently, a
Received 26 February 2015 so-called duct-membrane silencer has attracted much research attention due to its simple configuration
Received in revised form 19 September 2015 and potential application, however, the current studies are merely limited to the cases in which just the
Accepted 30 November 2015
classical boundary conditions are considered. Actually, as an important factor affecting the modal char-
acteristics of the membrane, and the existing studies are not enough to fully understand the vibro-
acoustic characteristics of such silencer with complicated boundary conditions. Motivated by this, in this
Keywords:
paper, the structural–acoustic coupling model of duct-membrane system is established by a modified
Silencer
Membrane
Fourier series method in combination with Rayleigh–Ritz procedure, in which the transverse elastic
Duct acoustics boundary restraints are taken into account. Energy principle is formulated for the vibro-acoustic coupling
Boundary restraints of such duct-membrane silencer to obtain the system matrix equation. Numerical results are then pre-
sented to validate the proposed model, and the influence of boundary restraining stiffness on sound
attenuation performance is also studied. To the best of authors’ knowledge, this work represents the first
time that the elastic boundary restraints have been considered for such duct-membrane silencing system.
Ó 2015 Elsevier Ltd. All rights reserved.

1. Introduction works well for the medium-to-high-frequency range. On the other

The problem of low frequency noise control in duct, such as
below 200 Hz, is often encountered in many engineering occasions,
such as air conditioning and ventilation system [1]. Based on
whether the secondary external energy is consumed or not, the
noise control techniques can be categorized into two types: active
and passive noise control. For the first one, it usually includes error
sensor and secondary actuator, as well as the electronic controller
[2]. Such system is effective for the low frequency noise control,
however, it is believed that the introduction of these electronic
devices will make the system complicated and expensive, at some
circumstance, there will even be stability and reliability issues.
Then, the realization of low frequency noise control through the
passive means will be much more reliable and attractive.
For the traditional passive noise control, in essence, it attenu-
ates the sound propagation in duct through dissipating the incom-
ing sound into heat or the impedance mismatch. The first one is
also called resistance silencer, where the sound absorbing material
such as fiber is lined in the duct. The fiber damping effect will dis-
sipate the sound energy into heat. But this kind of silencer just
⇑ Corresponding author. Tel.: +86 451 82589199 307; fax: +86 451 82589199
311.
E-mail address: dujingtao@hrbeu.edu.cn (J. Du).
hand, the impedance silencer is much more effective in low fre-
quency, such as the expansion-chamber-type muffler. But, as sum-
marized by Huang [3], it also has some problems, such as the
existence of passbands, the bulkiness and the aerodynamic loss.
So, low frequency duct noise is very difficult to deal with by tradi-
tional passive technique.
In recent years, Huang [3] proposed a novel duct-membrane
silencer, which is based on the structural–acoustic coupling to
achieve low frequency noise attenuation in duct. The tensioned
membrane with clamped boundary condition is used as segment
of the duct wall to reflect the grazing incident noise. Since there
is no need of lining material and/or cross section variation like
other traditional passive silencer does, such duct-membrane silen-
cer is environmentally friendly, and has good flow-through charac-
teristics. His investigation showed that the combination of noise
reflection and damping of slow flexural waves can be very power-
ful in a broad frequency band. Subsequently, Huang theoretically
and experimentally studied the membrane to be further backed
by rigid-walled cavities, as should be the case in a practical silencer
design, to form the so-called ‘drum-like’ or ‘drum silencer’ [4–6],
which can achieve a satisfactory attenuation from low frequency
to medium over an octave band. Huang and Choy [7] further
extended the study into three-dimensional model in which the

http://dx.doi.org/10.1016/j.apacoust.2015.11.020
0003-682X/Ó 2015 Elsevier Ltd. All rights reserved.
 J. Du et al. / Applied Acoustics 105 (2016) 156–163 157

membrane has all four edges fixed. They found that the lateral ten-
sion was always detrimental to the silencing performance. Simi- Pi Pd
larly, the silencer model with its segment of rigid duct replaced duct h
y
by a cavity-backed panel is also analyzed by Huang, in which both L
simply supported and clamped boundary conditions are investi- O x
membrane
gated and compared [8,9]. The results show that, for a uniform k1 k2
plate, the optimal stop band is narrower than that of the simply
supported configuration. Moreover, Choy and Huang [10] also
Fig. 1. A duct-membrane silencer with elastic boundary restraints.
explored the effect of flow on the drum-like silencer, their experi-
mental results indicate that for the flow speed common in ventila-
tion applications, there is no flow-induced flexural instability, and the familiar fixed boundary condition will be obtained. When med-
the possible noise radiated by the turbulence-induced vibration is ian value is set, it will be actually the elastic boundary restraint.
insignificant.
In the aforementioned studies, most of them are confined to the 2.2. Energy formulation of the duct-membrane coupling system
classical boundary conditions for the flexible structure, such as
simply supported and clamped. In fact, as one type of structural For the vibro-acoustic coupling system illustrated in Fig. 1,
factors, boundary condition has an important effect on the modal mathematically, there will be two ways for the problem descrip-
characteristics of flexible structure as well as its vibro-acoustic tion. For the first one, the governing differential equations and
coupling system. As found by Wang et al. [9], the boundary condi- boundary/coupling conditions are simultaneously used to
tions also has significant influence on the silencing performance. formulate the whole coupled system. The other one is the integral
From the practical point of view, the flexible structure will be form, namely energy formulation. When the admissible function is
not absolutely restrained in the ideal cases, while they are actually constructed smooth sufficiently, the later solution will be
elastically restrained. On the other hand, a better understanding on equivalent to the first one. In this work, the energy formulation
the boundary restraint effect will be helpful for the optimal design is employed for the structural–acoustic description of duct-
of such silencer. membrane system. The inherent benefit for such energy formula-
Motivated by the limitation in literature, a two-dimensional tion is that the current model can be easily extended to include
duct-membrane coupling model is established, in which the tensed multiple vibro-acoustic subsystems.
membrane is elastically restrained in transverse direction at both The system Lagrangian for the current duct-membrane coupling
ends. Different with the current studies that derive the solution model can be written as
from the differential governing equation, the energy principle is
L ¼ U  T þ WP ð1Þ
formulated for the whole vibro-acoustic coupling system [11,12],
with the sound radiation process taken into account in terms of in which, U and T are respectively the total potential energy and
the work done for the vibrating membrane. An improved Fourier total kinetic energy associated with the transverse vibration of ten-
series method previously developed for the in-plane vibration sile string. WP is the work done by sound pressure across the upper
analysis of rectangular plate [13] is employed for the construction surface of string, and it represents the structural–acoustic interac-
of admissible function, with all unknown expansions coefficients tion between the subsystems of duct and string.
determined via Rayleigh–Ritz procedure. Numerical results are For the total potential energy U, it actually includes two parts,
then presented to validate the proposed model by comparing the namely, strain energy due to transverse deformation and the elas-
calculated results with those from other approach in literature. tic potential energy stored in the elastic boundary restraints. It can
The influence of boundary restraints on the silencing performance be expressed as
of duct-membrane system is studied and discussed based on the Z  2
developed model. 1 L
@u 1
U¼ F dx þ ½k1 u2 ð0Þ þ k2 u2 ðLÞ ð2Þ
2 0 @x 2
here, F is the horizontal tensing force applied to the both ends of
2. Theoretical formulation
string.
The total kinetic energy associated with the string transverse
2.1. Model description
vibration is
Z  2
Let us consider a duct-membrane silencer, with its cross section 1 L
@uðx; tÞ
T¼ q dx ð3Þ
along the longitudinal direction shown in Fig. 1, in which a seg- 2 0 @t
ment of duct is covered by the tensing membrane with elastic
boundary restraint. For the two-dimensional duct-membrane cou- in which, q is the string mass density of unit length.
pling system studied in this work, the string model will be During the calculation of work done by the sound pressure WP,
employed accordingly. The coordinate system is also presented, the pressure across the upper surface of the string is the superpo-
in which the origin O is located on the left end of the membrane. sition of two parts, namely
L is the membrane length, and h is the duct height. In the duct P ¼ Pi þ P rad ð4Þ
upstream, there is an incident plane sound wave defined as
here, Pi is the sound pressure propagating from the duct upstream,
pi ¼ eik0 xþixt , in which x = 2pf is the angular frequency of the inci-
and Prad is the sound radiation into the duct due to the string flex-
dent sound wave, and k0 is equal to x/c0 with c0 as the sound
ural vibration.
speed. For the 1-D membrane, elastic boundary springs are
The radiated sound pressure from the vibrating structure in
assumed to restrain the transverse vibration of string at both ends.
duct can be predicted by using the following formula [14]:
k1 and k2 respectively denote the spring stiffness coefficients on
Z
the left and right end of the string. Any classical boundary condi- q0 X
1 L 
wm ðy0 Þv ðx0 ; y0 Þ  Hðx  x0 Þeikm ðxx Þ
0
Prad ðx; yÞ ¼ cm wm ðyÞ
tion can be easily realized by setting the stiffness coefficient into 2h m¼0 0
infinity or zero. For example, when both two spring stiffnesses 0  0
are taken as infinity (a very large number in numerical calculation), þHðx0  xÞeþikm ðxx Þ dx ð5Þ
158 J. Du et al. / Applied Acoustics 105 (2016) 156–163

in which, v(x0 , y0 ) is the velocity distribution across the vibrating For the vibrating string, different with beam or rod, the struc-
string, and the coordinate system (x0 , y0 ) is the source point position, tural flexibility associated with the transverse vibration is gener-
since for the radiated sound pressure calculation, the integration is ated by the tensing force applied at its both ends. This is also
carried out on the string surface, namely, y0 = 0, and x0 varies from 0 reflected in the boundary equilibrium equations for the elastic
to L. Another coordinate system (x, y) represents the observing point restraint, namely
position in the duct. H is the Heaviside function, m is the modal
order number of the duct. When m is equal to 0, it is then 0 order @uðx; tÞ
F ¼ k1 uð0; tÞ; at x ¼ 0 ð11Þ
plane wave; for the case of m > 1, it will be higher order wave. @x x¼0

When the excitation frequency is lower than the first cut-on fre- @uðx; tÞ
F ¼ k2 uðL; tÞ; at x ¼ L ð12Þ
quency of the duct c0/2h, one will have @x x¼L
ic0 here, the k1 and k2 are the elastic boundary spring stiffnesses along
cm ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ; ð6Þ
2 the transverse direction at each end.
ðmp=k0 hÞ  1
x As we can see that, when the standard Fourier series Eq. (10) is
km ¼ ; ð7Þ submitted into the elastic boundary conditions Eqs. (11) and (12),
cm
the boundary discontinuity issue will be encountered, and it will
and significantly deteriorate the solution convergence and accuracy.
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi mp  In order to overcome this issue, the following improved version
wm ðyÞ ¼ 2  d0m cos y : ð8Þ is constructed for the string flexural displacement:
h
here, d is the Kronenic delta function. X
1
uðxÞ ¼ an cos kn x þ b1 f1 ðxÞ þ b2 f2 ðxÞ ð13Þ
The corresponding work done by sound pressure on upper sur-
n¼0
face of the vibrating string can be written as
Z L Z L in which, f1(x) and f2(x) are two supplementary functions, with b1
1
WP ¼ Pi udx þ Prad udx ð9Þ and b2 being the corresponding weight coefficients. Here, these
0 0 2 two function are chosen as follows:
in which, u is the transverse displacement of string. It should be x 2 x2 x 
noted that the radiated sound pressure Prad in Eq. (9) is actually f1 ðxÞ ¼ x 1 ; f2 ðxÞ ¼ 1 ð14a; bÞ
the function of vibration response of the string Eq. (5), which means L L L
the vibration and acoustic sub-systems are well coupled together. The purpose of introducing these two auxiliary polynomials
Such full vibro-acoustic coupling will be also reflected in the follow- into the standard Fourier series is to remove the differential dis-
ing system matrix equation. continuities at both ends of string. To better illustrate this, Fig. 2
is presented to show the function distribution of the auxiliary poly-
2.3. Improved Fourier series solution of the coupled system nomials as well as their first order derivatives. From this Figure, it
is easy to observe that:
Once the energy description of duct-membrane system is for-
mulated, the remaining task is to construct the appropriate admis- f1 ð0Þ ¼ f1 ðLÞ ¼ f01 ðLÞ ¼ 0; f01 ð0Þ ¼ 1 ð15a; bÞ
sible function, and then all the unknown coefficients can be found
via Rayleigh–Ritz procedure. For the classical fixed boundary con- f2 ð0Þ ¼ f2 ðLÞ ¼ f02 ð0Þ ¼ 0; f02 ðLÞ ¼ 1 ð16a; bÞ
dition, the transverse displacement of vibrating string is usually
For this improved Fourier series expression of string vibrating
expanded as the Fourier series, namely
displacement, the boundary differential discontinuities associated
X
1 with the traditional Fourier series can be removed successfully,
uðxÞ ¼ an cos kn x ð10Þ and it is now sufficiently smooth in the entire solving interval from
n¼0
0 to L including both two end points.
here, an is the Fourier expansion coefficient, and kn is defined as Substituting Eq. (13) into Eq. (1), and minimizing the Lagran-
np/L. gian with respect to the all unknown coefficients, one is able to

1 1
ζ 1(x) ζ 2(x)
0.8 0.8
ζ '1(x) ζ '2(x)
0.6 0.6

0.4 0.4

0.2 0.2

0 0

-0.2 -0.2

-0.4 -0.4
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
x/L x/L

Fig. 2. The graphs of two supplement functions f1(x) and f2(x), as well as their first order derivatives f0 1(x) and f0 2(x).
 J. Du et al. / Applied Acoustics 105 (2016) 156–163 159

4.5 160
1 2 3 4 5 6 7 8 9 10 11
4 current
Ref.[3] 150
3.5

Vibration velocity (dB)

3 140

2.5
|v(x)|

130
2

1.5 120

1
110
0.5

0 100
0 0.1 0.2 0.3 0.4 0.5 0 50 100 150 200 250 300 350 400
x(m) Frequency (Hz)

Fig. 3. Vibration velocity distribution across the string under the upstream plane Fig. 4. Vibration velocity frequency response of the string at the point x = 13L/30.
sound wave excitation at f = 340 Hz. 2 3
2 X
km L4 ðcos np  eikm L Þðcos np þ cos n0 pÞ ikm L2 Dðn0 ; nÞ5
M
fP11 gnþ1;n0 þ1 ¼ 24     þ ;
2
m¼0
2 2
n02 p2  km L2 n2 p2  km L2 n 2 p 2  km L 2
yield a set of linear equations which can be written in a matrix
form as ð25Þ
8
ðK  x2 M þ ixPÞC ¼ Pi ð17Þ 0
<0 n – n
>
0
in which, Dðn ; nÞ ¼ 2 n0 ¼ n – 0 ;
L
ð26Þ
>
:
T L n0 ¼ n ¼ 0
C ¼ fa0 ; a1 ; a2 ; . . . ; an ; b1 ; b2 g ; ð18Þ

 and
K11 K12
K¼ ; ð19Þ ik0 L2 ð1  eikL cos npÞ
K21 K22 fPi1 gnþ1;1 ¼ 2
: ð27Þ
n2 p2  k0 L2


M11 M12
M¼ ; ð20Þ Once all the matrices and the load vector are calculated, the
M21 M22 Fourier series coefficient vector C can be easily obtained as

 1
P11 P12 C ¼ ðK  x2 M þ ixPÞ Pi ð28Þ
P¼ ; ð21Þ
P21 P22 With the solved expansion coefficients, the transverse vibration
and displacement at any point across string is directly determined from

 Eq. (13).
Pi1
Pi ¼ : ð22Þ
Pi2 2.4. Sound response of the far-field downstream in duct
where K and M are the stiffness and mass matrix of the elastically
restrained string system in vacuo, respectively; P is the contribution For the downstream sound response in duct, it actually includes
matrix of radiated sound from string in duct environment, Pi is two parts, one is the incident upstream sound wave, and the other
external load vector caused by duct upstream sound pressure, C is is sound wave radiated from the vibrating string into the duct. For
a vector that contains all the unknown Fourier expansion coeffi- the case of that the excitation frequency is much lower than the
cients in Eq. (13). critical frequency of duct, the radiated sound wave that propagates
In numerical calculations, all the Fourier series in Eq. (17), will into the far-filed downstream along the duct will be just the zero
have to be truncated to m = M, n = n0 = N. The following representa- order wave, namely the plane sound wave. Then, the sound radia-
tive expressions are given to illustrate how to form the matrices tion field in the far-field downstream can be written in the follow-
used in Eq. (17). ing simplified form.
Z Z L
L q0 c 0
fK11 gnþ1;n0 þ1 ¼ Fkn kn0 sin kn0 x sin kn xdx þ k1 þ k2 ð1Þnþn ;
0
ð23Þ Prad ðx; yÞjx!þ1 ¼ eik0 x v ðx0 Þeik x dx0
0
0
ð29Þ
0
2h 0

Z L
In order to characterize the sound attenuation performance of
fM11 gnþ1;n0 þ1 ¼ q cos kn0 x cos kn xdx; ð24Þ such 2-D duct-membrane silencer, the transmission loss is used
0 with its definition as

Table 1
Comparison of the coupled resonant frequencies with those in vacuo.

Cases Resonant frequency (Hz)

1 2 3 4 5 6 7 8 9 10
In-vacuo 34 68 102 136 170 204 238 272 306 340
Coupled 27.03 56.57 88.61 120.6 153.2 184.7 218.3 253.3 285.9 319.9
160 J. Du et al. / Applied Acoustics 105 (2016) 156–163

Z
Pi þ Prad L
TL ¼ 20log10 0
0
ð30Þ fR2 g1;1 ¼ f1 ðx0 Þeik0 x dx ; ð34Þ
Pi 0

Substituting Eq. (29) into Eq. (30), one will obtain

and
Z
q c0 L
0
TL ¼ 20log10 1 þ 0 v ðx Þe
0 ik0 x0
dx ð31Þ Z L
2h 0 0
0
fR3 g1;1 ¼ f2 ðx0 Þeik0 x dx : ð35Þ
0
With the assumption of harmonic oscillation, the vibrating
string velocity can be calculated from displacement field informa- In this work, as a primary study, the 2-D theoretical model is
tion with the relationship of v(x) = ixu(x), namely established for investigating the boundary restraint influence on
q c0 the sound attenuation performance of such silencer. However, it

TL ¼ 20log10 1 þ ix 0 fR1 ; R2 ; R3 gC ð32Þ should be noted that the current method can be extended to the
2h
3-D case straightforwardly. For example, a segment of three-
in which, dimensional rectangular duct is replaced by a elastically restrained
Z L flexible panel backed by acoustical cavity. Then, the 2-D and 3-D
0 0
fR1 g1;nþ1 ¼ cos kn x0 eik0 x dx ; ð33Þ improved Fourier series can be constructed for the panel vibration
0
and sound field inside the cavity to consider the boundary and

Mode
The first four in-vacuo modes The first four coupled modes
order
0.15 0.15

0.1 0.1
Displacement mode

0.05 0.05

0 0
1
-0.05 -0.05

-0.1 -0.1

0 0.1 0.2 0.3 0.4 0.5 0 0.1 0.2 0.3 0.4 0.5
x (m) x (m)
0.15 0.15

0.1 0.1
Displacement mode

0.05 0.05

0 0
2
-0.05 -0.05

-0.1 -0.1

0 0.1 0.2 0.3 0.4 0.5 0 0.1 0.2 0.3 0.4 0.5
x (m) x (m)
0.15 0.15

0.1 0.1
Displacement mode

0.05 0.05

0 0
3
-0.05 -0.05

-0.1 -0.1

0 0.1 0.2 0.3 0.4 0.5 0 0.1 0.2 0.3 0.4 0.5
x (m) x (m)
0.15 0.15

0.1 0.1
Displacement mode

0.05 0.05

0 0
4
-0.05 -0.05

-0.1 -0.1

0 0.1 0.2 0.3 0.4 0.5 0 0.1 0.2 0.3 0.4 0.5
x (m) x (m)

Fig. 5. Comparison of the first four coupled mode shapes with those in vacuo.
 J. Du et al. / Applied Acoustics 105 (2016) 156–163 161

pressure–velocity continuities across the panel edge and struc-
tural–acoustic coupling interface. In order to take the cavity effect
into account, the sound energy and work done on the interface will
be further included in the coupling system Lagrangian.
3. Numerical results and discussion
In this section, several numerical examples will be presented to
verify the proposed model and analyze the influence of elastic
boundary restraint on sound attenuation of such duct-membrane
system. In order to validate the proposed model, the results will
be first compared with those predicted by Huang using other
approach [3], and the simulation model parameters are kept the
same as those used in his former study, namely, duct height
h = 0.1 m, string length L = 5h, string mass density per unit length
q = 0.05 kg/m, tensing force in string F = 57.8 N/m, air mass density
q0 = 1.255 kg/m3. In the numerical simulation, wave mode m = 20
is used in the calculation of sound radiation into duct from string,
and the improved Fourier series is truncated at n = n0 = 60.
Fig. 3 shows the comparison of vibration velocity distribution
on the string calculated from these two methods, and the excita-
tion frequency of upstream plane wave is f = 340 Hz with unit
sound pressure amplitude. It can be seen that the velocity distribu-
tion is not uniform across the whole string, the fluctuation implies
that modal behavior of string should play a important role in such
vibro-acoustic coupling phenomenon. It can be also observed that
the agreement between these two results are excellent, the cor-
rectness and reliability of the proposed model are then verified.
Moreover, in the current model, the fixed boundary condition of
string can be easily achieved by setting the transverse restraining
spring coefficients as k1 = k2 = 1  1012 N/m in simulation
attention to the influence of elastic boundary restraints on the
transmission loss of such duct-string system. In current study,
the most general case of boundary condition for flexible string
has been considered, any boundary restraint can be easily obtained
by setting restraining stiffness accordingly. Compared with other
approaches in literature, there is no need to re-formulate the the-
oretical equation or modify the simulation code when the bound-
ary condition is changed. This model will be more suitable and
convenient for the study of boundary restraints effect on the
silencing performance.
Fig. 7 demonstrates the transmission loss of duct-string cou-
pling system under various boundary restraining stiffnesses. In
Fig. 7(a), the left end of string is elastically restrained with different
spring stiffnesses, namely k1 = 101, 103, 105 and 107 N/m, while the
right end is still fixed with the k2 = 1012 N/m representing the
infinity. It can be observed that the boundary restraint has a signif-
icant influence on transmission loss. The releasing process is not
uniform, at the beginning, the effect is not obvious. It means that
there is a sensitive zone for the boundary restraining stiffness,
when the value is smaller than k1 = 105 N/m, there will be a great
enhancement for the transmission loss of such duct-string cou-
pling system. For Fig. 7(b), the results is for the variation of right
hand spring stiffness with the left end fixed. It can be seen that
the curves are almost the same as those in Fig. 7(a), which implies
50
Current
45
Ref.[ 3 ]
40
Transmission loss (dB)

computation. 35
Fig. 4 presents the vibration velocity of the string at point
30
x = 13L/30 within frequency range from 0 to 400 Hz. The vertical
dot lines in this figure show the resonant frequencies of the same 25
string in vacuo. It can be found that the coupled resonant peaks
20
shift to some extent comparing with those in vacuo. Tabulated in
Table 1 is the corresponding resonant frequency comparison 15
between these two cases, while the displacement mode shapes
10
are also presented in Fig. 5. For the modal information of coupled
system, the resonant frequencies are picked up from the resonant 5
peaks of frequency response curve, and the mode shape is derived (a)
0
by calculating the displacement response distribution under exter- 0 200 400 600 800 1000
nal excitation at these resonant frequencies. For the string sub- Frequency (Hz)
system of coupled system, it can be found that certain variation
for the mode shape is occurred due to the vibro-acoustic coupling 50
between the duct and string. Both the mode order and mode shape 45 Current
are modified significantly, and the amplitude of vibration mode on Ref.[ 3 ]
each side are much smaller than those at the middle of string. 40
Transmission loss (dB)

Plotted in Fig. 6 are the transmission loss of the duct-string sys-

35
tem, while the Fig. 6(a) is for the single string case, and the Fig. 6(b)
is the transmission loss for the case of double strings. The results 30
from other approach presented in Ref. [3] are also given for com-
25
parison. In the current model, the transmission loss for double
strings system can be easily treated in the string Lagrangian 20
through introducing the terms of mutual work done by sound radi-
15
ation from each string. From the figure, it can be seen again that
the transmission loss prediction compares well with those 10
obtained in Ref. [3]. Then, the reliability of the proposed model
5
has been validated again. From these two figures, it can be obvi- (b)
ously seen that the introduction of another string into the duct 0
0 200 400 600 800 1000 1200
on the top surface will effectively widen the frequency range of
Frequency (Hz)
transmission loss for such coupled duct-string system.
The aforementioned study is carried out for the case of string Fig. 6. Transmission loss of the duct-string coupling system: (a) single string case;
with classical fixed boundary condition, now let us move our (b) double strings case.
162 J. Du et al. / Applied Acoustics 105 (2016) 156–163

that the influence of these two symmetric boundary restraints are k1 = k2. As we can observe that such case will lead a much larger
the same for the transmission loss. Presented in Fig. 7(c) is the noise reduction comparing the above two cases in which the
influence for the elastic boundary restraints on both side with boundary restraint is released singly.
To get a whole picture of boundary condition effect on the
transmission loss, the contours of varying spring stiffness continu-
ously are given in Fig. 8. Due to the above symmetric reason, here
70 just the transmission loss with one end elastically restrained and
the other fixed is given in Fig. 8(a). The result corresponding to
k1=101N/m
the simultaneous variation of both ends with equal restraining
60
k1=103N/m stiffness is demonstrated in Fig. 8(b). From these two, it can be
k1=105N/m obviously seen that how the boundary restraint affect the noise
Transmission loss (dB)

50 reduction performance and where the sensitivity zone is. From

k1=107N/m
the optimal standpoint of view, proper selecting the boundary
40 restraint stiffness will be much helpful to achieve a better trans-
mission loss for such duct-membrane silencer.
30 From the practical point of view, the flexible connection at the
boundaries may be much easier for the plate structure, which is
usually clamped by bolts. When the bolts is tightened with less
20
k2=1012N/m sufficient strength, then the boundary condition is not absolutely
clamped, while it is elastically restrained with finite supporting
10 stiffness. On the other hand, in the future, the controlled boundary
(a) supporting stiffness apparatus may be developed based on the
0 smart material such as magnetorheological elastomer.
0 200 400 600 800 1000
Frequency (Hz)

70

k2=101N/m
60
k2=103N/m
k2=105N/m
Transmission loss (dB)

50
k2=107N/m

40

30

20 k1=1012N/m

10
(b)
0
0 200 400 600 800 1000
Frequency (Hz)

70

k1=k2=101N/m
60
k1=k2=103N/m
k1=k2=105N/m
Transmission loss (dB)

50
k1=k2=107N/m
40

30

20

10
(c)
0
0 200 400 600 800 1000
Frequency (Hz)

Fig. 7. Influence of boundary restraining stiffness on the transmission loss of duct-

string coupling system: (a) variation of k1 with k2 = 1; (b) variation of k2 with Fig. 8. Contour of transmission loss under continuously varying boundary restrain-
k1 = 1; (c) variation of k1 = k2. ing stiffness of the flexible string.
 J. Du et al. / Applied Acoustics 105 (2016) 156–163
4. Conclusion
In this paper, a two-dimensional theoretical model for predict-
ing silencing performance of a duct-membrane coupling system is
established, with a special emphasis on the elastic boundary
restraint of the flexible membrane. The energy description is firstly
formulated for the whole vibro-acoustic coupling system, and a
modified Fourier series is constructed for the admissible function
with aim to improve the convergence and accuracy of the series
solution. The coupled system response matrix is further derived
in the combination with Rayleigh–Ritz procedure.
Numerical examples are presented to demonstrate the effec-
tiveness of the current model and study the influence of boundary
restraint on the silencing performance of duct-membrane silencer.
The reliability of the proposed model is validated through the
excellent agreement of response prediction between the current
one and other approach in literature. The vibro-acoustic coupling
phenomenon can be also observed from the significant variation
of resonant frequencies and mode shapes comparing with their
counterparts in vacuo. Based on the validated theoretical model,
the effect of boundary restraint on the silencing performance is
discussed. The results show that there exists a sensitive range for
the restraining stiffness, and setting the leading and trailing edge
solely will generate an equivalent silencing enhancement. Releas-
ing the both two edges will lead to a much larger noise reduction
window. This implies that proper selection of boundary restraining
stiffness will be helpful to obtain a better transmission loss for
such duct-membrane system.
Compared with all the related studies in literature, the theoret-
ical model and the solution method described in the present paper
have the following distinct features: (a) the energy principle
description is firstly established, and the current study can be con-
veniently extended to include other multiple structural/acoustic
sub-systems; (b) the boundary condition is unifiedly considered
the same as other structural parameters, there is no need to refor-
mulate the theoretical model when boundary restraints is changed.
This model will be more suitable for studying the influence of
boundary restraining stiffness on silencing performance of such
163
duct-membrane system. Although the focus of this work is put
on the 2-D model, the current method can be extended to the 3-
D cases straightforwardly.
Acknowledgements
This work was supported by the National Natural Science Foun-
dation of China (Grant no. 11202056) and the Fundamental
Research Funds for the Central Universities (Grant no.
HEUCFQ1402).
References
[1] Munjal ML. Acoustics of ducts and mufflers. New York: Wiley; 1987.
[2] Burgess JC. Active adaptive sound control in a duct: a computer simulation. J
Acoust Soc Am 1981;70(3):715–26.
[3] Huang LX. A theoretical study of duct noise control by flexible panels. J Acoust
Soc Am 1999;106(4):1801–9.
[4] Huang LX. Modal analysis of a drumlike silencer. J Acoust Soc Am 2002;112
(5):2014–25.
[5] Choy YS, Huang LX. Experimental studies of a drumlike silencer. J Acoust Soc
Am 2002;112(5):2026–35.
[6] Huang LX. Parametric study of a drum-like silencer. J Sound Vib
2004;269:467–88.
[7] Huang LX, Choy YS. Vibroacoustics of three-dimensional drum silencer. J
Acoust Soc Am 2005;118(4):2313–20.
[8] Huang LX. Broadband sound reflection by plates covering side-branch cavities
in a duct. J Acoust Soc Am 2006;119(5):2628–38.
[9] Wang CQ, Han J, Huang LX. Optimization of a clamped plate silencer. J Acoust
Soc Am 2007;121(2):949–60.
[10] Choy YS, Huang LX. Effect of flow on the drumlike silencer. J Acoust Soc Am
2005;118(5):3077–85.
[11] Du JT, Li WL, Liu ZG, Xu HA, Ji ZL. Acoustic analysis of a rectangular cavity with
general impedance boundary conditions. J Acoust Soc Am 2011;130
(2):807–17.
[12] Du JT, Li WL, Xu HA, Liu ZG. Vibro-acoustic analysis of a rectangular cavity
bounded by a flexible panel with elastically restrained edges. J Acoust Soc Am
2012;131:2799–810.
[13] Du JT, Li WL, Jin GY, Yang TJ, Liu ZG. An analytical method for the in-plane
vibration analysis of rectangular plates with elastically restrained edges. J
Sound Vib 2007;306:908–27.
[14] Doak PE. Excitation, transmission and radiation of sound from source
distributions in hard-walled ducts of finite length (I): the effects of duct
cross-section geometry and source distribution space-time pattern. J Sound
Vib 1973;31(1):1–72.