Transmission Loss 1

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 6

Plate-type acoustic metamaterial with cavities coupled via an orifice for enhanced

sound transmission loss


Linus Yinn Leng Ang, Yong Khiang Koh, and Heow Pueh Lee

Citation: Appl. Phys. Lett. 112, 051903 (2018);


View online: https://doi.org/10.1063/1.5019602
View Table of Contents: http://aip.scitation.org/toc/apl/112/5
Published by the American Institute of Physics

Articles you may be interested in


High efficiency and broadband acoustic diodes
Applied Physics Letters 112, 051902 (2018); 10.1063/1.5020698

Acoustic metamaterial panel for both fluid passage and broadband soundproofing in the audible frequency range
Applied Physics Letters 112, 041903 (2018); 10.1063/1.5004605

Acoustic perfect absorption and broadband insulation achieved by double-zero metamaterials


Applied Physics Letters 112, 021901 (2018); 10.1063/1.5018180

Asymmetric absorber with multiband and broadband for low-frequency sound


Applied Physics Letters 111, 143502 (2017); 10.1063/1.4998516

Perspective: Acoustic metamaterials in transition


Journal of Applied Physics 123, 090901 (2017); 10.1063/1.5007682

Broadband sound transmission loss of a large-scale membrane-type acoustic metamaterial for low-frequency
noise control
Applied Physics Letters 111, 041903 (2017); 10.1063/1.4995405
APPLIED PHYSICS LETTERS 112, 051903 (2018)

Plate-type acoustic metamaterial with cavities coupled via an orifice


for enhanced sound transmission loss
Linus Yinn Leng Ang,1,2,a) Yong Khiang Koh,2 and Heow Pueh Lee1
1
Department of Mechanical Engineering, National University of Singapore, 9 Engineering Drive 1,
Singapore 117575, Singapore
2
Kinetics Design and Manufacturing, Singapore Technologies Kinetics Ltd., 249 Jalan Boon Lay,
Singapore 619523, Singapore
(Received 14 December 2017; accepted 17 January 2018; published online 29 January 2018)
Membrane-type acoustic metamaterials generally involve a heavy platelet attached to a pretensioned
membrane. Their acoustical performance is characterised solely based on the resonant behaviour of
the membrane-platelet assembly. However, typical designs may pose manufacturing issues if
extended in scale for industrial applications. Examples include the spatial consistency of the plate-
let(s), the uniformity of the membrane pretension, and the durability of the membrane—not to men-
tion stress relaxation. As large-scale designs are recently gaining interest, it is imperative to address
the shortcomings for manufacturability. This study presents a proof-of-concept design using tension-
less membranes without any platelets. The results showed that the acoustical performance could be
complemented by the coupling effect between two enclosed cavities via an orifice. The orifice diame-
ter could serve as a tuning parameter for broadband or narrowband transmission loss at selected fre-
quencies. Consequently, the proposed design could address the shortcomings of membrane-type
acoustic metamaterials and complement their acoustical performance with the additional feature.
Published by AIP Publishing. https://doi.org/10.1063/1.5019602

Membrane-type acoustic metamaterials have been stud- shown that plastic membranes may eliminate pretension and
ied extensively over the past decade.1–11 Their designs gen- provide better durability in contrast to rubber membranes.23,24
erally involve a heavy platelet attached to a pretensioned The acoustical performance, however, is still governed solely
rubber membrane. If properly designed, these metamaterials by the membrane resonances.
may be a potential alternative for noise control in the indus- This study—numerical and experimental—considers a
try.12 For example, Sui et al.13 showed improved noise plate-type acoustic metamaterial showing that the acoustical
reduction by adding a pretensioned latex membrane in a hon- performance could be complemented by the coupling effect
eycomb sandwich panel. This improvement was achieved between two enclosed cavities via an orifice in addition to
without much weight (low material density) and spatial the resonant behaviour of the membranes.
(small thickness) penalties—beneficial in vehicle and aircraft Printed from polylactide filaments via fused deposition
cabins. But careful design is necessary so that the acoustical modelling (MOMENT Moment S), each specimen was assem-
performance of a small-scale specimen is not compromised bled from three components and two Mylar films. Figure 1(a)
by the global dynamic response of a large-scale specimen.14 shows the three-dimensional (3D) rendering of the specimen
Recently, more complex designs have been reported to alongside the dimension labels. A film (dm ¼ 127 lm thick)
improve the acoustical performance. For example, a con- was first adhered on one side of the inner frame by an
straint structure15 could be used to suppress certain eigenmo- industrial-grade double-sided tape (Nitto Denko No. 500). The
des of the membrane or a perforated membrane16 could be tape width was larger than Lo to avoid discontinuity, which
used to introduce the effect of a Helmholtz resonator. could cause air leakages from the cavities. Next, the excess
Alternatively, active controls could also be introduced to membrane was removed by cutting. The same steps were
tune the acoustical performance.16–19 repeated to adhere the membrane on the other side of the inner
The acoustical performance of membrane-type acoustic frame. Finally, an outer frame (3 mm thick) was adhered on
metamaterials depends on the resonant behaviour of the mem- the exposed surface of each membrane. The outer frames
brane with or without the platelet(s). To achieve the desired served to emulate a pinned boundary condition along the mem-
outcome, tuning of the membrane and the platelet parameters branes’ edges and prevent accidental damage to the regions
is necessary. However, typical designs may pose manufactur- adhered to the inner frame. The frame outer Lo and inner Li
ing concerns if extended in scale for industrial applications. lengths were 6 and 4 cm, respectively. The smaller Ls and
Examples include the uniformity of the membrane pretension, larger Ll cavities had widths of 1 and 2.5 cm, respectively. The
the spatial consistency of the platelet(s), and the durability of partition thickness tp was 5 mm. The orifice was situated at the
the membrane—not to mention stress relaxation. With recent centre point of the partition. Several diameters were investi-
interest in large-scale designs,20–23 it is imperative to address gated (D ¼ 0, 3, 5, and 7 mm). Note that D ¼ 0 mm served as
the shortcomings for manufacturability. Recent studies have the reference specimen to identify the orifice’s effect. Overall,
the specimen had a total depth of d ¼ 20 þ 2 dm mm. The
a)
Electronic addresses: ylang@u.nus.edu and anglinus@stengg.com specimen had a surface density of 2.54 kg/m2 when D ¼ 0 mm,

0003-6951/2018/112(5)/051903/5/$30.00 112, 051903-1 Published by AIP Publishing.


051903-2 Ang, Koh, and Lee Appl. Phys. Lett. 112, 051903 (2018)

FIG. 1. (a) 3D rendering of the specimen alongside the dimension labels. (b) Set-up of the specimen and the holder in the impedance tube (viewed from the
downstream side). (c) 3D numerical model of the impedance tube. The inset shows the holder and the specimen with the first membrane and the interior fluid
domain removed for clarity.

2.51 kg/m2 when D ¼ 3 mm, 2.46 kg/m2 when D ¼ 5 mm, and depending on the measured frequency range. Numerically,
2.39 kg/m2 when D ¼ 7 mm. this distance was found to have a negligible influence on the
The experiment was conducted based on ASTM results. Hence, the distance was kept at 8 cm in the numerical
E2611.25 The specimen was mounted in the impedance tube model. Table I lists the material properties specified in the
(BSWA SW422) by an in-house designed holder. The holder simulation. All values were taken from either a handbook27
was made from a highly dense material (stainless steel) to or the manufacturer’s datasheet except for the structural loss
ensure that its dynamic behaviour would not influence the factor of Mylar, which was measured via dynamic mechani-
results in the frequency range of interest (0.1–1.6 kHz). Two cal analysis by Singapore Institute of Manufacturing
mounting frames were used to secure the specimen in the Technology. The density and the bulk modulus of air were
holder. Precautions were taken to minimize sound leakages: 1.18 kg/m3 and 142 kPa, respectively.
(1) tapes were applied along the contacting edges between An incident wave loading of a unit magnitude (1 Pa)
the specimen and the holder and on the bolt holes; (2) model- was specified to propagate from the upstream to the anechoic
ling clay was applied along the outer edges of the mounting end (non-reflecting) of the downstream tube [Fig. 1(c)]. Zero
frames that were in contact with the holder and along the displacement was specified in the x- and y-directions on the
outer edges of the holder that were in contact with the outer surface of the holder to emulate the constraints due to
impedance tube. Figure 1(b) shows the described set-up. the tube walls. Tie constraints were specified on all acoustic-
To reduce experimental errors that could be attributed to structure interfaces using the surface-to-surface approach to
the installation of the holder in the impedance tube, the achieve an acoustic-structure interaction. The constraints
installation was repeated five times where a set of three mea- were specified according to the software’s recommendations
surements was recorded for each repetition. Subsequently, where the slave surface should preferably be of a lower wave
the datasets were used to compute the expanded uncertainty speed in relative to the master surface.28 Solid-to-shell cou-
U% as given by26 pling was specified between the membranes’ edges and the
" # contacting surfaces of the frames. The remaining boundaries
XM
U % ¼ k % sx þ2 2
bk ; (1) of the fluid domains were set as sound hard. The fluid and
k¼1 the solid domains were discretised by quadratic elements
(AC3D20 and C3D20R, respectively), while the membranes
where k% ¼ 2 for a confidence level of 95%; sx denotes the were discretised by linear shell elements (S4R). Note that
standard deviation; M and bk denote the number of system- quadratic shell elements (S8R) would be suitable only for
atic error sources and the standard deviation estimate of the thick plates.28 All domains were discretised with at least six
k-th systematic error source. In this study, the microphones elements per wavelength.
were considered as the only systematic error source where Complex pressure values were extracted from the micro-
bk ¼ 60.3 dB according to the manufacturer’s datasheet. phone nodes [Fig. 1(c)] to compute the transmission loss
The numerical model was developed in ABAQUS 2017, (TL) based on the transfer matrix method as given by25
involving three main segments—upstream tube, specimen,
and downstream tube [Fig. 1(c)]. The dimensions of the TL ¼ 20 log10 jsj; (2)
upstream and downstream tubes (10 cm diameter) and the TABLE I. Material properties specified in the simulation.
microphone positions with respect to the specimen were kept
identical to the test set-up except for the omission of the Density Young’s Poisson’s Structural
mounting frames. This omission was simplified by specify- Material (kg/m3) modulus (GPa) ratio loss factor
ing zero displacement in the z-direction on the frame surfa-
Stainless steel 7800 206 0.28 …
ces that were in contact with the mounting frames in the Polylactide 1250 2.1 0.36 …
experiment. Experimentally, the distance between each pair Mylar 1380 5.1 0.38 0.03
of upstream/downstream microphones was either 8 or 30 cm
051903-3 Ang, Koh, and Lee Appl. Phys. Lett. 112, 051903 (2018)

where s denotes the transmission coefficient as given by25 modulus, respectively. In this case, the homogenization
scheme—not considered in this study—may be applied to
2eikd determine both parameters.30
s¼ ; (3)
T11 þ T12 =qo co þ qo co T21 þ T22 When D ¼ 3 mm, one new peak (653 Hz) and one new
dip (637 Hz) were predicted [Fig. 2(b)]. This phenomenon
where k denotes the wavenumber; Tij denotes the transfer was due to the resonant behaviour of both cavities coupled
matrix element in row i and column j; qo denotes the density via the orifice [View A in Fig. 2(b)]. As a result, when one
of air; and co denotes the sound speed in air. cavity was expanding, the other cavity was contracting. The
Figure 2(a) shows the measured and simulated TL remaining peak and dips were attributed to that when
curves of the specimen when D ¼ 0 mm. Note that the nor- D ¼ 0 mm [Fig. 2(a)]. Good agreement in the overall trend
mal incidence mass law is included only as a reference with was observed between the measurement and the simulation.
respect to an unbounded panel of equivalent surface density However, the new peak and dip were absent in the measure-
and does not provide a direct indication of the acoustical ment. This absence could be reasoned by the fact that fused
benefit and drawback of the small-scale specimen.29 As this deposition modelling is known to produce orifices with poor
specimen was used as a benchmark to identify the orifice’s cylindricity compared to other additive manufacturing tech-
effect later, its acoustical performance was not tuned for niques—such as stereolithography or selective laser sinter-
low-frequency noise control applications. Comparing both ing—or conventional machining methods.31 Hence, the
curves, good agreement in the overall trend was observed geometrical inaccuracy with respect to the small orifice
with four dips (491, 1000, 1169, and 1428 Hz) and one peak diameter (3 mm) could have influenced the resonant behav-
(1079 Hz). Experimentally, the peak and the dips occurred at iour due to the orifice, resulting in the missing peak and dip.
around the same frequency with a maximum deviation of Also, the acoustic boundary layer thickness could be large
2.9%. The mode shapes at 491 and 1169 Hz corresponded relative to the orifice diameter. Hence, viscous losses due to
to the in-phase translational motion of the membranes in the the interaction with the acoustic boundary layer—not consid-
normal direction (dipolar resonance), while the mode shapes ered numerically—could also “smear out” the new peak and
at 1000 and 1428 Hz corresponded to the coupling between dip.32 These hypotheses can be reinforced by the measured
the membranes and the larger cavity. This coupling resulted TL curves when larger orifice diameters (D ¼ 5 and 7 mm)
in an out-of-phase translational motion of the membranes in were considered [Figs. 2(c) and 2(d), respectively].
the normal direction (monopolar resonance). Dipolar and The missing peak and dip were now reflected in the mea-
monopolar resonances could suggest the presence of surement at 874 and 806 Hz, respectively, when D ¼ 5 mm,
negative effective mass density and negative effective bulk and at 948 and 868 Hz, respectively, when D ¼ 7 mm.

FIG. 2. Simulated, measured, and analytical TL curves of the specimen when (a) D ¼ 0 mm; (b) D ¼ 3 mm; (c) D ¼ 5 mm; and (d) D ¼ 7 mm alongside the
mode shapes at the eigenfrequencies. Views A, B, and C show the top view of the mode shape. The shaded region represents the limits of the experimental
uncertainty at a confidence level of 95%. (e) Simulated normal displacement (magnitude) ratio R when D ¼ 3 mm alongside the displacement profile of the
membranes at each peak frequency. Blue and red dashed lines—extending to (b)—represent the dip and the peak frequencies, respectively. (f) Change in TL
between D ¼ 5 mm and D ¼ 0 mm. The shaded region highlights the orifice’s effect.
051903-4 Ang, Koh, and Lee Appl. Phys. Lett. 112, 051903 (2018)

Comparing the mode shapes in Figs. 2(c) and 2(d) with Fig.
2(b), the newly captured peak and dip were indeed due to the
resonant behaviour of the two cavities coupled via the orifice.
The orifice’s effect can be associated with a “tonraum”
resonator.33 Assuming that the cavities’ boundaries are rigid,
the cavity-orifice-cavity system can be seen as a one-
dimensional spring-mass-spring system. Knowing the acous-
tic stiffness of each cavity and the mass of the neck,34 the
system’s natural frequency fn can be estimated by
 0:5 FIG. 3. Relationship between the orifice diameter and the corresponding TL
co AðVs þ Vl Þ peak and dip frequencies.
fn ¼ ; (4)
2p Vs Vl L0
Having understood the physics behind the TL curves,
where A and L0 denote the cross-sectional area and the effec- the design aspects of the specimen are next discussed for tun-
tive length of the neck, respectively; and Vs(l) denotes the ing purposes. Recalling the results, an upward shift in the TL
volume of the smaller (larger) cavity. Since fn / A0.5, an peak and dip frequencies was observed when the orifice
increase in D would lead to a higher natural frequency. This diameter was increased. This relationship is shown in Fig. 3.
understanding supports the observations where the dip fre- Both trendlines showed an incremental shift in frequency for
quency shifted from 637 to 833 Hz, and to 891 Hz, when D a large orifice. For a small orifice, the frequency gap between
was increased from 3 to 5 mm, and to 7 mm, respectively the trendlines was closer compared to that of a large orifice.
[Figs. 2(b)–2(d)]. This information may be useful when tuning the specimen
In Figs. 2(a)–2(d), the discrepancies in peak/dip fre- for a given scenario. For example, with an orifice diameter
quency between the simulation and the measurement could of 5–6 mm, the specimen may be potential as a see-through
be due to the idealization of boundary conditions in the noise barrier at construction sites.36 As this study focused on
numerical model35 and the deviation in material properties in the orifice’s effect to enhance TL, the specimens were not
contrast to that given in the handbook27 or manufacturer’s tuned for low-frequency noise control applications. For this
datasheet. For the discrepancies in TL between the simula- consideration, other parameters can be tuned apart from the
tion and the measurement, it could be due to the loud- orifice diameter. For example, the acoustic stiffness of each
speaker’s limitation to generate a loud noise signal in the cavity and the flexural stiffness of each membrane could be
impedance tube, resulting in a maximum TL of only reduced to shift the TL peaks and dips to the lower frequen-
40–45 dB. cies. If extended in scale, larger specimens could also be
Till now, the focus has been placed on understanding considered because the size is no longer constrained by the
the physics behind the dip frequencies but not the peak fre- impedance tube. Consequently, with larger membrane area
quencies. The discussion will now focus on the latter by ana- and cavity volume, the TL peaks and dips would occur in the
lysing the normal displacement (magnitude) ratio of the low-frequency range.
second to the first membrane based on D ¼ 3 mm as an In conclusion, this study has shown that the acoustical
example [Fig. 2(e)]. Mathematically, the ratio R is given by performance of a plate-type acoustic metamaterial could be
U ?;m2 =U ?;m1 , where U ? denotes the average normal dis- complemented by the coupling effect between two cavities
placement (magnitude) of the entire membrane and sub- via an orifice in addition to the resonant behaviour of the
scripts m1 and m2 denote the first and second membranes, membranes. This consideration could address the shortcom-
respectively. ings of typical membrane-type acoustic metamaterials—such
A ratio of 1 would imply that both membranes had the as the membranes’ durability, stress relaxation, pretension,
same displacement profile regardless of their relative phase, and the spatial consistency of the platelet(s)—which may
suggesting a resonant behaviour of the specimen as indicated pose manufacturing issues if extended in scale for industrial
by the blue dashed lines in Figs. 2(b) and 2(e). At the peak applications.
frequencies (red dashed lines), the ratios were less than 1,
suggesting that the second membrane was nearly motionless The authors acknowledge the financial support provided
relative to the first membrane. This understanding could be by the Singapore Economic Development Board (EDB) and
reinforced by the displacement profile of the membranes Singapore Technologies Kinetics Ltd. (ST Kinetics) under
[Fig. 2(e)] where the second membrane functioned like a the EDB Industrial Postgraduate Program (EDB-IPP) (Grant
“noise shield”, minimizing sound transmission. Significant No. COY-15-IPP/120005). The authors thank Dr. Le Quan
TL improvement due to the orifice’s effect could also be Ngoc Tran for performing dynamic mechanical analysis on
achieved. For example, when D ¼ 5 mm, the TL improved the Mylar film. The authors thank Nitto Denko (Singapore),
by up to 35 dB despite a drop in TL by up to 17.2 dB at the Pte. Ltd. for the sponsorship of customized adhesive tapes.
dip frequency [Fig. 2(f)]. It must be emphasized that, if
extended in scale, careful design is necessary. Else, the 1
C. J. Naify, C. M. Chang, G. McKnight, and S. R. Nutt, J. Appl. Phys.
acoustical performance observed in small-scale specimens 108, 114905 (2010).
2
C. J. Naify, C. M. Chang, G. McKnight, and S. R. Nutt, J. Acoust. Soc.
may be considerably different in large-scale specimens due
Am. 132, 2784 (2012).
to the influence of the boundary conditions and the global 3
C. J. Naify, C. M. Chang, G. McKnight, and S. R. Nutt, J. Appl. Phys.
dynamic response of the structure.14 110, 124903 (2011).
051903-5 Ang, Koh, and Lee Appl. Phys. Lett. 112, 051903 (2018)

4 22
C. J. Naify, C. M. Chang, G. McKnight, F. Scheulen, and S. R. Nutt, P. Marinova, S. Lippert, and O. von Estorff, J. Acoust. Soc. Am. 142,
J. Appl. Phys. 109, 104902 (2011). 2400 (2017).
5 23
J. Mei, G. Ma, M. Yang, Z. Yang, W. Wen, and P. Sheng, Nat. Commun. S. Varanasi, J. S. Bolton, and T. H. Siegmund, J. Acoust. Soc. Am. 141,
3, 756 (2012). 602 (2017).
6 24
Z. Yang, H. M. Dai, N. H. Chan, G. Ma, and P. Sheng, Appl. Phys. Lett. F. Ma, M. Huang, and J. H. Wu, J. Appl. Phys. 121, 015102 (2017).
25
96, 041906 (2010). ASTM, ASTM E2611-09: Standard Test Method for Measurement of
7
F. Ma, J. H. Wu, M. Huang, W. Zhang, and S. Zhang, J. Phys. D: Appl. Normal Incidence Sound Transmission of Acoustical Materials Based on
Phys. 48, 175105 (2015). the Transfer Matrix Method (ASTM International, West Conshohocken,
8
F. Ma, J. H. Wu, and M. Huang, J. Phys. D: Appl. Phys. 48, 465305 (2015). PA, USA, 2009).
9 26
Y. Zhang, J. Wen, H. Zhao, D. Yu, L. Cai, and X. Wen, J. Appl. Phys. H. W. Coleman and W. G. Steele, Experimentation, Validation, and
114, 063515 (2013). Uncertainty Analysis for Engineers, 3rd ed. (John Wiley & Sons,
10
G. Ma, M. Yang, Z. Yang, and P. Sheng, Appl. Phys. Lett. 103, 011903 Hoboken, NJ, USA, 2009), Chap. 2, pp. 29–60.
27
(2013). F. Cardarelli, Materials Handbook: A Concise Desktop Reference, 2nd ed.
11
M. Yang, G. Ma, Z. Yang, and P. Sheng, Phys. Rev. Lett. 110, 134301 (2013). (Springer, London, England, 2008).
12 28
L. Y. L. Ang, Y. K. Koh, and H. P. Lee, Int. J. Appl. Mech. 8, 1650072 SIMULIA, Abaqus Analysis User’s Guide (SIMULIA, Providence, RI,
(2016). USA, 2017).
13 29
N. Sui, X. Yan, T. Y. Huang, J. Xu, F. G. Yuan, and Y. Jing, Appl. Phys. F. J. Fahy and P. Gardonio, Sound and Structural Vibration: Radiation,
Lett. 106, 171905 (2015). Transmission and Response, 2nd ed. (Academic Press, Cambridge, MA,
14
A. Peiffer, M. Grunewald, and P. Lempereur, Appl. Phys. Lett. 107, USA, 2007), Chap. 5, pp. 277–373.
30
216101 (2015). M. Yang, G. Ma, Y. Wu, Z. Yang, and P. Sheng, Phys. Rev. B 89, 064309
15
X. Wang, H. Zhao, X. Luo, and Z. Huang, Appl. Phys. Lett. 108, 041905 (2014).
31
(2016). P. Shah, R. Racasan, and P. Bills, Case Study Nondestr. Test. Eval. 6, 69
16
F. Langfeldt, H. Kemsies, W. Gleine, and O. von Estorff, Phys. Lett. A (2016).
32
381, 1457 (2017). D. T. Blackstock, Fundamentals of Physical Acoustics (John Wiley &
17
H. Zhang, Y. Xiao, J. Wen, D. Yu, and X. Wen, Appl. Phys. Lett. 108, Sons, Hoboken, NJ, USA, 2000).
33
141902 (2016). I. Malecki, Physical Foundations of Technical Acoustics, 1st ed.
18
X. Chen, X. Xu, S. Ai, H. Chen, Y. Pei, and X. Zhou, Appl. Phys. Lett. (Pergamon Press, Oxford, England, 1969), Chap. 9, pp. 375–424.
34
105, 071913 (2014). L. E. Kinsler, A. R. Frey, A. B. Coppens, and J. V. Sanders, Fundamentals
19
S. Xiao, G. Ma, Y. Li, Z. Yang, and P. Sheng, Appl. Phys. Lett. 106, of Acoustics, 4th ed. (John Wiley & Sons, Hoboken, NJ, USA, 2000),
091904 (2015). Chap. 10, pp. 272–301.
20 35
L. Y. L. Ang, Y. K. Koh, and H. P. Lee, Appl. Phys. Lett. 111, 041903 (2017). R. D. Cook, D. S. Malkus, M. E. Plesha, and R. J. Witt, Concepts and
21
F. Langfeldt, W. Gleine, and O. von Estorff, in Proceedings of the INTER- Applications of Finite Element Analysis, 4th ed. (John Wiley & Sons,
NOISE 2016, edited by W. Kropp (Deutsche Gesellschaft Fuer Akustik, Hoboken, NJ, USA, 2002).
36
Hamburg, Germany, 2016), pp. 7551–7562. H. P. Lee, Z. Wang, and K. M. Lim, Build. Acoust. 24, 21 (2016).

You might also like