Whole

Prediction and optimisation of the acoustic performance of
mufflers for sleep apnoea devices
Author:
Jones, Peter William
Publication Date:
2011
DOI:
https://doi.org/10.26190/unsworks/15269
License:
https://creativecommons.org/licenses/by-nc-nd/3.0/au/
Link to license to see what you are allowed to do with this resource.
Downloaded from http://hdl.handle.net/1959.4/51649 in https://

unsworks.unsw.edu.au on 2022-04-12
PREDICTIO
ON AND OPTIMISATIO
ON OF THE
ACOUSTIC
C PERFORMANCE OF MUFFLERS
M
FOR
R SLEEP APNOEA DEVIICES
By
Peter William Jones
A thesis submitted in fulfilment
of the requirements for the degree of
Doctor of Philosophy
Schoool of Mechanical and Manufacturing Engineeering
T University of New South Wales, Australia

The
December 2011
THE UNIVERSITY OF NEW SOUTH WALES
Thesis/Dissertation Sheet
Surname or Family name: JONES
First name: PETER Other name/s: WILLIAM
Abbreviation for degree as given in the University calendar: PHD
School: MECHANICAL AND MANUFACTURING ENGINEERING Faculty: ENGINEERING
Title: PREDICTION AND OPTIMISATION OF THE ACOUSTIC PERFORMANCE OF MUFFLERS FOR SLEEP APNOEA DEVICES
Abstract
Noise from continuous positive airway pressure (CPAP) flow generators is controlled using very small, irregularly shaped mufflers which must
attenuate noise at frequencies up to 5 kHz. This thesis investigates computational and experimental techniques to predict the acoustic performance
of these mufflers and presents a computational technique to efficiently optimise the muffler design in order to maximise the performance.
The work presented in this thesis can be roughly grouped as (1) acoustic modelling, (2) experimental measurements and (3) parametric design
optimisation. In the first group, three different analytical techniques were reviewed in the context of modelling the acoustic performance of a simple
expansion chamber reactive muffler. Three-dimensional acoustic finite element analysis (FEA) models of simple geometries and complex CPAP
device muffler geometries were then developed. Polyurethane foam inserts were incorporated into the FEA models of the muffler designs. The
computational results of both the reactive and resistive designs were compared to the analytical results and were successfully validated against
experimental data.
The experimental component of the thesis includes measurement of the acoustic performance of mufflers and acoustic characterisation of
polyurethane foams. The transmission loss of the simple and complex mufflers was measured using a two-microphone acoustic pulse method. The
acoustic performance of the mufflers was also measured in the presence of mean flow to examine the influence of air flow on the transmission loss.
The acoustic properties of two polyurethane foams were obtained in an impedance tube using a two cavity method. A rig for measuring the airflow
resistivity of the porous materials was designed, constructed and commissioned. The combined acoustic characterisation of the foams was
incorporated into the FEA models, enabling the modelling of muffler designs which contain absorptive elements.
In the final part of the thesis, a methodology to efficiently optimise muffler configurations to maximise their acoustic performance was developed.
This involved the development of an integrated model that couples various optimisation tools available through the MATLAB interface with the
acoustic FEA capabilities of COMSOL. The reactive and resistive components of a prototype CPAP device muffler were independently optimised
for acoustic performance. Results were also obtained for the transmission loss of the muffler for which the reactive and resistive components were
simultaneously optimised. The merits of each approach are discussed.
Declaration relating to disposition of project thesis/dissertation
I hereby grant to the University of New South Wales or its agents the right to archive and to make available my thesis or dissertation in whole or in
part in the University libraries in all forms of media, now or here after known, subject to the provisions of the Copyright Act 1968. I retain all property
rights, such as patent rights. I also retain the right to use in future works (such as articles or books) all or part of this thesis or dissertation.
I also authorise University Microfilms to use the 350 word abstract of my thesis in Dissertation Abstracts International (this is applicable to doctoral
theses only)
Signature Witness Date
The University recognises that there may be exceptional circumstances requiring restrictions on copying or conditions on use. Requests for
restriction for a period of up to 2 years must be made in writing. Requests for a longer period of restriction may be considered in exceptional
circumstances and require the approval of the Dean of Graduate Research.
FOR OFFICE USE ONLY Date of completion of requirements for Award:
THIS SHEET IS TO BE GLUED TO THE INSIDE FRONT COVER OF THE THESIS

Originality Statement
I hereby declare that this submission is my own work and to the best of my
knowledge it contains no materials previously published or written by another
person, or substantial proportions of material which have been accepted for
the award of any other degree or diploma at UNSW or any other educational
institution, except where due acknowledgement is made in the thesis. Any
contribution made to the research by others, with whom I have worked at
UNSW or elsewhere, is explicitly acknowledged in the thesis. I also declare
that the intellectual content of this thesis is the product of my own work,
except to the extent that assistance from others in the project's design and
conception or in style, presentation and linguistic expression is
acknowledged.
Signed ……………………………………………..............
Date ……………………………………………..............
ii
Copyright Statement
I hereby grant the University of New South Wales or its agents the right to archive and to
make available my thesis or dissertation in whole or part in the University libraries in all
forms of media, now or here after known, subject to the provisions of the Copyright Act
1968. I retain all proprietary rights, such as patent rights. I also retain the right to use in
future works (such as articles or books) all or part of this thesis or dissertation. I also
authorise University Microfilms to use the 350 word abstract of my thesis in Dissertation
Abstract International. I have either used no substantial portions of copyright material in
my thesis or I have obtained permission to use copyright material; where permission has
not been granted I have applied/will apply for a partial restriction of the digital copy of
my thesis or dissertation.
Signed ……………………………………………..............
Date ……………………………………………..............
Authenticity Statement
I certify that the Library deposit digital copy is a direct equivalent of the final officially
approved version of my thesis. No emendation of content has occurred and if there are
any minor variations in formatting, they are the result of the conversion to digital format.
Signed ……………………………………………..............
Date ……………………………………………..............
iii
Abstract
Obstructive sleep apnoea affects approximately 10-25% of the adult population. It is

most frequently managed by applying an elevated pressure to the airway during sleep;
however compliance with this continuous positive airway pressure (CPAP) therapy is
poor. Low compliance is partly attributed to the noise emitted by the flow generating
device. Noise from the flow generator is controlled using mufflers situated in the air flow
paths to and from the fan. These mufflers are very small, irregularly shaped and must
attenuate noise at frequencies up to 5 kHz. This thesis investigates computational and
experimental techniques to predict the acoustic performance of these mufflers and
presents a computational technique to efficiently optimise the muffler design in order to
maximise the performance.
The work presented in this thesis can be roughly grouped as (1) acoustic modelling,
(2) experimental measurements and (3) parametric design optimisation. In the first group,
three different analytical techniques were reviewed in the context of modelling the
acoustic performance of a simple expansion chamber reactive muffler. The techniques
investigated were the continuity of pressure and volume velocity method, the impedance
method (also known as transmission line theory), and the transfer matrix method (also
known as the 2-port approach). Three-dimensional acoustic finite element analysis
models of simple geometries (cylindrical expansion chamber mufflers) and complex
geometries (one production CPAP device muffler and two prototype CPAP device
mufflers) were then developed. Polyurethane foam inserts were incorporated into the
finite element analysis models of the muffler designs. Transmission loss was evaluated
directly, by considering attenuation of acoustic power, and indirectly using the transfer
matrix method. The computational results of both the reactive and resistive designs were
compared to the analytical results and were successfully validated against experimental
data.
The experimental component of the thesis includes measurement of the acoustic

performance of mufflers and acoustic characterisation of polyurethane foams. The
transmission loss of the simple and complex mufflers was measured using a two-
microphone acoustic pulse method. As CPAP devices operate with varying air flows, the
iv
transmission loss of two simple expansion chamber mufflers and two CPAP device
mufflers were also measured in the presence of mean flow to examine the influence of air
flow on the acoustic performance. The acoustic properties of two polyurethane foams
were obtained in an impedance tube using a two cavity method. A rig for measuring the
airflow resistivity of the porous materials was designed, constructed and commissioned.
The airflow resistivity characteristics of the foam types was then combined with the
impedance tube results to produce a set of data for use in the finite element
computational models, enabling the modelling of muffler designs which contain
absorptive elements.
In the final part of the thesis, a methodology to efficiently optimise muffler

configurations to maximise their acoustic performance was formulated. This involved the
development of an integrated model that couples various optimisation tools available
through the MATLAB interface (both native and third party) with the acoustic finite
element analysis capabilities of COMSOL. The reactive and resistive components of a
prototype CPAP device muffler were independently optimised for acoustic performance.
Results were also obtained for the transmission loss of the muffler for which the reactive
and resistive components were simultaneously optimised. The merits of each approach
are discussed.
v
Acknowledgements
Firstly, I would like to express my deep gratitude to my wife Stacey for the sacrifices that
she has made throughout the long process of completing this thesis, especially the final
six months when I must have seemed so absent. I hope that I can repay your generosity.
I would like to thank my supervisor Assoc. Prof. Nicole Kessissoglou for her boundless
enthusiasm and keen editorial eye. When I felt that time was running out, her
encouragement kept me on track with what I had originally set out to achieve. The scope
of this thesis would be less complete were it not for that encouragement.
I would also like to thank the Vibration and Acoustics laboratory manager Mr. Russell
Overhall for his assistance and for sharing his expertise with the equipment used in the
experimental component of this thesis. His dedication to providing timely support to
students deserves special mention.
Lastly, I would like to thank my parents for their love and support, and for understanding
that my increasingly less frequent contact did not mean that I cared any less.
Contributions
This work was carried out with funding provided by the Australian Research Council and
ResMed under ARC Linkage Project LP0669543. This support is gratefully
acknowledged.
vi
Table of Contents
Originality Statement ..................................................................................................... ii
Copyright Statement ...................................................................................................... iii
Authenticity Statement .................................................................................................. iii
Abstract .................................................................................................................. iv
Acknowledgements ........................................................................................................ vi
Contributions.................................................................................................................. vi
Table of Contents .......................................................................................................... vii
List of Figures ................................................................................................................. xi
List of Tables ................................................................................................................ xxi
List of Symbols ........................................................................................................... xxiii
Chapter 1 Introduction and Literature Review...................................................... 1
1.1 Introduction ............................................................................................... 1
1.2 Obstructive Sleep Apnoea ......................................................................... 1
1.3 Background on Mufflers ........................................................................... 3
1.4 CPAP Device Mufflers .............................................................................. 5
1.5 Literature Review ...................................................................................... 6
1.5.1 Analytical approaches ................................................................. 7
1.5.2 Numerical approaches ................................................................. 9
1.5.3 Porous materials ........................................................................ 11
1.5.4 Design optimisation .................................................................. 13
1.6 Contribution to Research ......................................................................... 16
1.7 Thesis Overview ...................................................................................... 18
vii
Chapter 2 Transmission Loss of a Reactive Muffler ............................................ 20
2.1 Introduction ............................................................................................. 20
2.2 Analytical Methods ................................................................................. 20
2.2.1 Single chamber muffler ............................................................. 21
2.2.2 Multiple chamber muffler ......................................................... 30
2.3 Computational Approaches ..................................................................... 33
2.3.1 Direct approach ......................................................................... 35
2.3.2 Indirect approach ....................................................................... 38
2.4 Experimental Approach ........................................................................... 40
2.5 Generic CPAP Device Expansion Chamber Muffler .............................. 46
Chapter 3 Transmission Loss of a Resistive Muffler ........................................... 50
3.1 Introduction ............................................................................................. 50
3.2 Computational Modelling of Porous Materials ....................................... 50
3.3 Experimental Methods ............................................................................ 55
3.3.1 Characteristic impedance and propagation coefficient ............. 55
3.3.2 Airflow resistivity ..................................................................... 62
3.4 Equivalent Fluid Parameters ................................................................... 64
3.5 Muffler Transmission Loss ..................................................................... 68
Chapter 4 Transmission Loss of CPAP Device Mufflers ..................................... 73
4.1 Introduction ............................................................................................. 73
4.2 Muffler Designs ....................................................................................... 73
4.3 Finite Element Modelling of CPAP Device Mufflers ............................. 76
4.3.1 ANSYS ...................................................................................... 76
4.3.2 COMSOL .................................................................................. 77
4.3.3 Transmission Loss of the CPAP Device Mufflers .................... 78
viii
4.4 Reactive CPAP Device Muffler Designs ................................................ 78
4.4.2 Integrated chamber muffler ....................................................... 81
4.4.3 Interconnected three-chamber muffler ...................................... 81
4.4.4 Comparison of muffler designs ................................................. 84
4.5 Resistive CPAP Device Muffler Designs ................................................ 85
4.5.2 Integrated chamber muffler ....................................................... 89
Chapter 5 Optimisation of Muffler Acoustic Performance ................................. 90
5.1 Introduction ............................................................................................. 90
5.2 Optimisation Methodology...................................................................... 90
5.3 Reactive Expansion Chamber Muffler .................................................... 94
5.4 Expansion Chamber Muffler With Foam Lining .................................... 98
5.5 CPAP Device Muffler ............................................................................. 99
5.5.1 Optimisation strategy .............................................................. 104
5.5.2 Muffler geometry optimisation ............................................... 107
5.5.3 Foam thickness optimisation ................................................... 109
5.5.4 Simultaneous geometry and foam optimisation ...................... 113
Chapter 6 Effect of Mean Flow on Transmission Loss ...................................... 120
6.1 Introduction ........................................................................................... 120
6.2 CPAP Device Flow Requirements ........................................................ 120
6.3 Experimental Approach ......................................................................... 121
6.4 Transmission Loss with Mean Flow ..................................................... 124
6.5 Flow Generated Noise ........................................................................... 128
ix
Chapter 7 Summary and Future Work ............................................................... 136
7.1 Summary ............................................................................................... 136
7.1.1 Prediction and measurement of muffler acoustic

performance............................................................................. 136
7.1.2 Optimisation of muffler acoustic performance ....................... 139
7.1.3 Measurement of muffler acoustic performance with mean

flow ......................................................................................... 141
7.2 Recommendations for Future Work ...................................................... 142
7.2.1 Computational models ............................................................ 142
7.2.2 Experimental methods ............................................................. 143
7.2.3 Further design applications ..................................................... 145
References ............................................................................................................... 146
Appendix A Impedance Method Analysis of a Multiple Chamber Muffler ........ 158
Appendix B Incorporating Revised Coefficients into Original Delany-Bazley

Functions .............................................................................................. 164
Appendix C Optimisation Algorithm Assessment ................................................. 168
C.1 Optimisation Algorithms ....................................................................... 168
C.1.1 “Optimize” / Nelder-Mead ...................................................... 169
C.1.2 Genetic algorithm .................................................................... 169
C.1.3 Simulated annealing ................................................................ 169
C.1.4 Particle swarm optimisation .................................................... 170
C.2 Reactive Expansion Chamber Muffler .................................................. 170
C.3 Expansion Chamber Muffler with Foam Lining ................................... 175
Publications Arising from this Thesis ....................................................................... 180
x
List of Figures
Figure 1.1 Mechanism of obstructive sleep apnoea .................................................... 2
Figure 1.2 ResMed S9 continuous positive airway pressure (CPAP) device ............. 2
Figure 1.3 Muffler acoustic performance metrics – insertion loss and

transmission loss........................................................................................ 4
Figure 1.4 CPAP device chassis showing inlet mufflers (red), blower chamber
(green) and outlet muffler (yellow). Arrows indicate air flow path. ......... 5
Figure 1.5 Comparison of material parameters used in three porous material

models ..................................................................................................... 14
Figure 2.1 A cylindrical expansion chamber reactive muffler ................................. 22
Figure 2.2 Transmission loss of the single 86 mm diameter expansion

chamber muffler, comparing continuity, impedance and transfer
matrix analytical methods ....................................................................... 31
Figure 2.3 A multiple expansion chamber reactive muffler ..................................... 31
Figure 2.4 Transmission loss of the multiple expansion chamber muffler,

comparing impedance and transfer matrix analytical methods ............... 34

chamber muffler, multiple expansion chamber muffler and single
chamber muffler times a factor of three – analytical results using
TMM with Karal correction factors ........................................................ 34

chamber muffler, comparing analytical results and FEA results ............ 36
Figure 2.7 Pressure fields in the single 86 mm diameter expansion chamber

muffler at frequencies coinciding with peak transmission loss –
FEA results .............................................................................................. 37
Figure 2.8 Transmission loss of the multi-chamber muffler, comparing

analytical results and FEA results ........................................................... 37
xi
chamber muffler, comparing FEA results obtained using the direct
approach (acoustic power) and indirect approach (transfer matrix
method) .................................................................................................... 40
Figure 2.10 Transfer matrix parameters of the single 86 mm diameter

expansion chamber muffler, comparing analytical results (obtained
with and without Karal correction factors) and FEA results ................... 42
Figure 2.11 Schematic diagram of the two-microphone acoustic pulse

experimental set-up (dimensions in mm) ................................................ 43
Figure 2.12 Photograph of the two-microphone acoustic pulse experimental set-

up showing horn drivers, microphones and muffler................................ 43
Figure 2.13 Upstream microphone (ͳ) time history showing rectangular FFT
weighting function ................................................................................... 44
Figure 2.14 Downstream microphone (ʹ) time history showing exponential

FFT weighting function ........................................................................... 44

chamber muffler, comparing analytical results, FEA results and
experimental results ................................................................................. 46
Figure 2.16 A side branch Helmholtz resonator ......................................................... 47
Figure 2.17 Pressure field in the single 114 mm diameter expansion chamber
muffler at the frequency coinciding with the peak transmission loss
– FEA results ........................................................................................... 48

chamber muffler, comparing analytical results and FEA results ............ 49

chamber muffler, comparing FEA results and experimental results ....... 49
Figure 3.1 Photographs of the polyurethane foam specimens – light grey (left)
and dark grey (right) ................................................................................ 55
Figure 3.2 Schematic diagram of the impedance tube configuration ....................... 56
xii
Figure 3.3 Photograph of the Brüel & Kjær LAN-XI data acquisition system
and Type 4206 impedance tube ............................................................... 57
Figure 3.4 Normalised complex characteristic impedance of the light grey

foam and dark grey foam – experimental data ........................................ 60
Figure 3.5 Complex propagation constant of the light grey foam and dark grey
foam – experimental data ........................................................................ 60
Figure 3.6 Normalised complex characteristic impedance of the light grey

foam, comparing numerical model values and experimental data .......... 61
Figure 3.7 Propagation constant of the light grey foam, comparing numerical
model values and experimental data ....................................................... 61
Figure 3.8 Schematic diagram of the airflow resistivity experimental set-up .......... 62
Figure 3.9 Photograph of the airflow resistivity experimental set-up ...................... 63
Figure 3.10 Airflow resistivity of the light and dark grey foams – experimental
results....................................................................................................... 64
Figure 3.11 Equivalent fluid speed of sound in the light and dark grey foams –
numerical model values obtained using the derived Delany-Bazley
coefficients .............................................................................................. 65
Figure 3.12 Equivalent fluid density of the light and dark grey foams –
coefficients .............................................................................................. 65
Figure 3.13 Equivalent fluid speed of sound in the light grey foam, comparing
numerical model values obtained using original and derived
Delany-Bazley coefficients ..................................................................... 67
Figure 3.14 Equivalent fluid density of the light grey foam, comparing
numerical model values obtained using original and derived
Delany-Bazley coefficients ..................................................................... 67
numerical model values obtained using derived coefficients in the
original Delany-Bazley functions and restated functions (no airflow
resistivity) ................................................................................................ 69
xiii
Figure 3.16 Equivalent fluid density of the light grey foam, comparing
numerical model values obtained using derived coefficients in the
original Delany-Bazley functions and restated functions (no airflow
resistivity) ................................................................................................ 69

chamber muffler with and without light grey foam insert,
comparing FEA results and experimental results .................................... 71

chamber muffler with foam inserts, comparing FEA results and
Figure 3.19 Transmission loss of the light grey foam-filled single 114 mm
diameter expansion chamber muffler, comparing FEA results
obtained using original and derived Delany-Bazley coefficients and
Figure 3.20 Transmission loss of the dark grey foam-filled single 114 mm
diameter expansion chamber muffler, comparing FEA results
obtained using original and derived Delany-Bazley coefficients and
Figure 4.1 Air volume of the production single chamber CPAP device muffler ..... 74
Figure 4.2 Air volume of the prototype integrated multi-chamber CPAP

device muffler.......................................................................................... 74
Figure 4.3 Air volume of the prototype interconnected three-chamber CPAP

device muffler.......................................................................................... 75
Figure 4.4 Transmission loss of the single chamber CPAP device muffler,
comparing ANSYS and COMSOL FEA results ..................................... 79
Figure 4.5 Transmission loss of the integrated chamber CPAP device muffler,
comparing ANSYS and COMSOL FEA results ..................................... 79
Figure 4.6 Transmission loss of the interconnected three-chamber CPAP

device muffler, comparing ANSYS and COMSOL FEA results ............ 80
xiv
comparing analytical results and experimental results ............................ 82
Figure 4.8 Transmission loss of the single chamber CPAP device muffler

device muffler, comparing FEA results and experimental results .......... 83
Figure 4.11 Transmission loss of the single chamber, integrated chamber and
interconnected three-chamber CPAP device muffler designs –
Figure 4.12 Finite element model geometry of the single chamber CPAP device
muffler showing the location of the foam insert (dark grey) .................. 86
Figure 4.13 Photograph of the single chamber CPAP device muffler with foam
insert ........................................................................................................ 86
Figure 4.14 Finite element model geometry of the integrated chamber CPAP
device muffler showing the location of the foam insert (dark grey) ....... 87
Figure 4.15 Transmission loss of the single chamber CPAP device muffler with
and without the light grey foam insert, comparing FEA results and
light and dark grey foam inserts, comparing FEA results and
Figure 4.17 Transmission loss of the integrated CPAP device muffler,

comparing performance without foam and with the first chamber
foam-filled - FEA results ......................................................................... 89
Figure 5.1 Flow diagram showing the integrated COMSOL-MATLAB

optimisation approach ............................................................................. 93
Figure 5.2 A cylindrical expansion chamber reactive muffler showing the

optimisation domain ................................................................................ 95
xv
Figure 5.3 Fitness landscape for the cylindrical expansion chamber reactive
muffler at 1 kHz excitation frequency..................................................... 97
Figure 5.4 Fitness landscape with objective function values (-TL) assessed by
the genetic algorithm solver during convergence – cylindrical
expansion chamber reactive muffler (white dot = convergence) ............ 97
Figure 5.5 A cylindrical expansion chamber muffler with foam lining showing
the optimisation domain .......................................................................... 98
Figure 5.6 Fitness landscape for the cylindrical expansion chamber muffler
with foam lining at 1 kHz excitation frequency .................................... 100
Figure 5.7 Fitness landscape with objective function values (-TL) assessed by
expansion chamber muffler with foam lining (white dot =
convergence) ......................................................................................... 100
Figure 5.8 Prototype CPAP device muffler showing optimisation control

variables................................................................................................. 101
Figure 5.9 Prototype CPAP device muffler showing foam constraints .................. 103
Figure 5.10 Transmission loss of the prototype CPAP device muffler using the
pre-optimised (original) geometry, comparing the reactive muffler
(no foam) and the muffler with foam inserts occupying the lower
half of chambers 2 and 3 – FEA results ................................................ 104
Figure 5.11 Pressure isosurfaces in the prototype CPAP device muffler at 1 kHz
excitation frequency – pre-optimised reactive muffler (no foam) –
FEA results ............................................................................................ 105
excitation frequency – pre-optimised muffler with foam in lower
half of chambers 2 and 3 – FEA results ................................................ 105
Figure 5.13 Sample CPAP fan noise source spectrum ............................................. 107
Figure 5.14 Transmission loss of the reactive prototype CPAP device muffler
(no foam), comparing original dimensions and optimised geometry
– FEA results ......................................................................................... 110
xvi
Figure 5.15 Transmission loss of the geometry-optimised prototype CPAP
device muffler, comparing the reactive muffler only and the
muffler with foam inserts in the lower half of chambers 2 and 3 –
FEA results ............................................................................................ 110
Figure 5.16 Transmission loss of the prototype CPAP device muffler,

comparing foam thickness optimisation in the low frequency band,
high frequency band and combined low and high frequency bands –
FEA results ............................................................................................ 112
Figure 5.17 Relationship between total volume of foam added to the muffler
and the average optimisation frequency ................................................ 113

comparing sequential and simultaneous optimisation of the
geometry and foam thickness – FEA results ......................................... 116

comparing simultaneous optimisation of the geometry and foam
thickness using different weighting in each of the frequency bands
– FEA results ......................................................................................... 116

comparing simultaneous optimisation of the geometry and foam
thickness in two frequency bands (low and high) and across the
entire 5 kHz frequency range – FEA results ......................................... 117
Figure 6.1 CPAP device blower air delivery requirements .................................... 121
Figure 6.2 Schematic diagram of the silencer box located at the inlet of the
two-microphone acoustic pulse experimental rig (dimensions in
mm) ....................................................................................................... 122
Figure 6.3 Photograph of the blower, silencer box and bell mouth ........................ 122
Figure 6.4 Transmission loss of the silencer box – experimental results ............... 123

chamber muffler, comparing the effect of varying mean air flow –
experimental results ............................................................................... 126
xvii
chamber muffler, comparing the effect of varying mean air flow –
comparing the effect of varying mean air flow – experimental
results..................................................................................................... 127

device muffler, comparing the effect of varying mean air flow –

chamber muffler with the light grey foam insert, comparing the
effect of varying mean air flow – experimental results ......................... 129
the light grey foam insert, comparing the effect of varying mean air
flow – experimental results ................................................................... 129
Figure 6.11 Signal-to-noise ratio measured at the downstream (ʹ) microphone

(using peak pressures), comparing the effect of varying mean air
flow ........................................................................................................ 130
Figure 6.12 Power autospectrum of flow noise measured at the upstream (ͳ)
microphone (with a straight-through duct fitted in place of a
muffler), comparing the effect of varying mean air flow ...................... 131
Figure 6.13 Average sound power measured at the downstream (ʹ)

microphone in the absence of a generated acoustic pulse,
comparing the contribution made by each muffler design .................... 133
Figure 6.14 Schematic diagram of the measurement locations used to assess

blower noise and silencer box performance .......................................... 134
Figure 6.15 Blower noise autospectra measured at 120 L/min, comparing noise
in the free field with noise in the duct adjacent to blower and noise
in the duct downstream of the silencer box ........................................... 135
Figure A.1 A multiple expansion chamber reactive muffler ................................... 158
xviii
Figure C.1 A cylindrical expansion chamber reactive muffler showing the
optimisation domain .............................................................................. 171
Figure C.2 Fitness landscape for the cylindrical expansion chamber reactive
muffler at 1 kHz excitation frequency................................................... 171
Figure C.3 Fitness landscape with objective function (-TL) values assessed by
the Optimize solver during convergence – cylindrical expansion
chamber reactive muffler (magenta dot = start; white dot =
convergence) ......................................................................................... 173
expansion chamber reactive muffler (white dot = convergence) .......... 173
the simulated annealing solver during convergence – cylindrical
expansion chamber reactive muffler (white dot = convergence) .......... 174
the particle swarm optimisation solver during convergence –
cylindrical expansion chamber reactive muffler (white dot =
convergence) ......................................................................................... 174
Figure C.7 A cylindrical expansion chamber muffler with foam lining showing
the optimisation domain ........................................................................ 175
Figure C.8 Fitness landscape for the cylindrical expansion chamber muffler
with foam lining at 1 kHz excitation frequency .................................... 176
Figure C.9 Fitness landscape with objective function values (-TL) assessed by
the Optimize solver during convergence – cylindrical expansion
chamber muffler with foam lining (green dot = start; white dot =
convergence) ......................................................................................... 178
convergence) ......................................................................................... 178
xix
the simulated annealing solver during convergence – cylindrical
convergence) ......................................................................................... 179
the particle swarm optimisation solver during convergence –
cylindrical expansion chamber muffler with foam lining (white dot
= convergence) ...................................................................................... 179
xx
List of Tables
Table 3.1 Derived Delany-Bazley function coefficients (revised form) ................. 58
Table 3.2 Foam airflow resistivity of the light and dark grey foams –
experimental data .................................................................................... 63
Table 3.3 Delany-Bazley function coefficients, comparing experimentally

derived values for the light and dark grey foams and the original
Delany-Bazley values .............................................................................. 66
Table 4.1 Dimensions of three CPAP device muffler designs ................................ 75
Table 5.1 Pre-optimisation dimensions and bounds of the optimisation

control variables for the prototype CPAP device muffler ..................... 102
Table 5.2 Original and optimised control variables for the prototype CPAP
device muffler – muffler geometry optimisation (reactive design
only)....................................................................................................... 108
Table 5.3 Optimised control variables for the prototype CPAP device muffler
– foam thickness optimisation only (using optimised muffler
geometry)............................................................................................... 111
Table 5.4 Optimised control variables for the prototype CPAP device
muffler, comparing sequential and simultaneous geometry and
foam optimisation .................................................................................. 115
Table 5.5 Optimised control variables for the prototype CPAP device
muffler, comparing simultaneous geometry and foam optimisation
in two frequency bands (low and high) and across the entire 5 kHz
frequency range ..................................................................................... 118
Table 5.6 Summary of averaged transmission loss across the entire 5 kHz
frequency range for the various prototype CPAP device muffler
design optimisations .............................................................................. 119
Table 6.1 Indicative flow velocity through muffler chambers at a volumetric

air flow rate of 120 L/min ..................................................................... 125
xxi
Table 6.2 Linear curve-fit data for predicting the self-generation of noise due
to various mufflers as a function of volume flow ................................. 132
Table B.1 Relationships for use of the revised model coefficients in the
original Delany-Bazley functions .......................................................... 166
Table B.2 Delany-Bazley function coefficients for the light and dark grey
foams (revised form) ............................................................................. 166
Table B.3 Delany-Bazley function coefficients for the light and dark grey
foams (derived)...................................................................................... 167
Table C.1 Optimisation results for the cylindrical expansion chamber reactive
muffler, comparing four optimisation algorithms ................................. 172
Table C.2 Optimisation results for the cylindrical expansion chamber muffler
with foam lining, comparing four optimisation algorithms .................. 177
xxii
List of Symbols
‫ܣ‬ǡ ‫ܤ‬ǡ ‫ܥ‬ǡ ‫ܦ‬ Transfer matrix parameters
‫ כܣ‬ǡ ‫ כ ܤ‬ǡ ‫ כ ܥ‬ǡ ‫כ ܦ‬ Transfer matrix parameters (improved method)
ܽ Acceleration
ܿ Complex speed of sound
‫ܥ‬ଵ െ ‫଼ܥ‬ Delany-Bazley function coefficients
ଵ െ ଵ଴ Design control variables
݀ Diameter
݂ Frequency in Hz
݂௖௢ Cut-on frequency
݂௜ Behaviour constraints
݂௢ Objective function
‫ܩ‬௫௫ Autospectrum of signal ‫ݔ‬
‫ܩ‬௫௬ Cross-spectrum of two signals ‫ ݔ‬and ‫ݕ‬
‫ܪ‬௡ Helmholtz number
‫ܪ‬ଵ Estimation of the frequency response function
‫ܪ‬ଵଶ Transfer function from microphone position 1 to position 2
݆ Imaginary unit ξെͳ
݇ Wave number
‫ܮ‬ Length
‫ܮ‬௄ Analogous acoustical inductance (Karal correction factor)
‫ܮ‬ௐ Sound power level
݉ Expansion ratio
‫ܯ‬ Mach number
‫݌‬ǡ ࢖ሺ‫ݔ‬ǡ ‫ݐ‬ሻ Acoustic pressure
xxiii
‫۾‬ Complex pressure (amplitude and phase is grouped)
Peak pressure amplitude
ο Static pressure drop
ܳ Volume flow
‫ݎ‬ Radius
‫ݎ‬௙ Airflow resistivity
ܴ Real component of characteristic impedance
ܴଶ Coefficient of determination
‫ݏ‬ Microphone separation distance
ܵ Cross-sectional area
‫ݐ‬ Time
‫ݐ‬௙ Thickness of foam sample
ܶ௧ Sound power transmission coefficient
‫ݑ‬ǡ ࢛ሺ‫ݔ‬ǡ ‫ݐ‬ሻ Particle velocity
ܷ Volume velocity
ܸ௠௔௫ Maximum allowable foam volume
ܹ Sound power
‫ݔ‬ሺ‫ݐ‬ሻ Upstream acoustic pressure time history
࢞ Vector of design control variables
ܺ Imaginary component of characteristic impedance
ܺሺ݂ሻ Fourier transform of ‫ݔ‬ሺ‫ݐ‬ሻ
ܺ ‫ כ‬ሺ݂ሻ Complex conjugate of ܺሺ݂ሻ
‫ݕ‬ሺ‫ݐ‬ሻ Downstream acoustic pressure time history
ܻሺ݂ሻ Fourier transform of ‫ݕ‬ሺ‫ݐ‬ሻ
‫ݖ‬ Characteristic impedance
ܼ Acoustic impedance
xxiv
ߙ Attenuation constant
ᇱ ሺߙ
ߙ௠௡ Roots of the Bessel function ௠ ௠௡ ሻ ൌ Ͳ
ߚ Phase constant
ߛ Propagation constant
ߣ Wavelength
ߩ Complex density
߮ Fractional weighting factor
߶ Radian phase angle
߱ Radian frequency
Subscripts
ܽ Air
ܾ Branch (resonator)
ܿ Chamber (expansion), cavity (resonator)
݀ Duct
݂ Foam
݅ Incident
݅݊ Inlet
݉ Muffler
‫ݐ݌݋‬ Optimum
‫ݐݑ݋‬ Outlet
‫݌‬ Pipe
‫ݐ‬ Transmitted
൅ǡ െ Direction of positive and negative travelling waves
xxv
Chapter 1: Introduction and Literature Review
Chapter 1
Introduction and Literature Review
1.1 Introduction
Sleep apnoea affects approximately 10-25% of the adult population and has strong links
to stroke and heart failure, as well as a strong association with obesity and diabetes [1].
The most effective therapy for treating this condition involves the application of an
elevated pressure to the airway during sleep. Compliance with this treatment however is
poor, with half of patients discontinuing therapy within the first year. A significant root
cause identified for such low compliance is the noise emitted by the flow generating
device [2]. This thesis investigates the prediction of the acoustic performance of mufflers
used in these medical respiratory devices and presents computational techniques to
facilitate optimisation of the performance of the muffler designs.
In this chapter, the physiology of sleep apnoea and how it is most widely treated is
discussed. Some muffler fundamentals and performance metrics are presented. A
literature review then follows which summarises previous work undertaken on the
analytical and numerical modelling of mufflers and recent efforts to combine numerical
modelling techniques with non-linear optimisation algorithms. The chapter concludes by
presenting the contribution to research and thesis layout.
1.2 Obstructive Sleep Apnoea
Obstructive sleep apnoea (OSA) is a medical condition whereby the smooth muscles of
the upper airway lose sufficient condition during sleep that the airway becomes
constricted, resulting in partial or complete obstruction of the airway [3]. OSA can be
successfully managed through the application of a positive pressure to the airway. This
elevated airway pressure is delivered via a mask using a flow generator comprising
power supply, motor, fan, humidifier and control circuitry. The most widely used flow
generator configuration maintains a constant airway pressure throughout the breathing
cycle and is referred to as a continuous positive airway pressure (CPAP) device. More
1
sophisticated configurations are able to vary the pressure between the inspiration phase
and expiration phase of the breathing cycle, or continuously monitor the breathing
pattern and automatically adjust the level of positive airway pressure.
CPAP devices function by using a small motor/fan unit to produce the elevated airway
pressure. The fan runs at speeds up to 30,000 rpm and, because of the high and varying
speed, generates noise which is annoying to the user and sleep partner. Noise from the
flow generator is controlled using mufflers situated in the air flow paths to and from the
fan.
Unobstructed breathing Obstructed breathing
Figure 1.1 Mechanism of obstructive sleep apnoea [4]
Figure 1.2 ResMed S9 continuous positive airway pressure (CPAP) device [4]
2
These devices typically operate in a bedroom environment which has a very low ambient
noise level. Noise reduction through the reliable prediction and optimisation of muffler
designs is an important issue in the development of these devices in order to improve the
user experience and maximise compliance with the CPAP therapy. A significant
challenge is to optimise the design of these mufflers within size, material and cost
constraints, while ensuring that the finished product complies with the various regulatory
codes which are applicable to medical devices.
1.3 Background on Mufflers
Mufflers can be broadly classified as being reactive or resistive (dissipative) although

designs often comprise elements from both categories. Reactive mufflers suppress noise
through the mechanism of reflection of acoustic energy while resistive mufflers absorb
acoustic energy by dissipating it into heat [5]. For mufflers of a similar size, reactive
mufflers tend to be more effective at lower frequencies while resistive mufflers are more
effective at higher frequencies. Typical reactive muffler components include expansion
chambers, resonant cavities, internal baffles, tuned branches, extended inlet and outlet
tubes and perforated tubes. Resistive mufflers utilise sound absorbing materials such as
glass fibre, wool or foams.
The acoustic performance of mufflers is commonly described using two, frequency

dependent quantities – the insertion loss (IL) and the transmission loss (TL). Insertion
loss considers the downstream sound power levels within a duct before (‫ܮ‬ௐଵ ) and after
(‫ܮ‬ௐଶ ) connection of a muffler, and is commonly used in industrial applications to
quantify the acoustic benefit obtained following muffler installation. This measure is able
to account for the in situ acoustic performance of the muffler. In addition, the effects of
insertion such as alteration of the source sound power and, where flow is present, the
self-generation of noise caused by the muffler, are also taken into account. While
insertion loss provides a measure of the acoustic performance of a muffler in a given
situation, it is unable to provide any comparison of acoustic performance between
mufflers installed in different configurations. In contrast, transmission loss considers the
sound power level of the acoustic wave incident on the muffler (‫ܮ‬ௐ௜ ) less that of the
wave transmitted from the muffler (‫ܮ‬ௐ௧ ). The measurement of transmission loss requires
an anechoic termination and does not account for the behaviour of the source.
Transmission loss thus depends only on the device and not on the device and its
3
installation. Although this independence from installation-specific influences makes

transmission loss the preferred performance metric for the comparison of muffler
designs, it is not generally sufficient to predict the actual performance of a given muffler
in a specific system.
Source ‫ܮ‬ௐଵ
‫ܮ‬ௐ௜ ‫ܮ‬ௐ௧
Source ‫ܮ‬ௐଶ
ܶ‫ ܮ‬ൌ ‫ܮ‬ௐ௜ െ ‫ܮ‬ௐ௧
ܹ௧
ܶ‫ ܮ‬ൌ െͳͲ ଵ଴ ൬ ൰
ܹ௜
‫ ܮܫ‬ൌ ‫ܮ‬ௐଵ െ ‫ܮ‬ௐଶ
Figure 1.3 Muffler acoustic performance metrics – insertion loss and transmission
loss
Noise is produced in the blower chamber of a CPAP device. The flow-generating fan, or
blower, is typically of the mixed-flow variety and may comprise multiple stages. The fan
performance varies throughout the respiration cycle and is also a function of the required
supply pressure and system flow resistance. Sound passing through the inlet muffler exits
into free space while sound passing through the outlet muffler exits into a long, flexible
hose. This hose is acoustically coupled via a mask to the user’s lungs. While the
termination of the inlet muffler could be assumed to be close to the characteristic
impedance, the termination of the outlet muffler is considerably more complicated and
will be influenced by physical properties of the hose and the user’s lung physiology. The
termination impedance will also vary throughout the respiration cycle.
The actual performance of the CPAP device mufflers will be affected by the in situ
source and termination impedances and is more accurately described by the insertion
loss. However, in order to predict the insertion loss using computational models, these
impedances must first be accurately defined. This is beyond the scope of this thesis.
4
Furthermore, replicating a chosen scenario with source and termination impedances for
experimental validation would be problematic. Hence, transmission loss has been
adopted as the performance metric in the modelling and optimisation in this thesis.
1.4 CPAP Device Mufflers
The mufflers that are incorporated into CPAP devices are very small, irregularly shaped
and often consist of several interconnected volumes. While they are predominantly
reactive, dissipative materials are often incorporated into the design to enhance the
acoustic performance and to extend the attenuated frequency range up to 5 kHz.
Figure 1.4 CPAP device chassis showing inlet mufflers (red), blower chamber
(green) and outlet muffler (yellow). Arrows indicate air flow path.
Two dimensionless numbers are useful for examining the aero-acoustic scale of muffler
and duct systems. The Helmholtz number, ‫ܪ‬௡ , provides a measure of the size of the
system as measured in sound wavelengths and gives an insight into the characteristics of
the sound propagation within the system [6]. The Helmholtz number is given by
5
ʹߨ݂‫ݎ‬
‫ܪ‬௡ ൌ ݇‫ ݎ‬ൌ (1.1)
ܿ
where ݇ is the wavenumber, ‫ ݎ‬is the equivalent radius of the duct, ݂ is the frequency and
ܿ is the speed of sound. An example of a current CPAP device chassis is shown in
Fig. 1.4. The three muffler chambers are highlighted and have dimensions in the range
30-90 mm while the interconnecting ducts are approximately 20 mm in diameter. The
maximum Helmholtz numbers for this design occur at the upper limit of the attenuated
frequency range and are 0.92 in the interconnecting ducts and 4.12 in the muffler
chambers. As non-plane waves start propagating at Helmholtz numbers approximately
equal to 1.84, the sound propagation in the interconnecting ducts will be essentially
planar while non-plane effects must be accounted for in the muffler chambers [7].
௨
The Mach number ‫ ܯ‬ൌ ௖
provides a measure of the air velocity ‫ ݑ‬in the system as
measured against the speed of sound and can be used to predict the impact of mean flow
on the attenuating performance of the muffler [6]. The maximum air delivery
requirement for the CPAP device shown in Figs. 1.2 and 1.4 is 110 L/min and all of this
air is required to pass through each of the mufflers in the CPAP device air path. The
smallest cross-section in the muffler assembly occurs in the interconnecting ducts and
results in a maximum air velocity of 7 m/s, or a Mach number of 0.02. As this is small
(<<0.1) the mean flow effects will be negligible [8].
CPAP device muffler chambers are often integrated into the design of the device chassis
rather than in the form of discrete components. The chassis shown in Fig. 1.4 is formed
from thin-walled plastic and adjacent muffler chambers share common walls. The effect
of through-wall sound transmission between chambers due to fluid-structure interaction
may be significant. Similarly, the effect of sound radiation from the chassis structure to
the surrounding environment will also have a bearing on the acoustic performance of the
device.
1.5 Literature Review
A large amount of work has been published on the prediction of muffler performance
since the 1950s encompassing one-dimensional analytical models, experimentally
6
derived empirical approximations and three-dimensional numerical approaches. Much of

the earlier literature has been documented by Jones [9] and Munjal [10]. In more recent
years there has also been an increasing amount of literature on the application of
computational design optimisation techniques in the field of muffler acoustics.
1.5.1 Analytical approaches
One of the earliest examples of substantial work on muffler design is the NACA report
by Davis et al. [11] who conducted a thorough investigation into the attenuation
characteristics of 77 different mufflers. They used transmission line theory (also known
as the impedance method) to derive equations for the attenuation of expansion chamber
and resonator mufflers. Theoretical predictions were compared against experimental data
for single chamber and multi-chamber muffler designs which included expansion
chambers, resonant cavities, and combinations of these elements.
The most common type of linear acoustic model used to predict the performance of
mufflers applies classical equivalent electrical filter (network) theory and is most widely
known as the transfer matrix method, although it is also referred to as the two port
approach or 4-pole parameter method [10, 12]. Karal [13] determined the equivalent
acoustic element to correct for discontinuities and constriction of circular cross sections.
Igarashi et al. [14, 15] derived the 4-pole parameters for common muffler components
including single and series expansion chamber elements, a side branch, resonator, an
expansion chamber element with extended inlet and outlet tubes and a Quincke tube
element. They observed discrepancies between the measured 4-pole parameters and those
estimated from the one dimensional theory even for the simplest cavity type element.
Davies [16] reviewed the equivalent electrical network approach and provided a good
discussion of aspects of the underlying physics which should be considered when
deciding the suitability of this approach to a particular muffler application.
As muffler design elements become more complex, the electro-acoustic analogy

approach has been combined with empirical approximations. Sullivan [17, 18] presents a
method for modelling a perforated tube which is able to describe the nonlinear
impedance of the perforation. The effect of mean flow on the impedance was also
discussed.
7
The flow of air through a muffler affects the acoustic performance of the system by
means of convective, refractive, and flow-induced mechanisms [10, 19]. The convection
and refraction mechanisms influence the propagation of acoustic waves and are
associated with the bulk movement of the fluid and the formation of shear layers,
respectively. The convective effect is important at all frequencies while the refractive
effect is strongest at the higher frequencies. The flow-induced effects result in the self-
generation of noise and are associated with the formation of unsteady vortex fields, such
as those produced by free jet turbulence and flow separation. The flow induced losses are
most important at low frequencies and low Mach numbers. Alfredson and Davies [20]
examined the performance of area discontinuities and branched systems in internal
combustion engine exhaust silencers. Based on their theoretical predictions and
measurements, they concluded that neglecting the mean flow would lead to an over-
estimation of the attenuation produced by a muffler. However, several authors have
derived transfer matrices for a range of muffler elements in the presence of mean flow
and found that the effect of flow on the performance of most elements was only slight for
Mach numbers normally encountered in exhaust mufflers [21-23]. Of the elements
examined, Munjal [21] found that only extended inlet and branch resonators would be
appreciably different and, even then, only at or around the tuned frequency. Fakuda et al.
[24] found that the flow-acoustic interactions are normally negligible if the mean flow
velocity is less than 10 m/s (or less than a Mach number ‫ ܯ‬of approximately 0.03). Ih
and Lee [8] state that when the mean flow velocity is less than Mach 0.1, the convective
contributions can be considered to be negligible second order quantities and flow-
generated noise may be disregarded. Peat [25] revisited the work of Karal to examine the
impact of mean flow on the acoustical impedance at duct discontinuities and found that
mean flow did not significantly affect the impedance at velocities less than Mach 0.3.
The earlier literature often makes the assumption that only plane waves propagate within
the muffler [10]. While this simplification reduces the complexity of the algebra
required, the derived one-dimensional equations are unable to accurately predict muffler
performance when higher order modes are present [26, 27]. These equations are thus
unreliable at higher frequencies and for analysing designs having larger cross sections
and/or geometric complexity [28]. While this weakness greatly limits the usefulness of
the analytical approaches in the context of commercial muffler designs, they remain
attractive due to their simplicity and ease of use when compared to the computationally
8
intensive numerical approaches. To overcome the plane wave limitation, several authors
have investigated the effects of higher order modes on the performance of expansion
chamber designs [10, 26, 29-32]. Åbom [26] derived the 4-pole parameters for expansion
chambers with extended inlet/outlet ducts which include the higher order mode effects.
This derivation clearly illustrates the complexity of the algebra required. Munjal [33]
presents a collocation method which attempts to address the plane wave limitation by
evaluating the relative amplitudes of the various modes but concludes that while the
method is relatively simple, it is difficult to apply to irregular shapes. A further
disadvantage of the transfer matrix approach is the difficulty encountered when trying to
manage non-sequential (parallel or multiply connected) elements. Although several
different approaches to address the problem have been presented it remains a challenge
[34-36].
1.5.2 Numerical approaches
Using analytical models, muffler designs are limited to combinations of basic elements.
Modelling of complex muffler features can be achieved using numerical discretisation
techniques. The most widely reported numerical approaches used to predict the
performance of mufflers include the finite element method (FEM), the boundary element
method (BEM), and computational fluid dynamics (CFD). The finite difference method
was also initially investigated for the prediction of muffler acoustics but was found to be
computationally inefficient and expensive compared to the finite element method
[37, 38]. A comparison of the various numerical methods has been given by Bilawchuk
and Fyfe [39] who concluded that the combination of the FEM and 3-point method of
Wu and Wan [40] was advantageous over other considered methods.
One of the earliest applications of the finite element method to predict the acoustic
performance of mufflers is the work of Young and Crocker [28], who applied the
variational formulation of Gladwell [41] to the modelling of a simple cylindrical
expansion chamber. Craggs [42] subsequently improved the flexibility of the method by
formulating an isoparametric hexahedral element which allowed the method to be applied
to mufflers having non-rectangular sections. The modelling results were compared to
those predicted using the electro-acoustic analogy and it was concluded that the FEM
was accurate for transmission loss analysis provided that the finite element mesh size
was suitably refined. Furthermore, Craggs noted that the FEM was flexible in its
9
application and allowed complex systems to be easily represented. Barbieri et al. [43]
combined the Galerkin FEM with the improved 4-pole parameter method described by
Wu et al. [44] to predict the performance of expansion chambers. They conducted a
review of sources of error in the FEM and found their results to be very close to those
obtained previously for the same muffler geometry by Selamet and Ji [45] using the
BEM.
As the finite element method involves discretisation of the entire computational domain,
different element types may be used throughout a non-homogeneous domain to facilitate
the modelling of absorptive materials, mean flow and mean temperature gradient
[39, 46]. Peat and Rathi [47, 48] successfully added mean flow to a finite element model
of a bulk reacting dissipative silencer. The perforated pipe separating the air from the
porous material was not explicitly modelled on the assumption that the porosity of such a
screen is so large that its acoustic effect is negligible. There was little difference found
between the transmission loss results calculated using an assumed uniform axial flow and
those calculated using a more realistic computed flow field. In contrast to the finite
element method, the boundary element method requires that only the perimeter of the
domain be discretised, leading to a reduction in the number of equations to be solved.
However, the formulation of the BEM coefficient matrix is less structured than that of
the FEM and can actually lead to greatly increased computational time as the number of
nodes increases [39, 49]. Harari and Hughes [49] concluded that finite element methods
are robust and more economical on most practical interior configurations.
One of the first applications of the boundary element method to muffler analysis is by
Seybert and Cheng [50, 51]. Ji et al. [52] extended the BEM by incorporating the
convective effects of mean flow. Wu et al. [53] used a direct mixed-body boundary
element method to predict the attenuation of packed silencers including perforated plates
and absorbing material. Their method eliminated the requirement to subdivide the
volume into different regions for each acoustic media (air and absorbent). Selamet et al.
[54] investigated a hybrid system that comprised two identical dissipative silencers with
a reactive chamber in between, combining the effects of a resonant cavity, absorptive
material and a perforated tube. The combination of elements was found to increase the
transmission loss at low frequencies while high duct porosity can improve the acoustic
behaviour of dissipative silencers at high frequencies.
10
Literature pertaining to the application of computational fluid dynamics to the modelling

of mufflers reveals that this method is almost exclusively used to assess the mean flow
behaviour and to predict pressure losses through the muffler system. The CFD method is
considerably disadvantaged over the finite and boundary element methods due to the
requirement for very fine computational mesh, long computational run times required for
transient analyses and large storage requirements. However, this additional effort may be
justified in situations where it is necessary to consider strong non-linear and/or mean
flow effects. Cooke [55] used the energy efflux time history to predict the noise
attenuation of a multi-chambered muffler attached to an auto-cannon. The Euler
equations of compressible flow were solved in one dimension and the computational
phase of the analysis took several days to complete using a supercomputer. Jebasinski
and Eberspacher [56] used a one-dimensional commercial software program, Ricardo
Wave, to calculate the transmission loss of the simple expansion chamber muffler
previously modelled by Selamet and Radavich [27]. However, the software is unable to
simulate absorptive material and uses an undisclosed “YJunction” element to simulate
three-dimensional wave behaviour using the one-dimensional CFD code. The most
comprehensive implementation of the CFD approach to date with regard to muffler
acoustics is by Middelberg [57]. A three-dimensional unsteady Reynolds-averaged
Navier-Stokes (U-RANS) approach was used to examine both the acoustic and mean
flow performance of several expansion chamber mufflers. The numerical results were
found to show good agreement with published experimental results, especially at
frequencies below 1 kHz. More recently, Obikane [58] calculated the transmission loss of
an internally baffled muffler incorporating perforated pipe using the exact compressible
Navier-Stokes equation at two forced oscillation frequencies. The calculated transmission
loss was similar to experimental results obtained by Gerges et al. [59].
1.5.3 Porous materials
The acoustic performance of mufflers is generally enhanced by the inclusion of

dissipative materials. Porous media are typically modelled as an equivalent fluid with
either a rigid or limp frame, or using a poroelastic model based on the theory of Biot
[60, 61]. Equivalent fluids may be described by a few complex and frequency dependent
material parameters corresponding to the speed of sound, mean density, characteristic
impedance and propagation constant.
11
Foundation work on the theoretical approach to describe sound propagation in porous

materials was presented by Zwikker and Kosten [62], who used a continuum model and
introduced the concept of effective density and bulk modulus. Biot [60, 61] incorporated
the effects of frame elasticity, where the skeleton of the material supports coupled
interaction between the solid and fluid phases. A key element of Biot’s work was the
identification of the existence of three types of sound wave which simultaneously
propagate within continuous materials: two compression (longitudinal) waves and one
shear (transverse) wave. The complete description of poroelastic dynamics in Biot’s
theory requires knowledge or measurement of nine material parameters corresponding to
Young’s modulus, density, loss factor, Poisson’s ratio, porosity, tortuosity, viscous
length, thermal length and flow resistivity. Biot’s theory is valid for any porous material
at any frequency. The model is capable of predicting resonant behaviour of the material
structure and is well suited to the modelling of polyurethane foams. However, the
mathematical complexity and computational requirements associated with Biot’s model
are significant. Lambert [63] studied the work of Biot [60], Zwikker and Kosten [62] and
Beranek [64] to formulate a theoretical approach to model sound propagation in low and
medium airflow resistance foams. It was predicted that the coupling between the elastic
frame wave and sound waves in the fluid could be ignored at frequencies greater than
10 Hz. This work was extended by Allard et al. [65] to high flow resistance foams.
Bolton et al. [66] applied Biot’s theory to the prediction of transmission loss through
double panels lined with polyurethane foam. The ability of the two-dimensional model to
account for the three wave types was demonstrated by comparing bonded and unbonded
methods of mounting the foam.
Morse and Ingard [67] developed generic acoustic models for rigid and limp frame
porous materials. Rigid frame models assume that the material is formed by a number of
pores within a motionless solid. The packing geometry of the structure determines the
fluid dynamics in the interstitial spaces and the attenuation of acoustic waves. Models of
rigid behaviour are more appropriate to describe stiff materials such as glass fibres or
foams made from plastic or metal. Limp frame models assume that the stiffness of the
solid phase can be neglected but incorporate the influence of the moving frame through
the addition of inertial loading on the acoustic wave. Models of limp behaviour are better
suited to soft porous materials such as woven wool and cotton fabrics. It is important to
note that the choice of model is frequency dependent, as the solid phase stiffens with
12
frequency due to inertia effects. Allard and Champoux [68] used the general frequency
dependence of the viscous forces in porous materials proposed by Johnson et al. [69] to
produce expressions incorporating five macroscopic properties of the porous material,
corresponding to porosity, tortuosity, viscous length, thermal length and flow resistivity.
Delany and Bazley [70] provided the first empirical equations for equivalent fluids based
on measured acoustic properties. Measured values of characteristic impedance and
propagation coefficient for a range of fibrous materials, normalised as a function of
frequency and divided by flow resistance, were presented as simple power law functions.
Miki [71] found that the Delany-Bazley model produced a non-physical prediction at low
frequencies and amended the original equation regression coefficients. Further work was
done by Bies and Hansen [72] and Mechel [73] to correct and extend the Delany-Bazley
method beyond the bounds recommended by the original authors. Attenborough [74]
observed that the normalising parameter used by Delany and Bazley appeared in the
theoretical expressions for any pore shape and concluded that empirical relationships of
the form proposed by Delany and Bazley should be valid for non-fibrous porous
materials. He did however also note that frame elasticity presents an additional
complication and that the coefficients in the model would be unique to each type of
porous material. The approach used by Delany and Bazley has since been applied by
others to a range of polyurethane foams and has resulted in the derivation of further sets
of regression coefficients [75-77]. Kidner and Hansen [78] reviewed several porous
material models and concluded that, although the Biot model provides a more complete
description of poroelastic dynamics, rigid frame models and the Delany-Bazley empirical
approach are applicable for many noise control materials. Figure 1.5 lists the material
parameters used in the various porous material models.
1.5.4 Design optimisation
The past decade has seen an increase in the number of publications combining the use of
non-linear optimisation algorithms with numerical modelling approaches to optimise the
acoustic performance of mufflers. The majority of the literature describes either a
parametric or shape optimisation approach; however there is a growing body of work that
has been conducted using topology optimisation.
13
• Young’s Modulus Biot

• Density
• Loss Factor
• Poisson’s Ratio
• Porosity Johnson-Champoux-Allard
• Tortuosity
• Viscous Length
• Thermal Length
• Flow Resistivity Delany-Bazley
Figure 1.5 Comparison of material parameters used in three porous material models
One of the pioneering works in parametric muffler optimisation is that of Bernhard [79],
who formulated an approach using transfer matrix analysis to obtain design sensitivity
information for mufflers. The insertion loss of side-branch and Helmholtz resonators and
an expansion chamber muffler was optimised using two-dimensional finite element
models at a single frequency. It was acknowledged that practical design problems would
require calculation of the objective function across a broad frequency range, resulting in
long analysis times.
Chang et al. [80-84] have applied a common approach in their work to optimise muffler
designs comprising varying combinations of single and multiple expansion chambers,
extended inlets and outlets, and perforated tubes in cross flow. The models are assembled
using the transfer matrix method and are based on the analytical plane wave matrix
representation of each component. The models also allow for the inclusion of mean flow.
Optimisation is performed using either a mathematical gradient method, a genetic
algorithm (GA) or simulated annealing (SA) approach. The effect of altering the control
parameters of the GA and SA methods is assessed and discussed in detail.
Seo and Kim [85] combined the electro-acoustic analytical approach with the transfer
matrix method to optimise serial and parallel arrangements of resonators along a duct.
Rather than seek to maximise the peak transmission loss, they aimed to broaden the
14
narrow band characteristics of the array in the low frequency range. This was achieved
by defining the objective function to be the difference between the maximum and
minimum transmission loss values calculated within the target frequency range, subject
to geometric constraints. Barbieri and Barbieri [86] used an axi-symmetric finite element
model combined with the improved four parameters method to conduct parametric
optimisation of a cylindrical expansion chamber with extended inlet and outlet ducts. The
problem was solved using Zoutendijk’s feasible directions method [87], as it was
considered to be robust and efficient for problems with non-linear constraints.
More recently, de Lima et al. [88] performed parametric and shape optimisations on
cylindrical single and double expansion chamber mufflers having extended inlet/outlet
ducts. The parametric optimisation was used to vary the extended lengths of the
inlet/outlet ducts to enhance attenuation at the first two resonant frequencies of the empty
cavity in the absence of extended inlet/outlet ducts. The profiles of the extended sections
of the inlet/outlet ducts were approximated using cubic curves. Shape optimisation was
then used to vary the curve profile in order to modify the transmission loss in specific
frequency ranges. The genetic algorithm was used to optimise the objective function and
the authors recommended the use of this algorithm for similar work. The final designs
were measured experimentally and the results showed good agreement with numerical
results.
Although literature describing topology optimisation appeared in the late 1980s, it was
almost exclusively focused on the solution of structural optimisation problems [89]. In
acoustic topology optimisation, two materials (air and solid) are distributed throughout
the design domain in such a way that an objective function is minimised. A design
variable is introduced which facilitates the continuous distribution of density and bulk
modulus between the two extremes of air and solid [90]. Various strategies exist to
ensure binary values of this design variable in the final solution [91]. As topology
optimisation is based on element-wise parametrisation of material properties – in contrast
to the parametrisation of specific geometric features used in shape optimisation – this
approach provides great design freedom and does not rely on an intuition-based initial
geometry.
Jensen and Sigmund [90] first presented the application of the topology optimisation
approach in muffler acoustics. They optimised a channel with two symmetrically placed
15
reflection chambers over a set of discrete target frequencies with an optimisation

objective of minimising the largest of these values. Duhring et al. [92] modelled three-
dimensional room acoustic problems and outdoor sound barriers and solved the
optimisation problem using the gradient-based method of moving asymptotes [93].
Although not strictly muffler related, the paper presents a clear discussion of the
methodology and discusses difficulties which are relevant to muffler optimisation
applications. The solution was found to be sensitive to the initial choice of volume
fraction. Difficulty was encountered in obtaining a solution that was independent of the
mesh and consisted only of air or solid regions. The presence of small details, holes or
regions of solid material which are unconnected to any surface result in designs which
would be impractical to manufacture. An image morphology-based filter can be
employed to eliminate these problems but adds further complexity to the solution [94]. It
was found that the method was not suitable for high frequencies due to difficulty
obtaining a sufficiently fine finite element mesh resolution and the existence of many
local minima. Lee and Kim [95] applied the topology optimisation approach to the layout
of partitions inside an expansion chamber muffler. They used a two-dimensional model
with planar symmetry and a different interpolation function to facilitate continuous
material properties.
1.6 Contribution to Research
The objectives of this thesis are to develop computational and experimental techniques to
predict and measure the acoustic performance of mufflers for respiratory medical
applications, as well as to develop computational techniques to facilitate optimisation of
the performance of the muffler designs. Specifically, the research relates to the reduction
of noise from sleep apnoea devices and will contribute to knowledge by producing:
• computational models to predict the acoustic performance of irregular shaped

mufflers such as those used for sleep apnoea devices;
• a methodology to efficiently optimise muffler configurations and internal surface
properties, that is, both reactive and resistive elements, to maximise acoustic
performance; and
• an experimental procedure to measure muffler acoustic performance in the
presence of mean flow.
16
While the literature reveals that considerable research has been carried out on the
prediction of the acoustic performance of mufflers, the focus has been almost exclusively
on industrial applications including automotive exhausts, reciprocating plant, and air/gas
handling systems. Literature pertaining to the computational optimisation of muffler
designs has only recently started to emerge and tends to be limited to simple reactive
geometries. Respiratory medical devices that are used for the treatment of obstructive
sleep apnoea present some unique challenges that set them apart from many of the
industrial muffler applications that have been studied to date. These challenges include:
• the need for very small scale mufflers due to size constraints;
• the need to consider muffler acoustic performance over a wide frequency range –
up to at least 5 kHz – thus requiring optimal integration of both reactive and
resistive components;
• the inability to regard all surfaces to be acoustically hard due to the use of plastics
in the manufacture of the expansion chamber and the use of flexible inlet and
outlet piping; and
• the requirement to consider the impacts of flow induced noise due to the presence
of small scale components, high expansion ratios and variable air flow.
The work contained in this thesis accounts for the effects of higher order modes by using
a three-dimensional finite element approach and presents the simultaneous optimisation
of reactive and resistive design features. The optimised muffler design is more intricate
than those presented in the previous papers (Refs. 80-88) and complex inter-relationships
exist between multiple parametric control variables and the underlying muffler geometry.
This thesis also presents a comparison of four metaheuristic optimisation algorithms for
use in optimising the acoustic performance of mufflers.
Although the focus of this project is on the reduction of noise from sleep apnoea devices,
the outcomes have relevance to the design of mufflers for a broad range of applications.
For example, a tool that can predict and optimise the acoustic performance of irregular
shaped mufflers would be of benefit where industrial mufflers must be incorporated into
irregular and/or confined spaces. This may allow increased design flexibility in
applications such as locomotives and high speed ferries.
17
1.7 Thesis Overview
Chapter 1 presents an introduction to the physiology of sleep apnoea, muffler

fundamentals and acoustic silencer performance metrics. A literature review then follows
which summarises the prediction and optimisation of muffler acoustics. The literature is
separated into four parts corresponding to analytical approaches, numerical approaches,
porous materials and design optimisation.
In Chapter 2, the transmission loss of a reactive expansion chamber muffler is predicted

analytically, computationally and experimentally. Three analytical approaches are
reviewed, corresponding to the continuity of pressure and volume velocity method, the
impedance method (also known as transmission line theory) and the transfer matrix
method. A finite element analysis model of the muffler is also developed. Results of the
muffler transmission loss obtained computationally are compared with those obtained
from the established analytical techniques. The transmission loss of the muffler was
measured using a two-microphone acoustic pulse method. Experimental results are
compared with the analytical and computational results.
Chapter 3 incorporates polyurethane foam into the muffler designs presented in

Chapter 2. A modification to the Delany-Bazley empirical expressions is presented which
facilitates the acoustic characterisation of dissipative material without requiring
knowledge of the airflow resistivity. The acoustic characteristics of two polyurethane
foams were obtained experimentally using a two-cavity impedance tube method and the
corresponding properties were incorporated into finite element models. Results of the
transmission loss of the foam-filled muffler obtained from the finite element models are
presented and compared with results obtained experimentally.
In Chapter 4, the computational methodologies described in Chapters 2 and 3 are

extended by applying these techniques to the more complex geometries required for sleep
apnoea assisted breathing devices. The transmission loss of three CPAP device muffler
designs is predicted analytically, computationally and experimentally. Finite element
models of each reactive muffler are developed and evaluated against results obtained
using the analytical techniques presented in Chapter 2. The acoustic characteristics of the
two polyurethane foams presented in Chapter 3 are incorporated into finite element
models of two of the CPAP muffler designs. The transmission loss of each resistive
18
muffler was measured experimentally and the results compared with those obtained
computationally.
Chapter 5 investigates computational optimisation of two muffler designs corresponding

to a cylindrical expansion chamber muffler and a prototype CPAP device muffler. The
reactive and resistive components of the mufflers were independently optimised for
acoustic performance. Results were also obtained for the transmission loss of the
prototype CPAP device muffler for which the reactive and resistive components were
simultaneously optimised.
In Chapter 6, the transmission loss of four muffler designs corresponding to two

cylindrical expansion chamber mufflers and two CPAP device mufflers is measured in
the presence of mean flow. The performance of two of the muffler designs with
polyurethane foam inserts fitted is also measured. The transmission loss in the presence
of mean flow is compared with experimental results obtained in the absence of flow. The
influence of blower and aeroacoustic noise on the experimental results is discussed.
Chapter 7 presents a summary of the major findings of this research and offers
recommendations for future work.
19
Chapter 2: Transmission Loss of a Reactive Muffler
Chapter 2
Transmission Loss of a Reactive Muffler
2.1 Introduction
In this chapter, the transmission loss of a reactive expansion chamber muffler is predicted
analytically, computationally and experimentally. A finite element model of the muffler
is developed and compared with results obtained from established analytical techniques.
Two acoustic finite element analysis approaches are presented which include a direct
calculation of the attenuation of acoustic power, and an indirect approach where transfer
matrix parameters are obtained from the finite element solution and subsequently used in
a transmission loss calculation. While the latter approach requires more effort, it has the
advantage of obtaining parameters which characterise the muffler behaviour and can be
used in further modelling without the need to repeat the finite element analysis.
Experiments are conducted based on the two-microphone acoustic pulse method for
comparison with both the analytical and computational results.
2.2 Analytical Methods
Three analytical approaches are reviewed, corresponding to the continuity of pressure

and volume velocity method, the impedance method (also known as transmission line
theory) and the transfer matrix method. These methods are based on one-dimensional
plane wave theory and are limited in the range of frequencies over which they are valid.
The practical application of these methods is restricted to simple muffler geometries.
For one-dimensional plane wave theory, the cut-on frequency ݂௖௢ at which the first non-
plane wave propagates in a circular pipe of diameter ݀ is given by [10, 67]
ߙ௠ǡ௡ ܿ
݂௖௢ ൌ ቀ ቁǡ ߙ௠ǡ௡ ൌ ߙଵǡ଴ ൌ ͳǤͺͶͳ͵ (2.1)
ߨ ݀
ᇱ
where ߙ௠ǡ௡ are the roots of the Bessel function ௠ ൫ߙ௠ǡ௡ ൯ ൌ Ͳ. The modes are designated
by ሺ݉ǡ ݊ሻ, where the integer ݉ indicates the number of diametrical nodal lines and the
20
integer ݊ indicates the number of annular nodal circles. Equation (2.1) corresponds to
excitation of the first diametral mode (1, 0) within the pipe. The next higher-order mode
starts propagating when ߙ ൌ 3.8318 and corresponds to the first circumferential mode
௖
(0, 1) [27]. Given that ߣ ൌ ௙ where ܿ is the speed of sound in the fluid and ߣ is the
wavelength, the transition can also be expressed as ݀௖௢ ൌ ͲǤͷͺ͸ߣ. If the pipe diameter is
greater than ݀௖௢ then the plane wave assumption is no longer valid, that is, the waves are
not truly confined in the pipe but instead behave as though they were travelling in an
unbounded medium. Considering air as the working fluid, assuming a constant ambient
temperature of 20oC, and expressing the pipe diameter in units of metres, the plane wave
ଶ଴ଵ
frequency range is limited to ݂ ൏ ௗ .
೘ೌೣ
For cross-sectional dimensions commonly associated with automotive applications, one

dimensional analysis is only applicable for frequencies less than 1 kHz (even after
considering the effects of higher gas temperatures). However, as the internal cavities in
CPAP devices are significantly smaller than those found in automotive mufflers, the
range of frequencies where the plane wave assumption can be considered to be valid will
extend to higher frequencies. CPAP device mufflers tend to have characteristic
dimensions between 30 mm and 90 mm. While the plane wave cut-on frequency for the
smaller cavities is approximately 5 kHz, the inclusion of the larger cavities in the air path
limits the range of plane wave validity for the muffler system to frequencies below
approximately 2.2 kHz. Although higher order modes start to propagate at frequencies
greater than the cut-on frequency of the larger cavities, energy propagation still remains
dominated by the energy contained in the plane waves up to a frequency of about twice
the cut-on frequency of the first higher order mode.
2.2.1 Single chamber muffler
Figure 2.1 shows one dimensional harmonic plane acoustic waves of angular frequency
߱ and wave number ݇ propagating in a simple expansion chamber reactive muffler of
cross sectional area ܵ஻ . The outlet pipe is considered to be anechoically terminated. The
complex representation of the acoustic pressure is obtained by the superposition of the
acoustic pressures associated with the positive and negative travelling one dimensional
21
plane acoustic waves and is given by [10]
࢖ሺ‫ݔ‬ǡ ‫ݐ‬ሻ ൌ ‫۾‬ା ݁ ௝ሺఠ௧ି௞௫ሻ ൅ ‫ ݁ ି۾‬௝ሺఠ௧ା௞௫ሻ (2.2)
The amplitude and phase information has been grouped as ‫ ۾‬ൌ ݁ ௝థ . ‫۾‬ା and ‫ ି۾‬are
respectively the complex pressures of the positive and negative travelling waves. The
corresponding particle velocities associated with the positive and negative travelling
waves are related to the acoustic pressures by the characteristic impedance ߩܿ, where ߩ is
the density of the fluid. Thus the complex representation of the particle velocity is given
by [10]
‫۾‬ା ௝ሺఠ௧ି௞௫ሻ ‫ ି۾‬௝ሺఠ௧ା௞௫ሻ

࢛ሺ‫ݔ‬ǡ ‫ݐ‬ሻ ൌ ݁ െ ݁ (2.3)
ߩܿ ߩܿ
The sound transmission loss, TL, can be found from ܶ‫ ܮ‬ൌ െͳͲ ଵ଴ ܶ௧ where ܶ௧ is the
sound power transmission coefficient and is defined as the ratio of the power of the
transmitted wave to the power of the incident wave. Hence, for the muffler shown in
Fig. 2.1, TL can be obtained as
ȁ‫۾‬௧ଶ ȁ ܵ஼
ܶ‫ ܮ‬ൌ െͳͲ ଵ଴ ቈ ଶ ቉ (2.4)
ห‫۾‬௜ ห ܵ஺
where ܵ஺ and ܵ஼ are respectively the cross sectional area of the pipes at the muffler inlet
and outlet.
ܵ஺ ܵ஻ ܵ஼
‫ܘ‬஻ା
‫ܘ‬௜
‫ܘ‬௧
‫ܘ‬௥ ͳ ʹ ͵ Ͷ
‫ܘ‬஻ି
‫ܮ‬
‫ݔ‬
Figure 2.1 A cylindrical expansion chamber reactive muffler
22
2.2.1.1 Continuity of pressure and volume velocity method
The continuity method is based on the premise that acoustic pressure and volume
velocity are maintained at a change of section [10]. At x = L, equality of acoustic
pressure and volume velocity leads to
‫۾‬஻ା ݁ ି௝௞௅ ൅ ‫۾‬஻ି ݁ ௝௞௅ ൌ ‫۾‬௧ ݁ ି௝௞௅ (2.5)
‫۾‬஻ା ି௝௞௅ ‫۾‬஻ି ௝௞௅ ‫۾‬௧

ܵ஻ ݁ െ ܵ஻ ݁ ൌ ܵ஼ ݁ ି௝௞௅ (2.6)
ߩܿ ߩܿ ߩܿ
These two equations can be used to determine ‫۾‬஻ା and ‫۾‬஻ି in terms of ‫۾‬௧ .
ͳ ܵ஼
‫۾‬஻ା ൌ ൬ͳ ൅ ൰ ‫۾‬௧ (2.7)
ʹ ܵ஻
ͳ ܵ஼
‫۾‬஻ି ൌ ൬ͳ െ ൰ ‫۾‬௧ ݁ ି௝ଶ௞௅ (2.8)
ʹ ܵ஻
At the change of section at ‫ ݔ‬ൌ Ͳ, equality of acoustic pressure and volume velocity
respectively lead to
‫۾‬௜ ൅ ‫۾‬௥ ൌ ‫۾‬஻ା ൅ ‫۾‬஻ି (2.9)
ܵ஺ ܵ஻
ሺ‫۾‬௜ െ ‫۾‬௥ ሻ ൌ ሺ‫ ۾‬െ ‫۾‬஻ି ሻ (2.10)
ߩܿ ߩܿ ஻ା
Using Eqs. (2.9) and (2.10), ‫۾‬௜ can be expressed in terms of ‫۾‬஻ା and ‫۾‬஻ି as
ͳ ܵ஻ ͳ ܵ஻
‫۾‬௜ ൌ ൬ͳ ൅ ൰ ‫۾‬஻ା ൅ ൬ͳ െ ൰ ‫۾‬஻ି (2.11)
ʹ ܵ஺ ʹ ܵ஺
Substitution of the expression for ‫۾‬஻ା and ‫۾‬஻ି given by Eqs. (2.7) and (2.8) into
23
Eq. (2.11) and rearranging allows ‫۾‬௜ and ‫۾‬௧ to be related by
‫۾‬௧ Ͷ
ൌ (2.12a)
‫۾‬௜ ሺܴଵ ൅ ܴଶ ሻ ൅ ሺܴଵ െ ܴଶ ሻ݁ ି௝ଶ௞௅
or rearranged as
‫۾‬௧ ʹሺሺ݇‫ܮ‬ሻ ൅ ݆ ሺ݇‫ܮ‬ሻሻ

ൌ (2.12b)
‫۾‬௜ ܴଵ ሺ݇‫ܮ‬ሻ ൅ ݆ܴଶ ሺ݇‫ܮ‬ሻ
ௌ ௌ ௌ
where ܴଵ ൌ ͳ ൅ ௌ಴ and ܴଶ ൌ ௌ಴ ൅ ௌಳ . For the ideal case where no energy loss occurs at
ಲ ಳ ಲ
points of reflection, incident and reflected waves will be in phase and Eq. (2.4) can be
used to obtain the transmission loss as
ܵ
Ͷ ቀܵ஼ ቁ
஺
ܶ‫ ܮ‬ൌ െͳͲ ଵ଴ ൦ ଶ ൪ (2.13)
ܴଵ ଶ ሺ݇‫ܮ‬ሻ ൅ ܴଶଶ ଶ ሺ݇‫ܮ‬ሻ
For the common case where inlet and outlet pipes have the same cross-sectional area
ሺܵ஺ ൌ ܵ஼ ሻ, Eq. (2.13) can be written as
ଶ
ଶ ሺ݇‫ܮ‬ሻ
ͳ ͳ
ܶ‫ ܮ‬ൌ ͳͲ ଵ଴ ቈ ൅ ൬ ൅ ݉൰ ଶ ሺ݇‫ܮ‬ሻ቉ (2.14)
Ͷ ݉
ௌ
where ݉ ൌ ௌಳ .
ಲ
2.2.1.2 Impedance method
The impedance method (also known as transmission line theory) is based on the premise
that equality of acoustic impedance ܼ is maintained at a change of section. Development
of the final equation for transmission loss requires calculating the impedance and
pressure at each of the locations 1 to 4 as identified in Fig. 2.1. With an anechoic
termination at the outlet duct, the specific acoustic impedance at point 4 is equal to the
characteristic impedance ሺ‫ݖ‬ସ ൌ ߩܿሻ and the acoustic pressure at point 4 is equal to the
transmitted pressure (‫۾‬ସ ൌ ‫۾‬௧ ).
24
Impedance calculations proceed from the outlet (point 4) to the inlet (point 1). The
acoustic impedance at point 4 is given by
‫ݖ‬ସ ߩܿ
ܼସ ൌ ൌ (2.15)
ܵ஼ ܵ஼
The acoustic pressures and volume velocities at points 3 and 4 are equal, hence the
acoustic impedances at these points are equal ሺܼଷ ൌ ܼସ ሻ. The specific acoustic
impedance at point 3 is then given by ‫ݖ‬ଷ ൌ ܼଷ ܵ஻ or
ܵ஻
‫ݖ‬ଷ ൌ ߩܿ ൬ ൰ (2.16)
ܵ஼
The specific acoustic impedance at point 2 can be found from that at point 3 by
considering the impedance formula for undamped plane acoustic waves in a gas column
in what follows. At‫ ݔ‬ൌ ‫( ܮ‬corresponding to point 3 in Fig. 2.1), the specific acoustic
impedance is given by
࢖ଷ ‫۾‬஻ା ݁ ି௝௞௅ ൅ ‫۾‬஻ି ݁ ௝௞௅

‫ݖ‬ଷ ൌ ൌ ߩܿ ቆ ቇ (2.17)
࢛ଷ ‫۾‬஻ା ݁ ି௝௞௅ െ ‫۾‬஻ି ݁ ௝௞௅
which can be rearranged to give
ߩܿ
ቀͳ െ ‫ ݖ‬ቁ
ଷ ି௝ଶ௞௅
‫۾‬஻ି ൌ ‫۾‬஻ା ߩܿ ݁ (2.18)
ቀͳ ൅ ‫ ݖ‬ቁ
ଷ
Similarly, at ‫ ݔ‬ൌ Ͳ (corresponding to point 2 in Fig. 2.1), the specific acoustic impedance
can be found as
࢖ଶ ‫۾‬஻ା ൅ ‫۾‬஻ି
‫ݖ‬ଶ ൌ ൌ ߩܿ ൬ ൰ (2.19)
࢛ଶ ‫۾‬஻ା െ ‫۾‬஻ି
25
Substitution of Eq. (2.18) into Eq. (2.19) yields
ߩܿ ௝௞௅ ߩܿ
ቀͳ ൅ ቁ ݁ ൅ ቀͳ െ ቁ ݁ ି௝௞௅
‫ݖ‬ଷ ‫ݖ‬ଷ
‫ݖ‬ଶ ൌ ߩܿ ቎ ߩܿ ௝௞௅ ߩܿ ି௝௞௅ ቏ (2.20)
ቀͳ ൅ ቁ ݁ െ ቀͳ െ ቁ ݁
‫ݖ‬ଷ ‫ݖ‬ଷ
and finally, substitution of Eq. (2.16) into Eq. (2.20) yields
ܵ ܵ
ቀͳ ൅ ܵ஼ ቁ ݁ ௝௞௅ ൅ ቀͳ െ ܵ஼ ቁ ݁ ି௝௞௅
஻ ஻
‫ݖ‬ଶ ൌ ߩܿ ൦ ൪ (2.21)
ܵ ܵ
ቀͳ ൅ ܵ஼ ቁ ݁ ௝௞௅ െ ቀͳ െ ܵ஼ ቁ ݁ ି௝௞௅
஻ ஻
The specific acoustic impedance at point 2 is also given by ‫ݖ‬ଶ ൌ ܼଶ ܵ஻ . The acoustic
pressure and volume velocity at points 1 and 2 are equal, hence the acoustic impedances
at these points are equal ሺܼଵ ൌ ܼଶ ሻ. The specific acoustic impedance at point 1 can then
be obtained as
ܵ஼ ௝௞௅ ܵ஼ ି௝௞௅
ܵ஺ ቀͳ ൅ ܵ஻ ቁ ݁ ൅ ቀͳ െ ܵ஻ ቁ ݁
‫ݖ‬ଵ ൌ ߩܿ ൬ ൰ ൦ ൪ (2.22)
ܵ஻ ቀͳ ൅ ܵ஼ ቁ ݁ ௝௞௅ െ ቀͳ െ ܵ஼ ቁ ݁ ି௝௞௅
ܵ஻ ܵ஻
The acoustic pressures at points 1 to 4 must now be found. The complex pressure at
‫ ݔ‬ൌ Ͳ is ‫۾‬ଵ ൌ ‫۾‬௜ ൅ ‫۾‬௥ and the specific acoustic impedance at point 1 is
࢖ଵ ‫۾‬௜ ൅ ‫۾‬௥
‫ݖ‬ଵ ൌ ൌ ߩܿ ൬ ൰ (2.23)
࢛ଵ ‫۾‬௜ െ ‫۾‬௥
The acoustic pressure at point 1 can then be obtained as
ʹ‫۾‬௜
‫۾‬ଵ ൌ ߩܿ (2.24)
ଵ
Equality of the acoustic pressures at points 1 and 2 leads to ‫۾‬ଶ ൌ ‫۾‬ଵ.
26
The complex pressures at point 2 ሺ‫ ݔ‬ൌ Ͳሻ and point 3 ሺ‫ ݔ‬ൌ ‫ܮ‬ሻ can be obtained by
applying Eq. (2.2) to show
‫۾‬ଶ ൌ ‫۾‬୆ା ൅ ‫۾‬୆ି (2.25)
‫۾‬ଷ ݁ ି௝௞௅ ൌ ‫۾‬୆ା ݁ ି௝௞௅ ൅ ‫۾‬୆ି ݁ ௝௞௅ (2.26)
Rearranging Eqs. (2.25) and (2.26) and substituting into Eq. (2.19) obtains the acoustic
pressure at point 3 as
‫۾‬ଶ ߩܿ ߩܿ
‫۾‬ଷ ൌ ൤൬ͳ ൅ ൰ ൅ ൬ͳ െ ൰ ݁ ௝ଶ௞௅ ൨ (2.27)
ʹ ‫ݖ‬ଶ ‫ݖ‬ଶ
From equality of the acoustic pressures at points 1 and 2 ሺ‫۾‬ଵ ൌ ‫۾‬ଶ ሻ and at points 3 and 4
ௌ
ሺ‫۾‬ଷ ൌ ‫۾‬ସ ሻ, and recalling that ‫ݖ‬ଶ ൌ ‫ݖ‬ଵ ቀ ಳ ቁ, Eq. (2.27) can be rewritten as
ௌ ಲ
‫۾‬ଵ ܵ஺ ߩܿ ܵ஺ ߩܿ ௝ଶ௞௅
‫۾‬ସ ൌ ൤൬ͳ ൅ ൰ ൅ ൬ͳ െ ൰݁ ൨ (2.28)
ʹ ܵ஻ ‫ݖ‬ଵ ܵ஻ ‫ݖ‬ଵ
Finally, Eq. (2.24) can be substituted into Eq. (2.28) to give
ܵ ߩܿ ܵ ߩܿ
‫۾‬௜ ቂቀͳ ൅ ܵ஺ ‫ ݖ‬ቁ ൅ ቀͳ െ ܵ஺ ‫ ݖ‬ቁ ݁ ௝ଶ௞௅ ቃ
஻ ଵ ஻ ଵ
‫۾‬ସ ൌ ߩܿ (2.29)
ଵ
Substitution of Eq. (2.22) into Eq. (2.29) and recalling ‫۾‬ସ ൌ ‫۾‬௧ allows ‫۾‬௜ and ‫۾‬௧ to be
related by
‫۾‬௧ Ͷ
ൌ
‫۾‬௜ ቀͳ ൅ ܵ஼ ൅ ܵ஼ ൅ ܵ஻ ቁ ൅ ቀͳ ൅ ܵ஼ െ ܵ஼ െ ܵ஻ ቁ ݁ ି௝ଶ௞௅ (2.30)
ܵ஺ ܵ஻ ܵ஺ ܵ஺ ܵ஻ ܵ஺
Using previous definitions for ܴଵ and ܴଶ , Eq. (2.30) becomes
‫۾‬௧ Ͷ
ൌ (2.31)
‫۾‬௜ ሺܴଵ ൅ ܴଶ ሻ ൅ ሺܴଵ െ ܴଶ ሻ݁ ି௝ଶ௞௅
27
Equation (2.31) is identical to Eq. (2.12a). The resulting equation for sound transmission
loss is thus identical to that obtained using the continuity of pressure and volume velocity
method.
2.2.1.3 Transfer matrix method
In contrast to the previous two analytical approaches, the transfer matrix method does not
require the explicit calculation of acoustic properties throughout the muffler air path but
instead requires decomposition of the system into its constituent acoustic components.
Each of these components are then characterised by a ʹ ൈ ʹ transfer matrix which relates
two variables at planes on either side the component [10, 28]. The transfer matrices for
common components found in muffler systems are well defined and can be readily
combined by successive multiplication of the subsystem matrices to form a single overall
matrix for the system [10, 14, 15, 17, 28].
The following general transfer matrix may be written to relate the state variables on
either side of an acoustic system.
‫݌‬௡ ‫ܣ‬ ‫ܤ‬ ‫݌‬௡ାଵ

ቈ ቉ൌቈ ቉ቈ ቉ (2.32)
ܷ௡ ‫ܥ‬ ‫ܷ ܦ‬௡ାଵ
‫ ݌‬is the acoustic pressure and ܷ is the volume velocity (product of the cross sectional
area and particle velocity). For the case of the simple expansion chamber reactive muffler
shown in Fig. 2.1, only the transfer matrices for a uniform tube, a sudden expansion and
a sudden contraction need to be considered. The transfer matrix for a uniform tube for the
two state variables chosen in this case is given by [10, 14]
ߩܿ
‫ۍ‬ ሺ݇‫ܮ‬ሻ ݆ ሺ݇‫ܮ‬ሻ‫ې‬
‫ܣ‬ ‫ܤ‬ ܵ஻
ቈ ቉ൌ‫ܵ ێ‬ ‫ۑ‬ (2.33)
‫ܥ‬ ‫ܦ‬ ‫ ݆ێ‬஻ ሺ݇‫ܮ‬ሻ ሺ݇‫ܮ‬ሻ ‫ۑ‬
‫ܿߩ ۏ‬ ‫ے‬
When the muffler cross section is small compared with the wavelength, and in the
absence of air flow, the sudden expansion and contraction at the ends of the expansion
chamber may be represented by simple unit matrices [10]. When this assumption cannot
be made, additional elements incorporating Karal correction factors should be
introduced [15]. These correction factors account for the additional impedance, referred
28
to by Karal as the discontinuity inductance, which is introduced by the pressure

discontinuity caused by the abrupt change in cross section. The correction matrix is given
by
‫ܣ‬ ‫ܤ‬ ͳ ݆߱‫ܮ‬௄ ͺߩ ‫ݎ‬௣

ቈ ቉ൌቈ ቉ǡ ‫ܮ‬௄ ൌ ଶ
‫ܪ‬൬ ൰ (2.34, 2.35)
‫ܥ‬ ‫ܦ‬ Ͳ ͳ ͵ߨ ‫ݎ‬௣ ‫ݎ‬௖
where ‫ܮ‬௄ is the analogous acoustical inductance, ‫ݎ‬௣ is the radius of the pipe,‫ݎ‬௖ is the
௥೛
radius of the expansion chamber and ‫ ܪ‬ቀ ቁ is given by Fig. 2b in Miwa and Igarashi
௥ ೎
[15] or may be approximated by
‫ݎ‬௣ ‫ݎ‬௣ ଷ ‫ݎ‬௣ ଶ ‫ݎ‬௣

‫ ܪ‬൬ ൰ ൌ ͲǤ͸ͺͷ͹ ൬ ൰ െ ͲǤͶ͵ͳʹ ൬ ൰ െ ͳǤʹͷͲͳ ൬ ൰ ൅ ͳǤͲͲͲ͸ (2.36)
‫ݎ‬௖ ‫ݎ‬௖ ‫ݎ‬௖ ‫ݎ‬௖
With reference to the simple expansion chamber reactive muffler shown in Fig. 2.1, the
final transfer matrix, denoted by subscript T, can be derived using simple matrix
multiplication of the appropriate combination of the above matrices.
ߩܿ
‫்ܣ‬ ‫்ܤ‬ ‫ۍ‬ ሺ݇‫ܮ‬ሻ ݆ ሺ݇‫ܮ‬ሻ‫ې‬
ͳ ݆߱‫ܮ‬௄ଵǡଶ ܵ஻ ͳ ݆߱‫ܮ‬௄ଷǡସ
ቈ ቉ൌቈ ቉‫ܵ ێ‬ ‫ۑ‬ቈ ቉ (2.37)
‫்ܥ‬ ‫்ܦ‬ Ͳ ͳ ‫ ݆ێ‬஻ ሺ݇‫ܮ‬ሻ ሺ݇‫ܮ‬ሻ ‫Ͳ ۑ‬ ͳ
where ‫ܮ‬௄ଵǡଶ and ‫ܮ‬௄ଷǡସ are the analogous acoustical inductances (also called Karal
correction factors) for the discontinuities at the inlet and outlet pipes, respectively.
For the case of a non-reflecting termination of the system, the transmission loss
incorporating the transfer matrix constants is given by [10]
ͳ ܵ஺ ଵΤଶ ܵ஼ ߩܿ ܵ஼
ܶ‫ ܮ‬ൌ ʹͲ ଵ଴ ቈ ൬ ൰ ฬ‫ ்ܣ‬൅ ‫ ்ܤ‬൬ ൰ ൅ ‫ ்ܥ‬൬ ൰ ൅ ‫ ்ܦ‬൬ ൰ฬ቉ (2.38)
ʹ ܵ஼ ߩܿ ܵ஺ ܵ஺
Modelling the sudden expansion and contraction at the discontinuities with unit matrices,
Eq. (2.37) simplifies to that of Eq. (2.33). Substituting this into Eq. (2.38) and resolving
29
the real and imaginary components yields
ͳ ܵ஺
ܶ‫ ܮ‬ൌ ͳͲ ଵ଴ ൤ ൬ ൰ ሺܴଵଶ ଶ ሺ݇‫ܮ‬ሻ ൅ ܴଶଶ ଶ ሺ݇‫ܮ‬ሻሻ൨ (2.39)
Ͷ ܵ஼
where ܴଵ and ܴଶ have been defined previously. Equation (2.39) is identical to the result
obtained using the continuity of pressure and volume velocity method given by
Eq. (2.13).
2.2.1.4 Comparison of analytical methods
Figure 2.2 shows transmission loss predictions for a single cylindrical expansion
chamber muffler having 86 mm chamber diameter and 156 mm chamber length. Both
inlet and outlet pipes have 18 mm diameter and 20 mm length. The upper limit of the
plane wave frequency range corresponds to the 2.3 kHz cut-on frequency for the
expansion chamber. The continuity of pressure and volume velocity method and the
impedance method are represented by a single curve as they produced identical equations
in their derivation of transmission loss. It can be shown that the transfer matrix method
(TMM) also yields an identical result when the effects of the expansion at the chamber
inlet and contraction at the chamber outlet are neglected. The prediction obtained using
the transfer matrix method with Karal correction factors (TMM + Karal) results in an
increase in transmission loss with increasing frequency compared to the other methods.
2.2.2 Multiple chamber muffler
A multi-chamber muffler is now considered which comprises three expansion chambers

and two interconnecting ducts, as shown in Fig. 2.3. The outlet pipe is considered to be
anechoically terminated. The impedance and transfer matrix methods are used to predict
the muffler transmission loss in what follows.
2.2.2.1 Impedance method
Development of the final equation for transmission loss follows a similar process to that
for the single chamber muffler by calculating the impedance and pressure at each of the
locations 1 to 12 as identified in Fig. 2.3.
30
30
Continuity, Impedance
25 TMM
TMM + Karal
Transmission Loss (dB)
20
15
10
0
0 1,000 2,000 3,000 4,000 5,000
Frequency (Hz)
Figure 2.2 Transmission loss of the single 86 mm diameter expansion chamber

muffler, comparing continuity, impedance and transfer matrix analytical
methods
ܵ஺ ܵ஻ ܵ஼ ܵ஽ ܵா ܵி ܵீ
‫ܘ‬௜ ‫ܘ‬௧
‫ܘ‬୰ ͳ ʹ ͵ Ͷ ͷ ͸ ͹ ͺ ͻ ͳͲ ͳͳ ͳʹ
‫ܮ‬஻ ‫ܮ‬஼ ‫ܮ‬஽ ‫ܮ‬ா ‫ܮ‬ி
Figure 2.3 A multiple expansion chamber reactive muffler
31
With an anechoic termination at the outlet duct, the specific acoustic impedance at point
12 is equal to the characteristic impedance ሺ‫ݖ‬ଵଶ ൌ ߩܿሻ and the acoustic pressure at point
12 is equal to the transmitted pressure (‫۾‬ଵଶ ൌ ‫۾‬௧ ).The complete development of the
transmission loss equation is presented in Appendix A. The sound transmission loss is
ȁ‫۾‬௧ଶ ȁ ܵீ
ܶ‫ ܮ‬ൌ െͳͲ ଵ଴ ቈ ଶ ቉ (2.40)
ห‫۾‬௜ ห ܵ஺
2.2.2.2 Transfer matrix method
The overall transfer matrix for the multi-chamber muffler, denoted by subscript T, is
derived using simple matrix multiplication of the appropriate combination of matrices for
five uniform tubes, three expansion discontinuities and three contraction discontinuities,
and is of the form
ߩܿ
‫்ܣ‬ ‫்ܤ‬ ‫ۍ‬ ሺ݇‫ܮ‬஻ ሻ ݆ ሺ݇‫ܮ‬஻ ሻ‫ې‬
ͳ ݆߱‫ܮ‬௄ଵǡଶ ܵ஻ ͳ ݆߱‫ܮ‬௄ଷǡସ
ቈ ቉= ቈ ቉‫ܵ ێ‬ ‫ۑ‬ቈ ቉ൈ
‫்ܥ‬ ‫்ܦ‬ Ͳ ͳ ‫ ݆ێ‬஻ ሺ݇‫ ܮ‬ሻ ሺ݇‫ܮ‬ ሻ ‫Ͳ ۑ‬ ͳ
஻ ஻
ߩܿ ߩܿ
‫ ۍ‬ሺ݇‫ܮ‬஼ ሻ ݆ ሺ݇‫ܮ‬஼ ሻ‫ې‬
ܵ஼ ͳ ݆߱‫ܮ‬௄ହǡ଺ ‫ۍ‬ ሺ݇‫ܮ‬஽ ሻ ݆ ሺ݇‫ܮ‬஽ ሻ‫ې‬
ܵ஽
‫ێ‬ ‫ۑ‬ቈ ቉ ‫ێ‬ ‫ۑ‬ൈ
‫ܵ ݆ێ‬஼ ሺ݇‫ ܮ‬ሻ ‫ێ‬ ܵ஽
஼ ሺ݇‫ܮ‬஼ ሻ ‫ۑ‬ Ͳ ͳ ݆ ሺ݇‫ܮ‬஽ ሻ ሺ݇‫ܮ‬஽ ሻ ‫ۑ‬
‫ܿߩ ۏ‬ ‫ے‬ ‫ܿߩ ۏ‬ ‫ے‬
ߩܿ
‫ۍ‬ ሺ݇‫ܮ‬ா ሻ ݆ ሺ݇‫ܮ‬ா ሻ‫ې‬
ͳ ݆߱‫ܮ‬௄଻ǡ଼ ܵா ͳ ݆߱‫ܮ‬௄ଽǡଵ଴
ቈ ቉‫ܵ ێ‬ ‫ۑ‬ቈ ቉ൈ
Ͳ ͳ ‫ ݆ێ‬ா ሺ݇‫ ܮ‬ሻ ሺ݇‫ܮ‬ ሻ ‫Ͳ ۑ‬ ͳ
ா ா
ߩܿ
‫ ۍ‬ሺ݇‫ܮ‬ி ሻ ݆ ሺ݇‫ܮ‬ி ሻ‫ې‬
ܵி ͳ ݆߱‫ܮ‬௄ଵଵǡଵଶ
‫ێ‬ ‫ۑ‬ቈ ቉ (2.41)
‫ܵ ݆ێ‬ி ሺ݇‫ ܮ‬ሻ ሺ݇‫ܮ‬ி ሻ ‫ۑ‬ Ͳ ͳ
ி
where ‫ܮ‬௄ଵǡଶ to ‫ܮ‬௄ଵଵǡଵଶ are Karal correction factors for each of the area discontinuities. ‫ܮ‬஻
to ‫ܮ‬ி and ܵ஻ to ܵி are respectively the lengths and cross-sectional areas of each
component shown in Fig. 2.3.
32
2.2.2.3 Comparison of analytical methods
Figure 2.4 shows transmission loss predictions for the multi-chamber muffler, where
each expansion chamber has the dimensions of diameter 86 mm and length 156 mm, that
is, having dimensions identical to the single muffler presented in Fig. 2.1. The inlet and
outlet pipes as well as the interconnecting pipes have 18 mm diameter and 20 mm length.
The impedance method and transfer matrix method yield identical results. The prediction
obtained using the transfer matrix method with Karal correction factors applied at the
discontinuities undergoes shifts in both frequency and transmission loss although the
trends are less predictable than those observed for the single chamber muffler.
Figure 2.5 shows the transmission loss prediction for the multi-chamber muffler
compared to the prediction for the single expansion chamber of the same dimensions.
Both predictions use the transfer matrix method with Karal correction factors applied. It
can be seen that the frequency spacing of the transmission loss domes remains the same
but the multi-chamber muffler result exhibits significant variation in the transmission
loss maxima.
A third curve has been added to Fig. 2.5 by simply multiplying the transmission loss of
the single chamber muffler by a factor of three. It can be observed that the transmission
loss obtained using a muffler comprising three chambers is not three times the loss
obtained using a single chamber, that is, losses due to consecutive chambers cannot be
simply added together. This is attributed to the fact that the assumed anechoic
termination at the outlet of the single muffler is not valid for the first or second chamber
in the three chamber muffler, and the need to account for any further contribution made
by the interconnecting ducts.
2.3 Computational Approaches
The transmission loss of mufflers can be evaluated using the finite element analysis
(FEA) method either directly, by considering attenuation of acoustic power, or indirectly
using the transfer matrix method. While the first approach is the simpler of the two, the
second approach obtains the transfer matrix parameters for the muffler system which can
be used to characterise the system.
33
90
Impedance
80
TMM
70
TMM + Karal
60
50
40
30
20
10
0
0 1,000 2,000 3,000 4,000 5,000
Frequency (Hz)
Figure 2.4 Transmission loss of the multiple expansion chamber muffler, comparing
impedance and transfer matrix analytical methods
90
Single chamber
80 Single chamber result x 3
Three chambers
70
60
50
40
30
20
10
0
0 1,000 2,000 3,000 4,000 5,000
Frequency (Hz)

muffler, multiple expansion chamber muffler and single chamber muffler
times a factor of three – analytical results using TMM with Karal
correction factors
34
2.3.1 Direct approach
The direct approach for evaluating the transmission loss only requires calculation of
sound power at the inlet and outlet of the finite element model.
Using COMSOL Multiphysics (v4.0a) as the FEA package, it is possible to define two
integration variables
ܲ௜ଶ ܲ௧ଶ
ܹ௜ ൌ න ݀‫ݏ‬ ܹ௧ ൌ න ݀‫ݏ‬ (2.42, 2.43)
ʹߩܿ ʹߩܿ
௜௡௟௘௧ ௢௨௧௟௘௧
where ܹ௜ and ܹ௧ are the incident and transmitted sound power, respectively. ܲ௜ and ܲ௧
are the incident and transmitted peak pressure amplitude, respectively. These integration
variables are evaluated at the inlet and outlet surfaces of the model, respectively.
A global variable can then be defined as
ܹ௧
ܶ‫ ܮ‬ൌ െͳͲ ଵ଴ ൬ ൰ (2.44)
ܹ௜
By applying a harmonic plane wave pressure field at the model inlet and an anechoic
boundary condition to the model outlet, the transmission loss can be readily calculated at
each frequency of interest. The remaining surfaces of the model are treated as being
acoustically hard. The model domain is meshed using second-order tetrahedral
Lagrangian elements having a maximum size of 10 mm, ensuring greater than six
elements per wavelength at the upper frequency of 5 kHz [96].
Figure 2.6 shows transmission loss for a single chamber muffler of dimensions given
previously. Comparing the analytical and FEA results, similar trends can be observed.
However, the FEA result departs significantly from the analytical solutions at frequencies
greater than 4.6 kHz. This deviation occurs at a significantly higher frequency than the
2.3 kHz cut-on frequency of the first diametral mode. Selamet and Radavich [27] found
that the one-dimensional repeating dome behaviour presented by cylindrical expansion
chambers existed only below the onset of the first circumferential mode and that no
domes extended above that frequency. They presented the following relationship which
35
௞௅
yields the number of repeating domes, గ
, as
݇‫ߙʹ ܮ‬௠ǡ௡ ‫ܮ‬

൏ ൬ ൰ (2.45)
ߨ ߨ ݀
The root obtained at the first circumferential mode is ߙ଴ǡଵ ൌ ͵Ǥͺ͵ʹ. The number of
repeating domes for an 86 mm diameter muffler having a chamber length of 156 mm is
predicted to be four (less than 4.4), which is consistent with the number of domes
observed in Fig. 2.6.
Figure 2.7 shows the acoustic pressure field within the chamber at frequencies coinciding
with transmission loss minima in Fig. 2.6. Plane waves are obvious throughout the
expansion chamber at the lower frequencies. Higher mode behaviour appears at the
chamber inlet and outlet at medium frequencies but plane waves still dominate within the
chamber. At the highest frequency, notably within a region where the analytical and FEA
results show no agreement, the pressure field is dominated by higher modes.
60
Continuity, Impedance
50 TMM + Karal
FEA
40
30
20
10
0
0 1,000 2,000 3,000 4,000 5,000
Frequency (Hz)

muffler, comparing analytical results and FEA results
36
Figure 2.7 Pressure fields in the single 86 mm diameter expansion chamber muffler
at frequencies coinciding with peak transmission loss – FEA results
160
Impedance
140
TMM + Karal
120 FEA
100
80
60
40
20
0
0 1,000 2,000 3,000 4,000 5,000
Frequency (Hz)
Figure 2.8 Transmission loss of the multi-chamber muffler, comparing analytical

results and FEA results
Figure 2.8 shows transmission loss predictions for the multi-chamber muffler. As
observed previously with the single 86 mm diameter muffler, the transfer matrix method
incorporating Karal correction factors produces the best correlation with the FEA results
over the entire frequency range. The departure between the analytical and finite element
37
results above 4.4 kHz is repeated in the multi-chamber design and is attributed to the
onset of the first circumferential mode.
2.3.2 Indirect approach
The indirect approach for evaluating the transmission loss of a muffler applies the
transfer matrix numerical approach using the finite element method. Kim and Soedel [97]
and Wu et al. [44] present an improved method for the calculation of the four pole
parameters in Eq. (2.32), which re-states the general transfer matrix equation as an
impedance matrix equation of the form
‫݌‬௡ ‫כܣ‬ ‫כܤ‬ ‫ݑ‬௡

ቈ ቉ൌቈ ቉ቈ ቉ (2.46)
‫݌‬௡ାଵ ‫כܥ‬ ‫ݑ כܦ‬௡ାଵ
Particle velocity ‫ ݑ‬is now used in place of the volume velocity and pressure and velocity
terms are grouped on opposite sides of the equality. In this form, only the velocity
boundary condition is used in the calculations and the matrix parameters can be obtained
by calculating (or measuring) pressures alone. The improved method offers several
advantages over the original method when applied with the finite element method to
evaluate transmission loss [98]. Equation (2.46) can be expanded to show
‫݌‬ଵ െ ‫ݑ כ ܤ‬ଶ ‫݌‬ଵ െ ‫ݑ כܣ‬ଵ

‫ כܣ‬ൌ ‫ כܤ‬ൌ (2.47, 2.48)
‫ݑ‬ଵ ‫ݑ‬ଶ
‫כ‬
‫݌‬ଶ െ ‫ݑ כܦ‬ଶ ‫כ‬
‫݌‬ଶ െ ‫ݑ כ ܥ‬ଵ
‫ ܥ‬ൌ ‫ ܦ‬ൌ (2.49, 2.50)
‫ݑ‬ଵ ‫ݑ‬ଶ
The transfer matrix parameters may be calculated from the pressures at the model inlet
and outlet by applying the following two velocity boundary condition cases
Case 1: ‫ כܣ‬ൌ ‫݌‬ଵ ȁ௨భୀଵǡ௨మ ୀ଴ ‫ כ ܥ‬ൌ ‫݌‬ଶ ȁ௨భ ୀଵǡ௨మ ୀ଴ (2.51, 2.52)
Case 2: ‫ כ ܤ‬ൌ ‫݌‬ଵ ȁ௨భ ୀ଴ǡ௨మ ୀଵ ‫ כܦ‬ൌ ‫݌‬ଶ ȁ௨భ ୀ଴ǡ௨మ ୀଵ (2.53, 2.54)
38
It is possible to calculate the original transfer matrix parameters necessary for use in the
transmission loss equation (2.38) by combining Eqs. (2.32) and (2.46) and using
ܷ௡ ൌ ‫ݑ‬௡ ܵ௡ to show
‫כܣ‬ ‫ͳ כܦ כܣ‬
‫ܣ‬ൌ ‫ ܤ‬ൌ ൬‫ כ ܤ‬െ ൰൬ ൰ (2.55, 2.56)
‫כܥ‬ ‫כܥ‬ ܵଶ
ͳ െ‫ܵ כܦ‬ଵ
‫ ܥ‬ൌ ൬ ‫ כ‬൰ ሺܵଵሻ ‫ ܦ‬ൌ ൬ ‫ כ‬൰൬ ൰ (2.57, 2.58)
‫ܥ‬ ‫ܥ‬ ܵଶ
While each boundary condition case requires a harmonic input velocity to be specified,
not all FEA packages support this option. In these instances it is possible to apply either a
ି௝௨
harmonic displacement ቀ‫ ݔ‬ൌ ఠ
ቁ or acceleration ሺܽ ൌ ݆‫߱ݑ‬ሻ instead [99]. The inlet
boundary condition can thus be specified with a velocity magnitude equal to unity
ሺ‫ ݑ‬ൌ ͳሻ. Boundary condition cases are applied at discrete frequencies through the range
of interest, with the number of computational runs being dictated by the resolution
required. Pressure data are obtained at the muffler inlet and outlet. It must be recognised
that when a real value of velocity is used at the muffler inlet, the pressures resulting from
the finite element analysis are imaginary and the velocities are real. Subsequently, the
transfer matrix parameters ሺ‫ܣ‬ǡ ‫ܤ‬ǡ ‫ܥ‬ǡ ‫ܦ‬ሻ derived using Eqs. (2.51) to (2.58) are complex.
As ‫ ܣ‬and ‫ ܦ‬are real (imaginary/imaginary and real/real ratios respectively), and ‫ ܤ‬and ‫ܥ‬
are imaginary (imaginary/real and real/imaginary ratios respectively), Eq. (2.38) can be
restated as
ଵΤଶ
ଵΤଶ ଶ ଶ
ͳ ܵ஺ ܵ஼ ܵ஼ ߩܿ
ܶ‫ ܮ‬ൌ ʹͲ ଵ଴ ൦ ൬ ൰ ቎ቆ‫ ்ܣ‬൅ ‫ ்ܦ‬൬ ൰ቇ ൅ ൭‫ ்ܤ‬൬ ൰ ൅ ‫ ்ܥ‬൬ ൰൱ ቏ ൪ (2.59)
ʹ ܵ஼ ܵ஺ ߩܿ ܵ஺
The parameters calculated at each frequency are substituted into Eq. (2.59) to obtain the
transmission loss spectrum over the desired frequency range.
Figure 2.9 shows the transmission loss predicted for the single chamber muffler using
both finite element approaches. It can be seen that both the direct approach based on the
acoustic power and the indirect approach using the transfer matrix method yield identical
results.
39
60
FEA - Direct approach (acoustic power)
50 FEA - Indirect approach (TMM)

40
30
20
10
0
0 1,000 2,000 3,000 4,000 5,000
Frequency (Hz)

muffler, comparing FEA results obtained using the direct approach
(acoustic power) and indirect approach (transfer matrix method)
Figure 2.10 compares the transfer matrix parameters ሺ‫ܣ‬ǡ ‫ܤ‬ǡ ‫ܥ‬ǡ ‫ܦ‬ሻ obtained from the finite
element analysis of the single chamber muffler with those obtained analytically and
given by Eq. (2.37). By including Karal correction factors in the analytical solution, it
can be seen that there is close agreement between the analytical and finite element
solutions up to approximately 4.4 kHz. The departure between the analytical and finite
element results at higher frequencies is attributed to the presence of higher order modes.
2.4 Experimental Approach
Methods that use a short duration acoustic pulse or acoustic impulse to determine the
transmission loss of mufflers are well described in literature [100-103]. Figure 2.11
shows a schematic diagram of the two-microphone experimental set-up used in the
current work. Figure 2.12 shows a photograph of the experimental test rig.
A transient acoustic pulse was generated by a Brüel & Kjær LAN-XI Pulse front end and
fed to two TU-650 horn drivers via a PA-25E power amplifier. The pulse propagated
down the 18 mm diameter PVC conduit where it was measured by the upstream
40
microphone ‫ܯ‬ଵ before continuing to the inlet of the muffler. The pressure of the
corresponding pulse transmitted from the outlet of the muffler was measured by the
downstream microphone ‫ܯ‬ଶ . The resulting time histories recorded by the two
microphones were processed using windowing to extract only those time periods that
captured the initial positive travelling wave.
Rectangular windowing with 1 ms leading and trailing cosine tapers was applied to the
time history measured by ‫ܯ‬ଵ (Fig. 2.13). Although the 8 ms window appears to be
excessively long, the recorded pulse at ‫ܯ‬ଵ does not decay to background noise levels
prior to closure of the window. Varying the length of the window between 3 ms and 8 ms
produced no discernable change in the transfer function data. The 8 ms window length
was thus selected to capture the full initial pulse data whilst excluding a small reflected
pulse at 59 ms.
The system response undergoes more complex decay than the excitation pulse and has a
significantly longer duration. The decaying signal also makes it difficult to identify the
point at which to terminate the time record. In the case of the simpler mufflers, the choice
of an exponential window allows a slightly longer window to be nominated to be certain
of capturing the full system response while attenuating any noise contamination
introduced at the end of the time record. In the case of the three-chamber muffler,
interactions between the chambers meant that there was insufficient time for the response
to decay to background levels prior to the arrival of a reflected pulse at the downstream
microphone. Simply truncating the time record to exclude the reflected pulse will result
in a leakage error in the measurement (truncation in time, leakage in frequency) [104].
Using an exponential window forces the response to decay completely within the record
length and avoids leakage due to truncation. Exponential windowing with a 3 ms leading
cosine taper and 4 ms decay constant (߬) was applied to the time history measured by ‫ܯ‬ଶ
(Fig. 2.14).
41
50 1.E+08
FEA TMM FEA TMM
TMM + Karal TMM + Karal
25 5.E+07
A
0 0.E+00
B (kg m-2 s-1)

-25 -5.E+07
-50 -1.E+08
0 1 2 3 4 5 0 1 2 3 4 5
Frequency (kHz) Frequency (kHz)
2.E-05 50
FEA TMM FEA TMM
TMM + Karal TMM + Karal
1.E-05 25
D
0.E+00 0
C (m2 s kg-1)
-1.E-05 -25
-2.E-05 -50
0 1 2 3 4 5 0 1 2 3 4 5
Frequency (kHz) Frequency (kHz)
Figure 2.10 Transfer matrix parameters of the single 86 mm diameter expansion chamber muffler, comparing analytical results (obtained
with and without Karal correction factors) and FEA results
3,200 320 3,200 ~130 320 3,200
Horn drivers
Muffler
‫ܯ‬ଵ ‫ܯ‬ଶ
Amplifier Pulse generator
PC Signal analyser
Figure 2.11 Schematic diagram of the two-microphone acoustic pulse experimental

set-up (dimensions in mm)
Figure 2.12 Photograph of the two-microphone acoustic pulse experimental set-up

showing horn drivers, microphones and muffler
43
400
Signal
300 Weighting
Acoustic pressure (Pa)
200
100
-100 Downstream reflection

from muffler
-200
0.045 0.050 0.055 0.060 0.065 0.070
Time (sec)
Figure 2.13 Upstream microphone (ͳ) time history showing rectangular FFT
weighting function
40
Signal
30 Weighting
Acoustic pressure (Pa)
20
10
-10
Upstream reflection
from horn drivers
-20
0.050 0.060 0.070 0.080 0.090
Time (sec)
Figure 2.14 Downstream microphone (ʹ) time history showing exponential FFT
weighting function
44
Utilising long lengths of pipe in the system provided sufficient time delay (approximately
15 ms) between the arrival of the initial pulse and the reflected waves generated at the
muffler and pipe ends to facilitate the data extraction. Applying appropriate windowing
to the downstream microphone time history in the case of multi-chamber mufflers proved
to be especially challenging. Interaction between the muffler chambers produced a series
of pulses at the muffler outlet. The combined duration of these output pulses approached
the time taken for the first reflected pulse to return to the outlet microphone. Judicious
selection of window length and time delay constant was required to capture sufficient
transmitted energy without also capturing any of the reflected pulse. The extracted time
histories at both microphones were captured for 100 individual pulses. The time histories
were then Fourier Transformed and the results averaged in the frequency domain.
As the acoustic path between the two microphones includes 3.5 m of narrow PVC
conduit in addition to the muffler, the attenuation due to the conduit must be subtracted
from the total attenuation in order to obtain the attenuation due to the muffler alone. The
experiment was repeated using an equivalent length of conduit in place of the muffler
and the transmission loss of the muffler was then obtained using [105]:
ܺ௠ ሺ݂ሻ ܺௗ ሺ݂ሻ
ܶ‫ ܮ‬ൌ ʹͲ ଵ଴ ቆቤ ቤቇ െ ʹͲ ଵ଴ ቆቤ ቤቇ (2.60)
ܻ௠ ሺ݂ሻ ܻௗ ሺ݂ሻ
ܺ௠ ሺ݂ሻ and ܻ௠ ሺ݂ሻ are respectively the Fourier transforms of the time histories of the
incident and transmitted waves obtained with the muffler in position, while ܺௗ ሺ݂ሻ and
ܻௗ ሺ݂ሻ are the Fourier transforms of the corresponding time histories obtained with the
length of conduit (duct) in place of the muffler.
Figure 2.15 contains the transmission loss obtained experimentally for the single 86 mm
diameter expansion chamber muffler, the transmission loss predicted using the analytical
transfer matrix method (including Karal correction factors) and the transmission loss
predicted using the direct finite element method. The FEA results show excellent
agreement with the experimental results over the frequency range assessed. The slight
frequency shift between the results is attributed to dimensional inaccuracy in assembling
the physical muffler and/or air temperature difference between the experiment and the
finite element model. As a guide, the frequency shift observed at 4.4 kHz would result
45
from less than a 2 mm inaccuracy in the 156 mm length of the expansion chamber or 5oC
temperature difference.
60
TMM + Karal
50 FEA
Experimental
40
30
20
10
0
0 1,000 2,000 3,000 4,000 5,000
Frequency (Hz)

muffler, comparing analytical results, FEA results and experimental
results
2.5 Generic CPAP Device Expansion Chamber Muffler
Expansion chambers within CPAP device mufflers typically have large expansion ratios
(as high as 40) and short length-to-diameter ratios (less than one). Mufflers having large
expansion ratios and short chamber lengths exhibit transmission loss behaviour that
departs from that predicted by Eq. (2.39). In such cases, the behaviour is more closely
predicted using the transfer matrix representation for a Helmholtz resonator. Miwa and
Igarashi [15] define the transfer matrix parameters for a resonator comprised of a side
branch with a closed cavity, as shown in Fig. 2.16, as
ሺ݇‫ܮ‬ௗ ሻ ሺ݇‫ܮ‬ௗ ሻ
‫ ܣ‬ൌ ‫ ܦ‬ൌ ଶ ሺ݇‫ܮ‬ௗ ሻ െ ଶሺ݇‫ܮ‬ௗ ሻ െ ܴ (2.61)
ܵௗ
46
ܵௗ ଶሺ݇‫ܮ‬ௗ ሻ
‫ ܤ‬൬ ൰ ൌ ʹ ሺ݇‫ܮ‬ௗ ሻ ሺ݇‫ܮ‬ௗ ሻ െ ܴ (2.62)
݆ߩܿ ܵௗ
ܵௗ ଶ ሺ݇‫ܮ‬ௗ ሻ
‫ ܥ‬൬ ൰ ൌ ʹ ሺ݇‫ܮ‬ௗ ሻ ሺ݇‫ܮ‬ௗ ሻ ൅ ܴ (2.63)
݆ߩܿ ܵௗ
ܵ௕ ሺ݇‫ܮ‬௕ ሻ ሺ݇‫ܮ‬௖ ሻ ൅ ܵ௖ ሺ݇‫ܮ‬௕ ሻ ሺ݇‫ܮ‬௖ ሻ

ܴൌ (2.64)
ሺ݇‫ܮ‬௕ ሻ ሺ݇‫ܮ‬௖ ሻ െ ሺܵ௖ Τܵ௕ ሻ ሺ݇‫ܮ‬௕ ሻ ሺ݇‫ܮ‬௖ ሻ
where ‫ܮ‬ௗ is the length of the main duct extending equally from either side of the
resonator, ܵௗ is the cross-section of the main duct, ‫ܮ‬௕ is the length of the side branch
(measured from main pipe centreline to start of cavity), ܵ௕ is the cross-section of the
connecting branch, ‫ܮ‬௖ is the length of the resonator cavity and ܵ௖ is the cross-section of
the cavity.
ܵ௖
‫ܮ‬௖
ܵ௕
ܵௗ ܵௗ
‫ܮ‬௕
‫ܮ‬ௗ ‫ܮ‬ௗ
Figure 2.16 A side branch Helmholtz resonator
To examine the effect of length-to-diameter ratio for a generic CPAP muffler device, a
single cylindrical expansion chamber muffler having 114 mm chamber diameter and
45 mm chamber length is now presented. Both inlet and outlet pipes have 18 mm
diameter and 20 mm length. This muffler has a significantly smaller length-to-diameter
ratio of 0.4 compared to that of the single chamber muffler examined previously (of
47
ratio 1.8). Equation (2.45) predicts that this muffler will have no complete transmission
loss domes. The cut-on frequency of the first diametral mode and the first circumferential
mode of the expansion chamber are 1.8 kHz and 3.7 kHz, respectively. Figure 2.17
shows the acoustic pressure field within the chamber at the resonant frequency. The short
length of the expansion chamber is insufficient to allow the pressure waves to become
planar along the chamber axis.
Figure 2.17 Pressure field in the single 114 mm diameter expansion chamber muffler
at the frequency coinciding with the peak transmission loss – FEA results
Figure 2.18 shows transmission loss predictions for the 114 mm diameter chamber
muffler. The FEA result departs significantly from the analytical solution, given by the
transfer matrix method with Karal correction factors, at frequencies greater than 1 kHz
and the transmission loss spectrum is dominated by a large resonant peak. Figure 2.19
presents the transmission loss obtained experimentally for the muffler and the
transmission loss predicted using the direct finite element method. The FEA results show
good agreement with the experimental results over the frequency range assessed. The
difference between the computational and experimental results in the vicinity of 4 kHz is
attributed to the FFT windowing.
48
70
TMM + Karal
60
FEA
50
40
30
20
10
0
0 1,000 2,000 3,000 4,000 5,000
Frequency (Hz)
muffler, comparing analytical results and FEA results
70
FEA
60
Experimental
50
40
30
20
10
0
0 1,000 2,000 3,000 4,000 5,000
Frequency (Hz)
muffler, comparing FEA results and experimental results
49
Chapter 3: Transmission Loss of a Resistive Muffler
Chapter 3
Transmission Loss of a Resistive Muffler
3.1 Introduction
In this chapter, a modification to the Delany-Bazley empirical expressions is presented

which facilitates the acoustic characterisation of polyurethane foam without requiring
knowledge of the airflow resistivity of the material. The acoustic characteristics of two
polyurethane foams were obtained experimentally and the corresponding properties
incorporated into finite element models of a single expansion chamber muffler design.
Results of the transmission loss of the foam-filled muffler obtained from the finite
element models are presented. The transmission loss of the muffler was measured using
the two-microphone acoustic pulse method presented in the previous chapter.
Experimental results are compared with results obtained computationally.
3.2 Computational Modelling of Porous Materials
Using the equivalent fluid approach, the properties of porous materials such as foam are
represented by a complex speed of sound ܿ௙ and a complex mean density ߩ௙ [46, 106]
߱
ܿ௙ ൌ ݆ (3.1)
ߛ
ܼ௖ ߛ
ߩ௙ ൌ െ݆ (3.2)
߱
where ܼ௖ is the characteristic impedance, ߛ is the propagation constant, ߱ is the radian

frequency and ݆ ൌ ξെͳ is the imaginary unit. The characteristic impedance and
propagation constant are complex and frequency dependent and can be written in terms
50
of their real and imaginary components as
ܼ௖ ൌ ܴ ൅ ݆ܺ (3.3)
ߛ ൌ ߙ ൅ ݆ߚ (3.4)
where ܴ, ܺ, ߙ and ߚ are all real functions of frequency. The propagation constant
consists of a real component ߙ, corresponding to the attenuation constant, and an
imaginary component ߚ, which corresponds to the phase constant [107]. In terms of the
real and imaginary components of the characteristic impedance and propagation constant,
the speed of sound and mean density can then be described by
߱ߚ ߱ߙ
ܿ௙ ൌ ൤ ൨൅݆൤ ଶ ൨ (3.5)
ሺߙ ଶ
൅ߚ ሻ
ଶ ሺߙ ൅ ߚ ଶ ሻ
ሺܴߚ ൅ ܺߙሻ ሺܺߚ െ ܴߙሻ

ߩ௙ ൌ ቈ ቉൅݆ቈ ቉ (3.6)
߱ ߱
Whilst the equivalent fluid properties can be obtained from experimental measurement of
complex ܼ௖ and ߛ, the input data to a computational finite element model will not be
continuous as they are sampled at discrete frequencies. Continuous data can be obtained
by curve fitting the experimental data to empirical power-law functions in terms of ܼ௖
and ߛ, such as those originally proposed by Delany and Bazley [70], and presented using
ఘೌ ௙
the non-dimensional normalising parameter ௥೑
[72].
ି଴Ǥ଻ହସ଴ ି଴Ǥ଻ଷଶ଴
ߩ௔ ݂ ߩ௔ ݂
ܼ௖ ൌ ߩ௔ ܿ௔ ൥ͳ ൅ ͲǤͲͷ͹ͳ ቆ ቇ െ ݆ ‫Ͳ כ‬ǤͲͺ͹Ͳ ቆ ቇ ൩ (3.7)
‫ݎ‬௙ ‫ݎ‬௙
ି଴Ǥହଽହ଴ ି଴Ǥ଻
ʹߨ݂ ߩ௔ ݂ ߩ௔ ݂
ߛൌ൬ ൰ ൥ͲǤͳͺͻͲ ቆ ቇ ൅ ݆ ൭ͳ ൅ ͲǤͲͻ͹ͺ ቆ ቇ ൱൩ (3.8)
ܿ௔ ‫ݎ‬௙ ‫ݎ‬௙
ߩ௔ and ܿ௔ are respectively the density and speed of sound in air, ݂ is the frequency in Hz
and ‫ݎ‬௙ is the airflow resistivity in MKS units.
51
The Delany-Bazley relationships are only considered to be valid for the following
range [70]
݂
ͲǤͲͳ ൑ ൑ͳ (3.9)
‫ݎ‬௙
It is well known that these empirical relations are unable to encapsulate the acoustic
behaviour of foams sufficiently to be regarded as universally applicable [74].
The Delany-Bazley functions can be expressed in terms of coefficients ‫ܥ‬ଵ to ‫ ଼ܥ‬by
஼మ ஼
ߩ௔ ݂ ߩ௔ ݂ ర
ܼ௖ ൌ ߩ௔ ܿ௔ ൥ͳ ൅ ‫ܥ‬ଵ ቆ ቇ െ ݆‫ܥ‬ଷ ቆ ቇ ൩ (3.10)
‫ݎ‬௙ ‫ݎ‬௙
஼ల ஼
ʹߨ݂ ߩ௔ ݂ ߩ௔ ݂ ఴ
ߛൌ൬ ൰ ൥‫ܥ‬ହ ቆ ቇ ൅ ݆ ൭ͳ ൅ ‫ ଻ܥ‬ቆ ቇ ൱൩ (3.11)
ܿ௔ ‫ݎ‬௙ ‫ݎ‬௙
Acoustic characterisation of a porous material can now be achieved by experimentally

obtaining the characteristic impedance ܼ௖ , propagation constant ߛand airflow resistivity
‫ݎ‬௙ of the material and fitting the data to Eqs. (3.10) and (3.11) to obtain the unknown
coefficients.
In terms of the real and imaginary components of ܼ௖ and ߛ, Eqs. (3.10) and (3.11) can be
re-stated as
ܴ ߩ௔ ݂
ଵ଴ ൬ െ ͳ൰ ൌ ‫ܥ‬ଶ ଵ଴ ቆ ቇ ൅ ଵ଴ ሺ‫ܥ‬ଵ ሻ (3.12)
ߩ௔ ܿ௔ ‫ݎ‬௙
െܺ ߩ௔ ݂
ଵ଴ ൬ ൰ ൌ ‫ܥ‬ସ ଵ଴ ቆ ቇ ൅ ଵ଴ ሺ‫ܥ‬ଷ ሻ (3.13)
ߩ௔ ܿ௔ ‫ݎ‬௙
ߙܿ௔ ߩ௔ ݂
ଵ଴ ൬ ൰ ൌ ‫ ଺ܥ‬ଵ଴ ቆ ቇ ൅ ଵ଴ ሺ‫ܥ‬ହ ሻ (3.14)
ʹߨ݂ ‫ݎ‬௙
ߚܿ௔ ߩ௔ ݂
ଵ଴ ൬ െ ͳ൰ ൌ ‫ ଼ܥ‬ଵ଴ ቆ ቇ ൅ ଵ଴ ሺ‫ ଻ܥ‬ሻ (3.15)
ʹߨ݂ ‫ݎ‬௙
52
As Eqs. (3.12) to (3.15) are of the form ‫ ݕ‬ൌ ݉‫ ݔ‬൅ ܾ, linear trend lines can be fitted to
experimental data obtained for ܼ௖ and ߛ. The coefficients ‫ܥ‬ଵ to ‫ ଼ܥ‬are then extracted
from the curves and substituted into Eqs. (3.10) and (3.11). Several authors have derived
different sets of coefficients for poroelastic materials however they are only relevant to
the particular foam for which they were obtained [71, 75, 76, 108].
The characteristic impedance and propagation constant of a porous material are easily
measured using an impedance tube and applying the transfer function method to a two-
cavity approach as described in Ref. [109]. While airflow resistivity can also be
measured, a method is described herein which simplifies the acoustic characterisation of
porous materials by enabling the coefficients in the Delany-Bazley functions to be
determined without the requirement to obtain the airflow resistivity. Equations (3.12) to
(3.15) can be re-arranged to give
஼
ܴ ߩ௔ మ
ଵ଴ ൬ െ ͳ൰ ൌ ‫ܥ‬ଶ ଵ଴ ݂ ൅ ଵ଴ ൥‫ܥ‬ଵ ቆ ቇ ൩ (3.16)
ߩ௔ ܿ௔ ‫ݎ‬௙
஼
െܺ ߩ௔ ర
ଵ଴ ൬ ൰ ൌ ‫ܥ‬ସ ଵ଴ ݂ ൅ ଵ଴ ൥‫ܥ‬ଷ ቆ ቇ ൩ (3.17)
ߩ௔ ܿ௔ ‫ݎ‬௙
஼
ߙܿ௔ ߩ௔ ల
ଵ଴ ൬ ൰ ൌ ‫ ଺ܥ‬ଵ଴ ݂ ൅ ଵ଴ ൥‫ܥ‬ହ ቆ ቇ ൩ (3.18)
ʹߨ݂ ‫ݎ‬௙
஼
ߚܿ௔ ߩ௔ ఴ
ଵ଴ ൬ െ ͳ൰ ൌ ‫ ଼ܥ‬ଵ଴ ݂ ൅ ଵ଴ ൥‫ ଻ܥ‬ቆ ቇ ൩ (3.19)
ʹߨ݂ ‫ݎ‬௙
Equations (3.16) to (3.19) are still of the form ‫ ݕ‬ൌ ݉‫ ݔ‬൅ ܾexcept now the x-axis values
are only a function of frequency, that is, ‫ݎ‬௙ is no longer needed in order to plot the data. It
is possible to obtain the equation coefficients ‫ܥ‬ଶ , ‫ܥ‬ସ , ‫ ଺ܥ‬and ‫ ଼ܥ‬directly from the slope of
linear trend lines fitted through the impedance tube experimental data. Coefficients ‫ܥ‬ଵ ,
‫ܥ‬ଷ , ‫ܥ‬ହ and ‫ ଻ܥ‬can then be obtained as a function of flow resistivity ‫ݎ‬௙ and air density ߩ௔
from the y-axis intercepts obtained from these trend lines.
53
From Eqs. (3.16) to (3.19) it can be shown that the y-intercepts produce
‫ݎ‬௙ ஼మ
‫ܥ‬ଵ ൌ ͳͲ௕ೃ ൬ ൰ (3.20)
ߩ௔
‫ݎ‬௙ ஼ర
‫ܥ‬ଷ ൌ ͳͲ௕೉ ൬ ൰ (3.21)
ߩ௔
௕ഀ
‫ݎ‬௙ ஼ల
‫ܥ‬ହ ൌ ͳͲ ൬ ൰ (3.22)
ߩ௔
‫ݎ‬௙ ஼ఴ
‫ ଻ܥ‬ൌ ͳͲ௕ഁ ൬ ൰ (3.23)
ߩ௔
where ܾோ , ܾ௑ , ܾఈ and ܾఉ are respectively the y-intercepts from the curve fits for the real
(ܴ) and imaginary (ܺ) parts of the characteristic impedance and the real (ߙ) and
imaginary (ߚ) parts of the propagation constant. Equations (3.20) and (3.21) can be
substituted directly into Eq. (3.10), and similarly, Eqs. (3.22) and (3.23) can be
substituted into Eq. (3.11) to produce
ܼ௖ ൌ ߩ௔ ܿ௔ ሾͳ ൅ ͳͲ௕ೃ ݂ ஼మ െ ݆ͳͲ௕೉ ݂ ஼ర ሿ (3.24)
ʹߨ݂
ߛൌ൬ ൰ ൣͳͲ௕ഀ ݂ ஼ల ൅ ݆൫ͳ ൅ ͳͲ௕ഁ ݂ ஼ఴ ൯൧ (3.25)
ܿ௔
The coefficients ‫ܥ‬ଶ , ‫ܥ‬ସ , ‫ ଺ܥ‬and ‫ ଼ܥ‬and y-intercepts ܾோ , ܾ௑ , ܾఈ and ܾఉ are extracted from
the curve fits and substituted into Eqs. (3.24) and (3.25), producing expressions for ܼ௖
and ߛ as a function of air density, speed of sound in air and frequency. There is no longer
any requirement to obtain the airflow resistivity of the material.
It has been observed that for use of the original Delany-Bazley functions in terms of the
eight unknown coefficients as given by Eqs. (3.10) and (3.11), the coefficients ‫ܥ‬ଵ , ‫ܥ‬ଷ , ‫ܥ‬ହ
and ‫ ଻ܥ‬can also be obtained in the absence of airflow resistivity data using an arbitrary
value for ‫ݎ‬௙ , provided the same arbitrary value for ‫ݎ‬௙ is also used in the Delany-Bazley
equations. This is described in more detail in Appendix B. While Lee et al. [110]
presented equations which are very similar to Eqs. (3.24) and (3.25), the procedure
described in Appendix B is novel. It allows the coefficients obtained in the absence of
54
airflow resistivity data to be directly used in the standard Delany-Bazley models

incorporated into commercial acoustic FEA software packages.
3.3 Experimental Methods
The experimental procedure to obtain the characteristic impedance ܼ௖ , propagation

constant ߛ and airflow resistivity ‫ݎ‬௙ of different foam materials is described in what
follows. Two different polyurethane foam materials were selected for comparison. The
first foam (light grey) has an apparent density of 34 kg/m3 and is a material currently
being used in CPAP device mufflers. The second foam (dark grey) has an apparent
density of 23 kg/m3 and is commonly used in protective packaging.
Figure 3.1 Photographs of the polyurethane foam specimens – light grey (left) and
dark grey (right)
3.3.1 Characteristic impedance and propagation coefficient
The characteristic impedance and propagation coefficient of porous materials can be

measured by applying the transfer function method to a two-cavity approach [109]. A
sample of homogeneous porous material of thickness ‫ݐ‬௙ was positioned within a Brüel &
Kjær Type 4206 impedance tube and against the front face of a moveable plunger. The
plunger was then withdrawn away from the sample, producing an air cavity with a
known air cavity depth ‫ ܮ‬between the rear face of the sample and the plunger, as shown
in Fig. 3.2. A random signal was fed to the loudspeaker of the impedance tube and the
normal surface acoustic impedance of the sample was measured in accordance with
ISO 10534-2 [111].
55
The transfer function ‫ܪ‬ଵଶ from microphone position 1 to position 2, defined by the
௣
complex ratio ௣మ, was measured using a two channel Fast Fourier transform. The surface
భ
acoustic impedance ܼ଴ is then obtained by [112]
‫ܪ‬ଵଶ sinሾ݇௔ ሺ‫ܮ‬௫ ൅ ‫ܦ‬௫ ሻሿ െ sinሺ݇௔ ‫ܮ‬௫ ሻ

ܼ଴ ൌ ݆ܼ௔ ቊ ቋ (3.26)
cosሺ݇௔ ‫ܮ‬௫ ሻെ‫ܪ‬ଵଶ cosሾ݇௔ ሺ‫ܮ‬௫ ൅ ‫ܦ‬௫ ሻሿ
where ݇௔ is the wavenumber of air and ܼ௔ ( ൌ ߩ௔ ܿ௔ ) is the characteristic impedance of

air.
Ͳ ͳ
‫ܦ‬௫ ‫ܮ‬௫ ‫ݐ‬௙ Movable
Speaker plunger
Mic. 1 Mic. 2
Impedance tube Porous sample Air cavity
Figure 3.2 Schematic diagram of the impedance tube configuration
The impedance tube plunger was withdrawn a further distance and the measurement
procedure was repeated at depth ‫ܮ‬ᇱ to obtain ܼ଴ᇱ . The theoretical impedance of closed
tubes with depths ‫ ܮ‬and ‫ܮ‬ᇱ is given by [109]
ܼଵ ൌ െ݆ܼ௔ ሺ݇௔ ‫ܮ‬ሻ (3.27)
ܼଵƍ ൌ െ݆ܼ௔ ሺ݇௔ ‫ܮ‬ƍ ሻ (3.28)
The characteristic impedance and propagation constant of the material can then be
56
calculated by [109]
ܼ଴ ܼ଴ƍ ൫ܼଵ െ ܼଵƍ ൯ െ ܼଵ ܼଵƍ ൫ܼ଴ െ ܼ଴ƍ ൯

ܼ௖ ൌ േඨ (3.29)
൫ܼଵ െ ܼଵƍ ൯ െ ൫ܼ଴ െ ܼ଴ƍ ൯
ͳ ܼ௢ ൅ ܼ௖ ܼଵ െ ܼ௖
ߛ ൌ ቆ ቇ ln ൤൬ ൰൬ ൰൨ (3.30)
ʹ‫ݐ‬௙ ܼ௢ െ ܼ௖ ܼଵ ൅ ܼ௖
where the sign in Eq. (3.29) is selected so that the real part of ܼ௖ is positive.
Figure 3.3 Photograph of the Brüel & Kjær LAN-XI data acquisition system and
Type 4206 impedance tube
The normal surface impedance for each foam type was measured and calculated using the
test method described in ISO 10534-2 [111]. Measurements were obtained at four cavity
depths corresponding to 25, 50, 75 and 100 mm, using samples of 25 mm thickness. The
characteristic impedance and propagation coefficient were calculated for each of the
cavity combinations 25/50, 50/75 and 75/100 (in mm) using Eqs. (3.29) and (3.30) and
57
the results for the three combinations were averaged. The coefficients ‫ܥ‬ଶ , ‫ܥ‬ସ , ‫ ଺ܥ‬and ‫଼ܥ‬
and y-intercepts ܾோ , ܾ௑ , ܾఈ and ܾఉ were extracted from the experimental data for ܼ௖ and
ߛ, and are summarised in Table 3.1. These coefficients and y-intercepts can now be
substituted into Eqs. (3.24) and (3.25) to obtain continuous data for ܼ௖ and ߛ for input
into a finite element model.
Table 3.1 Derived Delany-Bazley function coefficients (revised form)
Dark grey foam Light grey foam

Parameters
Coefficient R2 Coefficient R2
ܾோ 0.0595 0.8597
ܴ ൌ ሺܼ௖ ሻ 0.58 0.92
‫ܥ‬ଶ -0.2249 -0.3659
ܾ௑ 0.6822 1.3567
ܺ ൌ ሺܼ௖ ሻ 0.72 0.88
‫ܥ‬ସ -0.4851 -0.6144
ܾఈ 1.1095 1.4337
ߙ ൌ ሺߛሻ 0.98 0.99
‫ ଺ܥ‬ -0.5416 -0.5728
ܾఉ 0.4634 1.2015
ߚ ൌ ሺߛሻ 0.96 0.97
‫ ଼ܥ‬ -0.3111 -0.4657
Results in Table 3.1 show that the propagation coefficient of both foam types correlate
well with the frequency, producing correlation coefficients between 0.96 and 0.99. The
characteristic impedance of the light grey foam also correlates well, producing
correlation coefficients between 0.88 and 0.92. These observations are consistent with
the findings of Wu [76] who reported correlation coefficients between 0.85 and 0.99 for
porous plastic open-celled foams. While the correlation coefficients for the characteristic
impedance of the dark grey foam are less encouraging (0.58 and 0.72), this result is
attributed to sample preparation and the influence of circumferential edge constraint on
the impedance tube measurements [66, 113-116]. Samples of both foam types were
prepared using the same set of cutting dies. Cutting an undersized sample leads to a loose
fit within the impedance tube and the possibility that the sample will undergo sliding
58
movement. This movement may occur during impedance tube assembly or it may be
driven by acoustic excitation. Cutting an oversized sample leads to some compression of
the sample when locating it within the impedance tube. This compression reduces the
likelihood that the sample will slide within the tube, however it also changes the stiffness
and porosity of the sample which leads to erroneous results. Although each die was
manufactured with an internal diameter 1.0 mm larger than that of the corresponding
impedance tube, the cutting process produced samples which were only very marginally
oversized. The minimal compression of these samples, when located within the
impedance tubes, produces only slight resistance to sliding. The structure of the dark
grey foam is less dense, less elastic and has a significantly lower coefficient of friction
than the light grey foam. These differences in physical attributes result in the dark grey
foam samples being more susceptible to movement within the impedance tube, especially
at resonant frequencies.
The experimental results for the normalised complex characteristic impedance and
complex propagation constant of the light grey and dark grey foams are shown in
Figs. 3.4 and 3.5, respectively. The trends exhibited by the properties of each of the
foams are very similar across the entire frequency range. Both the characteristic
impedance and the propagation constant of the light grey foam are greater than that of the
dark grey foam across the measured frequency range.
The numerical and experimental values for the characteristic impedance and propagation
constant of the light grey foam are presented in Figs. 3.6 and 3.7, respectively. The
numerical values, obtained using Eqs. (3.24) and (3.25) and the coefficients summarised
in Table 3.1, show excellent agreement with the experimental data. This is not
unexpected as the numerical values were derived using the same set of experimental data
and the correlation coefficients were good. Although not presented here, the numerical
and experimental values for ܼ௖ and ߛ for the dark grey foam also show comparable
agreement, with only slight deviation between the numerical and experimental results at
frequencies below 250 Hz. This deviation is attributed to the lower correlation
coefficients for the dark grey foam as given in Table 3.1.
59
3.5

(light grey foam) (light grey foam)
Characteristic Impedance (normalised)
2.5
(dark grey foam) (dark grey foam)

1.5
0.5
-0.5
-1.5
-2.5
0 1,000 2,000 3,000 4,000 5,000
Frequency (Hz)
Figure 3.4 Normalised complex characteristic impedance of the light grey foam and
dark grey foam – experimental data
140
Ƚ (light grey foam) Ⱦ (light grey foam)
120
Ƚ (dark grey foam) Ⱦ (dark grey foam)
100
Propagation constant
80
60
40
20
0
0 1,000 2,000 3,000 4,000 5,000
Frequency (Hz)
Figure 3.5 Complex propagation constant of the light grey foam and dark grey foam
– experimental data
60
3.5

(numerical) (numerical)
Characteristic Impedance (normalised)
2.5
(experimental) (experimental)

1.5
0.5
-0.5
-1.5
-2.5
0 1,000 2,000 3,000 4,000 5,000
Frequency (Hz)
Figure 3.6 Normalised complex characteristic impedance of the light grey foam,
comparing numerical model values and experimental data
140
Ƚ (numerical) Ⱦ (numerical)
120
Ƚ (experimental) Ⱦ (experimental)
100
Propagation constant
80
60
40
20
0
0 1,000 2,000 3,000 4,000 5,000
Frequency (Hz)
Figure 3.7 Propagation constant of the light grey foam, comparing numerical model
values and experimental data
61
3.3.2 Airflow resistivity
The airflow resistivity of a homogeneous material is given by
οܲ
‫ݎ‬௙ ൌ (3.31)
‫ݐ‬௙ ‫ݑ‬
where οܲ is the static pressure drop across the material, ‫ݐ‬௙ is the sample thickness and ‫ݑ‬
is the linear velocity of air passing through it [117]. Airflow resistivity for each foam
type was measured according to the direct airflow method described in ISO 9053 [117].
A unidirectional airflow was passed through cylindrical samples having 25 mm thickness
and 100 mm diameter and the resulting pressure drop between the two free faces of the
sample was measured. A schematic diagram and photo of the airflow resistivity
experimental rig are respectively shown in Figs. 3.8 and 3.9.
Data was also recorded at linear airflow velocities greater than the 4 mm/s upper limit
recommended by ISO 9053 to establish the effect of turbulent flow on the apparent
airflow resistivity for the foams being studied. The values for airflow resistivity
calculated using data within the laminar range are presented in Table 3.2 and it can be
seen that the measured airflow resistivity of the two foam types is significantly different.
This finding is consistent with the observed difference in surface pore sizes and spacing.
Sample holder
Differential
pressure
Sample P
measuring
device
Flow Flow
Flow diffuser
generator meter
Figure 3.8 Schematic diagram of the airflow resistivity experimental set-up
62
Figure 3.9 Photograph of the airflow resistivity experimental set-up
Figure 3.10 shows that the apparent airflow resistivity for the light grey foam increases
as the linear airflow is increased beyond the laminar region, while the apparent airflow
resistivity of the dark grey foam remains largely unaffected. This difference in observed
behaviour is significant as the Delany-Bazley method uses a single value for flow
resistivity to characterise the porous material.
Table 3.2 Foam airflow resistivity of the light and dark grey foams – experimental
data
Flow resistivity 95% confidence

Foam
(Rayls/m) interval
Light grey (acoustic) 8,445 182

Dark grey (non-acoustic) 2,652 36
The experimentally measured characteristic impedance ܼ௖ , propagation constant ߛand

airflow resistivity ‫ݎ‬௙ are now curve fitted to Eqs. (3.10) and (3.11) to obtain the unknown
63
coefficients ‫ܥ‬ଵ to ‫ ଼ܥ‬. From Eq. (3.9) the valid frequency ranges for the Delany-Bazley
power law functions are 25 to 2,650 Hz for the dark grey foam and 85 to 8,450 Hz for the
light grey foam.
(x10^3) 11
Turbulent
10
9
Airflow resistivity (Pa.s/m^2)
7 Light grey foam
6 Dark grey foam
1
0 50 100 150 200 250
Linear airflow velocity(mm/s)
Figure 3.10 Airflow resistivity of the light and dark grey foams – experimental results
3.4 Equivalent Fluid Parameters
Equations (3.1) and (3.2) were used to obtain the complex speed of sound and complex
density of the two foam materials using the derived coefficients (‫ܥ‬ଵ to ‫ ) ଼ܥ‬for the
Delany-Bazley empirical formulae given by Eqs. (3.10) and (3.11) and the measured
airflow resistivity. The experimentally derived coefficients for both foam types are listed
in Table 3.3. Figures 3.11 and 3.12 respectively show the equivalent fluid speed of sound
and equivalent fluid density for both foam materials. The speed of sound is greater in the
dark grey foam than it is in the light foam and the dark grey foam has a lesser density
than the light grey foam across the frequency range. These trends are consistent with the
measured apparent density and observation of porosity and stiffness of each material.
64
400
Dark grey foam (real) Dark grey foam (imaginary)
350
Light grey foam (real) Light grey foam (imaginary)
300
Speed of sound (m/s)
250
200
150
100
50
0
0 1,000 2,000 3,000 4,000 5,000
Frequency (Hz)
Figure 3.11 Equivalent fluid speed of sound in the light and dark grey foams –
coefficients
10
0
Density (kg/m^3)
-5
-10
-15
Dark grey foam (real) Dark grey foam (imaginary)
-20
Light grey foam (real) Light grey foam (imaginary)
-25
0 1,000 2,000 3,000 4,000 5,000
Frequency (Hz)
Figure 3.12 Equivalent fluid density of the light and dark grey foams – numerical
model values obtained using the derived Delany-Bazley coefficients
65
Table 3.3 Delany-Bazley function coefficients, comparing experimentally derived

values for the light and dark grey foams and the original Delany-Bazley
values
Dark grey Light grey Delany-

Parameters
foam foam Bazley
‫ܥ‬ଵ 0.2022 0.2814 0.0571

ܴ ൌ ሺܼ௖ ሻ
‫ܥ‬ଶ -0.2249 -0.3659 -0.7540
‫ܥ‬ଷ 0.1138 0.0974 0.0870

ܺ ൌ ሺܼ௖ ሻ
‫ܥ‬ସ -0.4851 -0.6144 -0.7320
‫ܥ‬ହ 0.1969 0.1682 0.1890

ߙ ൌ ሺߛሻ
‫ ଺ܥ‬ -0.5416 -0.5728 -0.5950
‫ ଻ܥ‬ 0.2634 0.2549 0.0978

ߚ ൌ ሺߛሻ
‫ ଼ܥ‬ -0.3111 -0.4657 -0.700
Figures 3.13 and 3.14 respectively compare the equivalent fluid speed of sound and
equivalent fluid density for the light grey foam using the original Delany-Bazley
coefficients as per Eqs. (3.7) and (3.8), and the derived coefficients as per Eqs. (3.10),
(3.11) and Table 3.3. In both cases, the measured airflow resistivity of the light grey
foam was used in the Delany-Bazley empirical formulae. The speed of sound in the light
grey foam predicted using the original Delany-Bazley coefficients shows poor agreement
with the results using the derived coefficients. Although the real component of the speed
of sound predicted using the original coefficients aligns with that predicted using the
derived coefficients at low frequencies, it quickly increases at a greater rate. The
imaginary component predicted using the original coefficients is approximately 50%
greater than that predicted using the derived coefficients across the majority of the
frequency range. Use of the original Delany-Bazley coefficients will result in an over-
prediction of the equivalent fluid speed of sound for the light grey foam. Figure 3.14
shows that use of the original Delany-Bazley coefficients to predict the equivalent fluid
density of the light grey foam will result in an under-prediction of the density.
66
400
Derived coefficients (real) Derived coefficients (imaginary)
350
Original coefficients (real) Original coefficients (imaginary)
300
250
200
150
100
50
0
0 1,000 2,000 3,000 4,000 5,000
Frequency (Hz)
numerical model values obtained using original and derived Delany-
Bazley coefficients
10
0
Density (kg/m^3)
-5
-10
-15
Derived coefficients (real) Derived coefficients (imaginary)

-20
Original coefficients (real) Original coefficients (imaginary)
-25
0 1,000 2,000 3,000 4,000 5,000
Frequency (Hz)
Figure 3.14 Equivalent fluid density of the light grey foam, comparing numerical
model values obtained using original and derived Delany-Bazley
coefficients
67
Figures 3.15 and 3.16 respectively show the equivalent fluid speed of sound and
equivalent fluid density for the light grey foam using (i) the derived Delany-Bazley
coefficients listed in Table 3.3 and (ii) the revised coefficients ‫ܥ‬ଶ , ‫ܥ‬ସ , ‫ ଺ܥ‬, ‫ ଼ܥ‬, ܾோ , ܾ௑ , ܾఈ ,
ܾఉ listed in Table 3.1. The former case (i) requires the measured airflow resistivity as per
Eqs. (3.10) and (3.11), whilst case (ii) does not as per Eqs. (3.24) and (3.25). The results
predicted for both material properties using the two approaches are identical and
confirms that the latter approach is able to produce valid predictions of the foam
equivalent fluid properties without the requirement to obtain the airflow resistivity of the
material.
3.5 Muffler Transmission Loss
The acoustic finite element model of the single 114 mm diameter expansion chamber
muffler, introduced in Chapter 2, was modified to incorporate both air-filled and foam-
filled regions. The foam inserts were modelled using the Delany-Bazley formulation
described by Eqs. (3.10) and (3.11). The experimentally derived coefficients listed in
Table 3.3 were used with the Delany-Bazley functions to obtain the equivalent fluid
parameters. Transmission loss is calculated directly in the finite element model using the
acoustic power at the inlet and outlet ports of the muffler. Experiments were also
conducted using the two-microphone acoustic pulse method described in Chapter 2. The
results of the transmission loss of the muffler obtained computationally and
experimentally are compared in what follows.
Figure 3.17 presents the transmission loss obtained experimentally for the 114 mm
diameter cylindrical expansion chamber muffler, with and without a light grey foam
insert present, and the transmission loss predicted by the finite element model. The FEA
results show good agreement with the experimental results over the frequency range
assessed. The inclusion of the foam insert has little impact in the lower frequency range
up to 2 kHz. In the medium frequency band (between 2 and 3 kHz), the inclusion of foam
has significantly reduced the attenuation due to the peak at around 2.5 kHz. At higher
frequencies (>3 kHz), the inclusion of foam produces a 5-10 dB increase in performance.
68
400
ĞůĂŶǇͲĂǌůĞǇĨŽƌŵ;ƌĞĂůͿ ĞůĂŶǇͲĂǌůĞǇĨŽƌŵ;ŝŵĂŐŝŶĂƌǇͿ
350
ZĞǀŝƐĞĚĨŽƌŵ;ƌĞĂůͿ ZĞǀŝƐĞĚĨŽƌŵ;ŝŵĂŐŝŶĂƌǇͿ
300
250
200
150
100
50
0
0 1,000 2,000 3,000 4,000 5,000
Frequency (Hz)
numerical model values obtained using derived coefficients in the original
Delany-Bazley functions and restated functions (no airflow resistivity)
10
0
Density (kg/m^3)
-5
-10
-15
ĞůĂŶǇͲĂǌůĞǇĨŽƌŵ;ƌĞĂůͿ ĞůĂŶǇͲĂǌůĞǇĨŽƌŵ;ŝŵĂŐŝŶĂƌǇͿ
-20
ZĞǀŝƐĞĚĨŽƌŵ;ƌĞĂůͿ ZĞǀŝƐĞĚĨŽƌŵ;ŝŵĂŐŝŶĂƌǇͿ
-25
0 1,000 2,000 3,000 4,000 5,000
Frequency (Hz)
Figure 3.16 Equivalent fluid density of the light grey foam, comparing numerical
model values obtained using derived coefficients in the original Delany-
Bazley functions and restated functions (no airflow resistivity)
69
The transmission loss averaged over the entire frequency range with and without the
inclusion of the light grey foam is 22.6 and 22.0 dB, respectively, demonstrating that
while the impact of the foam insert varies across the frequency range it has a negligible
impact on the averaged transmission loss of this muffler.
Figure 3.18 compares the transmission loss obtained computationally and experimentally
for the cylindrical muffler using the light and dark grey foams. The results show that both
foam inserts have a very similar impact on the acoustic performance of this muffler
design despite a difference in apparent density of approximately 50%. While the FEA
results show good agreement with the experimental results for the light grey foam, the
correlation for the results obtained using dark grey foam is not especially close in the
medium frequency range. This is attributed to the use of two, 22 mm thick dark grey
foam disks as it was not possible to obtain a sample of this foam type in the required
45 mm thickness. The use of two disks in the muffler chamber introduces a discontinuity
in the foam structure which disrupts the transmission of structural vibration through the
frame of the foam inserts. The combined thickness of the two disks is 1 mm less than the
expansion chamber length, further exacerbating the effect of this discontinuity. The light
grey foam insert consisted of a single 45 mm thick disk.
Figures 3.19 and 3.20 compare the transmission loss obtained computationally using the
original Delany-Bazley coefficients in Eqs. (3.7) and (3.8) with that obtained using the
derived coefficients listed in Table 3.3. These figures present the results for the light grey
and dark grey foam-filled cylindrical muffler, respectively. The transmission loss
obtained experimentally for each of the foam types is also shown. For both foam types,
the computational results obtained using the derived coefficients are more closely aligned
to the experimental measurements than those obtained using the original coefficients.
70
70
Empty muffler - FEA Light grey filled - FEA
60
Empty muffler - Experimental Light grey filled - Experimental
50
40
30
20
10
0
0 1,000 2,000 3,000 4,000 5,000
Frequency (Hz)
muffler with and without light grey foam insert, comparing FEA results
and experimental results
70
Light grey filled - FEA Dark grey filled - FEA
60
Light grey filled - Experimental Dark grey filled - Experimental
50
40
30
20
10
0
0 1,000 2,000 3,000 4,000 5,000
Frequency (Hz)
muffler with foam inserts, comparing FEA results and experimental results
71
70
FEA - original Delany-Bazley coefficients
60
FEA - derived coefficients
50 Experimental
40
30
20
10
0
0 1,000 2,000 3,000 4,000 5,000
Frequency (Hz)
Figure 3.19 Transmission loss of the light grey foam-filled single 114 mm diameter
expansion chamber muffler, comparing FEA results obtained using
original and derived Delany-Bazley coefficients and experimental results
70
FEA - original Delany-Bazley coefficients
60
FEA - derived coefficients
50 Experimental
40
30
20
10
0
0 1,000 2,000 3,000 4,000 5,000
Frequency (Hz)
Figure 3.20 Transmission loss of the dark grey foam-filled single 114 mm diameter
expansion chamber muffler, comparing FEA results obtained using
original and derived Delany-Bazley coefficients and experimental results
72
Chapter 4: Transmission Loss of CPAP Device Mufflers
Chapter 4
Transmission Loss of CPAP Device Mufflers
4.1 Introduction
In this chapter, the transmission loss of one production and two prototype CPAP device
muffler designs are predicted analytically, computationally and experimentally. Finite
element models of each reactive muffler are developed and evaluated against results
obtained using the analytical techniques presented in Chapter 2. Two acoustic finite
element analysis software packages are compared to validate the computational results
and to assess the feasibility of using commercially available programs for muffler
analysis. The transmission loss of each CPAP device muffler was measured using a two-
microphone acoustic pulse method for comparison with both the analytical and
computational results.
The acoustic characteristics of the two polyurethane foams presented in Chapter 3 are
incorporated into finite element models of the production CPAP device muffler and one
of the prototype designs. The transmission loss of each resistive muffler was
experimentally measured and compared to results obtained computationally.
4.2 Muffler Designs
The first design shown in Fig. 4.1 consists of a single unbaffled chamber having coaxial
inlet and outlet ports. The close proximity between the inlet and outlet ports creates a
narrow, short opening between the ports and the muffler chamber. While the level of
geometric detail in the design is high, the underlying configuration is that of a Helmholtz
resonator. The acoustic characteristics of this design are well known and, as such, it
serves as a suitable design for initial comparison between the analytical, computational
and experimental results. The characteristic dimension of the muffler chamber will result
in the propagation of higher order modes within the chamber at frequencies within the
desired attenuation range.
73
The second design shown in Fig. 4.2 consists of two integrated chambers and presents a
complex path between the inlet and outlet ports. Air flowing through the device is
deflected around a vertical internal baffle before passing through a narrow slot into the
final chamber.
Figure 4.1 Air volume of the production single chamber CPAP device muffler
Figure 4.2 Air volume of the prototype integrated multi-chamber CPAP device
muffler
74
The third design shown in Fig. 4.3 consists of three interconnected expansion chambers
each having orthogonal inlet and outlet ports. The chambers are geometrically simple and
contain no internal baffles. A cylindrical pipe of 43 mm length and 18 mm internal
diameter connects each chamber to isolate through-wall noise transmission between
adjacent chambers. The dimensions of each muffler design are given in Table 4.1. The
lengths are measured in the direction normal to the chamber inlet and the cross-sectional
areas are calculated by dividing the total chamber volume by its length.
Figure 4.3 Air volume of the prototype interconnected three-chamber CPAP device
muffler
Table 4.1 Dimensions of three CPAP device muffler designs
Chamber 1 Chamber 2 Chamber 3

Design
‫ܮ‬ଵ (mm) ܵଵ (mm2) ‫ܮ‬ଶ (mm) ܵଶ (mm2) ‫ܮ‬ଷ (mm) ܵଷ (mm2)
1 35 5,728 - - - -
2 50 6,489 28 2,770 - -
3 40 5,920 47 7,077 27 7,247
75
4.3 Finite Element Modelling of CPAP Device Mufflers
Computational models of each muffler design were developed using two commercially
available finite element analysis software packages – ANSYS and COMSOL.
Replication of the analyses using the two packages was undertaken to validate the
computational results and to compare the relative capabilities of each of these products in
the context of modelling small reactive type mufflers. Two different boundary conditions
were chosen in order to compare the direct (acoustic power) and indirect (transfer matrix)
approaches for evaluating the transmission loss. COMSOL was used exclusively for
modelling the resistive muffler designs due to limitations in the ANSYS package,
specifically an inability to support the complex density required in the modelling of
foam.
4.3.1 ANSYS
Transmission loss was calculated in ANSYS (version 11.0) by applying the transfer
matrix methodology using ANSYS Parametric Design Language (APDL), a scripting
language that may be used to customise the FEA workflow. The form of the transfer
matrix used was
‫݌‬ଵ ‫ܣ‬ ‫ܤ‬ ‫݌‬ଶ

ቈ ቉ൌቈ ቉ቈ ቉ (4.1)
‫ݑ‬ଵ ‫ܥ‬ ‫ݑ ܦ‬ଶ
where ‫ ݌‬and ‫ ݑ‬are the acoustic pressure and the particle velocity, respectively. The
transfer matrix parameters were calculated using two separate analyses which share the
same inlet boundary condition ሺ‫ݑ‬ଵ ൌ ͳሻ but apply the two different downstream
boundary conditions given by cases 1 and 2.
‫݌‬ଵ ‫ݑ‬ଵ
Case 1: ‫ܣ‬ൌ ȁ ‫ܥ‬ൌ ȁ (4.2, 4.3)
‫݌‬ଶ ௨భ ୀଵǡ௨మ ୀ଴ ‫݌‬ଶ ௨భ ୀଵǡ௨మୀ଴
‫݌‬ଵ ‫ݑ‬ଵ
Case 2: ‫ܤ‬ൌ ȁ ‫ܦ‬ൌ ȁ (4.4, 4.5)
‫ݑ‬ଶ ௨భ ୀଵǡ௣మ ୀ଴ ‫ݑ‬ଶ ௨భ ୀଵǡ௣మ ୀ଴
Case 1 represents a rigidly fixed end termination while case 2 represents a totally free
termination. Pressure and velocity data at the muffler inlet and the muffler outlet were
76
obtained over the desired frequency range. The transfer matrix parameters were
calculated and the transmission loss was then obtained using
ͳ ܵ௜௡ ଵΤଶ ‫ܤ‬

ܶ‫ ܮ‬ൌ ʹͲ ଵ଴ ቈ ൬ ൰ ฬ‫ ܣ‬൅ ൅ ‫ ܿߩܥ‬൅ ‫ܦ‬ฬ቉ (4.6)
ʹ ܵ௢௨௧ ߩܿ
where ܵ௜௡ and ܵ௢௨௧ are the cross-sectional areas of the inlet duct and outlet duct,
respectively.
Each model was meshed using tetrahedral FLUID30 elements with mesh controls applied
to adequately resolve the fine details and tight radii in the muffler geometries. The
resulting mesh size for the single chamber muffler produced 25 elements per acoustic
wavelength at the upper bound of the frequency range being analysed (limiting case).
This is very high compared a widely accepted reasonable minimum mesh density of 5 or
6 elements per wavelength [96]. The mesh sizes for the integrated muffler and three
chamber muffler designs were not as fine as that of the single chamber muffler but were
controlled to be finer than the 6 elements per wavelength guidance. A harmonic input
velocity equal to unity was specified for each case using an equivalent displacement
boundary condition. The fluid (air) was assumed to be non-flowing and inviscid.
Acoustic damping was not applied at the fluid-structure interface, that is, the walls were
treated as acoustically hard boundaries.
4.3.2 COMSOL
Transmission loss was calculated directly in COMSOL (version 4.0a) using the acoustic
power at the inlet and outlet of the acoustic system.
ȁ௧ଶ ȁ ܵ௢௨௧
ܶ‫ ܮ‬ൌ െͳͲ ଵ଴ ቈ ଶ ቉ (4.7)
ห௜ ห ܵ௜௡
This method of calculating transmission loss offers a significant advantage over the
transfer matrix methodology in that only one load case is required, thereby reducing
computational effort. The corresponding disadvantage with this method is that it does not
allow the transfer matrix for the acoustic system to be obtained. While the solution time
per frequency is reduced significantly as a percentage when calculating transmission loss
77
directly, the absolute differences in computational run times were not significant due to
the small size of the mufflers and corresponding finite element models.
A model of each muffler design was developed using COMSOL. Each model was
meshed using Lagrange-quadratic elements with controls applied to achieve a mesh
consistent in size and distribution to that generated in the corresponding ANSYS model.
A harmonic pressure of 1 Pa was specified at the muffler inlet and a radiation boundary
condition was applied at the muffler inlet and muffler outlet. Acoustic damping was not
applied at the fluid-structure interfaces to allow comparison with the ANSYS results.
4.3.3 Transmission Loss of the CPAP Device Mufflers
Results of the transmission loss of each CPAP device muffler obtained using the two
different finite element analysis software packages are presented in what follows.
Figures 4.4 to 4.6 compare the transmission loss results obtained for the single chamber,
integrated chamber and interconnected chamber mufflers, respectively. In each figure,
both ANSYS and COMSOL are consistent in their prediction of the resonant frequencies
and prediction of the magnitude of the transmission loss at each resonance. The
difference in the magnitude of the peak transmission loss at around 3.2 kHz for the
integrated chamber muffler in Fig. 4.5 is attributed to the very narrow width of the peak
and the underlying frequency step resolution used in the two analyses. Differences in
results obtained from the software packages can also be attributed to (i) the finite element
formulation used, that is, linear elements in ANSYS versus quadratic Lagrange elements
in COMSOL, (ii) the different approaches used to obtain transmission loss (the use of
transfer matrices versus sound power reduction), and (iii) the different computational
meshes (density and refinement) that inherently result from automated tetrahedral mesh
generation tools.
4.4 Reactive CPAP Device Muffler Designs
The transmission loss of the single production and two prototype CPAP device muffler
designs were experimentally measured using the two-microphone acoustic pulse method
described in Chapter 2. Results are compared with those obtained computationally using
the COMSOL software package. The results obtained for the single chamber muffler are
78
80
ANSYS
70
COMSOL
60
50
40
30
20
10
0
0 1,000 2,000 3,000 4,000
Frequency (Hz)
Figure 4.4 Transmission loss of the single chamber CPAP device muffler, comparing
ANSYS and COMSOL FEA results
100
ANSYS
80 COMSOL
60
40
20
0
0 1,000 2,000 3,000 4,000
Frequency (Hz)
comparing ANSYS and COMSOL FEA results
79
140
ANSYS
120
COMSOL
100
80
60
40
20
0
0 1,000 2,000 3,000 4,000
Frequency (Hz)
Figure 4.6 Transmission loss of the interconnected three-chamber CPAP device

muffler, comparing ANSYS and COMSOL FEA results
also compared to predictions obtained using the analytical transfer matrix method in
order to demonstrate that the analytical approach is inadequate for complex muffler
designs.
The analytical transfer matrix approach and the corresponding matrices for an expansion
chamber and a resonator comprising a side branch with a closed cavity were presented in
Chapter 2. Figure 4.7 compares the transmission loss obtained experimentally for the
single chamber muffler and the transmission loss predicted by the analytical transfer
matrix method, both as an expansion chamber and side branch Helmholtz resonator.
While both analytical models exhibit a poor correlation with the experimental results, the
resonator model aligns more closely with the measured results than the expansion
chamber model. This is attributed to the close proximity between the coaxial inlet and
outlet ports. The poor correlation between analytical predictions and experimental results
demonstrates that the one dimensional analytical methods are unreliable beyond very
simple designs.
80
Figure 4.8 presents the transmission loss obtained experimentally for the single chamber
muffler and computationally by the finite element model. The FEA results show
excellent agreement with the experimental results over the frequency range assessed,
with the exception that the magnitude at resonant frequencies is over-predicted by the
model. This is attributed to the assumptions of totally rigid walls and no acoustic
damping in the finite element model. Reducing wall compliance has been shown to
reduce resonant frequencies in the transmission loss results [118]. The peaks and troughs
in the FEA results are also significantly sharper than those measured experimentally.
This is attributed to the assumption of an inviscid working fluid in the FEA models
which has the effect of understating the acoustic damping. The computational analyses
were conducted using frequency increments of 10 Hz whereas the resolution of the
experimental FFT results was 1 Hz. The coarser resolution of the FEA results was
required to maintain reasonable computational run times. Increasing the frequency
resolution of the FEA models may further sharpen the appearance of the peaks and
troughs but is unlikely to otherwise significantly alter the transmission loss curve.
4.4.2 Integrated chamber muffler
Figure 4.9 compares the transmission loss obtained experimentally for the integrated
chamber muffler and computationally by the finite element model. The FEA results show
good agreement up to 2.3 kHz. The underlying trend followed by the experimental and
FEA results is similar over the remaining higher frequency range with the exception of
the large double peak predicted by the finite element model.
4.4.3 Interconnected three-chamber muffler
Figure 4.10 compares the transmission loss obtained experimentally for the
interconnected chamber muffler and computationally by the finite element model. The
FEA results show good agreement with the experimental results for the majority of the
frequency range. The oscillation in the experimental results at 600 Hz is attributed to the
effect of wall vibration. A similar, albeit smaller, oscillation can be observed in the
experimental results for the integrated chamber muffler (Fig. 4.9). These two mufflers are
of similar scale and manufactured from the same prototyping material, while the single
chamber muffler was manufactured from a more rigid ABS plastic. Departure between
81
Chapter 4: Transmission Loss of CPAP
C Device Mufflers
Figure 4.7 Transm

mission loss of the single chamber CPAP deviice muffler, comparing
analyticcal results and experimental results
70
FEA (CO
OMSOL)
60 Experime
ental
50
40
30
20
10
0
0 1,000 2,000 3,00
00 4,000
Frequency (Hz)
Figure 4.8 Transm

mission loss of the single chamber CPAP deviice muffler comparing
FEA ressults and experimental results
82
100
FEA (COMSOL)
80 Experimental
60
40
20
0
0 1,000 2,000 3,000 4,000
Frequency (Hz)
comparing FEA results and experimental results
140
FEA (COMSOL)
120
Experimental
100
80
60
40
20
0
0 1,000 2,000 3,000 4,000
Frequency (Hz)

muffler, comparing FEA results and experimental results
83
the experimental and FEA results at higher frequencies is attributed to experimental

limitations with the initial acoustic pulse generation and subsequent rejection of reflected
pulses. The two-microphone method requires weighting of the time domain results to
capture the initial incident and transmitted acoustic pulses while excluding any
subsequent reflections. The interaction between adjacent muffler chambers increased the
time decay of the transmitted pressure signal and complicates the separation of the initial
transmitted pulse from subsequent reflected pulses. Furthermore, during experimental
testing it was noted that pressures in the FFT spectrum for the downstream microphone at
frequencies greater than 500 Hz were less than 20 x 10-6 Pa, resulting in poor coherence
and low signal-to-noise ratios.
4.4.4 Comparison of muffler designs
Figure 4.11 compares the transmission loss obtained experimentally for each of the three
muffler designs considered. The frequency response of the single chamber muffler design
is dominated by a series of narrow transmission loss peaks. The first of these is attributed
to the proximity between the inlet and outlet ports behaving as a Helmholtz resonator,
while the higher frequency peaks are attributed to the various cross-modes. Variants of
this design may be useful where discrete frequencies are to be targeted but it has limited
application over a broad frequency range. The average transmission loss for the
integrated chamber design is similar to that of the single chamber design despite a 60%
increase in total chamber volume. However, the frequency response is more uniform and
the design outperforms the single chamber muffler across much of the frequency range.
The interconnected muffler design provides the greatest transmission loss of the three
designs. This is attributed to the combined effect of three discrete chambers, the isolation
provided by the interconnecting pipes, and a 250% increase in total chamber volume
compared to the single chamber design.
84
100 Interconnected chamber
Integrated chamber
80 Single chamber
60
40
20
0
0 1,000 2,000 3,000 4,000
Frequency (Hz)
Figure 4.11 Transmission loss of the single chamber, integrated chamber and
interconnected three-chamber CPAP device muffler designs –
experimental results
4.5 Resistive CPAP Device Muffler Designs
A polyurethane foam insert was incorporated into two of the CPAP device muffler
designs corresponding to the single chamber and integrated chamber mufflers.
Figure 4.12 shows a schematic diagram of the single chamber CPAP device muffler and
the foam insert, which occupies the majority of the chamber volume as shown by the
grey shaded area. It is important to note that this insert does not intrude into the direct
path between the inlet and outlet ports. A photograph of the single chamber muffler with
the foam insert is given in Fig. 4.13. Figure 4.14 shows a schematic diagram of the
integrated chamber muffler and the foam insert, which completely fills the volume of the
first chamber. Sound waves entering from the inlet port of the muffler must travel
through the foam prior to reaching the outlet port.
85
Chapter 4: Transmission Loss of CPAP
C Device Mufflers
Figure 4.12 Finite element

e model geometry of the single chhamber CPAP device
muffler showing the location of the foam insert (darkk grey)
Figure 4.13 Photoggraph of the single chamber CPAP device muuffler with foam insert
86
Figure 4.14 Finite element model geometry of the integrated chamber CPAP device
muffler showing the location of the foam insert (dark grey)
Acoustic finite element models of each of the muffler designs were developed using the
COMSOL finite element analysis package. The two polyurethane foam materials
corresponding to the light and dark grey foams which were characterised in Chapter 3
were used. The foam inserts were modelled using the Delany-Bazley functions with
derived coefficients that were obtained experimentally for each the two foam materials.
Figure 4.15 compares the transmission loss obtained computationally and experimentally
for the single chamber muffler with the light grey foam insert present. The transmission
loss predicted by the finite element model for the muffler without any foam insert is also
shown for comparison. The FEA results show good agreement with the experimental
results over the frequency range assessed. The inclusion of the foam insert results in a
dramatic reduction in transmission loss at the resonant peaks but also results in a
transmission loss of at least 10 dB over a broadband frequency range beyond the first
resonant peak centred at 800 Hz. Figure 4.16 compares the transmission loss obtained
computationally and experimentally for the single chamber muffler using the light and
dark grey foams. The results show that the foam inserts have a very similar impact on the
acoustic performance of this muffler design despite a difference in apparent density of
approximately 50%.
87
70
Empty muffler - FEA (COMSOL)
60 Light grey foam insert - FEA (COMSOL)
Light grey foam insert - Experimental
50
40
30
20
10
0
0 1,000 2,000 3,000 4,000
Frequency (Hz)
Figure 4.15 Transmission loss of the single chamber CPAP device muffler with and
without the light grey foam insert, comparing FEA results and
30
Light grey foam insert - FEA (COMSOL)
25 Light grey foam insert - Experimental

Dark grey foam insert - FEA (COMSOL)
20 Dark grey foam insert - Experimental
15
10
0
0 1,000 2,000 3,000 4,000
Frequency (Hz)
Figure 4.16 Transmission loss of the single chamber CPAP device muffler with light
and dark grey foam inserts, comparing FEA results and experimental
results
88
4.5.2 Integrated chamber muffler
Figure 4.17 compares the transmission loss predicted computationally for the integrated
chamber muffler both with and without the first chamber filled with foam. The FEA
results show that the presence of foam has little effect on the acoustic performance of the
muffler below 500 Hz but contributes significantly to increased transmission loss at
higher frequencies. In contrast to the observations made in respect to the single chamber
muffler, the results show that the two foam materials make differing contributions to the
acoustic performance of this muffler design. This observation is attributed to differences
between the muffler designs in regards to the location of the foam inserts. In the case of
the integrated chamber design, sound waves travelling between the inlet and outlet ports
are required to pass through approximately 20 cm of the foam while in the single
chamber design, they only graze the surface of the foam insert. The greater contribution
made by the light grey foam is consistent with the higher apparent density and flow
resistivity when compared to the dark grey foam.
80
Empty muffler - FEA (COMSOL)
70
Light grey foam insert - FEA (COMSOL)
60 Dark grey foam insert - FEA (COMSOL)

50
40
30
20
10
0
0 1,000 2,000 3,000 4,000
Frequency (Hz)
Figure 4.17 Transmission loss of the integrated CPAP device muffler, comparing
performance without foam and with the first chamber foam-filled - FEA
results
89
Chapter 5: Optimisation of Muffler Acoustic Performance
Chapter 5
Optimisation of Muffler Acoustic Performance
5.1 Introduction
In this chapter, computational optimisation techniques are applied to maximise muffler

acoustic performance. Optimisation of the geometry of a cylindrical expansion chamber
reactive muffler was initially examined. The inclusion of foam was then considered and
optimisation of the foam dimensions was performed. A similar process was then applied
to optimise the acoustic performance of a prototype CPAP device muffler comprising
three interconnected chambers, with and without the presence of foam. The reactive and
resistive components of the CPAP device muffler were independently optimised for
acoustic performance at various frequency bands. Results are also obtained for the
transmission loss of the CPAP device muffler for which the reactive and resistive
components were simultaneously optimised.
5.2 Optimisation Methodology
The solution of a standard nonlinear optimisation problem is of the general form [119]:
minimise ݂௢ ሺ࢞ሻ
subject to ݂௜ ሺ࢞ሻ ൑ Ͳ ݅ ൌ ͳǡ ǥ ǡ ݉ (5.1)
‫ݔ‬௝௠௜௡ ൑ ‫ݔ‬௝ ൑ ‫ݔ‬௝௠௔௫ ݆ ൌ ͳǡ ǥ ǡ ݊
where ݂௢ is the objective function, ݂௜ are behaviour constraints, ݉ is the number of

constraints and ࢞ is a vector of ݊ design control variables ‫ݔ‬௝ . Two alternative objective
functions were considered for the muffler optimisation: (i) minimising the acoustic
power at the muffler outlet, and (ii) maximising the transmission loss across the muffler.
Variation in the magnitude of the outlet power across the fitness domain is greater than
the variation in the magnitude of the transmission loss due to the logarithmic function in
90
the calculation of transmission loss. Whilst the larger gradients present in the outlet
power function would assist gradient-based solvers to locate minima, gradients are not a
significant consideration when using metaheuristic solvers. The magnitude of the outlet
power from small mufflers can result in very small objective function values, leading to
numerical errors in the calculations. In addition, the range of outlet power function
values calculated within a bounded solution domain can be many orders of magnitude
(~1030), producing further numerical challenges. Hence, transmission loss was adopted as
the most suitable performance measure to incorporate into an objective function. The
objective function for optimisation at multiple frequencies is of the form
௤ ௤
݂௢ ሺ࢞ሻ ൌ ෍ ߮௞ ܶ‫ܮ‬ሺ࢞ǡ ݂௞ ሻǡ ෍ ߮௞ ൌ ͳ (5.2)

௞ୀଵ ௞ୀଵ
where the vector ࢞ ൌ ሾଵ ǡ ǥ ǡ ௡ ሿ contains the control variables, CV, that are used to
define the geometry of the muffler components. The vectors ࢌ ൌ ൣ݂ଵ ǡ ǥ ǡ ݂௤ ൧ and
࣐ ൌ ൣ߮ଵ ǡ ǥ ǡ ߮௤ ൧ contain the frequencies in the range being evaluated and the fractional
weighting applied to the transmission loss at each frequency, respectively. ‫ ݍ‬is the
number of frequencies being evaluated and ݊ is the number of control variables. This
approach is consistent with that taken by Lee and Kim [95] and Airaksinen and
Heikkola [120]. The latter also applied a limiting value ܶ‫ܮ‬௠௔௫ to the transmission loss
calculated at each frequency in order to suppress the influence of narrow infinite peaks in
the transmission loss spectrum. Incorporating the limiting value into Eq. (5.2) gives
݂௢ ሺ࢞ሻ ൌ ෍ ߮௞ ሺܶ‫ܮ‬ሺ࢞ǡ ݂௞ ሻǡ ܶ‫ܮ‬௠௔௫ ሻ (5.3)

௞ୀଵ
In the analyses that follow, the acoustic performance of reactive mufflers is optimised by
assigning geometric dimensions such as expansion chamber length and/or diameter as the
design control variables. In the case of a resistive muffler where a foam insert is present,
dimensions such as the foam thickness and/or the length that the lining extends along an
expansion chamber are the assigned control variables. Upper and lower bounds are
applied to the control variables to restrict the design space. More complex behaviour
constraints may also be applied. For example, the following non-linear constraint can be
91
applied to limit the total volume of foam that can be used in a lining that is applied to the
wall of an expansion chamber:
ߨ‫ܮ‬௙ ‫ݐ‬௙ ൫݀௖ െ ‫ݐ‬௙ ൯ െ ܸ௠௔௫ ൑ Ͳ (5.4)
where ‫ܮ‬௙ and ‫ݐ‬௙ are respectively the length and thickness of the foam, ݀௖ is the diameter
of the expansion chamber and ܸ௠௔௫ is the maximum allowable foam volume. Alterations
to the muffler geometry may affect the level of self-generated noise produced by the
presence of mean flow. Further, the geometric changes may impact on the system flow
resistance and affect the aeroacoustic behaviour of the flow-generating fan. Neither of
these flow-related effects is considered in the optimisation analyses that follow.
Text in Section 5.2 has been redacted due to confidentiality requirements

associated with Australian Research Council Linkage Project LP0669543.
The redaction has been approved by the Dean of Graduate Research.
Public access to the redacted text will be available from 9th December 2013.
92

The redaction has been approved by the Dean of Graduate Research.
Public access to the redacted text will be available from 9th December 2013.
93

A range
of optimisation algorithms were investigated corresponding to the Nelder-Mead, genetic
algorithm, simulated annealing and particle swarm optimisation approaches. A
comparison of their performance for optimisation of mufflers, with and without the
presence of foam, is summarised in Appendix C. The genetic algorithm (GA) was
identified as the most robust approach that produced the global minimum of the
transmission loss for each case. Hence, only results obtained using the GA are presented
for the optimisation analyses in this chapter.
5.3 Reactive Expansion Chamber Muffler
A cylindrical expansion chamber reactive muffler was initially examined for the
optimisation study. The chamber diameter and/or the chamber length were used as the
design variables and were permitted to vary within predetermined bounds in order to
obtain an optimised design. The initial chamber length was selected on the basis that the
plane wave analytical solution provides a good prediction of the acoustic performance of
the muffler over the permitted range of the design variables and over the frequency range
of interest. The muffler geometry is shown in Fig. 5.2 and is assigned initial dimensions
of ݀௖ = 60 mm, ‫ܮ‬௖ = 250 mm, ݀௜௢ = 20 mm and ‫ܮ‬௜௢ = 20 mm. The diameter ݀௖ and length
‫ܮ‬௖ of the expansion chamber are permitted to vary during optimisation of the muffler.
The diameter ݀௜௢ and length ‫ܮ‬௜௢ of the inlet and outlet ducts remain fixed. The frequency
at which the first non-plane wave propagates in the expansion chamber is given by
௖
ͲǤͷͺ͸ͳሺௗ ሻ [67]. The cut-on frequency for the selected muffler geometry is thus
೎
3,350 Hz. The optimisation analyses for this muffler are restricted to frequencies below
2 kHz to ensure that plane wave theory provides a reliable prediction of acoustic
performance.
94
Maximum size
ȟ݀Ȁʹ ȟ‫ܮ‬
Minimum size
݀௜௢ ݀௖ ݀௜௢
‫ܮ‬௜௢ ‫ܮ‬௖ ‫ܮ‬௜௢
Figure 5.2 A cylindrical expansion chamber reactive muffler showing the

optimisation domain
For the case where the inlet and outlet pipes have the same cross-sectional area, the
transmission loss ܶ‫ ܮ‬can be calculated using
ଶ
ͳ ͳ
ܶ‫ ܮ‬ൌ ͳͲ ଵ଴ ቈଶ ሺ݇‫ܮ‬௖ ሻ ൅ ൬ ൅ ݉൰ ଶሺ݇‫ܮ‬௖ ሻ቉ (5.5)
Ͷ ݉
ଶగ௙ ௗ ଶ
where ݇ ൌ ቀ ௖
ቁ is the wave number and ݉ ൌ ቀௗ ೎ ቁ is the expansion ratio of the
೔೚
expansion chamber to the inlet duct. Maximum transmission loss occurs when
ሺʹ݊ െ ͳሻߨ
݇‫ܮ‬௖ ൌ ǡ ݊ ൌ ͳǡ ʹǡ ͵ǡ ǥ (5.6)
ʹ
The frequencies ݂௢௣௧ at which the maximum transmission loss occurs for a muffler
chamber of length ‫ܮ‬௖ or, alternatively, the optimised chamber length ‫ܮ‬௢௣௧ that produces
the maximum transmission loss for a given frequency ݂ can be obtained as
ሺʹ݊ െ ͳሻܿ
݂௢௣௧ ൌ ǡ ݊ ൌ ͳǡ ʹǡ ͵ǡ ǥ (5.7)
Ͷ‫ܮ‬௖
95
ሺʹ݊ െ ͳሻܿ
‫ܮ‬௢௣௧ ൌ ǡ ݊ ൌ ͳǡ ʹǡ ͵ǡ ǥ (5.8)
Ͷ݂
The maximum transmission loss can be obtained by substituting Eq. (5.6) into Eq. (5.5)
and is given by
݉ଶ ൅ ͳ
ܶ‫ܮ‬௠௔௫ ൌ ʹͲ ଵ଴ ቈ ቉ (5.9)
ʹ݉
The following optimisation is performed at an excitation frequency of 1 kHz. For the

muffler parameters given here, Eq. (5.8) predicts that the maximum transmission loss
will occur when the chamber length is an odd multiple of 85.8 mm. The first three
resonant lengths are thus 85.8 mm, 257.4 mm and 429.0 mm. The design optimisation
was bounded by constraining the chamber diameter ݀௖ to a range between 45 mm and
75 mm (±15 mm from the initial design) and the length ‫ܮ‬௖ to a range between 205 mm
and 295 mm (±45 mm from the initial design). The constraints applied to the range of
possible chamber lengths ensure that the feasible design space contains only one length-
related transmission minima. From Eq. (5.9) it can be observed that the maximum
transmission loss is independent of length and increases with increasing expansion ratio.
Hence the maximum transmission loss is expected to coincide with the maximum
chamber diameter of 75 mm. The corresponding maximum transmission loss predicted
analytically is 17.0 dB.
A fitness landscape was initially created by stepping the change in diameter ο݀ and
change in length ο‫ ܮ‬in integer increments across their respective bounded ranges and
using COMSOL to calculate the transmission loss at an excitation frequency of 1 kHz.
From Fig. 5.3 it can be seen that the topology is smooth and that only one minimum is
present in the bounded domain. Optimisation results for the muffler using the genetic
algorithm are obtained as ሺο݀ǡ ο‫ܮ‬ǡ ܶ‫ܮ‬ሻ ൌ (15.0 mm, 8.2 mm, 17.1 dB), which are in
close agreement with the predicted theoretical values of (15.0 mm, 7.4 mm, 17.0 dB).
The corresponding chamber lengths for the FEA and theoretical results are 258.2 mm and
257.4 mm, respectively, and the difference in these two lengths is only 0.3%. The
objective function values assessed by the GA solver during convergence are presented in
Fig. 5.4.
96
Figure 5.3 Fitness landscape for the cylindrical expansion chamber reactive muffler
at 1 kHz excitation frequency
Figure 5.4 Fitness landscape with objective function values (-TL) assessed by the
genetic algorithm solver during convergence – cylindrical expansion
chamber reactive muffler (white dot = convergence)
97
5.4 Expansion Chamber Muffler With Foam Lining
A similar cylindrical expansion chamber muffler now lined with foam on the chamber
wall was optimised for acoustic performance. The original chamber dimensions of
݀௖ = 60 mm and ‫ܮ‬௖ = 250 mm were used. The foam has the same acoustic properties as
the light grey foam that was characterised in Chapter 3. No foam was applied to the inlet
and outlet ducts. The region occupied by the lining is specified by the foam thickness and
the length that the foam lining extends along the expansion chamber, commencing at the
chamber inlet. The muffler geometry is shown in Fig. 5.5. The design optimisation was
bounded by constraining the foam thickness ‫ݐ‬௙ to a range between 5 mm and 25 mm and
the length ‫ܮ‬௙ to a range between 5 mm and 245 mm. The bounds were selected so that
the foam insert could never collapse to zero volume, or fully occupy the chamber volume
thus causing the unobstructed region through the centre of the muffler to collapse to zero
volume. Both geometric extremes risk the generation of poor quality finite element mesh
and unreliable predictions.
‫ܮ‬௙
Minimum size ‫ݐ‬௙

Foam
݀௜௢ ݀௜௢
Maximum size
݀௖
‫ܮ‬௜௢ ‫ܮ‬௖ ‫ܮ‬௜௢
Figure 5.5 A cylindrical expansion chamber muffler with foam lining showing the
optimisation domain
The same excitation frequency of 1 kHz as in the case for the reactive muffler was used.
A fitness landscape was created by stepping the foam thickness and length in increments
of 5 mm and 25 mm respectively across their bounded ranges. The fitness values were
expanded to consider the case where the foam insert was absent (‫ݐ‬௙ = 0), and for
98
increments of length where the foam insert occupied the full muffler cross-section,
corresponding to ‫ݐ‬௙ = 30 mm. The additional values were only used to assist with
visualisation of the fitness landscape as they do not fall within the feasible design space.
From Fig. 5.6 it can be seen that the topology is smooth and that the global minimum is
clearly defined. The optimisation results using the genetic algorithm are obtained as
൫‫ݐ‬௙ ǡ ‫ܮ‬௙ ǡ ܶ‫ܮ‬൯ ൌ (25.0 mm, 237.9 mm, 25.0 dB). The corresponding field of objective
function values assessed by the GA solver is shown in Fig. 5.7, and clearly shows that
the global minimum in the fitness domain is attained. As the optimisation has converged
near the upper bounds of both the thickness and length of the lining, this result
demonstrates that foam can be used to enhance the acoustic performance of the muffler at
1 kHz. However, two maxima are present in the bounded domain which reveals that the
inclusion of a foam lining can also degrade the acoustic performance of the muffler at the
same frequency.
5.5 CPAP Device Muffler
A range of optimisation analyses were applied to the three-chamber prototype CPAP

device muffler presented in Chapter 4. The interconnecting ducts of the original design
have been reduced from 43 mm to 2 mm in length. The resulting compact muffler,
hereafter referred to as the pre-optimisation design, is shown in Fig. 5.8. Eight
dimensions are assigned as geometric control variables (CV) as shown in Fig. 5.8 and are
utilised in the optimisation process. CV1 defines the length of the interconnecting ducts
between successive chambers. CV2 and CV3 define the height and diameter of
chamber 1, respectively. CV4 defines the height of chamber 2 while CV5 and CV6
together define the chamber depth. CV7 and CV8 define the height and width of
chamber 3, respectively. A further two control variables, CV9 and CV10, are used to
define the thickness of foam inserts which are located at the bottom of chambers 2 and 3,
respectively.
The horizontal dimension of chamber 2 that is normal to CV5 is constrained to be equal

to the diameter of chamber 1. The rear face of chamber 3 is co-planar with the axis of the
inlet duct. The interconnecting duct between chambers 1 and 2 is located in the centre of
the rectangular face at the outlet of chamber 1. The interconnecting duct between
99
Figure 5.6 Fitness landscape for the cylindrical expansion chamber muffler with
foam lining at 1 kHz excitation frequency
Figure 5.7 Fitness landscape with objective function values (-TL) assessed by the
chamber muffler with foam lining (white dot = convergence)
100
chambers 2 and 3 shares the same vertical offset from the top face as the previous duct
and is offset horizontally from the front face by the distance CV6 (see Fig. 5.8). The
outlet duct shares the same vertical offset from the top face as the previous ducts and is
located on the vertical centreline of the front face. Whilst the lengths of the
interconnecting ducts can vary, the duct diameters remain fixed.
Complex inter-relationships exist between the control variables and the underlying
muffler geometry. Changing a variable attached to one component of the system has the
potential to influence the acoustic behaviour of other components by affecting expansion
ratios, chamber volumes, and/or duct locations. Such complexity in the optimisation
problem is representative of the practical reality brought about by the compact nature of
these mufflers.
Chamber 1
Chamber 3
CV3
CV8
CV5
CV2
CV6
CV7 CV1 = interconnecting

CV4
duct length
Chamber 2
Figure 5.8 Prototype CPAP device muffler showing optimisation control variables
101
Table 5.1 presents the dimensions of the pre-optimisation CPAP device muffler design.
The dimensions are the same as those initially presented in Chapter 4 with the exception
of the reduced interconnecting duct lengths. Table 5.1 also presents the lower and upper
bounds placed on the ten control variables during the optimisation analyses. The upper
bounds for the two control variables which are used to define the foam thickness at the
bottom of chamber 2 and chamber 3 are not fixed values. Instead, these bounds are
described in terms of the difference between the height of chamber 1 (CV2) and the
height of the chamber containing the foam insert (CV4 and CV7 respectively). As any, or
all, of these chamber heights may be permitted to vary during the optimisation process,
the upper bounds on the foam thickness will also change. The foam constraints are
illustrated in Fig. 5.9.
Table 5.1 Pre-optimisation dimensions and bounds of the optimisation control

variables for the prototype CPAP device muffler
Control Pre-optimisation Lower bound Upper bound

variable dimension (mm) (mm) (mm)
CV1 2 2 10
CV2 40 30 60
CV3 84 60 100
CV4 80 45 100
CV5 25 10 45
CV6 44 30 55
CV7 80 45 100
CV8 30 30 50
CV9 NIL 5 CV4 – CV2
CV10 NIL 5 CV7 – CV2
102
Chamber 3 Chamber 2
CV7 CV4 CV2
CV9 maximum
Chamber 1
CV9
CV10 maximum CV9 minimum (5 mm)

CV10
CV10 minimum (5 mm)
Figure 5.9 Prototype CPAP device muffler showing foam constraints
Using the original dimensions of the muffler listed in Table 5.1, the transmission loss of
the muffler with and without the presence of foam was calculated and the results are
presented in Fig. 5.10. In the pre-optimised design, the foam inserts occupy the lower
half of chambers 2 and 3. The transmission loss of the muffler in the absence of foam
shows poor performance below 800 Hz and an uneven performance at higher
frequencies. The inclusion of foam reduces the transmission loss peaks around 1 kHz and
2 kHz, but in general improves performance at higher frequencies.
The pre-optimised muffler dimensions correspond to a limiting plane wave cut-on

frequency of 1.3 kHz. Figures 5.11 and 5.12 show isosurfaces of acoustic pressure
produced by a 1 kHz excitation for the pre-optimised design in the absence and presence
of the foam inserts, respectively. Although the excitation frequency is below the plane
wave cut-on frequency, the surfaces reveal highly non-planar wave behaviour. This is
attributed to the large expansion ratios and short chamber lengths, as well as the
orthogonal alignment of the inlet and outlet ducts. The figures also highlight the
transition between the air and foam regions.
103
5.5.1 Optimisation strategy
Three approaches to optimise the acoustic performance of a muffler were considered.

The first approach is to take an existing or proposed design and optimise it such that the
muffler performance is either more uniformly broadband or biased in favour of a
particular frequency band. This approach can be undertaken without detailed knowledge
of the noise source. The second approach is to consider the sound power spectrum of a
known noise source and to identify discrete frequencies or narrow frequency bands
where attenuation is more desirable. The muffler design can then be optimised to best
attenuate these target frequencies. The third approach is to combine the transmission loss
of the muffler design with the sound power spectrum of the noise source to produce an
indicative attenuated power spectrum. Discrete frequencies or narrow frequency bands
present in the attenuated spectrum can be identified for attenuation and the muffler
design then optimised to attain a goal that is closely aligned with how the noise impacts
on the user.
140
Pre-optimisation geometry
120
Pre-optimisation geometry plus foam inserts (50% of V2 and V3)
100
80
60
40
20
0
0 1,000 2,000 3,000 4,000 5,000
Frequency (Hz)
Figure 5.10 Transmission loss of the prototype CPAP device muffler using the pre-
optimised (original) geometry, comparing the reactive muffler (no foam)
and the muffler with foam inserts occupying the lower half of chambers 2
and 3 – FEA results
104
excitation frequency – pre-optimised reactive muffler (no foam) – FEA
results
excitation frequency – pre-optimised muffler with foam in lower half of
chambers 2 and 3 – FEA results
105
The second approach, corresponding to the consideration of the fan noise source
spectrum to identify the frequencies where attenuation is desirable, has been adopted in
the work that follows.
Maximising the averaged transmission loss of a muffler across a broad frequency range
would ideally involve performing calculations over a very finely discretised frequency
spectrum. The computational expense of this approach is prohibitive. In order to obtain
more practical computational run times, the strategy adopted in this thesis was to:
• consider the sound power spectrum of an appropriate noise source;

• identify the frequency range over which the averaged transmission loss is to be
obtained;
• identify specific frequency bands within that range where targeted optimisation
would make the greatest contribution; and
• optimise the design using the targeted frequency bands with the aim of achieving
the maximum averaged transmission loss across the full frequency range of
interest.
Figure 5.13 presents a typical noise source spectrum generated by the fan of a CPAP
device for a frequency range up to 5 kHz. Beyond 5 kHz, the sound power level of the
noise source decreases with increasing frequency and thus does not significantly
contribute to the radiated noise. The noise spectrum clearly shows two frequency bands
within this range to be targeted by the muffler. The first band extends between 500 and
1200 Hz and is centred on the low frequency peak. The second band is broader and spans
the higher frequency region between 2 and 5 kHz. Attenuation in these two bands will be
dominated by the reactive geometry and acoustically absorptive foam, respectively.
The various optimisation analyses for the prototype CPAP device muffler are described
in what follows. The reactive geometry of the muffler was initially optimised to attenuate
the peak noise around 1 kHz in the lower frequency band. This geometry was then fixed
and the resistive component of the muffler, corresponding to the two foam inserts, was
optimised to attenuate the broadband noise at higher frequencies (2–5 kHz). A third
optimisation was performed which simultaneously optimised the muffler geometry and
the thickness of the foam inserts and included frequencies from both the lower and higher
frequency bands. The transmission loss resulting from optimisation of the reactive and
106
resistive muffler components separately is compared with the transmission loss obtained
from the simultaneous optimisation run.
REACTIVE
Sound Power Level (dBA)
RESISTIVE (FOAM)
0 1,000 2,000 3,000 4,000 5,000

Frequency (Hz)
Figure 5.13 Sample CPAP fan noise source spectrum
5.5.2 Muffler geometry optimisation
Optimisation of the geometry of the reactive component of the prototype CPAP device
muffler was conducted considering eight equally weighted frequencies in the lower
frequency band, from 500 to 1200 Hz in 100 Hz increments. The control variables CV1
to CV8 and the bounds shown in Table 5.1 were used. The pre-optimised (original) and
optimised dimensions that were obtained using the genetic algorithm are presented in
Table 5.2. Comparison of the optimised values with the lower and upper bounds in
Table 5.1 shows that all optimised dimensions have increased yet only three of the eight
dimensions are within 5 mm of their upper bound. Only the length of the interconnecting
ducts, CV1, has optimised to its upper bound value. Optimisation of the reactive muffler
geometry results in an increase in the volume of each chamber and an increase in the
expansion ratios.
107
Table 5.2 Original and optimised control variables for the prototype CPAP device
muffler – muffler geometry optimisation (reactive design only)
Original Optimised
Control variable
dimensions (mm) dimensions (mm)
CV1 2 10
CV2 40 52
CV3 84 90
CV4 80 94
CV5 25 42
CV6 44 49
CV7 80 89
CV8 30 47
Averaged ܶ‫ܮ‬
33.6 (dB) 62.5 (dB)
(0.5 – 1.2 kHz)
Averaged ܶ‫ܮ‬
36.3 (dB) 54.9 (dB)
(entire 5 kHz range)
The transmission loss of the muffler in the absence of foam, using the original
dimensions and using the optimised geometry, is shown in Fig. 5.14. The low frequency
band from 500 to 1200 Hz over which the optimisation was carried out is also
highlighted. Values for the averaged transmission loss for the pre- and post optimised
reactive muffler geometry in the low frequency band and for the entire 5 kHz frequency
range are given in Table 5.2. Optimisation of the muffler geometry in the low frequency
band has resulted in an increase in transmission loss of nearly 30 dB in that frequency
band, as well as a significant increase in transmission loss at higher frequencies.
Using the optimised geometry, foam was then inserted to occupy the lower half of
chambers 2 and 3, as in the case for the results presented in Fig. 5.10. Comparison of
Fig. 5.15 with Fig. 5.10 confirms that the presence of foam reduces the transmission loss
peaks below 2 kHz, but results in a more uniform transmission loss across the higher
108
frequency range. The averaged transmission loss across the entire 5 kHz frequency range
for the pre- and post optimised muffler geometries with foam inserts included are
41.8 dB and 56.3 dB, respectively. Comparison of these two values and the results
presented in Figs. 5.10 and 5.15 shows that optimisation of the muffler geometry alone
has already produced a significant increase in transmission loss across the entire
frequency range.
5.5.3 Foam thickness optimisation
Using the optimised muffler geometry dimensions listed in Table 5.2, the thickness of the
foam inserts located in chambers 2 and 3 was then optimised. Three frequency regions
were considered – the low frequency band only (from 500 to 1200 Hz), the high
frequency band only (from 2 to 5 kHz), and the low frequency and high frequency bands
combined. During the foam optimisation, only the foam thickness in chambers 2 and 3
corresponding to control variables CV9 and CV10 was allowed to vary, that is, control
variables CV1 to CV8 were fixed. A minimum foam thickness of 5 mm, as indicated by
the lower bound in Table 5.1, was enforced in order to prevent the portion of either
chamber occupied by foam from collapsing to zero volume and risking the generation of
poor quality finite element mesh. Upper bounds were applied which allowed the foam
inserts to extend up to the plane defined by the bottom surface of chamber 1. This
ensured that chambers 2 and 3 always retained foam-free regions having a minimum
height corresponding to the height of chamber 1.
Commencing with the low frequency band, optimisation was carried out across the eight
equally spaced frequencies considered previously in the reactive muffler geometry
optimisation, that is, from 500 to 1200 Hz in 100 Hz increments. The second
optimisation run was performed across the high frequency band only and considered
seven equally weighted frequencies from 2 to 5 kHz in 500 Hz increments. The third
optimisation run was then conducted combining the narrow low frequency band with the
broad high frequency band. Hence, a total of fifteen equally weighted frequencies were
considered. The optimised foam thickness for chambers 2 and 3 obtained using the
genetic algorithm are summarised in Table 5.3 and the resulting transmission loss curves
are presented in Fig. 5.16.
109
140
Geometry prior to optimisation
120 Optimised geometry (reactive only)
100
80
60
40
20
0
0 1,000 2,000 3,000 4,000 5,000
Frequency (Hz)
Figure 5.14 Transmission loss of the reactive prototype CPAP device muffler (no
foam), comparing original dimensions and optimised geometry – FEA
results
140
Optimised geometry (reactive only)
Optimised geometry plus foam in lower half of Vol. 2 and Vol. 3
120
100
80
60
40
20
0
0 1,000 2,000 3,000 4,000 5,000
Frequency (Hz)
Figure 5.15 Transmission loss of the geometry-optimised prototype CPAP device

muffler, comparing the reactive muffler only and the muffler with foam
inserts in the lower half of chambers 2 and 3 – FEA results
110
Table 5.3 shows that optimisation performed in the low frequency band produces the
greatest averaged transmission loss of the three optimisation runs for that frequency
band. Similarly, optimisation in the high frequency band produces the greatest averaged
transmission loss for that frequency band. Optimisation across both the low and high
frequency bands produces similar averaged transmission loss values in the individual and
combined frequency bands.
Table 5.3 Optimised control variables for the prototype CPAP device muffler – foam
thickness optimisation only (using optimised muffler geometry)
Frequency bands for foam optimisation

Control variable
(mm)
Low frequency High frequency Both frequency
band band bands
CV9 34 42 41
CV10 5 30 16
Averaged ܶ‫ܮ‬
59.9 (dB) 55.9 (dB) 57.7 (dB)
(0.5–1.2 kHz)
Averaged ܶ‫ܮ‬
59.4 (dB) 63.3 (dB) 62.5 (dB)
(2–5 kHz)
Averaged ܶ‫ܮ‬
59.7 (dB) 59.4 (dB) 59.9 (dB)
(0.5–1.2 and 2–5 kHz)
Averaged ܶ‫ܮ‬
55.7 (dB) 56.6 (dB) 56.2 (dB)
Figure 5.16 shows that the transmission loss peak below 1 kHz is greatest when the
thickness of the foam inserts is optimised in the low frequency band, and reduces as the
optimisation shifts to the high frequency band. This is due to the fact that optimising the
foam inserts in the low frequency band results in the smallest values for the foam
thickness, thus increasing the effective resonant reactive volume. The decrease in
performance of the muffler at frequencies up to 2 kHz when optimisation is carried out in
the high frequency band is attributed to the increased thickness of foam in chambers 2
111
and 3, thus reducing the effective resonant reactive volume. The best performance of the
muffler across the entire frequency range occurs when the foam thickness optimisation is
performed in the high frequency band (from 2 to 5 kHz). This is due to the presence of
the foam inserts making the greatest positive contribution to the muffler transmission loss
at higher frequencies while reducing performance at the lower frequencies to a lesser
extent. Optimisation in the high frequency band produces the thickest foam inserts of the
three optimisations performed, as shown in Table 5.3.
140
0.5 - 1.2 kHz
120 0.5 - 1.2 kHz & 2 - 5 kHz

2 - 5 kHz
100
80
60
40
20
0
0 1,000 2,000 3,000 4,000 5,000
Frequency (Hz)
Figure 5.16 Transmission loss of the prototype CPAP device muffler, comparing foam
thickness optimisation in the low frequency band, high frequency band
and combined low and high frequency bands – FEA results
Two further foam thickness optimisation runs were performed corresponding to a narrow
high frequency band (3.5 kHz to 4.5 kHz in 100 Hz increments) and the full frequency
range (500 Hz to 5 kHz in 500 Hz increments). The total volume of foam added to the
muffler as a function of the average frequency in the optimisation frequency range is
shown in Fig. 5.17. This figure indicates that the amount of foam added to the muffler
increases linearly with the average frequency. At average frequencies greater than 2 kHz,
the increase in foam volume takes place almost entirely in chamber 3.
112
550
Foam added to Chambers 2 and 3 (mm3/1000) 2 - 5 kHz
500
450 0.5 - 1.2 kHz

plus 2 - 5 kHz
400
3.5 - 4.5 kHz
350
0.5 - 5 kHz
300
250
200
0.5 - 1.2 kHz
150
0 1,000 2,000 3,000 4,000 5,000
Average Frequency (Hz)
Figure 5.17 Relationship between total volume of foam added to the muffler and the
average optimisation frequency
5.5.4 Simultaneous geometry and foam optimisation
Commencing with the pre-optimised (original) dimensions for the control variables listed
in Table 5.1, the reactive muffler geometry and thickness of foam inserts located in
chambers 2 and 3 of the prototype CPAP device muffler were optimised simultaneously.
All control variables (CV1 to CV10) were allowed to vary within their lower and upper
bounds, also listed in Table 5.1. Optimisation was conducted considering the same
fifteen discrete frequencies in the low and high frequency bands that were used earlier for
the sequential reactive geometry and foam thickness optimisations. The lower frequency
band extended from 500 to 1200 Hz in 100 Hz increments while the higher frequency
band extended from 2 to 5 kHz in 500 Hz increments. Three optimisations were
performed using these frequency bands. In the first optimisation, all fifteen frequencies
received equal weighting. In the second optimisation, the eight frequencies within the
lower frequency band were assigned twice the weighting of the seven frequencies in the
upper band. In the third optimisation, the frequencies in the upper frequency band were
assigned twice the weighting of the frequencies in the lower band.
113
The dimensions of the control variables obtained using the simultaneous geometry and
foam optimisation in the low and high frequency bands are presented in Table 5.4,
alongside those obtained by sequentially optimising the reactive geometry in the low
frequency band and then optimising the foam thickness in the high frequency band.
Comparison of the results obtained from the simultaneous optimisation using equally
weighted frequencies with those from the sequential optimisation shows both significant
variation in the final designs and difference in the averaged transmission losses. The
averaged transmission loss across the entire frequency range of 62.2 dB and 56.6 dB for
the simultaneously and sequentially optimised mufflers, respectively, shows that
simultaneous optimisation of the reactive and resistive components of the muffler results
in the best overall acoustic performance. Figure 5.18 compares the transmission loss for
the simultaneous and sequential optimisations at equally weighted frequencies in the low
and high frequency bands. The simultaneous geometry and foam optimised design results
in greater transmission loss in the low and high frequency ranges.
The averaged transmission loss for the simultaneous optimisations conducted using two
times frequency weighting in the low or high frequency bands are also presented in
Table 5.4. Figure 5.19 presents a comparison of the transmission loss for the three
simultaneous optimisations. The low frequency biased result demonstrates a slightly
better transmission loss below 2 kHz, but is accompanied by a reduction in performance
in the high frequency band. Results show that the application of a weighting bias to the
frequency bands does not produce any significant benefit, and in fact reduces the
averaged transmission loss across the entire frequency range in both cases.
114
Table 5.4 Optimised control variables for the prototype CPAP device muffler,
comparing sequential and simultaneous geometry and foam optimisation
Simultaneous geometry and foam optimisation

Sequential
Control variable geometry 2x weighted 2x weighted
Equally
(mm) then foam in low in high
weighted
optimisation frequency frequency
frequencies
band band
CV1 10 10 10 10
CV2 52 56 58 43
CV3 90 100 95 100
CV4 94 91 97 96
CV5 42 27 43 28
CV6 49 39 55 52
CV7 89 80 96 76
CV8 47 49 46 50
CV9 42 35 28 52
CV10 30 5 9 6
Averaged ܶ‫ܮ‬
59.4 (dB) 68.4 (dB) 69.0 (dB) 64.8 (dB)
(0.5–1.2 and 2–5 kHz)
Averaged ܶ‫ܮ‬
56.6 (dB) 62.2 (dB) 59.9 (dB) 56.4 (dB)
115
140
Sequential optimisation (geometry then foam)
120 Simultaneous geometry and foam optimisation
100
80
60
40
20
0
0 1,000 2,000 3,000 4,000 5,000
Frequency (Hz)
Figure 5.18 Transmission loss of the prototype CPAP device muffler, comparing
sequential and simultaneous optimisation of the geometry and foam
thickness – FEA results
140
Equal weight at each frequency 2x low frequency band bias
120 2x high frequency band bias
100
80
60
40
20
0
0 1,000 2,000 3,000 4,000 5,000
Frequency (Hz)
simultaneous optimisation of the geometry and foam thickness using
different weighting in each of the frequency bands – FEA results
116
A final optimisation run was conducted which considered 25 equally weighted

frequencies across the entire frequency range, extending from 200 Hz to 5 kHz in 200 Hz
increments. Table 5.5 compares the dimensions of the control variables obtained for the
simultaneous optimisation using the 15 equally weighted frequencies in the low and high
frequency bands, with those obtained for simultaneous optimisation using the 25 equally
weighted frequencies across the entire 5 kHz range. The corresponding transmission loss
of the two muffler designs is shown in Fig. 5.20. Comparison of the results shows that
both approaches have produced similar muffler chamber dimensions but have converged
on dramatically different values for the thickness of the foam inserts. A slightly better
performance in averaged transmission loss is achieved for optimisation across the entire
frequency range.
140
Both frequency bands
120
Entire frequency range
100
80
60
40
20
0
0 1,000 2,000 3,000 4,000 5,000
Frequency (Hz)
simultaneous optimisation of the geometry and foam thickness in two
frequency bands (low and high) and across the entire 5 kHz frequency
range – FEA results
117
Table 5.5 Optimised control variables for the prototype CPAP device muffler,
comparing simultaneous geometry and foam optimisation in two
frequency bands (low and high) and across the entire 5 kHz frequency
range
15 equally weighted 25 equally weighted

frequencies across frequencies across
Control variable (mm)
the low and high the entire 5 kHz
frequency bands range
CV1 10 10
CV2 56 60
CV3 100 100
CV4 91 86
CV5 27 25
CV6 39 41
CV7 80 91
CV8 49 48
CV9 35 21
CV10 5 21
Averaged ܶ‫ܮ‬
68.4 (dB) 65.1 (dB)
(0.5 – 1.2 and 2–5 kHz)
Averaged ܶ‫ܮ‬
62.2 (dB) 64.0 (dB)
Table 5.6 presents a summary of the transmission loss results averaged over the entire
5 kHz frequency range for the various sequential and simultaneous optimisations of the
prototype CPAP device muffler. Also presented in Table 5.6 is the averaged transmission
loss for the original (pre-optimised) geometry and for the case when only the reactive
muffler geometry is optimised and then foam is inserted into the lower half of
chambers 2 and 3. Optimisation of the muffler design has resulted in at least a 14 dB
118
improvement in performance, as indicated by the averaged transmission loss across the

entire frequency range, compared with that of the pre-optimised design. Simultaneous
optimisation of the muffler reactive geometry and thickness of foam inserts produced a
greater improvement in muffler acoustic performance than that obtained by sequentially
optimising the reactive and resistive muffler components. In particular, simultaneous
optimisation over both the low and high frequency bands yields a design having an
averaged transmission loss for the entire frequency range which is 5.6 dB greater than
that obtained by optimising the geometry in the low frequency band followed by
optimising the foam in the high frequency band. The final optimisation which was
performed using equally weighted frequencies throughout the entire 5 kHz frequency
range resulted in the greatest averaged transmission loss of 64.0 dB. However, the
computational time required for the full range optimisation was three times that required
for previous optimisation runs.
Table 5.6 Summary of averaged transmission loss across the entire 5 kHz frequency
range for the various prototype CPAP device muffler design optimisations
Optimisation Geometry Foam Averaged ܶ‫ܮ‬
Fixed – Lower half

NIL Pre-optimisation 41.8 (dB)
chambers 2 and 3
Fixed – Lower half

Geometry only Low band a 56.3 (dB)
chambers 2 and 3
Sequential
(geometry then Low band a High band b 56.6 (dB)
foam)
Simultaneous Both low band and high band 62.2 (dB)

geometry and
foam Full frequency range 64.0 (dB)
a
Low frequency band: 500 – 1200 Hz
b
High frequency band: 2 – 5 kHz
119
Chapter 6: Effect of Mean Flow on Transmission Loss
Chapter 6
Effect of Mean Flow on Transmission Loss
6.1 Introduction
In this chapter, the transmission loss of four mufflers is measured experimentally in the
presence of a range of duct air flow rates. The muffler designs include the two simple
expansion chamber mufflers that were presented in Chapter 2 and two of the CPAP
device mufflers presented in Chapter 4. The performance of two of the muffler designs
with foam inserts fitted is also measured. The air delivery requirements of a CPAP
device blower are examined to quantify the appropriate mean flow velocities. The
transmission loss of each muffler in the presence of mean flow is compared with
experimental results obtained in the absence of flow. The influence of blower and
aeroacoustic noise on the experimental results is discussed.
6.2 CPAP Device Flow Requirements
A CPAP device delivers an elevated pressure to the airway via a mask which forms a
positive seal against the user’s face. Vent holes are located in the mask to allow expired
breath to exit from the mask and to limit the volume of expired air which is forced up the
hose towards the blower. Any expired air which remains in the hose and mask at the end
of the exhalation phase of the breathing cycle will be re-breathed during the subsequent
inhalation phase. In order to manage the re-breathing of gas and the possibility for carbon
dioxide accumulation within the breathing circuit, the blower is required to deliver air
flow in excess of the demands of breathing. Figure 6.1 shows the blower flow demand
during a single breath for a healthy adult at rest and during a period of elevated
respiration. The average blower air delivery requirement is 35 L/min and 50 L/min for
the typical and elevated demand cases, respectively. The maximum air delivery
requirement is 110 L/min and occurs during the inhalation phase of the elevated demand
breath cycle. All of the air supplied by the blower passes through the muffler/s in the
120
CPAP device air path. Assuming a typical air path internal diameter of 18 mm, the
maximum air delivery corresponds to a duct velocity of 7 m/s, or a Mach number of 0.02.
120
Typical adult breath cycle
100
Elevated demand breath cycle
80
Blower Flow (L/min)
60
40
20
0
0 1 2 3 4
-20
Time (sec)
Figure 6.1 CPAP device blower air delivery requirements
The convective and diffractive effects of mean flow alter the attenuation produced by a
muffler [10, 20]. Mean flow also changes the acoustic properties of porous materials
through scattering caused by the formation of vortices. When the flow direction
coincides with the direction of sound propagation, the presence of mean flow usually
decreases attenuation at lower frequencies and increases it at higher frequencies.
Although neglecting the effects of mean flow leads to an incorrect estimation of the
attenuation, the effect of flow on the performance of most muffler design elements has
been shown to be only slight for Mach numbers normally encountered in exhaust
mufflers [21]. Flow-acoustic interactions can generally be considered negligible below
‫ ܯ‬൏ ͲǤ͵ although limits have also been suggested as low as ‫ ܯ‬൏ ͲǤͲʹ [21, 23-25, 123].
6.3 Experimental Approach
Transmission loss measurements were conducted using the experimental rig described in
Chapter 2. The rig was augmented by the addition of a small high-speed blower and a
121
silencer box, both located at the inlet end of the rig as shown in Fig. 6.2. A photograph of
the blower and silencer box is shown in Fig. 6.3.
600 ~ 23,000
600
‫ܯ‬ଵ
Blower To
Acoustic muffler
lining
Horn drivers
Silencer box
Figure 6.2 Schematic diagram of the silencer box located at the inlet of the two-
microphone acoustic pulse experimental rig (dimensions in mm)
Figure 6.3 Photograph of the blower, silencer box and bell mouth
122
The centrifugal blower selected is typical of those used in portable ventilators and was
capable of delivering an air flow of 120 L/min against a total system resistance of
4.5 kPa. The air from the blower was directed into the silencer box where it entered into
a bell mouth before passing through to the inlet of the original rig. The silencer box was
lined with 50 mm thick mineral wool insulation to reduce any pressure fluctuations and
airborne noise generated by the blower. The bell mouth had a 130 mm inlet diameter
with a 41 mm bell internal radius. The purpose of the bell mouth is to smooth the entry of
air into the 18 mm diameter PVC duct and reduce any aeroacoustic noise generated by
the transition. The transmission loss of the silencer box was obtained experimentally and
is presented in Fig. 6.4. The results show that the silencer box is capable of attenuating
the fan-generated noise by at least 50 dB across the frequency range of interest.
100
80
60
40
20
0
0 1,000 2,000 3,000 4,000 5,000
Frequency (Hz)
Figure 6.4 Transmission loss of the silencer box – experimental results
When measurements are conducted in the presence of mean flow, the microphones are
subjected to pressure fluctuations associated with the flow. These flow-induced pressure
fluctuations are random and are not correlated with the acoustic transients generated by
the horn drivers. Both the upstream ‫ݔ‬ሺ‫ݐ‬ሻ and downstream ‫ݕ‬ሺ‫ݐ‬ሻ acoustic pressure time
histories become contaminated with noise. The presence of this noise has more
significant implications for the measurement of the downstream transient acoustic
123
pressure time history as the pulse has been attenuated by the muffler. The effects of the
aeroacoustic noise can be partially ameliorated by averaging the time histories of 100
generated pulses. The effects of noise in the downstream signal can be further addressed
using the ‫ܪ‬ଵ estimation of the frequency response function, which is given as the ratio
between the cross-spectrum of the two signals ‫ݔ‬ሺ‫ݐ‬ሻ and ‫ݕ‬ሺ‫ݐ‬ሻ to the autospectrum of the
upstream signal ‫ݔ‬ሺ‫ݐ‬ሻ [105].
ܻሺ݂ሻ ܺ ‫ כ‬ሺ݂ሻ ‫ܩ‬௫௬ ሺ݂ሻ

‫ܪ‬ଵ ሺ݂ሻ ൌ ή ൌ (6.1)
ܺሺ݂ሻ ܺ ‫ כ‬ሺ݂ሻ ‫ܩ‬௫௫ ሺ݂ሻ
ܺሺ݂ሻ and ܻሺ݂ሻ are the Fourier transforms of the upstream ‫ݔ‬ሺ‫ݐ‬ሻ and downstream ‫ݕ‬ሺ‫ݐ‬ሻ
pressure signals, respectively. ܺ ‫ כ‬ሺ݂ሻ is the complex conjugate of ܺሺ݂ሻ. ‫ܩ‬௫௬ ሺ݂ሻ
represents the cross-spectrum of two signals ‫ݔ‬ሺ‫ݐ‬ሻ and ‫ݕ‬ሺ‫ݐ‬ሻ. ‫ܩ‬௫௫ is the autospectrum of
signal ‫ݔ‬ሺ‫ݐ‬ሻ. The ‫ܪ‬ଵ algorithm is able to provide a good estimate of the frequency
response function even with coherence values as low as 0.1.
6.4 Transmission Loss with Mean Flow
The transmission loss of four muffler designs was experimentally measured using the
two-microphone acoustic pulse method described in Chapter 2. The designs consisted of:
• the single expansion chamber muffler of 86 mm diameter and 156 mm length;

• the single expansion chamber muffler of 114 mm diameter and 45 mm length;
• the single chamber CPAP device muffler (see Fig. 4.1); and
• the interconnected three chamber CPAP device muffler (see Fig. 4.3).
Measurements were conducted at volumetric air flow rates of 30, 60, 90 and 120 L/min.
An indicative flow velocity through the chambers and interconnecting ducts of each
muffler, at a volumetric air flow rate of 120 L/min, is summarised in Table 6.1. For the
purpose of calculating the velocities in this table it has been assumed that the air flow is
distributed uniformly across the chamber/duct cross-section. As the interconnected three-
chamber muffler has orthogonal inlet/outlet ducts, two velocity values have been
provided for each chamber, corresponding to the cross-sections normal to the inlet and
outlet ducts respectively. Flow velocities corresponding to 30, 60 and 90 L/min may be
readily calculated by scaling.
124
Table 6.1 Indicative flow velocity through muffler chambers at a volumetric air flow
rate of 120 L/min
Duct Chamber 1 Chamber 2 Chamber 3

Muffler design velocity velocity velocity velocity
(m/s) (m/s) (m/s) (m/s)
86 mm diameter single
7.9 0.3 --- ---
expansion chamber

7.9 0.2 --- ---
expansion chamber
CPAP device, single

7.9 0.4 --- ---
chamber
CPAP device, interconnected

7.9 0.4-0.6 0.3-0.5 0.3-0.9
three chambers
Figures 6.5 to 6.8 present the transmission loss obtained experimentally for each of the
four mufflers. The results show that increasing the air flow velocity has a negligible
impact on the transmission loss curves with the exception of at the resonant frequency
peaks. At these locations, an increase of air flow velocity leads to a reduction in
transmission loss.
The transmission loss of the single expansion chamber muffler of 114 mm diameter and
the single chamber CPAP device muffler was also experimentally measured with the
light grey foam inserts. It was not possible to measure the transmission loss of the
interconnected three-chamber CPAP device muffler with foam inserts as it was supplied
as a sealed unit without foam inserts fitted. Figure 6.9 presents the results for the 114 mm
diameter muffler and shows that the transmission loss increases significantly with
increasing flow velocity. This is attributed to the scattering effects produced by the air
flow passing through the porous foam insert. It was not possible to obtain results at an air
flow rate of 120 L/min as the system resistance exceeded the capability of the blower.
125
60
No flow 30 L/min
50 60 L/min 90 L/min
120 L/min
40
30
20
10
0
0 1,000 2,000 3,000 4,000 5,000
Frequency (Hz)

muffler, comparing the effect of varying mean air flow – experimental
results
60
No flow 30 L/min
120 L/min
40
30
20
10
0
0 1,000 2,000 3,000 4,000 5,000
Frequency (Hz)
results
126
60
No flow 30 L/min
120 L/min
40
30
20
10
0
0 1,000 2,000 3,000 4,000 5,000
Frequency (Hz)
Figure 6.7 Transmission loss of the single chamber CPAP device muffler, comparing
the effect of varying mean air flow – experimental results
100
80
60
40
20
No flow 30 L/min 60 L/min
90 L/min 120 L/min

0
0 1,000 2,000 3,000 4,000 5,000
Frequency (Hz)

results
127
Figure 6.10 presents the results for the single chamber CPAP device muffler and shows
that mean flow has a negligible impact on the transmission loss for this muffler. This
observation is consistent with the behaviour of the reactive-only mufflers in the presence
of mean flow but is unexpected as it is well-known that grazing flow can significantly
affect the transmission loss in a lined duct. The negligible impact on transmission loss is
attributed to the positioning of the foam insert and the low grazing velocity. The insert
does not intrude into the air path between the muffler inlet and outlet as shown in
Fig. 4.12, and only surrounds 40% of the circumference of the flow path. As the distance
between the inlet and outlet ports is less than 10 mm, there is minimal foam surface area
that is exposed to grazing flow. The maximum grazing velocity, corresponding to a
volumetric air flow rate of 120 L/min, is likely to be less than 1 m/s, or ‫ ܯ‬ൌ ͲǤͲͲ͵.
6.5 Flow Generated Noise
Increasing the mean flow for a range of airflow rates up to 120 L/min resulted in
diminished coherence between the signals recorded by the upstream and downstream
microphones. This effect was especially apparent at frequencies below 1 kHz and at
resonant frequencies. In the case of the three-chamber CPAP device muffler, poor
coherence was also shown to have affected the transmission loss measurements at higher
frequencies when the flow rate was greater than 60 L/min. Flow noise at the two
microphones will be uncorrelated when the microphone separation distance ‫ ݏ‬satisfies the
following condition [124]:
‫ܿܯ‬
‫ݏ‬൐ (6.2)
݂
The limiting case for the measurements that were discussed in the previous section
requires a microphone separation distance greater than 0.13 m to satisfy Eq. (6.2). As the
distance between the upstream and downstream microphones exceeds this requirement,
the flow noise at the two microphones can be considered to be uncorrelated across the
entire frequency range of interest. The reduced coherence as mean flow is increased is
thus attributed to increasing levels of background noise.
128
60
No flow 30 L/min
50
60 L/min 90 L/min
40
30
20
10
0
0 1,000 2,000 3,000 4,000 5,000
Frequency (Hz)
muffler with the light grey foam insert, comparing the effect of varying
mean air flow – experimental results
30
No flow 30 L/min
60 L/min 90 L/min
120 L/min
20
10
0
0 1,000 2,000 3,000 4,000 5,000
Frequency (Hz)
Figure 6.10 Transmission loss of the single chamber CPAP device muffler with the
light grey foam insert, comparing the effect of varying mean air flow –
129
Figure 6.11 shows the signal-to-noise ratio (SNR) at the downstream microphone ‫ܯ‬ଶ
with no muffler present (straight-through pipe) and when the four reactive mufflers are
present in the system. The SNR is calculated as the ratio between the maximum value of
the pressure resulting from the generated acoustic pulse and the maximum value of
pressure fluctuations present in the absence of any pulse. The figure shows that the SNR
reduces rapidly with increasing mean flow. In the case of the three chamber CPAP device
muffler, the SNR is less than 3 dB at 90 and 120 L/min. Measurements taken at such low
SNR levels will be unreliable.
50
No flow 30 Lpm 60 Lpm 90 Lpm 120 Lpm
40
Signal-to-noise ratio (dB)
30
20
10
0
Straight-through 86mm diameter 114mm diameter Single chamber Three chamber
pipe expansion expansion CPAP device CPAP device
chamber chamber muffler muffler
Figure 6.11 Signal-to-noise ratio measured at the downstream (ʹ) microphone (using
peak pressures), comparing the effect of varying mean air flow
Sources of background noise were examined to quantify the relative contribution made
by each source and to determine how the effects of the noise can be minimised. Three
primary sources of noise associated with the presence of air flow are [23, 125-130]:
• aeroacoustic noise within the duct due to turbulence generated by wall friction
and duct fittings;
• aeroacoustic noise due to turbulence generated by air flow passing through the
muffler; and
• mechanical and aeroacoustic noise generated by the blower.
130
In order to assess the distribution of acoustic power of the background noise in the
frequency domain, pressure time histories were measured for a range of mean flows in
the absence of any generated acoustic pulse. A straight-through length of pipe was
inserted into the test rig in the place of a muffler. Figure 6.12 shows the acoustic power
spectrum at the upstream microphone ‫ܯ‬ଵ for the range of mean flows considered. The
flow-generated noise increases between 30 and 40 dB depending on the frequency as the
flow is increased up to the maximum 120 L/min. The figure shows that the greatest
contribution to the level of background noise occurs at the lower frequencies.
100
No flow
90
30 L/min
80
Autospectrum (dB re: 400E-12)
60 L/min
70 90 L/min
120 L/min
60
50
40
30
20
10
0
0 1,000 2,000 3,000 4,000 5,000
Frequency (Hz)
Figure 6.12 Power autospectrum of flow noise measured at the upstream (ͳ)
microphone (with a straight-through duct fitted in place of a muffler),
comparing the effect of varying mean air flow
The contribution to aeroacoustic noise due to the presence of a muffler in the system was
examined by measuring pressure time histories for a range of mean flows in the absence
of any generated acoustic pulse. The magnitude of the pressure was then averaged across
each time history and converted to acoustic power. Measurements were also performed
with the straight-through pipe inserted into the test rig to provide a reference against
which to compare each of the four reactive mufflers. The data shows very strongly linear
131
relationships between the average acoustic power ܹ and the volume flow ܳ raised to an
exponent ݊.
ܹ ൌ ݉ܳ ௡ (6.3)
where ݉ is the gradient of the line of best fit. Equation (6.3) can be re-stated in terms of
the sound power level as
‫ܮ‬ௐ ൌ ሺͳͲ݊ሻ ଵ଴ ܳ ൅ ͳͲ ଵ଴ ݉ െ ͳʹͲ (6.4)
The exponents range between 3.7 and 4.8, depending on the muffler that is being
measured, and are summarised in Table 6.2. The range of exponent values is consistent
with self-generation of noise in duct systems [7].
Table 6.2 Linear curve-fit data for predicting the self-generation of noise due to
various mufflers as a function of volume flow
Exponent Gradient
Design
݊ ݉
Straight-through pipe 4.17 4.51E+05
4.44 4.69E+06
expansion chamber

4.39 2.21E+06
expansion chamber
CPAP device, single chamber 3.67 2.07E+04
CPAP device, interconnected

4.76 4.45E+07
three chambers
Figure 6.13 shows the average acoustic power recorded at the downstream microphone
‫ܯ‬ଶ with the straight-through pipe present and for each of the four reactive mufflers. The
data has been plotted by applying a constant exponent of 4.4 to the volume flow, being
an average of the muffler exponents, and obtaining corresponding linear curve fits for
132
each muffler. The figure shows the linear relationship between the average power and the
volume flow raised to the power of 4.4, with sound power increasing with increasing
flow rate. Although an averaged exponent has been used, the curves produced using the
ܳ ସǤସ power relationship still provide a very close fit to the data, with coefficients of
determination ܴ ଶ greater than 0.995. It can be seen that the magnitude of the average
sound power increases at a greater rate for mufflers having increasingly complex
geometry. This is attributed to the higher level of flow turbulence imparted by complex
geometries. This trend is further exaggerated if the best-fit exponents and curve data for
each muffler is used. Comparison of the results for each muffler against the result
obtained using the straight-through pipe indicates that the aeroacoustic contribution made
by each of the single-chamber mufflers is less than the contribution which may be
attributed to the ducts and fittings of the experimental rig.
7.E-06
Straight-through pipe
6.E-06 86mm diameter expansion chamber
114mm diameter expansion chamber
5.E-06 Single chamber CPAP device muffler
Average Power (W)
Three chamber CPAP device muffler

4.E-06
3.E-06
2.E-06
1.E-06
0.E+00
0.0E+00 5.0E-13 1.0E-12 1.5E-12
Flow4.4 (m3/s)4.4
Figure 6.13 Average sound power measured at the downstream (ʹ) microphone in
the absence of a generated acoustic pulse, comparing the contribution
made by each muffler design
133
In order to assess the contribution made by the blower to background noise, pressure was
measured adjacent to the blower ሺ‫ܯ‬஻ ሻ and immediately downstream of the silencer box
ሺ‫ܯ‬ௌ ሻ, as shown in Fig. 6.14. Measurements were also made in the free field at a distance
of 1 m directly in front of the blower inlet. The autospectra recorded at each location for
a mean flow of 120 L/min were then compared.
Blower
‫ܯ‬஻
‫ܯ‬ௌ
To
1m
horn drivers
Silencer box
‫ܯ‬௙
Figure 6.14 Schematic diagram of the measurement locations used to assess blower
noise and silencer box performance
Figure 6.15 shows that there is little correlation between the blower frequency spectrum
and the spectrum on the downstream side of the silencer box, confirming that the silencer
box is able to effectively attenuate the blower-generated noise. Narrow spikes coinciding
with the blower rotation frequency and subsequent harmonics are still present in the
downstream signal, however they are greatly attenuated and are unlikely to affect muffler
transmission loss measurements. A smooth decrease of the amplitude of the downstream
spectrum with increasing frequency is consistent with the presence of random
turbulence [130].
It is concluded that the source of the background noise that is causing difficulty with the
measurement of muffler transmission loss is aeroacoustic in nature and as such
unavoidable. The limitations that this places on the measurement of transmission loss
using the two microphone acoustic pulse method can be addressed by increasing the
power of the generated pulse.
134
140
Free field Before box After box
120
Autospectrum (dB re: 400E-12)
100
80
60
40
20
0
0 1,000 2,000 3,000 4,000 5,000
Frequency (Hz)
Figure 6.15 Blower noise autospectra measured at 120 L/min, comparing noise in the
free field with noise in the duct adjacent to blower and noise in the duct
downstream of the silencer box
135
Chapter 7: Summary and Future Work
Chapter 7
Summary and Future Work
This thesis has presented computational and experimental techniques to predict and
measure the acoustic performance of irregular shaped mufflers used in positive airway
pressure device applications. Computational techniques were also developed to facilitate
optimisation of the muffler configurations and internal surface properties, that is, both
reactive and resistive elements, to maximise acoustic performance. Finally, an
experimental procedure was developed to measure muffler acoustic performance in the
presence of mean flow.
7.1 Summary
7.1.1 Prediction and measurement of muffler acoustic performance
7.1.1.1 Reactive mufflers
The transmission loss of single and multiple expansion chamber reactive mufflers was
predicted analytically using (i) the continuity of pressure and volume velocity method,
(ii) the impedance method and (iii) the transfer matrix method. The continuity method
and the impedance method were shown to produce identical equations for prediction of
the transmission loss for both muffler designs. The transfer matrix method also yielded
an identical result when the effects of expansion at the chamber inlet and contraction at
the chamber outlet were neglected. Accounting for these end effects by introducing Karal
correction factors resulted in shifts in both the frequency and magnitude of the predicted
transmission loss when compared to the other two methods.
A finite element analysis (FEA) model of each of the muffler designs was developed and
two acoustic finite element approaches were compared. The first approach used a direct
calculation of the attenuation of acoustic power, while the second approach obtained
transfer matrix parameters from the finite element solution and then used these
parameters to calculate the transmission loss. Both approaches were shown to yield
136
identical transmission loss predictions. The FEA results produced close agreement with
the repeating dome behaviour that was predicted by the analytical methods for
frequencies below the onset of the first circumferential mode. Of the three analytical
methods considered, the transfer matrix method incorporating Karal correction factors
exhibited the closest agreement with the FEA results. At frequencies greater than the
onset of the first circumferential mode, the FEA results departed significantly from the
repeating dome behaviour predicted by the analytical solutions. This departure was
attributed to the presence of higher order modes which cannot be accounted for using
plane wave theory. The transmission loss of the single chamber muffler was
experimentally measured using a two-microphone acoustic pulse method. The FEA
results showed excellent agreement with the experimental results over the frequency
range assessed.
Expansion chambers within CPAP device mufflers typically have large expansion ratios
and short length-to-diameter ratios. The transmission loss of a cylindrical expansion
chamber muffler having dimensions similar to a generic CPAP device muffler was
predicted analytically and computationally. The FEA results departed significantly from
the analytical solution across most of the frequency range. The transmission loss of the
generic CPAP device muffler was measured and the FEA results showed good agreement
with the experimental results over the frequency range assessed.
7.1.1.2 Resistive mufflers
A two-cavity impedance tube approach was used to experimentally obtain the

characteristic impedance and propagation constant of two polyurethane foams,
corresponding to a light grey foam and a dark grey foam. The data was fitted to empirical
power-law functions based on those originally proposed by Delany and Bazley [70]. A
method was presented which simplified the acoustic characterisation of the porous
materials by enabling the coefficients in the Delany-Bazley functions to be determined
without the requirement to obtain the airflow resistivity. The numerical values for the
characteristic impedance and propagation constant predicted using the experimentally
derived coefficients showed excellent agreement with the experimental data. In
comparison, use of the original Delany-Bazley coefficients was shown to result in an
over-prediction of the equivalent fluid speed of sound and an under-prediction of the
density for the light grey foam.
137
Acoustic properties of the two polyurethane foams were incorporated into the finite
element models by applying an equivalent fluid approach using the Delany-Bazley
formulation with the experimentally derived coefficients. The finite element model of the
generic CPAP device muffler was modified to incorporate both air-filled and foam-filled
regions. The transmission loss of the muffler was measured and the FEA results showed
good agreement with the experimental results over the frequency range assessed. The two
foams used in the inserts were shown to have a similar impact on the acoustic
performance of this muffler design despite a difference in apparent density of
approximately 50%.
The transmission loss of the generic CPAP device muffler was also predicted
computationally using the original Delany-Bazley coefficients. The FEA results obtained
using the experimentally derived coefficients were shown to be more closely aligned to
the experimental measurements than those obtained using the original coefficients.
7.1.1.3 CPAP device mufflers
The computational methodologies developed for the reactive and resistive mufflers were
applied to the more complex muffler geometries used in sleep apnoea assisted breathing
devices. The transmission loss of three CPAP device muffler designs was predicted
analytically, computationally and experimentally. The mufflers included a single
chamber design, an integrated design comprising two chambers with a vertical internal
baffle and a design comprising three interconnected chambers. Comparison of the
transmission loss obtained analytically and experimentally for each of the muffler
designs showed poor correlation and confirmed that the one dimensional analytical
methods are unreliable beyond very simple designs.
The FEA results obtained for each of the muffler designs showed good agreement with
the experimental results for the majority of the frequency range, with the exception that
the magnitude at resonant frequencies was over-predicted by the models. This was
attributed to the lack of acoustic damping in the finite element models. Departure
between the experimental and FEA results at higher frequencies was attributed to
experimental limitations.
The acoustic characteristics of the two polyurethane foams were incorporated into finite
element models of the single chamber muffler and the integrated two-chamber muffler.
138
The transmission loss of each resistive muffler was also measured experimentally. The
FEA results obtained for both of the muffler designs showed good agreement with the
experimental results over the frequency range assessed. The inclusion of foam inserts had
little effect on the acoustic performance at low frequencies but contributed significantly
to increased transmission loss at higher frequencies. The presence of foam also resulted
in a dramatic reduction in the transmission loss at resonant peaks.
The two different foam types were shown to have a similar impact on the acoustic
performance of the single chamber muffler design despite a difference in apparent
density of approximately 50%. This observation was attributed to the location of the
foam insert such that it did not obstruct the path between the inlet and outlet ports. In
contrast, the results for the integrated chamber muffler showed that the two foam
materials made differing contributions to the acoustic performance of this muffler design.
This observation was attributed to the presence of a significant volume of foam in the
direct path between the inlet and outlet ports of the first of the two integrated chambers.
The differing contribution made by each of the foams was consistent with the differences
in apparent density and flow resistivity.
7.1.2 Optimisation of muffler acoustic performance
Optimisation of the acoustic performance of mufflers was performed by developing an

approach which integrated the finite element analysis capabilities of the COMSOL
software package with the optimisation tools contained within MATLAB. A range of
optimisation algorithms were investigated corresponding to the Nelder-Mead, genetic
algorithm, simulated annealing and particle swarm optimisation approaches. The genetic
algorithm was selected to optimise the muffler acoustic performance. The transmission
loss of mufflers was optimised by using geometric dimensions such as expansion
chamber length and/or diameter as the design control variables. In the case of a resistive
muffler where a foam insert was present, dimensions such as the foam thickness and/or
the length that the lining extends along an expansion chamber were used as the assigned
control variables.
Optimisation of the geometry of a cylindrical expansion chamber reactive muffler was

initially examined. Results that were obtained using the genetic algorithm were shown to
be in close agreement with the predicted theoretical values. A similar cylindrical
139
expansion chamber muffler was lined with foam on the chamber wall and optimisation of
the foam dimensions was performed. Optimisation results were shown to coincide with
the global transmission loss maximum in the fitness domain. The presence of one
transmission loss maximum and two transmission loss minima in the bounded domain
highlighted that the inclusion of a foam lining in this muffler could both enhance and
degrade the muffler acoustic performance, depending on the selected combination of
foam thickness and length.
A similar process was then applied to optimise the acoustic performance of a prototype
CPAP device muffler comprising three interconnected chambers, with and without the
presence of foam inserts. Ten control variables were used to define the muffler geometry
and the thickness of foam inserts during the optimisation analyses. Optimisation was
performed using a number of discrete frequencies within targeted frequency bands with
the aim of achieving the maximum averaged transmission loss across the full 5 kHz
frequency range of interest.
The reactive and resistive components of the CPAP device muffler were initially
independently optimised using various frequency bands. Optimisation of the muffler
geometry in the low frequency band resulted in a significant increase in transmission loss
across the full frequency range when compared to the pre-optimised design. Optimisation
of the thickness of the foam inserts in the high frequency band using the optimised
geometry produced an additional increase in the averaged transmission loss. Results were
then obtained for the transmission loss of the CPAP device muffler for which the reactive
and resistive components were simultaneously optimised. The averaged transmission loss
over the entire frequency range assessed resulting from the simultaneous geometry and
foam optimisation was greater than the loss obtained by optimising the reactive and
resistive components sequentially. A final optimisation was performed using equally
weighted frequencies throughout the entire 5 kHz frequency range. While this
optimisation produced the greatest averaged transmission loss, the increase over that
achieved using the simultaneous optimisation with targeted frequency bands was only
marginal yet tripled the required computational time.
140
7.1.3 Measurement of muffler acoustic performance with mean flow
The maximum airflow rate supplied by a CPAP device blower was shown to produce a
corresponding duct velocity of 7 m/s, or Mach 0.02. The transmission loss of four
mufflers was measured experimentally at duct velocities up to 8 m/s. The muffler designs
included two simple expansion chamber mufflers, the single chamber CPAP device
muffler and the interconnected three chamber CPAP device muffler. Comparison of the
transmission loss in the presence of mean flow for each of the mufflers with experimental
results obtained in the absence of flow showed that increasing the air flow velocity had a
negligible impact on the transmission loss with the exception of the resonant frequency
peaks. At these locations, an increase of air flow velocity lead to a reduction in
transmission loss.
Polyurethane foam inserts were fitted to one of the single expansion chamber mufflers
and to the single chamber CPAP device muffler. The acoustic performance was
experimentally measured in the presence of mean flow. Comparison of the transmission
loss in the presence of mean flow with experimental results obtained in the absence of
flow showed that the transmission loss of the single expansion chamber muffler
increased significantly with increasing flow velocity. This was attributed to the scattering
effects produced by air passing through the porous foam insert. In contrast, mean flow
had a negligible impact on the transmission loss of the single chamber CPAP device
muffler despite the presence of the foam insert. This was attributed to the positioning of
the foam insert to one side of the air path between the muffler inlet and outlet, as the flow
is not required to pass through the foam.
Increasing the mean flow resulted in diminished coherence between the signals recorded
by the upstream and downstream microphones. The increased flow also resulted in
diminished signal-to-noise ratios at the downstream microphone and was attributed to
increasing levels of background noise. Sources of noise were assessed and included those
due to (i) flow turbulence generated by wall friction and duct fittings within the
experimental test rig, (ii) turbulence generated by complex muffler geometries and
(iii) mechanical and aeroacoustic noise generated by the blower. It was concluded that
the background noise which was causing difficulty with the measurement of muffler
transmission loss was predominantly due to flow turbulence generated by wall friction
and duct fittings.
141
7.2 Recommendations for Future Work
Suggestions for enhancements to the finite element models and experimental techniques
presented in this thesis are described in what follows.
7.2.1 Computational models
The finite element models that were developed in this thesis included the simplifying
assumption that the working fluid was inviscid. This assumption leads to under-damped
behaviour as evidenced by much sharper peaks and troughs in the transmission loss
results. Adjusting the damping in the FEA models so that the peaks are similar in
sharpness to the experimentally measured ones would increase the accuracy of the
computational models of the CPAP device mufflers. Damping may be added by defining
the speed of sound in the fluid to be a complex property and setting a small imaginary
component [131-133].
The finite element models also included the simplifying assumption that the fluid-
structure interface was totally rigid, that is, the walls were regarded to be acoustically
hard boundaries. The tendency of the FEA models to over-predict the magnitude of the
transmission loss at resonant frequencies was attributed to this assumption. CPAP device
mufflers are increasingly being manufactured with thin-walled plastic sections. In
addition, the compact nature of CPAP devices often requires that adjacent muffler
chambers share common walls. The complete prediction of the acoustic performance of
these mufflers requires that transfer of energy through the structure be accounted for.
Incorporating non-rigid boundary physics into the FEA models would increase the
accuracy of the computational models of the CPAP device mufflers.
As consumers become increasingly aware of the impact of noise, competitive pressure

will increase the requirement for timely and efficient optimisation of muffler designs.
The growing body of literature on the subject of computational design optimisation
shows that this is an important research area. Although the optimisation work carried out
in this thesis included evaluation of the relative performance of a number of optimisation
algorithms, default/recommended settings were used for many of the controlling
parameters. Results showed that there are efficiencies to be gained by further
investigation of suitable optimisation algorithms and approaches.
142
7.2.2 Experimental methods
7.2.2.1 Pulse generation
The two-microphone method used to measure the transmission loss of mufflers in this
thesis required the generation of an intense short duration acoustic pulse. The transient
acoustic pulse that was used consisted of a half sinusoid which was fed to two horn
drivers mounted directly opposite each other and flush with the side of the main duct.
During experimental testing, it was noted that the transmitted pulse from the more
complex muffler designs was very low, resulting in poor coherence and marginal signal-
to-noise ratios. The experimental technique would benefit from generation of a higher
intensity impulse which provides sufficient energy across the frequency range of interest,
however any improvement is limited by the risk of introducing non-linear effects. Further
improvement in pulse intensity may be possible by orienting the horn drivers such that
their generated pulse is directed towards the muffler rather than normal to the main duct.
7.2.2.2 Separation of initial and reflected pulses
The initial transient acoustic pulse is reflected at a number of locations throughout the
experimental rig. As the experimental technique used in this thesis calculates the
attenuation of the initial positive-travelling wave, temporal windowing of the upstream
and downstream microphone signals is required to capture the initial incident and
transmitted acoustic pulses while excluding any subsequent reflections. Interaction
between the adjacent chambers in the more complex muffler designs produced a series of
pulses at the muffler outlet, increasing the time decay of the transmitted pressure signal.
The combined duration of these pulses approached the time taken for the first reflected
pulse to return to the downstream microphone, complicating the separation of the
transmitted and reflected pulses. While the dimensions of the experimental rig used in
this thesis were adequate for single chamber mufflers, better resolution of the results for
the more complex muffler designs could be achieved with increased dimensions.
7.2.2.3 Alternative transmission loss experimental approaches
Alternative experimental techniques could be used to overcome the limitations of the

two-microphone method when applied to the measurement of the transmission loss of
CPAP device mufflers. The four-pole (two-port) approach [134-136] is widely applied in
143
duct and muffler applications. An extensively cited four-pole approach is the two-source
location method described by Munjal and Doige [137].
7.2.2.4 Preparation of foam samples
In this thesis, measurements of the acoustic properties of polyurethane foam using an

impedance tube were observed to be affected by sample preparation and the influence of
the circumferential edge constraint. Oversized samples result in compression and
distortion of the foam structure when mounted in the impedance tube. Undersized
samples are poorly constrained and may slide within the tube, also producing erroneous
results. Preparation of cylindrical samples of polyurethane foam is complicated by the
flexible nature of the material. The quality of the acoustic characterisation of the foam
could be enhanced by close attention to sample preparation and the use of appropriate
cutting tools and/or techniques.
7.2.2.5 Flow generated noise
Increasing the mean airflow through the experimental rig resulted in reduced signal-to-
noise ratios, particularly at the downstream microphone, as well as diminished coherence
between the signals recorded by the upstream and downstream microphones. These
observations were attributed to increasing levels of background noise. It was concluded
that the noise was aeroacoustic in nature and that the largest contribution was produced
by the ducts and fittings of the experimental rig. Increasing the diameter of the duct used
throughout the experimental rig would reduce the flow velocity and associated
aeroacoustic noise in the rig. However, appropriate duct transitions would be required at
the muffler and microphone connections to minimise impedance mismatches. In addition,
the bell mouth located at the discharge from the silencer box had a relatively tight bell
internal radius. The profile of the bell could be improved to result in a smoother entry of
air into the duct and thereby reduce any aeroacoustic noise generated by the silencer-duct
transition. Alternatively, increasing the power of the generated acoustic pulse would
increase the signal-to-noise ratio and overcome the limitations that background noise
places on the measurement of transmission loss using the two microphone acoustic pulse
method.
The upstream and downstream microphones were mounted perpendicular to the main
conduit such that the face of the microphones was recessed 4 mm from the inner surface
144
of the conduit but was otherwise exposed to the effects of the mean airflow. The
aeroacoustic noise on the microphones could be reduced by mounting each microphone
in a short, closed-end tube attached to the duct wall and placing a foam or mylar cover
over the entrance to the tube where it is attached to the duct. The cover should be shaped
so as to present a flush surface on the inside of the conduit to prevent the formation of
turbulent eddies at the entrance to the microphone holder.
7.2.3 Further design applications
The finite element models that were developed in this thesis treated all boundaries,
except the inlet and outlet ports, as being acoustically hard. Due to the compact nature of
CPAP device mufflers, adjacent muffler chambers may share common walls. Further
work is recommended to investigate the effects of through-wall acoustical coupling of
muffler chambers, especially where the adjacent chambers are located on opposite sides
of the flow generating fan. This effect might be accentuated by the replacement of the
common wall with a flexible membrane.
The focus of the finite element models that were developed in this thesis was reduction
of the conducted noise due to the flow generating fan, that is, the noise which propagates
down the air path and along the hose to the mask. Further work is recommended to also
consider the radiated noise, that is, the acoustic energy that is transferred through the
chamber walls and then radiated into the free field. This might provide future benefits in
guiding the structural design of the chassis of CPAP devices.
The acoustic finite element models were demonstrated to provide a good prediction of
the performance of mufflers in sleep apnoea devices. The optimisation methodology
showed that encouraging improvement in performance can be achieved at the design
stage of product development. These modelling and optimisation approaches have
relevance to the design of mufflers for a broad range of applications. For example, a tool
that can predict and optimise the acoustic performance of irregular shaped mufflers
would be of benefit where industrial mufflers must be incorporated into irregular and/or
confined spaces. Hence, computational acoustic optimisation models may allow
increased design flexibility in applications such as locomotives and high speed ferries.
145
References
References
[1] Young, T. “Rationale, design, and findings from the Wisconsin Sleep Cohort
Study: Toward understanding the total societal burden of sleep-disordered
breathing”, Journal of Clinical Sleep Medicine, 4, 37-46 (2009)
[2] Stuart, M. “Sleep apnoea devices: The changing of the guard”, Start-up, 15(10),
2-9 (2010)
[3] Thorpy, M.J. (ed.) International Classification of Sleep Disorders: Diagnostic
and Coding Manual, American Sleep Disorders Association, Minnesota, USA,
1990
[4] ResMed (2010) “What is Sleep Apnoea?” <http://www.resmed.com> Accessed
20 April 2011
[5] Barron, R.F. Industrial Noise Control and Acoustics, Marcel Dekker, New
York, 2003
[6] Morfey, C.L. Dictionary of Acoustics, Academic Press, London, 2001
[7] Bies, D.A. and Hansen, C.H. Engineering Noise Control: Theory and practice,
Unwin Hyman, London, 1988
[8] Ih, J.G. and Lee, B.H. “Theoretical prediction of the transmission loss of
circular reversing chamber mufflers”, Journal of Sound and Vibration, 112, 261-
272 (1987)
[9] Jones, A.D. “Modelling the exhaust noise radiated from reciprocating internal
combustion engines – A literature review”, Noise Control Engineering Journal,
23, 12-31 (1984)
[10] Munjal, M.L. Acoustics of Ducts and Mufflers, John Wiley and Sons, New
York, 1987
[11] Davis, R.D., Stokes, G.M., Moore, D. and Stevens, G.L. “Theoretical and
experimental Investigation of mufflers with comments on engine exhaust
muffler design”, TN 1192, National Advisory Committee for Aeronautics
(NACA), 1954
[12] Harris, C.M. (ed.) Handbook of Noise Control, McGraw-Hill, New York, 1957
146
References
[13] Karal, F.C. “The analogous acoustical impedance for discontinuities and
constriction of circular cross section”, Journal of the Acoustical Society of
America, 25, 327-334 (1953)
[14] Igarashi, J. and Toyama, M. “Fundamentals of Acoustic Silencers: (I) Theory
and experiment of acoustic low-pass filters”, (pp. 223-241), Report No. 339,
Aeronautical Research Institute, University of Tokyo, 1958
[15] Miwa, T. and Igarashi, J. “Fundamentals of Acoustic Silencers: (II)
Determination of four terminal constants of acoustical elements”, (pp. 67-85),
Report No. 344, Aeronautical Research Institute, University of Tokyo, 1959
[16] Davies, P.O.A.L. “Realistic models for predicting sound propagation in flow
duct systems”, Noise Control Engineering Journal, 40, 135-141 (1993)
[17] Sullivan, J.W. “A method for modelling perforated tube muffler components: (I)
Theory”, Journal of the Acoustical Society of America, 66, 772-778 (1979)
[18] Sullivan, J.W. “A method for modelling perforated tube muffler components:
(II) Applications”, Journal of the Acoustical Society of America, 66, 779-788
(1979)
[19] Crocker, M.J. Handbook of Acoustics, John Wiley and Sons, New York, 1998
[20] Alfredson, R.J. and Davies, P.O.A.L. “Performance of exhaust silencer
components”, Journal of Sound and Vibration, 15, 175-196 (1971)
[21] Munjal, M.L. “Velocity ratio-cum-transfer matrix method for the evaluation of a
muffler with mean flow”, Journal of Sound and Vibration, 39, 105-119 (1975)
[22] Peat, K.S. “Evaluation of four-pole parameters for ducts with flow by the finite
element method”, Journal of Sound and Vibration, 84, 389-395 (1982)
[23] Munjal, M.L. and Prasad, M.G. “On plane wave propagation in a uniform pipe
in the presence of a mean flow and a temperature gradient”, Journal of the
Acoustical Society of America, 80, 1501-1506 (1986)
[24] Fukuda, M., Kojima, N. and Iwaish, T. “A study on the mufflers with flow”,
Transactions of the Japanese Society of Mechanical Engineers, 48(B), 1-9
(1983)
[25] Peat, K.S. “The acoustical impedance at discontinuities of ducts in the presence
of mean flow”, Journal of Sound and Vibration, 127, 123-132 (1988)
[26] Åbom, M. “Derivation of four pole parameters including higher order mode
effects for expansion chamber mufflers with extended inlet and outlet”, Journal
of Sound and Vibration, 137, 403-418 (1990)
147
References
[27] Selamet, A. and Radavich, P.M. “The effect of length on the acoustic
attenuation performance of concentric expansion chambers: An analytical,
computational and experimental investigation”, Journal of Sound and Vibration,
201(4), 407-426 (1997)
[28] Young, C-I.J. and Crocker, M.J. “Prediction of transmission loss in mufflers by
the finite element method”, Journal of the Acoustical Society of America, 57,
144-148 (1975)
[29] Eriksson, L.J. “Higher order mode effects in circular ducts and expansion
chambers”, Journal of the Acoustical Society of America, 68, 545-560 (1980)
[30] Ih, J.G. and Lee, B.H. “Analysis of higher order mode effects in the circular
expansion chamber with mean flow”, Journal of the Acoustical Society of
America, 77, 1377-1388 (1985)
[31] Yi, S.I. and Lee, B.H. “Three dimensional acoustic analysis of circular
expansion chambers with a side inlet and a side outlet”, Journal of the
[32] Yi, S.I. and Lee, B.H. “Three dimensional acoustic analysis of a circular
expansion chamber with side inlet and end outlet”, Journal of the Acoustical
Society of America, 81, 1279-1287 (1987)
[33] Munjal, M.L. “A simple numerical method for three-dimensional analysis of
simple expansion chamber mufflers of rectangular as well as circular cross-
section with stationary medium”, Journal of Sound and Vibration, 116, 71-88
(1987)
[34] Depollier, C., Kergomard, J. and Lesueur, J.C. “Propagation of low frequency
acoustic waves in periodic 2-D lattice of tubes”, Journal of Sound and
Vibration, 142, 153-170 (1990)
[35] Glav, R. and Åbom, M. “A general formulism for analysing acoustic 2-port
networks”, Journal of Sound and Vibration, 202, 739-747 (1997)
[36] Dowling, J.F. and Peat, K.S. “An algorithm for the efficient acoustic analysis of
silencers of any general geometry”, Applied Acoustics, 65, 211-227 (2004)
[37] Baumeister, K.J. “Numerical techniques in linear duct acoustics – A status
report”, TM-81553 (E-513), National Aeronautics and Space Administration,
1980
148
References
[38] Sahasrabudhe, A.D., Ramu, S.A. and Munjal, M.L. “Matrix condensation and
transfer matrix techniques in the 3-D analysis of expansion chamber mufflers”,
Journal of Sound and Vibration, 147(3), 371-394 (1991)
[39] Bilawchuk, S. and Fyfe, K.R. “Comparison and implementation of the various
numerical methods used for calculating transmission loss in silencer systems”,
Applied Acoustics, 64, 903-916 (2003)
[40] Wu, T.W. and Wan, G.C. “Muffler performance studies using a direct mixed-
body boundary element method and a three-point method for evaluating
transmission loss”, ASME Journal of Vibration and Acoustics, 118, 479-484
(1996)
[41] Gladwell, G.M.L. “A variational formulation of damped acousto-structural
vibration problems”, Journal of Sound and Vibration, 4, 172-186 (1966)
[42] Craggs, A. “A finite element method for damped acoustic systems: An
application to evaluate the performance of reactive mufflers”, Journal of Sound
and Vibration, 48(3), 377-392 (1976)
[43] Barbieri, R., Barbieri, N. and de Lima, K.F. “Application of the Galerkin-FEM
and the improved four-pole parameter method to predict acoustic performance
of expansion chambers”, Journal of Sound and Vibration, 276, 1101-1107
(2004)
[44] Wu, T.W., Zhang, P. and Cheng, C.Y.R. “Boundary element analysis of
mufflers with an improved method for deriving the four-pole parameters”,
[45] Selamet, A. and Ji, Z.L. “Acoustic attenuation performance of circular
expansion chambers with extended inlet/outlet”, Journal of Sound and
Vibration, 223, 197-212 (1999)
[46] Mehdizadeh, O.Z. and Paraschivoiu, M. “A three-dimensional finite element
approach for predicting the transmission loss in mufflers and silencers with no
mean flow”, Applied Acoustics, 66, 902-918 (2005)
[47] Peat, K.S. and Rathi, K.L. “A finite element analysis of the convected acoustic
wave motion in dissipative silencers”, Journal of Sound and Vibration, 184(3),
529-545 (1995)
[48] Kirby, R. “A comparison between analytical and numerical methods for
modelling automotive silencers with mean flow”, Journal of Sound and
Vibration, 325, 565-582 (2009)
149
References
[49] Harari, I. and Hughes, T.J.R. “Cost comparison of boundary element and finite
element methods for problems of time-harmonic acoustics”, Computer Methods
in Applied Mechanics and Engineering, 97(1), 77-102 (1992)
[50] Seybert, A.F. and Cheng, C.Y.R. “Application of the boundary element method
to acoustic cavity response and muffler analysis”, Journal of Vibration,
Acoustics, Stress, and Reliability in Design, 109(1), 15-21 (1987)
[51] Cheng, C.Y.R., Seybert, A.F. and Wu, T.W. “A multi-domain boundary element
solution for silencer and muffler performance prediction”, Journal of Sound and
Vibration, 151(1), 119-129 (1991)
[52] Ji, Z., Ma, Q. and Zhang, Z. “Application of the boundary element method to
predicting acoustic performance of expansion chamber mufflers with mean
flow”, Journal of Sound and Vibration, 173(1), 57-71 (1994)
[53] Wu, T.W., Cheng, C.Y.R. and Zhang, P. “A direct mixed-body boundary
element method for packed silencers”, Journal of the Acoustical Society of
America, 111(6), 2566-2572 (2002)
[54] Selamet, A., Lee, I.J. and Huff, N.T. “Acoustic attenuation of hybrid silencers”,
[55] Cooke, C.H. “Super-computer simulation of a multi-chambered muffler”,
Numerical Methods for Partial Differential Equations, 5(2), 97-106 (1989)
[56] Jebasinski, R. and Eberspacher, J. “Calculation of the tail-pipe noise of exhaust
systems with Wave”, Ricardo Software International User Conference, 1 March
1996, Detroit, Michigan
[57] Middelberg, J.M. Prediction and validation of the mean flow and the acoustic
performance of reactive mufflers using computational fluid dynamics, PhD
thesis, The University of New South Wales, Sydney, 2008
[58] Obikane, Y. “Aeroacoustic Simulation in Automobile Muffler by Using the
Exact Compressible Navier-Stokes Equation”, in Choi, H., Choi, H.G.,
Yoo, J.Y. (eds.) Computational Fluid Dynamics 2008: Part 4, Springer, Berlin-
Heidelberg, 2009
[59] Gerges, S.N.Y., Jordan, R., Thieme, F.A., Bento Coelho, J.L. and Arenas, J.P.
“Muffler modelling by transfer matrix method and experimental verification”,
Journal of the Brazilian Society of Mechanical Science and Engineering, 27(2),
132-140 (2005)
150
References
[60] Biot, M.A. “Theory of propagation of elastic wave in a fluid saturated porous
solid”, Journal of the Acoustical Society of America, 28, 168-178 (1956)
[61] Biot, M.A. “Mechanics of deformation and acoustic propagation in porous
media”, Journal of Applied Physics, 33, 1482-1498 (1962)
[62] Zwikker, C. and Kosten, C.W. Sound Absorbing Materials, Elsevier, New York,
1949
[63] Lambert, R.F. “Propagation of sound in highly porous open-cell foams”,
Journal of the Acoustical Society of America, 73, 1131-1138 (1983)
[64] Beranek, L.L. “Acoustical properties of homogeneous, isotropic rigid tiles and
flexible blankets”, Journal of the Acoustical Society of America, 19(4), 556-568
(1947)
[65] Allard, J.F., Aknine, A. and Depollier, C. “Acoustical properties of partially
reticulated foams with high and medium flow resistance”, Journal of the
[66] Bolton, J.S., Shiau, N.M. and Kang, Y.J. “Sound transmission through
multipanel structures lined with elastic porous materials”, Journal of Sound and
Vibration, 191(3), 317–347, (1996)
[67] Morse, P.M. and Ingard, K.U. Theoretical Acoustics, McGraw-Hill, New York,
1968
[68] Allard, J.F. and Champoux, Y. “New empirical equations for sound propagation
in rigid frame fibrous materials”, Journal of the Acoustical Society of America,
91(6), 3346-3353 (1992)
[69] Johnson, D.L., Koplik, J. and Dashen, R. “Theory of dynamic permeability and
tortuosity in fluid-saturated porous media”, Journal of Fluid Mechanics, 176,
379-402 (1987)
[70] Delany, M.E. and Bazley, E.N. “Acoustical properties of fibrous absorbent
materials”, Applied Acoustics, 3, 105-116 (1970)
[71] Miki, Y. “Acoustical properties of porous materials – Modifications of Delany-
Bazley models”, Journal of the Acoustical Society of Japan, 11(1), 19-28 (1990)
[72] Bies, D.A. and Hansen, C.H. “Flow resistance information for acoustical
design”, Applied Acoustics, 13, 357-391 (1980)
[73] Mechel, F.P. “Chapter 8 - Sound-absorbing materials and sound absorbers” in
Noise and Vibration Control Engineering, Beranek, L.L. and Vér, I.L. (eds.),
John Wiley and Sons, New York, 1992
151
References
[74] Attenborough, K. “Acoustical characteristics of porous materials”, Physics

Reports, 82(3), 179-227 (1982)
[75] Dunn, I.P. and Davern, W.A. “Calculation of acoustic impedance of multi-layer
absorbers”, Applied Acoustics, 19, 321-334 (1986)
[76] Wu, Q. “Empirical relations between acoustical properties and flow resistivity
of porous plastic open-cell foam”, Applied Acoustics, 25, 141-148 (1988)
[77] Ling, M.K. “Technical note – Impedance of polyurethane foams”, Applied
Acoustics, 34, 221-224 (1991)
[78] Kidner, M.R.F. and Hansen, C.H. “A comparison and review of theories of the
acoustics of porous materials”, International Journal of Acoustics and
Vibration, 13(3), 112-119 (2008)
[79] Bernhard, J.R. “A finite element method for synthesis of acoustical shapes”,
Journal of Sound and Vibration, 89, 55-65 (1985)
[80] Yeh, L-J., Chang, Y-C., Chiu, M-C. and Lai, G-J. “GA optimization on multi-
segments muffler under space constraints”, Applied Acoustics, 65, 521-543
(2004)
[81] Chang, Y-C., Yeh, L-J. and Chiu, M-C. “Shape optimization on constrained
single-chamber muffler by using GA method and mathematical gradient
method”, International Journal of Acoustics and Vibration, 10(1), 17-25 (2005)
[82] Yeh, L-J., Chang, Y-C., Chiu, M-C. “Numerical studies on constrained venting
system with reactive mufflers by GA optimization”, International Journal for
Numerical Methods in Engineering, 65, 1165-1185 (2006)
[83] Chiu, M-C. and Chang, Y-C. “Numerical studies on venting system with multi-
chamber perforated mufflers by GA optimization”, Applied Acoustics, 69, 1017-
1037 (2008)
[84] Chiu, M-C. and Chang, Y-C. “Shape optimisation of multi-chamber cross-flow
mufflers by SA optimization”, Journal of Sound and Vibration, 312, 526-550
(2008)
[85] Seo, S-H. and Kim, Y-H. “Silencer design by using array resonators for low-
frequency band noise reduction”, Journal of the Acoustical Society of America,
118(4), 2332-2338 (2005)
[86] Barbieri, R. and Barbieri, N. “Finite element acoustic simulation based shape
optimization of a muffler”, Applied Acoustics, 67, 346-357 (2006)
152
References
[87] Zoutendijk, G. Methods of Feasible Directions: A Study in Linear and

Nonlinear Programming, Elsevier, New York, 1960
[88] de Lima, K.F., Lenzi, A. and Barbieri, R. “The study of reactive silencers by
shape and parametric optimization techniques”, Applied Acoustics, 72, 142-150
(2011)
[89] Bendsøe, M.P. and Kikuchi, N. “Generating optimal topologies in structural
design using a homogenization method”, Computer Methods in Applied
Mechanics and Engineering, 71(2), 197-224 (1988)
[90] Jensen, J.S. and Sigmund, O. “Systematic design of acoustic devices by
topology optimization”, Proceedings of the 12th International Congress on
Sound and Vibration, 11-14 July 2005, Lisbon, Portugal
[91] Bendsøe, M.P. and Sigmund, O. Topology Optimization: Theory, Methods and
Applications, Springer Verlag, Berlin, 2003
[92] Dühring, M.B., Jensen, J.S. and Sigmund, O. “Acoustic design by topology
optimization”, Journal of Sound and Vibration, 317, 557-575 (2008)
[93] Svanberg, K. “The method of moving asymptotes – a new method for structural
optimization”, International Journal for Numerical Methods in Engineering, 24,
359-373 (1987)
[94] Sigmund, O. “Morphology-based black and white filters for topology
optimization”, Structural and Multidisciplinary Optimization, 33, 401-424
(2007)
[95] Lee, J.W. and Kim, Y.Y. “Topology optimization of muffler internal partitions
for improving acoustical attenuation performance”, International Journal for
Numerical Methods in Engineering, 80, 455-477 (2009)
[96] Marburg, S. “Six elements per wavelength. Is that enough?”, Journal of
Computational Acoustics, 10, 25-51 (2002)
[97] Kim, J. and Soedel, W. “Analysis of gas pulsations in multiply connected three-
dimensional acoustic cavities with special attention to natural mode or wave
cancellation effects”, Journal of Sound and Vibration, 131, 103-114 (1989)
[98] Barbieri, N., Barbieri, R. and de Lima, K.F. “Errors in transmission loss
prediction – the bispectrum and kurtosis approaches”, Mechanical Systems and
Signal Processing 18, 223-233 (2004)
153
References
[99] Imaoka, S. (2004) “Sheldon’s ANSYS Tips and Tricks: Acoustic Elements and
Boundary Conditions”, Memo Number STI:05/01B. <http://ansys.net> Accessed
20 April 2011
[100] Seybert, A.F. and Ross, D.F. “Experimental determination of acoustic properties
using a two-microphone random excitation technique”, Journal of the
Acoustical Society of America, 61(5), 1362-1370 (1977)
[101] Cummings, A. and Chang, I. -J. “Sound attenuation of a finite length dissipative
flow duct silencer with internal mean flow in the absorbent”, Journal of Sound
and Vibration, 127(1), 1-17 (1988)
[102] Payri, F., Desantes, J. M. and Broatch, A. “Modified impulse method for the
measurement of the frequency response of acoustic filters to weakly nonlinear
transient excitations”, Journal of the Acoustical Society of America, 107(2), 731-
738 (2000)
[103] Kirby, R. “Simplified techniques for predicting the transmission loss of a
circular dissipative silencer”, Journal of Sound and Vibration, 243(3), 403-426
(2001)
[104] Døssing, O. Structural Testing: Part 1: Mechanical Mobility Measurements,
Brüel & Kjær, Nærum, 1988
[105] Randall, R.B. Frequency Analysis, 3rd Ed., Brüel & Kjær, Copenhagen, 1987
[106] Seybert, A.F., Seman, R.A. and Lattuca, M.D. “Boundary element prediction of
sound propagation in ducts containing bulk absorbing materials”, Journal of
Vibration and Acoustics – Transactions of the ASME, 120, 976-981 (1998)
[107] Scott, R.A. “The absorption of sound in a homogeneous porous medium”,
Proceedings of the Physical Society, 58(2), 165-183 (1946)
[108] Mechel, F.P. “Absorption cross section of absorber cylinders”, Journal of Sound
and Vibration, 107, 131-148 (1986)
[109] Utsuno, H., Tanaka, T., Fujikawa, T. and Seybert, A.F. “Transfer function
method for measuring characteristic impedance and propagation constant of
porous materials”, Journal of the Acoustical Society of America, 86(2), 637-643
(1989)
[110] Lee, I., Selamet, A. and Huff, N. "Acoustic impedance of perforations in contact
with fibrous material", Journal of the Acoustical Society of America, 119(5),
2785-2797 (2006)
154
References
[111] ISO 10534-2:1998: Acoustics – Determination of sound absorption coefficient

and impedance in impedance tubes – Part 2: Transfer-function method,
International Organisation for Standardization, Switzerland, 1998
[112] Munjal, M.L. and Doige, A.G. “The two-microphone method incorporating the
effects of mean flow and acoustic damping”, Journal of Sound and Vibration,
137(1), 135-138 (1990)
[113] Chung, J.Y. and Blaser, D.A. “Transfer function method of measuring in-duct
acoustic properties. II. Experiment,” Journal of the Acoustical Society of
America, 68, 914-921 (1980)
[114] Cummings, A. “Impedance tube measurements on porous media: The effect of
air-gaps around the sample”, Journal of Sound and Vibration, 151, 63–75
(1991)
[115] Kang, Y.J. and Bolton, J.S. “Finite element modelling of isotropic elastic porous
materials coupled with acoustical finite elements”, Journal of the Acoustical
Society of America, 98, 635–643 (1995)
[116] Song, B.H., Bolton, J.S. and Kang, Y.J. “Effect of circumferential edge
constraint on the acoustical properties of glass fibre materials”, Journal of the
Acoustical Society of America, 110, 2902–2916 (2001)
[117] ISO 9053:1991: Acoustics – Materials for acoustical applications –
Determination of airflow resistance, International Organisation for
Standardization, Switzerland, 1991
[118] Munjal, M.L., Venkatesham, B. and Tiwari, M. “Four-Pole Parameters of a
Rectangular Expansion Chamber with Yielding Walls”, Proceedings of the 16th
International Congress on Sound and Vibration, 5-9 July 2009, Krakow, Poland
[119] Svanberg, K. “Optimization of geometry in truss design”, Computer Methods in
Applied Mechanics and Engineering, 28(1), 63-80 (1981)
[120] Airaksinen, T. And Heikkola, E. Multiobjective muffler shape optimization with
hybrid acoustics modelling. In Reports of the Department of Mathematical
Information Technology - Series B. Scientific Computing, B 6/2010
[121] Oldenhuis, R.P.S. (2009) “Optimize (Version 3)” [MATLAB script], Delft
University of Technology, Netherlands. <http://www.mathworks.com>
Retrieved 13 April 2011
155
References
[122] Donckels, B. (2006) “PSO” [MATLAB script], Ghent University, Belgium.

< http://www.mathworks.com> Retrieved 13 April 2011
[123] Iqbal, M.A., Willson, T.K. and Thomas, R.J. The Control of Noise in Ventilation
Systems: A Designer's Guide, Atkins Research & Development Corp, Spon,
London, 1977
[124] Åbom, M. and Bodén, H. “Error analysis of two-microphone measurements in
ducts with flow”, Journal of the Acoustical Society of America, 83(6), 2429-
2438 (1988)
[125] Chartered Institution of Building Services, CIBS Guide B12: Sound Control,
HPC, Great Britain, 1972
[126] Barber, A. (ed.) Handbook of Noise and Vibration Control, 6th Ed., Elsevier
Advanced Technology, England, 1992
[127] Marks, T.M. “Explicit formulas for the calculation of regenerated noise in
ducts”, Proceedings of the 33rd International Congress and Exposition on Noise
Control Engineering, 22-25 August 2004, Prague, Czech Republic
[128] Granneman, J.H. and Jansen, R.P.M. “Pipe noise”, Hydrocarbon Engineering,
8(12), 49-53 (2003)
[129] Ver, I.L. “Prediction scheme for self generated noise of silencers”, Proceedings
of the 1972 International Conference on Noise Control Engineering, 4-6
October 1972, Washington, D.C.
[130] English, E.J. and Holland, K.R. “Aeroacoustic sound generation in simple
expansion chambers”, Journal of the Acoustical Society of America, 128(5),
2589-2595 (2010)
[131] Bouillard, Ph., Lacroix, V and De Bel, E. “A wave-oriented meshless
formulation for acoustical and vibro-acoustical applications”, Wave Motion, 39,
295–305 (2004)
[132] Herrin, D.W. and Seybert, A.F. Numerical Methods for Low Frequency HVAC
Noise Applications (1218-TRP), Department of Mechanical Engineering,
University of Kentucky, 2006
[133] Kinsler, L.E., Frey, A.R., Coppens, A.B. and Sanders, J.V. Fundamentals of
Acoustics, 4th Ed., John Wiley and Sons, New York, 2000
[134] Chung, J.Y. and Blaser, D.A., “Transfer Function Method of Measuring In-duct
Acoustic Properties, I: Theory”, Journal of the Acoustical Society of America,
68, 907-913 (1980)
156
References
[135] Chung, J.Y. and Blaser, D.A., “Transfer Function Method of Measuring In-duct
Acoustic Properties, II: Experiment”, Journal of the Acoustical Society of
America, 68, 914-921 (1980)
[136] Jones, P. and Kessissoglou, N. “Measurement and prediction of the acoustic
performance of mufflers for sleep apnoea devices”, Proceedings of the 16th
International Congress on Sound and Vibration, 5-9 July 2009, Krakow, Poland
[137] Munjal, M.L. and Doige A.G., “Theory of a Two Source-location Method for
Direct Experimental Evaluation of the Four-pole Parameters of an Aeroacoustic
Element,” Journal of Sound and Vibration, 141(2), 323-333 (1990)
[138] Nelder, J.A. and Mead, R. “A Simplex Method for Function Minimization”, The
Computer Journal, 7(4), 308-313 (1965)
[139] Whitley, D. “A genetic algorithm tutorial’, Statistics and Computing, 4(2), 65–
85 (1994)
[140] The MathWorks, Inc. (2009) “Ga – Revision 1.1.6.2” [MATLAB script],
MATLAB (R2010a)
[141] Kirkpatrick, S., Gelatt, C.D. and Vecchi, M.P. “Optimization by simulated
annealing”, Science, 220(4598), 671-680 (1983)
[142] The MathWorks, Inc. (2009) “Simulannealbnd – Revision 1.1.6.2” [MATLAB
script], MATLAB (R2010a)
[143] Kennedy, J. and Eberhart, R. "Particle Swarm Optimization". Proceedings of the
IEEE International Conference on Neural Networks, 27 Nov 1995 - 01 Dec
1995, Perth, Australia
[144] Clerc, M. and Kennedy, J. “The particle swarm - explosion, stability, and
convergence in a multidimensional complex space”, IEEE Transactions on
Evolutionary Computation, 6(1), 58-73 (2002)
157
Appendix A: Impedance Method Analysis of a Multiple Chamber Muffler
Appendix A
Impedance Method Analysis of a Multiple

Chamber Muffler
A multi-chamber muffler is considered which comprises three expansion chambers and

interconnecting ducts, as shown in Fig. A.1. The outlet pipe is considered to be
anechoically terminated. Development of the final equation for transmission loss follows
a similar process to that for the single chamber muffler by calculating the impedance and
pressure at each of the locations 1 to 12 as identified in Fig. A.1. With an anechoic
termination at the outlet duct, the specific acoustic impedance at point 12 is equal to the
characteristic impedance ሺ‫ݖ‬ଵଶ ൌ ߩܿሻ and the acoustic pressure at point 12 is equal to the
transmitted pressure (‫۾‬ଵଶ ൌ ‫۾‬௧ ).
ܵ஺ ܵ஻ ܵ஼ ܵ஽ ܵா ܵி ܵீ
‫ܘ‬௜ ‫ܘ‬௧
‫ܘ‬୰ ͳ ʹ ͵ Ͷ ͷ ͸ ͹ ͺ ͻ ͳͲ ͳͳ ͳʹ
‫ܮ‬஻ ‫ܮ‬஼ ‫ܮ‬஽ ‫ܮ‬ா ‫ܮ‬ி
Figure A.1 A multiple expansion chamber reactive muffler
The complex representation of the acoustic pressure and particle velocity are given by
࢖ሺ‫ݔ‬ǡ ‫ݐ‬ሻ ൌ ‫۾‬ା ݁ ௝ሺఠ௧ି௞௫ሻ ൅ ‫ ݁ ି۾‬௝ሺఠ௧ା௞௫ሻ (A.1)
‫۾‬ା ௝ሺఠ௧ି௞௫ሻ ‫ ି۾‬௝ሺఠ௧ା௞௫ሻ

࢛ሺ‫ݔ‬ǡ ‫ݐ‬ሻ ൌ ݁ െ ݁ (A.2)
ߩܿ ߩܿ
158
Impedance calculations proceed from the outlet (point 12) to the inlet (point 1). The
acoustic impedance at point 12 is given by
‫ݖ‬ଵଶ
ܼଵଶ ൌ (A.3)
ܵீ
The acoustic pressures and volume velocities at points 11 and 12 are equal; hence the
acoustic impedances at these points are equal ሺܼଵଵ ൌ ܼଵଶ ሻ. The specific acoustic
impedance at point 11 is then given by ‫ݖ‬ଵଵ ൌ ܼଵଵ ܵி or
ܵி
‫ݖ‬ଵଵ ൌ ‫ݖ‬ଵଶ ൬ ൰ (A.4)
ܵீ
considering the impedance formula for undamped plane acoustic waves in a gas column.
At‫ݔ‬ி ൌ ‫ܮ‬ி , corresponding to point 11 on Fig. A.1, the specific acoustic impedance is
given by
࢖ଵଵ ‫۾‬ிା ݁ ௝ሺఠ௧ି௞௅ಷሻ ൅ ‫۾‬ிି ݁ ௝ሺఠ௧ା௞௅ಷሻ

‫ݖ‬ଵଵ ൌ ൌ ߩܿ ቆ ቇ (A.5)
࢛ଵଵ ‫۾‬ிା ݁ ௝ሺఠ௧ି௞௅ಷሻ െ ‫۾‬ிି ݁ ௝ሺఠ௧ା௞௅ಷሻ
which can be rearranged to give
ߩܿ
ቀͳ െ ‫ ݖ‬ቁ
ଵଵ ି௝ଶ௞௅ಷ
‫۾‬ிି ൌ ‫۾‬ிା ߩܿ ݁ (A.6)
ଵଵ
Similarly, at ‫ݔ‬ி ൌ Ͳ, corresponding to point 10 on Fig. A.1, the specific acoustic

impedance can be found as
࢖ଵ଴ ‫۾‬ிା ൅ ‫۾‬ிି

‫ݖ‬ଵ଴ ൌ ൌ ߩܿ ൬ ൰ (A.7)
࢛ଵ଴ ‫۾‬ிା െ ‫۾‬ிି
159
Substitution of Eq. (A.6) into Eq. (A.7) yields
ߩܿ ௝௞௅ಷ ߩܿ ି௝௞௅ಷ
ቀͳ ൅ ቁ݁ ൅ ቀͳ െ ቁ݁
‫ݖ‬ଵଵ ‫ݖ‬ଵଵ
‫ݖ‬ଵ଴ ൌ ߩܿ ቎ ߩܿ ௝௞௅ಷ ߩܿ ି௝௞௅ಷ ቏ (A.8)
ቀͳ ൅ ቁ݁ െ ቀͳ െ ቁ݁
‫ݖ‬ଵଵ ‫ݖ‬ଵଵ
and finally, substitution of Eq. (A.4) into Eq. (A.8) yields
ߩܿ ܵ ߩܿ ܵ
ቀͳ ൅ ‫ ீܵ ݖ‬ቁ ݁ ௝௞௅ಷ ൅ ቀͳ െ ‫ ீܵ ݖ‬ቁ ݁ ି௝௞௅ಷ
ଵଶ ி ଵଶ ி
‫ݖ‬ଵ଴ ൌ ߩܿ ൦ ൪ (A.9)
ߩܿ ܵீ ௝௞௅ಷ ߩܿ ܵீ ି௝௞௅ಷ
ቀͳ ൅ ‫ ܵ ݖ‬ቁ ݁ െ ቀͳ െ ‫ ܵ ݖ‬ቁ ݁
ଵଶ ி ଵଶ ி
The specific acoustic impedance at point 10 is also given by ‫ݖ‬ଵ଴ ൌ ܼଵ଴ ܵி . The acoustic
pressure and volume velocity at points 9 and 10 are equal; hence the acoustic impedances
at these points are equal ሺܼଽ ൌ ܼଵ଴ ሻ. The specific acoustic impedance at point 9 can then
be obtained as
ߩܿ ܵீ ௝௞௅ಷ ߩܿ ܵ
ܵா ቀͳ ൅ ‫ݖ‬ଵଶ ܵி ቁ ݁ ൅ ቀͳ െ ‫ ீܵ ݖ‬ቁ ݁ ି௝௞௅ಷ
ଵଶ ி
‫ݖ‬ଽ ൌ ߩܿ ൬ ൰ ൦ ൪ (A.10)
ܵி ቀͳ ൅ ߩܿ ܵீ ቁ ݁ ௝௞௅ಷ െ ቀͳ െ ߩܿ ܵீ ቁ ݁ ି௝௞௅ಷ
‫ݖ‬ଵଶ ܵி ‫ݖ‬ଵଶ ܵி
The specific acoustic impedance at point 8 can be found from that at point 9 by following
the derivation of Eq. (A.8) to show that
ߩܿ ߩܿ
ቀͳ ൅ ‫ ݖ‬ቁ ݁ ௝௞௅ಶ ൅ ቀͳ െ ቁ ݁ ି௝௞௅ಶ
ଽ ‫ݖ‬ଽ
‫ ଼ݖ‬ൌ ߩܿ ቎ ߩܿ ௝௞௅ಶ ߩܿ ି௝௞௅ಶ ቏ (A.11)
ቀͳ ൅ ‫ ݖ‬ቁ ݁ െ ቀͳ െ ‫ ݖ‬ቁ ݁
ଽ ଽ
replicating the process that was applied to the terminal expansion chamber.
160
Following the derivation of Eq. (A.10) it can be shown that
ߩܿ ܵா ௝௞௅ವ ߩܿ ܵா ି௝௞௅ವ
ܵ஼ ቀͳ ൅ ‫ܵ ଼ݖ‬஽ ቁ ݁ ൅ ቀͳ െ
‫ܵ ଼ݖ‬஽
ቁ݁
‫ݖ‬ହ ൌ ߩܿ ൬ ൰ ൦ ൪ (A.12)
ܵ஽ ቀͳ ൅ ߩܿ ܵா ቁ ݁ ௝௞௅ವ െ ቀͳ െ ߩܿ ܵா ቁ ݁ ି௝௞௅ವ
‫ܵ ଼ݖ‬஽ ‫ܵ ଼ݖ‬஽
The specific acoustic impedance at point 4 can be found from that at point 5 by following
the derivation of Eq. (A.8) to show that
ߩܿ ߩܿ
ቀͳ ൅ ‫ ݖ‬ቁ ݁ ௝௞௅಴ ൅ ቀͳ െ ‫ ݖ‬ቁ ݁ ି௝௞௅಴
ହ ହ
‫ݖ‬ସ ൌ ߩܿ ቎ ߩܿ ߩܿ ቏ (A.13)
ቀͳ ൅ ‫ ݖ‬ቁ ݁ ௝௞௅಴ െ ቀͳ െ ‫ ݖ‬ቁ ݁ ି௝௞௅಴
ହ ହ
replicating the process that was applied to the terminal expansion chamber. Following
the derivation of Eq. (A.10) it can be shown that
ߩܿ ܵ஼ ௝௞௅ಳ ߩܿ ܵ
ܵ஺ ቀͳ ൅ ‫ݖ‬ସ ܵ஻ ቁ ݁ ൅ ቀͳ െ ‫ܵ ݖ‬஼ ቁ ݁ ି௝௞௅ಳ
ସ ஻
‫ݖ‬ଵ ൌ ߩܿ ൬ ൰ ൦ ൪ (A.14)
ܵ஻ ቀͳ ൅ ߩܿ ܵ஼ ቁ ݁ ௝௞௅ಳ െ ቀͳ െ ߩܿ ܵ஼ ቁ ݁ ି௝௞௅ಳ
‫ݖ‬ସ ܵ஻ ‫ݖ‬ସ ܵ஻
Applying the anechoic termination boundary condition ሺ‫ݖ‬ଵଶ ൌ ߩܿሻ it is now possible to
use Eqs. (A.10) to (A.14) to calculate ‫ݖ‬ଽ ǡ ‫ ଼ݖ‬ǡ ‫ݖ‬ହ ǡ ‫ݖ‬ସ and ‫ݖ‬ଵ .
The acoustic pressures at points 1 to 12 must now be found. The complex pressure at
‫ݔ‬஻ ൌ Ͳ is ‫۾‬ଵ ൌ ‫۾‬௜ ൅ ‫۾‬௥ and the specific acoustic impedance at point 1 is
࢖ଵ ‫۾‬௜ ൅ ‫۾‬௥
‫ݖ‬ଵ ൌ ൌ ߩܿ ൬ ൰ (A.15)
࢛ଵ ‫۾‬௜ െ ‫۾‬௥
The acoustic pressure at point 1 can then be obtained as
ʹ‫۾‬௜
‫۾‬ଵ ൌ ߩܿ (A.16)
ଵ
161
Equality of the acoustic pressures at points 1 and 2 leads to ‫۾‬ଶ ൌ ‫۾‬ଵ. The complex
pressures at point 2 ሺ‫ݔ‬஻ ൌ Ͳሻ and point 3 ሺ‫ݔ‬஻ ൌ ‫ܮ‬஻ ሻ can be obtained by applying
Eq. (A.1) to show
‫۾‬ଶ ൌ ‫۾‬୆ା ൅ ‫۾‬୆ି (A.17)
‫۾‬ଷ ݁ ି௝௞௅ಳ ൌ ‫۾‬୆ା ݁ ି௝௞௅ಳ ൅ ‫۾‬୆ି ݁ ௝௞௅ಳ (A.18)
At point 2 ሺ‫ݔ‬஻ ൌ Ͳሻ, Eqs. (A.1) and (A.2) can be used to show that the specific acoustic
impedance is
࢖ଶ ‫۾‬஻ା ൅ ‫۾‬஻ି
‫ݖ‬ଶ ൌ ൌ ߩܿ ൬ ൰ (A.19)
࢛ଶ ‫۾‬஻ା െ ‫۾‬஻ି
Rearranging Eqs. (A.17) and (A.18) and substituting into Eq. (A.19), the acoustic
pressure at point 3 is obtained as
‫۾‬ଶ ߩܿ ߩܿ
‫۾‬ଷ ൌ ൤൬ͳ ൅ ൰ ൅ ൬ͳ െ ൰ ݁ ௝ଶ௞௅ಳ ൨ (A.20)
ʹ ‫ݖ‬ଶ ‫ݖ‬ଶ
From equality of the acoustic pressures at points 1 and 2 ሺ‫۾‬ଵ ൌ ‫۾‬ଶ ሻ and at points 3 and 4
ௌ
ሺ‫۾‬ଷ ൌ ‫۾‬ସ ሻ, and recalling that ‫ݖ‬ଶ ൌ ‫ݖ‬ଵ ቀ ಳ ቁ, Eq. (A.20) can be rewritten as
ௌ ಲ
‫۾‬ଵ ܵ஺ ߩܿ ܵ஺ ߩܿ ௝ଶ௞௅
‫۾‬ସ ൌ ൤൬ͳ ൅ ൰ ൅ ൬ͳ െ ൰݁ ಳ൨ (A.21)
ʹ ܵ஻ ‫ݖ‬ଵ ܵ஻ ‫ݖ‬ଵ
The acoustic pressure at point 5 can be found from that at point 4 by following the
derivation process of Eq. (A.20) to show that
‫۾‬ସ ߩܿ ߩܿ
‫۾‬ହ ൌ ൤൬ͳ ൅ ൰ ൅ ൬ͳ െ ൰ ݁ ௝ଶ௞௅಴ ൨ (A.22)
ʹ ‫ݖ‬ସ ‫ݖ‬ସ
The acoustic pressure at point 8 can be found from that at point 5 by replicating the
process that was applied to the initial expansion chamber.
162
Following the derivation of Eq. (A.21) it can be shown that
‫۾‬ହ ܵ஼ ߩܿ ܵ஼ ߩܿ ௝ଶ௞௅
‫ ଼۾‬ൌ ൤൬ͳ ൅ ൰ ൅ ൬ͳ െ ൰݁ ವ൨ (A.23)
ʹ ܵ஽ ‫ݖ‬ହ ܵ஽ ‫ݖ‬ହ
The acoustic pressure at point 9 can be found from that at point 8 by following the
derivation process of Eq. (A.20) to show that
‫଼۾‬ ߩܿ ߩܿ
‫۾‬ଽ ൌ ൤൬ͳ ൅ ൰ ൅ ൬ͳ െ ൰ ݁ ௝ଶ௞௅ಶ ൨ (A.24)
ʹ ‫଼ݖ‬ ‫଼ݖ‬
The acoustic pressure at point 12 can be found from that at point 9 by replicating the
process that was applied to the initial expansion chamber. Following the derivation of
Eq. (A.21) it can be shown that
‫۾‬ଽ ܵா ߩܿ ܵா ߩܿ ௝ଶ௞௅
‫۾‬ଵଶ ൌ ൤൬ͳ ൅ ൰ ൅ ൬ͳ െ ൰݁ ಷ൨ (A.25)
ʹ ܵி ‫ݖ‬ଽ ܵி ‫ݖ‬ଽ
Applying the anechoic termination boundary condition gives ‫۾‬ଵଶ ൌ ‫۾‬௧ and it is now
possible to use Eq. (A.16) and Eqs. (A.21) to (A.25) and the specific impedances
‫۾‬೟
obtained previously ሺ‫ݖ‬ଽ ǡ ‫ ଼ݖ‬ǡ ‫ݖ‬ହ ǡ ‫ݖ‬ସ ǡ ‫ݖ‬ଵ ሻ to calculate . The sound transmission loss can be
‫۾‬೔
found using
ȁ‫۾‬௧ଶ ȁ ܵீ
ܶ‫ ܮ‬ൌ െͳͲ ଵ଴ ቈ ଶ ቉ (A.26)
ห‫۾‬௜ ห ܵ஺
163
Appendix B: Incorporating Revised Coefficients into Original Delany-Bazley Functions
Appendix B
Incorporating Revised Coefficients into Original

Delany-Bazley Functions
The Delany-Bazley functions can be expressed in terms of coefficients ‫ܥ‬ଵ to ‫ ଼ܥ‬by
஼మ ஼
ߩ௔ ݂ ߩ௔ ݂ ర
ܼ௖ ൌ ߩ௔ ܿ௔ ൥ͳ ൅ ‫ܥ‬ଵ ቆ ቇ െ ݆‫ܥ‬ଷ ቆ ቇ ൩ (B.1)
‫ݎ‬௙ ‫ݎ‬௙
஼ల ஼
ʹߨ݂ ߩ௔ ݂ ߩ௔ ݂ ఴ
ߛൌ൬ ൰ ൥‫ܥ‬ହ ቆ ቇ ൅ ݆ ൭ͳ ൅ ‫ ଻ܥ‬ቆ ቇ ൱൩ (B.2)
ܿ௔ ‫ݎ‬௙ ‫ݎ‬௙
A simplified approach can facilitate the acoustic characterisation of porous materials by

enabling the coefficients in the Delany-Bazley functions to be determined without the
requirement to obtain the airflow resistivity. In this approach the characteristic
impedance and propagation constant are expressed in terms of the original coefficients
‫ܥ‬ଶ ǡ ‫ܥ‬ସ ǡ ‫ ଺ܥ‬ǡ ‫ ଼ܥ‬and new coefficients ܾோ ǡ ܾ௑ ǡ ܾఈ ǡ ܾఉ as
ܼ௖ ൌ ߩ௔ ܿ௔ ሾͳ ൅ ͳͲ௕ೃ ݂ ஼మ െ ݆ͳͲ௕೉ ݂ ஼ర ሿ (B.3)
ʹߨ݂
ߛൌ൬ ൰ ൣͳͲ௕ഀ ݂ ஼ల ൅ ݆൫ͳ ൅ ͳͲ௕ഁ ݂ ஼ఴ ൯൧ (B.4)
ܿ௔
The coefficients ‫ܥ‬ଶ ǡ ‫ܥ‬ସ ǡ ‫ ଺ܥ‬ǡ ‫ ଼ܥ‬obtained using the revised approach may be used directly
with the original Delany-Bazley model (Eqs. (B.1) and (B.2)).
164
If the airflow resistivity of the material is known, the coefficients ‫ܥ‬ଵ ǡ ‫ܥ‬ଷ ǡ ‫ܥ‬ହ ǡ ‫ ଻ܥ‬used in
the original model can be readily calculated from the y-intercepts ܾோ ǡ ܾ௑ ǡ ܾఈ ǡ ܾఉ using
‫ݎ‬௙ ஼మ ‫ݎ‬௙ ஼ర
‫ܥ‬ଵ ൌ ͳͲ௕ೃ ൬ ൰ ‫ܥ‬ଷ ൌ ͳͲ௕೉ ൬ ൰ (B.5, B.6)
ߩ௔ ߩ௔
‫ݎ‬௙ ஼ల ‫ݎ‬௙ ஼ఴ
‫ܥ‬ହ ൌ ͳͲ௕ഀ ൬ ൰ ‫ ଻ܥ‬ൌ ͳͲ௕ഁ ൬ ൰ (B.7, B.8)
ߩ௔ ߩ௔
In the absence of the flow resistivity it is still possible to use the original form of the
model by selecting an arbitrary value ‫ݎ‬௙ᇱ to enable the calculation of resistivity-free
coefficients ‫ܥ‬ଵᇱ ǡ ‫ܥ‬ଷᇱ ǡ ‫ܥ‬ହᇱ ǡ ‫଻ܥ‬ᇱ using Eqs. (B.5) to (B.8). These resistivity-free coefficients can
then be used in place of the original coefficients ‫ܥ‬ଵ ǡ ‫ܥ‬ଷ ǡ ‫ܥ‬ହ ǡ ‫ ଻ܥ‬in Eqs. (B.1) and (B.2) as
long as they are only used in conjunction with the nominated arbitrary airflow resistivity
‫ݎ‬௙ᇱ . It is important to realise that the resistivity-free coefficient values cannot be directly
compared to the corresponding original Delany-Bazley coefficients.
The magnitude of the bracketed term ሺ‫ݎ‬୤ Τߩ௔ ሻ contained in Eqs. (B.5) to (B.8) can be
large (typically in the range 103 – 105 for polyurethane foams), making the calculation of
‫ܥ‬ଵ ǡ ‫ܥ‬ଷ ǡ ‫ܥ‬ହ ǡ ‫ ଻ܥ‬very sensitive to the precision of exponents ‫ܥ‬ଶ ǡ ‫ܥ‬ସ ǡ ‫ ଺ܥ‬ǡ ‫ ଼ܥ‬. This sensitivity
may be eliminated by setting ‫ݎ‬௙ᇱ ൌ ߩ௔ as the arbitrary value for airflow resistivity.
Equations (B.5) to (B.8) then simplify to become:
‫ܥ‬ଵƍ ൌ ͳͲ௕ೃ ‫ܥ‬ଷƍ ൌ ͳͲ௕೉ (B.9, B.10)
‫ܥ‬ହƍ ൌ ͳͲ௕‫ן‬ ‫଻ܥ‬ƍ ൌ ͳͲ௕ഁ (B.11, B.12)
The relationships required to use the revised coefficients in the original form of the
Delany-Bazley model, where the airflow resistivity is unknown, are summarised in
Table B.1.
165
Table B.1 Relationships for use of the revised model coefficients in the original
Delany-Bazley functions
Original ‫ݎ‬୤ ‫ܥ‬ଵ ‫ܥ‬ଶ ‫ܥ‬ଷ ‫ܥ‬ସ ‫ܥ‬ହ ‫଺ܥ‬ ‫଻ܥ‬ ‫଼ܥ‬
Use ߩ௔ ͳͲ௕ೃ ‫ܥ‬ଶ ͳͲ௕೉ ‫ܥ‬ସ ͳͲ௕‫ן‬ ‫଺ܥ‬ ͳͲ௕ഁ ‫଼ܥ‬
The revised form of the Delany-Bazley coefficients was calculated for the light grey and
dark grey foam materials, introduced in Chapter 3, and are shown in Table B.2. The
airflow resistivity of these foam types was measured and found to be 8,445 Rayls/m and
2,652 Rayls/m respectively. Assuming an air density of 1.18 kg/m3, these airflow
resistivity values can be used with Eqs. (B.5) to (B.8) to convert the coefficients
summarised in Table B.2 into a set of coefficients that are consistent with the original
Delany and Bazley equations. These results are presented in Table B.3.
Table B.2 Delany-Bazley function coefficients for the light and dark grey foams
(revised form)
Dark grey Light grey

Parameters
foam foam
ܾோ 0.0595 0.8597
ܴ ൌ ሺܼ௖ ሻ
‫ܥ‬ଶ -0.2249 -0.3659
ܾ௑ 0.6822 1.3567
ܺ ൌ ሺܼ௖ ሻ
‫ܥ‬ସ -0.4851 -0.6144
ܾఈ 1.1095 1.4337
ߙ ൌ ሺߛሻ
‫ ଺ܥ‬ -0.5416 -0.5728
ܾఉ 0.4634 1.2015
ߚ ൌ ሺߛሻ
‫ ଼ܥ‬ -0.3111 -0.4657
166
Table B.3 Delany-Bazley function coefficients for the light and dark grey foams
(derived)
Dark grey Light grey

Parameters
foam foam
‫ܥ‬ଵ 0.2022 0.2814

ܴ ൌ ሺܼ௖ ሻ
‫ܥ‬ଶ -0.2249 -0.3659
‫ܥ‬ଷ 0.1138 0.0974

ܺ ൌ ሺܼ௖ ሻ
‫ܥ‬ସ -0.4851 -0.6144
‫ܥ‬ହ 0.1969 0.1682

ߙ ൌ ሺߛሻ
‫ ଺ܥ‬ -0.5416 -0.5728
‫ ଻ܥ‬ 0.2634 0.2549

ߚ ൌ ሺߛሻ
‫ ଼ܥ‬ -0.3111 -0.4657
167
Appendix C: Optimisation Algorithm Assessment
Appendix C
Optimisation Algorithm Assessment
Four metaheuristic optimisation algorithms were assessed for use in optimising the
acoustic performance of mufflers in this thesis. The algorithms consisted of the Nelder-
Mead algorithm, genetic algorithm, simulated annealing and particle swarm optimisation
approaches. The relative performance of each algorithm is compared by conducting
parametric optimisation of two muffler designs. The first design is a cylindrical
expansion chamber reactive muffler which is optimised for acoustic performance by
varying the diameter and length of the chamber within pre-determined bounds. The
second design consists of a similar cylindrical expansion chamber muffler internally
lined with polyurethane foam on the chamber wall. The geometry of the muffler remains
fixed and the design is optimised by varying the thickness and length of the foam insert
within pre-determined bounds.
C.1 Optimisation Algorithms
The four optimisation approaches considered here do not rely on gradient information
and are also commonly referred to as being stochastic, derivative-free, or zero order. By
making no assumptions about the problem being solved, these methods are capable of
performing a very wide search of the feasible design space. Although less susceptible to
becoming trapped by a local minima than gradient-based methods, the metaheuristic
approach is still unable to guarantee that the converged solution is the optimal solution.
MATLAB scripts, either as native MATLAB functions or in the form of stand-alone

functions written by third parties, are readily available which implement each of these
approaches [121, 122, 140, 142]. The optimisation methodology described in Chapter 5
is formulated such that it can call any of these functions to manage the iterative stage of
the optimisation.
168
C.1.1 “Optimize” / Nelder-Mead
The Nelder-Mead algorithm within the “Optimize” function can be used to optimise
linear and non-linear problems and builds on the native MATLAB “fminsearch”
algorithm by adding constraints (equality or inequality) to the feasible domain [121]. It
uses the Nelder-Mead downhill simplex algorithm which is a direct search method that
does not rely on numerical or analytical gradients [138]. Although not gradient-based, the
Optimize function is sensitive to the choice of starting values assigned to the variables
and may converge to non-global minima.
C.1.2 Genetic algorithm
Based on the theory of evolution, the genetic algorithm (GA) proceeds by forming an
initial population of possible solutions selected, often randomly, from the feasible
solution domain [139]. The fitness of the population is evaluated using a defined
objective function and the population is improved through an iterative process of
selection, mutation and crossover. The algorithm terminates when a satisfactory fitness
level has been attained but may also be halted using criteria such as reaching a maximum
number of generations. The genetic algorithm used in this work was the standard
MATLAB “ga” function [140].
C.1.3 Simulated annealing
Taking its name from the process of metallurgical annealing, this approach is an
adaptation of a Monte Carlo method used to generate sample states of a thermodynamic
system [141]. The simulated annealing (SA) approach commences with a vector of initial
values for the design control variables. At each step of the optimisation, the current
solution is replaced by a random solution whose probability of being chosen depends on
(i) the difference between the corresponding objective function values and (ii) a global
“temperature” parameter that is gradually decreased during the process. The combined
effect of these dependencies is to produce subsequent solutions which are initially almost
random but tend towards a gradient-based result as the temperature is reduced. At pre-
determined intervals, the temperature is elevated to its maximum value and the cooling
process is repeated. This cyclic approach provides the opportunity for the solution to free
itself from a local minimum. The simulated annealing algorithm considered here was the
standard MATLAB “simulannealbnd” function [142].
169
C.1.4 Particle swarm optimisation
The particle swarm optimisation (PSO) approach is based on the collective intelligence
of a large group of intercommunicating entities [143]. The PSO approach commences
with a population (the swarm) of candidate solutions (the particles). Each individual
particle has a current location, current fitness value, a best location based on a prior
optimum fitness value, and knowledge of the best location encountered by other
members of the swarm. The problem is optimised by moving the particles within the
feasible solution domain based on this data, which is continually updated as locations
with better objective function values are found. The particle swarm optimisation
algorithm that was assessed in the current work is based on the PSO algorithm originally
introduced by Kennedy and Eberhart [143] and incorporating a constriction factor as
introduced by Clerc and Kennedy [122, 144].
C.2 Reactive Expansion Chamber Muffler
The cylindrical expansion chamber muffler geometry is shown in Fig. C.1 and is
assigned initial dimensions of ݀௖ = 60 mm, ‫ܮ‬௖ = 250 mm, ݀௜௢ = 20 mm and ‫ܮ‬௜௢ = 20 mm.
The diameter ݀௖ and length ‫ܮ‬௖ of the chamber are constrained to a range between 45 mm
and 75 mm (±15 mm from the initial design) and 205 mm and 295 mm (±45 mm from
the initial design), respectively. These constraints ensure that the feasible design space
contains only one transmission loss maximum. The diameter ݀௜௢ and length ‫ܮ‬௜௢ of the
inlet and outlet ducts remain fixed.
A fitness landscape was initially created by stepping the change in diameter ο݀ and
change in length ο‫ ܮ‬in integer increments across their respective bounded ranges and
using COMSOL to calculate the transmission loss at an excitation frequency of 1 kHz.
From Fig. C.2 it can be seen that the topology is smooth and that only one minimum is
present in the bounded domain.
The optimisation was performed at an excitation frequency of 1 kHz. All solvers were
run with the same settings (e.g. convergence tolerance) where possible and using default
values for the remaining settings. Improved performance from each of the algorithms can
be expected by refining these settings however this is beyond the scope of the current
work.
170
Maximum size
ȟ݀Ȁʹ ȟ‫ܮ‬
Minimum size
݀௜௢ ݀௖ ݀௜௢
‫ܮ‬௜௢ ‫ܮ‬௖ ‫ܮ‬௜௢
Figure C.1 A cylindrical expansion chamber reactive muffler showing the

optimisation domain
Figure C.2 Fitness landscape for the cylindrical expansion chamber reactive muffler
at 1 kHz excitation frequency
171
The optimisation results are presented in Table C.1. All four algorithms converged on the
same result and are in close agreement with the theoretical values that are predicted using
Eqs. (5.8) and (5.9). The convergence paths and/or intermediate populations for each of
the solutions are overlaid on the fitness landscape in Figs. C.3 to C.6. The convergence
path of the Optimize solver corresponding to the Nelder-Mead algorithm shows that
while it is not a gradient-based solver per se, the convergence behaviour is influenced by
the gradient of the fitness landscape. The Optimize solver achieved convergence in
significantly less time than any of the other solvers. The wide distribution of intermediate
populations over the fitness landscape for the GA, SA and PSO solvers shows that these
solvers are all capable of thoroughly exploring the feasible solution domain during the
optimisation process. The highly scattered distribution of intermediate populations
produced by the SA solver initially appears to be the least efficient however this is not
supported by the normalised convergence times shown in Table C.1. The scattering is
attributed to the regular “re-heating” cycles that are intended to free the solution from
local minima. The GA was most efficient while the PSO algorithm was least efficient.
Table C.1 Optimisation results for the cylindrical expansion chamber reactive
muffler, comparing four optimisation algorithms
Theoretical
Optimize GA SA PSO
values
ܶ‫( ܮ‬dB) 17.0 17.1 17.1 17.1 17.1
ο݀ (mm) 15.0 15.0 15.0 15.0 15.0
ο‫( ܮ‬mm) 7.4 8.2 8.2 8.2 8.2
Normalised time --- 0.16 1.00 1.29 1.42
Optimised? --- YES YES YES YES
172
Figure C.3 Fitness landscape with objective function (-TL) values assessed by the
Optimize solver during convergence – cylindrical expansion chamber
reactive muffler (magenta dot = start; white dot = convergence)
173
simulated annealing solver during convergence – cylindrical expansion
particle swarm optimisation solver during convergence – cylindrical
expansion chamber reactive muffler (white dot = convergence)
174
C.3 Expansion Chamber Muffler with Foam Lining
A similar cylindrical expansion chamber muffler now lined internally with foam on the
chamber wall is shown in Fig. C.7. The muffler was assigned the same initial dimensions
that were applied to the reactive muffler and these remained fixed throughout the
optimisation. The region occupied by the lining was specified by the foam thickness ‫ݐ‬௙
and the length ‫ܮ‬௙ that the foam lining extends along the expansion chamber, commencing
at the chamber inlet. The foam has the same acoustic properties as the light grey foam
that was characterised in Chapter 3. No foam was applied to the inlet and outlet ducts.
The design optimisation was bounded by constraining the foam thickness to a range
between 5 mm and 25 mm and the length to a range between 5 mm and 245 mm. These
constraints were selected so that the foam insert could never collapse to zero volume, or
fully occupy the chamber volume thus causing the unobstructed region through the centre
of the muffler to collapse to zero volume. Both geometric extremes risk the generation of
poor quality finite element mesh and unreliable predictions.
‫ܮ‬௙
Minimum size ‫ݐ‬௙

Foam
݀௜௢ ݀௜௢
Maximum size
݀௖
‫ܮ‬௜௢ ‫ܮ‬௖ ‫ܮ‬௜௢
Figure C.7 A cylindrical expansion chamber muffler with foam lining showing the
optimisation domain
A fitness landscape was created by stepping the foam thickness and length in increments
of 5 mm and 25 mm respectively across their bounded ranges. The fitness values were
expanded to consider the case where the foam insert was absent (‫ݐ‬௙ = 0), and for
increments of length where the foam insert occupied the full muffler cross-section,
175
corresponding to ‫ݐ‬௙ = 30 mm. The additional values were only used to assist with
visualisation of the fitness landscape as they do not fall within the feasible design space.
From Fig. C.8 it can be seen that the topology is smooth and that the global minimum is
clearly defined.
Figure C.8 Fitness landscape for the cylindrical expansion chamber muffler with
foam lining at 1 kHz excitation frequency
The optimisation was performed at an excitation frequency of 1 kHz. All solvers were
run with the same settings where possible and using default values for the remaining
settings. The optimisation results are presented in Table C.2. The convergence paths
and/or intermediate populations for each of the solutions are overlaid on the fitness
landscape in Figs. C.9 to C.12. The genetic algorithm and the simulated annealing and
particle swarm optimisation approaches converged with similar results. Although the
Optimize solver achieved convergence in significantly less time than any of the other
solvers, it failed to locate the global minimum and converged instead at a local minimum.
The permitted range of the two control variables encloses two maxima and two minima
within the feasible design space (indicated by the black box on Figs. C.9 to C.12). The
topology of the fitness domain is also such that a gradient-influenced solver can become
176
trapped in a local minimum. The fitness domain topology is expected to become

increasingly complex as more elaborate muffler designs are considered and as a greater
number of control variables are introduced into the optimisation problem. Hence, the
Optimize algorithm is considered to be unsuited to parametric optimisation of practical
muffler designs.
The genetic algorithm, simulated annealing and particle swarm optimisation approaches
all performed a wide search of the feasible design space for both muffler optimisations
and converged with similar results. The genetic algorithm required 15-20% less time to
converge than the simulated annealing and particle swarm optimisation approaches. As
the genetic algorithm was also observed to be more computationally robust than the SA
and PSO approaches, the genetic algorithm was selected for use in optimising the
acoustic performance of mufflers in this thesis (see Chapter 5).
Table C.2 Optimisation results for the cylindrical expansion chamber muffler with
foam lining, comparing four optimisation algorithms
Optimize GA SA PSO
ܶ‫( ܮ‬dB) 13.8 25.0 25.2 25.2
Thickness(mm) 25.0 25.0 25.0 25.0
Length (mm) 11.8 237.9 245.0 245.0
Normalised time 0.13 1.00 > 2.08 a 1.19
Optimised? NO YES YES YES
a
Solution was manually terminated as practical convergence had been attained.
177
Figure C.9 Fitness landscape with objective function values (-TL) assessed by the
Optimize solver during convergence – cylindrical expansion chamber
muffler with foam lining (green dot = start; white dot = convergence)
178
simulated annealing solver during convergence – cylindrical expansion
particle swarm optimisation solver during convergence – cylindrical
expansion chamber muffler with foam lining (white dot = convergence)
179
Publications Arising from this Thesis
Publications Arising from this Thesis
[1] Jones, P. and Kessissoglou, N. “Computational optimisation of the acoustic

performance of mufflers for sleep apnoea devices”, Proceedings of Acoustics
2011, 2-4 November 2011, Gold Coast, Australia.
performance of poroelastic foam filled mufflers for sleep apnoea devices”,
Proceedings of the 20th International Congress on Acoustics, 23-27 August
2010, Sydney, Australia.
[3] Jones, P.W. “Prediction of the acoustic performance of small poroelastic foam
filled mufflers: A case study”, Acoustics Australia, 38(2), 73-79 (2010).
performance of poroelastic foam filled mufflers for sleep apnoea devices”,
Proceedings of the 17th International Congress on Sound and Vibration, 18-22
July 2010, Cairo, Egypt.
[5] Jones, P.W. and Kessissoglou, N.J. “A numerical and experimental study of the
transmission loss of mufflers used in respiratory medical devices”, Acoustics
Australia, 38(1), 13-19 (2010).
[6] Jones, P. and Kessissoglou, N. “A comparison of current commercial acoustic
FEA software for small complex muffler geometries: Prediction vs.
Experiment”, Proceedings of Acoustics 2009, 23-25 November 2009, Adelaide,
Australia.
[7] Jones, P.W. and Kessissoglou, N.J. “Transmission loss of mufflers for sleep
apnoea devices, Proceedings of Inter-noise 2009, 23-26 August 2009, Ottawa,
Canada.
performance of mufflers for sleep apnoea devices”, Proceedings of the 16th
International Congress on Sound and Vibration, 5-9 July 2009, Krakow,
Poland.
[9] Jones, P. and Kessissoglou, N. “Prediction of the acoustic performance of
mufflers for sleep apnoea devices: Preliminary study”, Proceedings of the 15th
International Congress on Sound and Vibration, 6-10 July 2008, Daejeon,
Korea.
180

Whole

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Whole

Uploaded by

Copyright:

Available Formats

Prediction and optimisation of the acoustic performance of

mufflers for sleep apnoea devices

Downloaded from http://hdl.handle.net/1959.4/51649 in https://

Peter William Jones

A thesis submitted in fulfilment

of the requirements for the degree of

Schoool of Mechanical and Manufacturing Engineeering

T University of New South Wales, Australia

Surname or Family name: JONES

First name: PETER Other name/s: WILLIAM

Abbreviation for degree as given in the University calendar: PHD

School: MECHANICAL AND MANUFACTURING ENGINEERING Faculty: ENGINEERING

Declaration relating to disposition of project thesis/dissertation

Signature Witness Date

FOR OFFICE USE ONLY Date of completion of requirements for Award:

THIS SHEET IS TO BE GLUED TO THE INSIDE FRONT COVER OF THE THESIS

Obstructive sleep apnoea affects approximately 10-25% of the adult population. It is

The experimental component of the thesis includes measurement of the acoustic

In the final part of the thesis, a methodology to efficiently optimise muffler

Originality Statement ..................................................................................................... ii

Copyright Statement ...................................................................................................... iii

Authenticity Statement .................................................................................................. iii

Abstract .................................................................................................................. iv

Acknowledgements ........................................................................................................ vi

Table of Contents .......................................................................................................... vii

List of Figures ................................................................................................................. xi

List of Tables ................................................................................................................ xxi

List of Symbols ........................................................................................................... xxiii

Chapter 1 Introduction and Literature Review...................................................... 1

1.1 Introduction ............................................................................................... 1

1.2 Obstructive Sleep Apnoea ......................................................................... 1

1.3 Background on Mufflers ........................................................................... 3

1.4 CPAP Device Mufflers .............................................................................. 5

1.5 Literature Review ...................................................................................... 6

1.5.1 Analytical approaches ................................................................. 7

1.5.2 Numerical approaches ................................................................. 9

1.5.3 Porous materials ........................................................................ 11

1.5.4 Design optimisation .................................................................. 13

1.6 Contribution to Research ......................................................................... 16

1.7 Thesis Overview ...................................................................................... 18

2.1 Introduction ............................................................................................. 20

2.2 Analytical Methods ................................................................................. 20

2.2.1 Single chamber muffler ............................................................. 21

2.2.2 Multiple chamber muffler ......................................................... 30

2.3 Computational Approaches ..................................................................... 33

2.3.1 Direct approach ......................................................................... 35

2.3.2 Indirect approach ....................................................................... 38

2.4 Experimental Approach ........................................................................... 40

2.5 Generic CPAP Device Expansion Chamber Muffler .............................. 46

Chapter 3 Transmission Loss of a Resistive Muffler ........................................... 50

3.1 Introduction ............................................................................................. 50

3.2 Computational Modelling of Porous Materials ....................................... 50

3.3 Experimental Methods ............................................................................ 55

3.3.1 Characteristic impedance and propagation coefficient ............. 55

3.3.2 Airflow resistivity ..................................................................... 62

3.4 Equivalent Fluid Parameters ................................................................... 64

3.5 Muffler Transmission Loss ..................................................................... 68

Chapter 4 Transmission Loss of CPAP Device Mufflers ..................................... 73

4.1 Introduction ............................................................................................. 73

4.2 Muffler Designs ....................................................................................... 73

4.3 Finite Element Modelling of CPAP Device Mufflers ............................. 76

4.3.1 ANSYS ...................................................................................... 76

4.3.2 COMSOL .................................................................................. 77

4.3.3 Transmission Loss of the CPAP Device Mufflers .................... 78

Figure 2.14 Downstream microphone (ʹ) time history showing exponential