By Reading and Using The Thesis, The Reader Understands and Agrees To The Following Terms

Copyright Undertaking
This thesis is protected by copyright, with all rights reserved.
By reading and using the thesis, the reader understands and agrees to the following terms:
1. The reader will abide by the rules and legal ordinances governing copyright regarding the
use of the thesis.
2. The reader will use the thesis for the purpose of research or private study only and not for
distribution or further reproduction or any other purpose.
3. The reader agrees to indemnify and hold the University harmless from and against any loss,
damage, cost, liability or expenses arising from copyright infringement or unauthorized
usage.
IMPORTANT
If you have reasons to believe that any materials in this thesis are deemed not suitable to be
distributed in this form, or a copyright owner having difficulty with the material being included in
our database, please contact lbsys@polyu.edu.hk providing details. The Library will look into
your claim and consider taking remedial action upon receipt of the written requests.
Pao Yue-kong Library, The Hong Kong Polytechnic University, Hung Hom, Kowloon, Hong Kong
http://www.lib.polyu.edu.hk
OPTIMIZATION OF THE COMFORT OF
COMPRESSION SPORTS BRAS
LIANG RUIXIN
PhD
The Hong Kong Polytechnic University
2020
THE HONG KONG POLYTECHNIC UNIVERSITY
INSTITUTE OF TEXTILES AND CLOTHING
OPTIMIZATION OF THE COMFORT OF

COMPRESSION SPORTS BRAS
LIANG RUIXIN
A thesis submitted in partial fulfilment of the requirements for
the degree of Doctor of Philosophy
January 2020
CERTIFICATE OF ORIGINALITY
I hereby declare that this thesis is my own work and that, to the best of
my knowledge and belief, it reproduces no material previously published or written,
nor material that has been accepted for the award of any other degree or
diploma, except where due acknowledgment has been made in the text.
…………….…………………………………………. (Signed)
…………LIANG Ruixin…………………. (Name of Student)

ABSTRACT
Sports bra is an undergarment for women designed to provide extra support, control breast
movement and provide protection to the breasts during exercises. Comfortable and functional
sports bras are the strong needs and essentials for women, especially for physically active
women. However, there is limited literature to develop efficient, valid and reliable computer-
aided method to simulate the contact conditions between females and sports bras; and there is
limited literature to combine numerical simulation and comfort pressure evaluation to
determine an optimal material parameter of sports bra fabrics. Systematic studies on this area
are needed to provide design considerations for intimate apparel industry and arouse attention
to women’s health.
Therefore, the key aim of this study was to establish and validate a new scientific method to
combine finite element (FE) modelling and subjective perception of comfort pressure to
determine the optimum pressure and fabric parameters for the comfort of sports bras. The
specific objectives were to build appropriate FE models of female bodies and sports bras to
simulate pressure distribution on body exerted by a sports bra, and to determine the range of
comfort pressure wearing a sports bra, then to identify the optimum material parameter of
sports bra fabrics for optimum pressure comfort.
This study consists of four parts to achieve these objectives. The first one is building an
accurate FE model of female body, which has a rigid torso, hyperelastic breasts, and
subcutaneous tissues. The coefficients of the Mooney-Rivlin material, which describe the
female breasts, are determined by examining the differences between the FE-modelled results
and experimental data. The second is developing the FE contact model of a female body and
sports bra to calculate the static and dynamic contact pressure and dynamic displacement of
the breasts. The contact model is developed by a new method called ’first shrink then expand’
and validated by motion capturing experiment. The calculated root mean square errors are less
than 1%, which shows a good agreement between the FE results and experimental data. The
third is determining comfort pressure range by subjective questionnaire about comfortable
sensation and objective pressure tests. The pressure feelings are analysed not only in static
condition (standing or sitting) but also in dynamic condition (running). The optimum contact
pressures on the shoulders, underarm and at the bottom of the bra cup are 2.08 ±0.82 kPa, 2.66
± 0.83 kPa and 0.58 ± 0.57 kPa respectively based on the results of Chapter 5. The fourth
objective is making prediction of contact conditions using finite element (FE) and machine
learning (ML) approaches. FE results are used to provide the database for the ML model. A
comparison between the results obtained from the FE and ML methods shows a good
agreement between the two methods. The effect of different sports bra components on the
contact pressures is investigated through the results of ML model, which is that the shoulder
strap and bra cup can greatly influence the contact pressure, while the back panel and elastic
bra band do not have much influence the contact pressure.
This study provides a robust computational method to optimize the design of sports bras. This
method can also potentially be applied to solve the contact problem between relatively rigid
and hyperelastic materials, or other problems in the area of comfort, elastic materials, intimate
apparel and sports activities. The optimization of sports bra can address a common need and
contribute to the wellbeing of women globally.

PUBLICATIONS ARISING FROM THE THESIS
Journal articles
1. Liang, R., Yip, J., Yu, W., Chen, L., & Lau, N. M. (2019). Numerical simulation of
nonlinear material behaviour: Application to sports bra design. Materials & Design,
108177.
2. Liang, R., Yip, J., Yu, W., Chen, L., & Lau, N. M. Computational modelling
methods for sports bra-human interactions. International Journal of Clothing
Science and Technology.
3. Liang, R., Yip, J., Yu, W., Chen, L., & Lau, N. M. Finite element-based machine
learning method to predict breast displacement during running. AATCC Journal of
Research, Under review
Conference presentation and publications
1. Liang, R., Yip, J., Yu, W., Chen, L., & Lau, N. M. (2018). Evaluation of stress
distribution of the body exerted by compression sports bra using Finite-element
body and sports bra sub-models. The 91st Textile Institute World Conference 2018.
2. Liang, R., Yip, J., Yu, W., Chen, L., & Lau, N. M. (2019). Finite element-based
machine learning method to predict breast displacement during running. The 2nd
Artificial Intelligence on Fashion and Textile International Conference 2019.

ACKNOWLEDGEMENTS
I would like to take this opportunity to deliver my deep and sincere gratitude to my
chief supervisor, Dr. Joanne Yip, for her invaluable academic advice, professional
guidance, extraordinary patience and consistent encouragement throughout my PhD
study. I greatly appreciate for guiding me to learn a systematic research methodology
and develop logic modes of thought. With her great help of providing me with precious
suggestion and inspiration of new ideas, I have the confidence to face all of the
problems and have conducted a series of research studies to finish this thesis in the past
three years. These helps will be valuable to my future study and work.
Besides, I am extremely grateful for my co-supervisor, Prof. Winnie Yu, for her kind
guidance, constructive suggestions, warm help, inspiring comments, as well as
interesting discussions throughout the whole PhD study period. Without her help, I
would have never had the chance to study at The Hong Kong Polytechnic University.
I would also like to express my special thanks to my co-supervisor, Dr. Lihua Chen, for
her professional comments and guidance for the research methodology.
I would like to acknowledge my sincere gratitude to my co-supervisor, Dr. Newman
Lau, for his encouragement and the help to the experiments during my study.
I would like to acknowledge the Hong Kong Research Grant Council (RGC) for funding
this research [PolyU 152510/16E] and also thank the Institute of Textile and Clothing
(ITC) at The Hong Kong Polytechnic University for providing me with excellent
research conditions and environment during the last three years.

I would like to give my most grateful appreciation to Dr. Yue Sun, Ms. Shichen Zhang,
Mr. Wanzhong Jiao and Dr. Yiqing Cai for their support and professional suggestions.
Finally, I would like to give my special thanks to my parents, Mr. Zejun Liang and Ms.
Cuilan Wang, and my families for their spiritual encouragement and loving support
during my study.
TABLE OF CONTENTS
page
Chapter 1 Introduction ................................................................................................... 1
1.1 Research background ................................................................................... 1
1.2 Statement problem ............................................................................................ 6
1.3 Research objectives........................................................................................... 7
1.4 Project originality and significance .................................................................. 8
1.5 Research methodology ...................................................................................... 8
1.6 Outline of thesis .............................................................................................. 10
1.7 Limitation of study.......................................................................................... 12
Chapter 2 Literature Review ........................................................................................ 15
2.1 Introduction ..................................................................................................... 15
2.2 Biomechanical properties of breasts ............................................................... 15
2.2.1 Anatomy of breasts ............................................................................... 15
2.2.2 Experimental techniques to characterize breast tissues ........................ 17
2.2.3 Material properties of breasts................................................................ 20
2.3 Sports bra ........................................................................................................ 22
2.3.1 Types of sports bras............................................................................... 24
2.3.2 Effects of sports bras ............................................................................. 27
2.3.3 Pressure exerted by a sports bra ............................................................ 33
2.3.4 Pressure comfort evaluation .................................................................. 36
2.4 Finite element model of breasts ...................................................................... 37
2.5 Interactions between body and garment ......................................................... 39

2.6 Machine learning approach ............................................................................. 43
2.7 Motion capture experiment ............................................................................. 45
2.7.1 Motion capture equipment .................................................................... 45
2.7.2 Studied points........................................................................................ 47
2.7.3 Reference system .................................................................................. 48
2.8 Conclusions ..................................................................................................... 50
Chapter 3 Development of finite element model of body............................................ 53
3.1 Introduction ..................................................................................................... 53
3.2 Subject-specific FE model of body obtained by 3D laser scanning ............... 54
3.2.1 Subjects ................................................................................................. 54
3.2.2 3D laser scanning .................................................................................. 57
3.3 Motion capturing............................................................................................. 59
3.4 Finite element model....................................................................................... 64
3.4.1 Geometric modelling ............................................................................ 64
3.4.2 Mesh creation ........................................................................................ 65
3.4.3 Coefficients of Mooney-Rivlin material ............................................... 68
3.4.4 Damping ratio ....................................................................................... 73
3.5 Conclusions ..................................................................................................... 79
Chapter 4 Development of finite element contact model ............................................ 81
4.1 Introduction ..................................................................................................... 81
4.2 Experimental work .......................................................................................... 82
4.2.1 Tensile strength test of sports bra components ..................................... 83

4.2.2 Contact pressure test ............................................................................. 84
4.3 Finite element contact model .......................................................................... 86
4.3.1 Sports bra model ................................................................................... 86
4.3.2 Loading and boundary conditions ......................................................... 88
4.4 Contact pressure .............................................................................................. 92
4.4.1 Static contact pressure........................................................................... 92
4.4.2 Dynamic contact pressure ..................................................................... 95
4.5 Dynamic breast displacement with sports bra ................................................ 96
4.6 Modal analysis .............................................................................................. 100
4.7 Conclusions ................................................................................................... 103
Chapter 5 Determining pressure comfort range ......................................................... 104
5.1 Introduction ................................................................................................... 104
5.2 Experimental work ........................................................................................ 105
5.2.1 Subjects ............................................................................................... 105
5.2.2 Experiment .......................................................................................... 107
5.3 Results and discussion .................................................................................. 111
5.4 Conclusions ................................................................................................... 121
Chapter 6 Prediction of contact pressure and dynamic displacement during running
.................................................................................................................................... 123
6.1 Introduction ................................................................................................... 123
6.2 Finite element-based machine learning approach to predict contact pressure
............................................................................................................................. 123
6.2.1 Dataset generation............................................................................... 124
6.2.2 ANN structure ..................................................................................... 127
6.2.3 Validation ............................................................................................ 134
6.3 Finite element models of different subjects .................................................. 138
6.4 Conclusions ................................................................................................... 147
Chapter 7 Conclusions and Recommendations for future work ................................ 149
7.1 Introduction ................................................................................................... 149
7.2 Development of finite element models ......................................................... 149
7.3 Comfort pressure evaluation of sports bra .................................................... 150
7.4 Prediction of contact pressures between the sports bra and female body ..... 151
7.5 Limitations and future work.......................................................................... 152

LIST OF FIGURES
page
Figure 1.1 Compression sports bra ................................................................................ 1
Figure 1.2 Encapsulation sports bra ............................................................................... 1
Figure 1.3 Process of creating an individual-specific breast model. ............................. 4
Figure 1.4 Research framework ................................................................................... 10
Figure 1.5 Outline of thesis .......................................................................................... 12
Figure 2.1 Anatomy of breast....................................................................................... 17
Figure 2.2 Principle of elastography ............................................................................ 18
Figure 2.3 Breast tissue modulus measurement system............................................... 20
Figure 2.4 Two main types of sports bras .................................................................... 25
Figure 2.5 Combination sports bra............................................................................... 26
Figure 2.6 Structure of a sports bra .............................................................................. 27
Figure 2.7 Back design of sports bras: (a) cross-over; (b) racer-back; (c) vertical centre;
(d) straight back; and (e) U-back ................................................................................. 27
Figure 2.8 Measurement system of clothing pressure ................................................. 34
Figure 2.9 Different types of pressure sensors............................................................. 35
Figure 2.10 Measured pressure points ......................................................................... 36
Figure 2.11 Testing principles of KES ......................................................................... 40
Figure 2.12 Garment pressure distribution .................................................................. 41
Figure 2.13 Reverse mapping of garment pressure distribution onto paper pattern .... 41
Figure 2.14 FE model of female body and bra ............................................................ 43
Figure 2.15 Markers to measure breast motion ........................................................... 47
Figure 2.16 Marker positions to measure breast motion.............................................. 48
Figure 2.17 Reference body landmarks ....................................................................... 49
Figure 3.1 Marker positions for defining boundaries of breast ................................... 57
Figure 3.2 3D body scanning of braless subject .......................................................... 59
Figure 3.3 Digital camera for motion capture .............................................................. 60
Figure 3.4 Floor plan of motion capture experiment ................................................... 60
Figure 3.5 Position of markers on breasts - 1: sternum; 3: right nipple; 8: left nipple;
and 2, 4, 5, and 6 are placed 4 cm away from 3 and 8 ................................................. 61
Figure 3.6 T-shaped wand used with retro-reflective markers ..................................... 63
Figure 3.7. Free vibration of breast .............................................................................. 63
Figure 3.8 Initial geometric model............................................................................... 65
Figure 3.9 Element 157 ................................................................................................ 66
Figure 3.10 Mesh of model of body............................................................................. 66
Figure 3.11 Steps of dynamic analysis......................................................................... 70
Figure 3.12 Dynamic displacements of breasts: experiment vs. FEM ........................ 72
Figure 3.13 Time vs. displacement of breasts .............................................................. 74
Figure 3.14 Time-displacement curves of braless condition and calculated damping
ratios............................................................................................................................. 76
Figure 3.15 Experimental curve of free vibration for S13 ........................................... 77
Figure 3.16 Magnified experimental curve .................................................................. 79

Figure 4.1 Sports bra samples ...................................................................................... 83
Figure 4.2 Stretching of sports bra ............................................................................... 84
Figure 4.3 Instron 4411 ................................................................................................ 84
Figure 4.4 Pressure test positions................................................................................. 85
Figure 4.5 Element 138 ................................................................................................ 86
Figure 4.6 FE model of compression sports bra .......................................................... 87
Figure 4.7 Boundary conditions of contact model ....................................................... 90
Figure 4.8 Distribution of contact pressure between Bra 1 and S01 ........................... 93
Figure 4.9 Distribution of contact pressure between Bra 2 and S01 ........................... 93
Figure 4.10 Distribution of contact pressure between Bra 3 and S01 ......................... 93
Figure 4.11 Distribution of contact pressure between Bra 4 and S01.......................... 94
Figure 4.12 Distribution of contact pressure between Bra 5 and S01 ......................... 94
Figure 4.13 Simulation and experimentally obtained breast displacement during two
speeds of running: Bra 1 .............................................................................................. 97

Figure 4.18 First mode shape ..................................................................................... 102
Figure 5.1 Questionnaire on perceived comfort sensation......................................... 109
Figure 5.2 Experimental procedure............................................................................ 110
Figure 5.3 Testing contact pressure of different bra components on body ................ 111
Figure 6.1 Regression process of Bayesian regularization algorithm........................ 128
Figure 6.2 Performance of Bayesian regularization algorithm .................................. 129
Figure 6.3 Regression process of Scaled conjugate gradient algorithm .................... 131
Figure 6.4 Performance of Scaled conjugate gradient algorithm .............................. 131
Figure 6.5 Regression process of Levenberg-Marquardt algorithm .......................... 133
Figure 6.6 Performance of Levenberg-Marquardt algorithm..................................... 134
Figure 6.7 ANNs structure ......................................................................................... 134
Figure 6.8 Effects of fabric of sports bra components with different elastic moduli. 138
Figure 6.9 Distribution of contact pressure of different subjects ............................... 140
Figure 6.10 Change in contact pressure with increase in elastic modulus of bra strap12
Figure 6.11 Change in contact pressure with increase in elastic modulus of bra cup
.................................................................................................................................... 144
LIST OF TABLES
page
Table 2.1 Summary of Young’s modulus values of breast tissues ............................... 21
Table 2.2 Properties of fabrics used for sports bras ..................................................... 23
Table 2.3 Important fiber properties in standard climate (22 ℃, r.h. =65%) .............. 24
Table 2.4 Summary of studies on effects of sports bras............................................... 28
Table 2.5 Reference body landmarks ........................................................................... 50
Table 3.1 Demographics and body measurements of subjects .................................... 55
Table 3.2 Description of markers for defining boundaries of breast ........................... 58
Table 3.3 Results of different mesh sizes ..................................................................... 67
Table 3.4 Material coefficients..................................................................................... 70
Table 3.5 RMSEs of different material coefficients..................................................... 72
Table 3.6 Material coefficients of different subjects and RMSEs ............................... 73
Table 4.1 Measured points ........................................................................................... 85
Table 4.2 Material properties of sports bra components .............................................. 87
Table 4.3 Comparison of numerically and experimentally obtained contact pressure 94
Table 4.4 RMSE percentage of simulated breast displacement ................................... 99
Table 4.5 Frequencies of breast motion during running ............................................ 101
Table 4.6 Natural frequencies with and without a sports bra ..................................... 101
Table 5.1 Subject information .................................................................................... 106
Table 5.2 Mean scores and contact pressures ............................................................ 112
Table 5.3 Correlation coefficients among overall score, scores for perceived pressure of
different parts of bra, and contact pressure values ..................................................... 112
Table 5.4 ANOVA table of influence of contact pressure on scores of perceived
pressures in static condition ....................................................................................... 114
Table 5.5 ANOVA table of influence of contact pressure on scores of perceived
pressures in dynamic condition .................................................................................. 117
Table 5.6 No. of conditions in three groups ............................................................... 121
Table 5.7 Mean pressure values ................................................................................. 121
Table 6.1 Finite element parameters and results ........................................................ 125
Table 6.2 RMSE between FE and ANN results ......................................................... 135
Table 6.3 Number of meshes of FE model of different subjects................................ 139
Table 6.4 Simulated contact pressure of different subjects ........................................ 141
Table 6.5 Correlation coefficients between contact pressure and subject particulars 142
Table 6.6 Increase in contact pressure with higher elastic modulus of bra strap and bra
cup .............................................................................................................................. 145

ABBREVIATIONS
AI Artificial intelligence
ANOVA Analysis of variance
ANNs Artificial neural networks
BI Inner-breast
BL Left nipple
BO Outer-breast
BR Right nipple
CCDs Charge-coupled devices
CT Computed tomography
FAST Fabric assurance by simple testing
FE Finite element
FEA Finite element analysis

K
KES Kawabata evaluation systems
ML Machine learning
MR Magnetic resonance
MRE Magnetic resonance elastography
MRI Magnetic resonance imaging
OCTE Optical coherence tomographic elastography
RTE Real-time ultrasound elastography
SWMs Support vector machines
RMSE Root mean square error
2D Two-dimensional
3D Three-dimensional
Chapter 1 Introduction
1.1 Research Background
Physically active women are often invariably frustrated in their attempts to find a comfortable
and supportive sports bra, and it is even more critical for plus-size women. A sports bra is an
undergarment for women designed to provide extra support, minimize breast movement and
provide protection to the breasts during sporting activities (Niemczyk, Arnold, & Wang, 2017).
There are normally two kinds of sports bras, which are compression and encapsulation sports
bras. The former are designed to press the breasts against the chest wall, and similar in
appearance to a short vest (Figure 1.1). The latter are designed to support the breasts
individually, and similar in appearance to daily wear bras (as shown in Figure 1.2) (Yu et al.,
2014). Generally, compression sports bras offer less support than encapsulation sports bras, so
they are more suitable for smaller breast women. However, White, Scurr and Smith (2009)
concluded that there is no significant difference between these two kinds of sports bras in their
ability to limit breast movement.
Figure 1.1 Compression sports bra Figure 1.2 Encapsulation sports bra
Breast tissues or ligaments very easily move on the chest wall during physical activities when
a woman is merely walking slowly, because the breasts have no substantial anatomical support
such as bones or muscles. Movement can therefore cause injury to the breasts and tissues that
surround them due to the stress and internal forces. In order to limit the amount of breast
-1-
movement during sports, wearing a sports bra can enhance the comfort of the breasts during
exercise. However, if the sports bra exerts excessive pressure onto the breasts, the result would
be wear discomfort. An inappropriate sports bra design may also have physical and
physiological effects on the body.
The aim of this study is to use the nonlinear finite element analysis (FEA) to predict the
optimum level of comfort in compression sports bras. Using an FEA to find the optimum design
parameters has already been widely done so in the engineering domain, such as to examine the
load of reinforced concrete beams (Ngo & Scordelis, 1967), viscoelastic behaviour of car tire
materials (Nandi et al., 2014), ocean circulation problems (Fix, 1975), stress and strain inside
heel pads (Luo et al., 2011), and design of therapeutic footwear (Spirka et al., 2014). However,
it is very challenging to analyse the comfort level of sports bras because the interactions
between the fabric which is elastic and the breasts which are viscoelastic are complex. It is
difficult to measure and quantify breast viscoelasticity, the interaction between human skin and
garment, and the nonlinear mechanical behavior of the breasts during static and dynamic
loading; therefore, it becomes a difficult task to construct a biomechanical interaction model.
Furthermore, the other difficulties involved are the large variations in breast size and shape
between different individuals. Finally, sports bras usually have at least two layers of different
fabrics for support, but women have different sensations towards pressure and comfort so there
is currently no scientific means of determining the minimum amount of pressure that can
provide the necessary support to the breasts.
Existing finite element (FE) models of the breasts have used published data, or derived the
mechanical properties of the breasts by carrying out indentation tests with ultrasound on a small
area of the breast. The results are then validated through comparisons with ex vivo data. (Lu et
al., 2009) It has also been noted that the Young’s modulus values are higher than the actual
values in the indentation tests due to the breast viscoelasticity. For simplicity, the material
-2-
properties of the human body are assumed to be the average value of the material properties of
the fat and glands. Therefore, it is obvious that constructing a more accurate FE model by using
the respective material properties of important structures such as the glands, ligaments and
pectoral muscles could improve the accuracy of the modelling. Rajagopal et al. (2008) provided
a new modelling framework for mapping tissue locations across medical images, such as
mammograms, and magnetic resonance and ultrasound images. However, this method is too
expensive when a large number of experiments need to be done.
-3-
Figure 1.3 Process of creating an individual-specific breast model. (a) Segmentation of tissue
boundaries from a magnetic resonance (MR) image. Light spots represent the skin surface,
and dark spots represent the muscle surface. (b) Dataset of skin (light spots), and muscle
(dark spots) after segmenting an entire MR image set. (c) Generic finite element breast
geometry and breast skin and muscle datasets. (d) Finite element model fitted to skin and
muscle surface data. (e) Fitted model without surface rendering to show the volumes of the
fitted elements. (Rajagopal et al., 2008)
-4-
As for bra fit, research work that uses FEA to examine the mechanical interaction between
sports bras and the human body is still in its infancy. The available studies mostly use a rigid
body, which makes it easier to carry out calculations. For instance, Ishimaru et al. (2011)
developed a numerical-analysis-based technique to simulate clothing pressure on a rigid
mannequin. FEA is also widely used in the biomechanical field. For example, Dai et al. (2006)
developed an FE model to simulate foot-sock-insole contact. However, the human foot has
more bones and joints than the breasts, which means that it is more rigid and incompressible.
If a rigid model is also used for the breasts, the result would not be reliable. Li, Zhang and
Yeung (2003) built an FE model of the female body, which consists of breasts with elasticity
and a rigid body. However, they estimated the material properties of the body from published
data in the literature and the breast density from the weight and volume of the breasts. Moreover,
their single-material and single-layer bra structure are overly simplified because commercial
compression sports bras mostly have two layers of material and elastic bra bands.
Scientific analyses of the pressure comfort range while wearing sports bras have been seriously
lacking. While skin pressure can be measured by using a pressure sensor for a single point, the
evaluation of the pressure distribution is not possible and the measured pressure values of the
breasts with soft tissues are not precise because they may be less than a practical measurable
range. According to previous research, the maximum pressure tolerated by a human being,
which will not cause health problems, is 3.920 kPa (Wang et al., 2006). Kobayashi et al. (2011)
stated that clothing pressure should be as low as possible. However, the pressure on the breasts
while wearing a sports bra must be high enough to support the breasts in order to limit
movement.
-5-
In this study, a nonlinear FEA is proposed to predict the optimum level of comfort of
compression sports bras and the focus is the optimization of the fabric properties of the sports
bras. To achieve this aim, a new scientific and reliable system that uses FE simulation will be
developed and validated to determine the optimum pressure and the fabric properties for the
comfort of compression sports bras. In order to build a realistic model of the female body, the
compression modulus and damping coefficients of the breasts will be measured through a
breast vibration analysis. The results from the FEA show a good agreement with the
experimental results.
1.2 Statement Problem
Compression sports bras compress the breasts against the chest wall to limit their movement
during physical activity, which has very important impacts on health because the compression
could lead to pain. There are also other problems with sports bra designs that influence the
effect and comfort of bra wear for women.
1) Designing conventional bras is a time-consuming process and prone to human errors.
Sports bra designs involve many steps in a long process, such as developing the design
prototype, making the pattern, selecting the fabric, and carrying out fitting. Improper
construction methods and/or use of fabric in any of the bra design procedures may fail to satisfy
the requirements of a good sports bra. In addition, this process highly depends on the
experience and expertise of the designer, who has to fully understand the relationship between
the sports bra features and feelings of the wearer.
2) There is uncertainty of the material parameters of the breast and breast boundaries.
Previous studies have suggested that it is critical to build authentic FE models of the human
body, which can be done by capturing images of the body with a 3D laser scanner. Mooney-
-6-
Rivlin material models are considered to be appropriate for describing the material properties
of the breasts. However, it is still uncertain whether the breast boundaries are well established
enough to build a geometric model and further work is therefore required to measure the
nonlinear hyperelastic parameters of the different types of breast tissues.
3) It is challenging to simulate the contact between the body and sports bra.
The interaction between the elastic fabrics used in sports bra and breast viscoelasticity is highly
complex. Moreover, computational simulation of the contact between the body and sports bra
usually have penetration and stress concentration problems. Therefore, the few previous studies
that have done work in this area have unrealistically modelled bras only as a single-layer
garment imposed onto a rigid body model or arbitrarily-defined breasts with soft tissues.
4）There is a challenge for objective quantification of comfort in bra wearing.
How to quantify the feeling of comfort while wearing a sports bra is a question for this kind of
study. The common solution is using a questionnaire to let the subject score his/her feeling.
Questions and analysis method need to be well designed.
1.3 Research Objectives
This project aims to use a nonlinear FEA to predict the optimum level of comfort for
compression sports bras. The specific objectives of the work in this thesis are summarized as
follows.
1. To develop FE sub-models of the various breast shapes by using in-vivo material properties;
2. To design and develop an FE sub-model of a sports bra based on the design specifications
and directly measured fabric properties;
3. To simulate the stress distribution on the body exerted by a sports bra with different fabrics
based on a sample of women;
-7-
4. To determine the range of the pressure comfort values of sports bras made of different fabrics;
and
5. To predict the contact pressures between sports bra and body, and also identify the optimum
fabric elastic modulus for optimum pressure comfort.
1.4 Project Originality and Significance
This study systemically investigates the mechanical properties of breasts and builds FE models
of the female body and sports bras. The FE contact model used to simulate the distribution of
contact pressure between the body and a sports bra, and breast movement due to dynamic
conditions when a sports bra worn, is found to have good accuracy. A machine learning (ML)
model of the contact pressure has been established. This successful endeavour provides a more
efficient, accurate and robust strategy for solving not only the complex problems of body-bra
interactions but also other design applications where the properties of the materials are highly
nonlinear and viscoelastic. This method can potentially provide the basis for future research
work on apparel comfort, elastic materials, intimate apparel and sports activities. Improved
sports bra designs will also contribute to the wellbeing of the female population globally.
1.5 Research Methodology
To realize the aim and objectives, this study follows a research framework and uses the
research methodologies that are discussed below.
This study involves both experimental work and computer modelling, in which five tasks are
carried out. Task 1 is building the sub-models of the body. To do so, three-dimensional (3D)
laser scanning experiments of solicited female subjects who participate on a voluntary basis
are first carried out. Then, the scanned images are computationally processed, such as
smoothing and removing the irrelevant parts, to obtain subject-specific geometric models.
-8-
Following that, motion capturing is carried out (experimental component) to record the
movement of braless breasts during running and when subjected to vibration. The material
properties of these breasts are then analysed based on the motion capture data.
Task 2 is building the sub-model of the sports bra. First, a geometric model of the sports bra is
extracted from a gravity-free model of the body of a female subject with different
configurations of the components found in a commercial compression sports bra. The
compression sports bra has a high neckline to limit the amount of upward movement of the
breasts, and wide shoulder straps to distribute the pressure over the shoulders to the large back
panel of the bra. The mechanical properties of the sports bra fabrics are also tested by using an
Instron 4411 tester at a constant-rate-of-extension.
Task 3 is simulating the pressure distribution over the body when the sports bra is worn. After
the FE models of the body and sports bra are obtained in Tasks 1 and 2, FE contact models are
then built and calculated to achieve Task 3. To avoid the penetration problem in the contact
models, a new method is used in which the model of the object of interest is first reduced and
then expanded (hereinafter ‘first reduce, then expand’) to successfully simulate the contact
conditions. This method will be explained in more detail in the coming chapters. An FE-based
ML approach is also used to build a prediction model of the contact pressure between the body
and sports bra, which can further enhance the efficiency of the modelling.
Task 4 is determining the range of the pressure comfort. In order to achieve Task 4, the female
subjects are requested to complete a questionnaire on their perceived comfort during the wear
trials for both the static and dynamic conditions. Systematic and scientific analyses of their
ratings would then be carried out.
Task 5 is identifying the optimal elastic modulus of the sports bra fabrics, which means
determining the elastic modulus value of the fabric which provides the optimum amount of
-9-
contact pressure between the sports bra and body. The use of the numerical contact models
obtained in Task 3 and pressure comfort range obtained in Task 4 would determine the
optimum performance of the sports bra with the least amount of pressure exerted onto the body.
To enhance efficiency, FE based machine learning method can be used to predict contact
pressures between sports bra and body.
A flow chart of the five tasks and the research framework are illustrated in Figure 1.4.
Figure 1.4 Research framework
1.6 Outline of thesis
There is a total of seven chapters in this thesis. Chapter 1 discusses the background information,
problem statement, objectives, project originality and significance and outline of this study.
Chapter 2 is the literature review and presents the current literature on the biomechanical
properties of breasts. Then, an analysis of the functions, types and features of existing sports
bras are presented. Following this, there is a systematic review of the methods and findings of
- 10 -
previous experiments and computer simulations of human bodies and contact problems. The
knowledge gaps are presented which provide the basis of the research objectives and methods
of this study.
Chapter 3 describes how to build an accurate FE model of female body. The FE body models
consist of a rigid torso, a layer of soft subcutaneous tissues, and two hyperelastic breasts. Two
experiments are required. The first one is 3D laser scanning to obtain the subject-specific
geometric models. The second experiment is motion capturing to record the movement of the
subjects and, then, determine the material properties of the breasts. This chapter is the
foundation of the whole study.
Chapter 4 presents the method of developing FE contact model between the female body and
sports bra. A new method called ‘first shrink then expand’ is used to avoid penetration problem.
The body model is first shrunk and a gap between the body model and sports bra model is made.
Secondly, the body model expands to contact with the sports bra model and the distribution of
contact pressures can be analysed. This FE contact model is verified by dynamic breast’s
displacements when a sports bra is donned with contact pressure. The error between the
experimental and FE results is acceptable. To further enhance the function of sports bras, modal
analysis in bra and braless conditions is made.
Chapter 5 reports how to determine the comfort pressure range of wearing a sports bra. The
method combines the objective pressure tests and subjective comfort assessment. Moreover,
the comparison between static and dynamic conditions is made. The findings of this chapter
can be used as guidance of compression sports bra design.
Chapter 6 presents the FE contact models of different female subjects and their results are
compared. Furthermore, a machine learning approach is used to investigate the contact
conditions between the body and sports bra to raise efficiency. The accuracy of the established
- 11 -
machine learning model is validated by comparing with FE results. Then, the effect of each
sports bra component on contact pressures is easily investigated and, thus, provide design
considerations for sports bra.
Chapter 7 concludes the work and provides recommendations for future work and studies.
A flowchart of the thesis chapters is shown in Figure 1.5.
Figure 1.5 Outline of thesis
1.7 Limitation of study
As with all studies, this study also has its limitations, including a focus on the methods to
predict the contact conditions between sports bras and the wearer. Moreover, only fifteen
- 12 -
subjects participated in the experimental component which is a rather smaller sample size. The
study only considers a sports bra; both the FE and ML models are based on the same
compression sports bra which has a vest-style. Hence, the results can only be applied to this
type of sports bra. However, encapsulation style sports bras are also popular in the market.
Therefore, future works should also consider other types of sports bras.
Breasts have a very complex structure and different types of tissues. However, detailed FE
models of the breasts make very little difference to the contact conditions between the sports
bras and the body. In addition, it is very time-consuming to build FE models. Hence, it is
assumed in the study that the breast is a homogeneous object made of soft matter with elastic
damping properties.
As for the experiments, breast motion with a sports bra donned is recorded through tracking
markers placed on the sports bra, and not on the surface of the breasts. Although the bra is
fitted well onto the female subject before the experiment takes place, there still may be slight
relative movements from the breasts and the bra during running. However, the working
principle of these passive markers is to reflect the infrared rays back to the cameras, so that the
markers cannot be placed on the surface of the breasts as they will be covered by the bra.
- 13 -
Chapter 2 Literature Review
2.1 Introduction
This chapter is the literature review, which aims to identify key publications that are relevant
to this study. The chapter includes studies on biomechanical models of breasts, breast
kinematics, sports bra designs, and the interaction between the human body and garments. First,
the anatomy of the breasts and evaluation of the function and comfort levels when wearing a
sports bra are reviewed. Then, previous studies on FE models of the breasts are systematically
reviewed. Also, interaction between the human body and garment are referenced. Finally,
research studies on the effect of sports bra design on sports activities with the use of motion
capture are reviewed.
2.2 Biomechanical properties of breasts
2.2.1 Anatomy of breasts
The breasts are essential for producing and secreting milk for newborn babies, so they are an
important organ for women. The breasts are a subcutaneous gland located on the anterior chest
wall and usually extend from the level of the second rib to that of the sixth rib. The breasts
consist of layers of two types of tissues: adipose and glandular tissues (Drake, Vogl & Mitchell,
2009). One-third of the breasts is fatty tissue and the amount can vary significantly between
individuals due to the age and nutritional status. The superficial layer of the breasts (fascia) is
separated from the skin by 0.5 to 2.5 cm of subcutaneous fat (Genfen & Dilmoney, 2007). The
other two-thirds is glandular tissue, which supports lactation. A female adult breast contains
15 to 20 lobules and ducts with diameters that range from 2 to 4.5 mm, which converge at the
nipple (Azar, Metaxas & Schnall, 2000).
15
There are no muscles in the breasts. Cooper’s ligaments support the breasts and keep them in
position. The breasts can be firmly attached to the skin because of these ligaments. Cooper’s
ligaments extend from the fascia layer outwards to the skin in many directions, which look like
tentacle in appearance. The mechanical properties of the Cooper’s ligaments have only been
approximated in previous studies (Genfen & Dilmoney, 2007). The anatomy of a female breast
is shown in Figure 2.1.
Glandular tissue is affected by estrogen. As women age towards menopause, their estrogen
levels begin to decline. Therefore, glandular tissue atrophies and eventually disappears (Azar,
Metaxas & Schnall, 2000). The superficial fascia, suspensory ligaments and skin take the
responsibility of supporting the breasts and will also change in the level of fibrous tissue with
time due to gravity load. These changes will have effects on the biomechanical properties of
the breasts. For example, stretching of the Cooper’s ligaments and reductions in the coupling
between the breasts and the surrounding tissues can be found with increasing age. Breast cancer,
which comprises about 25% of all types of cancer in women, can also affect the elasticity of
the breasts (Stewart & Wild, 2014; Ramião et al., 2016). Research work on the biomechanical
properties of the breasts therefore play an important role in the early detection of breast cancer.
16
Figure 2.1 Anatomy of breast (Ellis & Mahadevan, 2013)
2.2.2 Experimental techniques to characterize breast tissue
This section discusses the studies that have carried out experiments on the material properties
of the breast tissues. The experimental techniques are usually divided into two types: in vivo
and ex vivo techniques. The former test the entire breast by using relatively small loads to avoid
injuring or damaging the breast. The latter test tissues that are removed from the breasts. Using
ex vivo extracted tissues allows tests or measurements to be carried out that would otherwise
not be possible or ethical in living female subjects.
Elastography scanning is the most commonly used in vivo technique in most studies.
Elastography scanning is a non-invasive imaging technique usually used to detect tumours,
which was developed in the late 1980s (Lerner, 1987; Lerner et al., 1988). There are various
methods of carrying out elastography scans, including mammography, indentation testing,
magnetic resonance elastography (MRE), ultrasound, computed tomography (CT), optical
17
coherence tomographic elastography (OCTE) and real-time ultrasound elastography (RTE)
(Venkatesh & Ehman, 2014). The basic theory behind elastography scanning is measuring the
deformation of breast tissue by applying a very minimal amount of pressure onto the surface
of the body, which is shown in Figure 2.2 (Shiina, 2013). As elastography scanning can
measure the response of the breast tissues to applied pressure, the technique can be used to
analyse the mechanical properties of the breast tissues. On the other hand, elastography
scanning provides high levels of specificity and sensitivity so that it can be used to detect breast
tumours and distinguish the tumours from healthy tissue (Wilson, Robinson & Dadd, 2000).
Palpation is another method for detecting breast tumours in clinical examinations. However,
this technique mostly depends on the experience of the doctor and is only useful for finding
large tumours (Ramião et al., 2016).
Figure 2.2 Principle of elastography (Shiina, 2013)
18
Ex vivo methods of examining breast tissue require the tissue to be tested in a very short amount
of time after removal from the body (Samani & Plewes, 2007). Samani et al. (2007) developed
a computer-controlled loading system to measure the elastic modulus of normal breast tissue
and breast tumour slices, which is shown in Figure 2.3. Matsumura et al. (2009) and Umemoto
et al. (2014) provided a measurement system for breast tissues which consists of a materials
testing machine (INSTRON 3342) and 10-N load cells with a φ3 mm cylindrical indenter and
computer for analytical use. As for the experiment samples, both studies used 5-mm slices from
a surgically removed specimen immediately after segmental resectioning.
19
Figure 2.3 Breast tissue modulus measurement system (a) Schematic of the block shape
tissue indentation apparatus, (b) photo of the block shape tissue indentation apparatus, (c)
schematic of the tissue slice indentation apparatus and (d) photo of the tissue slice indentation
apparatus (Samani et al., 2007)
2.2.3 Material properties of breasts
This section reviews several studies on the biomechanical properties of breast tissues. The
female breasts are viscoelastic, anisotropic, and inhomogeneous with nonlinear behaviour. The
20
breasts can experience large deformations because they are not physically constrained and the
breast tissues have low stiffness (Fung & Cowin, 1994). However, many studies have assumed
that the breasts are linear and homogeneous elastic for simplification purposes (Samani et al.,
2003; Van Hounten et al., 2003; Umemoto et al., 2014). There have only been a few studies
done to test the nonlinear properties of the breasts. For instance, Krouskop et al. (1988)
investigated the mechanical behaviour of 142 breast tissue samples under compression loading.
However, they did not attempt to characterize the nonlinear parameters. Samani and Plewes
(2004) developed a system to measure the nonlinear hyperelastic parameters of soft tissues by
examining their force-displacement response via an inversion technique that uses a nonlinear
FE model. The results showed that the elastic modulus of the breast tissues widely range from
0.5 to 50 kPa. However, this system can only measure the properties of small block shaped
specimens of the breasts. Table 2.1 summarizes the Young’s modulus values of the breast
tissues based on nine studies in the literature.
The serious lack of studies on nonlinear hyperelastic parameters of different types of breast
tissues means that further work is required to measure the material parameters to develop a
database that can be used in the FE simulation of breast displacement.
Table 2.1 Summary of Young’s modulus values of breast tissues
Young’s modulus (kPa)

Source
Fatty tissue Glandular tissue
Samani et al. (2001) 3.25 3.24
McKnight et al. (2002) 3.3 7.5
Van Houten et al. (2003) 1 to 20 2 to 37
Samani et al. (2003) 1.8 1.9
21
Xydeas et al. (2005) 1.2 1.2
Matsumura et al. (2009) 0.7 to 17.3 0.8 to 15.4
Srivastava et al. (2011) 4.17
Umemoto et al. (2014) 0.69 to 19.08 0.73 to 16.99
Breast density is one of the most influential factors on breast health. However, it is difficult to
precisely measure breast density in vivo; therefore, breast density of young women remains
highly unknown (Dorgan et al., 2012). Breast density varies widely because of the different
percentage of constituent fat and tissue in each individual and also due to age. McGhee et al.
(2013) determined the breast mass by directly measuring the volume and multiplying the
volume by using a breast density coefficient in order to evaluate the influence of breast mass
on vertical breast displacement. Vandeweyer and Hertens (2002) calculated the average density
of breast fat and glandular tissue is 0.50 × 103 kg·m-3 and 1.06 × 103 kg·m-3, respectively.
2.3 Sports bra
A good sports bra should provide adequate support and also limit breast movement both
vertically and laterally relative to the body through an underwire or a strong elastic strap
(Haycock, Shierman & Gillette, 1978). Runner Lisa Lindahl and costume designer Polly Smith
in the US succeeded in creating the first sports bra by sewing two jockstraps together in the
late 1970s (Shea, 2017) which originated from the frustration felt by Lisa by the inadequacy of
her daily wear bra for jogging. This first sports bra reduced breast motion whilst running
(Schuster, 1979). Fabric that was subsequently used for sports bras greatly affected the effect
and effectiveness of support to limit breast motion (Page & Steele, 1999). In the last 40 years,
technology in women’s intimate apparel has developed rapidly with the development of
advanced materials, such as Lycra, elastane and Coolmax® (Bowles, Steele & Munro, 2012).
22
These relatively new materials are light in weight with high strength, and their specific
properties are shown in Table 2.2 and 2.3 in contrast to cotton.
Table 2.2 Properties of fabrics used for sports bras (Yu & Zhou, 2016)
Fiber
Property
Cotton Coolmax Spandex(Lycra) Polyamide(Nylon) Polyester
Elasticity Low High High High High
Stretch
Low High High High High
ability
Recovery Low High High High High
Strength High High Low High High
Comfort High High High Low Low
Table 2.3 Important fiber properties in standard climate (22 ℃, r.h. =65%) (Shishoo, R. ed.,
2005)
Fiber
Polyester
Properties Spandex Polyamide
Cotton (Polyethylenete
(Lycra) (Nylon)
rephthalate)
Density
1.50-1.54 1.1-1.3 1.14 1.36-1.41
(g/cm3)
23
Break
elongation 6-10 400-700 15-60 25-55
(%)
Tensile
strength 20-50 5-70 35-90 25-95
(cN/tex)
Initial
modulus
(cross- 45-90 0.006-0.012 0.6-6 3.4-21
section)
(GPa)
2.3.1 Types of sports bras
There are mainly two kinds of sports bras: compression and encapsulation sports bras. The
latter are designed to support each breast separately, which is more effective for larger breast
women (see Figure 2.4a). The former are designed to flatten the breasts to provide support
(Figure 2.4b). Zhou et al. (2009) investigated the shock absorption performance of 4 types of
sports bras by evaluating the reduced percentage of controlled breast displacement and
provided various suggestions on designing sports bras. They claimed that a higher percentage
of polyamide, and higher bra neckline and wider shoulder straps could help to reduce breast
motion. While White et al. (2009) found no significant differences in breast displacement
between the two kinds of sports bras, McGhee, Steele and Munro (2010) found that greater
compression of compression sports bras can provide significantly increased breast and bra
comfort compared to standard encapsulation sports bras during physical activity for larger
24
breast women. However, the study by White et al. (2009) only examined one encapsulation
sports bra and one compression sports bra. To determine the effect of a sports bra, a variety of
different factors should be examined, and not only the type of sports bra.
(a) Encapsulation sports bra (b) Compression sports bra
Figure 2.4 Two main types of sports bras (Yu & Zhou, 2016)
Currently, there are also combination sports bras in the market (see Figure 2.5) which
incorporate the characteristics of both the compression and encapsulation sports bras. This bra
design is more effective because it combines the advantages of both bras. In addition, mode
specific sports bras, such as sports bras for ball game, have become popular in the last few
years.
25
Figure 2.5 Combination sports bra (Recreational Equipment, Inc. 2019)
To design a sports bra, the components needed are the bra cups, shoulder straps, back wing,
underband, etc.; see Figure 2.6. There are two kinds of bra cups in sports bras: traditional cup-
and-sewn and molded foam cups. The latter are assumed to provide better support for the
breasts than the former (Wu et al., 2012). The shoulder strap design also affects the comfort
and function of a sports bra. They are supposed to hold the bra in place. However, wearers
sometimes tighten the straps so that there is more support of their breasts, which may exert a
high amount of pressure onto their shoulders and cause discomfort (Bowles & Steele, 2013).
Therefore, it is essential to use a well-designed shoulder strap for a good sports bra. Another
cause of discomfort is slipping of the shoulder straps. This problem can be prevented by a good
bra back design. There are primarily five types of bra back designs in the market (see Figure
2.7), including cross-over; racer-back; vertical centre; straight back; and U-back. Among them,
the cross-over, racer-back and vertical centre designs can prevent the shoulder straps from
slipping off the shoulders. The bra underband supports most of the breast weight so it is mostly
made of strong elastic material. The underband should fit the wearer well because an overly
loose underband means that the straps will have to support more of the breast weight, which
causes discomfort. An overly tight underband will induce excessive amounts of pressure onto
the wearer (McGhee, 2008).
26
Figure 2.6 Structure of a sports bra (Yu & Zhou, 2016)
Figure 2.7 Back design of sports bras: (a) cross-over; (b) racer-back; (c) vertical centre; (d)
straight back; and (e) U-back (Yu & Zhou, 2016)
2.3.2 Effects of sports bras
Breast motion can cause more serious health problems for larger breast women. The
effectiveness of reducing their breast motion and discomfort range from wearing an everyday
t-shirt and daily wear bras to supportive sports bras, which has been widely studied in terms of
breast displacement, breast velocity and breast acceleration. This section reviews the literature
on the effects of sports bras during physical activities. Table 2.4 provides the details on some
of the studies in this area. Note however, that the studies that are reviewed in this section focus
on younger women.
27
Table 2.4 Summary of studies on effects of sports bras
Source No. of Age Bra size Bra style Equipment Activity
subjects (years
old)
Lorentzn 59 18 to 60 A to D Eight 16 mm Jumping
and types of Photo-
Lawson, sports sonics
1987 bras action
master
camera
Mason et 3 17 to 21 B and C Sports Two Running
al., 1999 bra, Photo- at 7 and
fashion sonics 10 km/h;
bra, crop biomech- Jumping
top anics 500,
16 mm
cine
cameras
Starr et 6 23 to 37 C to DD Three Video Running
al., 2005 types of camera
sports
bras
28
Mats- 2 22 A and C Sports Two video Jumping
jmoto, bra, cameras
2006 adhesive
bra, daily
bra,
strapless
bra
McGhee 16 19 to 43 C to J Crop top Camcor- Running
and der at 10
Steele, MV600i km/h
2007 digital
video
camera,
Poolcam
video
camera
White et 8 24.8 ± 6.4 D Daily Five Running
al., 2009 wear bra, ProReflex at 10.08
sports infrared km/h
bras cameras
Scut et 15 24 ± 4.8 D Daily Five 100- Running
al., 2009 wear bra, Hz at 5.04
sports bra ProReflex and
infrared 10.08
cameras km/h
29
Scurr et 15 25.1 ± 4.8 D Daily Eight 200- Running
al., 2009 wear bra, Hz Oqus at 10.08
sports bra infrared km/h
cameras
McGhee 20 31± 8 C to F Sports Two Opto- Running
et al., bra, TRAK at 8.3±
2010 experi- 3020 1.3 km/h
ment bra, position
placebo sensors
bra
White et 6 24.8 ± 7.3 D Everyday Five Running
al., 2011 bra and ProReflex at 3.1
sports infrared m/s
bras cameras
Zhou et 11 24 to 40 B to D Sports bra Vicon Running
al., 2012 motion at 7 km/h
analysis
system
Jang et 6 20 B and C Sports Vicon Running
al., 2013 bras motion at 4, 7,
capture and 10
systems km/h
McGhee 15 19 to 44 D+ Sports Two Opto- Running
et al., bra, TRAK at 8-9
2013 3020 km/h
30
fashion Position
bra Sensors
Jamkrajn 30 22.3± 2.7 B and C Sports Eight Running
g, & bras opto- at 3, 6,
Limroonn electronic and 9
greungrat cameras km/h
, 2014
Coltman 23 22.3 ± 2.6 D to E Encap- Optotrak Running
et al., sulation Certus® at
2015 sports bra motion 9.1 ± 0.3
capture km/h
system
Chen et 21 24±2.5 B Sports bra Running
al., 2016 at 5, 7.5,
and 10
km/h
Nolte et 17 22.4±2.9 B and C Sports Qualisys Running
al., 2016 bra, Motion and
partici- Capture jumping
pant’s Analysis
own bra hardware
Arch et 6 26±7 n/a Sports bra 13-camera Running
al., 2018 motion at 6
analysis miles per
system
31
hour (9.7
km/h)
Scurr, White and Hedger (2011) reported the effectiveness of sports bras in reducing the
amplitude of breast displacement during walking and running, and the direction of the breast
displacement is not affected. Scurr, White and Hedger (2010) also examined vertical breast
displacement, velocity, and acceleration with the female subjects running at 2.8 m·s-1 without
a bra, and wearing an everyday bra and a sports bra. They found that breast velocity has the
strongest relationship with comfort. Hence, the improvement of sports bra designs may need
to consider reductions in breast velocity. These studies are all based on observations during
short durations of running on a treadmill. Scurr, White and Hedger (2011) compared breast
support requirements during overground and treadmill running and suggested that the findings
on breast motion during treadmill running are also applicable to overground running.
Nowadays, sports bras are produced with different styles which are designed based on customer
demand, such as nursing sports bras and yoga bras. However, little attention has been paid in
the literature on the effectiveness of sports bras during other kinds of exercise activities.
According to previous studies, sports bras can effectively reduce breast motion and they
recommend that such bras should be worn during physical activities. The prevalence of sports
bra wear among women has been examined by many scholars. For instance, Bowles, Steele
and Munro (2008) surveyed 413 Australian women between 20 and 35 years old on their habits
of sports bra usage. The results indicated less than half of the participants would choose a sports
bras during physical activity and the larger breast women are more than twice as likely to wear
sports bras in comparison to the smaller breast women. McGhee, Steele and Munro (2010)
conducted a survey with physically active adolescent females between 14 and 18 years old and
found that only 13% wear a sports bras while exercising. The reason for the lack of wear could
32
be the discomfort. Lawson and Lorentzen (1990) compared seven sports bras on quantitative
and subjective measures of support and comfort with 59 female subjects during jogging. The
completed questionnaires indicated that the most functional sports bra which can effectively
limit breast motion also tends to be the most uncomfortable one. Bowles, Steele and Munro,
(2012) also found the main features of commercial sports bras that causes discomfort are the
shoulder straps slipping or cutting into the shoulders, and the perceived tightness of the sports
bra around the chest. Although their research only involved Australian women, it is still a
valuable reference for designing sports bras. Further studies are required to improve the
shoulder strap designs of sports bra without affecting the ability of the bra to reduce breast
motion or impeding athletic performance.
2.3.3 Pressure exerted by a sports bra
Pressure is a critical element of this study. The perceived comfort when wearing a sports bra
is associated with the pressure exerted onto the body. The importance of the pressure comfort
of bras is a concern in the field of apparel because bras have direct contact with the skin which
is sensitive. However, only a few studies have actually investigated the pressure comfort of
bras on the breasts (Okabe & Kurokawa, 2004).
The predominant measurement devices to measure pressure values are air-pack pressure
sensors and electrical pressure sensors. Wang, Chen and Lin (2009) used an AMI-3037-5S air-
pack type contact surface pressure measuring system as shown in Figure 2.8. The AMI-3037-
5S has an air sac with a thickness of about 2 mm, sensing parts to induce pressure, voltage
amplifier to process signals, and sensor to output the voltage signal. The precision of the device
is 0 to 14 ± 0.1 kPa and 14 to 34 ± 0.25 kPa (Liu et al., 2013).
33
Figure 2.8 Measurement system of clothing pressure (Wang, Chen & Lin, 2009)
An example of a commercially available electrical pressure sensing system is FlexiForce.
FlexiForce sensors are ultra-thin (0.008 inch) and constructed of a two layer substrate of
polyester/polyamide films. Another common pressure measurement system is the Novel
Pliance system, which is composed of pressure sensors (shown in Figures 2.9a and 2.9b), a
multi-channel analyzer, calibration device and computer software. This system can measure
different points on human skin because of the different sized sensors. Moreover, the elasticity
of the sensors can ensure consistency around contoured areas on the body without wrinkling
(Yu et al., 2014).
34
a. Single sensors b. 2×2 matrix sensor
Figure 2.9 Different types of pressure sensors
The tested points for pressure measurements significantly vary in the literature. Wang, Chen
and Lin (2009) tested four points along the right side of a bra underwire to determine the
pressure values of bra underwires. Morooka et al. (2005) proposed seven points of
measurement, which include the underwire directly inferior to the nipple, front side panel,
bottom of the bra band inferior to the front panel, point of intersection of the bottom of the bra
band and side seam, cross point between the underbust and scapulae lines, back of the wing,
and shoulder straps. Zheng, Yu and Fan (2009) followed Morooka et al. (2005) to select 11
points for a pressure evaluation of the body, which are shown in Figure 2.10.
35
Figure 2.10 Measured pressure points (Zheng, Yu & Fan, 2009)
2.3.4 Pressure comfort evaluation
It is important to investigate the relationship between comfort and clothing pressure to
design effective sports bras. While sports bras can effectively reduce breast displacement
through pressure, an appropriate amount of pressure would provide higher perceived comfort,
a secure feeling and greater support. However, excessive skin pressure may cause discomfort
to the wearer (Yu, 2011)
The range of pressure which is acceptable to wearers has been investigated in previous studies.
Denton (1972) found that the pressure comfort range of a garment is 1.96 to 3.92 kPa and
would differ on different areas of the body. Costantakos and Watkins (1982) found a clear
relationship between clothing pressure and wearer comfort, in which a greater amount of
pressure would result in a rating of lower perceived comfort. They provided the basis to see
how different combined features affect both pressure and comfort. However, they only focused
on the relative perceived comfort with two bras, which would not be able to provide a real
range of pressure comfort values.
Coltman, McGhee and Steele (2015) related the rated discomfort of bras that have different
types of shoulder straps to the pressure exerted onto the shoulders. They found that the cross-
36
back bra, which induces a pressure of 4.9 kPa, can lead to the discomfort score of 2.1 (the
condition that caused the most discomfort and the maximum score for discomfort was 10).
However, there is no significant relationship between discomfort and pressure due to the
inaccuracy of the pressure sensor and multiple influential factors. Gho, Steele and Munro (2008)
found a correlation between bra discomfort and bra pressure. They showed that the bra that
exerted the highest amount of pressure (2.2 N/cm²) can lead to the highest score for discomfort
(5), and the lowest pressure (<0.2 N/cm²) can lead to the lowest score for discomfort (<1).
Although correlation was found, they did not provide a pressure comfort range.
All of the discussed studies have neglected that while pressure causes discomfort, a certain
amount of pressure is necessary for support and a sense of security (Yu, 2011). The relationship
between bra comfort and bra pressure is not supposed to be a linear relationship. Instead, the
minimum pressure comfort might have a value over zero. Liu et al. (2017) used both subjective
fit tests and objective tension tests on bras of the same size and the same materials to conduct
an evaluation on the perceived comfort of the underband. The pressure comfort for a 75B bra
size was found to range from 2.0 – 2.7 kPa.
Liu, Wang and Hong (2017) provided numerically obtained data on clothing pressure which
could be applied to evaluate garment wear comfort. Compared with traditional measurement
methods, the virtual method of measuring clothing pressure is lower in cost, uses less time,
increases efficiency, easily accessible and higher in accuracy.
2.4 Finite element model of breasts
FE methods have been widely used to predict the amount of breast displacement in different
conditions for various applications. Azar, Metaxas and Schnall (2002) developed a deformable
FE model of the breasts based on magnetic resonance (MR) data to predict the position of
37
tumours during the procedure. The use of MR data can reduce calibration errors and ensure
registration quality (Zhang et al., 2007). However, magnetic resonance imaging (MRI) is
relatively expensive and not always necessary. Del Palomar et al. (2008) used X-ray computed
tomography (CT) imaging, which also provides a clear image of the breast geometry, to extract
the fatty and fibroglandular tissue regions.
Some researchers assume that the breasts are a homogenous body and have modelled them
accordingly. Dufaye et al. (2013) presented a virtual deformable model of the breasts of a
female subject with the geometry constructed by using MR data. All of the constituents of the
model of the breasts were set to be quasi-incompressible, isotropic, hyperelastic, and
homogenous. The displacement of the breast surface was calculated by using FE modelling
under the effect of gravity and compared with manually measured displacements of identified
points marked on the skin and 3D scanned images in a standing position. However, their FE
model used linear material properties, which cannot simulate large deformations of the breasts.
Eder et al. (2014) compared 12 different material properties of an FE model of the breasts based
on the previous literature. The results of the FE calculations which were compared to the
corresponding real geometric measures of the breasts from 3D scanned images to determine
the accuracy of the FE simulation showed that the Neo-Hookean hyper-elastic models proposed
by Tanner et al. (2006) and Rajagopal et al. (2008) are the most appropriate models to describe
the viscoelastic behaviour of the breasts under gravity loading among the 12 FE models. The
study also indicated that the mechanical properties of the breast materials can remarkably affect
the simulation results.
Chen et al. (2013) built a 3D non-linear dynamic FE model to simulate the large displacement
of the breasts. The FE model, which uses a hyper-elastic neo-Hookean material, is composed
of a thorax with two breasts and three layers of skin with different mechanical properties
38
(density, Young’s modulus, Poisson’s ratio and shear modulus). They successfully predicted
breast displacement at three different natural frequencies with good accuracy. However, the
Young’s modulus was not validated and the boundary conditions for the walking and running
activities are not accurate due to arm movement.
2.5 Interactions between body and garment
Friction between a garment and human skin can lead to hot spots (warm reddened areas; Brueck,
2000). The adverse consequences of the interaction between a garment and skin could range
from a tickling feeling to the formation of rashes and blisters on the skin (Sulzberger et al.,
1966). Therefore, a more in-depth understanding and investigation of the interactions between
the body and garment under various conditions are very important for overall wellbeing.
It is important to accurately model not only the body but also the garment fabric when
addressing the interaction between the garment and body. Hence, a scientific investigation and
understanding of the fabric mechanical properties are essential in this study. The modelling of
fabrics and textiles by using FE modelling has been well established with the use of both
geometric and mechanical models (Glaessgen et al., 1996; Boisse et al., 1997; Tarfaoui &
Akesbi, 2001; Hamila et al., 2009; Liu, Koric & Kontsos, 2019). The first precise testing
technique of fabric mechanical properties is the Kawabata evaluation systems (KES), which
test the tensile, shear and bending properties, compressibility, and surface properties (i.e.
frictional coefficient). The testing principles are shown in Figure 2.11. Another testing
technique is the Fabric Assurance by Simple Testing (FAST), which is used to measure the
parameters that affect the quality of the fabric and the appearance of the garment (Hu, 2004).
After testing the mechanical properties of the fabric, they can be imported into FE software for
modelling (Tarfaoui & Akesbi, 2001).
39
Figure 2.11 Testing principles of KES (Hu, 2004)
As the aim in this study is to model the physical interaction between a garment and the human
body, it is important to accurately define deformation with computer efficiency (Delingette,
1998). Calculation of the clothing pressure of a human body model with the use of the FE
method has been studied since the 1990s but with shortcomings. For instance, Niwaya, Imaoka
and Shibuya (1996) developed an FE technique to estimate the garment pressure distribution
and reversely mapped this distribution onto a paper pattern (see Figure 2.12 and 2.13) in order
to show the preferred modification. However, this technique did not involve a model that
calculates the pressure of a tightly fitting garment that comes into contact with the body.
40
Figure 2.12 Garment pressure distribution (Niwaya, Imaoka & Shibuya, 1996)
Figure 2.13 Reverse mapping of garment pressure distribution onto paper pattern (Niwaya,
Imaoka & Shibuya, 1996)
Since the human body varies in rigidity depending on the part of the body, the widely-used
simplified methods in the literature presuppose that the garment is worn by a mannequin.
Ishimaru et al. (2011) modelled the human body as a rigid body without taking into account
the deformations of the human body. They proposed a two-step method to fit a garment onto
41
the body by placing the fabric in contact with an intermediary, followed by removal of the
intermediary and placing the fabric in contact with the human body model. This calculation of
clothing pressure is a good example for reference purposes. However, the model is too simple
and the use of a rigid body is not accurate enough to determine the deformation and pressure
from the interaction between the body and garment.
Liu et al. (2007) used a simplified leg model which included soft tissues and bones and a 3D
stock model to investigate the mechanical interaction between the leg and stocking. They
applied an automated surface-to-surface contact option. The developed model showed a good
agreement with reasonable pressure ranges in a comparison with tested pressure values, which
can be a reference for this study. Lin et al. (2011) developed an FE model of a male leg and a
sportswear garment to predict the interface contact pressure distribution. The material
properties were considered to be the same for the leg model, which was 20 kPa for the Young’s
modulus and 1.03 mg/mm3 for the Poisson’s ratio. However, this leg model is oversimplified
and cannot realistically represent the male leg. Xing et al. (2006) provided an FE model on
interactions between the skin and clothing sleeve during arm rotation. The forearm model was
idealized as a cylindrical shape and the entire FE model consisted of a fabric sleeve, skin,
muscle, and bone which approximate the actual structure. However, the arm and leg structures
are very different from the breasts which are composed of skin, fat and glandular tissue.
There are studies that have investigated the interaction between the human body and a sports
bra by using FE analysis. For instance Li, Zhang and Yeung (2003) presented a 3D
biomechanical model of a female body and a bra (shown in Figure 2.14) to determine the effects
of the bra on the dynamic deformation of the breasts while the subject was walking at a constant
speed. Only the breast is elastic and the other parts are rigid, which is not true in real life
situations. They also did not consider gravity, which means that there is no initial stress of the
bra in their model. Bel-Brunon et al. (2014) studied the influence of a sports bra on the stress
42
and strain of the breasts with an FE model by taking images of the breasts under water to model
a non-gravity configuration of the breasts. However, they used single-layer and single-material
bra models which cannot be used to evaluate the tested results of commercial bras as they have
at least two layers with different material properties.
Figure 2.14 FE model of female body and bra (Li, Zhang & Yeung, 2003)
2.6 Machine learning approach
The previous sections have reviewed many studies in the literature that use FE modelling to
describe the biomechanical behaviours of breasts and interactions between the breasts and bras
with great accuracy. However, building an FE model is very time consuming and usually
subject specific. This limits the practicality of using an FE model for product design. In
particular, traditional sports bra designs are very complex and prone to human error. Hence, an
ML approach is considered in this study.
ML has been studied since the late 1950s (Martens, 1959). It is considered to be artificial
intelligence (AI), which means that the computer (machine) has the ability to automatically
learn and extract an algorithm for the task, even though identifying the complete process may
not be possible (Alpaydin, 2009). The generally accepted definition of ML is: “A computer
43
program is said to learn from experience E with respect to some class of tasks T and
performance measure P, if its performance at tasks in T, as measured by P, improves with
experience E”. (Michalski, Carbonell & Mitchell, 1985) Currently, ML is used in a wide variety
of applications. (Mullainathan & Spiess, 2017; Bottou, Curtis & Nocedal, 2018; Atta et al.,
2019; Huang, Kwok & Zhou, 2019). Although this area does not have standardized
classifications, there are four classifications of ML algorithms that are mainly used: supervised,
unsupervised, semi-supervised, and reinforcement learning. In supervised learning (Kotsiantis,
2007; Zhang & Tsai, 2006), the computer is given a data set and already knows the correct
output, as the computer already understands that there is a relationship between the input and
output. Unsupervised learning (Celebi & Aydin, 2016) can be used to approach problems with
little or no idea of the results, and derive structure from data in which there is not necessarily
an understanding of the effect of the variables. Semi-supervised learning (Chapelle, Schölkopf
& Zien, 2006) is used when algorithms work with a given data set with missing information,
and there is still the need to learn from large amounts of received data. Finally, there is
reinforcement learning, which is used when learning is based on external feedback given either
by a thinking entity or the environment.
Today, there are countless ML algorithms in the literature and the number of ML algorithms is
still growing fast with the development of computer technology. Recently, ML approaches have
been proposed in the literature to build models that predict the biomechanical behaviors of
breasts, which are mainly used for breast cancer diagnoses. Kharya, Dubey and Soni (2013)
summarized dozens of academic papers on the diagnosis and prognosis of breast cancer by
using ML techniques. They provided the benefits and limitations of four ML approaches:
decision trees, artificial neural networks (ANNs), support vector machines (SWMs), and
Bayesian networks. ANNs are mostly used as a predictive technique that replicates human
thinking and often used in medical predictions as compared to decision trees which are a tool
44
that resembles a tree of the decision and all possible outcomes. SVMs are mainly used in
computational biology such as microarray data analysis, to transform data and categorize new
examples. Finally, Bayesian networks are suitable for carrying out predictions under uncertain
situations with incomplete data. Shan et al. (2016) also used four ML approaches which include
decision trees, ANNs, SVMs, and random forests to develop a computer-aided diagnosis
system for 283 ultrasound images of breasts. They compared the results and found that SVM
has the best performance. However, ML techniques have not been clinically applied, even
though the techniques have developed rapidly in recent years. This is mainly because of the
lack of standardization and requirement of large datasets (Reig et al., 2019). Therefore,
applications of ML techniques still need to be studied in greater depth.
2.7 Motion capture experiment
The aim of motion capture experiments is to record the movement of specific objects in a
concise and usable manner. To record the movement of breasts, female subjects need to wear
reflective markers to capture the motion by identifying the position or angle between markers.
Previous studies in the literature on breast motion are important because they have already (i)
evaluated the research methods and identified the problems of studying the dynamic movement
of breasts; (ii) compared breast displacement, velocities, accelerations and trajectories under
braless conditions and when a bra is donned to evaluate the effect of sports bras; and (iii)
determined future research directions for breast motion studies (Zhou, 2011).
2.7.1 Motion capture equipment
To record breast motion, it is convenient to use standard cameras because they are low in cost
and easily accessible. Haycock, Shierman and Gillette (1978) used a movie camera to shoot
45
images on 16 mm film and calculated breast motion through the images. Only one camera is
necessary to capture the two-dimensional (2D) Cartesian coordinates of the markers on the
subject. To acquire the 3D coordinates of the markers, Mason, Page and Fallon (1999) used
two 16 mm high speed cine cameras that filmed at 100 frames per second and 3D digitizers or
charge-coupled devices (CCDs). The cameras were placed 10 m directly to the front and the
left of the subject to shoot the frontal and side views of the breasts. An event indicator, which
could be activated each time that the right foot of the subject made contact with the treadmill,
was also set up on the side of the subject in order to acquire the rhythm of the stride. Video
cameras are also widely used. McGhee, Power and Steele (2007) placed a digital video camera
2.5 m directly in front of the subject to analyse exercise-induced vertical breast displacement.
Fuseya and Matsumoto (2006) obtained 3D vibration data of the breasts by using two video
cameras. McGhee and Steele (2010) used two position sensors positioned 3.6 m away from the
subject in the frontal and sagittal planes, which obtained 3D data on breast motion. Furthermore,
this motion capture system was already pre-configured and pre-calibrated, which reduced the
time needed to conduct the experiment. Starr et al. (2005) collected motion data by using video
cameras which were then stored and analyzed frame-by-frame by using the Peak Motus®
Motion Measurement System. With the advancement of motion capture technology, analysis
systems and software, such as Vicon and ProReflex, are user-friendly resources for processing
raw experimental data.
Breast displacement is studied by capturing the displacement found by using reflective circular
markers that are placed onto the body of subjects (Starr et al., 2005). These retro-reflective
markers are passive, and easily adhered onto the skin of subjects or surface of sports bras.
Haake and Scurr (2010) placed hemispherical retro-reflective markers with a 5-mm radius on
the right nipple of their subjects, as well as on the superior border of the manubrium of the
sternum, and the left and right anterior inferior aspects of the 10th ribs, which are shown in
46
Figure 2.15. However, McGhee and Steele (2010) used infrared light-emitting diodes with a 2-
mm diameter instead of reflective markers. The 3D coordinates of all of the diodes placed on
both nipples and other sarco-iliac joints of the subject were tracked by using two OptoTARAK
3020 position sensors positioned 2.6 m away from the subject in the frontal and sagittal planes.
Figure 2.15 Markers to measure breast motion (Haake & Scurr, 2010)
2.7.2 Studied points
Most studies in the literature have focused on the nipples (Zhou, Yu & Ng, 2011). However,
nipple movement is not accurate enough to represent breast motion because breasts have
complex structures which have viscous damping and hyper-elastic behaviours. Hence, Okabe
(2002) analysed 5 to 25 markers placed on the breasts. Moreover, Zhou, Yu and Ng (2012)
used four breast markers attached to the inner-breast (BI), outer-breast (BO), left nipple (BL)
and right nipple (BR) of the subject. The positions of the markers in their study are shown in
Figure 2.16.
47
Figure 2. 16 Marker positions to measure breast motion (BI: inner-breast , BO: outer-breast,
BL: left nipple and BR: right nipple) (Zhou, Yu & Ng, 2012)
2.7.3 Reference system
It is also important to define a reference object and determine the number and positions of body
reference markers to calculate the relative movement of the breasts. However, the optimal
number and location of the reference markers that would adequately represent the thorax
remain unclear (Zhou, Yu & Ng, 2012). According to Della Croce et al. (2005), human
movement analyses in principle can rely on any anatomical based reference frame. However,
the ideal reference frame should be based on easily identifiable and traceable landmarks (Conti
et al., 2008). Veeger et al. (2003) suggested bony landmarks above bony ridges, such as the
scapular spine or the humeral shaft, to compose the reference frame. McGhee, Power and Steele
(2007) only used one marker to determine the torso displacement. However, the motion of a
rigid body in 3D space is described by 6 degrees of freedom movements and rotations about
three perpendicular axes. Therefore, one landmark on the body of a subject cannot be used as
a proper reference system. Starr et al. (2005) added two more markers to describe torso
48
displacement. White, Scurr and Smith (2009) created a reference grid for markers placed on of
the left and right clavicles and anterior superior iliac spines. In addition, Scurr, White and
Hedger (2009) used four reference points to describe body motion more accurately. The body
landmarks which are used in 17 studies are shown in Figure 2.17 and listed in Table 2.5 (Zhou,
2011).
Figure 2.17 Reference body landmarks (Zhou, 2011)
49
Table 2.5 Reference body landmarks (Zhou, 2011)
Landmark Description of anatomical landmark
a Left clavicle center or clavicle directly superior to the nipple
b Lateral point of acromion processes
c Point on front arm
d Sternum, jugular, or suprasternal notch
e Sternal angle, or sternum at the 3rd rib level
f Lower sternum, or cross point of underbust line and body median
g Iliac crests
h Anterior superior iliac spines
i Anterior inferior aspects of the 10th rib
j Spinous process of the 12th thoracic vertebra
k 5th lumbar vertebra
Although these landmarks are easy identifiable, there is one limitation which is the lack of
accuracy. For example, the pelvis would rotate when the subject is running and the position of
the clavicle center would change when the arm is swinging, all of which would affect the
accuracy of the marker placed on that body part.
2.8 Conclusion
The literature review provided above reveals the following knowledge gaps:
1). The female breasts are viscoelastic, anisotropic, and inhomogeneous with nonlinear
behavior. However, to simplify the problem, the breasts are always assumed to be isotropic
50
and homogenous. The lack of studies on the nonlinear hyperelastic parameters of different
breast tissues is obvious, and the range of measured Young’s modulus values of the breasts
is still large, or from 0.5 to 50 kPa.
2) There are a lack of studies that evaluate the comfort of wearing a sports bra and its
effectiveness during different sports activities. There are different styles of sports bras,
which are designed based on customer demand. However, women have different perceived
comfort and pressure. It is therefore difficult to quantify the level of comfort. Previous
studies have mainly used experiments in which treadmill running is carried out, so
evaluations during other kinds of sporting activities should be investigated.
3) Existing FE body and garment models are not applicable to the real human body and
garments. This is a widely-used simplified method which presupposes that the garment is
donned onto a mannequin. However, a rigid body is not accurate enough for calculating the
deformation and pressure from the interaction between the body and garment. The human
body has subcutaneous adipose tissue which is not rigid, and these soft tissues can reduce
the contact pressure.
4) Gravity load has also been neglected by many researchers, which means there is no initial
pressure of the bras in their modelling.
5) The ML approach is widely used in many areas because it is efficient and accurate.
However, it is rarely used in intimate apparel production and design.
6) The FE models of single-layer and single-material bras in previous studies cannot be
used to evaluate the tested results of commercial bras as they have at least two layers with
different material properties.
Therefore, this study aims to develop a novel and valid method to investigate the nonlinear
material parameters of the breasts and model the contact conditions between the body and a
51
compression sports bra. With precise FE models, any kind of physical activity can be analyzed.
Answers to a questionnaire by subjects on the level of comfort of a sports bra are collected and
organized. Using both experimental work and computer modelling, this study intends to
provide useful information for future designs of sports bras.
52
Chapter 3 Development of finite element model of body
3.1 Introduction
The coupling of finite element (FE) modelling and practical experiments to build precise
numerical models has been widely done in the engineering disciplines to validate the
behaviours, relationships and properties of both living and non-living entities, such as beams
(Ghaffarianjam & Abolbashari, 2010), manufactured biomaterials (Simoneau et al., 2017),
human muscles (Wu et al., 2013) and bone structures (Jang & Kim, 2010). Recently,
advancements in computational modelling have extended FE modelling to the breasts, which
is commonly done in clinical or industry practices, such as detecting breast anomalies (Xi et
al., 2012; Jin, He & Liu, 2014), predicting breast displacement (Kuhlmann et al., 2013; Chen
at al., 2013), and assisting with breast surgery (Patete et al., 2013). However, the mechanical
behaviours of soft tissues, such as the breasts, are in fact more complex than most engineering
materials and structures because breasts have viscoelastic and anisotropic properties (Freutel
et al., 2014). Hence, it is a complex task to determine the material coefficients of breasts
because of their hyper-elasticity and non-linear behaviour.
In this study, two experiments have been conducted to build a precise subject-specific FE
model of the body. Geometric models of the bodies and breasts are constructed based on
experiments carried out with 3D laser scanning. Then, the effects of gravity are removed for
the initial state of the breasts by subjecting the model to an inverse gravity load. The dynamic
state of the free vibration of the breasts and running are recorded by motion capture to calculate
and validate the material coefficients of the breasts.
53
3.2 Subject-specific FE model of body obtained by 3D laser scanning
3.2.1 Subjects
Women with large and hypertrophic breasts face more difficulties because their breasts jiggle
and move, thus interfering with their ability to take part in daily and physical activities
(Coltman et al., 2019). In this study, fifteen female subjects are recruited on a voluntary basis
to obtain geometric models of the female body and breasts. The selection criteria are that they
have to be healthy, have not underwent any previous breast surgeries, and are not pregnant or
lactating. The youngest subject is 19 years old and the oldest subject is 60 years old. Their bra
cup size includes 32B (three subjects), 32C (three subjects), 34C (three subjects), 34D (three
subjects), 36C (one subject), 38C (one subject), and 38D (one subject). Their demographics
and body measurements are listed in Table 3.1. The subjects were informed about the purpose
of the study, and then had to give informed consent form before taking part in the experiment.
The experimental procedure was approved by the Human Ethics Committee of the Hong Kong
Polytechnic University (HSEARS20151207004).
54
Table 3.1 Demographics and body measurements of subjects
Horizontal cup line (cm) Breast root (cm) Cup size

Height Weight Full bust Under bust
Subject Age BMI (commercial
(cm) (kg) L R L R (cm) (cm)
bra)
S01 50 163.0 55.5 20.9 23.5 23.8 26.0 25.7 95.0 74.0 34D
S02 26 168.0 70.7 25.0 23.0 22.8 26.0 25.8 97.9 83.2 38C
S03 26 159.0 57.3 22.7 22.5 22.7 23.5 23.0 81.7 75.0 34C
S04 33 153.0 53.5 22.9 21.0 20.5 26.0 25.3 85.2 70.6 32C
S05 24 162.0 52.3 20.0 17.5 18.0 19.0 20.2 82.2 72.0 32B
S06 60 148.0 50.3 23 19.5 20.3 24.1 23.8 82.5 71.8 32C
S07 32 160.0 51.5 20.1 22.0 21.7 22.5 22.7 81.8 72.5 32B
S08 59 153.0 63.0 26.9 21.6 21.7 24.0 23.8 97.5 82.5 38D
S09 54 154.0 55.5 23.4 19.4 19.0 25.8 26.0 89.5 75.5 34C
S10 19 162.0 61.9 23.6 22.5 24.6 27.5 27.5 92.7 72.7 34D
S11 59 162.0 70.9 27.0 20.8 28.5 27.0 24.1 90.0 79.1 32C
S12 54 161.0 61.2 23.6 19.8 20.0 25.5 25.0 92.9 77.0 36C
55
S13 23 160.0 60 23.4 21.5 21.8 24.8 20.8 91.8 76.0 34D
S14 24 162.0 52.3 19.9 18.0 17.0 22.5 22.0 79.7 69.0 32B
S15 25 159.0 55.4 21.9 21.6 21.8 27.5 27.3 88.0 73.9 34C
Average 37.9 159.1 58.1 23.0 21.0 21.6 24.8 24.2 88.6 75.0
Standard
15.9 5.0 6.5 2.2 1.8 2.8 2.2 2.2 6.1 4.1
deviation
56
3.2.2 3D laser scanning
The subjects underwent scanning with a 3D laser body scanner (Vitus, Human Solutions, Germany)
which was set to scan at a high resolution of 300 pixels/cm2 to construct geometric models of their
body and breasts. The body scanner has a control unit and 4 sets of cameras which can move to capture
the full image of the subjects. The subjects were to stand upright with both arms open and breathe
naturally during the scanning process. The scanned data clouds were saved in ASCII format and
processed with an image processing software (Rapidform XOR, INUS Technology, Seoul, Korea).
Before the scanning operation was carried out, ten 5-mm markers were placed onto the skin of the
subject to define the boundaries of the breasts. The marker positions are shown in Figure 3.1 and listed
in Table 3.2.
Figure 3.1 Marker positions for defining boundaries of breast
57
Table 3.2 Description of markers for defining boundaries of breast
Marker Position
Marker 1 Lateral point of the breast
Marker 2 Inferior point of the breast
Marker 3 Inner-most point of the breast
Marker 4 Point between Markers 1 and 2
Marker 5 Point between Markers 2 and 3
Marker 6 Upper-most point of the breast
Markers 7, 8 Point inserted between Markers 2 and 6
Markers 9,
Point inserted between Markers 1 and 6
10
After adhering the markers onto the subjects, the subjects were scanned in the braless condition and
while wearing a sports bra. This was repeated three times. A sample of the scanned image is shown in
Figure 3.2.
58
Figure 3.2 3D body scanning of braless subject
3.3 Motion capturing

Twelve digital cameras (Eagle Motion Analysis Corporation, USA) were used to record the dynamic
movement of the breasts of the subjects. One of the cameras is shown in Figure 3.3. Movement was
recorded during running and breast vibration. Figure 3.4 shows the experimental setup in the laboratory.
A treadmill was placed at the centre of the room and 12 digital cameras were hung close to the celling
and surrounding the treadmill.
59
Figure 3.3 Digital camera for motion capture (Motion Analysis, n.d.)
Figure 3.4 Floor plan of motion capture experiment
To determine the material coefficients of the breasts, the captured motion was running braless on a
treadmill at a constant speed. Therefore, the breast displacement only depended on the running speed,
and material and geometric properties of the breasts. Spherical passive retro-reflective 12.5-mm
markers were placed onto the skin of the subject which reflect infrared light to capture the
displacements. It is worth noting that these markers are different from the markers used in the 3D
60
scanning experiment, which are just semi-spherical and placed on the skin of the subject to help to
identify the boundaries of the breasts while processing the geometric model. The retro-reflective
markers cannot be used for the 3D scanning experiment, as they cannot be scanned. Eleven markers
were used in the motion capture experiment, which are shown in Figure 3.5. One of the markers was
placed on the torso of the subject to record the movement of the subject. The other markers were placed
onto the breasts.
Figure 3.5 Position of markers on breasts - 1: sternum; 3: right nipple; 8: left nipple; and 2, 4, 5, and
6 are placed 4 cm away from 3 and 8 (Liang et al., 2019)
The first step of the experiment was the calibrate the cameras. A wand wave was done by waving a T-
shaped wand (see Figure 3.6). The positioning of the T-shaped wand during calibration allows the
direction of possible subject movement to be specified by the retro-reflective markers in three
directions: left and right (x coordinate plane), up forward and backward (y coordinate plane), and up
and down (z coordinate plane) which was then calibrated in EVaRT (version 5) software (Motion
Analysis Corporation, USA) (Park et al., 2014). The initial posture for the subject was to stand still
61
and look straight ahead. Then, the subject was asked to raise her breasts gently with her hands and hold
them still. Next, she was to quickly let go of her breasts so that they vibrated from free falling due to
gravity load and damping forces (Cai et al., 2018). Free vibration takes place when a force is exerted
onto a structure or part and disturbed momentarily, and then allowed to move freely with no restraint.
In this case, the focus is the free vibration of the breasts due to their own elasticity. That is, the breasts
are falling freely due to gravity load (Landau & Lifshitz, 1969). These vibrations then gradually reduce
or decay, their frequency or intensity changes, or all together stops, and the structure or part (and in
this case, the breasts) returns to static equilibrium; see Figure 3.7 (Chen et al., 2013). Damping ratio
is to measure how the vibrations gradually reduce until the system rests in its equilibrium position.
Here free-vibration testing of the breasts allows the digital cameras to capture the information to plot
the 3D time-displacement of the nipples. It is also important to measure how much the vibrations decay
or are reduced from one bounce of the breasts to the next, which is measured by using the damping
ratio. The damping ratio is then inputted into FE software to simulate viscosity, which is important for
describing the nonlinear behaviour of the breasts. To obtain the damping ratio of the breasts, the
displacement of the breasts was measured during the free vibration tests with two markers attached to
each of the nipples, in which the 3D time-dependent displacement was recorded and then also
processed with EVaRT before calculating the damping ratio of each subject.
Secondly, the braless subject was asked to stand upright and still with both arms open. Her body was
in a ‘T’ position as the initial static condition of each movement cycle. Thirdly, the subject had to run
on the treadmill until she reached a steady speed. The time-dependent coordinates of the eleven
markers were recorded in terms of the x-mediolateral (+ve left), y-vertical (+ve up), and z-
anteroposterior (+ve front) directions. The motion capture results were filtered, cleaned and smoothed
by using the EVaRT software to remove noise before analysing the dynamic breast displacement in
the braless condition during running.
62
Figure 3.6 T-shaped wand used with retro-reflective markers
Figure 3.7. Free vibration of breast
63
3.4 Finite element model
3.4.1 Geometric modelling
The raw scanned point cloud of the body is irregular and has holes due to missing scanned data. To
address this problem, Geomagic® software (Geomagic Studio 12, USA) was used to fill the holes and
smooth the surface of the geometric model. Geomagic® is a reverse engineering software that creates
3D parametric models. Its ability to automatically trim, join, cut or subtract surfaces and solids makes
it simple to comprehensively process geometric models.
In this study, some of the body parts, such as the legs and arms, have no impacts for analysing breast
motion. Hence, the first step in processing the geometric model was to remove these irrelevant body
parts (as shown in Figure 3.8). The geometric model then comprises two breasts, 10-mm of
subcutaneous tissue and a torso without the internal components of the breast, such as the fat tissues
and glands. The two breasts were separated from the body by using Boolean operations according to
the determined breast boundaries. The torso was built through offsetting operations on the body
without the two breasts and the 10-mm of subcutaneous tissue was also built through a Boolean
operation. The operations were realized with Rapidform XOR software (Rapidform XOR, INUS
Technology, Seoul, Korea). Rapidform XOR is a software package that allows both 3D scanned data
processing and parametric solid modelling to give users the ability to create intelligent geometric
models from point clouds and polygon meshes. In practice, after importing the scanned data into
Geomagic Studio 12, we first used the build-in function “Fill Single” with a setting named “curvature”
to fill holes that are within the region of interest of our later analysis. The function requires manual
selection of the holes that need to be filled. For each selected hole, the software automatically generates
new mesh points within the hole such that the new mesh matches the curvature of the surrounding
64
mesh given by the scanned raw data. After filling the holes, we smoothed the surface of the model
using another build-in function “Remove Spikes”, which is designed to find the single-point spikes on
the given mesh and flatten them. Figure 3.2 and Figure 3.8 show our scanned raw data and the model
obtained after the hole-filling and smoothing processes respectively. As expected, both processes do
not alter the qualitative features of the scanned object, whereas they reduce the noises caused by the
imperfection of the scanning process significantly.
Figure 3.8 Initial geometric model
3.4.2 Mesh creation
An FE software, Marc 2014.2.0 (MSC Software Corporation, US), is used in this study, which has
excellent capability of simulating the nonlinear behaviour of FE models. Three-dimensional 4-node
tetrahedral solid elements (Element 157) which have quadratic displacement behaviour and can model
irregular meshes were used to model the breasts and body. The element form and meshes of the model
of the body are shown in Figures 3.9 and 3.10, respectively.
65
Figure 3.9 Element 157 Figure 3.10 Mesh of model of body
Previous studies have proven that 4-node tetrahedral elements can simulate large deformations of the
female breasts (Martínez-Martínez et al., 2017; Sun et al., 2019; Hsu et al., 2011). The FE model of
Subject S01 was first established. As 4-node tetrahedral elements have the problem of being overly
stiff, 10-node quadratic elements were also considered for building the FE model. However, the model
would have had 203,066 nodes in comparison to the 32,159 nodes with 4-node tetrahedral elements,
which is a very large number of nodes and would require more computation time. Finally, the accuracy
of the calculated results is defined by the root mean square error (RMSE), which is calculated by using
Equation (3.1):
1 ∆𝑌𝑌𝑒𝑒𝑒𝑒𝑒𝑒,𝑖𝑖 −∆𝑌𝑌𝐹𝐹𝐹𝐹𝐹𝐹,𝑖𝑖 2
RMSE =
𝑛𝑛
�∑𝑛𝑛𝑖𝑖=1( ∆𝑌𝑌𝑒𝑒𝑒𝑒𝑒𝑒,𝑖𝑖
) (3.1)
where ∆Yexp is the experimental change of displacement of the subject’s nipple in the y-direction at
every frame, ∆YFEM is the change of displacement of the subject’s nipple in the y-direction from the
FE analysis result at every frame s, and n is the number of sample data points.
Equation (3.1) only considers the displacement in the y-direction. During treadmill running in this
study, there are displacements in the x- and z-directions. However, they are smaller in comparison to
66
the predominant displacement in the y-direction. Also, there are relatively more errors in the x- and z-
directions due to shoulder rotation and body waggle. Thus, the displacements in the x- and z-directions
were ignored. The calculated RMSE between the FE results and experimental data with the use of 4-
node elements was 0.5942%, and the RMSE with the use of 10-node elements was 0.4897%. Therefore,
the computation time would be longer if 10-node elements are used and the accuracy of the results
would not be much significantly improved.
In order to examine the mesh sensitivity of the FE model, five FE models with different mesh element
sizes (2 mm, 4 mm, 6 mm, 8 mm and 10 mm) were built to determine the influence of the mesh density.
It was observed that there is only a 0.26% variation in the dynamic breast displacement between the
results with the use of 4 mm elements and 6 mm elements. However, the calculation time for the model
with 4 mm elements is 3 times more than that with the 6 mm elements. Therefore, 6 mm elements were
used for both accuracy and efficiency of computation time.
Table 3.3 Results of different mesh sizes
Number of elements RMSE of breast

Computation
Mesh size displacement
Body Breasts time
during running
2 mm 1,986,290 385,301 About 1 month
4 mm 302,501 56,970 About 6 days 0.5349%
6 mm 119,510 22,931 About 2 days 0.5942%
8 mm 49,880 11,086 25.5 h 0.9319%
10 mm 23,916 5870 13.5 h 1.0241%
67
From the mesh sensitivity study and comparison of the two element types, 4-node tetrahedral
elements that are 6 mm in size were used to mesh the geometric model of the female subjects.
3.4.3 Coefficients of Mooney-Rivlin material
To simulate breast movement accurately, it is critical to precisely determine the material parameters.
Following previous studies (Sun et al., 2019; Liang et al., 2019; Samani et al., 2003), a Mooney-Rivlin
material model with 5 coefficients (C01, C02, C10, C11, and C20) was used to construct the breasts with
Marc 2014.2.0 (MSC Software Corporation, US). This material model has been proven to have the
ability to simulate large deformations and nonlinear mechanical behaviours of soft tissues.
The generalized Mooney-Rivlin polynomial function of strain energy, which results in a better
agreement with the test data for both unfilled as well as filled rubber in previous studies, can be written
as:
𝑁𝑁 𝑁𝑁
𝑖𝑖 𝑗𝑗
1 (3.2)
W = � 𝐶𝐶𝑖𝑖𝑖𝑖 (𝐼𝐼1 − 3) (𝐼𝐼2 − 3) + � (𝐼𝐼 − 1)2 𝑖𝑖
𝐷𝐷𝑖𝑖 3
𝑖𝑖,𝑗𝑗=1 𝑖𝑖−1
where W is the strain energy potential; I3 is the elastic volume ratio or third strain invariant; Cij is a
factor related to the shear behaviour of the material; Di is the compressibility behaviour of the material;
and N is order of the polynomial. I1 and I2 are the first and second strain invariants of the components
of the left Cauchy-Green deformation tensor B, which is written in Equations 3.3 and 3.4.
68
I1 = tr(B) (3.3)
1
I2 = [(tr(B))2 − tr(B2 )] (3.4)
2
where B=F∙FT, and F is the deformation gradient.
To accurately describe the breasts, 5 coefficients (C01, C02, C10, C11, and C20), were used to define the
Mooney-Rivlin material properties. A series of computing analyses were carried out to determine the
most appropriate coefficients. The boundary conditions of these analyses were gravity loaded to the
breasts and the boundary displacement was extracted from the torso part based on the motion capture
experiment. The steps of dynamic analysis were shown in Figure 3.11. The first step was to include a
time dependent displacement history using the table option in the boundary condition menu, and the
displacement was based on motion capture experiment as we mentioned. Then, the second step was
creating the loadcase for the dynamic analysis. The next step was creating the job and submitting it.
The final step was postprocessing the results by opening the post file and reviewing the results.
69
Figure 3.11 Steps of dynamic analysis
The differences between the displacement of the bare breast condition in the experiment and that in
the FE analysis results were compared and minimized. The initial sets of the coefficients of the
Mooney-Rivlin material model for the breasts and layer of subcutaneous tissues were based on those
in Samani et al. (2003).
Table 3.4 Material coefficients (Samani et al. 2003)
Component of model Coefficient Value/ kPa
C01 0.31
C02 3.8
Breasts C10 0.3
C11 2.25
C20 4.72
C01 5
Layer of subcutaneous tissue
C10 5
70
Three steps were carried out to obtain the optimal material coefficients of the breasts. The first step
was to determine the optimal range of material coefficients for the FE model of the body without a
sports bra. The FE model that uses the Mooney-Rivlin material with smaller coefficients should show
a smaller amplitude during vibrating. It was observed that the experimentally measured amplitude of
the breast motion is 10 mm less than the simulated amplitude by using coefficients that are 5 times
smaller than the initial set of coefficients (was C01 = 0.060 kPa, C02 = 0.944 kPa, C10 = 0.062 kPa, C11
= 0.450 kPa, and C20 = 0.160 kPa), and 4 mm larger than the simulated amplitude by using coefficients
that are 2 times smaller (was C01 = 0.150 kPa, C02 = 2.360 kPa, C10 = 0.155 kPa, C11 = 1.125 kPa, and
C20 = 1.900 kPa) than the initial set of coefficients. Therefore, the optimal material coefficients should
be in this range. The second step was to input the coefficients of the breasts that were within this range
into the FE model. A compensation factor α was defined to divide the initial set of material coefficients.
The corresponding RMSEs are shown in Table 3.5. The third step was to find the minimized RMSE
result for the different material coefficients. Table 3.5 shows that the optimal material coefficients are
close to the coefficients when the compensation factor is 3. By adjusting the coefficients, the set with
C01 = 0.108 kPa, C02 = 1.18 kPa, C10 = 0.094 kPa, C11 = 0.82 kPa, and C20 = 0.84 kPa has the lowest
RMSE, which is 0.0405%. Therefore, this set is selected for the optimal material coefficients of the
breasts. With appropriate material coefficients, the dynamic displacements of the braless breasts are
very similar to those of the experimental data, which is shown in Figure 3.12.
71
Table 3.5 RMSEs of different material coefficients
Compensation
5 4.5 4 3.5 3 2.5 2
factor α
RMSE /% 3.34 1.68 1.39 0.32 0.21 2.86 6.77
Figure 3.12 Dynamic displacements of breasts: experiment vs. FEM
The same method was applied for each subject, and the results of Mooney-Rivlin material coefficients
are shown in Table 3.6. According to Chen et al. (2013), when the RMSE is less than 1%, the error is
acceptable and the corresponding material coefficients are selected.
72
Table 3.6 Material coefficients of different subjects and RMSEs
Subject C10/ kPa C01/ kPa C11/ kPa C20/ kPa C02/ kPa RMSE %
S01 0.094 0.108 0.820 1.180 0.840 0.045
S02 0.103 0.103 0.820 0.860 0.860 0.152
S03 0.102 0.082 0.820 0.820 0.840 0.132
S04 0.090 0.098 0.940 1.280 0.940 0.297
S05 0.109 0.099 0.920 0.980 0.940 0.110
S06 0.074 0.088 0.820 1.180 0.840 0.069
S07 0.104 0.108 0.620 0.640 0.980 0.140
S08 0.086 0.084 0.820 0.830 0.920 0.180
S09 0.084 0.100 0.620 1.080 0.810 0.290
S10 0.102 0.103 0.820 0.870 0.920 0.162
S11 0.094 0.100 0.620 0.580 0.560 0.220
S12 0.062 0.064 0.830 0.640 0.540 0.120
S13 0.096 0.102 1.010 1.030 1.000 0.170
S14 0.105 0.089 0.940 0.980 0.102 0.150
S15 0.135 0.162 0.930 2.430 0.870 0.070
3.4.4 Damping ratio
A typical time vs. displacement graph of the free-vibration of the breasts that shows the motion decay
of this viscous material is provided in Figure 3.13.
73
Figure 3.13 Time vs. displacement of breasts
The viscous damping ratio can be calculated by using a logarithmic decrement equation. The
logarithmic decrement δ is the natural log of the amplitude ratio of two adjacent wavelengths of
displacement in the free decay of vibration, and the equation can be written as:
𝑦𝑦𝑛𝑛
δ = ln ( ) (3.5)
𝑦𝑦𝑛𝑛+1
where yn is defined in the time-displacement curve of motion decay in Figure 3.10. In this study, n is
equal to 1.
The damping ratio ζ can be calculated with the following equation:
74
δ
ζ= (3.6)
�(2𝜋𝜋)2 +δ2
The breasts were lifted to different initial heights so that 3 rounds of free-vibration testing were
conducted. It was necessary to conduct more than one round of testing because it is difficult to capture
the damping movement of the breasts due to their limitations in volume. Time vs. displacement was
plotted from the results of the free-vibration tests for the left nipple of S01 which were captured by the
infrared cameras. To ensure the reliability of the data, two curves were selected to calculate the
damping ratio of the breasts, which are shown in Figure 3.14.
75
Braless Damping
Time-displacement curve
condition ratio
1 0.289
2 0.257
Figure 3.14 Time-displacement curves of braless condition and calculated damping ratios
According to Equations 3.5 and 3.6, the damping ratio of the braless breasts can be calculated by using
the time-displacement curves of the breast vibration. The mean value of the viscous damping ratio is
confirmed to be 0.273.
Occasionally, the subjects were slightly shaking, so most of the plotted experimental results are similar
to those of S13 in Figure 3.15. The relative position z does not become a constant after the first peak
of oscillation, which means that the subject was not standing still for a long enough time. This problem
76
was found in every performed experiment. Therefore, it is not possible to determine the equilibrium
position by plotting this kind of experimental curve with Equations 3.5 and 3.6.
Figure 3.15 Experimental curve of free vibration for S13
In this work, the value in which the acceleration becomes 0 where the first peak declines is used.
Mathematically, the spring is at rest in the equilibrium position. Then, the equation of motion becomes
Equation 3.1, which is not zero. However, according to Cai et al. (2018), the spring constant for the
lower part of the breasts is large because this part is denser than the upper part of the breasts. Thus, a
small deformation of the lower part of the breasts can balance the right part of Equation 3.1 and render
the acceleration equal to zero. In other words, this small deformation acts as an evaluation of the
uncertainty of this method in finding the equilibrium position. In addition, the area of decline after the
first peak is chosen because there is the lowest speed of vibration in this area.
77
mÿ = −cẏ (3.1)
Figure 3.16 is a magnification of the experimental curve and the calculated acceleration, 𝑦𝑦̈ . In this
figure, the dashed black line shows the time when the acceleration is equal to zero during the decline
of the first peak. The dashed blue line shows the displacement value at the selected equilibrium time,
which is the estimated equilibrium position. For this experimental data set, the equilibrium position is
-166.9 mm. After determining the equilibrium position, the damping ratio of the breasts can be
calculated to build the FE model. To evaluate the uncertainty of this method, the damping ratio and
spring constant in Cai et al. (2018) were applied and the small deformation mentioned before is
calculated to be -0.45 mm. In comparison to the overall displacement of the breasts, this uncertainty
is an acceptable value.
78
Figure 3.16 Magnified experimental curve
3.5 Conclusions
This chapter has systematically presented the research methods and results to carry out the first
objective of the project. Building an accurate FE model of the female body is the foundation of this
project. The FE model of the body is constructed with a rigid torso, breasts with hyperelasticity, and a
layer of subcutaneous tissues. The geometric model of the body is obtained through 3D laser scanning
79
for authenticity. The Mooney-Rivlin material coefficients of the breasts are determined by examining
the differences between the FE-modelled results and experimental data, and the damping ratio is
directly calculated from the free-vibration test data. The method proposed in this chapter provides an
effective way to determine the properties of the breasts to predict breast deformation and then, used to
analyse the contact mechanism of the breasts and sports bra.
80
Chapter 4 Development of finite element contact model
4.1 Introduction
In order to limit breast motion during physical activities, commercial sports bras tend to compress the
breasts against the ribcage, but the excessive pressure exerted onto the skin can cause discomfort and
is physiologically distressing. Designing garments that address overall wellbeing is fast becoming an
emerging research topic. However, scientific analyses of the mechanical interactions between sports
bras and the human body have been seriously lacking because it is challenging to do so, due to the
difficulties of modelling and optimising the nonlinear interactions between the breasts which are
viscoelastic and fabrics that are elastic.
As for increasing the wear comfort of sports bras, a direct method to optimise wear comfort is to reduce
the amount of contact pressure between the sports bra and body of the wearer but under the premises
of ensuring function. Previous studies have typically focussed on the interaction between a sports bra
and the body based on experiments with human subjects (Coltman, McGhee & Steele, 2015). However,
fabricating sports bra samples and fitting them onto female subjects are time consuming processes and
prone to human error. In the last few decades, FE modelling has been used to solve this problem. For
example, the FE model in a study by Li et al. (2003) was built by using B-splines for a numerical
analysis, as opposed to images of real human bodies. Nevertheless, their study lacked authenticity as
parts of the breast are modelled as elastic components and the other parts of the body as rigid
components. They also did not consider gravity, which means that there is no initial pressure of the
bra in their model. In addition, Bel-Brunon et al. (2014) developed an FE contact model and calculated
the normalised von Mises stresses for a bra worn without gravity. Their results showed that the vertical
displacement of the nipples with a bra worn is greater than that without a bra worn, which is peculiar.
81
The deficiencies of previous studies therefore show the difficulties of using FE analyses to solve
problems on the contact between breasts and bras. The primary challenge is the penetration problem
in modelling the contact mechanics.
To overcome these shortcomings, this chapter presents the development of a FE contact model of the
female body and sports bra to simulate the static and dynamic contact conditions and dynamic
displacement, and calculate the natural frequencies of the breasts when a sports bra is worn as the
support system. The FE models were based on the actual female body and sports bra, and nonlinear
material model was used to describe the movement of the breasts. In particular, a new method was
used to avoid penetration problem. This chapter provides a better understanding of the contact between
a compression sports bra and the body, which can improve the development process of sports bras
without the need to involve humans through trial and error. Moreover, the new method provides a
more efficient, accurate and robust strategy to solve not only the complex problems of body-bra
interactions but also other design applications where the properties of the materials are highly nonlinear
and viscoelastic in nature.
4.2 Experimental work
To simulate the contact conditions between the sports bra and the wearer, three experiments were
carried out. The first involved a tensile strength test to obtain the material parameters of the sports bra
components by using an Instron 4411 tester. The second experiment was a contact pressure test to
validate the boundary conditions of the FE contact model by using a pressure sensor system. The third
was a motion capture experiment of the subjects while they wore the sports bra to validate the results
of the FE contact model, and the RMSE between the experimental and FE dynamic displacements was
calculated to measure the accuracy of the FE contact model. The procedures of the motion capture
82
process have been described in detail in the last chapter and the only addition here is that the subject
is wearing a sports bra. Therefore, the procedures for the motion capture experiment will not be
discussed in this chapter.
4.2.1 Tensile strength test of sports bra components
The five sports bras selected for this study are shown in Figure 4.1, which have different impact ratings
(high, medium and low impacts). The mechanical properties of the sports bra components were tested
by using an Instron 4411 tester at a constant-rate-of-extension; see Figure 4.2. The sports bras were
examined as 4 components, namely, the shoulder straps, back panel, bra cup, and elastic bra band; see
Figure 4.3. Tthe different components were individually sewn into loop samples. The samples were
then loaded at a specified rate to a pre-set tension and unloaded at a specified rate until the loop did
not have any tension. This cycle was repeated 5 times to find the mean value.
Figure 4.1 Sports bra samples
83
Figure 4.2 Stretching of sports bra
Figure 4.3 Instron 4411
4.2.2 Contact pressure test
The actual contact pressure between the body and sports bra was tested by using a pressure sensor
system (Novel Pliance-X). Two different types of sensors were used to measure 5 points on the body:
84
the left and right shoulders (Points 1 and 2), left and right underarms (Points 3 and 4), and bottom of
the left bra cup (Point 5). Points 1 to 4 were tested by using a single sensor and Point 5 with a 2*2
matrix sensor because of the large area that needed to be measured. The tested positions and the
equipment are shown in Figure 4.4 and Table 4.1. The contact pressure changed when the subject
inhaled and exhaled, and each time that she made a small movement. In this study, the tested pressure
is calculated by averaging the pressure over the stationary phase.
Figure 4.4 Pressure test positions
Table 4.1 Measured points
Point Location
Point 1 Left shoulder
Point 2 Right shoulder
Point 3 Left underarm
Point 4 Right underarm
Point 5 Bottom of left bra cup
85
4.3 Finite element contact model
4.3.1 Sports bra model
The strain-stress behaviour of the fabric of the sports bra components in the direction of the stretch
primarily influences the amount of contact pressure. The direction of the stretch of the strap, bra cup
and back panel is in the wale direction, while that of the elastic bra band is in the course direction.
Hence, the FE model of the sports bra fabric was assumed to be isotropic, and the Young’s moduli
from the experimental data are shown in Table 4.2 (Yu et al., 2016). The geometric model of the sports
bra was extracted from the gravity-free model of the body. The initial shape of the FE model of the
body was affected by the gravity load. A gravity-free model was then constructed by using a force with
the same magnitude as gravity but in the opposite direction. The entire model of the bra was meshed
by using 6 mm quadrilateral elements. The type of element used is Element 138 in MSC. Marc software,
which is shown in Figure 4.5. As shown in the figure, the first two element base vectors (V1 and V2)
lie along the plane of the three corner nodes and the third vector (V3) is determined by using the right-
hand system from the first two vectors. This model is shown in Figure 4.6.
Figure 4.5 Element 138
86
Figure 4.6 FE model of compression sports bra
Table 4.2 Material properties of sports bra components

Component Young’s modulus /MPa
Strap 2.3159
Bra cup fabric 0.3344

Bra 1
Elastic bra band 1.5732
Back panel 4.2027
Strap 1.8372

Bra 2
Back panel 0.5973
Strap 7.2167

Bra 3
Back panel 0.6363
Bra4 Strap 5.8943
87
Back panel 0.7771
Strap 1.3324

Bra 5
Back panel 0.5502
4.3.2 Loading and boundary conditions
The FE model of the body was initially based on the subject in the standing position and the shape of
the breasts would be affected by the gravity load. A force with the same magnitude as gravity but in
the opposite direction was added so that the breasts were based on a static equilibrium position, which
means that no external force was applied to them.
The greatest challenge that needed to be addressed was penetration in the contact model. To solve this
problem, the model of the body was first reduced and then expanded, which involves the following
steps:
a) the size of the original model of the body and bra was geometrically reduced. The scale factor for
shrinking was 0.97 in each direction (x-, y- and z-directions). Sports bras expand during wear. Hence,
the model of the bra in the initial state had to be smaller than the original model;
b) the model of the body was reduced, during which the gap between the models of the body and the
sports bra was noted;
88
c) the model of the body was geometrically expanded, which was done in Marc (MSC Marc 2014.2.0,
US) by adding a thermal expansion property, which means that the volume increases with temperature.
The expansion ended when the size of the model of the body reached the original size and calculated
by using Equations 4.1 to 4.4; and
d) gravity load was applied to the breasts.
The key step of this method which is to first shrink and then expand the model is to determine the scale
factor for shrinking, which is 0.97 in this study. Different amounts of contraction (using scale factors
of 0.92, 0.93, 0.94, 0.95, 0.96, 0.97 and 0.98) were attempted in each direction (x-, y- and z-directions)
for both the body and the sports bra. In order to ensure that each part of the FE model uniformly shrunk,
the volume of the entire FE model was centred and scaled based on the centroid of the model. Hence,
the scale factor for shrinking was the same in all three directions. The end of the expansion was when
the size of the model of the body reached the original size (i.e. from 0.983 to 1). Therefore, the amount
of expansion depends on the scale factor for shrinking. However, the amount of expansion is limited
because of the penetration problem and excessive deformation of the FE meshes. After a few
simulations, the smallest scale factor for shrinking is 0.92. Then, seven trials of simulations with
different scale factors for shrinking (i.e., 0.92, 0.93, 0.94, 0.95, 0.96, 0.97 and 0.98) were conducted.
Finally, the simulated static contact pressure which used a scale factor of 0.97 was found to mostly
agree with the experimental value. Therefore, the scale factor used for shrinking is 0.97. The static
contact pressure between the sports bra and the body was analysed by using the steps described above.
The entire process is shown in Figure 4.7.
The reason why this method could solve the penetration problem was that the overlap between the bra
and the breast initially was too large. Therefore, FEM could not give a physically reasonable solution
if there was no gap at the first stage.
89
Figure 4.7 Boundary conditions of contact model
Assuming that the initial temperature in the FE models is T0, the initial volume of the model of the
body is V0, and the volumetric coefficient of the thermal expansion is αv , which results in
1 ∆𝑉𝑉
α𝑣𝑣 = (4.1)
𝑉𝑉0 ∆𝑇𝑇
90
where ∆𝑉𝑉 is the variation in the size of the model of the body and ∆𝑇𝑇 is the temperature variation.
T=k∗t (4.2)
where T is the temperature, k is a user-defined proportionality coefficient and t is time. Hence,
Equation (5) can be rewritten as follows:
𝑉𝑉𝑠𝑠 −𝑉𝑉0 𝑉𝑉𝑠𝑠 −𝑉𝑉0

α𝑣𝑣 = = (4.3)
𝑉𝑉0 (𝑇𝑇𝑠𝑠 −𝑇𝑇0 ) 𝑉𝑉0 (𝑘𝑘𝑡𝑡𝑠𝑠 −𝑇𝑇0 )
1 𝑉𝑉𝑆𝑆 −𝑉𝑉0
t1 = ( + 𝑇𝑇0 ) (4.4)
𝑘𝑘 α𝑣𝑣 𝑉𝑉0
where ts is the end of the expansion of the model of the body, Ts is the temperature at time ts and Vs is
the size of the model of the body at time ts.
The dynamic FE contact model was used to simulate the breast displacement and contact pressure
during running. As the breasts were attached to the torso part of the body with the glue contact option,
the boundary displacement extracted from the motion capture experiment was applied to the torso part.
The applied boundary displacement was simplified to two degrees of freedom (y- and z-directions).
91
4.4 Contact pressure
4.4.1 Static contact pressure
The distribution of the contact pressure in the different sports bras is very similar. The highest contact
pressure was found on the shoulders caused by the shoulder strap, and under the arms caused by the
elastic bra band. Studies in the literature state that the bra band supports most of the weight of the
breasts, so the bra band is mostly fabricated with strong elastic material, and the shoulder straps are
supposed to hold the bra in place. Therefore, they can induce relatively high contact pressure, which
is especially true for larger-breasted women, such as S01 among the subjects in this study. The highest
contact pressure between the bra cup and the breasts can be observed at the bottom of the breasts,
which is caused by the gravity load. The values of the simulated contact pressure at the bottom of the
bra cup approximate those of the experiment, which indicates that taking gravity load into
consideration is a good idea. With accurate material coefficients, the FE model results agree with the
experimental results. However, most commercial sports bras have two layers. The sports bra model
was simplified to one layer, which means that the results could be somewhat inaccurate.
Figures 4.8 to 4.12 show the distribution of the contact pressure between the sports bras and S01 with
the static FE contact models. The numerically and experimentally obtained contact pressure values are
summarised in Table 4.3. It can be observed that the differences in contact pressure between FEM and
experiment were small except Bra 3. Bra 3 has three layers of the bra cup, so the contact pressures
were larger in the experimental results. This shows a need for future work that more FE models of
sports bras with different structure should be built based on the experience in this study.
92
Figure 4.8 Distribution of contact pressure between Bra 1 and S01
93
Table 4.3 Comparison of numerically and experimentally obtained contact pressure

Contact pressure based on Contact pressure based on
Sports bra Position
FEM /kPa experiments /kPa
Shoulder 1.4 1.4
Bra 1 Underarm 2.0 2.0
Bottom of bra cup 0.8 0.7
94
Shoulder 1.4 1.4
Shoulder 1.7 1.6
Shoulder 2.1 2.1
Shoulder 1.5 1.4
4.4.2 Dynamic contact pressure
Not only can FE modelling accurately simulate the static contact pressure but can also be performed
to determine the contact pressure during running. Since dynamic contact pressure is difficult to
determine with the use of wired sensors, the FE method can be applied to solve this problem. The
dynamic results showed that the amount of the contact pressure caused by the bra strap and the elastic
bra band increases during movement. However, the contact pressure at the bottom of the bra cup is
slightly reduced, probably because the compression sports bra used in this study is a simple vest-style
bra. The breasts were mainly supported by the straps and elastic bra band during running. Additionally,
the changes increased with a faster speed of running. The greatest change in the contact pressure caused
by the strap and elastic bra band is 0.8259 and 0.3788 kPa, respectively. As the strap may cause the
most discomfort, studying its dynamic contact pressure is important when designing a sports bra.
95
4.5 Dynamic breast displacement with sports bra
In this study, FE modelling is carried out to predict the dynamic displacement of the breasts during
running at two constant speeds (5 and 10 km/h). To validate the FE contact model, the breast
displacement obtained from the simulations was compared with that based on the motion capture
experiment with the five sports bras in the y- and z-directions. The displacement in the x-direction was
ignored because the displacement of the torso and breasts in the x-direction very much relies on the
arms which move back and forth during running. A comparison between the simulation and
experimentally obtained displacement was made for each sports bra condition; see Figures 4.13 to 4.17.
The figures illustrate that the two methods show similar patterns of breast motion. The corresponding
RMSEs were less than 1%, which is acceptable (Chen et al., 2013) (Table 4.4). However, the absolute
error at the peak were relatively large. This may be because the FE model of the breasts were simplified
as homogeneous, instead of having different tissues. More detailed FE model of breast should be
developed in the future.
96
Figure 4.13 Simulation and experimentally obtained breast displacement during two speeds of
running: Bra 1
running: Bra 2
97
running: Bra 3
running: Bra 4
98
running: Bra 5
Table 4.4 RMSE percentage of simulated breast displacement

Bra Direction Condition - Running RMSE% Max absolute error (mm)
5 km/h 0.3310 15
Y
10 km/h 0.2633 6
Bra 1
5 km/h 0.5056 13
Z
10 km/h 0.5214 30
5 km/h 0.8781 12
Y
10 km/h 0.3091 10
Bra 2
5 km/h 0.1912 13
Z
10 km/h 0.4725 18
5 km/h 0.2907 25
Y
10 km/h 0.9177 17
Bra 3
5 km/h 0.2713 8
Z
10 km/h 0.6328 19
5 km/h 0.3068 15
Y
10 km/h 0.3323 16
Bra 4
5 km/h 0.8237 18
Z
10 km/h 0.2672 29
5 km/h 0.5942 16
Y
10 km/h 0.7416 18
Bra 5
5 km/h 0.7941 20
Z
10 km/h 0.5064 12
99
4.6 Modal analysis
To analyse the dynamic response of the breasts with and without a sports bra worn, a modal analysis
was carried out and the natural frequencies were calculated in MSC. Marc software. In reality,
resonance occurs when the frequency of motion matches the natural frequencies of the human body.
The female breasts are especially vulnerable and large deformations of the breasts caused by resonance
can lead to substantial discomfort, pain or even injury. However, resonance is undesirable so if the
natural frequencies of the breasts are known, this would help to prevent resonance (Balachandran,
2009).
The first 5 natural frequencies of the breasts of S01 with and without the most basic sports bra (Bra 5)
worn were calculated by considering the viscous damping effect. The frequencies are reported in Table
4.4. The findings from the modal analysis demonstrate the importance of wearing a sports bra during
physical activities. The step frequencies with different walking or running speeds were calculated by
using f=1/T, which ranged between 1.0 and 3.5 Hz; see Table 4.5. The natural frequencies in the braless
condition are all within this range too; see Table 4.6. Usually, the first mode shape has the largest
amplitude, which means that the breasts experience the largest deformation. So, it is possible that the
first mode shape points to a less comfortable condition as opposed to the other conditions. Figure 4.18
shows the first mode shape when a sports bra is not worn.
100
Table 4.5 Frequencies of breast motion during running
Frequency (s-1)
Condition
5 km/h (y) 10 km/h (y) 5 km/h (z) 10 km/h (z)
Breast FEM 2.84 3.21 1.07 1.49
Breast actual 2.85 3.17 1.12 1.63
Table 4.6 Natural frequencies with and without a sports bra
Condition
Vibration No bra Sports bra worn
Left breast Right breast
Frequency order Magnitude of frequency (Hz)
1st 2.074 2.075 5.585
2nd 2.198 2.203 7.242
3rd 2.533 2.569 7.575
4th 3.205 3.205 8.071
5th 3.298 3.307 8.801
101
(a) First mode shape of left breast (b) First mode shape of right breast
Figure 4.18 First mode shape
The natural frequencies of the breasts when a sports bra is worn as a support system were found to be
significantly higher than those in the braless condition. Since the step frequency during running easily
reaches the natural frequency of the breasts, resonance can occur. Moreover, the breasts will oscillate
with a much greater amplitude, which can cause high levels of discomfort. The human body and
sportswear combined are supposed to have a higher natural frequency than the step frequency during
exercise. A modal analysis can improve the understanding of breast motion and provide the basis for
predicting breast displacement in dynamic conditions. Therefore, a modal analysis can be used to
validate the function of a sports bra or any other sportswear. This analysis proves that the FE model is
an important and necessary tool for future studies that seek to address problems that involve soft tissues
and garments.
102
4.7 Conclusions
This chapter explicates in detail, a method of developing and using an FE contact model, the results
from the contact model, and the corresponding experimental work. To solve the penetration problem
with the use of a contact model, a method is created which first shrinks and then expands the model,
and used to simulate the contact conditions between the female body and sports bra. In the static
analysis, the simulated contact pressure values are in agreement with the results of the contact pressure
tests. The distribution of the contact pressure is clear and can be easily analysed with computer
modelling. In the dynamic analysis, breast motion is investigated and analysed when a sports bra is
donned with contact pressure. Based on the calculated errors between the results of the FE model and
experiments, the model is found to predict the contact conditions with good accuracy. The reliability
of this study is thus validated.
103
Chapter 5 Determining pressure comfort range
5.1 Introduction
The perceived comfort when wearing a sports bra is associated with the pressure exerted onto the skin
by the sports bra (Okabe & Kurokawa, 2004). A well-designed sports bra should provide an
appropriate amount of contact pressure which limits the movement of the breasts during physical
activities, and also gives a sense of security to the wearer. However, excessive pressure not only causes
discomfort but also results in various health problems, as documented in previous research (Lee, Hyun
& Tokura, 2001; Miyatsuji et al., 2002). Nevertheless, an evaluation of the range of contact pressure
that is tolerable while wearing a sports bra during running is still necessary as the results in previous
studies are mixed. Hence, the purpose of this chapter is to determine a range of contact pressure that
would be tolerable while wearing a sports bra during running.
In order to determine this range, two steps are carried out; one is subjective while the other is objective.
First, a questionnaire survey is carried out to determine the perceived comfort sensation. The design
of the questionnaire is based on two conditions: static and dynamic. The perceived subjective comfort
in the dynamic condition is important because sports bras are worn during dynamic activities, like
physical exercise. The other step is conducting pressure tests to obtain objective experimental data.
The aim is to provide a sports bra that is well fitting, which is based on the bust and under-bust
circumferences of a woman.
104
5.2 Experimental work
5.2.1 Subjects
To investigate the pressure comfort range of a sports bra that is perceived to be comfortable, twelve
Chinese female subjects were recruited on a volunteer basis to try on at least one sports bra and give
feedback on the perceived comfort. The corresponding contact pressures were also tested. The
selection criteria were that the subjects have to be healthy, have not underwent any previous breast
surgeries, and are not pregnant or lactating. The basic information of the subjects and their body
measurements are listed in Table 5.1. The youngest subject is 19 years old and the oldest subject is 50
years old in this experiment. The subjects were grouped based on their bra cup size: 32C (three
subjects), 34B (three subjects), 34C (four subjects), 34D (one subject), and 36C (one subject). The
subjects were informed about the purpose of the study, and then had to sign an informed consent form
before taking part in the experiment. The experimental procedure was approved by the Human Ethics
Committee of the Hong Kong Polytechnic University (HSEARS20151207004).
105
Table 5.1 Subject information
Breast Full Under Cup size (commercial

Age Height(cm) Weight(kg) BMI Horizontal
cup line Root Bust Bust bra)
(cm) (cm) (cm) (cm)
Subject L R L R
S01 21 149.0 42.3 19.1 17.8 17.9 24.0 24.1 76.3 66.2 32C
S02 21 161.0 51.7 19.9 21.0 20.9 21.1 20.5 85.0 68.4 32C
S03 20 155.0 52.7 21.6 18.5 19.5 24.6 24.4 85.8 75.6 34B
S04 23 163.0 55.8 21.0 20.2 19.1 23.8 24.1 84.5 71.8 34B
S05 21 167.0 60.9 21.8 23.1 21.5 20.3 20.3 86.2 70.9 34B
S06 24 155.0 48.6 20.2 19.2 18.9 22.3 22.5 83.5 69.8 34C
S07 42 166.0 61.2 22.2 22.5 21.4 20.6 21.7 91.5 77.2 36C
S08 19 158.0 56.6 22.7 21.4 22.1 22.8 23.5 90.2 74.5 34C
S09 50 163.0 55.5 20.9 23.5 23.8 26.0 25.7 95.0 74.0 34D
S10 24 148.0 56.2 21.7 22.6 21.9 23.1 21.5 91.5 77.5 34C
S11 22 156.0 52.5 21.6 20.3 20.0 22.0 21.5 83.0 72.0 32C
S12 23 167.0 58.6 21.0 22.0 21.5 22.7 22.0 92.0 75.0 34C
Average 25.83 159.0 54.38 21.14 21.01 20.71 22.78 22.65 87.04 72.74
Standard
deviation 9.70 6.30 5.09 0.98 1.77 1.59 1.60 1.62 4.96 3.36
106
5.2.2 Experiment
To prepare for the experiment, a questionnaire was first carried out to determine the
perceived comfort sensation while wearing a sports bra (Figure 5.1) and the sports bra
samples. The samples are vest-style sports bras, which is the same type of sports bra
discussed in Chapter 4.
The questionnaire has two parts: the first part asked for the basic information of the
subjects and their body measurements, while the second part assessed the comfort of
the wearer in both static and dynamic conditions. A nine-point Likert scale with 1 as
extremely uncomfortable and 9 as extremely comfortable was used to measure the
perceived comfort sensation.
The experiments were conducted in the laboratory at The Hong Kong Polytechnic
University. The temperature was controlled between a range 23℃ to 27℃, and the
relative humidity was 65%±3%. The wind speed was less than 1 m/s. This environment
ensured that the female subjects felt comfortable in their natural state, and that the
environment would not affect their evaluation of the perceived comfort.
The experiment procedures and approximate time required to carry out the experiment
are shown in Figure 5.2. After the subject had a good understanding of the aims and
procedures of the experiments, measurements were first taken of her horizontal cup line,
breast roots, full bust and under bust. The measurements of each subject are listed in
Table 5.1. Based on the measurements, the correct-size sports bra can then be provided
to the subject. Then, the female subjects donned their provided sports bra and they were
examined to ensure that the bra is well-fitting. Four different types of sports bras were
offered to the subjects, and they were randomly assigned their bra and number of bras.
The order of the wear trial for each subject was also randomized. Next, the contact
107
pressure between the body of the subject and their sports bra was determined by using
a pressure sensor system (Novel Pliance-X). The contact pressure from the shoulder
strap, elastic bra band and bra cup were tested, which are shown in Figure 5.3.
Pressure was exerted to the same three areas of each subject for a few minutes during
the contact pressure test and the subjects were not to make any large movements. Then,
the subjects were also required to rate the pressure comfort in the static condition. To
obtain the scores of the pressure comfort, each subject was required to run on a treadmill
at two constant speeds (5 km/h and 10 km/h) for a few minutes, and then rate the
pressure comfort for the dynamic component of the testing. The entire experiment
required about 50 minutes. The experimental data were used to study the effect of
contact pressure at the three positions toward pressure comfort by using statistical
analyses, which are correlation coefficient and analysis of variance (ANOVA).
108
Figure 5.1 Questionnaire on perceived comfort sensation
109
Figure 5.2 Experimental procedure
110
Figure 5.3 Testing contact pressure of different bra components on body
5.3 Results and discussion
Table 5.2 shows the mean scores of the perceived pressure from the subjects during standing
or sitting (static), and running (dynamic), and the contact pressure from the shoulder strap, bra
cup and elastic bra band based on the tests. The differences between the scores in the static and
dynamic conditions are evident. With the exception of the contact pressure caused by the elastic
bra band, the mean scores of perceived pressures increase with running rather than standing.
Moreover, the score for the perceived pressure from the elastic bra band is the same for both
the static and dynamic conditions, and less for the static condition. This means that the subjects
who felt less comfortable during standing or sitting felt that it is more comfortable to wear the
same sports bra for exercise. According to the study in Chapter 4, the amount of the contact
pressure caused by the bra strap and the elastic bra band increases during movement, and that
at the bottom of the bra cup is reduced. Therefore, the change in perceived pressure exerted by
the bra cup is mainly caused by a decrease in the pressure value during running. As for the
111
perceived pressure of the shoulder strap, a greater sense of safety and protection may be needed
during running, which can be provided by a higher-pressure value. Hence, the score of the
perceived pressure of the entire sports bra is increased.
Table 5.2 Mean scores and contact pressures
Mean scores of perceived
Bra component pressures Contact pressure (kPa)
Static Dynamic
Overall 6.5 ± 1.4 7.0 ± 1.7 NA
Shoulder strap 6.5 ± 2.1 8.0 ± 0.7 2.1 ± 0.9
Cup 6.0 ± 1.8 8.5 ± 0.7 0.7 ± 0.7
Elastic bra band 7.0 ± 1.5 7.0 ± 1.1 2.2 ± 1.3
Table 5.3 shows the correlation coefficients among the overall score of perceived pressure,
scores for the perceived pressure from the different parts of the bra, and the contact pressure
values in the static condition (significance at 0.05). The correlation coefficients were calculated
by using SPSS Statistics 25 (IBM, US).
Table 5.3 Correlation coefficients among overall score, scores for perceived pressure of
different parts of bra, and contact pressure values
Correlation coefficient
Pressure from strap 0.236
Pressure from bra band 0.561
112
Pressure from bra cup -0.114
Scores of perceived pressures from shoulder 0.805
straps
Scores of perceived pressures from elastic bra 0.748
band
Scores of perceived pressures from bra cup 0.676
It can be easily observed in Table 5.3 that the correlation between the overall score and the
contact pressure from the elastic bra band is high compared to the contact pressure from the
other bra components. Also, the perceived comfort of the shoulder strap is affected the most in
the overall score. This result corresponds with the results of a study by Bowles, Steele and
Munro (2012) who found that the slipping of the bra straps down the shoulders or the digging
of bra straps into the shoulders is the most uncomfortable feature of sports bras. The results of
the ANOVA are shown in Tables 5.4 and 5.5 for the static and dynamic conditions respectively.
In the static condition, the contact pressure caused by the shoulder strap has a significant (p ＜
0.05 ) impact on the scores of perceived pressure from the bra cup. The contact pressure caused
by the bra cup has a significant influence on the scores of perceived pressures from the elastic
bra band and overall scores of perceived pressures. In the dynamic condition, the contact
pressure caused by the shoulder strap significantly affects the scores of perceived pressures
from the elastic bra band. Moreover, the contact pressure caused by the elastic bra band
significantly affects the scores of perceived pressures from the bra cup. The contact pressure
from the bra cup could significantly affect the scores of perceived pressures from the shoulder
strap (Wong, Li, & Zhang, 2004). Hence, the contact pressure from all of the bra components
can have impacts on the perceived pressure and should be considered in the design of sports
bras.
113
Table 5.4 ANOVA table of influence of contact pressure on scores of perceived pressures in
static condition
Factor Attributive Sum of df Mean F Sig.
variable squares square
Pressure Overall Between 35.321 13 2.717 3.215 0.064
caused by scores of groups
strap perceived Within 5.917 7 0.845
pressures groups
Total 41.238 20
Scores of Between 71.351 13 2.717 2.597 0.105
perceived groups
pressures Within 14.792 7 0.845
from groups
shoulder Total 86.143 20
straps
perceived groups
from bra groups
cup Total 62.071 20
perceived groups
from groups
114
elastic bra Total 46.071 20
band
bra band perceived Within 10.625 5 2.125
pressures groups
Total 41.238 20
perceived groups
from groups
straps
perceived groups
from bra groups
cup Total 62.071 20
perceived groups
from groups
band
115
bra cup perceived Within 5.982 8 0.748
pressures groups
Total 41.238 20
perceived groups
from groups
straps
perceived groups
from bra groups
cup Total 62.071 20
perceived groups
pressure Within 5.339 8 0.667
from groups
band
116
Table 5.5 ANOVA table of influence of contact pressure on scores of perceived pressures in
dynamic condition
Factor Attributive Sum of df Mean F Sig.
variable squares square
bra strap perceived Within 9.167 7 1.833
pressures groups
Total 27.500 20
perceived groups
from groups
straps
perceived groups
from bra groups
cup Total 22.725 20
perceived groups
from groups
Total 25.500 20
117
elastic bra
band
bra band perceived Within 14.500 5 4.833
pressures groups
Total 27.500 20
perceived groups
from groups
straps
perceived groups
from bra groups
cup Total 22.725 20
perceived groups
from groups
band
118
bra cup perceived Within 9.500 8 4.750
pressures groups
Total 27.500 20
perceived groups
from groups
straps
perceived groups
pressure Within 0.667 8 0.333
from cup groups
Total 22.725 20
perceived groups
from groups
band
As it is difficult to accurately test the dynamic contact pressure and women usually feel more
comfortable with a sports bra, the analysis on the pressure comfort range focused on the static
119
condition. In this study, most of the subjects did not try only one sports bra. Hence, there were
21 answered questionnaire and corresponding tested pressures. These were divided into three
groups: Group 1 (score higher than 8), Group 2 (score higher than 6 but less than 8) and Group
3 (score less than 6). The no. of conditions is shown in Table 5.6. To obtain the pressure comfort
value, the mean contact pressure values in the three groups were calculated, which are shown
in Table 5.7. There is clearly an increase in the contact pressure values with reduced scores of
perceived pressures. The significance is 0.05, so that a 95% confidence interval of the mean
contact pressure value on the shoulders is not less than 1.26 kPa and not more than 2.90 kPa.
A 95% confidence interval of the mean contact pressure value of the underarm is not less than
1.83 kPa and not more than 3.49 kPa. A 95% confidence interval of the mean contact pressure
value at the bottom of the bra cup is not less than 0.01 kPa and not more than 1.15 kPa
(Tsujisaka et al., 2004). Hence, it is important to use appropriate fabrics and sports bra designs
to adjust the contact pressure values towards a range that would be tolerable. The contact
pressure values in Group 3 are an uncomfortable range. The contact pressure on the shoulders
and underarm and at the bottom of the bra cup is in the range of 3.30 ± 1.55 kPa, 4.00 ± 1.41
kPa and 1.43 ± 1.02 kPa respectively and can cause discomfort.
120
Table 5.6 No. of conditions in three groups
Contact pressure Group No. of conditions
Group 1 8
Pressure from bra strap Group 2 6
Group 3 7
Group 1 5
Pressure from bra band Group 2 11
Group 3 5
Group 1 12
Pressure from bra cup Group 2 6
Group 3 3
Table 5.7 Mean pressure values
Group 1 (kPa) Group 2 (kPa) Group 3 (kPa)
Pressure from bra strap 2.08 ± 0.82 2.83 ± 1.13 3.30 ± 1.55
Pressure from bra band 2.66 ± 0.83 3.27 ± 1.46 4.00 ± 1.41
Pressure from bra cup 0.58 ± 0.57 0.69 ± 0.79 1.43 ± 1.02
5.4 Conclusions
The relationship between the tested contact pressures (objective experimental data) and the
scores of perceived pressures (subjective perception) has been investigated and the tolerable
range of contact pressure is determined in this chapter. The scores of perceived comfort
increase from static (standing or sitting) to dynamic (running) conditions. As for the factors
that affect the level of comfort, the contact pressure in three positions, which are caused by the
121
shoulder strap, elastic bra band and bra cup, could have significant influence on the comfort
level for the different parts of the bra and in different conditions. Therefore, the contact pressure
needs to be considered in the design of sports bras and the findings can be used to optimize
designs. A contact pressure on the shoulders, underarm and at the bottom of the bra cup of 2.08
± 0.82 kPa, 2.66 ± 0.83 kPa and 0.58 ± 0.57 kPa respectively is most tolerable. The optimum
contact pressure allows optimum fabrics for sports bras to be determined.
122
Chapter 6 Prediction of contact pressure and dynamic displacement during running
6.1 Introduction
FE modelling has been successfully used with great accuracy to describe the biomechanical
behaviours of breasts and bra-breast interactions (see for example, Chen et al., 2013; Sun et al.,
2019a; Sun et al., 2019b). Although FE modelling provides detailed information and does not
require the involvement of human subjects, it is still computationally expensive and time
consuming in most cases. On the other hand, machine learning (ML) is currently used in a wide
variety of applications (Bottou et al., 2018; Atta et al., 2019; Huang et al., 2019). ML
algorithms can build a mathematical model with given data to make predictions without
background knowledge of the problem. That is, models trained to use ML algorithms based on
sample data can predict the desired quantities in real time with reasonably little bias and
variance. To further enhance efficiency, the ML approach is used in this study to investigate
the contact pressure from bra-breast interaction instead of the FE method or an FE model
because the modelling process requires hours or even days to calculate the information.
Nevertheless, FE modelling results of different female subjects are also presented in this
chapter to validate and compare the ML results. The generalisability of the results is therefore
ensured.
6.2 Finite element-based machine learning approach to predict contact pressure
Chapters 3 and 4 described in great detail the development of an FE model of the female body
and FE contact model between the body and sports bra. The contact conditions between a sports
bra and the body of a subject (S01) were simulated by using FE software, Marc 2014.2.0 (MSC
Software Corporation, US). The FE models provide good simulated results for the contact
123
pressure between the sports bra and the female body. In this chapter, the data obtained by the
FE models are used to train an ANN model. The results are obtained under two conditions:
1) The FE model of the body is based on the same subject (a 50-year-old woman with a bra
size of 75D), and
2) The FE models of the sports bra are based on the same design, which is a vest-style sports
bra.
6.2.1 Dataset generation
ANN models require three sets of data: a training set, a validation set and a testing set. These
were prepared by applying the data from the FE models. The function of the training set is to
provide a set of data to fit every training example of the ANN. The validation set is used to
evaluate the ANN model when training is being carried out and halt the training if there are too
many errors. The testing set is used to evaluate the final ANN model. The testing:
validation:training ratio in this study is 70:15:15.
Four parameters were used which represent the four parts of a sports bra (shoulder strap, back
panel, bra cup, and elastic bra band). Five levels of each parameter were used to investigate the
effect of the different material properties of the sports bra fabrics, which provided 54 = 625 full
factorial experimental runs. Collecting the results of more than six hundred runs is very time
consuming because each FE modelling process takes hours, or even days to carry out.
Therefore, the Taguchi method is used to reduce the number of experiments because it is a
powerful statistical method that reduces the number of full factorial runs with the use of
orthogonal arrays (OAs). In this study, a standard OA with four parameters and five levels is
used, and 25 FE models are simulated to provide the training set for the ANN model. These
runs were carried out with different sets of parameters and for different levels. As the training
124
set consisted of 70% of the sample data, ten more sets of data were needed, and of these ten,
five sets were calculated by using the tested elastic modulus of the five sports bra samples
whereas the other five sets were calculated with random material properties. Table 6.1 shows
the contact pressure at three locations on the sports bras which are obtained from the 35 FE
runs.
Table 6.1 Finite element parameters and results

Exp. no. FE parameters (MPa) FE results (KPa)
Contact
Contact
Young’s Young’s Young’s Young’s Contact pressure
pressure
modulus modulus modulus modulus pressure at
under
of bra of bra of back of bra on bottom
the
cup strap panel band shoulders of bra
arms
cup
1 0.2 2 1 1 1.230 0.2459 0.8607
2 0.2 4 2 2 1.349 0.2362 1.349
3 0.2 6 3 3 1.401 0.2802 1.401
4 0.2 8 4 4 1.612 0.3281 1.881
5 0.2 10 5 5 2.261 0.4230 1.938
6 0.4 2 2 3 2.034 0.3515 1.743
7 0.4 4 3 4 2.342 0.5313 2.342
8 0.4 6 4 5 2.623 0.5747 2.248
9 0.4 8 5 1 2.383 0.5983 1.589
10 0.4 10 1 2 2.705 0.4057 1.803
11 0.5 2 3 5 1.757 0.4562 2.109
125
12 0.5 4 4 1 2.126 0.5252 1.701
13 0.5 6 5 2 2.479 0.5361 1.983
14 0.5 8 1 3 2.979 0.4964 1.986
15 0.5 10 2 4 3.120 0.6201 2.080
16 0.6 2 4 2 1.936 0.5227 1.936
17 0.6 4 5 3 2.743 0.6428 2.400
18 0.6 6 1 4 3.018 0.5311 2.156
19 0.6 8 2 5 3.286 0.8286 2.300
20 0.6 10 3 1 2.775 0.6775 1.665
21 0.8 2 5 4 2.687 0.5375 2.687
22 0.8 4 1 5 2.730 0.6768 2.184
23 0.8 6 2 1 3.024 1.008 1.344
24 0.8 8 3 2 3.626 1.209 2.015
25 0.8 10 4 3 4.164 1.249 2.915
26 0.3344 2.3159 4.2027 1.5732 1.403 0.802 2.004
27 0.9626 1.8372 0.5973 3.0017 1.426 0.634 1.585
28 0.3101 7.2167 0.6363 2.5339 1.741 0.390 1.741
29 0.3597 5.8943 0.7771 1.7695 2.151 0.717 1.792
30 0.9474 1.3324 0.5502 0.9127 1.431 0.477 1.113
31 0.3100 1.8270 3.5000 1.5700 1.944 0.5326 1.166
32 0.3100 1.8270 0.4400 2.7500 2.322 0.5220 1.935
33 0.4000 7.2000 4.2000 3.0000 2.900 0.6451 2.250
34 0.3100 6.5000 0.4370 1.5700 2.100 0.4791 2.000
35 0.3100 5.5000 0.4370 1.5700 2.024 0.6287 1.214
126
6.2.2 ANN structure
After analysing the characteristics and applicable ranges of different algorithms, a multilayer
neural network that uses a backpropagation algorithm was chosen to train the ANN. Three
widely used and efficient algorithms were considered; i.e., the Bayesian regularisation, scaled
conjugate gradient and Levenberg-Marquardt algorithms. The basic principles and
performance of each algorithm are described as follows.
6.2.2.1 Bayesian regularisation algorithm
The Bayesian regularisation algorithm updates the weights and bias values by minimising the
combination of squared errors and weights. The weights are considered to be random variables.
Then, the density function for the weights can by written based on Bayes’ rule:
𝑃𝑃�𝐷𝐷 �𝑤𝑤, 𝛽𝛽, 𝑀𝑀 �𝑃𝑃�𝐷𝐷 �𝑤𝑤, 𝛼𝛼, 𝑀𝑀 �

𝑃𝑃(𝑤𝑤|𝐷𝐷, 𝛼𝛼, 𝛽𝛽, 𝑀𝑀) = (6.1)
𝑃𝑃�𝐷𝐷 �𝛼𝛼, 𝛽𝛽, 𝑀𝑀 �
where α and β are objective function parameters; D represents the data set; M is the particular
neural network model used; and w is the vector of the network weights.
The optimal weights have a maximum posterior probability 𝑃𝑃(𝑤𝑤|𝐷𝐷, 𝛼𝛼, 𝛽𝛽, 𝑀𝑀), which is the same
as minimising the regularised objective function 𝐹𝐹 = 𝛽𝛽𝐸𝐸𝐷𝐷 + 𝛼𝛼𝐸𝐸𝑤𝑤 .
where 𝐸𝐸𝐷𝐷 is the sum of the squared errors and 𝐸𝐸𝑤𝑤 is the sum of the squares of the network
weights.
Based on this theory, the parameters can be solved with:
𝛾𝛾
𝛼𝛼 𝑀𝑀𝑀𝑀 = (6.2)
2𝐸𝐸𝑤𝑤 (𝑤𝑤 𝑀𝑀𝑀𝑀 )
𝑛𝑛−𝛾𝛾
𝛽𝛽𝑀𝑀𝑀𝑀 = (6.3)
2𝐸𝐸𝐷𝐷 (𝑤𝑤 𝑀𝑀𝑀𝑀 )
127
𝛾𝛾 = 𝑁𝑁 − 2𝛼𝛼 𝑀𝑀𝑀𝑀 𝑡𝑡𝑡𝑡(𝐻𝐻 𝑀𝑀𝑃𝑃 )−1 (6.4)
where H is the Hessian matrix of the objective function; 𝛾𝛾 is the effective number of
parameters; and N is the total number of parameters in the network. The optimal values are
obtained for 𝛼𝛼 and 𝛽𝛽 at the minimum point by setting them equal to zero. A detailed description
and derivation are given in Foresee & Hagan (1997).
Figure 6.1 Regression process of Bayesian regularization algorithm
128
Figure 6.2 Performance of Bayesian regularization algorithm
Figure 6.1 shows the regression and R values with the Bayesian regularisation algorithm, and
Figure 6.2 shows its performance. It is evident that the Bayesian regularisation algorithm also
works well for this problem based on the two figures. The R value of all of the data is only
0.00018 which is less than that with the Levenberg-Marquardt algorithm (see below).
6.2.2.2 Scaled conjugate gradient algorithm
The scaled conjugate gradient algorithm is another efficient algorithm used in numerical
calculation and optimisation methods. This method uses a step size scaling mechanism, which
avoids a time-consuming line search for each learning iteration which is the case for the
Levenberg-Marquardt approach (Møller, 1993). This algorithm is an iterative method that is
used to numerically solve large scale linear and non-linear equations due to its high speed of
convergence and precision (Atkinson, 1988).
129
The process is as follows: solving a linear system 𝐴𝐴𝐴𝐴 = 𝑏𝑏 is equal to finding the minimum
1
value of the function 𝑓𝑓(𝑥𝑥) = 𝑥𝑥 𝑇𝑇 𝐴𝐴𝐴𝐴 − 𝑥𝑥 𝑇𝑇 𝑏𝑏. The residual of the kth step is 𝑟𝑟𝑘𝑘 = 𝑏𝑏 − 𝐴𝐴𝑥𝑥𝑘𝑘 . Then,
2
the conjugate vector 𝑝𝑝𝑘𝑘 can be written as:
𝑝𝑝𝑖𝑖𝑇𝑇 𝐴𝐴𝑟𝑟𝑘𝑘
𝑝𝑝𝑘𝑘 = 𝑟𝑟𝑘𝑘 − ∑𝑖𝑖<𝑘𝑘 𝑝𝑝𝑖𝑖 (6.5)
𝑝𝑝𝑖𝑖𝑇𝑇 𝐴𝐴𝑝𝑝𝑖𝑖
Then,
𝑥𝑥𝑘𝑘+1 = 𝑥𝑥𝑘𝑘 + 𝛼𝛼𝑘𝑘 𝑝𝑝𝑘𝑘 (6.6)
𝑝𝑝𝑘𝑘𝑇𝑇 𝑏𝑏 𝑝𝑝𝑘𝑘𝑇𝑇 (𝑟𝑟𝑘𝑘 +𝐴𝐴𝑥𝑥𝑘𝑘 ) 𝑝𝑝𝑘𝑘𝑇𝑇 𝑟𝑟𝑘𝑘

𝛼𝛼𝑘𝑘 = = = (6.7)
𝑝𝑝𝑘𝑘𝑇𝑇 𝐴𝐴𝑝𝑝𝑘𝑘 𝑝𝑝𝑘𝑘𝑇𝑇 𝐴𝐴𝑝𝑝𝑘𝑘 𝑝𝑝𝑘𝑘𝑇𝑇 𝐴𝐴𝑝𝑝𝑘𝑘
Therefore, the advantages of this algorithm are high speed of convergence and precision, which
is suitable to solve large scale linear and nonlinear equations. (Atkinson, 1988)
130
Figure 6.3 Regression process of Scaled conjugate gradient algorithm
Figure 6.4 Performance of Scaled conjugate gradient algorithm
131
Based on the two figures, it is evident that the Scaled conjugate gradient algorithm is not
suitable for this problem. The data points are relatively irregularly distributed, therefore the R
values shown in Figure 6.3 are relatively lower than those of the other two algorithms. The
regression curves shown in Figure 6.4 are not smooth for all the epochs, which means that there
may be a nonconvergence problem.
6.2.2.3 Levenberg-Marquardt algorithm
The Levenberg-Marquardt algorithm has both the advantages of Gauss-Newton algorithm and
gradient descent to find a local minimum. This algorithm was first proposed by a statistician in
the USA called Kenneth Levenberg, which was then rediscovered by another American
statistician, Donald Marquardt. In this algorithm, the Hessian matrix is approximated by using:
𝐻𝐻 = 𝐽𝐽𝑇𝑇 𝐽𝐽 (6.8)
Then, the gradient can be calculated with:
𝑔𝑔 = 𝐽𝐽𝑇𝑇 𝑒𝑒 (6.9)
where J is a Jacobian matrix that contains the first derivatives of the network errors with respect
to the weights and biases, and e is a vector of the network errors.
Through this approximation, the Levenberg-Marquardt algorithm can be written as:
𝑥𝑥𝑘𝑘+1 = 𝑥𝑥𝑘𝑘 − [𝐽𝐽𝑇𝑇 𝐽𝐽 + 𝜇𝜇𝜇𝜇]−1 𝐽𝐽𝑇𝑇 𝑒𝑒 (6.10)
where μ is a scalar. This scalar will decrease after a successful step (that is, there is a reduction
in performance function) (Hagan & Menhaj, 1994). This algorithm is very efficient in this way.
Figure 6.5 shows the regression and R values of each set of data. All of the R values are more
than 0.9, which points to a strong linear relationship. The performance of the Levenberg-
132
Marquardt algorithm is shown in Figure 6.6 through a measure called the mean squared error
(MSE). Figure 6.6 shows the overall regression trend and the curves that are parallel to the base
line. The curves are relatively smooth, which proves that this algorithm avoids the problems of
non-convergence and trapping in the local minima. Therefore, this algorithm which is one of
the fastest and most mature backpropagation algorithms, shows a superior performance and
will be used to train the ML model in this study. Furthermore, the multilayer structure includes
an input layer of four nodes (the four inputs are the Young’s moduli of the four parts of the
sports bras), two hidden layers, and an output layer of three nodes (the predicted contact
pressure at three different locations on the sports bras), as shown in Figure 6.7.
Figure 6.5 Regression process of Levenberg-Marquardt algorithm
133
Figure 6.6 Performance of Levenberg-Marquardt algorithm
Figure 6.7 ANNs structure
6.2.3 Validation
A comparison was made between the FE-simulated results and the ANN-predicted contact
pressure of all of the data sets. Table 6.2 shows the RMSEs between the FE and the ANN
results. The equation for the RMSE is shown as Equation (6.11). The RMSE values indicate
134
that there is only a small difference between the ANN and the FE results, especially for the
contact pressure on the shoulders.
1 𝑃𝑃𝐹𝐹𝐹𝐹𝐹𝐹,𝑖𝑖 −𝑃𝑃𝐴𝐴𝐴𝐴𝐴𝐴,𝑖𝑖 2
RMSE =
𝑛𝑛
�∑𝑛𝑛𝑖𝑖=1( 𝑃𝑃𝐹𝐹𝐹𝐹𝐹𝐹,𝑖𝑖
) (6.11)
where PFEM is the contact pressure calculated by using the FE method, PANN is the contact
pressure predicted by using ANN, and n is the number of sample data points.
Table 6.2 RMSE between FE and ANN results
Predicted contact pressure RMSE
Contact pressure on shoulders 0.0211
Contact pressure at the bottom of cup 0.0649
Contact pressure under the arms 0.0404
After the ANN model was established, obtaining all 625 sets of data or any desired result can
be simple and fast. Some of the ANN results were used to investigate the relationship between
the contact pressure and the elastic modulus of a sports bra, as shown in Figure 6.8. The plotted
information easily shows the effect of the elastic modulus of the shoulder strap, back panel, bra
cup, and elastic bra band on the contact pressure. The Young’s modulus of the other three parts
was selected at the same level (the highest or lowest level). It was found that the elasticity of
the shoulder strap and bra cup fabric have the greatest impact on contact pressure. However,
their impacts are reduced as the elastic modulus of the other three parts is reduced. The back
panel has basically no effect on the contact pressure. The elastic bra band effectively changes
135
the contact pressure under the arms. However, it has much less influence on the contact
pressure at the shoulders and bottom of the bra cup than the shoulder straps and cup fabric.
Hence, it is more important to choose relatively rigid fabrics for the shoulder straps and bra
cups when designing high-impact sports bras. When combined with research on the pressure
comfort of bras, the results obtained herein can directly be used as a guide for choosing sports
bra fabrics.
(a) Effect of elastic modulus of bra cups on contact pressure at three points (shoulders,
under the arms and bottom of bra cup)
136
(b) Effect of elastic modulus of bra straps on contact pressure at three points (shoulders,
(c) Effect of elastic modulus of back panel on contact pressure on shoulders, under the
arms and at bottom of bra cup
137
(d) Effect of elastic modulus of bra band on contact pressure at three points (shoulders,
Figure 6.8 Effects of fabric of sports bra components with different elastic moduli
6.3 Finite element models of different subjects
After the successful simulation of the contact conditions between the sports bra and the body
of S01, FE contact models for the other subjects were then built. The number of meshes differs
between each FE model due to the differences in the geometric models and pre-processing
steps, which are provided in Table 6.3.
Table 6.3 Number of meshes of FE model of different subjects
No. of meshes
Subject
Bra Torso Body Breast
S02 25,496 128,360 129,656 57,700
138
S03 33,662 176,482 167,104 68,758
S04 29,227 122,463 152,811 50,679
S05 19,496 105,098 116,784 30,272
S06 30,400 106,677 137,426 54,915
S07 22,366 119,965 132,759 54,866
S08 31,956 122,223 165,179 65,151
S09 28,330 114,340 134,670 39,530
S10 27,497 134,868 150,415 66,234
S11 36,769 129,644 173,188 61,732
S12 35,948 157,253 145,849 45,888
S13 26,915 112,622 153,270 61,966
S14 22,461 139,453 131,504 38,833
S15 30,678 122,644 161,538 74198
The material parameter for the FE models of the sports bra is the tested elastic modulus of
Sports Bra 5 as discussed in Chapter 4 because it is the simplest bra, and the geometric model
of the tested bra is most similar to Sports Bra 5. The contact pressure between the FE model of
the body of the fourteen other subjects and the sports bras was simulated by using the same
method which is to first shrink and then expand the model. The distribution of the contact
pressure of each subject is very similar; see Figure 6.9.
139
Figure 6.9 Distribution of contact pressure of different subjects
Due to the irregular shape of the model from the scanned images of the body and the complexity
of contact, it was challenging to conduct FE modelling of the contact between the sports bra
and different subjects. Hence, the meshes of the contact area were refined for most of the FE
contact models. The size of the meshes in the contact area is 4 mm and that of the other parts
is 6 mm. Figure 6.3 also shows that the meshes in the centre of the model are smaller except
for S15. This method improves the success rate of computation without incurring very much
additional computational time.
140
To further investigate the differences between the subjects, the contact pressure values of three
locations; that is, the shoulders, under the arms, and bottom of the bra cup, are simulated, which
are listed in Table 6.4.
Table 6.4 Simulated contact pressure of different subjects
Simulated contact pressure (kPa)
Subject Age Cup size Bottom of bra

Shoulders Under the arms
cup
S02 26 38C 1.2 0.9 0.9
S03 26 34C 1.3 1.1 0.7
S04 33 32C 1.0 0.8 0.6
S05 24 32B 1.9 1.5 1.0
S06 60 32C 1.3 1.1 0.9
S07 32 32B 1.3 1.1 0.6
S08 59 38D 1.9 1.7 0.9
S09 54 34C 2.5 1.9 1.1
S10 19 34D 1.1 0.9 0.6
S11 59 32C 2.0 1.6 0.8
S12 54 36C 0.9 0.7 0.4
S13 23 34D 1.1 2.0 1.0
S14 24 32B 1.9 1.5 1.0
S15 25 34C 2.0 2.1 0.8
141
Table 6.4 shows that the amount of contact pressure exerted by the same sports bra differs for
each subject. The largest difference is 1.5 kPa which is on the shoulders. There are no obvious
relationships among age, bra cup size and contact pressure. The correlation coefficients are
shown in Table 6.5 (significance level, 0.05).
Table 6.5 Correlation coefficients between contact pressure and subject particulars
Simulated contact pressure (kPa)

Subject Particulars
Shoulders Under the arms Bottom of bra cup
Age 0.263 0.031 0.005
Cup size 0.232 0.121 0.056
To investigate the effect of the elastic modulus of the different sports bra components on the
contact pressure that is shown in Figure 6.2, the other fourteen FE models of different subjects
were calculated with different elastic modulus of each part of the bra. Two different elastic
moduli (0.50 MPa and 1.33 MPa) were examined for the shoulder strap. The higher elastic
modulus causes the contact pressure at the three stated locations to increase significantly,
considering that the contact pressure is low. These changes validate previous findings.
142
(a) Change in contact pressure on shoulders (b) Change in contact pressure under the arms
(c) Change in contact pressure at bottom of bra cup
Figure 6.10 Change in contact pressure with increase in elastic modulus of bra strap
Similarly, two different elastic moduli of 0.50 MPa and 0.95 MPa were examined for the bra
cup. According to a previous finding, the elasticity of the bra cup fabric can greatly influence
the contact pressure. The changes in contact pressure as shown in Figure 6.10 are in agreement
with the predicted values from the relationship we discovered.
143
(a) Change in contact pressure on shoulders (b) Change in contact pressure under the arms
(c) Change in contact pressure at bottom of cup
Figure 6.11 Change in contact pressure with increase in elastic modulus of bra cup
Figures 6.10 and 6.11 both show that an increased elastic modulus of the shoulder strap and
bra cup can greatly change the contact pressure between the body and sports bra. To directly
shows the magnitude of the changes, Table 6.6 shows the amount of increase in contact
pressure with increases in the elastic modulus of the shoulder strap and bra cup.
144
Table 6.6 Increase in contact pressure with higher elastic modulus of bra strap and bra cup
Subject Location Change of contact pressure (%)
Increasing elastic modulus Increasing elastic modulus
of shoulder strap of bra cup
S02 Shoulders 16.7 16.7
Under arm 11.1 22.2
Under arm 11.1 45.4
Underarm 12.5 25.0
Under arm 46.6 20.0
Under arm 45.5 27.3
Under arm 36.4 36.4
Under arm 17.6 23.5
145
Under arm 21.1 26.3
Under arm 22.2 33.3
Underarm 12.5 18.8
S12 Shoulders 11.1 0
Under arm 14.3 14.3
Under arm 25.0 25.0
Under arm 13.3 20.0
Under arm 23.8 23.8
The elastic modulus of the elastic bra band for all of the FE models increases by 0.50 MPa,
wherein half of the FE models showed a 0.1 kPa increase in pressure on the shoulders and at
146
the bottom of the bra cup, and 65% of the FE models showed a 0.1 kPa increase in pressure
under the arms. As for the effect of the back panel, 25% of the FE models showed a 0.1 kPa
increase in pressure on the shoulders and half of the FE models showed a 0.1 kPa increase in
pressure under the arms when the elastic modulus of the back panel for all of the FE models is
increased by 0.5 MPa. In other cases, the simulated contact pressure remains unchanged.
Therefore, these results, which are that the shoulder strap and bra cup can greatly influence the
contact pressure, while the back panel and elastic bra band do not have much influence the
contact pressure, can be used to specify the material parameters of each component of the sports
bra as they impact the amount of contact pressure on different subjects who differ in age, cup
size and BMI. They can also be regarded as generalised results that can guide the future design
of sports bras.
6.4 Conclusions
This chapter first provides a description of an FE model-based ML method to predict the
contact pressure between a sports bra and wearer. With appropriate material coefficients and
boundary conditions, the experimental results show that FE simulation can accurately calculate
the contact pressure between a sports bra and a human. Hence, the FE results can be used as
input data to establish an ANN model. ANNs are an effective tool for making predictions.
Specifically, the ANN results of the contact pressure, which applies the Levenberg-Marquardt
learning algorithm, are in good agreement with the corresponding FE results based on the
calculated RMSEs. Therefore, the prediction system discussed in this chapter can be used to
calculate the contact pressure between a sports bra and wearer within a limit of allowable errors.
The contact pressure between a sports bra and fourteen different subjects is then measured. The
distribution of the contact pressure is very similar in all of the FE models of the different
147
subjects. Moreover, the effect of the different components of the sports bra on the contact
pressure is based on the same result for all of the subjects, which is that the shoulder strap and
bra cup can greatly influence the contact pressure, while the back panel and elastic bra band do
not have much influence the contact pressure. These results can be directly used for selecting
sports bra materials, and positively benefit a large percentage of the female population
worldwide in terms of their bra needs for physical activities.
The concepts in this chapter have a limitation, which is the sports bra style. Both the FE and
ANN models are based on the same vest-style compression sports bra. Encapsulation sports
bras are also popular in the market and future works should consider other styles of sports bras.
148
Chapter 7 Conclusions and Recommendations for Future Work
7.1 Introduction
Wearing a sports bra is important for controlling movement of the breasts during physical
activities due to the substantial displacement which can cause much discomfort and even injury.
However, if a sports bra exerts excessive pressure onto the body of the wearer, she may also
feel uncomfortable. Therefore, this thesis has provided a nonlinear FEA to predict the
mechanical behavior of the female breasts when in contact with a compression sports bra with
a focus on the optimization of the fabric properties of the sports bra. The major achievements
and important findings in this study are summarized as follows.
7.2 Development of finite element models
Chapters 3 and 4 have systematically presented the methods and process of developing FE
models of the body and the contact between the female body and sports bras. The latter includes
a rigid torso, breasts with hyperelasticity and layers of subcutaneous tissues. A 3D laser
scanning experiment has been conducted to obtain the geometric models of the body. The
material properties of the breasts are determined through motion capture experiments of
running braless and free vibration of the breasts. The Mooney-Rivlin material coefficients of
the breasts are determined by examining the differences between the FE-modelled results and
experimental data, and the damping ratio is directly calculated from the free-vibration test data.
The FE contact model of the body and sports bra is built by using a method which first shrinks
and then expands the model to avoid the penetration problem. This method has four steps: a)
first, the size of the original model of the body and bra is geometrically reduced; b) the model
149
of the body is then reduced, during which the gap between the models of the body and the
sports bra is noted; c) subsequently, the model of the body is geometrically expanded by adding
a thermal expansion property, which means that the volume increases with temperature; and d)
finally, a gravity load is applied to the breasts. In the static analysis, the simulated contact
pressure values are in agreement with the results of the contact pressure tests. In terms of the
distribution of contact pressure, the highest contact pressure is found on the shoulders which is
caused by the shoulder strap, and under the arms which is caused by the elastic bra band. The
highest contact pressure exerted by the bra cup is found at the bottom of the bra cup. In the
dynamic analysis, the displacement of breasts is investigated and analysed based on contact
with the sports bra. The calculated RMSEs from the results of the FE model and experiments
show that the contact model can predict the contact conditions with good accuracy. A modal
analysis has also been conducted by using the FE models and the corresponding natural
frequencies are calculated. The natural frequencies of the breasts with a sports bra are found to
be much higher than those of the braless condition, which confirms the necessity of wearing a
sports bra.
7.3 Pressure comfort evaluation of sports bra
Chapter 5 presents the relationship between the contact pressure and scores of perceived
comfort and pressure comfort range. The results are determined by using objective
experimental data and subjective perceptions. The objective experimental data are the tested
contact pressure at three positions (on the shoulders and underarm and at the bottom of the bra
cups) by using Novel Pliance-X pressure sensors. Furthermore, the subjective perceptions of
the bras are derived from a questionnaire that focused on how the subjects felt about the sports
bra samples. The relationship between these two results is investigated in both static and
150
dynamic conditions by using a statistical analyses, which are correlation coefficient and
ANOVA.
The statistical analyses showed that the overall score for perceived comfort is strongly
influenced by the contact pressure at the underarm and the shoulder strap affects the perceived
comfort the most in the overall score. The static and dynamic conditions provide different
findings. In the static condition, the contact pressure on the shoulders caused by the shoulder
straps has a significant (p ＜ 0.05 ) impact on the scores of perceived pressure from the bra
cup. The contact pressure at the bottom of the bra cup has a significant impact on the scores of
perceived pressure from the elastic bra band and overall scores of perceived pressure. In the
dynamic condition, the contact pressure on the shoulders from the shoulder straps significantly
affects the scores of perceived pressure from the elastic bra band. The contact pressure at the
underarm significantly affects the scores of perceived pressure from the bra cup. Finally, the
contact pressure at the bottom of the bra cup significantly affects the scores of perceived
pressure from the shoulder strap. As for the range of pressure comfort, the optimum contact
pressure or most tolerable amount of pressure on the shoulders and underarm and at the bottom
of the bra cup are in the ranges of 2.08 ± 0.82 kPa, 2.66 ± 0.83 kPa and 0.58 ± 0.57 kPa
respectively.
7.4 Prediction of contact pressure between sports bra and body
An FE model-based ML method to predict the contact pressure between the sports bra and
body is described in Chapter 6. The data to train an ANN model are obtained from the FE
results. Three backpropagation algorithms are then taken into consideration, but the Levenberg-
Marquardt algorithm has the best performance. In addition, the ANN results of the contact
151
pressure, which uses the Levenberg-Marquardt learning algorithm, are in good agreement with
the corresponding FE results based on the calculated RMSEs.
The contact pressure exerted onto the body of the other subjects by the sports bra is also
calculated in Chapter 6. The distribution of the contact pressure is very similar in all of the FE
models of the different subjects. The effect of the different components of the sports bra on the
contact pressure is then investigated, and it is found that the shoulder strap and bra cup can
greatly influence the contact pressure, while the back panel and elastic bra band do not
influence the contact pressure very much.
7.5 Limitations and future work
While there are important findings in this study, there are also some limitations of the work
done. First, there is the small number of subjects and the sports bra style used is limited. The
sample involves fifteen female subjects, and only one style of compression sports bra (vest
style) is considered and analysed. Therefore, the results of the FE analysis and ML models are
only applicable for vest-style compression sports bras. Future work needs to involve more or
other types of female subjects and different styles of sports bras, such as encapsulation sports
bras.
In addition, the FE models of the female body have been simplified in this study. The structure
of the female breasts is highly complex and has many different components. The FE models of
the breasts are however assumed to be homogeneous for less time costly pre-processing and
modelling calculations. The development of more complex FE models of the breasts that
include different components, such as fat and Cooper’s ligaments, are recommended to be
carried out in future studies. The material properties of each component should also be
determined.
152
Moreover, the displacements of the breast are represented by the displacements of the nipple,
instead of the entire breast. The movement with the different positions of the breast is slightly
different during physical activities due to the complex structure of the breasts. Therefore,
reflective markers should be placed in other positions or areas of the breast during the motion
capture experiment to record the movement of the entire breast in future studies.
Finally, contact pressures between the sports bras and subjects’ bodies at only three positions
have been considered. It was not enough for the design of a sports bra. For instance, how much
pressure the subjects’ backs can bare is also a question worth being analysed.
153
REFERENCES
Alpaydin, E. (2009). Introduction to machine learning. MIT press.

Atta, M., Abd-Elhady, A., Abu-Sinna, A., et al. (2019). Prediction of failure stages for double
lap joints using finite element analysis and artificial neural networks. Engineering Failure
Analysis, 97, 242-257.
Arch, E. S., Colón, S., & Richards, J. G. (2018). A Comprehensive method to measure 3-
dimensional bra motion during physical activity. Journal of Applied Biomechanics, 34(5), 392-
395.
Azar, F. S., Metaxas, D. N., Schnall, M. D. (2000). A finite element model of the breast for
predicting mechanical deformations during biopsy procedures. Paper presented at the
Mathematical Methods in Biomedical Image Analysis, 2000. Proceedings. IEEE Workshop on.
Azar, F. S., Metaxas, D. N., Schnall, M. D. (2002). Methods for modeling and predicting
mechanical deformations of the breast under external perturbations. Medical Image Analysis,
6(1), 1-27.
Balachandran, B. (2009). Vibrations (2nd ed.. ed.). Toronto, Ont.: Toronto, Ont.: Cengage
Learning.
Bel-Brunon, A., Bouten, L., Cornolo, J., & Morestin, F. (2014, July). Numerical modeling of
bra wear during running. In 11th World Congress on Computational Mechanics.
Bel-Brunon, A., Kehl, S., Martin, C., Uhlig, S., & Wall, W. A. (2014). Numerical identification
method for the non-linear viscoelastic compressible behavior of soft tissue using uniaxial
tensile tests and image registration–application to rat lung parenchyma. Journal of the
Mechanical Behavior of Biomedical Materials, 29, 360-374.
Boisse, P., Borr, M., Buet, K., & Cherouat, A. (1997). Finite element simulations of textile
composite forming including the biaxial fabric behaviour. Composites Part B: Engineering,
28(4), 453-464.
Bottou, L., Curtis, F. E., Nocedal, J. (2018). Optimization methods for large-scale machine
learning. Siam Review, 60(2), 223-311.
Bowles, K. A., Steele, J. R., & Munro, B. J. (2008). What are the breast support choices of
Australian women during physical activity? British Journal of Sports Medicine, 42(8), 670-
673.
Bowles, K. A., Steele, J. R., & Munro, B. J. (2012). Features of sports bras that deter their use
by Australian women. Journal of Science and Medicine in Sport, 15(3), 195-200.
Bowles, K. A., & Steele, J. R. (2013). Effects of strap cushions and strap orientation on comfort
and sports bra performance. Medicine and Science in Sports and Exercise, 45(6), 1113-1119.
Brueck, C. M. (2000). The role of topical lubrication in the prevention of skin friction in
physically challenged athletes. Journal of Sports Chiropractic and Rehabilitation, 14(2), 37-39.
Cai, Y., Chen, L., Yu, W., Zhou, J., Wan, F., Suh, M., & Chow, D. H. K. (2018). A piecewise
mass-spring-damper model of the human breast. Journal of Biomechanics, 67, 137-143.
Carbonell, J. G., Mitchell, T. M., & Michalski, R. S. (Eds.). (1984). Machine learning: An
artificial intelligence approach. Springer-Verlag.
Celebi, M. E., & Aydin, K. (Eds.). (2016). Unsupervised learning algorithms. Berlin: Springer
International Publishing.
Chapelle O., Schölkopf B., Zien A. (2006). Semi-Supervised learning (1st), The MIT Press.
Chen, L. H., Ng, S. P., Yu, W., Zhou, J., & Wan, K. F. (2013). A study of breast motion using
non-linear dynamic FE analysis. Ergonomics, 56(5), 868-878.
Chen, X., Gho, S. A., Wang, J., & Steele, J. R. (2016). Effect of sports bra type and gait speed
on breast discomfort, bra discomfort and perceived breast movement in Chinese women.
Ergonomics, 59(1), 130-142.
Coltman, C. E., McGhee, D. E., & Steele, J. R. (2015). Bra strap orientations and designs to
minimise bra strap discomfort and pressure during sport and exercise in women with large
breasts. Sports mMedicine-Open, 1(1), 21.
Coltman, C. E., Steele, J. R., McGhee, D. E. (2019). Does breast size affect how women
participate in physical activity? Journal of Science and Medicine in Sport, 22(3), 324-329.
Conti, G., Cristofolini, L., Juszczyk, M., Leardini, A., & Viceconti, M. (2008). Comparison of
three standard anatomical reference frames for the tibia–fibula complex. Journal of
Biomechanics, 41(16), 3384-3389.
Costantakos, A. V., Watkins, S. M. (1982). Pressure analysis as a design research technique
for increasing the comfort of nursing brassieres. Family and Consumer Sciences Research
Journal, 10(3), 271-278.
Dai, X. Q., Li, Y., Zhang, M., & Cheung, J. T. M. (2006). Effect of sock on biomechanical
responses of foot during walking. Clinical Biomechanics, 21(3), 314-321.
Del Palomar, A. P., Calvo, B., Herrero, J., López, J., & Doblaré, M. (2008). A finite element
model to accurately predict real deformations of the breast. Medical Engineering & Physics,
30(9), 1089-1097.
Delingette, H. (1998). Toward realistic soft-tissue modeling in medical simulation.
Proceedings of the IEEE, 86(3), 512-523.
Della Croce, U., Leardini, A., Chiari, L., & Cappozzo, A. (2005). Human movement analysis
using stereophotogrammetry: Part 4: assessment of anatomical landmark misplacement and its
effects on joint kinematics. Gait & Posture, 21(2), 226-237.
Denton, M. (1972). Fit, stretch and comfort. Textiles, 3(1), 12-17.
Dorgan, J. F., Klifa, C., Shepherd, J. A., et al. (2012). Height, adiposity and body fat
distribution and breast density in young women. Breast Cancer Research, 14(4), R107.
doi:10.1186/bcr3228
Drake, R., Vogl, A. W., Mitchell, A. W. (2009). Gary’s Anatomy for Students E-Book: Elsevier
Health Sciences.
Dufaye, G., Cherouat, A., Bachmann, J. M., & Borouchaki, H. (2013). Advanced finite element
modelling for the prediction of 3D breast deformation. European Journal of Computational
Mechanics/Revue Européenne de Mécanique Numérique, 22(2-4), 170-182.
Eder, M., Raith, S., Jalali, J., Volf, A., Settles, M., Machens, H. G., & Kovacs, L. (2014).
Comparison of different material models to simulate 3-D breast deformations using finite
element analysis. Annals of Biomedical Engineering, 42(4), 843-857.
Ellis, H., & Mahadevan, V. (2013). Anatomy and physiology of the breast. Surgery (Oxford),
31(1), 11-14.
Fix, G. J. (1975). Finite element models for ocean circulation problems. SIAM Journal on
Applied Mathematics, 29(3), 371-387.
Foresee, F. D., & Hagan, M. T. (1997, June). Gauss-Newton approximation to Bayesian
learning. In Proceedings of International Conference on Neural Networks (ICNN'97) (Vol. 3,
pp. 1930-1935). IEEE.
Freutel, M., Schmidt, H., Dürselen, L., Ignatius, A., & Galbusera, F. (2014). Finite element
modeling of soft tissues: material models, tissue interaction and challenges. Clinical
Biomechanics, 29(4), 363-372.
Fung, Y., Cowin, S. (1994). Biomechanics: mechanical properties of living tissues. Journal of
Applied Mechanics, 61, 1007
Fuseya, S., & Matsumoto, C. (2006). Three dimensional study on the vibration of breasts with
brassieres. Wayo Joshi Daigaku Kiyo. Kaseikeihen, 46, 1-12.
Gefen, A., Dilmoney, B. (2007). Mechanics of the normal woman's breast. Technology and
Health Care, 15(4), 259-271.
Ghaffarianjam, H. R., & Abolbashari, M. H. (2010). Performance of the evolutionary structural
optimization-based approaches with different criteria in the shape optimization of beams. Finite
Elements in Analysis and Design, 46(4), 348-356.
Gho, S. A., Steele, J. R. and Munro, B. J. (2008) Can reductions in bra band pressure increase
comfort during exercise in lumpectomy patients?. Emed Scientific Meeting 2008, Dundee,
Scotland, 28-31 July.
Glaessgen, E. H., Pastore, C. M., Griffin, O. H., & Birger, A. (1996). Geometrical and finite
element modelling of textile composites. Composites Part B: Engineering, 27(1), 43-50.
Haake, S., & Scurr, J. (2010). A dynamic model of the breast during exercise. Sports
Engineering, 12(4), 189-197.
Hagan, M. T., & Menhaj, M. B. (1994). Training feedforward networks with the Marquardt
algorithm. IEEE Transactions on Neural Networks, 5(6), 989-993.
Hamila, N., Boisse, P., Sabourin, F., & Brunet, M. (2009). A semi‐discrete shell finite element
for textile composite reinforcement forming simulation. International Journal for Numerical
Methods in Engineering, 79(12), 1443-1466.
Haycock, C., Shierman, G., Gillette, J. (1978). The female athlete–does her anatomy pose
problems. Paper presented at the Proceedings of the American Medical Association 19th
Conference on the Medical Aspects of Sports.
Hsu, C. M., Palmeri, M. L., Segars, W. P., Veress, A. I., & Dobbins, J. T. (2011). An analysis
of the mechanical parameters used for finite element compression of a high‐resolution 3D
breast phantom. Medical Physics, 38(10), 5756-5770.
Hu, J. (2004). Structure and mechanics of woven fabrics. Elsevier.
Huang, J., Kwok, T.-H., Zhou, C. (2019). Parametric design for human body modeling by
wireframe-assisted deep learning. Computer-Aided Design, 108, 19-29.
Ishimaru, S., Isogai, Y., Matsui, M., et al. (2011). Prediction method for clothing pressure
distribution by the numerical approach: attention to deformation by the extension of knitted
fabric. Textile Research Journal, 81(18), 1851-1870.
Jamkrajng, P., & Limroonngreungrat, W. (2014). Comparison of vertical breast displacement
between two different sports bra designs. Turkish Journal of Sport and Exercise, 16(2), 97-99.
Jang, I. G., & Kim, I. Y. (2010). Application of design space optimization to bone remodeling
simulation of trabecular architecture in human proximal femur for higher computational
efficiency. Finite Elements in Analysis and Design, 46(4), 311-319.
Jang, Y., Chun, J., Lee, H., & Han, B. (2013). A study on the design features for sports bra
styles according to treadmill running speeds and bra cup sizes. The Research Journal of the
Costume Culture, 21(1), 81-92.
Jin, C., He, Z., & Liu, J. (2014). MRI-based finite element simulation on radiofrequency
ablation of thyroid cancer. Computer Methods and Programs in Biomedicine, 113(2), 529-538.
Kharya, S., Dubey, D., & Soni, S. (2013). Predictive machine learning techniques for breast
cancer detection. International Journal of Computer Science and Information Technologies,
4(6), 1023-1028.
Kobayashi, T., Oi, S., Sato, M., et al. (2011). Analysis of clothing pressure on the human body.
Paper presented at the SIMULIA Customer Conference.
Kotsiantis, S. B., Zaharakis, I., & Pintelas, P. (2007). Supervised machine learning: A review
of classification techniques. Emerging Artificial Intelligence Applications in Computer
Engineering, 160, 3-24.
Krouskop, T. A., Wheeler, T. M., Kallel, F., et al. (1998). Elastic moduli of breast and prostate
tissues under compression. Ultrasonic Imaging, 20(4), 260-274.
Kuhlmann, M., Fear, E. C., Ramirez-Serrano, A., & Federico, S. (2013). Mechanical model of
the breast for the prediction of deformation during imaging. Medical Engineering & Physics,
35(4), 470-478.
Landau, L. D., Lifshitz, E. M. (1969). Mechanics.
Lawson, L., Lorentzen, D. (1990). Selected sports bras: comparisons of comfort and support.
Clothing and Textiles Research Journal, 8(4), 55-60.
Lee, Y. A., Hyun, K. J., & Tokura, H. (2001). Circadian rhythms of core body temperature and
urinary noradrenaline secretion under the influence of skin pressure due to foundation garments
worn during wakefulness. Biological Rhythm Research, 32(3), 389-400.
Lerner, R. M. (1987). Sono-elasticity in ultrasonic tissue characterization and echographic
imaging. In Proc. 7th Eur. Comm. Workshop, Nijmegen, The Netherlands, 1987.
Lerner, R. M., Parker, K. J., Holen, J., Gramiak, R., & Waag, R. C. (1988). Sono-elasticity:
Medical elasticity images derived from ultrasound signals in mechanically vibrated targets. In
Acoustical Imaging (pp. 317-327). Springer, Boston, MA.
Li, Y., Zhang, X., Yeung, K. (2003). A 3D biomechanical model for numerical simulation of
dynamic mechanical interactions of bra and breast during wear. Sen'i Gakkaishi, 59(1), 12-21.
Liang, R., Yip, J., Yu, W., Chen, L., & Lau, N. M. (2019). Numerical simulation of nonlinear
material behaviour: Application to sports bra design. Materials & Design, 108177.
Lin, Y., Choi, K. F., Luximon, A., Yao, L., Hu, J. Y., & Li, Y. (2011). Finite element modeling
of male leg and sportswear: contact pressure and clothing deformation. Textile Research
Journal, 81(14), 1470-1476.
Liu, C., Miao, F. X., Dong, X. Y., & Xu, B. (2018). Enhancing pressure comfort of a bra’s
under-band. Textile Research Journal, 88(19), 2250-2257.
Liu, D., Koric, S., & Kontsos, A. (2019). Parallelized finite element analysis of knitted textile
mechanical behavior. Journal of Engineering Materials and Technology, 141(2).
Liu, H., Chen, D., Wei, Q., & Pan, R. (2013). An investigation into the bust girth range of
pressure comfort garment based on elastic sports vest. Journal of the Textile Institute, 104(2),
223-230.
Liu, K., Wang, J., & Hong, Y. (2017). Wearing comfort analysis from aspect of numerical
garment pressure using 3D virtual-reality and data mining technology. International Journal of
Clothing Science and Technology, 29(2), 166-179.
Liu, R., Kwok, Y. L., Li, Y., Lao, T. T., Dai, X. Q., & Zhang, X. (2007). Numerical simulation
of internal stress profiles and three-dimensional deformations of lower extremity beneath
medical graduated compression stocking (GCS). Fibers and Polymers, 8(3), 302-308.
Lorentzen, D., & Lawson, L. (1987). Selected sports bras: a biomechanical analysis of breast
motion while jogging. The Physician and Sportsmedicine, 15(5), 128-139.
Lu, M. H., Yu, W., Huang, Q. H., et al. (2009). A hand-held indentation system for the
assessment of mechanical properties of soft tissues. IEEE Transactions on Instrumentation and
Measurement, 58(9), 3079-3085.
Luo, G., Houston, V. L., Garbarini, M. A., et al. (2011). Finite element analysis of heel pad
with insoles. Journal of Biomechanics, 44(8), 1559-1565.
Martens, H. H. (1959). Two notes on machine “learning”. Information and Control, 2(4), 364-
379.
Mason, B. R., Page, K. A., & Fallon, K. (1999). An analysis of movement and discomfort of
the female breast during exercise and the effects of breast support in three cases. Journal of
Science and Medicine in Sport, 2(2), 134-144.
Matsumura, T., Umemoto, T., Fujihara, Y., Ueno, E., Yamakawa, M., Shiina, T., & Mitake, T.
(2009, September). Measurement of elastic property of breast tissue for elasticity imaging. In
2009 IEEE International Ultrasonics Symposium (pp. 1451-1454). IEEE.
McKnight, A. L., Kugel, J. L., Rossman, P. J., Manduca, A., Hartmann, L. C., & Ehman, R. L.
(2002). MR elastography of breast cancer: preliminary results. American Journal of
Roentgenology, 178(6), 1411-1417.
McGhee, D. E., Power, B. M., & Steele, J. R. (2007). Does deep water running reduce exercise-
induced breast discomfort?. British Journal of Sports Medicine, 41(12), 879-883.
McGhee, D. (2008). Sports bra fitness.
McGhee, D. E., Steele, J. R., Munro, B. J. (2010). Education improves bra knowledge and fit,
and level of breast support in adolescent female athletes: a cluster- randomised trial. Journal of
Physiotherapy, 56(1), 19-24.
McGhee, D. E., & Steele, J. R. (2010). Breast elevation and compression decrease exercise-
induced breast discomfort. Medicine and Science in Sports and Exercise, 42(7), 1333-1338
McGhee, D. E., Steele, J. R., Zealey, W. J., & Takacs, G. J. (2013). Bra–breast forces generated
in women with large breasts while standing and during treadmill running: Implications for
sports bra design. Applied Ergonomics, 44(1), 112-118.
Miyatsuji, A., Matsumoto, T., Mitarai, S., Kotabe, T., Takeshima, T., & Watanuki, S. (2002).
Effects of clothing pressure caused by different types of brassieres on autonomic nervous
system activity evaluated by heart rate variability power spectral analysis. Journal of
Physiological Anthropology and Applied Human Science, 21(1), 67-74.
Møller, M. F. (1993). A scaled conjugate gradient algorithm for fast supervised learning.
Neural networks, 6(4), 525-533.
Motion Analysis (n.d.) Kestrel 4200. Retrieved (2015) from
https://www.motionanalysis.com/cameras/kestrel-4200
Morooka, H., Fukuda, R., Nakahashi, M., Morooka, H., & Sasaki, H. (2005). Clothing pressure
and wear feeling at under-bust part on a push-up type brassiere. Sen’i Gakkaishi, 61(2), 53-58.
Mullainathan, S., Spiess, J. (2017). Machine learning: an applied econometric approach.
Journal of Economic Perspectives, 31(2), 87-106.
Nandi, B., Dalrymple, T., Yao, J., & Lapczyk, I. (2014, September). Importance of capturing
non-linear viscoelastic material behavior in tire rolling simulations. In Meeting of the Tire
Society (pp. 8-10).
Ngo, D., Scordelis, A. (1967). Finite element analysis of reinforced concrete beams. Paper
presented at the Journal Proceedings.
Niemczyk, S. E., Arnold, L., Wang, L. (2017). Qualitative observations on the design of sports
bras for wear under body armour. KnE Engineering, 2(2), 342-352.
Niwaya, H., Imaoka, H., & Shibuya, A. (1996). Estimation of garment pressure distribution
and it's expression. Sen'i Gakkaishi, 52(2), 76-81.
Nolte, K., Burgoyne, S., Nolte, H., Van der Meulen, J., & Fletcher, L. (2016). The effectiveness
of a range of sports bras in reducing breast displacement during treadmill running and two-step
star jumping. Journal of Sports Medicine and Physical Fitness, 56(11), 1311-1317.
Okabe, K. (2002). Analysis of breast vibration while exercising and its reflection to vibration
proof design of sports brassiere. Descente Sports Science, 23, 180-188.
Okabe, K., & Kurokawa, T. (2004). Relationship between wearing comfort and clothing
pressure for designing comfortable brassieres. Bulletin of Japanese Society for the Science of
Design, 51(3), 31-38.
Patete, P., Iacono, M. I., Spadea, M. F., Trecate, G., Vergnaghi, D., Mainardi, L. T., & Baroni,
G. (2013). A multi-tissue mass-spring model for computer assisted breast surgery. Medical
Engineering & Physics, 35(1), 47-53.
Page, K.-A., Steele, J. (1999). Breast Motion and Sports Brassiere Design. Sports Medicine,
27(4), 205-211.
Park, H. S., Park, K., Kim, Y., & Choi, S. W. (2014). Deformation monitoring of a building
structure using a motion capture system. IEEE/ASME Transactions on Mechatronics, 20(5),
2276-2284.
Ramião, N. G., Martins, P. S., Rynkevic, R., Fernandes, A. A., Barroso, M., & Santos, D. C.
(2016). Biomechanical properties of breast tissue, a state-of-the-art review. Biomechanics and
Modeling in Mechanobiology, 15(5), 1307-1323.
Rajagopal, V., Lee, A., Chung, J. H., Warren, R., Highnam, R. P., Nash, M. P., & Nielsen, P.
M. (2008). Creating individual-specific biomechanical models of the breast for medical image
analysis. Academic Radiology, 15(11), 1425-1436.
Recreational Equipment, Inc. (2019) How to choose sports bras.
https://www.rei.com/learn/expert-advice/sports-bras.html
Reig, B., Heacock, L., Geras, K. J., & Moy, L. (2019). Machine learning in breast MRI. Journal
of Magnetic Resonance Imaging.
Samani, A., Bishop, J., Yaffe, M. J., & Plewes, D. B. (2001). Biomechanical 3-D finite element
modeling of the human breast using MRI data. IEEE Transactions on Medical Imaging, 20(4),
271-279.
Samani, A., Bishop, J., Luginbuhl, C., et al. (2003). Measuring the elastic modulus of ex vivo
small tissue samples. Physics in Medicine and Biology, 48(14), 2183.
Samani, A., Plewes, D. (2004). A method to measure the hyperelastic parameters of ex vivo
breast tissue samples. Physics in Medicine and Biology, 49(18), 4395.
Samani, A., & Plewes, D. (2007). An inverse problem solution for measuring the elastic
modulus of intact ex vivo breast tissue tumours. Physics in Medicine and Biology, 52(5), 1247
Samani, A., Zubovits, J., & Plewes, D. (2007). Elastic moduli of normal and pathological
human breast tissues: an inversion-technique-based investigation of 169 samples. Physics in
Medicine and Biology, 52(6), 1565
Schuster, K. (1979). Equipment update: jogging bras hit the streets. The Physician and Sports
Medicine, 7(4), 125-128.
Scurr, J., White, J., & Hedger, W. (2009). Breast displacement in three dimensions during the
walking and running gait cycles. Journal of Applied Biomechanics, 25(4), 322-329.
Scurr, J., White, J., Hedger, W. (2010). The effect of breast support on the kinematics of the
breast during the running gait cycle. Journal of Sports Sciences, 28(10), 1103-1109.
Scurr, J., White, J., Hedger, W. (2011). Supported and unsupported breast displacement in three
dimensions across treadmill activity levels. Journal of Sports Sciences, 29(1), 55-61.
Shan, J., Alam, S. K., Garra, B., Zhang, Y., & Ahmed, T. (2016). Computer-aided diagnosis
for breast ultrasound using computerized BI-RADS features and machine learning methods.
Ultrasound in Medicine and Biology, 42(4), 980-988.
Shea, M. 2017. The first sports bra was two jockstraps sewn together. New York Post.
https://nypost.com/2017/08/14/the-first-sports-bra-was-two-jockstraps-sewn-together/
Shiina, T. (2013). JSUM ultrasound elastography practice guidelines: basics and terminology.
Journal of Medical Ultrasonics, 40(4), 309-323.
Simoneau, C., Terriault, P., Jetté, B., Dumas, M., & Brailovski, V. (2017). Development of a
porous metallic femoral stem: Design, manufacturing, simulation and mechanical testing.
Materials & Design, 114, 546-556.
Spirka, T. A., Erdemir, A., Ewers Spaulding, S., et al. (2014). Simple finite element models for
use in the design of therapeutic footwear. Journal of Biomechanics, 47(12), 2948-2955.
Srivastava, A., Verma, Y., Rao, K. D., & Gupta, P. K. (2011). Determination of elastic
properties of resected human breast tissue samples using optical coherence tomographic
elastography. Strain, 47(1), 75-87.
Starr, C., Branson, D., Shehab, R., Farr, C., Ownbey, S., & Swinney, J. (2005). Biomechanical
analysis of a prototype sports bra. Journal of Textile and Apparel, Technology and management,
4(3), 1-14.
Stewart, B. W. K. P., & Wild, C. P. (2014). World cancer report, International Agency for
Research on Cancer (IARC), ISBN 978-92-832-0429-9.
Sulzberger, M. B., Cortese Jr, T. A., Fishman, L., Wiley, H. S., & Peyakovich, P. S. (1966).
Studies on blisters produced by friction: I. Results of linear rubbing and twisting technics.
Journal of Investigative Dermatology, 47(5), 456-465.
Sun, Y., Yick, K. L., Yu, W., Chen, L., Lau, N., Jiao, W., & Zhang, S. (2019). 3D bra and
human interactive modeling using finite element method for bra design. Computer-Aided
Design, 114, 13-27.
Sun, Y., Chen, L., Yick, K. L., Yu, W., Lau, N., & Jiao, W. (2019). Optimization method for
the determination of Mooney-Rivlin material coefficients of the human breasts in-vivo using
static and dynamic finite element models. Journal of the Mechanical Behavior of Biomedical
Materials, 90, 615-625.
Sutton R. S., Barto A. G. (1998). Reinforcement learning: An introduction (1st). Cambridge:
MIT Press.
Tanner, C., Schnabel, J. A., Hill, D. L., Hawkes, D. J., Leach, M. O., & Hose, D. R. (2006).
Factors influencing the accuracy of biomechanical breast models. Medical Physics, 33(6Part1),
1758-1769.
Tarfaoui, M., & Akesbi, S. (2001). Numerical study of the mechanical behaviour of textile
structures. International Journal of Clothing Science and Technology, 13(3/4), 166-175.
Tsujisaka, T., Azuma, Y., Matsumoto, Y. I., & Morooka, H. (2004). Comfort pressure of the
top part of men's socks. Textile Research Journal, 74(7), 598-602.
Umemoto, T., Ueno, E., Matsumura, T., Yamakawa, M., Bando, H., Mitake, T., & Shiina, T.
(2014). Ex vivo and in vivo assessment of the non-linearity of elasticity properties of breast
tissues for quantitative strain elastography. Ultrasound in Medicine & Biology, 40(8), 1755-
1768.
Van Houten, E. E., Doyley, M. M., Kennedy, F. E., et al. (2003). Initial in vivo experience with
steady ‐ state subzone ‐ based MR elastography of the human breast. Journal of Magnetic
Resonance Imaging, 17(1), 72-85.
Vandeweyer, E., Hertens, D. (2002). Quantification of glands and fat in breast tissue: An
experimental determination. Annals of Anatomy - Anatomischer Anzeiger, 184(2), 181-184.
Veeger, H. E. J., van dee Helm, F. C. T., Chadwick, E. K. J., & Magermans, D. (2003). Toward
standardized procedures for recording and describing 3-D shoulder movements. Behavior
Research Methods, Instruments, & Computers, 35(3), 440-446.
Venkatesh, S. K., & Ehman, R. L. (Eds.). (2014). Magnetic resonance elastography. Springer.
Wang, L. Z., Chen, D. S., & Lin, B. (2009). Analysis of pressure distribution of brassiere’s
under wires. Journal of Fiber Bioengineering and Informatics, 2(1), 19-24.
Wang, Y. P., Zhao, P., Gao, X. S., YANG, J., & WANG, J. Z. (2006). Objective test and
evaluation on pressure comfort for brassiere. Journal of Textile Research, 27(11), 90.
White, J., Scurr, J., Smith, N. A. (2009). The effect of breast support on kinetics during
overground running performance. Ergonomics, 52(4), 492-498.
White, J., Scurr, J., & Hedger, W. (2011). A comparison of three-dimensional breast
displacement and breast comfort during overground and treadmill running. Journal of Applied
Biomechanics, 27(1), 47-53.
Wilson, L. S., Robinson, D. E., & Dadd, M. J. (2000). Elastography-the movement begins.
Physics in Medicine and Biology, 45(6), 1409.
Wong S.W., Li Y., & Zhang X. (2004). Influence of Fabric Mechanical Property on Clothing
Dynamic Pressure Distribution and Pressure Comfort on Tight-Fit Sportswear. Sen'i Gakkai
Shi 60.10, 293-299. Web.
Wu, L., Yick, K. L., Ng, S. P., Yip, J., & Kong, K. H. (2012). Parametric design and process
parameter optimization for bra cup molding via response surface methodology. Expert Systems
with Applications, 39(1), 162-171.
Wu, T., Hung, A. P. L., Hunter, P., & Mithraratne, K. (2013). Modelling facial expressions: A
framework for simulating nonlinear soft tissue deformations using embedded 3D bons. Finite
Elements in Analysis and Design, 76, 63-70.
Xi, L., Li, X., Yao, L., Grobmyer, S., & Jiang, H. (2012). Design and evaluation of a hybrid
photoacoustic tomography and diffuse optical tomography system for breast cancer detection.
Medical Physics, 39(5), 2584-2594.
Xing, M. M., Sun, Z., Pan, N., Zhong, W., & Maibach, H. I. (2006). An EFE model on skin-
sleeve interactions during arm rotation. Journal of Biomechanical Engineering, 128(6), 872-
878.
Xydeas, T., Siegmann, K., Sinkus, R., Krainick-Strobel, U., Miller, S., & Claussen, C. D.
(2005). Magnetic resonance elastography of the breast: correlation of signal intensity data with
viscoelastic properties. Investigative Radiology, 40(7), 412-420.
Yu, W. (2011). Achieving comfort in intimate apparel. In Improving comfort in clothing (pp.
427-448). Woodhead Publishing.
Yu, W., Fan, J., Ng, S., et al. (2014). Innovation and technology of women's intimate apparel:
Woodhead Publishing.
Yu, W., & Zhou, J. (2016). Sports bras and breast kinetics. In Advances in women's intimate
apparel technology (pp. 135-146). Woodhead Publishing.
Zhang, D. (Ed.). (2006). Advances in machine learning applications in software engineering.
Igi Global.
Zhang, Y., Qiu, Y., Goldgof, D. B., Sarkar, S., & Li, L. (2007). 3D finite element modeling of
nonrigid breast deformation for feature registration in-ray and MR images. In 2007 IEEE
Workshop on Applications of Computer Vision (WACV'07) (pp. 38-38). IEEE.
Zheng, R., Yu, W., & Fan, J. (2009). Pressure evaluation of 3D seamless knitted bras and
conventional wired bras. Fibers and Polymers, 10(1), 124.
Zhou, J. (2011). New methods of evaluating breast motion in braless and sports bra conditions
(Doctoral dissertation, The Hong Kong Polytechnic University).
Zhou, J., Yu, W., Ng, S. P., & Hale, J. (2009). Evaluation of shock absorbing performance of
sports bras. Journal of Fiber Bioengineering and Informatics, 2(2), 108-113.
Zhou, J., Yu, W., & Ng, S. P. (2011). Methods of studying breast motion in sports bras: a
review. Textile Research Journal, 81(12), 1234-1248.
Zhou, J., Yu, W., & Ng, S. P. (2012). Studies of three-dimensional trajectories of breast
movement for better bra design. Textile Research Journal, 82(3), 242-254.

By Reading and Using The Thesis, The Reader Understands and Agrees To The Following Terms

Uploaded by

Document Information

Original Title

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

By Reading and Using The Thesis, The Reader Understands and Agrees To The Following Terms

Uploaded by

Copyright:

Available Formats

Copyright Undertaking

This thesis is protected by copyright, with all rights reserved.

The Hong Kong Polytechnic University

INSTITUTE OF TEXTILES AND CLOTHING

OPTIMIZATION OF THE COMFORT OF

A thesis submitted in partial fulfilment of the requirements for

the degree of Doctor of Philosophy

…………LIANG Ruixin…………………. (Name of Student)

limited literature to combine numerical simulation and comfort pressure evaluation to

sports bra fabrics for optimum pressure comfort.

third is determining comfort pressure range by subjective questionnaire about comfortable

bra band do not have much influence the contact pressure.

contribute to the wellbeing of women globally.

methods for sports bra-human interactions. International Journal of Clothing

Science and Technology.

learning method to predict breast displacement during running. AATCC Journal of

Research, Under review

Conference presentation and publications

distribution of the body exerted by compression sports bra using Finite-element

Artificial Intelligence on Fashion and Textile International Conference 2019.

guidance, extraordinary patience and consistent encouragement throughout my PhD

study. I greatly appreciate for guiding me to learn a systematic research methodology

guidance, constructive suggestions, warm help, inspiring comments, as well as

her professional comments and guidance for the research methodology.

I would like to acknowledge my sincere gratitude to my co-supervisor, Dr. Newman

research conditions and environment during the last three years.

Chapter 1 Introduction ................................................................................................... 1

1.1 Research background ................................................................................... 1

1.2 Statement problem ............................................................................................ 6

1.3 Research objectives........................................................................................... 7

1.4 Project originality and significance .................................................................. 8

1.5 Research methodology ...................................................................................... 8

1.6 Outline of thesis .............................................................................................. 10

1.7 Limitation of study.......................................................................................... 12

Chapter 2 Literature Review ........................................................................................ 15

2.1 Introduction ..................................................................................................... 15

2.2 Biomechanical properties of breasts ............................................................... 15

2.2.1 Anatomy of breasts ............................................................................... 15

2.2.2 Experimental techniques to characterize breast tissues ........................ 17

2.2.3 Material properties of breasts................................................................ 20

2.3 Sports bra ........................................................................................................ 22

2.3.1 Types of sports bras............................................................................... 24

2.3.2 Effects of sports bras ............................................................................. 27

2.3.3 Pressure exerted by a sports bra ............................................................ 33

2.3.4 Pressure comfort evaluation .................................................................. 36

2.4 Finite element model of breasts ...................................................................... 37

2.5 Interactions between body and garment ......................................................... 39

2.7 Motion capture experiment ............................................................................. 45

2.7.1 Motion capture equipment .................................................................... 45

2.7.2 Studied points........................................................................................ 47

2.7.3 Reference system .................................................................................. 48

2.8 Conclusions ..................................................................................................... 50

Chapter 3 Development of finite element model of body............................................ 53

3.1 Introduction ..................................................................................................... 53

3.2 Subject-specific FE model of body obtained by 3D laser scanning ............... 54

3.2.1 Subjects ................................................................................................. 54

3.2.2 3D laser scanning .................................................................................. 57

3.3 Motion capturing............................................................................................. 59

3.4 Finite element model....................................................................................... 64

3.4.1 Geometric modelling ............................................................................ 64

3.4.2 Mesh creation ........................................................................................ 65

3.4.3 Coefficients of Mooney-Rivlin material ............................................... 68

3.4.4 Damping ratio ....................................................................................... 73