Provided by the author(s) and University of Galway in accordance with publisher policies.

Please cite the

published version when available.

A computational investigation of the laser bonding of balloon

Title catheters

Author(s) Ryan, Enda

Publication 2015-09-28
Date

Item record http://hdl.handle.net/10379/5310

Downloaded 2024-04-06T03:17:16Z

Some rights reserved. For more information, please see the item record link above.
 A Computational Investigation of the Laser Bonding of
Balloon Catheters

By

Enda Ryan

A thesis submitted to the National University of Ireland as fulfilment

of the requirements for the Degree of Doctor of Philosophy

Department of Mechanical and Biomedical Engineering

National University of Ireland, Galway

November 2014

Supervisors of Research: Dr. Mark Bruzzi and Dr. Nathan Quinlan

Industrial Partner: Boston Scientific

Abstract

The balloon catheters produced by Boston Scientific are manufactured using a laser bonding
process. The behaviour of the materials during this laser welding process is not very well
understood. In this work a computational model of this process was created which will help
predict the behaviour of the materials during bonding and will increase understanding of the
laser bonding process.

The two parts being bonded are cylindrical and are both made of a thermoplastic polymer
called PEBAX®. During the bonding process, these cylindrical parts are surrounded by a heat-
shrink tubing which applies a pressure to the PEBAX® parts when heated. The heat and
pressure cause the PEBAX® to melt and flow. Modelling the heat-shrink tubing and the melt
flow during the laser welding process are key aims of this work.

A thermal model of the process was created by modelling the laser absorption and heat
transfer through the assembly. This was compared with previous work and validated with
experimental data. The PEBAX® was modelled as a viscoelastic material which transitions
from relatively rigid to very compliant as it passes its melting temperature.

Extensive experimental testing was performed to characterise the heat-shrink tubing. The
heat-shrink tubing was modelled using two shape memory models from the literature. These
two models were implemented into Abaqus FEA using two user defined material subroutines
(UMATs).

A thermo-mechanical finite element model of the bonding process was then created in
Abaqus FEA and the results were compared with experimental data. This model captures the
overall behaviour of the materials during the bonding process, providing predictions of melt
flow. The results are compared with the experimentally observed melt flow providing new
knowledge on the thermo-mechanical behaviour during the laser bonding process of balloon
catheters.

i
TABLE OF CONTENTS

1 INTRODUCTION ....................................................................................................................... 1

1.1 BACKGROUND AND MOTIVATION .................................................................................................. 1

1.1.1 Atherosclerosis ................................................................................................................. 2
1.1.2 Angioplasty ....................................................................................................................... 3
1.1.3 Modern balloon catheters ................................................................................................ 3
1.2 LASER WELDING......................................................................................................................... 4
1.3 PROXIMAL BOND ....................................................................................................................... 5
1.4 SHAPE MEMORY POLYMERS ......................................................................................................... 7
1.5 OBJECTIVES .............................................................................................................................. 8
1.6 THESIS OVERVIEW ...................................................................................................................... 9

2 LITERATURE REVIEW .............................................................................................................. 10

2.1 LASER WELDING OF POLYMERS.................................................................................................... 10

2.2 MODELLING OF LASER WELDING OF POLYMERS .............................................................................. 11
2.3 SHAPE MEMORY POLYMERS ....................................................................................................... 14
2.3.1 Applications and interest ................................................................................................ 15
2.3.2 Materials and categorization ......................................................................................... 16
2.3.3 Characterization ............................................................................................................. 17
2.4 MODELLING OF SMPS .............................................................................................................. 17
2.4.1 Thermo-viscoelastic modelling approaches.................................................................... 18
2.4.2 Phase transition modelling approaches ......................................................................... 20
2.4.3 Applications of SMP models ........................................................................................... 23
2.5 Summary and contribution ............................................................................................. 23

3 EXPERIMENTAL METHODS ..................................................................................................... 25

3.1 DYNAMIC MECHANICAL THERMAL ANALYSIS .................................................................................. 26

3.2 HEAT-SHRINK TESTS PERFORMED ON DMTA ................................................................................. 27
3.2.1 Frequency sweep ............................................................................................................ 27
3.2.2 Temperature dependence of elastic modulus ................................................................ 28
3.2.3 Thermal expansion coefficient α .................................................................................... 29
3.2.4 Plasticity ......................................................................................................................... 29
3.2.5 Anisotropy ...................................................................................................................... 29
3.2.6 Frozen fraction................................................................................................................ 29
3.2.7 Recovery force tests ........................................................................................................ 30
3.2.8 Stress relaxation tests..................................................................................................... 30
3.3 DIFFERENTIAL SCANNING CALORIMETRY........................................................................................ 30
3.4 MICROSCOPY .......................................................................................................................... 32
3.4.1 Polarized microscopy ...................................................................................................... 32
3.4.2 Optical coherence tomography ...................................................................................... 33
3.5 SHRINK FORCE TEST .................................................................................................................. 34
3.6 PRESSURE TEST........................................................................................................................ 35

4 RESULTS OF MATERIAL CHARACTERISATION ......................................................................... 38

4.1 MEASUREMENT OF ELASTIC MODULUS OF LARGE HEAT-SHRINK TUBE.................................................. 38

4.1.1 Elastic properties under dynamic loading ...................................................................... 38
4.1.2 Effect of frequency.......................................................................................................... 39

ii
 4.1.3 Static tests ...................................................................................................................... 41
4.1.5 Axial tests ....................................................................................................................... 49
4.1.5 Anisotropy ...................................................................................................................... 49
4.2 RECOVERY FORCE ..................................................................................................................... 50
4.3 VISCOELASTIC PROPERTIES ......................................................................................................... 52
4.3.1 Stress relaxation ............................................................................................................. 52
4.3.2 Creep .............................................................................................................................. 53
4.4 THERMAL EXPANSION COEFFICIENT .............................................................................................. 54
4.5 HANGER TEST.......................................................................................................................... 55
4.6 PRESSURE TEST ........................................................................................................................ 56
4.7 DETERMINATION OF MODULUS OF SMALL HEAT-SHRINK TUBES ......................................................... 59
4.8 DETERMINATION OF THE FROZEN FRACTION .................................................................................. 60
4.9 DISCUSSION ............................................................................................................................ 60

5 THERMAL MODELLING OF THE LASER WELDING OF BALLOON CATHETERS ............................ 62

5.1 LASER THERMAL ABSORPTION ..................................................................................................... 62

5.2 HEAT TRANSFER MODEL ............................................................................................................ 64
5.3 MODELLING OF THE STERLINGTM BALLOON CATHETER JOINT ASSEMBLY............................................... 68
5.4 COMPARISON OF COMPUTATIONAL MODEL WITH EXPERIMENTAL DATA .............................................. 75
5.5 DISCUSSION ............................................................................................................................ 79

6 IMPLEMENTATION OF THERMO-MECHANICAL MODEL OF SHAPE MEMORY POLYMER ........ 81

6.1 INTRODUCTION ....................................................................................................................... 81

6.2 REVIEW OF STORED STRAIN MODEL OF LIU ET AL. ........................................................................... 81
6.3 THE MULTIPLE NATURAL CONFIGURATIONS MODEL OF BAROT ET AL. .................................................. 83
6.4 IMPLEMENTATION OF SMP MODELS............................................................................................ 85
6.4.1 Implementation of stored strain model .......................................................................... 85
6.4.2 Implementation of multiple natural configurations model ............................................ 86
6.5 VERIFICATION ......................................................................................................................... 88
6.5.1 Verification of implementation of stored strain model .................................................. 88
6.5.2 Verification of implementation of multiple natural configurations model ..................... 89
6.6 COMPARISON OF UMATS WITH EXPERIMENT................................................................................ 94
6.6.1 Stored strain model ........................................................................................................ 95
6.6.2 Multiple natural configurations model ........................................................................... 96
6.7 MATERIAL MODEL FOR PEBAX® ................................................................................................ 98
6.7.1 Prony series .................................................................................................................... 98
6.7.2 Time-temperature superposition .................................................................................... 99
6.7.3 PEBAX® data ................................................................................................................. 101
6.8 DISCUSSION .......................................................................................................................... 106

7 THERMO-MECHANICAL MODELLING OF BALLOON CATHETER LASER WELDING PROCESS ... 108

7.1 INTRODUCTION ..................................................................................................................... 108

7.2 THERMO-MECHANICAL MODEL DEVELOPMENT ............................................................................ 109
7.2.1 Laser weld assembly geometry .................................................................................... 109
7.2.2 Mesh ............................................................................................................................. 111
7.2.3 Boundary conditions ..................................................................................................... 111
7.2.4 Thermomechanical loading .......................................................................................... 112
7.2.5 Contact ......................................................................................................................... 112
7.2.6 Materials ...................................................................................................................... 113
7.3 RESULTS............................................................................................................................... 113

iii
 7.3.1 Melt flow results ........................................................................................................... 113
7.3.2 Effect of friction ............................................................................................................ 118
7.3.3 Effect of initial geometry .............................................................................................. 121
7.3.4 Effect of heat-shrink modulus....................................................................................... 125
7.4 EXPERIMENTAL RESULTS .......................................................................................................... 125
7.5 DISCUSSION .......................................................................................................................... 130

CHAPTER 8 CONCLUDING REMARKS ...................................................................................... 133

8.1 CONCLUSIONS ....................................................................................................................... 133

8.2 LIMITATIONS OF THIS WORK ..................................................................................................... 135
8.3 DISCUSSION .......................................................................................................................... 136
8.4 FUTURE WORK ...................................................................................................................... 137

REFERENCES ................................................................................................................................. 139

iv
1 Introduction

The aim of the work in this thesis is to create a computational model of a laser bonding
process used in the manufacture of balloon catheters. The laser bonding process is used to
weld two thermoplastic polymer parts together. During the process, the laser heats the
polymer parts above their melting temperature and a heat-shrink tubing, which is placed
over the parts, applies a pressure which causes the polymers to flow. The heat-shrink tubing
is a type of shape memory polymer (SMP) and modelling this material accurately is a major
goal of this work. The model is used to predict the transient temperature distribution as well
as the stresses and strains which occur during the bonding procedure. The model uses the
commercial finite element analysis package Abaqus FEA.

This work:
1. Provides a better understanding of the effect of different material and experimental
parameters (such as heat transfer coefficients, laser power, etc.) on the temperature
distribution throughout the assembly.
2. Provides a better understanding of the effects of different material and experimental
parameters (such as modulus, viscosity, maximum temperature etc.) on the
deformation that occurs in the bond area.
3. Identify the most important parameters with regard to bond shape and strength.

This project will focus on one bond called the Proximal bond on one particular product (the
SterlingTM catheter). However, it is envisaged that this model can be easily adapted to other
bonds and other products.

1.1 Background and motivation

A catheter is a flexible tube that can be inserted into a body cavity or blood vessel. Catheters
are regularly used to drain or administer fluids or for the delivery of surgical instruments to
a specific location. In 1963 Thomas Fogarty invented a balloon catheter for removing blood
clots. The deflated balloon was inserted past the clot then inflated and pulled back dragging
the clot out of the body through the incision made for the catheter. Prior to this the only
treatment for clots was surgery which was a very risky procedure with up to 50% of patients

1
 CHAPTER 1

dying [1]. Fogarty’s catheter was the first successful example of “less invasive” surgery and
quickly became the standard way to remove clots.

Less invasive, and minimally invasive, surgery are terms used to describe procedures
performed using only a small incision or no incision at all. These procedures often have many
benefits such as less operative trauma, less pain and a quicker recovery [2]. These methods
are now being used to treat a range of diseases. One very common disease that is now
regularly treated using minimally invasive methods is atherosclerosis.

1.1.1 Atherosclerosis

Atherosclerosis is the build-up of fatty materials such as cholesterol on the artery walls. It is
a very common and very serious condition which can affect arteries throughout the body.
The build-up of the fatty material, called plaques, causes the artery wall to thicken and
become stiffer. This reduces blood flow through the artery which can cause various
problems. The plaques can build so much that they eventually block the artery completely
or pieces of plaque can break off and travel to smaller blood vessels causing blockages.

It is a chronic disease that can exist for decades without any symptoms and leads to heart
disease which is the leading cause of death in the western world [3]. The main cause of
atherosclerosis not yet known but it is thought to be due to inflammatory processes caused
by the body’s immune system responding to damage of the artery wall.

Figure 1.1 Schematic diagram showing build up of plaque in artery. Reproduced from [4].

2
 CHAPTER 1

1.1.2 Angioplasty

When a blood vessel has narrowed due to a

build-up of plaque a procedure called an
angioplasty can be used to mechanically widen
the obstructed blood vessel. The device used for
this procedure consists of a balloon on a guide
wire and is called a balloon catheter. The
deflated balloon is inserted into the artery and
positioned at the location of the blockage. The
balloon is then inflated and it compresses the
plaque and widens the blood vessel allowing
blood flow to return to normal.

The first angioplasty was performed by Charles

Dotter in the U.S. in 1964 [5]. However the
procedure did not become popular immediately
in the U.S. due in part to difficulties in
reproducing the techniques and the occurrence
of complications. In 1975 Andreas Gruentzig Figure 1.2 Angioplasty procedure. Reproduced from [6]
used Dotter techniques to develop a balloon catheter and in 1977 conduct the first coronary
angioplasty on an awake human.

Prior to this, atherosclerosis had only been treatable by invasive surgery. Angioplasty is now
one of the most commonly performed surgical procedures and is an excellent example of the
success and advantages of minimally invasive surgical techniques.

1.1.3 Modern balloon catheters

There is now a large market [7] for balloon catheters with a wide range of different designs
available for different uses. They are typically mass-produced from thermoplastics and are
inserted into the body along a guide-wire.

The diagram below shows the catheter studied in this work. It is a rapid exchange balloon
catheter called the Sterling™ balloon catheter and it is produced by Boston Scientific. This
catheter is made of a thermoplastic elastomer called PEBAX® 7033. PEBAX® is a registered
trade name for a group of polyether block amides produced by Arkema. The properties that

3
 CHAPTER 1

make PEBAX® a good choice for catheters are: its flexibility; its chemical resistance; its
biocompatibility; it is easily sterilized; its compliance with UPS class VI; its ability to be
extruded into very thin wall tubes; it is kink resistant; its smooth surface finish and it can be
compounded with radiopaque fillers [96][97]. The different parts for the catheters (such as
the balloon, the tip, the outer tubing) are produced separately and then they are welded
together with lasers. Figure 1.3 shows a schematic drawing of the balloon catheter
construction and a photograph (enlarged) of the proximal bond which will be the focus of
this work.

Figure 1.3 Sterling™ balloon catheter currently produced by Boston Scientific. [8].

1.2 Laser welding

There are many different techniques for bonding plastics but Boston Scientific has chosen to
use lasers as they provide a precise and easily controllable heat source. A common method
of laser welding is to use a laser which can travel through the first layer without being
absorbed (i.e. the material is transparent to this wavelength) but then gets absorbed at the
interface between the materials. It is absorbed either because the second layer is opaque or
an opaque dye is placed between the layers. This method is called Laser Transmission
Welding (LTW) and it produces a very localized heat field. However, the welding process used
by Boston Scientific employs heat and pressure and therefore the function of the laser is to
heat both the heat-shrink tubing and the PEBAX® layers. To do this Boston Scientific uses CO2
lasers which produce an infra-red beam with a wavelength of 10.6 µm and radiation at this

4
 CHAPTER 1

wavelength is absorbed readily by both the heat-shrink and the PEBAX®. Boston Scientific
have measured the absorption coefficients of the heat-shrink and the PEBAX® by measuring
the intensity of a CO2 laser after it had passed through PEBAX® and heat-shrink layers of
different thicknesses [94]. The absorption coefficient of the PEBAX® was measured as 18898
m-1, which means that the intensity of the light passing through the PEBAX® halves every 37
µm. The absorption coefficient of the heat-shrink was measured as 3543 m-1, which means
that the intensity of the light passing through the heat-shrink halves every 196 µm.

To achieve a circumferential joint, the joint assembly is rotated at high speed during the laser
welding process. This method ensures that the heat-shrink is heated thoroughly and that
enough PEBAX® is melted to produce the desired melt flow. The interaction of the laser beam
with the assembly produces a spatially-varying volumetric heat source. Modelling this heat
source is the first step in creating a thermal model of the bonding process.

1.3 Proximal bond

Figure 1.4 below shows a cross-sectional sketch of a balloon catheter at the point where the
balloon is joined to the main shaft of the catheter, called the outer tube. This area is called
the proximal bond as it closer to the surgeon during use. The bond at the other end of the
balloon is known as the distal bond. The work described in this thesis focuses on the proximal
bond because there is more data available on the proximal band and because this bond is
axisymmetric. The fact that this bond is axisymmetric makes it less computationally
expensive than modelling other bonds such as the port bond which would require a full 3D
model.

To form the proximal bond the balloon is placed over the outer tube. Heat-shrink is then
placed over the balloon. The purpose of the heat-shrink is to apply a pressure to the assembly
to help the balloon bond to the outer tube. An infra-red (IR) laser is used to heat up the
assembly. As the assembly heats up the heat-shrink tubing shrinks and applies a pressure to
the two layers of PEBAX® (the balloon and the outer tube). The main purpose of the heating
is to cause intermolecular diffusion from one part to the other. This mixing process is what
forms the bond when cooled.

The heating also causes the PEBAX® to melt and flow under the pressure from the heat-
shrink. This flow produces a smooth tapered surface, as shown in Figure 1.5, which aids in
the insertion and extraction of the tube. A model which could predict this melt flow would

5
 CHAPTER 1

be of great use to Boston Scientific and this is one of the major goals of this project. To model
the melt flow, it is first necessary to accurately model the heat-shrink tubing.

Heat-shrink
Balloon Laser light

Bond Area
Outer tube

Mandrel

Figure 1.4 Setup for welding of proximal bond. The assembly is rotated at 500 RPM during the laser welding
process.

Bond Area

Outer tube
Balloon

Figure 1.5. Tapered proximal bond after welding

6
 CHAPTER 1

1.4 Shape memory polymers

The shape-memory effect in polymers describes the ability to store a deformed shape
indefinitely and recover entirely the undeformed shape in response to specific
environmental stimulus [67]. The heat-shrink tubing used by Boston Scientific in the laser
bonding process is a type of shape-memory polymer (SMP). Heat-shrink tubing can be made
from a range of thermoplastics and the heat-shrink used by Boston Scientific is a polyolefin
tubing made by Raychem called RNF-100®. The thermomechanical cycle of a SMP is shown
schematically below in Figure 1.6. The shape of the polymer after forming (e.g. extrusion or
moulding) is called the original or permanent shape. In Figure 1.6 this is represented by the
shape in the top-centre. Going clockwise from the top-centre the material is heated above
its glass transition temperature, Tg, so that it is in a rubbery state, then it can be elastically
deformed. This causes a stress in the material but when cooled below Tg the material
becomes glassy and the bonds formed during this transition prevent the material recovering
to its original shape. The polymer will then remain in this deformed state (bottom-centre of
Figure 1.6) until it is heated above Tg. This heating breaks down the glassy bonds and causes
the polymer to re-enter the rubbery state. This releases the elastic energy stored in the
polymer which causes the material to revert back to its original shape.

SMPs have been used for years as heat-shrink tubing, packaging, etc. but SMPs are now being
developed for uses in medical devices, sensors and actuators. A review of the applications of
SMPs is given by Behl and Lendlein [9]. When SMPs only had very simple uses (heat-shrink
tubing for installing electrical wires, etc.) there was very little work done in understanding
their behaviour. But as they have started being developed for advanced applications, a
comprehensive understanding of their behaviour is now of great interest.

7
 CHAPTER 1

Figure 1.6 Shape Memory Polymer heat deformation cycle. [10]

Shape Memory Polymers can be activated at Tg, as mentioned above, but the melting point
of semi-crystalline networks can also be used to trigger shape recovery. This type of SMP is
often referred to as a crystallisable SMP (CSMP). Heat-shrink tubing falls into this category
of polymer. In CSMPs the polymer is semi-crystalline and partially cross-linked.

1.5 Objectives

The overall objective of this work is to develop an understanding of the thermo-mechanical

behaviour of the laser welding process of balloon catheters during laser welding assembly.
A computational finite element model is developed in this work to contribute to this
understanding. This will involve modelling the temperatures, stresses and melt flow during
the welding process. The overall objective has been divided into four main parts.

The first objective of this project is to create a thermal model which can accurately predict
the temperatures in the assembly during the laser welding process.

The second objective of this project is to create a model for the heat-shrink tubing. This
material will be modelled using a SMP model from the literature. To implement this model
in the commercial finite element code, Abaqus FEA, a user defined material model (UMAT)
is developed through programming a user subroutine. The material model is then used to

8
 CHAPTER 1

simulate the processing of the shape memory material to represent the stored elastic energy
as illustrated schematically in Figure 1.6.

The third objective of this work is to model the melt flow of the bonded region. This will
consist of combining the thermal model and heat-shrink model with a model that represents
the PEBAX® as it is heated to above its melt temperature.

The fourth objective is to use the model to analyse and achieve a better understanding of
the welding process. This improved understanding will help in the future design of catheters
and will potentially speed up the development process and reduce trial and error
experimentation when developing new products.

1.6 Thesis overview

In this thesis, experimental and computational methods are used to address the objectives
outlined above. Chapter 2 gives a detailed overview of the background literatures relevant
to this work and provides a context of previous work performed in the field and also of
methods developed to characterise and represent shape memory polymer materials.
Chapter 3 gives an overview of all the experimental testing methods used to characterise
and test the thermo-mechanical behaviour of the materials used in this work. The results of
these experimental characterisation tests are given in Chapter 4. These experimental results
provide inputs into the development and calibration of the shape memory polymer model
as well as providing data for validation of its performance.

Chapter 5 describes the investigation into the thermal modelling of the laser welding
process. The temperature profiles are examined and compared to those determined from
experimental tests. Chapter 6 describes the development of a user defined material
behaviour to represent the behaviour of the shape memory polymer, and also describes its
performance in accurately representing the thermo-mechanical behaviour of this two-phase
material. Chapter 7 goes on to describe the utilisation of this shape memory polymer UMAT
in thermo-mechanically modelling the welding process of the balloon catheter assembly and
the PEBAX® material as it transitions through its melting temperature. A discussion on the
overall thesis is given in Chapter 8 where conclusions are made and future work is described.

9
2 Literature Review

2.1 Laser welding of polymers

Laser beam welding of polymers was considered in the early years of laser technology [11]
but at the time there were less expensive solutions to the problem of polymer welding [12]
such as hot gas welding and friction welding. However, as polymers have found more and
more applications, welding demands arose which could not be easily met with conventional
technologies. Also the cost of laser systems has decreased rapidly in recent years while the
quality has improved substantially, leading to laser welding being used in many industries
such as automotive, electrical and medical devices [13].

During the laser welding of polymers the parts are clamped together and then heated with
the laser. The clamping holds the parts in place and is necessary to maintain a pressure at
the weld seam, which is important for a good bond. The heat generated by the laser
absorption causes the polymer parts to melt. This results in molecular diffusion between the
two parts and a solid joint forms as the polymer cools and solidifies. The two main forms of
laser welding of polymers are laser transmission welding and CO2 laser welding.

Laser Transmission Welding (LTW) is the most widely used laser welding technique and also
the most discussed in the literature. This technique usually uses Nd:YAG or diode lasers
which produce light in the near-infrared (NIR) part of the spectrum which can pass through
many polymers. In LTW one part must be transparent to the laser light while the other part
absorbs it. The parts are arranged so that the laser travels through the transparent part and
is absorbed at the boundary between the transparent part and the absorbent part. This heats
up the absorbent part which then heats the transparent part by conduction across the
boundary. This technique produces a weld with a small heat affected zone because it delivers
the heat directly to the joining area. This process was first described for welding automotive
components in [14]. To make the second part absorb the radiation, it is often necessary to
add an appropriate pigment [15, 16].

The major advantages of LTW include:

 It produces high quality welds.

 The rapid heating results in a short weld time.

10
 CHAPTER 2

 The energy input is very controlled and localised.

There are still challenges to overcome with LTW such as extending the range of colours of
weldable parts. This can be overcome by placing a certain absorbing material between the
(transparent) parts but this is not always acceptable due to polymer compatibility, toxicology
or cost efficiency.

LTW has been considered by Boston Scientific but they opted to use a different welding
technique instead. The welding technique used by Boston Scientific uses a CO2 laser. CO2
lasers produce light with a wavelength of 10.6 µm. Light at this wavelength is readily
absorbed by most plastics and this leads to rapid heating from the surface through the parts.
This rapid heating allows for very rapid processing of thin polymer films. Coelho et al. [17]
demonstrated welding of polymer film at speeds of up to 10 ms-1. Due to the opacity of most
plastics at 10.6 µm and the low thermal conductivities of plastics, CO2 lasers are only used
for welding thin plastic films. In thick samples there would be excessive heating at the surface
and not enough at the joint area. This is the main reason why CO2 lasers are not used as
extensively as diode (NIR) lasers. The main advantages of CO2 welding are that it is fast,
simple and can be used on any polymer.

2.2 Modelling of laser welding of polymers

Currently the majority of laser welding applications rely on experimentation to determine

process parameters. An accurate model would enable optimization of process parameters
without the need for extensive trial and error experimentation.

The modelling process needs to incorporate various factors such as laser distribution and
absorption, heat transfer temperature, pressure and many material properties that vary with
temperature. Models can be analytical, such as that described in [18], or more commonly
numerical finite element models.

The finite element method (FEM) is a useful tool for material processing simulation. There
has been a large amount of work done in applying FEM to the modelling of laser welding of
metals and a detailed literature review in this area was carried out by Mackwood and Crafer
[19]. However, there are a relatively small number of researchers who have used FEM to
model the laser welding of plastics. The main difference between welding plastics and metals
is that in metals the laser is absorbed within the first few nanometres whereas in plastics the

11
 CHAPTER 2

laser is absorbed over a scale of a few millimetres (for CO2). Also metals have much higher
thermal conductivities and melting points.

Mayboudi et al. [20] and Frick [21] created and tested an FEM model to predict the
temperature profile in laser transmission welding. When modelling the temperature profile
in a laser welded part, the first thing to consider is the laser absorption. Light absorption
within a solid can be described by the Beer-Lambert law:

𝐼(𝑧) = 𝐼𝑜 𝑒 −𝛼𝑧 (2.1)

where I(z) is the intensity of the beam as a function of depth into the material, I0 is the
intensity at z=0 and α is the absorption coefficient. This equation is appropriate in the case
of absorbent polymers and is widely used [18, 23 and 24]. However for polymers with a low
absorption coefficient, the scattering of the laser beam becomes an important issue as
demonstrated experimentally by Potente et al. [25] and Becker et al. [24]. Potente et al. [25]
defined a correction factor which took the scattering into account and supported their
results by comparing the experimentally observed melt layer thickness with the model
predictions. These materials scatter light because they are inhomogeneous. This can be due
to their semi-crystalline structure or the presence of impurities or additives. Ilie [26] has
modelled this scattering using the Mie theory and the Monte Carlo method. This numerical
model allows a good estimation of the laser beam profile after it has passed through the
transparent layer in LTW.

Most polymers have a high absorption coefficient for CO2 laser radiation and therefore the
Beer-Lambert Law can be applied. However, Coelho et al. [27] have found that polyethylene
samples thinner than 200 µm have high transparency since the attenuation length is larger
than the sample thickness. They proposed methodologies for determining the optical
properties (at 10.6 µm), which involves using an integrating sphere to determine reflectance
and transmittance and using these values to determine the complex refractive index. Geiger
et al., [28] also used integrating spheres to determine the optical properties of polymers for
laser welding and use these properties in an FE simulation of the temperature fields during
LTW. One conclusion from this work on polypropylene was that the change of absorption
coefficient with temperature had a negligible effect on influencing the melt pool geometry.
However, Van de Ven [23] stresses the importance of measuring α as a function of
temperature for the absorbing PVC part in LTW.

12
 CHAPTER 2

The heat transfer resulting from the laser absorption has also been investigated. Heat
transfer in a material can be modelled using the general heat conductivity equation [29]. In
three-dimensional Cartesian space this can be written as follows:

𝜕 𝜕𝑇 𝜕 𝜕𝑇 𝜕 𝜕𝑇 𝜕𝑇
𝜕𝑥
(𝑘𝑥 (𝑇) 𝜕𝑥) + 𝜕𝑦 (𝑘𝑦 (𝑇) 𝜕𝑦) + 𝜕𝑧 (𝑘𝑧 (𝑇) 𝜕𝑧 ) + 𝑄(𝑥, 𝑦, 𝑧, 𝑡) = 𝜌(𝑇)𝐶(𝑇) 𝜕𝑡 (2.2)

where ρ is the density, C is the specific heat capacity, T is the temperature, t is time, k is
thermal conductivity and Q(x,y,z,t) is the heat generation rate. In the case of laser welding
the heat generation rate is equal to the light absorption given by the Beer-Lambert law in
Equation 2.1 above.

This heat conduction equation can be solved numerically [20-22] or analytically [30] and
many researchers now report good agreement between the temperature fields predicted by
the models and those observed by experiment.

One interesting heat transfer result came from Coelho et al., [30]. They found that 30% more
energy needed to be applied in practice compared to the model. They concluded that
gaseous by-products released during the initial moments of processing were absorbing the
laser radiation just above the plastics surface. This led to the formation of a plasma on the
surface of the materials which was directly observed experimentally.

In addition to the temperature field during welding, the pressure is also of interest. External
pressure is applied to the parts to improve the weld quality. Also thermal expansion of the
two materials during the welding process can create pressures and tractions. Van de Ven and
Potente [22, 23] found that the pressure was not usually a significant factor in LTW as long
as excessive limits were not exceeded. Therefore, for many researchers the pressure has not
been a crucial factor. However, the CO2 welding process used in Boston Scientific is different
from most of the LTW processes because the surface material melts and flows due to the
heat and the pressure applied. Therefore, pressure is more likely to be an important factor.

A study of the residual stresses in polymers after LTW was conducted by Potente et al., 2008
[33]. In this work, material equations were developed to describe the elasto-viscoplasticity
at different temperatures and were included in a model developed using the Abaqus FEA
software. The main finding of this work was that the model could predict the influence of the
process parameters on the development of the residual stresses in the LTW parts.

An earlier paper by Potente and Fiegler [34] presents an Abaqus FEM model, along with
experimental data, of melt displacement and temperature profiles for LTW. An interesting

13
 CHAPTER 2

point to note from this work was the use of tracer particles to monitor the melt displacement
during the laser welding process. These tracer particles were Tungsten spheres (dia. 75-70
µm) and were detected using micro-focus radioscopy. This allowed the chronological and
spatial displacements of these particles to be recorded throughout the welding process. This
data was then compared to the calculated flow profile and the model was found to be in
good agreement with experiment.

Often the main objective of thermal modelling is to get a temperature profile at the end of
the heating phase, i.e. just as the laser is turned off (as this is when the temperatures are at
a maximum) and therefore there is no need to run the simulation after the laser is turned
off. However, an important point to note from literature [34] regarding the
thermomechanical modelling is that the cooling phase should also be modelled as stresses
and strains can change during this period.

The welding process in Boston Scientific uses heat-shrink tubing to provide an external
pressure on the assembly to aid welding. Heat-shrink tubing has not been characterised
extensively in the literature, probably because it is normally used for simple applications such
as installing electrical wires. Also, there are no studies in the literature where the thermo-
mechanical behaviour of heat-shrink tubing has been modelled, perhaps due to the
commercial sensitivities around revealing processing parameters. However, there is a
substantial amount of literature on the related area of shape memory polymers (SMPs). This
is discussed in the following section.

2.3 Shape memory polymers

As mentioned previously the heat-shrink tubing used by Boston Scientific is a type of shape
memory polymer (SMP) and modelling this is a major goal of this work. A large amount of
research has been published on characterizing and modelling SMPs. Various reviews exist
already [35-39] and they range from the broad overview by Mather et al. [36] to more
narrowly focused reviews such as the review of cardiovascular applications by Yakacki et al.
[35] and the review of multifunctional SMPs by Behl et al. [37].
Figure 2.1 is a graph of the stress and strain of a sample during an SMP cycle. In step 1 the
sample is stretched at a high temperature (50°C) to a strain of εm and is held at this strain for
some time. Stress-relaxation occurs and causes the stress to reduce while the strain and
temperature remain constant. The temperature is then reduced while the strain is kept

14
 CHAPTER 2

constant (step 2) and this fixes the sample in its temporary shape. When the sample is
released (step 3) the stress goes to zero. It remains in this temporary shape until it is heated
(step 4) which causes the sample to return to its original shape.

Figure 2.1 The stress-strain-temperature response of an SMP during a typical shape-memory cycle. The sample
is stretched (step 1), cooled (step 2), released (step 3) which forms its temporary shape and when it is reheated
(step 4) it returns to its original shape [9].

2.3.1 Applications and interest

Shape memory alloys (SMA) are currently in use in many industries. In particular, a nickel-
titanium alloy called Nitinol is used in many commercial applications. These materials have
many merits but also have some limitations which SMPs may be able to overcome.

One of the main advantages of SMPs over SMAs is that SMPs can have very large recoverable
strains of up to 400% [40] whereas SMAs maximum strain recovery is about 8% [41]. Also
SMPs are inexpensive to manufacture. They are malleable, damage-tolerant and the
recovery temperature can be adjusted by changing the chemical structure and composition
[35]. These properties mean that SMPs have potential applications in a wide variety of
industries.

15
 CHAPTER 2

One area where SMPs could have a large impact is in biomedical applications. Various
authors have investigated SMP stents [35, 42, 43]. Lendlein and Langer [44] created a
biodegradable SMP suture. Small et al. [45] describe a SMP device for the removal of blood
clots. Sokolowski et al. [46] discuss using SMPs in cardiac pacemakers, vascular grafts and
artificial hearts.

Sokolowski et al. [47] discusses SMP-based deployable structures for space applications.
Other uses include smart fabrics [48], self-disassembling mobile phones [49] and actuators
[50, 51]. A comprehensive list of SMP applications has been produced by Meng et al. [52]
along with a review of recent attempts to strengthen SMPs and Rousseau [53] has reviewed
progress toward overcoming the limitations of SMPs.

2.3.2 Materials and categorization

Liu et al. [54] classify SMPs in 4 categories based on differences on fixing mechanism and the
origin of permanent shape elasticity.

Covalently cross-linked glassy thermosets network

These SMPs are activated at their glass-transition temperature (Tg). Above this temperature
the material has a rubbery elasticity which provides for a high degree of shape recovery.

Covalently cross-linked semi-crystalline networks

This class of SMP uses the melting temperature of the semi-crystalline network to trigger
shape recovery. Heat-shrink is in this group.

Physically cross-linked glassy co-polymers

These SMPs have the advantage that their permanent shape can be altered as opposed to
covalently cross-linked SMPs whose permanent shape is fixed after the polymer has been
moulded or extruded.

Physically cross-linked semi-crystalline block co-polymers

These SMPs are activated at the melting temperature and can also have their permanent
shape reprocessed by thermal processing above 100°C.

16
 CHAPTER 2

2.3.3 Characterization

There has been a large amount of work done on the characterization of SMPs. Early work on
the properties of SMPs was performed by Liang et al. [56] and Hayashi et al. [57]. Tobushi et
al. [58] examined the thermomechanical properties of shape memory polyurethane films,
including modulus, yield stress, shape fixity and shape recovery. Lendlein et al. [41] discuss
the properties of various different SMPs. In particular they discuss how the molecular
structure of the polymer affects the strain recovery ratio and the mechanical properties. Gall
et al. [59] examined the storage and release of internal stress in a SMP cycle. Beblo et al. [60]
discuss anisotropic properties of SMPs which occur due to the application of a large strain.
Diani et al. [61] have recently described a new torsion test to study large deformation
recovery in SMPs. However, there is very little data available for the SMP used in this work
(RNF-100® heat-shrink tubing) so it was necessary to characterise the thermomechanical
properties in detail so that the material could be modelled accurately.

The most common methods for characterization of the mechanical properties involve a
dynamic mechanical thermal analyser (DMTA) [58, 62, 63, 64, 65, 66]. With the DMTA the
temperature-dependent properties of polymers can be assessed quickly and many different
types of tests can be performed with it. The DMTA can measure properties including, the
complex modulus as a function of temperature, recovery behaviour, viscoelasticity and yield
stress. Due to its prevalence in the literature and its versatility, DMA was chosen as the main
method for the characterization of the heat-shrink tubing.

2.4 Modelling of SMPs

Most work on SMPs has concentrated on experimental characterization especially in the

early years of SMP research. However, as SMPs are being used in increasingly complex and
ambitious ways, there is a need to be able to predict their behaviour accurately. Many
models have been published in the last few years with increasing complexity and accuracy.
Nguyen [67] has recently performed a comprehensive review of these models.

There are two main approaches to modelling SMPs 1) Models which are based on standard
viscoelasticity with various rheological elements 2) Models which are based on the existence
of two separate phases.

17
 CHAPTER 2

2.4.1 Thermo-viscoelastic modelling approaches

Many authors have tried to model SMPs using combinations of parallel and series
arrangements of rheological elements which have temperature-dependent constitutive
relations for the viscosities and/or moduli to affect strain storage and recovery. Tobushi et
al. [69] proposed a linear constitutive model but this was not valid for large strains. In order
to express behaviour at large strains this model was modified [70] to account for nonlinear
behaviour and the proposed theory is useful for determining some basic characteristics of
SMPs such as the amount of shape recovery and start and completion temperatures.

Morshedian et al. [71] also used temperature-dependent rheological elements (springs,

dashpots) to create a mechanical model of SMPs. This model was developed to model the
stress-strain-time behaviour of heat-shrinkable polyethylene. The proposed mechanical
model is shown below in Figure 2.2A. A focus of this work was stress relaxation which can
prevent the full dimensional recovery of heat-shrinkable materials. It was found that this
stress relaxation occurs during the strain holding/cooling time before the formation of
crystallites at the solidification point. If the elapsed time before crystallization is long enough
(in this case 30 h) then the elastic restoring force completely releases, resulting in zero
shrinkage. The mechanical model was capable of describing the experimental results
observed in shape memory induction and dissipation very well after some model parameters
were determined by experiment.

Nguyen et al. [72] developed a finite deformation model for amorphous SMPs which
incorporated a time-dependent model of the glass transition and a structural relaxation
model. This model is shown in Figure 2.2B. The model was implemented in an FE program
and it reproduced many of the important features of the stress and strain recovery response,
such as the hysteresis in the stress-temperature curve and the peak stress during reheating.
However some quantitative differences remain such as the unconstrained recovery response
and the strain recovery time.

Ghosh et al. [73] propose a two-network model which is based on the Helmholtz potential
(Figure 2.2C). This work highlights the sensitive dependence of the response on thermal
expansion and the importance of modelling the hysteresis in the yield stress during
temperature change. This model was developed for small-strain analysis and was
implemented in MATLAB. The model results agree well with experiment.

Diagrams of four SMP models based on rheological elements are shown below in Figure 2.2.

18
 CHAPTER 2

A B

C D

Figure 2.2 Four SMP models based on rheological elements A) Morshedian et al. [71], B) Nguyen et al. [72], C)
Ghosh et al. [73] and D) Westbrook et al. [74].

Westbrook et al. [74] builds on previous work by Castro et al. [75] and Nguyen et al. [72] to
create a new multi-branch model to capture the SM effect and various relaxation processes
of amorphous SMPs. This model is shown in Figure 2.2D and the results of this model are
compared with experimental results and are shown below in Figure 2.3. In this experiment a
cylindrical sample was heated to 60°C, then subjected to a compressive strain of 20% and
cooled to lock-in the deformation (cooling shown in Figure 2.3) . The sample was then
reheated (heating shown in Figure 2.3) while the strain was kept at 20% in a constrained
recovery experiment. As the sample is reheated the thermal expansion of the material
causes a large increase in the compressive stress. As the sample is heated further the
material becomes more compliant and the stress reduces.

19
 CHAPTER 2

Figure 2.3 Comparison between numerical simulation and experimental results for the stress response during
a constrained recovery shape memory cycle [74].

This model was implemented in 2D in a user subroutine (UMAT) for the FE package Abaqus
FEA. The model involves a significant number of parameters (exact number depends on
number of branches used) and a protocol is provided to identify these parameters. The
model simulations are compared with a range of thermomechanical experiments and show
very good agreement.

2.4.2 Phase transition modelling approaches

Many researchers take an alternative approach to modelling SMPs and use the concept of
two separate phases. This method was first introduced by Liu et al. [64]. Their model is based
on the thermodynamic concepts of entropy and internal energy. They adopt the concept of
“stored” strain into a simplified thermomechanical model. In this model the SMP material is
modelled as a mixture of two kinds of extreme phases, the “frozen phase” and the “active
phase”. In the frozen phase the material is relatively rigid while in the active phase the
material is easily deformed. An SMP cycle of this model is shown in Figure 2.4. This model
applies to SMPs which change shape at the Tg of the polymer. The volume fraction of each
phase changes with temperature. In the glassy state, below Tg, the frozen phase is the major
phase but as the material is heated the frozen fraction reduces and the active fraction
increases. Above Tg, when the active fraction has completely replaced the frozen fraction,
the polymer is in the rubbery state. The 1-D constitutive equation (excluding thermal
expansion) for this model is

20
 CHAPTER 2

𝜎 = 𝐸(𝜀 − 𝜀𝑆 ) (2.3)

where εs is the stored strain. This model uses a rule of mixtures to derive an equation for the
Young’s modulus of the polymer as a function of temperature.

1
E
F 1  F (2.4)

EF EA

Where EF and EA are the moduli of the frozen phase and active phase respectively and are
assumed to be constant and ϕf is the frozen fraction which can be calculated from:

1
F  1  (2.5)
1  c f Th  T 
n

In this equation, cf and n are variables which can be determined by experiment as

described in [64] and Th is an arbitrary temperature above Tg where the polymer is in the
full rubbery phase.

Initial State Deformed state

Heated
above Tm Deformed

Frozen State Active State Strain = εpre

E = EF E = EA σ = EA εpre

Fixed and EF ≥ E ≥ EA
cooled Fully cooled

Strain becomes stored in frozen phase as T σ = EF (εpre- εs)

decreases. εs increases. Frozen phase does not but εs ≈ εpre therefore σ ≈ 0
contribute to stress
Heated Full
Released
Recovery

When constraint is released As T increases stored strain Initial shape is

the material holds its shape decreases. σ = E (ε - εs) recovered
Therefore σ increases.
Figure 2.4 The stored strain model of Liu et al. [64] of a thermomechanical cycle of an SMP.

21
 CHAPTER 2

This model is demonstrated only for small strains. Chen et al. [76] continued on this idea to
develop a three-dimensional constitutive theory for large deformations.

The general concept of considering two phases within the polymer with different mechanical
properties is very common as evident from Nguyen’s recent review [67] of SMP models.
Barot et al. [77] use a multiple natural configurations approach along with the framework
that was developed recently for studying crystallization in polymers [78] to model the
behaviour of CSMPs. They formulate constitutive equations to describe the interaction
between an amorphous phase and a semi-crystalline phase. The modelling is divided into
four parts, namely, the rubbery phase (T>Tm), the semi-crystalline phase (T<Tm), the
crystallization process (on cooling through Tm) and the melting process (on heating through
Tm). The hyperelastic behaviour of different materials can be modelled within the Barot et
al. framework and it is demonstrated with the Neo-Hookean model. The equation for the
stress in a Neo-Hookean material is

𝜇
𝛔= 𝑑𝑒𝑣(𝐁∗ ) + 𝜅(𝐽 − 1)𝚰 (2.6)
𝐽

Where μ is the shear modulus, κ is the bulk modulus, J is the determinant of the deformation
gradient F, and B is the left Cauchy-Green deformation tensor and 𝚰 is the identity tensor.
The model was used to simulate a typical uniaxial cycle of deformation and the results
compare very well with the experimental data.

Kafka [79], following on from work on shape memory alloys [80], proposes a model which
assumes there are two continuous three-dimensional substructures in SMPs. One
substructure remains elastic with a constant Young’s modulus. The other substructure is
assumed to deform in an elastic-plastic-viscous way with Young’s modulus decreasing with
increasing temperature. This is a very general model and simulated results agree well with
experimental data but at the moment it is only valid for small strains.

Qi et al. [81] develop a 3D finite deformation model based on the two phases approach. The
hyperelastic behaviour of the rubbery phase is modelled using the Arruda-Boyce eight-chain
model which captures the material behaviour up to large strains. This model was
implemented in a UMAT and compared with uniaxial experiments. The model captured the
isothermal stress strain behaviour and the shape memory effects during heating and cooling.
However it did not capture the stress recovery implying the simple volume fraction evolution
rule which was used is not accurate enough.

22
 CHAPTER 2

2.4.3 Applications of SMP models

While many SMP models have been proposed there has been very little work done in
applying these analytical models to industrial applications through the use of numerical
models. In 2009 Reese et al. [43] developed a 3D SMP model and applied it to a SMP stent
through the use of computational finite element modelling. This work is purely qualitative
and is not compared to experimental data. Reese et al. stated “To our knowledge FE-based
simulations of 3-D [shape memory polymer] structures are not found in any previous
publication”. This is echoed by Ghosh in 2011 [73], “There is a paucity in experimental data
in the 3D response of [shape memory polymer] materials”.

2.5 Summary and contribution

The work presented in this thesis aims to contribute to the knowledge within the field
through applying the analytical models of Liu et al. [64] and Barot et al. [77] to simulate the
behaviour of shape memory polymers within 3D and axisymmetric models with the purpose
of applying these models to the laser welding of a Boston Scientific balloon catheter joint.
The model of Liu et al. was chosen for initial exploratory modelling work, primarily for its
simplicity. The Barot et al. model was chosen because it provided a flexible framework that
would enable the modelling of large deformations with a relatively simple hyperelastic
material model.

A lot of work has been done on characterisation and modelling of SMPs. However there is
very little data available for the RNF-100® heat-shrink tubing and there has been very little
work done on applying these models to industrial applications. The main three novel
contributions of this work are:

1. To accurately characterise and understand the thermomechanical behaviour of the

RNF-100® heat-shrink tubing used in the laser welding process. There is no data in
the literature on many of the properties needed to model this material such as
modulus as a function of temperature, modulus as a function of strain, viscoelasticity
and plasticity.
2. To use an SMP model in an industrial FE application and compare the results with
experimental data. To our knowledge this has not been done before.

23
 CHAPTER 2

3. This work represents the first thermomechanical simulation of the laser bonding of
balloon catheters. This model will combine heat-transfer, an SMP material and a
model of a molten polymer in order to predict the melt flow that occurs during the
welding process. There is no model like this in the literature.

This work has the potential to make a strong contribution to the field in terms of improving
the understanding of the welding process and reducing the need for trial and error
experimentation in the early product development stage of catheters.

24
3 Experimental methods

This chapter contains details of the experimental techniques used in this work. The majority
of the techniques described here are used to characterise the material properties of the
heat-shrink tubing. Some imaging techniques are also described which are used to image the
catheter after welding. The determined material properties are used as inputs in calibrating
the constitutive behaviour of the heat-shrink material through the development of a user
defined material model, as described in Chapter 6. The PEBAX® properties were obtained
from existing literature. Also, separate experimental data is used for validating the
performance of the computational model developed.

The majority of the techniques described here are well known and widely used in industry
and academia. However, a new experimental technique was also developed and validated to
characterise the pressure-volume relationship, and hence elastic modulus, of small-gauge
heat-shrink tube as a function of temperature. Another experiment, known as the hanger
test, as developed by Boston Scientific, is also described and used to characterise / validate
material properties.

It is crucial in any model to know the properties of all the materials involved as accurately as
possible. During the welding process the temperature of the heat-shrink reaches 240°C (as
measured by Boston Scientific the IR thermometer (Land M6)) and the PEBAX® reaches
approximately 200°C (model prediction [93]) and the whole welding process takes
approximately 20 seconds. As the properties of polymers depend on temperature and time
it is important to define the ranges relevant to the process. With this in mind a list of
properties needed was drawn up and a series of tests were devised to obtain these
properties. Extensive testing was carried out at 120°C because it was found that when testing
at higher temperatures, the material would discolour and it became stiffer. The high
temperatures and extended duration of the tests may have caused the sample to react with
the oxygen in the air. However it was found that the modulus is relatively insensitive to
temperature above the crystalline melting point of approximately 100°C (see Figure 4.4) and
so it was decided that it was acceptable to test at 120°C.

25
 CHAPTER 3

3.1 Dynamic mechanical thermal analysis

A dynamic mechanical thermal analyser (DMTA) measures the modulus and the damping of
a viscoelastic material, as a function of temperature, as it is subjected to a sinusoidal stress.
The modulus is the ratio between stress and strain and in a viscoelastic material the modulus,
E*, is a complex quantity,

E* = E' + iE'' (3.1)

where E'' is the loss modulus and E' is the storage modulus. E'' represents energy dissipated
during deformation of the sample, and E' refers to the ability of a material to store energy.

A viscoelastic material has properties which contain elements of both elastic and viscous
behaviour. In an elastic body, stress and strain are in phase, and stress is proportional to
strain. In viscous fluids stress and strain are 90° out of phase and stress is proportional to the
rate of strain. In a viscoelastic material under sinusoidal loading, the strain is between 0° and
90° out of phase. The strain lags behind the stress, this lag is represented by the phase angle
δ [82],

tan (δ) = E''/E' (3.2)

The sample is placed in the DMTA which exerts a periodic tensile force of known frequency
on it. This force is usually sinusoidal but complex waveforms can also be used. The
temperature is increased at a set rate as a function of time and the force and displacement
are measured as a function of temperature. Figure 3.1 shows a sample clamped within the
DMTA.

Figure 3.1 Setup for DMTA tensile test in which a sample is clamped

Dynamic mechanical thermal analysis can be used to determine the glass transition,
relaxation spectra, degree of crystallinity, molecular orientation, phase separation and
structural/morphological changes of copolymer blends. In this work the DMTA was mainly

26
 CHAPTER 3

used to determine the modulus of the heat-shrink in different conditions. The DMTA was
also used to determine the thermal expansion coefficient.

3.2 Heat-shrink tests performed on DMTA

As heat-shrink tubing is an SMP the parameters that needed to be measured depend on the
model that will be used. To decide on which model to use tests were performed to determine
the degree of plasticity and viscoelasticity.

The properties needed for the stored strain SMP model of Liu et al. [64] are

i. The modulus of the frozen phase

ii. The modulus of the active phase
iii. The frozen fraction
iv. The thermal expansion coefficient α

Once these properties were known the model could be calibrated. To validate the model a
recovery force experiment was performed.

The heat-shrink tubing used in the welding process has a diameter of 2-3 mm and a wall
thickness of approximately 0.4 mm which makes testing the properties very difficult with
standard approaches and techniques. Due to this a large size (diameter 70 mm) heat-shrink
tube was obtained for Boston Scientific. This large tube was made from the same material
and was processed in the same way as the smaller tube. The heat-shrink tubing was pre-
shrunk in an oven at 120°C to remove the shape memory effect for all of the tests except the
recovery force test. The large heat-shrink tubing was used to determine the qualitative
properties of the heat-shrink material however the final quantitative data used for the model
was determined from the small heat-shrink tubing which was the tubing used in the
experiments.

3.2.1 Frequency sweep

The DMTA was used to measure the variation of the modulus of the heat-shrink varied with
frequency. A rectangular sample measuring 8 mm x 30 mm x 1.2 mm was cut out of pre-
shrunk heat-shrink as shown in Figure 3.2. Five samples were measured and placed in the
DMTA.

27
 CHAPTER 3

1.2 mm

8 mm

30 mm

Figure 3.2 Sample preparation for large size heat-shrink tube in circumferential direction

The chamber was heated to 120°C and left for 5 minutes to come to thermal equilibrium.
Then a sinusoidal strain of 0.5 % was applied. The frequency of the applied strain was varied
from 0.1 Hz to 100 Hz. The load required is used to calculate the modulus and this is plotted
against frequency. This test was repeated for five samples at 120°C and five samples at 30°C.
This plot is used to assess the viscoelastic properties.

3.2.2 Temperature dependence of elastic modulus

Two methods were used to measure the elastic modulus as a function of temperature: A
dynamic method and a static method. For the dynamic method a rectangular sample of heat-
shrink was placed in the DMTA. The temperature was ramped from 20°C to 120°C at a rate
of 2°C/min. A small sinusoidal load of 0.5 % strain was applied to the sample and the force
required to maintain this load was recorded by the DMTA. The DMTA software calculated
the stress and strain from the applied force and sample dimensions and also calculated the
modulus as a function of temperature. Ten samples were used for each of these tests.

The static method also used rectangular samples in the DMTA. However, in the static method
a number of tensile tests were carried out at different temperatures and the modulus was
for each tensile test. In these tests the temperature was increased in intervals of 10°C and
then the temperature in the chamber was held constant for 5 minutes before the tensile test
was performed. The pause was to allow the sample to reach the same temperature as the
chamber. The tensile test was performed to a low strain value to minimize the plastic

28
 CHAPTER 3

behaviour. The modulus was calculated by measuring the slope of the stress/strain curve for
each tensile test, and then the modulus was plotted as a function of temperature.

3.2.3 Thermal expansion coefficient α

The DMTA was also used to measure the thermal expansion coefficient α. A sample of pre-
shrunk heat-shrink was placed in DMTA. A small constant load of 0.001 N was applied then
the sample was heated up at a rate of 2°C/min and the change in displacement was
measured. A load of zero would be ideal but this was not possible in the DMTA. The test was
carried out with a tensile load and with a compressive load to minimize the error introduced
by the material straining under the small load. This test was repeated for five samples.

3.2.4 Plasticity

The plastic behaviour of the heat-shrink was characterised in the DMTA by performing tensile
tests to various different strain levels. The stress/strain curves are the plotted for loading
and unloading and the amount of plasticity can be measured in the strain that is not
recovered during unloading. All plasticity tests were carried out at 120°C.

3.2.5 Anisotropy

The properties of polymer tubes are often different in different directions due to extrusion
processes. To measure this, tensile tests were performed on samples cut from the axial
direction and these where compared which samples cut from the circumferential directional.

3.2.6 Frozen fraction

The frozen fraction, ϕf, is a parameter used in the stored strain SMP model of Liu et al. [64]
and not an intrinsic material property. The method used by Liu et al. was to fit a curve to the
free strain recovery curve. This is a somewhat arbitrary curve to choose but it produces good
results. In this work another method was attempted using data from differential scanning
calorimetry (DSC). DSC measures energy flowing into a sample during heating at a constant
rate and reveals how much energy is absorbed or released during phase transitions. The

29
 CHAPTER 3

extra energy absorbed by the sample during the melting of the crystalline phase was used as
a measure of the frozen fraction.

The other method used was an approximation of the free strain recovery used by Liu et al.
where the strain is measured during the recovery of an unconstrained sample. The Liu et al.
experiment was performed with a video extensometer and a heated chamber but the
experiment used in this work was performed in the DMTA. A small sinusoidally varying load
was applied to a sample of heat-shrink and the temperature was ramped slowly. As the heat-
shrink recovered the DMTA records the change in displacement. This gives a strain recovery
as a function of temperature graph which was used to fit a curve to the frozen fraction.

3.2.7 Recovery force tests

To measure the recovery force of the large-size heat-shrink tubing, rectangular samples were
cut and placed in the DMTA. A small preload was applied and then the temperature was
ramped from room temperature up to 120°C at 2°C/min. When the crystalline phase of the
heat-shrink melts the rubbery phase tries to contract. This produces a force which is
measured by the DMTA. This test was repeated twenty times.

3.2.8 Stress relaxation tests

Stress relaxation experiments were performed to investigate the effect of time on the shrink
force. In these tests a sample is held at a constant strain in the DMTA and the drop in stress
is measured over time. The sample was heated to 120°C and then stretched to a strain of 5
%. The strain was held constant for 10 minutes and the stress was measured. These tests
were repeated three times.

3.3 Differential scanning calorimetry

In differential scanning calorimetry (DSC), the temperature of a sample and a reference pan
is raised at the same rate and the heat flow required to keep the samples at the same
temperature is measured. The difference in heat flow is then plotted as a function of
temperature. Changes in the heat flow indicate phase changes and these appear as peaks
and dips in a plot of heat transfer rate as a function of temperature. DSC can be used
determine heat capacity, purity, and degree of crystallinity as well as measuring the

30
 CHAPTER 3

enthalpies of chemical reactions and phase changes that occur during heating of a material.
A typical DSC output is shown below in Figure 3.3. The three phase changes shown in this
figure show the glass transition Tg, the crystallisation temperature Tc and the melting

temperature Tm.

Figure 3.3 Typical output from a DSC indicating he glass transition Tg, the crystallisation temperature Tc and
the melting temperature Tm.

The sample to be investigated is placed on an aluminium pan and a lid is crimped onto the
pan. This prevents the sample interacting with the nitrogen which is used to heat the furnace.
The samples used are typically 2-50 mg in mass. Another pan is prepared and left empty. This
is used as the reference pan. The chamber is small so it can heat up and cool down fast. The
temperature increases at a predetermined rate and is monitored by two thermocouples.
Heat flow differences between the sample and reference pan will occur due to exothermic
or endothermic reactions in the sample.

The heat capacity, Cp, is the quantity of heat, q, needed to increase a unit mass of a material
by 1°C and is calculated using Equation 3.3

Cp = q/ΔT (3.3)

where ΔT is the change in temperature. A step in the DSC curve indicates a change in heat
capacity.

Above the glass transition temperature, molecules have more freedom to move and so the
sample can absorb more heat. This leads to an increase in heat capacity, which is represented
by a downward step on the graph.

31
 CHAPTER 3

When the sample reaches a certain temperature, crystallisation begins as the molecules
begin to move in ordered arrangements. This means the material gives off heat and this is
represented by a large peak in the graph. The area under the peak is the latent energy of
crystallisation.

Melting occurs when the molecules have too much energy to be held in a crystal structure.
Extra energy is needed to melt the crystals and this result in an endothermic dip in the graph.
The area between the dip and the baseline represents the latent heat of fusion of the
material.

3.4 Microscopy

Boston Scientific use microscopy extensively to examine the quality and consistency of
catheter joint bonds. The two main types of microscopy used by Boston Scientific are
polarized microscopy and optical coherence tomography (OCT).

3.4.1 Polarized microscopy

In polarized microscopy a light microscope is used with two polarizing filters, one between
the light source and the sample the other between the sample and the eyepiece. The
polarizing filters are set up perpendicular to each other so no light passes through when the
sample is isotropic and amorphous. However if the sample consists of optically anisotropic
crystals, the light will be split into two rays polarised perpendicular to each other [83]. Some
of this light can then pass through the second polarizing filter which produces an image with
good contrast.

Sample preparation involves carefully cutting a section out of the catheter in the axial
direction. This method of imaging produces a very clear image of the melt pool. However,
this is a destructive technique and sample preparation is difficult.

32
 CHAPTER 3

Balloon

Outer tube

Bonded region

Non-melted region Melt pool

Figure 3.4 Image showing advantages of polarized light (lower image) over non-polarized light (upper image)

A comparison of regular microscopy and polarized microscopy is shown above in Figure 3.4.
The melt pool is clearly visible in the lower image as the darker region while the rest of the
catheter is much brighter. The upper image using non-polarized light produces no contrast
between the melted and non-melted regions showing the advantages of polarized
microscopy. The material in the catheter is anisotropic due to the extrusion process but when
the material is heated above its melting point, the anisotropic crystals melt and the material
becomes isotropic. This is the reason the melt pool is much darker than the non-melted
region when imaged with polarized microscopy.

3.4.2 Optical coherence tomography

Boston Scientific also uses optical coherence tomography (OCT) to image the bonded region.
This technique produces a cross-sectional slice image in transparent or translucent samples.
The major advantage of OCT over polarized microscopy is that OCT is non-destructive.

33
 CHAPTER 3

Non-melted Outer tube

Balloon Melt pool region

Edge of melt pool

Figure 3.5 OCT image of bonded region of catheter

An OCT image of a bond is shown above in Figure 3.5. The melt pool is clearly visible. Note:
the dimensions in the vertical direction below the surface are not accurate due to refraction
effects.

3.5 Shrink force test

To measure the shrink force of the heat-shrink tubing Boston Scientific developed a test
called the hanger test. In this test a short sample (2 mm) of heat-shrink tubing was placed on
two horizontal hangers which are held in a fixed position in a DMTA. The setup is shown
below in Figure 3.6. The temperature in the DMTA chamber was then ramped up to 150°C
over approximately 5 minutes which causes the tube to contract. As the tube contracts it
places a force on the hanger and this is measured by the DMTA. The results of this test
depend on the distance between the hangers so care has to be taken that the test is setup
accurately.

34
 CHAPTER 3

Heat-shrink tube
Hangers held in
clamps of DMTA
2 mm

Hanger O.D.
mm

0.8 mm

Figure 3.6 Schematic diagram of shrink force test used by Boston Scientific

The change in shrink force as the tube shrinks was also investigated. This was done by varying
the outside diameter of the hangers (hanger O.D. in Figure 3.6) and measuring the shrink
force as a function of the hanger O.D.

3.6 Pressure test

As mentioned in Section 3.2, the heat-shrink tubing used in the laser welding bonding
process has a diameter of 2 mm. This makes testing in the DMTA difficult. Therefore, a
custom designed pressure test was developed to measure the modulus of the small heat-
shrink tubing. The principle of the pressure test is to measure the pressure that the heat-
shrink tubing applies to a fluid during a constrained recovery. The dimensions of the tube
during the test do not change because the tube cannot compress the fixed volume of oil
inside the tube. The internal diameter of the tube in the expanded state is approximately 2
mm and the internal diameter of the recovered tube is approximately 1 mm. This test
assumes that when the heat-shrink tube is heated to 120°C it exerts a pressure that is
equivalent to the pressure that would be needed to expand the tube from the recovered
diameter to the expanded diameter. This is equivalent to heating the 2 mm tube, letting it
shrink to a stress-free state with a diameter of 1 mm and then pressurising it to expand it
again to 2 mm.

As the dimensions of the tube do not change during the pressure test, the pressure of the
fluid in the tube is proportional to the modulus of the tube and therefore the modulus can
be calculated from the pressure produced. The dimensions were supplied by Boston
Scientific [99]. An FE model of this test was used to calculate the modulus from the pressure
test results and this is described in Section 4.7.

35
 CHAPTER 3

The pressure is equal in magnitude to the radial stress σr at the inner surface of the tube and
assuming that there is no axial strain and the material is Neo-Hookean and incompressible,
the radial stress at the inner surface of a thick-walled cylinder undergoing finite deformation
can be calculated from the equation below [98].

𝐴2 − 𝑎2 1 1 𝐴𝑏
𝜎𝑟 = 𝐺 ( ( 2 − 2 ) + 𝑙𝑜𝑔 ) (3.4)
2 𝑎 𝑏 𝐵𝑎

In this equation G is the shear modulus, A and B are the internal and external radii before
the tube is expanded and a and b are the internal and external radii after expansion.

In this test a 200 mm length of tube was filled with oil. One end of the tube was sealed with
a clamp and the other end was connected to a pressure transducer. The pressure sensor
used was a Honeywell SCC SMT which required a supply current of 1 mA and produced an
output voltage of 2.9 mV for every 1 psi of pressure measured (including atmospheric
pressure). The tube was then immersed in bath of oil which had been heated to 120°C. Upon
immersion in the oil bath the heat-shrink tube is rapidly heated to above its crystalline
melting temperature. The pressure of the oil within the tube was measured as the heat-
shrink tube tried to contract. A schematic diagram of the pressure test set-up is shown below
in Figure 3.7.

Pressure
Oil at 120°C
sensor

Figure 3.7 Schematic diagram for pressure test

A sealing mechanism had to be created between the heat-shrink and the pressure sensor.
The configuration below in Figure 3.8 was decided upon. This allows the tubes to be changed
quickly and easily between tests while provided a leak-proof seal. As the nut is screwed in
towards the back-plate, it pushes the metal spacer which puts pressure on the rubber seal.
As the seal is compressed it pushes the heat-shrink tube against the small steel tube forming
a seal.

Once the sealed tube is lowered into the hot oil the pressure inside the tube increases quickly
and then drops off with time. To ensure a leak was not causing this drop in pressure, a leak

36
 CHAPTER 3

test was performed. This involved placing the sealed oil-filled tube on flat surface and putting
weights on the tube until the pressure inside was approximately the same as that seen when
the heat-shrink is placed in the hot oil.

Steel back plate

Threaded nut

Threaded bolt

Steel spacer
Pressure sensor

Heat-shrink tube
Steel tube

Rubber seal

Figure 3.8 Sealing mechanism developed for pressure test

Once leakage had been ruled out a series of tests were carried out on five different grades
of heat-shrink. These five types of tube have very similar dimensions and are made from the
same material but produce different shrink forces when tested on Boston Scientifics hanger
test. This is due to the varying levels of irradiation which the tubes received when they were
manufactured. Higher levels of irradiation produce more cross-linking and more cross-linking
results in a higher modulus. The modulus of the tubes is calculated from the pressure they
produce. To validate this method the modulus was then used as the input into a finite
element model of the hanger test. The results of this model were then compared with the
results of the hanger test performed by Boston Scientific.

37
4 Results of material characterisation

This chapter contains the results of the experiments on the heat-shrink tubing described in
Chapter 3 and the details of the material model used to model the PEBAX® which is the
material the balloon and outer tube are made from. The objective of this chapter is to
comprehensively characterise the polyolefin heat-shrink tubing and to obtain material
properties which are required for subsequent modelling. No experiments were performed
on the PEBAX® - PEBAX® data was obtained from the literature for the PEBAX® model. The
heat-shrink tubing was tested through DMTA and the pressure test as described in Chapter
3.

Two types of heat-shrink tube were tested, the large heat-shrink with a diameter of 70 mm
and small heat-shrink (Raychem tab 10 Lot No. 15867) with a diameter of 3 mm. The small
tubing is used in the welding process but due to its small size it was difficult to characterise.
Therefore, the large tube was used to characterise the behaviour of heat-shrink tubing
through more extensive testing. Various characteristics of the heat-shrink tubing were
investigated including:

i. The elastic behaviour

ii. The effect of viscoelasticity
iii. The effect of plasticity
iv. The recovery force
v. The thermal expansion coefficient

This chapter describes the material behaviour of PEBAX® through data obtained from the
literature and details on how this data was used to create a material model for PEBAX® to be
used in modelling the welding process.

4.1 Measurement of elastic modulus of large heat-shrink tube

4.1.1 Elastic properties under dynamic loading

The first method used to measure the modulus of the heat-shrink tube was a dynamic test
conducted with the use of a Dynamic Mechanical and Thermal Analyser (DMTA). This

38
 CHAPTER 4

involved applying a sinusoidal strain with amplitude of 0.5% at a frequency of 1 Hz and

measuring the force while increasing the temperature. These tests were performed on 10
rectangular samples cut from the large size heat-shrink tubing as shown in Figure 3.2. The
dimensions of the samples were measured with a digital callipers. The minimum and
maximum thickness of the samples were 1.18 mm and 1.22 mm respectively. Some samples
were shrunk before testing and some were tested in the stretched (as received) state. Results
are shown below in Figure 4.1 below.

160

140
Storage Modulus (MPa)

120
Preshrunk HS
100
As-recieved HS
80

60

40

20

0
20 40 60 80 100 120 140 160
Temperature (°C)

Figure 4.1 Storage modulus of large-diameter heat-shrink tubing as a function of temperature. Tests were
conducted at a frequency of 1 Hz and the heating rate is 2°C/min.

The modulus varies strongly with temperature up to approximately 120°C, after which it
stabilises. The modulus at room temperature is approximately 140 MPa and above 120°C the
modulus is less than 1 MPa. The curve for the pre-shrunk heat-shrink tubing is more linear
than that of the as-received tube. This is due to the shape memory effects in the as-received
tube which cause it to shrink during testing.

4.1.2 Effect of frequency

The effect of test frequency on the modulus was also investigated using the procedure
described in section 3.2.1. Results are shown in Figures 4.2 and 4.3 for DMTA tests over a
range of loading frequencies at constant temperature. In Figure 4.2 tests were performed at
120°C because it was found that when testing at higher temperatures, the material began to
react with the air in the chamber. However it was found that the modulus does not vary a
lot with the temperature above the crystalline melting point of approximately 100°C (as seen

39
 CHAPTER 4

in Figure 4.4, the modulus at 1 Hz decreases by 5 % when the temperature is increased from
120°C to 160°C), and so it was decided that it was acceptable to test at 120°C. Figure 4.4
below shows the variation in modulus over a smaller frequency range at three different
temperatures.

2.5
Storage Modulus (MPa)

1.5

0.5

0
0.1 1 10 100
Frequency (Hz)
Figure 4.2 Storage modulus of large-diameter heat-shrink tubing as a function of frequency. Tests were
performed at a constant temperature of 120°C with loading in the circumferential direction.

200

180
Storage Modulus (MPa)

160

140

120

100

80
0.1 1 10 100
Frequency (Hz)

Figure 4.3 Storage modulus of large-diameter heat-shrink tubing as a function of frequency. Tests were
performed at a constant temperature of 30°C with loading in the circumferential direction.

40
 CHAPTER 4

0.8

0.75
Storage Modulus (MPa)

0.7

0.65
120°C
0.6 140°C
160°C
0.55

0.5
1 10
Frequency (Hz)

Figure 4.4 Storage modulus of large-diameter heat-shrink tubing as a function of frequency for 3 temperatures.

These results show that the heat-shrink tubing is a viscoelastic material, as the modulus
increases slightly with higher frequency. The welding process takes place over approximately
20 s, and so the lower ranges of frequency are more relevant than the higher ranges. It is not
known what causes the increase in modulus between 20 Hz and 100 Hz in Figure 4.2, but
deformation at these high frequencies are not relevant to the laser welding process.

From Figure 4.4 is can be seen that the change in modulus from 120°C to 160°C is very small
at low frequencies but increases at higher frequencies. The change in modulus from 1 Hz to
10 Hz is approximately 20%. The highest temperature reached during the welding process is
approximately 245°C.

4.1.3 Static tests

Static tests were also performed in the DMTA to determine the stress-strain relationship.
This involved tensile testing at 120°C and at strain rates of 30%/min. The welding process
takes place over 20 seconds and during this time the diameter of the heat-shrinks halves.
The strain rate varies significantly over the 20 s and it varies throughout the heat-shrink and
therefore choosing a strain rate to do the tests at was difficult. The maximum strain rate in
the process occurs during the free recovery of the heat shrink at the centre of the laser,
before it comes into contact with the PEBAX®. In this region the strain reduces by
approximately 40% over approximately 2 s. However, the heat-shrink strain rate when the

41
 CHAPTER 4

PEBAX® begins to flow (the most important part of the process) is in the region of 30%/min
and lower. This is estimated from the change in radius of the outer tube during the welding
process. The radius of the outer tube decreases by approximately 0.05 mm over
approximately 10 s and this equates to a 5% change in strain over 10 s for the heat-shrink;
therefore a strain rate of 30% was chosen for these tests. Figure 4.5 below shows a stress-
strain curve for the loading and unloading of a rectangular sample of the large heat-shrink
tubing in the circumferential direction. It can be observed that the stress-strain relationship
is non-linear. Figure 6 shows the repeated loading of a single sample with increasing levels
of strain.

0.18

0.16

0.14

0.12

0.1
Stress (MPa)

0.08

0.06

0.04

0.02

0
0 10 20 30 40 50 60
% Strain

Figure 4.5 Tensile loading and unloading of heat-shrink at 120°C. Test performed in the circumferential
direction and with a strain rate of 30%/min.

42
 CHAPTER 4

0.14

0.12

0.1
Stress (MPa)

0.08

0.06

0.04

0.02

0
0 10 20 30 40 50
Strain (%)
Figure 4.6 Repeated loading of a single sample. Testing was conducted at 120°C.

These curves show that when the heat-shrink is stretched to 45% strain about 2.5% strain is
not recovered due to plasticity. This is quite a small amount of plasticity. At higher strain
values slipping in the clamps became significant.

Static tests were also performed at a range of temperatures to compare with the dynamic
(DMTA) results. Figure 4.7 below shows the conditions used for these tests and Figure 4.8
shows the results. Young’s modulus is calculated by measuring the slope of the stress/strain
curve, which at low strain levels is very linear.

43
 CHAPTER 4

14 140

12 120

10 100

Temperature (°C)
Strain (%)

8 80

6 60

4 40

Strain
2 20
Temperature

0 0
0 20 40 60 80 100 120
Time (min)
Figure 4.7 Tensile tests were performed at a number of temperatures. This graph shows the applied loading
conditions. The temperature was increased then left to come to equilibrium. Then a stress/strain test was
conducted at relatively small strains to determine the Young’s modulus. The strain rate used was 2%/min.

160

140

120

100
Modulus (MPa)

Dynamic
80 Static

60

40

20

0
0 20 40 60 80 100 120 140 160
Temperature (°C)

Figure 4.8 Comparison of static and dynamic test results. The static test results are from Figure 4.7 and the
dynamic test results are from Figure 4.1.

44
 CHAPTER 4

The static and dynamic results show a substantial difference this may be due to the samples
being at different temperatures because of the temperature of the sample in the dynamic
test lagging behind the temperature of the chamber.

Dogbone samples were created using a sample cutter to cut samples for loading in the
circumferential direction. The samples were then shrunk and tested. These samples showed
a much greater repeatability than the rectangular samples when stretched to high strains.
Figure 4.9 below shows 6 samples tested to large strains (>150% at centre).

0.40

0.35

0.30

0.25
Stress (MPa)

0.20

0.15

0.10

0.05

0.00
0 20 40 60 80 100 120 140 160 180
Strain (%)
Figure 4.9 Comparison of stress/strain curves for 6 rectangular samples.

The greater repeatability of the dogbone samples is demonstrated in Figure 4.10 and Figure
4.11 below. In Figure 4.10 the force is plotted as a function of length change for 6 dogbone
samples and in Figure 4.11 the same data is plotted as stress as a function of strain. The
stress/strain curves vary more than the force/length change curves because the strain is
calculated from the distance between the clamps and this distance varied slightly for each
test. Ideally the strain would be measured by video extensiometry measuring the strain at
the narrow part of the dogbone sample but this was not possible in the DMTA and also the
strain in the wide parts of the dogbone sample is significant.

45
 CHAPTER 4

1.2

0.8
Force (N)

0.6

0.4

0.2

0
0 2 4 6 8 10 12 14 16
Length change (mm)

Figure 4.10 Force as a function of the change in length for 6 dogbone samples. Original sample length was 7.9
mm.

0.35

0.3

0.25
Stress (MPa)

0.2

0.15

0.1

0.05

0
0 20 40 60 80 100 120 140 160
Strain %
Figure 4.11 Same data as above but converted to stress as a function of strain. Strain calculated using distance
between clamps as original sample length. This causes the increase in variation.

In order to use this data to create a material model for the heat-shrink tubing, a finite-
element model of the dogbone tensile test was created. The set-up of the model is shown
below in Figure 4.12. The test was modelled in 3-D using the same dimensions used in the
experimental tests. A Neo-Hookean hyperelastic model (see Equation 2.6) was used to model
the heat-shrink tubing and the stiffness was adjusted until the model data matched the
experimental data. Figure 4.13 is a comparison of the force as a function of length change
for the model and experimental data and it shows good agreement. The results are plotted

46
 CHAPTER 4

as force as a function of length change because the strain varies significantly over the sample
and is much higher at centre as shown in Figure 4.12.

Figure 4.12 Finite element model of tensile test of dogbone heat-shrink sample. A Neo-Hookean model was
used and the stiffness was adjusted to match the experimental data.

1.4

1.2

0.8
Force (N)

Experimental
0.6 Model with E= 0.44 MPa

0.4

0.2

0
0 2 4 6 8 10 12 14 16
Length change (mm)
Figure 4.12 Force as a function of length change for tensile test on large heat-shrink test samples. Comparison
of experimental tests with FE model of test. The Neo-Hookean hyperelastic model was used to model the heat-
shrink tubing.

47
 CHAPTER 4

These tests show that the stress/strain behaviour of the large heat-shrink tube (at 120°C and
in the circumferential direction) can be accurately modelled using an incompressible Neo-
Hookean model with a Young’s modulus of 0.44 MPa. Young’s modulus is converted to shear
modulus for use in Equation 2.6.

The plasticity of the material at large levels of deformation was investigated (Figure 4.13) by
stretching a number of samples to various strain levels, as described in section 3.2.4.

0.3

0.25

0.2
Stress (MPa)

0.15

0.1

0.05

0
0 10 20 30 40 50 60 70 80 90 100
Strain (%)
Figure 4.13 Graph showing loading and unloading of heat-shrink to large deformations. In these tests the force
is ramped at 0.5 N/min and strain is calculated from distance between clamps.

From this graph it can be seen that there is a relatively small amount of plasticity in the
material even when stretched to large strains. To quantify the plasticity the non-recoverable
strain was plotted as a function of the maximum strain (Figure 4.14).

10
Non-recoverable strain

0
0 20 40 60 80 100
Maximum strain
Figure 4.14 Graph of non-recoverable strain as a function of maximum strain. Same data as Figure 4.13.

48
 CHAPTER 4

From Figure 4.14 it can be seen that there is no clear yield point for this material below 90%
strain. The non-recoverable strain is approximately 7-9% of the maximum strain.

4.1.5 Axial tests

A small number of tests were performed in the axial direction. It was not possible to create
dogbone samples in the axial direction and so the tests were only performed to relatively
low strains. A small number of tests were also performed on the small size heat-shrink
tubing. Figure 4.15 below shows a comparison of the stress-strain curves for the large and
small tubes. Both samples were rectangular in shape and oriented in the axial direction.

0.25

0.2
Stress (MPa)

0.15

0.1
Small tube
large tube
0.05

0
0 5 10 15 20 25 30 35 40
Strain (%)

Figure 4.15 Comparison of stress versus strain curves for small and large heat-shrink tubing. Tests were
performed in the axial direction at 120°C.

This graph indicates there is a significant difference between the modulus of the small and
large tubes. According to this limited data, the modulus of the small tube is 55% larger than
the modulus of the large tube. The modulus was calculated from the slope of the two curves
in Figure 4.15.

4.1.5 Anisotropy

The level of anisotropy in the heat-shrink was quantified by measuring the modulus in the
axial and circumferential directions. The modulus was measured in the DMTA using the
frequency sweep described in section 3.2.1. The results of these tests are shown below in
Figure 4.16.

49
 CHAPTER 4

0.85

0.8

0.75
Modulus (MPa)

0.7

0.65

0.6 Axial
Circumferential
0.55

0.5
0.2 2 20
Frequency (Hz)
Figure 4.16 Comparison of dynamic modulus in axial and circumferential directions. Results are an average of
5 samples. The error bars are ± the standard deviation.

On average the modulus in the axial direction is approximately 11% higher than the modulus
in the circumferential direction. The tube is stiffer in the axial direction because the extrusion
process causes the polymer chains to align in the axial direction.

4.2 Recovery force

As described earlier, the heat-shrink tubing recovers to its original shape when heated. If the
material is held at constant strain then the heat-shrink generates a recovery force. This
recovery force is used to apply a pressure to the catheter assembly during welding. Modelling
this recovery force is a key part of creating a computational model of the laser bonding
process.

Many tests have been performed on the recovery force of the large heat-shrink tubing as
described in section 3.2.7. Figure 4.17 below shows that these tests are quite repeatable. In
these tests a small tensile strain of 0.1% was applied and the force required to maintain the
strain as the sample is heated was measured. This is the reason for the small positive stress
observed at low temperature. As the temperature increases, the force required initially
decreases and becomes negative (compressive) due to the thermal expansion of the sample.
As the material is heated to approximately 55°C, the crystalline phase begins to melt and the

50
 CHAPTER 4

sample begins to recover. Above approximately 55°C the tensile stress due to the shape
memory effect exceeds the thermal stress and the force required to keep the strain constant
becomes positive (tensile). The modulus decreases significantly with temperature and this
reduces the stress in the material. The bump in the curve around 100°C is not understood
but probably due to the processing conditions when creating the shape memory effect.

0.6

0.5

0.4
Stress (MPa)

0.3

0.2

0.1

-0.1
30 50 70 90 110 130
Temperature (°C)
Figure 4.17 Stress as a function of temperature for 13 samples of the large diameter heat-shrink tubing. In
these tests the strain was kept constant at 0.1% while the temperature was ramped at 2°C/min.

Heat-shrink that had been shrunk was heated, stretched and cooled to create “Re-expanded
heat-shrink”. Samples with different strains were created and tested to see how much of the
forming force was recovered. The recovery force graphs of these samples are shown below
in Figure 4.18.

51
 CHAPTER 4

1.6

1.2

0.8
Force(N)

0.4

Forming force of 1.6 N

-0.4 Forming force of 1.3 N

Forming force of 1.1 N

-0.8
20 40 60 80 100 120
Temperature (°C)
Figure 4.18 Recovery force of re-expanded heat-shrink as a function of forming force.

These graphs show that not all of the force needed to form the samples is recovered during
reheating. This is thought to be primarily due to plasticity in the material but stress relaxation
may also have an effect on the recovered force. The recovered force is approximately 80%
of the forming force, however, these results showed significant variation from test to test.

4.3 Viscoelastic properties

4.3.1 Stress relaxation

Stress relaxation experiments were performed as described in section 3.2.8. In this test a
sample is held at a constant strain in the DMTA at 120°C and the stress is measured over 10
mins. The results are shown below in Figure 4.19.

52
 CHAPTER 4

0.028

0.027

0.026 Stress
Stress (MPa)

Logarithmic fit
0.025

0.024

0.023

0.022
0.01 0.1 1 10
Time (mins)

Figure 4.19 Stress relaxation of heat-shrink tubing at 120°C with a strain of 5%.

From Figure 4.19 it can be seen that the stress decreases logarithmically with time. Overall,
the stress relaxation over the 10 minutes accounted for approximately 17% of the overall
stress. It should be recognised that in the welding process, the heat-shrink material recovers
its strain over 20 seconds at temperatures above 120°C, so it is therefore estimated that the
drop in stress over this time period is on the order of 10%.

4.3.2 Creep

In these tests the DMTA was used to apply a constant load to the material and monitor the
change in strain with time. A constant stress of 0.05 MPa was applied and dogbone samples
were used. The change in strain is shown below in Figure 4.20.

53
 CHAPTER 4

19

18

17
Strain (%)

16

15 Strain

14 Logarithmic fit

13

12
1 10 Time (s) 100 1000
Figure 4.20 Creep test showing strain increasing logarithmically. Dogbone samples used and result is an average
of 3 tests.

After 2 seconds the strain is approximately 14%. The strain then increases by 25% over 20
minutes in an approximately logarithmic fashion. It is estimated that for a constant stress
over time periods relevant to the welding process, the increase in strain of the material is
approximately 10%.

4.4 Thermal expansion coefficient

The thermal expansion coefficient of the heat-shrink, α, was determined by placing a sample
in the DMTA and applying a small constant load as described in section 3.3.1.3. The
temperature was then ramped up to 120°C very slowly (1°C/min) and the displacement
monitored. The slow heating rate was chosen to see how alpha changes during heating. It
can be seen that there is a large increase in length between 60 and 80°C.

54
 CHAPTER 4

9
8
7
6
5
Strain (%)

4
3 Strain (%)
2
Linear fit
1
0
30 50 70 90 110 130 150
Temperature (°C)
Figure 4.21 Strain as a function of temperature with slow heating rate (1°C/min) showing variation of α at
different temperatures.

The thermal expansion coefficient varies significantly during heating however, for the
purpose of calculating a single value it is assumed to be constant and this linear fit is shown
in Figure 4.21. The maximum difference between the linear fit and the experimental strain is
approximately 1% and because the strain level that occurs in the welding process is
approximately 100%, it was determined that the error introduced by this approximation was
small. The average thermal expansion coefficient over the range of 30°C to 120°C is 6.4 x 10-
4
/°C.

4.5 Hanger test

The hanger test was performed by engineers in Boston Scientific and is described in section
3.5. Short samples (2 mm) of heat-shrink tubing with internal diameter of 2 mm were placed
on hangers in the DMTA and the position of the hangers was kept constant while the
temperature in the chamber was increased to 150°C. As the heat-shrink contracts the force
it applies to the hanger is measured by the DMTA. This test was repeated with different
distances between the hangers. The results are shown below in Figure 4.22 and it can be
seen that the shrink force is approximately proportional to the hanger outside diameter (OD)
over quite a large range.

55
 CHAPTER 4

0.4

0.35

0.3
Shrink Force (N)

0.25

0.2
Shrink force
0.15 Linear fit

0.1

0.05

0
0.5 0.7 0.9 1.1 1.3 1.5 1.7 1.9
Hanger OD (mm)
Figure 4.22 Test performed by Boston Scientific showing linear relationship between force and hanger OD.

This graph shows that the outer diameter of the hangers needs to be measured very
accurately for this test. For example, if an outer diameter of 1 mm is used then an error of
0.02 mm would cause a 5% error in shrink force. This effect becomes smaller as the hanger
outer diameter increases and so a larger hanger OD should be used if possible.

4.6 Pressure test

The shrink force of the small heat-shrink tube was also measured using a pressure test as
described in section 3.6. In these tests a heat-shrink tube is filled with oil and then immersed
in oil at 120°C, while the pressure inside the tube is measured.

The first test performed was a leak test to ensure the seal between the tube and pressure
sensor was not leaking (also described in 3.6). The leak test involved placing a weight on the
sealed tube and measuring the pressure of the oil inside the tube. If there was a leak the
pressure would be expected to decrease. The results of the leak test are shown below in
Figure 4.23.

56
 CHAPTER 4

0.095

0.093

0.091
Pressure (MPa)

0.089

0.087 Pressure

0.085 Logarithmic fit

0.083

0.081
0 5 10 15 20
Time (mins)
Figure 4.23 Results of leak test showing increase in pressure with time. This test was performed at 20°C.

As the pressure in the tube does not decrease during the 20 min test it was assumed that the
seal was not leaking. The pressure increases by approximately 14% over 20 minutes during
this test. This increase is thought to be due to stress relaxation in the wall of the heat-shrink
tubing. As the weight is being supported by the elasticity of the tube and the pressure of the
oil, any stress relaxation in the tube will cause an increase in oil pressure.

The results of a pressure test are shown below in Figure 4.24. During the tests the pressure
reaches a maximum within a few seconds and then begins to drop off.

0.09
0.085
0.08
0.075
Pressure (MPa)

Pressure
0.07
Logarithmic fit
0.065
0.06
0.055
0.05
0.045
0.04
0 5 10 15 20 25 30 35 40 45
Time (mins)
Figure 4.24 Variation of pressure with time for pressure test. At t = 0 the heat-shrink tube was immersed in oil
at 120°C at the pressure inside the tube was recorded every 30 s for 40 minutes.

The pressure drops by 50% over 40 minutes. After 10 minutes the pressure has dropped by
34%. When this is compared with the results of the uniaxial stress relaxation experiment in

57
 CHAPTER 4

Figure 4.19, it can be seen that the pressure in the pressure test drops significantly faster
than the stress in the stress relaxation experiment. This is due to the fact that in the stress
relaxation experiment, the stress relaxation occurs in one dimension (circumferential),
whereas in the pressure test the stress relaxation occurs in three dimensions
(circumferential, axial and radial). Creep may also play a role in this pressure drop.

Five different grades of the small heat-shrink tubing were tested. The five tubes had
approximately the same dimensions but produced significantly different shrink forces when
measured on the hanger test. The shrink force value of the tube depends on the elastic
modulus at high temperatures. The modulus of the tube is determined by the amount of
cross-linking which in turn is determined by the amount of irradiation the material received
when being produced [84]. The five different tubes tested were exposed to five different
levels of irradiation and therefore produce different elastic moduli at high temperature.

Figure 4.25 below shows the results of the pressure test on the five different heat-shrink
grades. The pressure plotted is the maximum pressure of the oil in the heat-shrink tubing
which occurs immediately after being immersed in the hot oil bath. At least five different
samples were used for each heat-shrink grade. The error bars show the range which all
samples fell within.

0.14

0.12

0.1
Pressure (MPa)

SF 2.72
0.08
SF 6.67

0.06 SF 10.149
SF 12.09
0.04 SF 13.97
Linear fit
0.02

0
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16
Shrink Force (N)
Figure 4.25 Maximum pressure of the oil during pressure test plotted against shrink force values provided by
Boston Scientific. The error bars show the range all samples fell within. The linear fit shown has been set to
go through the origin.

58
 CHAPTER 4

The pressure produced by the heat-shrink during the pressure test is approximately linearly
proportional to the shrink force measured with the hanger test. This is expected as the
samples have the same dimensions and the only difference between the samples is the
modulus. The linear fit shown in Figure 4.25 has been set to go through the origin as it is
assumed that as the shrink force goes to zero the pressure would also go to zero.

4.7 Determination of modulus of small heat-shrink tubes

To calculate the modulus of the different tubes an FE model of the pressure test was created.
The FE model consisted of an axisymmetric model of an incompressible Neo-Hookean
hyperelastic tube (see Equation 2.6). The initial dimensions of the tube were those of a fully
recovered heat-shrink tube. An arbitrary modulus (1 MPa) was assumed initially and a
pressure load was applied to the tube to cause it to expand. Using an iteration process, the
pressure was varied until the dimensions of the tube after stretching matched those of the
as-received heat-shrink. This pressure was then recorded. As the pressure produced by an
elastic tube is linearly proportional the modulus, as shown in Equation 3.4, the ratio of the
pressure from the FE model to pressure from experimental pressure test is equal to the ratio
of the assumed modulus to the actual modulus of heat-shrink tubing used in the pressure
test. Using this process the modulus of the five tubes was calculated and the results are
shown in Table 4.1.

Results from pressure test

SF (g) P (MPa) E (MPa)

2.72 0.045 0.208

6.67 0.072 0.339
10.15 0.097 0.443
12.07 0.121 0.537
13.97 0.136 0.613
Table 4.1 Modulus values calculated from pressure test for the 5 different grades of heat-shrink

59
 CHAPTER 4

4.8 Determination of the frozen fraction

Two methods were used to calculate the frozen fraction, one using the DSC as described in
Section 3.2.6 and the other was using an approximation of the free strain recovery used by
Liu et al., also described in Section 3.2.6. The DSC method was used to try to connect the
frozen fraction to a physical parameter, namely the crystallinity. However using the frozen
fraction calculated from this method in the simulation of the recovery stress (see section
3.2.7) produced a result which did not agree well with the experimentally observed recovery
stress. Therefore the Liu et al. model of fitting a curve to the free strain recovery curve was
used and the results are shown below in Figure 4.26.

1.2

0.8
normalized strain

0.6
Frozen Fraction

0.4

0.2

0
0 20 40 60 80 100 120 140 160
Temperature (°C)

Figure 4.26. Normalized free strain recovery as a function of temperature. The frozen fraction curve was fitted
to the strain curve.

4.9 Discussion

In this chapter the results of the experimental characterisation of the heat-shrink tubing have
been presented. The majority of the testing was done using a large size heat-shrink tube in
the DMTA as the tube used in the welding process was difficult to test due to its small size.
The only test performed on the small size tube was a pressure test which measured the
pressure applied by the small tube when heated.

The modulus of the heat-shrink and the parameters which affect it were investigated. The
parameter with the largest effect on the modulus was the temperature. Heating the heat-

60
 CHAPTER 4

shrink from 20°C to 120°C causes the modulus to drop by a factor of 100. However, further
heating has a relatively small effect as shown in Figure 4.4.

The viscoelastic properties were investigated by measuring the modulus in a dynamic test at
different frequencies and also by performing stress relaxation and creep tests. It was found
that the modulus is dependent on frequency but over the range of values relevant to the
welding process the effect was quite small. The stress relaxation causes a drop in stress of
approximately 10% over the relevant time period (20 s). Due to these results, it was decided
that viscoelasticity could be left out of the heat-shrink model as a reasonable approximation.

Tensile tests were performed on the large-size heat-shrink using large strains (> 100%). A
hyperelastic model was fitted to the resulting stress/strain curve and it was found that the
Neo-Hookean model produced a very good fit.

The anisotropy was quantified by measuring the modulus in the circumferential and axial
directions. The axial modulus is approximately 10% larger than the circumferential modulus
but this difference was deemed small enough to be left out of the heat-shrink model. The
circumferential direction is more relevant to the process so data from this direction was used
to calibrate the model.

The elastic modulus of 5 different grades of small heat-shrink tube was calculated using data
from the pressure test. These values are used in Chapter 6 to model the shrink force test.
Due to the relatively small effects of viscoelasticity, plasticity and anisotropy it was decided
that the SMP could be modelled using a Neo-Hookean hyperelastic SMP model with the
modulus derived from the pressure test.

61
5 Thermal modelling of the laser welding of balloon
catheters

The purpose of the thermal model is to predict the temperatures throughout the balloon
catheter proximal joint assembly (shown in Figure 1.4) during the laser bonding process of
the PE balloon to the catheter shaft. Obtaining an understanding of the thermal profile
throughout the joint assembly is essential if the thermo-mechanical behaviour of the PE
material during welding is to be analysed. Preliminary thermal modelling of the balloon
catheter joint assembly was completed by engineers at Boston Scientific [93]. However, it
was deemed necessary to validate this work in order to gain a full understanding of the
thermal model and to be able to apply the thermal methods to different assembly
geometries. The thermal model consists of two parts, modelling the laser absorption and
modelling the heat flow through the assembly. These two aspects are discussed below in
sections 5.1 and 5.2 respectively.

5.1 Laser thermal absorption

When modelling laser welding, the first thing to consider is the heat source and its
representation within the model. The diagram below (Figure 5.1) shows the laser beam
interacting with the assembly. The assembly is rotating at approximately 500 rpm so the heat
generation calculated below is averaged circumferentially and assumed to be axi-symmetric.
The incoming continuous laser radiation source is represented within the model by dividing
it up into approximately 200 beams. Similarly, the cross section of the catheter assembly is
divided into approximately 500 annuli. Each beam is tracked through the assembly and the
volumetric heat source due to the absorption of the laser light, 𝑞̇ , is calculated by taking the
magnitude of the derivative of Equation 2.1 with respect to depth, z [20].

𝑞̇ = 𝛼𝐼0 𝑒 −∝𝑧

The laser beam is refracted at the surface of the assembly due to the change in refractive
index. The angle of refraction is calculated using Snell’s law [90]:

𝑛1 𝑆𝑖𝑛(𝜃1 ) = 𝑛2 𝑆𝑖𝑛(𝜃2 ) (5.1)

62
 CHAPTER 5

where n1 is the refractive index of air, n2 is the refractive index of the heat-shrink, θ1 is the
angle between the ray and the surface normal outside the assembly and θ2 is the angle
between the ray and the surface normal inside the assembly. Reflection at the outer surface
is accounted for using Fresnel’s equation [91]:

𝑛 𝑐𝑜𝑠𝜃 −𝑛 𝑐𝑜𝑠𝜃 2
𝐼 = 𝐼0 [1 − (𝑛2 𝑐𝑜𝑠𝜃1 +𝑛1 𝑐𝑜𝑠𝜃2 ) ] (5.2)
2 1 1 2

Where I is the intensity of the light that is transmitted and I0 is the intensity of the incident
beam of light.

Reflection and refraction between layers in the assembly is neglected in this model.
Reflection and absorption of the laser by the mandrel is also neglected (i.e. all light which
reaches the mandrel is assumed to be lost).

Laser Source

Figure 5.1 Laser interacting with the balloon catheter assembly shown in a cross sectional profile. The dark line
between the tubing and the balloon represents the area to be welded. The assembly is rotated to get an even
weld around the joint and the heat source is averaged circumferentially to account for this rotation.

63
 CHAPTER 5

This technique is known as ray tracing and the results of this are shown below in Figure 5.2.
The results are compared with previous Boston Scientific work. The engineers in Boston
Scientific used MATLAB to implement the ray tracing technique from first principles. They
then used a commercial ray tracing software package called TRACEPRO to verify their work.

8.0E+08

7.0E+08
Heat Generation rate (W/m^3)

6.0E+08 H.G.
5.0E+08
Previous BSC
4.0E+08 data

3.0E+08

2.0E+08

1.0E+08

0.0E+00
0 200 400 600 800 1000
Radial Position (µm)
Figure 5.2 Graph of Heat generation as a function of radial distance from the centre of the assembly. The line
labelled H.G. is the data obtained from the ray tracing technique while the data points represent results from
previous work by Boston Scientific.

These results have been averaged around the circumference of the assembly to obtain an
axisymmetric heat generation rate which takes into account the rotation of the assembly
[94]. There is very good agreement between the results obtained from this ray tracing
technique and the results of Boston Scientific. An equation was then fitted to this graph so
that it could be included in a thermal model.

5.2 Heat transfer model

A transient thermal model was created in the commercial finite element code Abaqus FEA
using the balloon catheter joint assembly geometry and the laser parameters used by Boston
Scientific in the laser welding process. The model is assumed to be axi-symmetric. The
geometry is axisymmetric and the heat generation has been averaged circumferentially to
provide an axi-symmetric loading. Heat generation is calculated using laser thermal
absorption data calculated using the ray-tracing method shown in Figure 5.2 above and the
heat losses of free convection and radiation are implemented. Forced convection due to the

64
 CHAPTER 5

rotation of the balloon catheter joint assembly was neglected. It is worth noting that the
mandrel conducts a large amount of heat away from the welded area due to the high thermal
conductivity of stainless steel and the relatively long length of the mandrel (approx. 1 m).
This is represented within the model. The entire assembly is represented as a single part,
hence neglecting contact effects and air gaps between the assembly parts. This single part is
then partitioned into different sections which are assigned the various material properties
(see Table 5.1 below). It is noted that the conductivities of the PEBAX® and heat-shrink vary
with temperature.

Material Density Specific Heat Conductivity Tmelt Latent heat α

Capacity of fusion
(kg m-3) (W m-1 K-1) (°C) (m-1)
(J kg-1 K-1) (J kg-1)

PEBAX® 1020 2520 For T < Tm 172 54428 18898

=0.336-0.0012*T

For T > Tm

=0.1296

Stainless 8000 500 16.2 - - -

Steel

Heat- 1020 2200 For T < 100°C - - 3543

shrink
=0.35-0.0011*T

For T > 100°C

=0.24

Table 5.1: List of material properties of the various materials used in the balloon catheter joint assembly.

The total heat flux out of the surface of the assembly is the sum of the convective and
radiative heat fluxes:

4
𝑞 = ℎ(𝑇𝑊 − 𝑇∞ ) + Ω𝜀(𝑇𝑊 − 𝑇∞4 ) (5.3)

Where q is the total heat flux, h is the heat transfer coefficient, 𝑇𝑊 is the surface temperature
of the assembly, 𝑇∞ is the external temperature, 𝜀 is the emissivity of the surface of the
assembly and Ω is the Stefan-Boltzmann constant [93]. The heat transfer coefficient was
calculated from various empirical relationships over an appropriate range of the Rayleigh

65
 CHAPTER 5

number [95] and its’ value was determined to be 30 W/m2/K. Boston Scientific determine
the value of the emissivity, 𝜀, through available experimental data to be 0.95 [93].

The assembly represented within the finite element model is assumed to have an initial
temperature of 20°C. The elements used were quadratic heat transfer elements and there
were 775 elements in the model. The heat input was implemented using the subroutine
HETVAL within the Abaqus FEA finite element code. This enables the creation of a spatially
varying volumetric heat source. Figure 5.3 below shows the meshed assembly, the mesh is
finer in the heated area because the thermal gradients are highest in this region. This mesh
is quite coarse; however, good agreement with previous work performed by Boston Scientific
[93] is achieved.

Figure 5.3 The meshed geometry used to represent the balloon catheter joint assembly. The heat-shrink outer
diameter is 1.6 mm. The mandrel radius is 0.36 mm. The outer thickness is 0.062 mm. The balloon waist
thickness is 0.046 mm.

As in all modelling approaches, some assumptions have been made and some phenomena
have been neglected. The effect of the air in the balloon heating up has been neglected. The
mandrel is coated in Teflon to stop the PEBAX® sticking to it. The effect of this Teflon coating
on the thermal properties has also been neglected. The intensity of the laser beam is
assumed to be uniform across the width of the beam which is 1.5 mm. It is also assumed that
the laser absorption in independent of temperature. In addition, as this is a thermal model
only, no material deformation and melt flow is represented within the model. This is
examined in detail in Chapter 7.

66
 CHAPTER 5

Figure 5.4 below compares the surface temperature (after 20 s) prediction of the model with
that of a Boston Scientific model. It can be seen that the models are in good agreement.

250

200
Previous BSC
work
Abaqus data
Temperature (°C)

150

100

50

0
-6 -4 -2 0 2 4 6
Axial distance from centre of laser (mm)

Figure 5.4 This graph shows the temperature variation along the surface of the heat-shrink.

Figure 5.5 shows how the temperature at the heat-shrink surface varies over time.
Simulations were run using various time steps, the influence of changing from 20 time steps
to 210 was quite small and any more time steps produce virtually no change.

67
 CHAPTER 5

250

Temperature (°C) 200

150
Previous BSC work
210 steps
100
20 steps

50

0
0 5 10 15 20 25
Time (s)
Figure 5.5 Variation of temperature of heat-shrink surface at centre of laser beam.

The results obtained from these simulations show that the temperatures are a little higher
during the welding process than in the previous work. The difference between the results
may be due to the time steps or the mesh density but is broadly in good agreement with
previous work.

The engineers in Boston Scientific have compared their results to experiment by using
polarized microscopy. This technique allows the shape of the melt region to be determined
and this can be compared with the melt region predicted from the model. There is good
agreement between the melt pool shape predicted by the Boston Scientific model and the
melt pool region observed on a welded catheter so this model is compared against the
Boston Scientific model and not directly with an experiment.

5.3 Modelling of the SterlingTM balloon catheter joint assembly

The Sterling™ balloon catheter is the catheter geometry examined in this work. More
specifically, the proximal bond of a catheter with a 3 mm balloon is represented within the
computational models of this work and in those of the thermo-mechanical investigation of
Chapter 7. There are a number of differences between the thermal model of the Sterling™
catheter and the thermal model in Section 5.2 including a different geometry, a moving laser

68
 CHAPTER 5

and a temperature feedback to control the laser power. These differences and their effects
are described here.

The first difference is the geometry. In the Sterling™ catheter, the balloon is inside the outer
tube as shown in Figure 5.6 and there is an overlap of 1.75 mm. Also, the end of the outer
tube is within the laser beam and this has to be taken into account when calculating the laser
absorption. In the thermal model in Section 5.2 the heat generation by laser absorption is
constant across the width of the laser beam but this is not the case for the Sterling™ model.
The right-hand-side of the laser passes through the outer but the left-hand-side does not as
shown in Figure 5.7. This results in two different heat generation curves, which are shown in
Figure 5.8.

Heat-shrink Outer tube

Balloon Mandrel

Mandrel radius: 400 µm

Teflon coating thickness: 6.35 µm
Balloon waist thickness: 46.4 µm
Balloon body wall thickness: 16 µm
Balloon body O.D.: 3.12 mm
Outer tube thickness: 114 µm
Heat-shrink O.D: 1967 µm

Figure 5.6 Representation of the model geometry for the Proximal joint assembly of the Sterling™ catheter.

69
 CHAPTER 5

x<0 x=0 x>0

Figure 5.7(A-B) Movement of laser during welding. The laser is stationary for the first 10 seconds of the
process in the position shown in A. It then moves 0.6 mm to the left over 8 seconds and finishes in the
position shown in B.

Another difference with the Sterling™ catheter model and the previous model, discussed in
Section 5.2, is that the laser is stationary in the previous model but in the Sterling™ catheter
model it moves. The laser is shown as the red region in Figure 5.7. The laser is stationary for
10 s in the position shown in Figure 5.7 A and then it moves 0.6 mm to the left over 8 s and
finishes in the position shown in Figure 5.7 B.

2.50E+07
Start of outer tube

2.00E+07
Start of balloon
Heat Generation (W/m^3)

Start of heat-
1.50E+07 shrink

1.00E+07

X>0
5.00E+06
X<0

0.00E+00
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1
Radial position (mm)
Figure 5.8 Heat generation as a function of radial position. X is the axial position shown in Figure 5.7. When X
< 0 (to the left of the end of the outer tube) the laser does not pass through the outer tube and when X > 0 it
does. Therefore, two heat generation curves are required.

70
 CHAPTER 5

The heat generation as a function of radial position is shown in Figure 5.8. There are two
heat generation curves because some of the laser passes through the outer tube and some
of it does not. These heat generation curves provide the thermal loading for the model. As
in the previous model, heat losses through free convection and radiation are implemented,
but forced convection due to the rotation of the assembly was neglected. The model consists
of four parts and there is assumed to be perfect conduction between the parts when they
are in contact therefore there is no contact resistance. This is achieved by setting the heat
transfer coefficient to a very large number, in this case 1x109 W/(m2K). This is a possible
limitation but as no data was available and the thermal model predictions were reasonably
accurate it was deemed acceptable. In reality there will be a thermal contact resistance that
decreases as the pressure increases. This thermal resistance would slightly reduce
conduction between the different layers. In this case that would mean the heat-shrink would
be slightly hotter and the PEBAX® slightly cooler.

The mesh used for the model is shown in Figure 5.9. The model used approximately 6,000
linear heat transfer axisymmetric elements (DCAX3 and DCAX4).

Heat-shrink Outer tube

Balloon Mandrel

Figure 5.9 Geometry and mesh for Sterling™ thermal model.

The material properties used in this model are summarised in Table 5.2 below.

71
 CHAPTER 5

Material ρ cP k TMELT ΔHFUS α

(kg/m3) (J/kg/K) (W/m/K) (°C) (J/kg) (1/m)
Mandrel 8000 477 + 0.2267*T 14.4 + 0.015 * T - - -
Teflon 2160 630+1.258*(T+273) 0.25 - - -
162-
PEBAX® 1020 -187+6.12*(T+273) 0.2 43000 18898
172
Heat- 0.35 − 0.00104 ∗ 𝑇 𝑖𝑓 0 < 𝑇 < 106
920 309+5.09*(T+273) ={ 50-102 100500 3543
shrink 0.24 𝑖𝑓 𝑇 ≥ 106

Table 5.2 Material properties used for SterlingTM thermal model. ρ is the density, cP is the specific heat capacity,
k is the conductivity, TMELT is the melt temperature, HFUS is the latent heat of fusion and α is the absorption
coefficient. The temperature, T, is measured in °C.

During the catheter assembly process at Boston Scientific, the Sterling™ catheter bond laser
welding process uses a feedback loop to control the temperature. An infra-red camera is
used to monitor the surface temperature of the heat-shrink during the laser welding process
and a PID (proportional-integral-derivative) control algorithm is used to control the laser
power. The power of the laser is adjusted to keep the temperature at a predetermined value.

Boston Scientific have modelled this PID control algorithm in detail and their results agree
very well with experimental measurements [93]. In this work the PID control algorithm has
not been implemented into the computational finite element models. Instead, the input
power to the laser (laser drive) has been approximated (from Boston Scientific’s model) as a
series of discrete steps. Figure 5.10 below shows the laser drive used by Boston Scientific
and the approximated one used in this work. The integral of the two curves is also plotted to
ensure the same amount of energy is going into both.

72
 CHAPTER 5

400 70

350 60
Area Under Drive curve (%*sec)
300
50
250
Integral of BSc Drive curve 40

% Drive
200 Integral of approx Drive
Drive 30
150 Drive approx
20
100

50 10

0 0
0 5 10 15 20 25
Time (sec)
Figure 5.10 Input drive for the laser during the SterlingTM catheter laser welding process. A curve (drive
approx) was fitted to Boston Scientific data [93] (drive). The integral of both curves was plotted to check that
the same amount of energy is going into both.

The output power of the laser as a function of the input drive is shown below in Figure
5.11.

1
0.9
0.8
Normalized Output Power

0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
0 20 40 60 80 100
% Drive
Figure 5.11 Normalized laser output power versus % input drive

The heat generation obtained from Figure 5.8 was multiplied by the normalized output
power to obtain the heat generation through the assembly as the input drive varied. The

73
 CHAPTER 5

thermal aspect of the welding process was then modelled using the finite element model
shown in Figure 5.8 and the results are shown in Figure 5.12 and 5.13 below.

Figure 5.12 Temperature distribution throughout the assembly at the end of the welding process.

Figure 5.12 above shows the temperature distribution predicted by the model at the end of
the welding process just before the laser is turned off. The surface temperature predicted by
the model is compared with the Boston Scientific model in Figure 5.13 below. The
temperature is outputted for 0.5 mm from the centre each side of the laser beam and is
averaged over this distance.

74
 CHAPTER 5

250

200
Abaqus

BSC data
Temperature (°C)

150

100

50

0
0 5 10 15 20 25 30
Time (s)
Figure 5.13 Temperature profile at the surface of the heat-shrink (averaged over 1 mm) plotted against time.
Comparison of Abaqus model and Boston Scientific model.

The temperature shown in Figure 5.13 is the surface temperature averaged over 1 mm (0.5
mm each side of the centre of the laser). The temperature ramps up to 215°C over the first
6 s and is then kept constant by the PID control algorithm which control the laser power. As
can be seen in Figure 5.13, the surface temperatures of the heat-shrink of the two models
agree very well.

5.4 Comparison of computational model with experimental data

The results of the computational thermal model of the laser welding process, described in
Section 5.3 were compared to the surface temperature profile of the heat-shrink, as shown
in Figure 5.13. However, the surface temperature is only one means of determining the
accuracy of the temperature profile within the bonding PE layers itself. As direct
experimental temperature measurements through the thickness of the bond were not
possible, the thermal model predictions were compared to optical analysing images of the
welded bond geometry after the laser welding process. The images were obtained through
the use of polarized microscopy and OCT images (as described in Chapter 3). Figure 5.14
below shows an image of the proximal bond taken of the SterlingTM catheter following the

75
 CHAPTER 5

welding process. The PE balloon is shown on the left and the outer PE tube is shown on the
right. This image only shows the surface of the bond. In order to get more detailed
information about the shape of the melt pool, it is necessary to cut a section out of the
bonded area. Figure 5.15 below is a polarized microscopy image of a Sterling™ proximal bond
which has been cut along the axial direction.

Balloon Outer tube

Proximal bond

Figure 5.14 OCT image of the proximal bond of the Sterling catheter following the laser welding process. The
dashed box shows where a section of the bond was removed for polarized microscopy to determine the
extent of the melt pool, as shown in Figure 5.15.

Bonded region Melt pool Non-melted

region Outer
tube

End of outer End of outer

after welding before welding Edge of melt pool Balloon
Melt flow length Non-bonded
1 mm region
| |
Figure 5.15 Polarized microscopy image of the proximal bond of the SterlingTM catheter following welding.

As outlined in Section 3.4.1, polarized microscopy produces images with a good contrast
between the relatively dark amorphous phase and the relatively bright crystalline phase. In
Figure 5.15 above, the slightly darker region is assumed to be the melt pool which became
amorphous after melting and the lighter region to the right is assumed to be the non-melted
region which retained its anisotropic crystallinity from the extrusion process. The boundary
between these two regions is clearly visible due to the different crystal structures. It is
assumed that this boundary represents the edge of the melt pool and that the maximum

76
 CHAPTER 5

temperature reached at this boundary was approximately 172°C, which is the melting point
of the PEBAX®. The melt flow region to the left of the image is also clearly visible. The non-
bonded region to the lower-right of the image is part of the balloon which has been
separated from the outer tube. This region has separated from the outer tube as a direct
result of removing the sample from the catheter and from the sample preparation for
imaging.

The bonded region consists of two distinct zones. In the bonded area on the left of the image,
the boundary between the outer tube and the balloon is no longer visible due to the melting
and mixing of the materials. The bonded area to the right has not melted and the boundary
between the outer tube and the balloon is still visible. However, this area is still bonded due
to the “autohesion” between polymers at high temperatures. Autohesion occurs when two
surfaces of the same elastomer bond when pressed together. It is caused by the diffusion of
polymer chain ends across the interface of the two surfaces [92]. The autohesive bond
strength increases with time, temperature and pressure [92].

The image shown in Figure 5.16 below shows the results of an OCT image of a Sterling™
proximal bond. As discussed in Chapter 3, this imaging technique has the advantages of being
non-destructive and there is no sample preparation which can affect the results. The
disadvantages of this technique are that the image is not as clear as the polarized microscopy
images and the dimensions in the vertical direction are affected by refraction effects and are
not considered to be accurate.

Melt pool Non-melted

Outer tube region
Balloon

Edge of melt pool

1 mm

Figure 5.16 OCT image of proximal bond for SterlingTM catheter following the welding process.

The contour plot below in Figure 5.17 shows the predicted area of the outer tube that
exceeded the melting temperature of 172°C during the computational modelling of the

77
 CHAPTER 5

welding process. The length of the melt pool predicted by the model is 1.0 mm. This is
compared with the experimental result in Figure 5.15. The length of the melt pool measured
in the experiments is approximately 1.6 mm, which includes the melt flow region which was
not represented within the thermal model in this work.

Melted region Non-melted region

(Above 172°C) (Below 172°C)

1 mm
Figure 5.17 The melt pool predicted by model. The red region represents the area that was heated to
temperatures above 172°C during the simulation of the welding process and the gray area represents
material with a predicted temperature below 172°C.

In the experimental image the melt pool at the edge of the outer tube encompasses almost
all of the outer tube and comes within approximately 0.02 mm of the balloon. The melt pool
predicted by the model comes within approximately 0.005 mm of the balloon.

Figure 5.18 below shows a graph of the maximum temperature reached at the interface of
the balloon and the outer tube. At the edge of the outer tube (x=0) the temperature reaches
approximately 170°C which is very close to the melting temperature of 172°C. In Figure 5.15
above, the axial position of the right-hand edge of the bonded region is approximately 0.83
mm from the initial edge of the outer tube (x=0). According to the model, the maximum

78
 CHAPTER 5

temperature reached at this point (x=0.83 mm) is 153°C. This implies that the minimum
temperature needed to bond the PEBAX® is 153°C.

190

180

170
Maximum Interface Temp (°C)

160 melt
temperature

150
End of
Balloon
140
Edge of
130 outer tube

120

110

100
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
Axial Position (mm)

Figure 5.18 Temperature at interface of balloon and outer for model of Sterling TM process. The region from x
= 0 to x = 1.75 mm represents the area where the outer tube overlaps the balloon.

5.5 Discussion

Thermal models were created for two separate laser bonding processes. The first model was
a preliminary investigation which involved reproducing the results of previous work by
engineers at Boston Scientific. The second thermal model was for the more complicated
Sterling™ welding process. Both models accurately reproduced the work previously
undertaking by Boston Scientific.

There are two methods available to check the accuracy of the model. The first is to compare
the melt pool length in the experiment with that predicted by the model. The second is to
compare the melt pool shape in the experiment with that predicted by the model. The melt
pool length predicted by the Sterling™ model is 1.0 mm while the observed melt flow length

79
 CHAPTER 5

is approximately 1.6 mm. One major reason the model under-predicts the size of the melt
pool is that there is no melt flow in this model.

To compare the melt pool shape the model data in Figure 5.17 is compared with the
experimental image in Figure 5.15. From these it can be seen that the overall shape of the
melt pool is similar. Also the melt pool depths can be compared. The melt pool depth
predicted by the model is approximately 0.015 mm larger than the melt pool depth seen in
the experiment. This is quite a small difference and gives confidence that the temperatures
predicted by the model are reasonably accurate.

Overall these models predict the temperature throughout the assembly quite accurately.
This is essential for the future thermomechanical models involving melt flow (Chapter 7) as
the properties of the materials vary significantly with temperature.

80
6 Implementation of thermo-mechanical model of shape
memory polymer

6.1 Introduction

The purpose of this chapter is to explain in detail two of the mathematical models of shape
memory behaviour reviewed in Chapter 2, and describe how they are implemented in the
finite element code of Abaqus FEA. These models are used to model the behaviour of the
heat-shrink tubing and therefore are an important part of this work.

Two shape memory polymer (SMP) models were implemented in Abaqus FEA. The first is the
stored strain model of Liu et al. [64] and the second is the multiple natural configurations
model of Barot et al. [77]. These two models were introduced in the literature review of
Chapter 2 and are described in more detail here along with their implementation into the
Abaqus code via a User Material (UMAT) subroutine.

6.2 Review of stored strain model of Liu et al.

Liu et al. [64] assumes that the polymer is a mixture of two kinds of extreme phases:
The “frozen phase” and the “active phase”. The frozen phase consists of relatively stiff
polymer chains which resist deformation and result in a high elastic modulus. As the material
heats up these polymer chains become more flexible and cannot resist deformation,
resulting in a much lower elastic modulus. This compliant phase is called the active phase.
The volume fraction of each of the phase depends only on the temperature, while the
thermomechanical properties of each phase stay the same.

This model uses the concept of “stored strain” to describe the shape memory effect. To
incorporate the shape memory effect into a material the material is heated, stretched and
then cooled. The strain that the material was stretched to is “stored” during cooling and
released upon reheating.

This model assumes that the stress at a given point is the same for the frozen phase and the
active phase. The strain in the material is given by a rule of mixtures addition of the strains
in the active and frozen phases,

81
 CHAPTER 6

𝛆 = (1 − 𝜑𝐹 )𝛆A + 𝜑𝐹 𝛆F (6.1)

where ϕf is the volume fraction of the frozen phase and εA, εF, and ε are the strains in the
active phase, the frozen phase and the overall material respectively.

Figure 6.1 This diagram shows a polymer in an intermediate state with a predominant active phase.

This model was developed within a small-strain framework and the constitutive equation is

σ = C (ε - εs - εT ) (6.2)

where σ is the Cauchy stress matrix, C is the stiffness matrix, ε is the strain, εS is the stored
strain and εT is the thermal strain. The stiffness matrix used here is the same as that used for
temperature-dependent linear elasticity, where the Young’s modulus, E, depends only on
temperature (not on strain). Liu et al. give an equation which relates E to the frozen fraction.
1
E
F 1  F (6.3)

EF EA
Where EF and EA are the moduli of the frozen phase and active phase respectively and are

assumed to be constant and ϕf is the frozen fraction which can be calculated from:

1
F  1  (6.4)
1  c f Th  T 
n

Cf and n are experimentally determined constants.

Liu et al. derive an expression for εS in terms of the strain in the frozen phase and volume
fraction. Using this expression for εS and the constitutive equation above they give a
differential equation which describes how the stored strain varies with temperature. In 1-D
this equation is

82
 CHAPTER 6

d S    S  T  d F 
   (6.5)
dT   F 1   F   dT 
E A   
 EF E A 
This can be solved to give an expression for εS. If we assume that during cooling the strain is
held at a constant value (called εpre) this expression is

  EF



  EF  E A  E F  
 S   pre 1     (6.6)
  A  E  E 
F F  E F  
 

When EF is much greater than EA then this expression for εS is approximately the same as

 S   preF (6.7)

which is easier to interpret.

6.3 The multiple natural configurations model of Barot et al.

Barot et al. [77] also employed a two-phase approach in their multiple natural configurations
model for SMPs. The rigid frozen phase dominates at low temperatures and the relatively
soft rubbery phase dominates at high temperatures. However, this model uses a finite-strain
formulation so it is better suited to modelling the large deformations which typically arise in
SMP applications. Another difference between this model and the stored strain model
described in Section 6.2 is that in this model both phases have the same deformation
gradient but the stress in each phase is different.

The multiple natural configurations model assumes that as the material freezes, the
deformed configuration of the material becomes the stress-free natural configuration of the
frozen phase. Therefore, the natural configuration (stress-free state) of the frozen phase can
be changed by heating, stretching and cooling, but the natural configuration of the active
phase remains the same. The changing of the natural configuration is how this model
captures the shape memory behaviour.

The stress is given by a rule of mixtures based on the volume fraction of each phase

𝛔 = (1 − 𝜑)𝛔𝐀 + 𝜑𝛔𝐅 (6.8)

83
 CHAPTER 6

where σF and σA are the stresses in the frozen and active phases respectively and φ is the
volume fraction of the frozen phase. The two phases are assumed to be Neo-Hookean. In the
Neo-Hookean model the stresses are computed from the deformation gradient F. In the
Barot et al model the active and the frozen phases have different deformation gradients FA
and FC respectively. When these are known the stress in each phase can be calculated and
then summed using the rule of mixtures above to give the overall stress in the material.

Figure 6.2 below shows a graphical representation of the various deformation gradients
during a shape memory cycle.

Reference B
configuration
F = FA
for active
phase
A
Reference
F = FDEF configuration
F= for frozen
deformation C phase
FC
gradient at F
any arbitrary
time
FDEF = deformation
D
F = FC.Fdef gradient of
-1
temporary shape
FC = F.(Fdef)

Figure 6.2 Diagram showing the deformation gradients used in the Barot et al. model of SMPs.

At state A, the material is in the active state and in its reference configuration (i.e. zero stress,
zero strain). When the material is stretched to point B the deformation of the active phase
results in a stress but the frozen phase does contribute to the stress as all the material is in
the active phase. The deformation gradient of the active phase FA is always the same as the
deformation gradient of the overall material, F.

When the material is cooled (point C) it forms a new reference configuration for the frozen
phase. At this point the deformation gradient of the frozen phase is equal to the identity
matrix, 𝐅𝐂 = 𝐈 (i.e. there is zero strain in the frozen phase) and the deformation of the active
phase is FDEF, the deformation gradient of the temporary shape. As the material is all in the
frozen phase, the stress contribution from the active phase is zero, and as the strain in the

84
 CHAPTER 6

frozen phase is zero, the overall stress in the material is zero. Therefore the material will
remain in this temporary shape until it is reheated.

Upon reheating the frozen phase begins to be replaced by the active phase and the stress
contribution from the active phase becomes significant causing the material to contract. At
point D the material is in an intermediate state between frozen and active. The deformation
gradient of the frozen phase, FC, is calculated from

𝐅𝐂 = 𝐅. (𝐅𝐃𝐄𝐅 )−𝟏 (6.9)

FC is then used to calculate the stress contribution from the frozen phase. The stress in the
frozen phase opposes the stress in the active phase which is trying to return the material to
its original shape (point A). Therefore the original shape is not recovered until all of the
material is in the active state.

6.4 Implementation of SMP models

The two SMP models were implemented in Abaqus FEA using a User Material (UMAT)
subroutine. Abaqus passes the strain (and many other variables) at the start of the increment
into the UMAT. The UMAT is required to output the updated stresses and the material
Jacobian matrix, ∆𝜕𝝈/∆𝜕𝜺. Abaqus then updates the strains based on the updated stress
and Jacobian provided by the UMAT the updated strain is then passed into the UMAT for the
next increment.

For both of these UMATs the Jacobian matrix is updated first and then this is used to update
the stresses.

6.4.1 Implementation of stored strain model

The constitutive equation for the small strain model is

σ = C (ε - εs - εT ) (6.10)

This equation was implemented in incremental form. In rate form the stress is updated by
calculating the change in stress for a given increment and adding it to the previous stress.

85
 CHAPTER 6

 new   old   (6.11)

Using index notation, the incremental form of the constitutive equation for 3 dimensions is

 ij   ij  kk
el
 2 ijel   ij  kk
el
 2 ijel (6.12)

Where 𝛿𝑖𝑗 is the Kronecker delta and where the elastic strain is given by

 ijel   ij   Sij  T ij (6.13)

The Jacobian matrix, 𝜕∆𝝈/𝜕∆𝜺, for this model is the same as that for simple 3-D elasticity

   2   0 0 0
 
    2  0 0 0
      2 0 0 0
  (6.14)
  0 0 0  0 0
 0 0 0 0  0
 
 0 0 0 0 0  

where λ and µ are Lamé’s constants.

λ and µ are related to the Young’s modulus, E, through

E
 (6.15)
1   1  2 

and
E
 (6.16)
21  
The Young’s modulus is a function of temperature (equation 6.3) therefore the Jacobian
matrix varies with temperature. Once λ and µ are determined for the new step the Jacobian
can be updated.

6.4.2 Implementation of multiple natural configurations model

The Neo-Hookean material model is one of the simplest hyperelastic models and its strain
energy density, Ψ, is defined as follows

86
 CHAPTER 6

𝜇 ∗ 𝜅
𝛹= (𝛪1 − 3) + (𝐽 − 1)2 (6.17)
2 2

where μ is the shear modulus and κ is the bulk modulus. J is the determinant of the
deformation gradient F, and is given by 𝐽 = 𝐷𝑒𝑡(𝐅) = 𝜆1 𝜆2 𝜆3 where λi are the principal
stretches. 𝛪1∗ is the first invariant of the left Cauchy-Green deformation tensor B (B=FFT) and
is given by

−2⁄
𝛪1∗ = 𝐽 2
3 (𝜆1 + 𝜆22 + 𝜆23 ) (6.18)

The stress is given by

𝜇
𝛔= 𝑑𝑒𝑣(𝐁∗ ) + 𝜅(𝐽 − 1)𝚰 (6.19)
𝐽

where 𝚰 is the identity matrix and dev(B) is the deviatoric part of B, that is

1
𝑑𝑒𝑣(𝐁∗ ) = 𝐁∗ − 𝛪1∗ 𝚰 (6.20)
3

and

−2⁄
𝐁∗ = 𝐽 3𝐁 (6.21)

Once the deformation gradients for both phases have been calculated (as described in
Section 6.3) these equations are used to calculate the stresses in each phase separately and
these stresses are summed using Equation 6.8 to give the overall stress. The volume fraction
was calculated using the same method as the Liu et al model as shown in Equation 6.4.

Thermal expansion was also included in the UMAT using the equation

𝐅 = 𝐅E 𝐅TH (6.22)

where F is the overall deformation gradient, FE is the deformation gradient due to elasticity
and FTH is the thermal deformation gradient given by

1 + 𝛼𝛥𝑇 0 0
𝐅𝑇𝐻 = ( 0 1 + 𝛼𝛥𝑇 0 ) (6.23)
0 0 1 + 𝛼𝛥𝑇

Therefore when calculating the elastic strains we need to use FE. This can be calculated from

𝐅E = 𝐅(𝐅TH )−1 (6.24)

87
 CHAPTER 6

FTH is calculated at the beginning of the increment, and then its inverse is multiplied by the
deformation gradient provided by Abaqus to give the elastic deformation gradient. This
elastic deformation gradient is then used in the subsequent calculation of stress.

6.5 Verification

Verification involves comparing the results of the UMAT with an analytical solution. This is
done to ensure there are no errors in the finite element implementation of the constitutive
model.

6.5.1 Verification of implementation of stored strain model

To test the UMAT of the stored strain model, a single-element 3-D model was created in
Abaqus FEA. This element was then put through a shape-memory cycle of heating, stretching,
cooling and reheating. The stretching was performed by applying a uniaxial strain, and the
resulting stress was compared to an analytical solution generated using the relevant
equations in a spreadsheet. The analytical solution to the small-strain formulation can be
calculated easily from the equations above in Section 6.2. The results below show the
imposed temperature and the resulting stress as a function of time during a shape memory
cycle.

88
 CHAPTER 6

350000 Analytical Stress 120

UMAT Stress
300000
100
Temperature
250000
80

Temperature (°C)
Stress (Pa)

200000
60
150000

40
100000

20
50000

0 0
0 2 4 6 8 10 12 14 16
Time (s)
Figure 6.3 This graph shows the stress during a shape memory cycle for two models, an analytical model and a
finite element model using the UMAT. Over the first two seconds the sample is heated then it is stretched.
After 4 s the temperature is reduced while the strain is kept constant. The stress goes to zero as the frozen
fraction increases. When the sample is reheated (after 10 s) the stress is recovered as the active phase
increases. Thermal expansion is not included in this test.

Figure 6.3 above shows that the UMAT and the analytical solution agree well. This gives
confidence that the UMAT has been coded correctly.

6.5.2 Verification of implementation of multiple natural configurations model

The analytical solution for the multiple natural configurations model is more complicated
than that for the stored strain model due to the finite deformation. This was also

implemented in a spreadsheet. For uniaxial tension 𝜆1 = 𝜆 and 𝜆2 = 𝜆3 = √𝐽/𝜆 and the

stress is given by

2𝜇 2
𝐽
𝜎11 = 5⁄ (𝜆 − ) + 𝜅(𝐽 − 1) (6.25)
3𝐽 3 𝜆

To solve this equation it is necessary to know the value of J. To find an equation for J the
stress in the lateral direction (𝜎22 = 𝜎33 ) is set at zero.

89
 CHAPTER 6

𝜇 𝐽
𝜎22 = 𝜎33 = 5 ( − 𝜆2 ) + 𝜅(𝐽 − 1) = 0 (6.26)
3𝐽 ⁄3 𝜆

This equation can then be solved numerically to provide a value for J which in turn can be
used to calculate the stress and the lateral stretch.

0
displacement in y direction (arbitrary units)

-0.05

-0.1

-0.15 Analytical

Abaqus
-0.2

-0.25

-0.3

-0.35
0 0.5 1 1.5 2 2.5
Change in length in x direction (arbritary units)
Figure 6.4 Comparison of lateral displacement for 2 models of a uniaxial tension test. One model is analytical
and one is a finite element model using the built-in Abaqus model for Neo-Hookean hyperelasticity.

90
 CHAPTER 6

1.8

1.6

1.4

1.2
Stress (MPa)

1 Theory
Abaqus
0.8

0.6

0.4

0.2

0
0 0.5 1 1.5 2 2.5
Change in length (arbritary units)

Figure 6.5 Comparison of stress for two models of a uniaxial tension test. One model is analytical and one is a
finite element model using the built-in Abaqus model for Neo-Hookean hyperelasticity. Poisson’s ratio in this
test is 0.4.

Figures 6.4 and 6.5 above compare results from a single-element uniaxial Abaqus simulation
with the analytical solution. These tests were performed to ensure the analytical model of
Neo-Hookean hyperelasticity was implemented correctly. There is no shape-memory effect
in these models.

The results of two models of a uniaxial test, conducted to very large strains and using a near
incompressible Poisson’s ratio (0.49999), are shown below in Figure 6.6.

91
 CHAPTER 6

14

12

10
Stress (MPa)

8
Analytical
6 Abaqus

0
0 1 2 3 4 5 6 7
Stretch
Figure 6.6 Comparison of stress for two models of a uniaxial tension test. One model is analytical and one is a
finite element model using the built-in Abaqus model for Neo-Hookean hyperelasticity. Poisson’s ratio in this
test is 0.49999.

The analytical results are compared with the built-in Abaqus FEA model for Neo-Hookean
hyperelasticity to ensure that the analytical model is solving the Neo-Hookean equations
correctly. These results all agree well so this means the analytical solution has been
implemented correctly. The analytical solution can now be used to verify that the UMAT has
been implemented correctly.

Figures 6.7 and 6.8 below show how the stress varies during a shape memory cycle according
to the model. These graphs compare the analytical solution with the finite element solution.
Both of these graphs represent a sample going through a shape memory cycle of being
heated, then stretched, then cooled, then reheated. Figure 6.8 includes the effect of thermal
expansion.

92
 CHAPTER 6

2 100
Abaqus S33
1.8 Analytical 90

1.6 Temperature 80

1.4 70

Temperature (°C)
1.2 60
Stress

1 50

0.8 40

0.6 30

0.4 20

0.2 10

0 0
0 1 2 3 4 5
Time (s)
Figure 6.7 Stress during a shape memory cycle. Sample starts at a high temperature is stretched then cooled.
As the sample is cooled the stress drops. When the sample is reheated (after 3 s) the stress is recovered. An
arbitrary stiffness was used.

2.5
Abaqus S33
2
Theory
1.5
Stress

0.5

0
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5
-0.5

-1
Time (s)

Figure 6.8 Stress during shape memory cycle including thermal expansion. Sample starts at a high temperature
is stretched then cooled. Sample released gradually from 2 s then clamped again at 3 s. After 3 s the sample is
reheated. The thermal expansion causes the stress to go negative (compressive) first, then the shape memory
effect causes the stress to increase and the stress applied during the stretching is recovered.

Figures 6.7 and 6.8 above show that the large strain UMAT agrees with the analytical
solution. This gives confidence that the UMAT has been coded correctly.

93
 CHAPTER 6

6.6 Comparison of UMATs with experiment

The two UMATs were tested by creating an Abaqus model of the recovery force experiment
(described in section 3.2.7) and comparing the results of the simulated and experimental
tests. The FE model is shown below in Figure 6.9. It consists of a thin material (same
dimensions as used in the experiment) which is meshed with 3-D quadratic elements and
stretched uniaxially (in the y direction) while at a high temperature. The material is then
cooled and due to the shape memory retains this shape. Boundary conditions are then
applied to prevent the material from contracting in the y direction and then it is reheated.
Due to symmetry, it was only necessary to model one quarter of the sample.

Temporary shape
Temporary shape at
at 20°C
120°C

Original
shape

Figure 6.9 FE model of recovery force experiment. The sample on the left was heated, stretched, cooled and
released. This left it in the stress-free elongated form shown in the centre. The sample is then constrained in
the y direction and reheated result in the stressed shape shown on the right. The contours show the Von-Mises
stress in MPa.

As the material heats up a stress is produced as the material tries to contract but is held by
the applied boundary conditions. The reaction force in the y-direction for each element is
then summed to produce the simulated shrink force which is then compared with
experiment.

94
 CHAPTER 6

6.6.1 Stored strain model

The results of the simulations of the recovery force test using the stored strain UMAT are
shown below. The test was done with a modulus of 0.44 MPa. This modulus was determined
in Section 4.13. Figure 4.12 shows the graph which was used to determine the modulus.

0.6

0.5

0.4
Experimental
Stress (MPa)

0.3
Simulated

0.2

0.1

0
20 40 60 80 100 120
-0.1
Temperature (°C)

Figure 6.10 Recovery stress of heat-shrink during in uniaxial tension test. Comparison of the
stored strain UMAT with experimental test.

5
4.5
4
3.5 Experimental
3 Simulated
Force (N)

2.5
2
1.5
1
0.5
0
-0.5 20 40 60 80 100 120
Temperature (°C)

Figure 6.11 Recovery force of heat-shrink during in uniaxial tension test. Comparison of the stored strain UMAT
with experimental test.

95
 CHAPTER 6

The stored strain model predicts the recovery stress quite well but the deformation of the
cross sectional area is unrealistic. This produces a force which does not agree with the
experimental results.

6.6.2 Multiple natural configurations model

The finite deformation UMAT was then used to model the same recovery force test used in
the previous section. The modulus E = 0.44 MPa was used as determined from Figure 4.12.

2.5

1.5

1 Experimental

Alpha = 200e-6 /°C

Force (N)

0.5
Alpha = 400e-6 /°C

0
0 20 40 60 80 100 120 140

-0.5

-1

-1.5
Temperature (°C)

Figure 6.12 Recovery force of heat-shrink during in uniaxial tension test. Comparison of multiple natural
configurations UMAT with experimental test. Two thermal expansion coefficients have been used and the
modulus used is 0.44 MPa.

The force at the end of the test (120°C) agrees well with experiment. The recovery force
during the test is highly dependent on the thermal expansion coefficient used but the force
at the end of the test is not. The measured thermal expansion coefficient of 4x10-4 °C-1
produces a negative force significantly larger than that observed in the experiment. With the
thermal expansion set to 2x10-4 °C-1 the simulated and experimental tests agree well.

The multiple natural configurations UMAT was then used to model the behaviour of the heat-
shrink tubing in a FE model of the hanger test described in Section 3.5. Due to the symmetry

96
 CHAPTER 6

of the hanger test it was possible to model just a quarter of the assembly. The heat-shrink
tubing is modelled using the large strain UMAT with the modulus calculated from the
pressure test section (Section 4.7). The model set-up is shown below in Figure 6.13. The
image on the left shows the model after the heat-shrink has been stretched out and cooled.
The image on the right shows the assembly at the end of the test after the heat-shrink has
been reheated and has contracted onto the hanger.

Figure 6.13 Schematic diagram of the finite element model of the hanger test.

0.16

0.14

0.12
Experimental
Shrink force (N)

0.1

0.08 Modelled

0.06

0.04

0.02

0
HS 1 HS 2 HS 3 HS 4 HS 5

Figure 6.14 Comparison of shrink force predicted by Abaqus model and shrink force measured by hanger test
experiment.

97
 CHAPTER 6

Comparison of the hanger test model and the experimental hanger test is shown above in
Figure 6.14. The simulation agrees well with the experiment for 4 out of the 5 heat-shrink
grades. However a higher shrink force was predicted than that measured by experiment for
the first tube (HS1). It should be recognized that as the shrink force gets smaller the
percentage error in both the pressure test and the shrink force test will increase. A
comparison of the shrink force predicted by the model and that measured by experiment is
given below in Table 6.1.

Shrink Force (N) Experimental Modelled Error (N) Error %

HS 1 0.027 0.039 0.0123 46.1
HS 2 0.065 0.067 0.002 3.1
HS 3 0.099 0.094 0.005 5.1
HS 4 0.118 0.117 0.001 0.85
HS 5 0.137 0.132 0.005 3.6
Table 6.1 Comparison of shrink force predicted by FE model using multiple natural configurations UMAT and
the shrink force measured by experiment.

6.7 Material model for PEBAX®

The manufacturers of PEBAX®, Arkemia, have made a large amount of PEBAX® properties
available online via the CAMPUS (computer aided material preselection by uniform
standards) [85] database. This was used to obtain PEBAX® properties below the melting point
and data for PEBAX® above the melting point was obtained from the literature [86].

The material model used to model the PEBAX® was a viscoelastic model using a Prony series
along with the time-temperature superposition to capture the behaviour of the material with
time and temperature.

6.7.1 Prony series

A Prony series [87] consists of a number of Maxwell elements connected in parallel and a
spring connected in parallel with the whole array (Figure 6.15).

98
 CHAPTER 6

Figure 6.15 A Prony series.

The Prony series can be used to model the stress relaxation behaviour for a polymer. A large
number of elements can be used to accurately capture the relaxation over an extended
period of time. The equation for how the modulus changes with time is

𝑁
−𝑡
𝐸(𝑡) = 𝐸0 − ∑ 𝐸𝑖 [1 − 𝑒 𝜏𝑖 ] (6.27)
𝑖=1

Where E0 is the instantaneous modulus, N is the number of elements in the model and Ei and
τi are the moduli and relaxation constants of relevant Maxwell element. Abaqus assumes
that the viscoelastic material is defined by a Prony series expansion of the dimensionless
relaxation modulus g(t)

𝑁
𝐸(𝑡) −𝑡
𝑔(𝑡) = = 1 − ∑ 𝑔𝑖 [1 − 𝑒 𝜏𝑖 ] (6.28)
𝐸0
𝑖=1

where gi range from 0 to 1 and the sum of gi over N must be less than 1. The relaxation
modulus terms, gi, define how much that element relaxes. For example if gi = 0 then this
element causes no reduction in the overall modulus of the material and if gi = 0.9 there will
be a 90% reduction in the overall modulus of the material. The time at which these
relaxations occur is determined by the relaxation constants, τi. To implement the Prony series
in Abaqus FEA, the two parameters required for each element are gi and τi.

6.7.2 Time-temperature superposition

The time-temperature superposition (TTS) principle [88] can be used to determine the
mechanical properties of temperature-dependent linear viscoelastic materials from known
properties at a reference temperature. The time-temperature superposition uses a reduced
time concept where the effect of increasing the temperature is assumed to be the same as
the effect of decreasing the time. The modulus as a function of time for three temperatures
is shown in Figure 6.16. If the modulus is known at one reference temperature (TREF) then

99
 CHAPTER 6

the modulus at higher temperatures (T1, T2) can be estimated by shifting the reference curve
to the left. This shift factor, aT, is illustrated in Figure 6.16. The modulus curve can also be
shifted to the right for lower temperatures if required. The TTS is usually used for amorphous
polymers whereas the PEBAX® is a semi-crystalline polymer. However as the material is
above its crystalline melting point it is all in the amorphous phase and so it was considered
acceptable.
Modulus

[E]TREF TREF < T1 < T2

T2 T1 TREF

log(aT)
[E]T1

t t/aT log (time)

Figure 6.16 Modulus as a function of time for three temperatures. The time-temperature superposition uses a
reduced time to represent the behaviour of the material at higher temperatures.

There is a uniform distance (log(aT)) between the TREF curve and the T1 curve. If the modulus
as a function of time is known at the temperature TREF then the modulus at T1 can be
calculated by using a time (t/aT) as shown in Figure 6.16. This means the modulus at a time t
and temperature T1 can be calculated using Equation 6.29.

𝑡
[𝐸(𝑡)] 𝑇1 = [𝐸 ( )] (6.29)
𝑎𝑇 𝑇𝑅𝐸𝐹

The time-temperature superposition was implemented in Abaqus FEA using a user

subroutine (UTRS) to control how much the curve was shifted. This reduced time concept for
temperature dependence is usually referred to as thermo-rheologically simple (TRS)
temperature dependence. The modulus calculated from the TTS is then used as the
instantaneous modulus E0 for the Prony series Equation 6.27.

100
 CHAPTER 6

6.7.3 PEBAX® data

The data needed to create a material model for PEBAX® was obtained from the literature
and from the manufactures’ (Arkema) website [89]. The manufactures website has DMTA
data up to 166°C and is shown in Figure 6.17 below. This was used to calibrate the PEBAX®
model while it was in the solid phase. Rheometry data obtained from the literature [86] was
used to calibrate the model in the molten phase.

1000

100
G' (MPa)

10

0.1
0 50 100 150 200
Temp (°C)
Figure 6.17 Variation of the shear storage modulus, G’, with temperature [89].

The rheometry data from the literature [86] is shown below in Figure 6.18. This data was for
PEBAX® 5533 while the PEBAX® used in the welding process is 7033. Viscosity data is
available for PEBAX® 7033 (and PEBAX® 5533) from Arkema but only for temperatures above
200°C. The viscosities of PEBAX® 5533 and PEBAX® 7033 above 200°C are very similar (as
seen in Figure 6.18) and it was assumed therefore that the rheometry data for PEBAX® 5533
could be applied to PEBAX® 7033.

101
 CHAPTER 6

10000 PEBAX 7033 at 200°C

PEBAX 7033 at 230°C
PEBAX 7033 at 260°C
PEBAX 5533 at 200°C
1000 PEBAX 5533 at 230°C
Viscosity (Pa s)

PEBAX 5533 at 260°C

100

10
1 10 100 1000 10000 100000
Shear rate(1/s)
Figure 6.18 Viscosity as a function of shear rate for two grades of PEBAX® at three different temperatures. Data
obtained from Campus database [85].

Figure 6.19 Rheometer data for PEBAX® 5533 showing modulus increasing rapidly with frequency and
decreasing with temperature [86]

The rheometer data in Figure 6.19 was converted to a modulus as a function of temperature
graph to allow comparison with DMTA data. To do this the frequency was converted to Hz
from rad/s and then the modulus at three frequencies (0.1 Hz, 1 Hz and 10 Hz) was plotted
for each of the five temperatures in Figure 6.20. The different grades of PEBAX® have
different melting temperatures with the PEBAX® 5533 melting 13°C lower than the PEBAX®
7033. It was assumed that shifting the temperature by 13°C would allow the PEBAX® 5533

102
 CHAPTER 6

data to be used to approximate the PEBAX® 7033. Therefore, the modulus of PEBAX® 5533
at 170°C was used to approximate the modulus of PEBAX® 7033 at 183°C and so on for the
other temperatures. The modulus as a function of temperature is shown below in Figure
6.20.

0.1

0.01

10 Hz
G' (MPa)

0.001
1 Hz
0.1 Hz

0.0001

0.00001
180 190 200 210 220
Temperature (°C)
Figure 6.20 Data used to approximate the shear modulus of PEBAX® 7033 above its melting temperature. Data
converted from [86].

The properties in the molten state can now be compared with the properties in the solid
state. However, there is a gap in data between 166°C and 183°C. As no data was available
for this region, interpolation was used to generate data. The modulus as a function of
temperature at 1 Hz is shown below in Figure 6.21 over the temperature ranges of interest.

103
 CHAPTER 6

100

10
DMTA
1 interpolated

Rheometry
G' (MPa)

0.1

0.01

0.001

0.0001
150 160 170 180 190 200 210 220
Temperature (°C)
Figure 6.21 Curve showing interpolated data around the melt temperature.

This graph was used to create a material model for the PEBAX®. First a Prony series was fit
to this data as shown in Figure 6.22.

100

10

1 Prony series
G' (MPa)

Experimental
0.1

0.01

0.001

0.0001
140 150 160 170 180 190 200 210 220 230
Temperature (°C)

Figure 6.22 Fitting of Prony series to experimental data.

This Prony series used six terms and they are given below in Table 6.2

104
 CHAPTER 6

Prony series term g (relaxation modulus) τi (relaxation constant)

1 0.899999 0.005
2 0.089999 0.005
3 0.0089999 0.1
4 0.00089999 0.5
5 8.9999x10-6 10
6 8.9999x10-6 10
Table 6.2. Terms uses for fitting Prony series to experimental data.

The time-temperature superposition (TTS) was used to implement the dependence of the
modulus on both time and temperature by shifting the curve in Figure 6.22 to the left or right
depending on time.

100
Experimental 10 Hz
TTS 10 Hz
10
Experimental 1 Hz
TTS 1 Hz
1
TTS 0.1 Hz
Experimental 0.1 Hz
0.1
G' (MPa)

0.01

0.001

0.0001

0.00001
140 150 160 170 180 190 200 210 220 230
Temperature (°C)
Figure 6.23 Fitting of material model to experimental data. G’ is the shear modulus.

A Prony series with six terms was used in conjunction with the TTS to fit a curve to the
experimental data (Figure 6.23).

As can be seen from Figure 6.23, when the material is below 155°C, the modulus is above 10
MPa. As the pressure applied by the heat-shrink is only on the order of 0.2 MPa, this means
that very little deformation will occur. Therefore it was assumed that the modulus below

105
 CHAPTER 6

155°C did not need to be accurately modelled. This made the process of fitting the Prony
series and TTS parameters easier.

6.8 Discussion

Two shape memory polymer models have been implemented in the finite element code
Abaqus FEA using two subroutines. The stored strain model is for small strains only while the
multiple natural configurations model can be used to model large strain problems. The two
models have been tested by comparing the shrink force produced in an experimental test
with the shrink force produced by a FE model of that test. These results show that the
multiple natural configurations model is able to replicate the behaviour of the heat-shrink
tubing while the stored strain model is not suitable for the deformation levels present in the
heat-shrink tubing. Also, data from literature has been used to create a material model for
the PEBAX®.

Only the multiple natural configurations model will be used in the full thermomechanical
model of the laser bonding process. Using the modulus determined from the pressure test
in Section 4.7, this model can reproduce the experimental shrink force from the hanger test
to within 5% for most cases. As described in Chapter Four, the effects of viscoelasticity,
plasticity and anisotropy are relatively small and are not modelled for ease of
implementation. Attempts to minimise the error introduced by these simplifications are
discussed below.

In Section 4.1.2 it was shown that the viscoelasticity of the heat-shrink tubing causes the
modulus to increase by 20% when the strain rate is increased by a factor of 10. This rate
dependent behaviour is not captured by this model but the modulus used in the model is
measured at approximately the same time scales as present in the welding process in an
attempt to minimize the error.

The plasticity of the heat-shrink also affects the accuracy of the model. The plastic properties
of heat-shrink are documented in Section 4.1.4. For this work the heat-shrink tubing was
received in its expanded state. All measurements of the tube were taken after this initial
expansion and the diameter of the tube before its initial expansion is not known. The initial
diameter of the tube used in the model is obtained by measuring the tube after shrinking.
This diameter will be slightly larger than the actual initial diameter due to the plasticity of
the tube. This means that while the model does not include plasticity the diameter of the
tube after shrinking is accurate.

106
 CHAPTER 6

A material model was developed to model the PEBAX® as it melts. A Prony series was used
along with the time-temperature superposition to model the viscoelastic behaviour of the
PEBAX®. In Figure 6.23 the fitting of the model to the experimental data is shown. As the
PEBAX® melts, the modulus drops very rapidly. When the temperature increases from 150°C
to 200°C, the modulus decreases by a factor of approximately 10,000. Also the behaviour of
the molten PEBAX® is very frequency dependent as it is effectively a viscous liquid. This
makes modelling this material as a solid model with a Lagrangian mesh very difficult.

107
7 Thermo-mechanical modelling of balloon catheter laser
welding process

7.1 Introduction

The purpose of this work was to investigate the laser welding of balloon catheters and to
gain insights into the deformation and flow characteristics, as well as the stresses and strains
associated with the process. In the previous chapters, various aspects of the welding process
were investigated. These include the thermal profile through the various layers of the
assembly (Chapter 5) to the behaviour of the heat-shrink used to induce pressure on the
PEBAX® layers to be welded (Chapter 6). In this chapter, these different aspects are
integrated to produce a full thermomechanical model of the laser welding process.

The particular bond that is modelled is the proximal bond of a catheter from a product line
called Sterling™. This bond was chosen as the most data was available and experiments could
be performed on this product more easily given the resources available at the time. It is
intended that the methods used here could be applied to model many other bonds across
different product lines.

In this chapter, full simulations of the entire welding process are described and performed.
Results of the laser welding model are presented with an emphasis on the melt flow length
of the PEBAX®. The melt flow length of the PEBAX® is defined as the distance from the distal
edge of the outer tube before welding to the distal edge of the outer tube after welding. The
melt flow length of the PEBAX® is chosen because it is the variable which is easiest to quantify
in experimental testing and hence as a means of validating the modelling approach used in
this work.

The impact of different parameters (friction, heat-shrink modulus and initial deformation)
on melt flow was investigated by performing simulations while varying the values of these
parameters. In the case of friction, no experimental data was available, and so a range of
values was chosen to produce a large variation of melt flow values as a means of determining
the sensitivity of friction values on the overall results. The pressure applied by the heat-
shrink tube on the assembly depends on the modulus of the heat-shrink. Therefore, the heat-

108
 CHAPTER 7

shrink modulus is an important parameter and its effect on the process was investigated by
performing computational simulations with different values.

To assess the accuracy of the models, experimental testing was performed. These
experiments involved welding catheters while varying the shrink force of the heat-shrink
tubing. These bonds were then imaged using optical coherence tomography (OCT) and the
melt flow measured. The experimental results were then compared with the predictions of
the simulations.

7.2 Thermo-mechanical model development

The full simulation of the laser bonding process incorporates all the processes discussed
previously in this work. The thermal profile during the laser welding process was described
in Chapter 5. In Chapter 4, the properties of the heat-shrink tubing were investigated and in
Chapter 6 the heat-shrink tubing was modelled with the development of a shape memory
polymer UMAT. This UMAT was validated in Chapter 6 and is incorporated into the thermo-
mechanical simulation in this work. The role of the heat-shrink material is to impose radial
pressure on the weld assembly with increasing temperature without itself forming part of
the balloon catheter weld. The constitutive behaviour of the Polyethylene material (PEBAX®)
used in the construction of both the catheter and the balloon was considered in Chapter 4.
The complexity in modelling this material behaviour was considered as it transitions from a
solid extrusion to a viscous flowing ‘liquid’. The overall material model used a temperature
dependent viscoelastic solid material to characterise the material behaviour. This material
model is used in the full simulations of the laser welding process in this chapter.

7.2.1 Laser weld assembly geometry

The basic construction of the geometry is shown in Figure 7.1, below. The geometry is
axisymmetric and the model is made up of four parts:

 the stainless steel cylindrical mandrel with a diameter of 0.8 mm

 the polyethylene (PEBAX®) balloon, with an internal diameter of 0.8 mm and wall
thickness of 0.046 mm
 the outer polyethylene (PEBAX®) tube, with an internal diameter of 0.89 mm and a
wall thickness of 0.114 mm

109
 CHAPTER 7

 and the EVA copolymer heat-shrink tubing, whose dimensions change upon reaching
a critical (phase transforming) temperature of approximately 100°C.

The length of the overlapped joint is 1.75 mm and the overall length of the assembly
modelled is 50 mm with an overall radius of approximately 1 mm.

In the finite element model constructed to reflect the assembly, the heat-shrink tubing is
initially represented in its ‘as formed’, stress free configuration. A temperature load is then
imposed on the heat-shrink material and simultaneously stretched to its nominal diameter
at which stage it is cooled so the shape memory has already been programmed into the
material at the point shown in Figure 7.1 below. This method of modelling the processing of
the heat-shrink material is described in more detail in Chapter 6.

A
)

Heat-shrink

Outer tube

Balloon

Mandrel

1.75 mm

C
Figure 7.1 Axisymmetric geometry of model for simulation of laser welding of Sterling TM balloon catheter.
Figures (A)-(C) show the assembly with increasing detail. The in-built small gap between the heat-shrink and
the outer tube is shown. When the heat-shrink is heated, it contracts and comes into contact with the outer.

110
 CHAPTER 7

For some of the simulations, the outer PEBAX® tube started off in a state with an assumed
initial deformation. This is described in more detail in section 7.3.3.

Heat-
shrink

Outer tube

Balloon

Mandrel

Figure 7.2 Part of the axisymmetric mesh used for a model of SterlingTM bonding process.

7.2.2 Mesh

The finite element mesh constructed had approximately 12,000 2D axisymmetric

quadrilateral and triangular (CAX4T and CAX3T) linear elements. The mesh density is highest
at the tip of the PEBAX® (shown in Figure 7.2) as this is the region where the most
deformation is predicted to occur. The outer tube was assumed to have a rounded geometry,
as representing the end of the outer tube with a ‘squared’ corner caused excessive
deformation at the corner elements, upon loading. A curved mesh was used to reduce this
excessive mesh deformation.

7.2.3 Boundary conditions

The left and right edges of the mandrel, balloon and heat-shrink are constrained in the axial
direction, while just the right edge of the outer tube is constrained axially. In these models
the mechanical load is not applied directly to the outer tube. Instead the heat-shrink is
subjected to a simulated temperature load of 100°C, then stretched using a displacement
boundary condition of 0.159 mm (to the position shown in Figure 7.2 above) and then cooled
back to room temperature while the displacement is maintained. This is the method used to

111
 CHAPTER 7

induce the shape memory polymer internal stress, as described in the development of this
thermo-mechanical behaviour of the heat-shrink in Chapter 6. Following the simulation of
the forming of the heat-shrink material - the laser then heats the assembly and the heat-
shrink tubing contracts onto the assembly. This produces the mechanical loading for the
model.

7.2.4 Thermomechanical loading

The thermal load applied to the geometry above in Figure 7.2 is the same as that described
in Section 5.3. This thermal load reflects the temperatures experienced by the materials
during the welding process. In this model, the conductivity of the PEBAX® in the outer tube
is set at a relatively high value of 1.0 Wm-1K-1. The thermal conductivity value used by Boston
scientific is 0.2 Wm-1K-1. The higher value of 1.0 was chosen as it greatly aids the generation
of a converged finite element solution through reducing the thermal gradient (and hence the
deformation gradient) within the melted PEBAX® layer. A realistic value of 0.2 Wm-1K-1 causes
the top of the outer tube to become much hotter than the lower part of the PEBAX® and this
causes the top to excessively shear and produce excessive deformation. The higher
conductivity value of 1.0 changes the thermal solution and this is a limitation of this
technique but it was deemed necessary to get the model to converge.

7.2.5 Contact

In this model, contact between the various layers of the assembly need to be simulated. This
is modelled using the built-in penalty contact model in the finite element code Abaqus
Version 6.11. This model assumes the pressure between the two surfaces to be zero when
the surfaces are more than 0.2 µm apart and it begins to increase linearly once the gap
becomes less than 0.2 µm. As the gap approaches zero, the pressure approaches 1 MPa. If
the gap becomes less than zero, a situation known as overclosure, the pressure continues to
increase linearly. However in this model the pressure does not exceed 1 MPa so overclosure
does not occur. The 1 MPa pressure was chosen to prevent overclosure as this often causes
convergence problems. The effect of the coefficient of friction was investigated over a range
of values to determine the sensitivity of the solution to frictional values.

112
 CHAPTER 7

7.2.6 Materials

The heat-shrink is modelled using the user defined material model (UMAT) for shape
memory polymers (SMP) described in Section 6.4. The modulus used for the majority of the
tests was 0.5 MPa as measured in Section 4.1.3. For some of the tests, the effect of modulus
was investigated to determine its effect on the melt flow. For these tests the modulus ranged
from 0.3 MPa to 1.3 MPa as this was the range used in an experimental investigation of melt
flow dependence on shrink force.

The PEBAX® is modelled as a temperature-dependent viscoelastic material as described in

Section 6.7. However, a boundary condition is placed on the lower PEBAX® layer (the balloon)
to prevent it from moving in order to simplify the computational analysis.

The mandrel is made of stainless steel with a thin coating of Teflon and both are modelled
as linear elastic materials with a Young’s modulus of 300 GPa. The thermal properties are
described in Section 5.3.

7.3 Results

7.3.1 Melt flow results

Figures 7.3, 7.4, 7.5 and 7.6 show the results of a melt flow model at specific time points.
This model uses a frictional value of 0.05 and a heat-shrink modulus of 0.5 MPa and is
presented here as an example of a melt flow simulation where the outer tube transitions
between a solid below the melt point to a viscous solid with a low modulus above the melt
temperature while being subjected to the pressure from the user defined heat-shrink
material model. The time points were selected to show the evolution of flow during the laser
welding process. The contour plot on the left, Figure 7.3 (A-D) and Figure 7.5 (A-D), shows
the temperature and the contour plot on the right, Figure 7.4 (A-D) and Figure 7.6 (A-D),
shows the logarithmic strain.

The first pair of images shows the assembly at a time of 4.7 s. At this stage the materials have
begun heating but the heat-shrink has not contracted yet. At a time of 7.0 s, enough of the
heat-shrink has been heated to a temperature above its crystalline melting point to allow it
to contract and come into contact with the PEBAX®. At this time point - the PEBAX® is still
below its melting point and hence there is minimal deformation. As the temperature in the

113
 CHAPTER 7

assembly rises, the PEBAX® begins to deform. By 8.7 s the PEBAX® has deformed significantly.
The laser power is reduced at this point as illustrated in Figure 5.11 and so the temperature
drops slightly and the PEBAX® deformation rate slows between 9 and 13 s. After 14 s the tip
of the PEBAX® begins to soften and deform due to thermal conductivity through the PEBAX®
layer with time. It is interesting to observe at this point that the heat-shrink tubing comes
into contact with the balloon (lower PEBAX® layer) in front of the deforming outer tube and
forms a barrier which prevents the type of rapid deformation / melt flow that occurred at
8.5 s. As the laser moves to the left the temperature of the PEBAX® tip increases allowing
further melt flow and from 16 s until the finish at 21.4 s the melt flow increases steadily.

114
 CHAPTER 7

Time (s) Temperature Logarithmic Strain

(A) (A)

t = 4.7

(B)
(B)
t = 7.0

(C) (C)
t = 8.5

(D) (D)

t = 8.7 Figure 7.3 (A-D) Temperature in assembly during first 8.7 s of welding. Figure 7.4 (A-D) Logarithmic strain in assembly during first 8.7 s of welding.

(°C)

115
 CHAPTER 7

(A)
t = 10.3 (A)

(B)
t = 14.1
(B)

(C)
t = 17.1 (C)

(D) (D)
t = 18.7

(E)
(E)
t = 21.4

Figure 7.5 (A-E) Temperature in assembly during welding from 10 s until finish at 21.4 s, Figure 7.6 (A-E) Logarithmic strain in assembly during welding from 10 s until finish at 21.4 s,
continued from figure 7.3. continued from figure 7.4.

116
 CHAPTER 7

The above deformation mechanisms are mapped in terms of melt flow against time in Figure
7.7 below.

0.45

0.4 Phase 1

0.35
Phase 2 Phase 3 Phase 4
0.3
Melt Flow (mm)

0.25

0.2

0.15

0.1

0.05

0
7 9 11 13 15 17 19 21
Time (sec)
Figure 7.7 Melt flow of PEBAX® during laser bonding simulation as a function of time. A coefficient of friction
of 0.05 was and the heat-shrink modulus was 0.5 MPa. These results are from the same model as the results in
Figures 7.4 - 7.7.

As illustrated in Figure 7.7, the melt flow process can be classified into four different phases
of deformation as a function of time:

 Phase 1: No significant deformation / flow occurs, as the PEBAX® layer is below its
melting point.
 Phase 2: PEBAX® flows rapidly, as the temperature induced by the laser causes the
PEBAX® to melt.
 Phase 3: Little or no flow of the PEBAX® as the outer PEBAX® layer tip cools slightly
and the heat-shrink deforms ahead of the tip causing an obstacle to the flow of the
polymer melt pool.
 Phase 4: PEBAX® flows again as the PEBAX® tip melts.

In the example above the PEBAX® slides between the heat-shrink and the outer tube quite
easily due to the low coefficient of friction (0.05). This results in a simulated melt flow length
of 0.43 mm.

117
 CHAPTER 7

7.3.2 Effect of friction

In order to investigate the effect of friction on the melt flow, simulations were performed
with various values of friction coefficients. The previous section showed how the PEBAX®
slides between the heat-shrink and the outer PEBAX® layer when the friction is assumed to
have a low value of 0.05. The example in Figure 7.8 below shows a melt flow simulation with
the same setup but using a coefficient of friction of 0.2.

In this case the PEBAX® slips very little between the heat-shrink and the inner PEBAX® layer.
The edges of the PEBAX® largely ‘stick’ to the heat-shrink and the outer tube, as the friction
is at the higher level of 0.2. As the material gets hotter as a direct result of the laser
absorption, the PEBAX® begins to demonstrate a boundary layer of flow as the polymer
displaces and flows through the centre while the edges remain stationary. This type of flow
results in a large amount of strain at the edges and relatively little in the centre as can be
seen in the contour plot in Figure 7.8. While the high friction flow produces more strain than
the low friction flow, it produces a considerably reduced overall melt flow length, despite
the simulations failing before completion.

118
 CHAPTER 7

(A)

(B)

(C)

Figure 7.8(A-C) Contour plot of logarithmic strain in assembly during simulation with a friction coefficient of
0.2. Figure 7.8A is the strain at 8.4 s, Figure 7.8B is the strain at 9.25 s and figure 7.8C is the strain at 15.6 s at
which point the simulation crashes due to excessive deformation.

119
 CHAPTER 7

Figure 7.9 below shows the effect of friction on melt flow for 4 different friction values
ranging from 0.05 to 0.2.

0.45

0.4

0.35 µ = 0.05

0.3 µ = 0.07
Melt Flow (mm)

µ = 0.10
0.25
µ = 0.20
0.2

0.15

0.1

0.05

0
7 9 11 13 15 17 19 21
Time (s)
Figure 7.9 Melt flow of PEBAX® over time for four different coefficients of friction (0.05, 0.07, 0.1 and 0.2). The
simulation ran to the end point of 21.4 s when the coefficient of friction was 0.05 but it failed due to excessive
deformation for the other three values of coefficient of friction.

It can be seen from examining Figure 7.9 above, that the initial heating phase (Phase 1) is
largely unaffected by the friction value as the temperature is too low for the PEBAX® to flow
or deform significantly under the heat-shrink pressure. The rapid deformation of the second
phase causes much more melt flow in the cases with lower friction. The extent of melt flow
depends heavily on the level of friction, highlighting the importance of friction in the melt
flow simulations. The duration of the third phase, which consists of little or no flow due to
the PEBAX® tip cooling as it get further from the heat source, also varies significantly with
friction. With a fiction of 0.05 the third phase lasts approximately 5 seconds and with a fiction
of 0.2 the third phase lasts approximately 1.5 seconds. This occurs because with a fiction
value of 0.05, the PEBAX® tip cools more as it moves further from the heat source than with
a fiction of 0.2.

120
 CHAPTER 7

It is also noted from Figure 7.9 that the simulation predictions above give data until different
time points. The data for the frictional values of 0.2, 0.1 and 0.07 represent simulation results
after 15.6, 17.7 and 18.7 seconds respectively. The only simulation to run to the end time-
point of 21.4 seconds was the one with the lowest coefficient of friction (μ = 0.05). The other
three simulations failed to complete due to the excessive deformation and strain caused by
the shear between the interface of the heat-shrink and the PEBAX®. This was determined to
occur when the logarithmic strain approached 1.5.

7.3.3 Effect of initial geometry

As there was considerable deformation and melt flow experienced, and as the material
model utilised a solid modelling approach to simulate the deformation of the PEBAX® as it
transitioned between a solid at temperatures below 172°C and a viscous liquid above this
temperature, considerable difficulties were experienced in completing the simulations over
the 21 second laser welding process.

To investigate if it was possible to run the simulations to the end without solution
convergence issues, the starting geometry was assumed to include some initial deformation.
The purpose of this initial deformation was to reduce the high localised strain levels in the
model. The figures below show the two deformed initial geometries examined. Figure 7.10(a)
represents an initially deformed geometry that represents melt flow of 0.05 mm from its
initial axial position and Figure 7.10(b) represents an initially deformed geometry that
represents melt flow of 0.1 mm. Both initial geometries assumed an in initial residual stress
state of zero.

(a)

121
 CHAPTER 7

(b)
Figure 7.10 Two geometries used to reduce strain levels in the model. Figure (a) shows an assumed initial
deformation of 0.05 mm and (b) shows an assumed initial deformation of 0.1 mm.

Simulations were performed on these two geometries in the same way as described
previously. The results are shown below in Figure 7.11 and Figure 7.12 for the two different
geometries respectively.

0.45

0.4 µ = 0.05

0.35 µ = 0.07

0.3 µ = 0.10
Melt Flow (mm)

0.25 µ = 0.20

0.2

0.15

0.1

0.05

0
7 9 11 13 15 17 19 21
Time (s)
Figure 7.11 Melt flow of PEBAX® over time for four different coefficients of friction using a geometry with an
initial deformation of 0.05 mm.

122
 CHAPTER 7

0.45
µ = 0.05
0.4
µ = 0.07
0.35
µ = 0.10
0.3
Melt Flow (mm)

µ = 0.20
0.25

0.2

0.15

0.1

0.05

0
7 9 11 13 15 17 19 21
Time (s)

Figure 7.12 Melt flow of PEBAX® over time for four different coefficients of friction using a geometry with an
initial deformation of 0.10mm.

In both of the above cases the simulation ran to completion for the three lower coefficients
of friction (0.05, 0.07 and 0.1) and experienced non-convergence issues associated with
significant mesh deformation for the highest coefficients of friction (0.2). The trends are very
similar in both graphs with the melt flow decreasing as the coefficient of friction increases.

The results below show how the assumption of including an initial stress-free geometry
affects the melt flow. The first set of curves (red curves in Figure 7.13) with a friction of 0.05
run to completion. It can be seen that the initial deformation does not have a significant
effect on melt flow length. With zero initial deformation the melt flow is 0.43 mm, whereas
assuming an initial stress free geometry representing 0.05 mm axial deformation predicts an
overall melt flow of 0.39 mm. Similarly, the simulations that assume an initial stress free
geometry representing 0.01 mm of axial deformation predict an overall melt flow of 0.42
mm.

This trend is repeated for the other coefficients of friction.

123
 CHAPTER 7

0.45

0.4

0.35

0.3

0.25
Melt Flow (mm)

0.2

0.15 Zero initial deformation (µ = 0.05)

0.05 mm initial deformation (µ = 0.05)
0.1 0.1 mm initial deformation (µ = 0.05)
Zero initial deformation (µ = 0.10)
0.05 0.05 mm initial deformation (µ = 0.10)
0.1 mm initial deformation (µ = 0.10)
0
7 9 11 13 15 17 19 21
Time (s)
Figure 7.13 Melt flow of PEBAX® over time for three different geometries and two different coefficients of
friction (0.05 and 0.10).

0.4

0.35

0.3

0.25
Melt Flow (mm)

0.2

0.15
Zero initial deformation (µ = 0.07)
0.1 0.05 mm initial deformation (µ = 0.07)
0.1 mm initial deformation (µ = 0.07)
Zero initial deformation (µ = 0.20)
0.05
0.05 mm initial deformation (µ = 0.20)
0.1 mm initial deformation (µ = 0.20)
0
7 9 11 13 15 17 19 21
Time (s)

Figure 7.14 Melt flow of PEBAX® over time for three different geometries and two different coefficients of
friction (0.07 and 0.20).

124
 CHAPTER 7

7.3.4 Effect of heat-shrink modulus

In investigating the effects of the heat-shrink modulus (and hence the level of stress imposed
by the heat-shrink) on the amount of melt flow of the outer tube, it was determined that the
heat-shrink with a higher modulus produced a larger shrink force, as would be expected. The
effect of changing the modulus was investigated using 4 different heat-shrink modulus
values. The melt flow length as a function of time, for the 4 different moduli, is shown below
in Figure 7.15.

0.5

0.45

0.4

0.35
Melt Flow (mm)

0.3
1.3 MPa
0.25
1.0 MPa
0.2 0.5 MPa
0.15 0.3 MPa
0.1

0.05

0
7 9 11 13 15 17 19 21
Time (s)

Figure 7.15 Melt flow of PEBAX® over time for four different heat-shrink moduli. These simulations used a
coefficient of friction of 0.05 and a geometry with no initial deformation.

Three of these simulations ran to completion while the one with the largest heat-shrink
modulus did not complete due to excessive mesh deformation. It is clear that the melt flow
increases with increasing modulus but the increase is relatively small. A heat-shrink modulus
of 0.3 MPa produces a melt flow of 0.39 mm while a heat-shrink modulus of 1.0 MPa
produces a melt flow of 0.49 mm. Therefore a greater than threefold increase in modulus
increases melt flow by approximately 25 %.

7.4 Experimental results

In order to attempt to validate the predictions of the model, a number of Sterling™ 3 mm

balloon catheters were welded and examined. Various heat-shrink tubing with different

125
 CHAPTER 7

moduli were used in these experimental tests to determine the validity of the computational
model predictions. The melt flow was measured experimentally using optical coherence
tomography (OCT) as described in section 3.2.3.

A total of 40 tests were performed, 10 each for 4 different heat-shrinks.

Melt pool Non-melted

Outer tube region

Balloon

Edge of melt pool

1 mm

Figure 7.16 OCT images of a SterlingTM balloon catheter welded using heat-shrink with a shrink force of 0.12 N.

The table below gives a summary of the results of the experimental results. The melt flow
length is defined as the distance from the tip of the PEBAX® after welding to its original
position. The melt pool length is defined as the distance from the tip of the PEBAX® after
welding to the proximal edge of the melt pool (right-hand edge highlighted in figure 7.16
above). The melt pool depth is included for information but the OCT data cannot measure
this accurately due to refraction effects.

126
 CHAPTER 7

0.9

0.8

0.7

0.6
Melt flow (mm)

0.5

0.4

0.3

0.2

0.1

0
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35
Shrink force (N)

Figure 7.17 Experimental measurements of melt flow of PEBAX® during welding for four different grades of
heat-shrink tubing. The shrink force is the value measured in the hanger test described in section 3.5. A total
of 10 experiments were performed for each of the 4 heat-shrink grades. The dashed line represents the average
values.

Shrink force Melt flow Melt Flow Std. Melt Pool Melt Pool
(N) Length (mm) Dev. (mm) Length (mm) Depth (mm)

0.065 0.58 0.09 1.48 0.14

0.118 0.67 0.07 1.69 0.15

0.185 0.69 0.06 1.69 0.14

0.288 0.68 0.05 1.75 0.15

Table 7.2 Summary of experimental results.

The experimental results shown in Figure 7.17 indicate that the pressure applied by the heat-
shrink tubing has a relatively small effect on the melt flow length. There is a 15 % increase in
melt flow from the first heat-shrink sample to the second. This is approximately the same as
the standard deviation from the experimental tests but after that there is only a small change
which is much less than the standard deviation of the tests. The first change is marginally
statistically significant but the subsequent changes are not.

127
 CHAPTER 7

As the shrink force increases the standard deviation of the melt flow length decreases. Also,
as the shrink force increases, the melt pool length increases slightly more than the melt flow
length, which indicates that the shrink force has a slightly larger effect on the temperature
profile in the bond than on the melt flow length.

0.8

0.7

0.6

0.5 Experimental
Melt Flow (mm)

FE model
0.4

0.3

0.2

0.1

0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6
Heat-shrink modulus (MPa)

Figure 7.18 Melt flow as a function of modulus as predicted by the model and measured by experiment. The
error bars show the standard deviation of the experimental results.

In Figure 7.18, the melt flow predicted by the model is compared with the melt flow
measured experimentally. The melt flow predicted by the model is approximately 30 % less
than that measured experimentally. The model predicts that the melt flow length increases
approximately linearly as the heat-shrink modulus increases, but the experimental data
shows the melt flow increasing initially then levelling off. To further examine these trends
the results in Figure 7.18 are normalized. The normalized results are plotted in Figure 7.19
and are rescaled and plotted again in Figure 7.20. The results are normalized by dividing all
the melt flow length values by the melt flow of the first sample and similarly, the modulus is
normalized by dividing the moduli values by the modulus of the first sample.

128
 CHAPTER 7

1.6

1.4

Melt Flow (Normalized) 1.2

1
Experimental
0.8
FE model
0.6

0.4

0.2

0
0 1 2 3 4 5
Heat-shrink modulus (Normalized)

Figure 7.19 Normalized melt flow as a function of normalized modulus as predicted by the model and measured
by experiment.

1.3

1.25
Melt Flow (Normalized)

1.2

1.15
Experimental

1.1 FE model

1.05

1
1 1.5 2 2.5 3 3.5 4 4.5 5
Heat-shrink modulus (Normalized)

Figure 7.20 Normalized melt flow as a function of normalized modulus as predicted by the model and measured
by experiment. Graph scaled to highlight trends.

In Figure 7.19 the normalized results show that there is quite a good agreement in the trends
between the model and experiment, as both curves show a small increase in melt flow with
heat-shrink modulus. The rescaled graph in Figure 7.20 highlights the differences in the
trends between the model and the experimental results. The model predicts melt flow
increasing approximately linearly as the modulus increases but the experimental results
show an initial increase in melt flow followed by relatively little change in melt flow as the
modulus is increased further.

129
 CHAPTER 7

The overall effect of the heat-shrink modulus on melt flow length is quite small for both
model and experiment. The model predicts that increasing the modulus by a factor of 3.3
increases melt flow by a factor of 1.25 and the experiment data shows that increasing the
modulus by a factor of 4.4 increases melt flow by a factor of 1.17.

7.5 Discussion

A finite element model of the laser bonding process of balloon catheters has been presented
in this chapter. The purpose of this model is to predict the thermomechanical behaviour of
the materials as they heat up and deform. The melt flow is the metric used to quantify the
amount of deformation that occurs. This is used to compare simulations with experiments
as well as to assess the effect of changing different parameters in the simulations.

The melt flow simulations all follow a similar pattern. A rapid increase in the melt flow length
is observed as the material reaches its melting temperature followed by a period with little
or no melt flow and then a final stage of steadily increasing melt flow as the tip of the outer
tube melts.

The effect of starting the simulation with the outer tube in a pre-deformed state was also
investigated. Starting with a pre-deformed geometry reduces mesh deformation and strain
in the model making it more likely the simulation will run to completion. By assuming an
initial stress free ‘deformed’ geometry, a very small effect on the predicted melt flow length
was observed. This implies that the stresses built up during the initial deformation do not
have a large impact on the melt flow length.

The effect of friction was determined to have an important role in predicting the melt flow
length. Altering the value of friction not only alters the melt flow length but also the
characteristics of the melt flow. With a low value of friction, the molten outer tube slides
easily between the balloon and the heat-shrink and this causes an increase in the predicted
melt flow length. Interestingly, for low coefficients of friction (0.05), the maximum strain
occurring in the outer tube is lower than the maximum strain for higher coefficients of
friction (0.2). With a high value of friction, the molten tube cannot slide between the balloon
and the heat-shrink and therefore the melt flow is significantly lower than in the low friction
case. Also as the surfaces of the outer polymer tube ‘stick’ to the heat-shrink and the balloon,
the material begins to flow through the centre of the outer tube representing boundary layer

130
 CHAPTER 7

flow. This causes a large amount of strain concentrated near the surfaces of the outer tube
(Figure 7.8C) even when the melt flow length is relatively small.

The effect of pressure (as imposed by the heat-shrink tubing) on the melt flow was also
investigated in this chapter. This was investigated by altering the value of the heat-shrink
modulus. As expected, it was found that the melt flow increased as the modulus of the heat-
shrink increased but not by a large amount. This is probably due to the fact that once the
heat-shrink diameter reduces in front of the outer tubing, it forms a barrier, restricting the
melt flow. Therefore, increases in the modulus of the heat-shrink not only results in increases
in the pressure pushing the PEBAX® forward but also in an axial pressure holding it back. This
would explain why, in Figure 7.15, there is a significant difference in the melt flow for the
different heat-shrink moduli during Phase 2 (the rapid deformation phase and before the
heat-shrink has come down in front of the deforming PEBAX® tip) but very little difference
in melt flow for Phases 3 and 4 (after the heat-shrink comes down in front of the deforming
PEBAX®). At the end of Phase 2 (approximately 9.1 s) the simulated melt flow lengths for the
three heat-shrink moduli (0.3 MPa, 0.5 MPa and 1.0 MPa) are 0.21 mm, 0.25 mm and 0.33
mm respectively. The melt flow during Phase 2 with a heat-shrink modulus of 1.0 MPa is 50
% larger than the melt flow with a heat-shrink modulus of 0.3 MPa. However, for Phases 3
and 4 the three simulations predict a further melt flow of approximately 0.18 mm
independent of the heat-shrink modulus.

Simulation results are in broad agreement with experimental measurements of melt flow
length in both trends and in quantitative predictions. However, the computational
simulations underestimate the melt flow length by approximately 30 %. There are a number
of factors which could explain this.

The higher thermal conductivity used in these simulations could affect the melt flow. The
higher conductivity was introduced to reduce the temperature gradient occurring across the
outer tube as this was causing excessive shear. As the mechanical properties of the PEBAX®
are very dependent on temperature, any change to the temperature profile could have a
significant effect on the melt flow.

Also, the value of the coefficient of friction was determined to have a large influence on the
simulated melt flow. In this work, the coefficients of friction varied from 0.05 to 0.20 and as
can be seen from Figure 7.9, the melt flow length decreases significantly as the coefficient of
friction is increased. As no experimental data for the coefficient of friction was available, a

131
 CHAPTER 7

value was chosen to produce a melt flow closest to that observed experimentally. The
coefficient of friction used in the comparisons with experimental data (Figures 7.18, 7.19 and
7.20) was 0.05. A lower value would have resulted in more melt flow, but attempts at these
simulations failed due to excessive mesh deformation.

The trend in the simulations is for the melt flow to increase slightly with increasing modulus
but the trend in the experimental results is for a small increase then a levelling off. However,
it should be emphasized that difference in trends is a relatively small difference on a
relatively small trend.

132
CHAPTER 8 Concluding remarks

The aim of this work was to create a model of the laser bonding process used by Boston
Scientific in the manufacturing of balloon catheters so that greater understanding of the flow
characteristics and hence quality of bonding of the balloon to the catheter could be achieved.
The laser bonding process uses a laser to heat an assembly containing the catheter parts to
be welded and a heat-shrink tubing. The heat-shrink tubing surrounds the catheter and
applies pressure to the assembly when heated. The catheter is made of an elastomeric
copolymer called PEBAX®. When the PEBAX® reaches its melting point it flows due to the
pressure applied by the heat-shrink. This flow results in a bond with a tapered profile. The
model developed provides a tool which can be used in future studies of the laser bonding
process so that the quality of bonding can be understood.

Prior to engaging on this research, there were many aspects of the bonding process which
were not well understood, including the thermo-mechanical behaviour of the heat-shrink
tubing, the thermo-mechanical deformation that occurs during the process and which
parameters have the largest effect on the bonding process. This thesis contributes
understanding to these through a combination approach of experimental testing and
computational modelling.

8.1 Conclusions

A large amount of research has been published on characterizing and modelling shape
memory polymers (SMPs). As shown in Chapter 2 there are many SMP models which are
appropriate for implementation into a computational FE code. However, there is an absence
of literature investigating SMP for industrial applications such as laser bonding processes and
this was also highlighted in Chapter 2. The two models implemented in this work are the
stored strain model of Liu et al. [64] and the multiple natural configurations model of Barot
et al. [77].

The thermo-mechanical properties of large-size heat-shrink tubing were characterised using

dynamic mechanical thermal analysis and the results are presented in Chapter 4. Various
properties were quantified including the elastic modulus as a function of temperature,

133
 CHAPTER 8

viscosity, plasticity, anisotropy and shrink force. It was found that the stress-strain
characteristics at high temperatures could be modelled with the Neo-Hookean hyperelastic
model.

A pressure test was designed and developed to measure the shrink pressure of the small
heat-shrink tubes as it transitions through its phase change temperature. This was a novel
experiment which was simple to perform and produced repeatable results. The pressure test
data was then used to determine the modulus of the small heat-shrink tubes, as their small
size made testing in a tensile tester very difficult.

Modelling the PEBAX® above and below its melt temperature using a solid modelling
approach proved difficult due to the very large difference in the stiffness between the solid
and molten PEBAX®. However, a model was successfully created using a Prony series and the
time-temperature superposition. This model was proved to be reasonably accurate over the
relevant time and temperature ranges.

A thermal model was described in Chapter 5 which predicts the temperature throughout the
assembly. The model was validated by comparing the predicted melt pool with experimental
data. It was found that the model predicted the melt pool quite accurately. It was necessary
to do this to determine the temperature profile as a basis for investigating the thermo-
mechanical behaviour in Chapter 7.

The two SMP models implemented into UMATs within ABAQUS were used to model the
shrink force experiments. The Barot et al. model of SMPs was able to accurately model the
shrink force of the heat-shrink, including the effects of thermal expansion. Thermal
expansion affects the force as a function of time curve but has a negligible effect on the final
force value. The Liu et al. model was not suitable for the large strains present in the heat-
shrink as it is deformed and recovers with imposed temperature.

The large thermal gradients across the PEBAX® caused a very large variation in stiffness. This
proved very difficult to model as it resulted in large localized shear strains which caused
convergence problems. The thermal conductivity of the PEBAX® was increased to reduce the
thermal gradients.

To deal with the large strains, there are a number of methods available within Abaqus. These
include remeshing, explicit analysis and coupled Eulerian Lagrangian analysis. The remeshing
technique was attempted but was ineffective because the remeshing cannot move the nodes
that are in contact. The explicit technique was also attempted but was unstable and also the

134
 CHAPTER 8

time-steps required were extremely small and this made the analysis very computationally
expensive. The coupled Eulerian Lagrangian technique is designed for fluid-solid interactions
and is often used for many large strain problems. However in this case the PEBAX® is
modelled as a solid and so this technique is not suitable. These techniques could be
potentially useful in future work if these limitations were overcome.

The large strains present in the thermo-mechanical model often caused convergence
problems. To reduce the strain levels some of the simulations were started with the PEBAX®
in a deformed state. It was found that this technique enabled the simulation to run for
longer, and it did not significantly affect the final melt flow prediction. This novel approach
worked reasonable well as a means of reducing strain levels and resulted in similar
predictions of melt flow with time.

The melt flow predicted by the thermo-mechanical model described in Chapter 7 is

reasonably accurate with the model predicting approximately 65% of the experimental melt
flow length. The coefficient of friction was found to be a very important parameter in the
melt flow model. The heat-shrink modulus did not have a large effect on the melt flow length.
This is due to the fact that the heat-shrink tube restricts the flow of the PEBAX® after it
contracts onto the assembly in front of the advancing melt pool. The experimental data
generated confirmed that the heat-shrink modulus had a modest impact on the melt flow
length, and this trend was accurately predicted by the computational model.

8.2 Limitations of this work

The thermo-mechanical model of the laser bonding process has a number of limitations.
Some of these limitations arise from assumptions, others from a lack of data and others from
the limits of the approach taken. Some of the main limitations of this work are outlined
below.

This model attempted to use a solid modelling approach to represent the PEBAX® as it melts
and flows. The validity of this is an open question. In reality the molten PEBAX® is a viscous
liquid but as the majority of the PEBAX® is solid for the majority of the time and as the melt
flow rates and times are low, a solid model approach was taken.

135
 CHAPTER 8

The model used for the PEBAX® was calibrated but not validated. Also, the calibration only
produced a reasonable degree of accuracy over a relatively small range of temperatures and
frequencies (but the most important range for the welding process). Inaccuracies in the
model predictions are possible due to these assumptions.

The coefficient of friction was found to have a large impact on the solution but no
experimental data on it was available. It had a very significant effect on both the qualitative
melt flow behaviour and the quantitative melt flow length and therefore the lack of
experimental data is a major limitation of this work.

The thermal conductivity of the outer tube was increased above values quoted in the
literature in order to reduce the thermal gradient across the tube which was causing
excessive distortion. This altered the initial flow behaviour of the PEBAX® in the model to
reduce the large amounts of shear which were occurring. The effect of this on the final melt
flow length is not known.

The strain levels observed in the computational model are very high and this often causes
convergence problems. This limited the number and range of the parameter studies that
could be undertaken with this complex model.

There is no experimental data of how the PEBAX® flows during the process. The only data is
the position of the edge of the melt pool after welding. The behaviour of the molten PEBAX®
may be significantly different to that predicted by the model.

The heat-shrink model implemented here does not include viscoelasticity, plasticity or
anisotropy, but these probably are considered to not have a significant effect on the melt
flow of the PEBAX®.

8.3 Discussion

The main objective of this work was to create a computational model of the catheter bonding
process and this has been achieved. The full thermomechanical model of the bonding
process is qualitatively accurate in that it predicts the shape of the bond after welding
reasonably accurately. The quantitative accuracy was assessed by comparing the melt flow
length predicted by the model with that measured experimentally. The model under predicts

136
 CHAPTER 8

the amount of melt flow by approximately 30% and possible reasons for this have been
discussed.

In order to create the thermo-mechanical model of the bonding process it was necessary to
model the thermo-mechanical behaviour of the heat-shrink tubing. Prior to this work the
details of the heat-shrink behaviour were poorly understood. In this work the heat-shrink
has been extensively characterised and a model which accurately predicts its behaviour has
been implemented into a computational FE code. This is a significant achievement and shows
that SMPs can be accurately modelled and implemented into an FE model of a complicated
industrial process.

The model of the bonding process provided a number of insights into the process which were
not obvious. These insights include, a higher coefficient of friction results in less melt flow
but produces higher strain values, the heat-shrink modulus has little impact on the melt flow
and that the coefficient of friction has a large effect on the amount of melt flow.

This work will be very useful to Boston Scientific in the future because it will help them
develop a better understanding of the bonding process. The effect of changing process
parameters can be modelled and this should reduce the amount of trial and error
experiments which are time-consuming and expensive. It is hoped that the better
understanding of the process can help to reduce the number of defects that occur in the
bonding process.

There are a number of aspects about this work which are novel. The heat-shrink tubing has
been tested extensively and characterised in more detail than previously available. The
pressure test developed to measure the shrink force of the small heat-shrink tube was also
novel. The use of a SMP model in real-world FE application is quite novel.

Overall this work provides a significant amount of new knowledge about the laser bonding
process and it provides a good foundation for further study.

8.4 Future work

There is a lot of future work which could be done to make this model more useful and to
increase understanding of the bonding process. From a modelling point of view more
parameter studies could be performed investigating the effect of different parameters such

137
 CHAPTER 8

as temperature, time and geometry on the solution. The accuracy of the heat-shrink model
could be further improved by including effects such as viscosity, plasticity and anisotropy.

The laser interaction with assembly could be integrated into the FE model along with the PID
control algorithm. This would allow for easily changing the geometry and would make the
model much more adaptable and useful.

The PEBAX® model could be validated. This could involve modelling a number of rheometry
tests and comparing the results with experimental data.

From an experimental point of view the coefficient of friction should be quantified, including
how it changes with temperature. An experimental study which stopped the welding process
at various time points could determine if the melt flow as a function of time predicted by the
model is realistic.

The bond modelled in this work was axisymmetric but some catheter bonds are not. The
methods used here could be applied to those bonds, although this would be computationally
expensive. This would be helpful because the melt flow which occurs in those non-
axisymmetric bonds is more complicated and less understood than the melt flow in the
axisymmetric bonds.

138
References

1. Brown David E., (2008) “Thomas Fogarty.” in ‘Inventing Modern America: From the
Microwave to the Mouse.’ The MIT Press.
2. Kilger E: ‘Intensive care after minimally invasive and conventional coronary surgery:
a prospective comparison’, Intensive care medicine, 27, pp 534-539, (2001).
3. [World Health Statistics 2009]. (2009). Retrieved from
http://www.who.int/whosis/whostat/2009/en/index.html
4. [Coronary Heart Disease]. (n.d.). Retrieved from
http://www.uchospitals.edu/online-library/content=P00207
5. Dotter C, Judkins M). ‘Transluminal treatment of arteriosclerotic obstruction.
Description of a new technique and a preliminary report of its applications’ 30 (5):
654–70 (1964).
6. [Cardiac & Heart Surgeries]. (n.d). Retrieved from
http://www.medelines.asia/treatment/97/Coronary_Angioplasty_PTCA_.html
7. Cardiac catheters – Global Trends, Estimates and Forecasts, 2012-2018, Research
and Markets, June 2014.
8. [Dilatation Catheter / Balloon]. (n.d.). Retrieved from
http://www.medicalexpo.com/prod/boston-scientific/product-74672-588291.html
9. Behl M., Lendlein A., ‘Shape memory polymers’, Materials Today 10 20-28 (2007).
10. Wornyo E. ‘Fabrication and Characterization of Shape Memory Polymers at small-
scales’
11. Silvers H. J. Jr, Wachtell S.: ‘Perforating, welding and cutting plastics film with a
continuous CO2 laser’, PA State University. Eng, Proc, pp 88-97, August 1970.
12. Wise R. J: ‘Thermal welding of polymers’, Woodhead Publishing Ltd., Cambridge, Oct
1999.
13. Russek U. A.: ‘Innovative Trends in Laser Beam Welding of Thermoplastics’. Proc of
2 Int conf on Lasers in Manufacturing 105-111 June 2003.
14. Toyota Jidosha, K. K.: ‘Laser beam welding of plastic plates’ Patent application
JP85213304 26 Sept 1985.
15. Kagan U. A., Bray R. G., Kuhn W.P., ‘Laser Transmission welding of semicrystalline
thermoplastics. Part I Optical characterization of nylon-based plastics’, 58th Annual
Technical Conference of the Society of Plastics Engineers. May 2000, pp 1171-1181.

139
16. Kagan U. A., Pinho G. P.: ‘Laser transmission welding of semi-crystalline
thermoplastics. Part II Analysis of mechanical performance of Welded nylon’. 58th
Annual Technical Conference of the Society of Plastics Engineers. May 2000 pp.1182-
1190.
17. Coelho J. P., Abreu M. A., Pires M. C., ‘High Speed laser welding of films’, Opt Laser
Eng. 34, 385-395 (2000).
18. Mayboudi L. S., Birk A. M., Zak G, ‘A 2D thermal model for laser transmission welding
of thermoplastics’. Laser material processing conference, ICALEO 2005.
19. Mackland A. P., Crafer R. C., ‘Thermal modelling of laser welding and related
processes: A literature review”. Opt laser tech 37, 99-115 (2005).
20. Mayboudi L. S., Birk A. M., Zak G., ‘Laser Transmission Welding of a Lap-joint:
Thermal Imaging Observations and Three-Dimensional Finite Element Modelling’
Jour of Heat Transfer 127, 1177-1186 (2007).
21. Frick T., 2007. ‘Computation of temperature fields for laser transmission welding of
plastics’. Proceedings of ICALEO 2007 Congress, Orlando, USA, pp. 596-601.
22. Casalino G., Ghorbel E., ‘Numerical model of CO2 laser welding of thermoplastic
polymers’, J of Mat. proc. tech. 207 63-71 (2008).
23. Van de Ven J. D., Erdman A.G., ‘Laser Transmission welding of thermoplastics – Part
I: Temperature and Pressure modelling’. J Manuf. Sci. Eng. 127 849-858 (2007).
24. Becker, F., Potente H., ‘A step towards understanding the heating phase of laser
transmission welding in polymers’. Polym. Eng. Sci. 42 (2) (2002).
25. Potente H., Korte J., Becker F, ‘Laser transmission welding of thermoplastics analysis
of heating phase’. J Reinf. Plast. Compos. 18 (10), 914-20 (1999).
26. Ilie, M., Cicola, E., Grevey, D., Mattei, S., Stoica, V, ‘Diode laser welding of ABS:
Experiments and process modeling’ Opt and Laser Tech 41, 608-614 (2004).
27. Coelho, J., Abren M. A., Rodrigues F. C., ‘Methodologies for determining
thermoplastic films optical parameters at 10.6µm laser wavelength’, Polymer Testing
23, 307-312 (2004).
28. Geiger M., Frick T., Schmidt M, ‘Optical properties of plastics and their role for the
modelling of the laser transmission welding process’. German Academic Society for
Production Engineering 3 49-55 (2008).
29. Incropera, F. P. and DeWitt, D. P. 1996, ‘Introduction to Heat Transfer’, 3rd ed., Wiley,
New York.

140
30. Coelho João M.P., Abreu M.A., Rodrigues F.C., ‘Thermal modelling CO laser radiation
transmission welding of superposed thermoplastic films’, Opt. Eng. 42 (11), 3365-
3373 (2003)
31. Van de Ven J., 2006, ‘Laser Transmission Welding of thermoplastics’, Thesis,
University of Minnesota, Twin Cities, Minneapolis, MN.
32. Potente H., Becker F., Fiegler G., and Korte J., ‘Investigations towards applications of
a New Technique on Laser Transmission Welding’, Weld World, 45 (5), pp. 2-7 (2001)
33. Potente H., Wilke L., Ridder H., Mohnken R., Shabon A., ‘Simulation of the Residual
Stresses in the contour Laser Welding of Thermoplastics’, Polym, Eng, Sci., 48 767-
773 (2008)
34. Potente H., Fiegler G., ‘An Approach to Model the Melt Displacement and
Temperature Profiles During the Laser Through–Transmission Welding of
Thermoplastics’, Polym. Eng. Sci, 46 1565-1575 (2006)
35. Yakacki C. M., Shandas R., Lanning C., Rech B., Eckstein A., Gall K., ‘Unconstrained
recovery characterization of shape-memory polymer networks for cardiovascular
applications’, Biomaterials, 28 2255-2263 (2007)
36. Mather P. T., Luo X., Rousseau I. A., ‘Shape memory polymer reasearch’, Annu. Rev.
Mater. Res. 39 445-471 (2009)
37. Behl M., Razzaq M. Y., Lendlein A., ‘Multifunctional Shape-Memory Polymers’, Adv.
Mater. 22, 3388-3410 (2010)
38. Serrano M. C., Ameer G. A., ’Recent insights into the biomedical applications of
shape-memory polymers’, Macromol. Biosci. 12, 1156-1171 (2012)
39. Beloshenko V. A., Varyukhin V. N, Voznyak Yu. V., ‘The shape memort effect in
polymers’, Russian chemical reviews, 74(3), 265-283 (2005)
40. Wei, Z. G., Sandstrom R., Miyazaki S., ‘Shape-memory materials and hybrid
composites for smart systems: Part 1 Shape-memory materials’, Journal of Material
ScienceScience, 33, 3743-3762 (1998)
41. Lendlein A., Kelch S., ’Shape-memory polymers’, Angew. Chem. Int. Ed. 41, 2034-
2057 (2002)
42. Xue L., Dai S., Li Z., ‘Biodegradable shape-memory block co-polymers for fast self-
expandable stents’, Biomaterials, 31 (32), 8132-8140 (2010)
43. Reese S., Bӧl M., Christ D., ‘Finite element-based multi-phase modelling of shape
memory polymer stents’, Comput. Methods Appl. Mech. Engrg., 199, 1276-1286
(2010)

141
44. Lendlein A., Langer R., ‘Biodegradable, elastic shape-memory polymers for potential
biomedical applications’, Science, 296, 1673-1676 (2002)
45. Ward Small, I. V., Singhal, P., Wilson, T. S., & Maitland, D. J. Biomedical applications
of thermally activated shape memory polymers. Journal of materials
chemistry, 20(17), 3356-3366 (2010)
46. Sokolowski W, Metcalfe A, Hayashi S, Yahia L, Raymond J., ‘Medical applications of
shape memory polymers,’ Biomed. Mater, 2:S23–27, (2007)
47. Sokolowski WM, Tan SC, ‘Advanced self-deployable structures for space
applications’, J. Spacecr. Rockets, 44, 750–54 (2007)
48. Mondal S., Hu J.L., ‘Temperature stimulating shape memory polyurethane for smart
clothing’, Indian J. Fibre Textile Res., 31 66 (2006)
49. Hussein, H., Harrison, D., ‘Investigation into the use of engineering polymers as
actuators to produce “automatic disassembly” of electronic products’, T. Bhamra,
Design and manufacture for sustainable development 2004, Wiley-VCH, Weinheim
(2004)
50. Yoo H.J., Jung Y.V., Sahoo N.G., Cho J.W., ‘Polyurethane-Carbon nanotube
nanocomposities prepared by In-Situ polymerization with electroactive shape
memory’, J. Macromol. Sci. Phys., 45, 441 (2006)
51. Maitland, D. J., Metzger, M. F., Schumann, D., Lee, A. and Wilson, T. S., ‘Photothermal
properties of shape memory polymer micro-actuators for treating stroke’ Lasers
Surg. Med., 30: 1–11 (2002)
52. Meng H., Li G., ‘A review of stimuli-responsive shape memory polymer composites’,
Polymer, 54 (9), 2199-2221 (2013)
53. Rousseau I. A., ‘Challenges of shape memory polymers: A review of progress toward
overcoming SMP’s limitations’, Polymer Engineering and Science, 48 2075-2089
(2008)
54. Meng H., Li G., ‘A review of stimuli-responsive shape memory polymer composites’,
Polymer, 54 (9), 2199-2221 (2013)
55. Liu C., Qin H., Mather P.T., ‘Review of progress in shape-memory polymers’, J. Mater.
Chem., 17 1543-1558 (2007).
56. Liang, C., C. A. Rogers, and E. Malafeew. "Preliminary investigation of shape memory
polymers and their hybrid composites." Smart structures and materials 97-105.
(1991)

142
57. Hayashi, S. "Technical Report on Shape Memory Polymers." Nagoya Research and
Development Center, Mitsubishi Heavy Industry, Inc (1990)
58. Tobushi H., Hashimoto T., Ito T., Hayashi S., Yamada E.,. ‘Thermomechanical
properties in a thin film of shape memory polymer of polyurethane series.’ J. Intell.
Mater. Syst. Struct., 9, 127-36 (1996)
59. Gall, K., Dunn, M. L., Liu, Y., Stefanic, G., & Balzar, D. ‘Internal stress storage in shape
memory polymer nanocomposites.’ Applied Physics Letters, 85(2), 290-292 (2004)
60. Beblo, R., & Weiland, L. M., ‘Strain induced anisotropic properties of shape memory
polymer.’ Smart Materials and Structures, 17(5), 055021. (2008)
61. Diani, J., Frédy, C., Gilormini, P., Merckel, Y., Régnier, G., & Rousseau, I. (2011). ‘A
torsion test for the study of the large deformation recovery of shape memory
polymers.’ Polymer Testing, 30(3), 335-341.
62. Ratna, D., & Karger-Kocsis, J., ‘Recent advances in shape memory polymers and
composites: a review’. Journal of Materials Science, 43(1), 254-269. (2008)
63. Chen, X., & Nguyen, T. D., ‘Influence of thermo-viscoelastic properties and loading
conditions on the recovery performance of shape memory polymers.’ Mechanics of
Materials, 43(3), 127-138 (2011)
64. Liu Y., Gall K., Dunn M.L., Greenberg A.R., Diani J., ‘Thermomechanics of shape
memory polymers: Uniaxial experiments and constitutive modelling’. Int. J. Plast. 22
279-313 (2006).
65. Gall, K., Yakacki, C. M., Liu, Y., Shandas, R., Willett, N., & Anseth, K. S. (2005).
Thermomechanics of the shape memory effect in polymers for biomedical
applications. Journal of Biomedical Materials Research Part A, 73(3), 339-348.
66. Chung, T., Romo-Uribe, A., & Mather, P. T. (2008). Two-way reversible shape
memory in a semi-crystalline network. Macromolecules, 41(1), 184-192.
67. Nguyen, T. D. (2013). Modeling shape-memory behaviour of polymers. Polymer
Reviews, 53(1), 130-152.
68. Wornyo E., Gall K., Yang F., King W., ‘Nanoindentation of shape memory polymer
networks’, Polymer 48, 3213-3225 (2007)
69. Tobushi H., Hashimoto T., Hayashi S., Yamada E., ‘Thermomechanical constitutive
modelling in shape memory polymer of polyurethane’. J. Intell. Mater. Syst. Struct.
8 711-718 (1997)
70. Tobushi H., Okumura K., Hayashi S., Ho N., ‘Thermomechanical constitutive model of
shape memory polymer’, Mech. Mat. 33, 545-554 (2001).

143
71. Morshedian J., Khonakdar H., Rasouli S., ‘Modelling of Shape Memory Induction and
Recovery in Heat-Shrinkable Polymers’, Macromol. Theory Simul. 14 428-434 (2005).
72. Nguyen, T. D., Jerry Qi, H., Castro, F., & Long, K. N. (2008). ‘A thermoviscoelastic
model for amorphous shape memory polymers: incorporating structural and stress
relaxation’, Journal of the Mechanics and Physics of Solids, 56(9), 2792-2814.
73. Ghosh, P., & Srinivasa, A. R. (2011). A two-network thermomechanical model of a
shape memory polymer. International Journal of Engineering Science, 49(9), 823-
838.
74. Westbrook, K. K., Kao, P. H., Castro, F., Ding, Y., & Jerry Qi, H. (2011). ‘A 3D finite
deformation constitutive model for amorphous shape memory polymers: a multi-
branch modelling approach for non-equilibrium relaxation processes.’ Mechanics of
Materials, 43(12), 853-869.
75. Castro F., Westbrook K. K., Long K. N., Shandas R., & Qi H. J. (2010). Effects of thermal
rates on the thermomechanical behaviours of amorphous shape memory
polymers. Mechanics of Time-Dependent Materials, 14(3), 219-241.
76. Chen Y., Lagoudas D., ‘A constitutive theory for shape memory polymers. Part 1:
Large deformations’, J. Mech. Phys. Solids, 56 1752-1765 (2008).
77. Barot G., Rao I.J., Rajagopal K.R., ‘Constitutive modelling of the mechanics associated
with crystallisable shape memory polymers’, Z.A.M.P. 57 652-681 (2006).
78. Rao I.J., Rajagopal K.R., ‘A thermodynamic framework for the study of crystallization
in polymers’, Z.A.M.P. 53 365-406 (2002).
79. Kafka V., ‘Shape memory polymers: A mesoscale model of the internal mechanism
leading to the S.M. phenomena’, Int. J. Plast., 24 1533-1548 (2008)
80. Kafka V., (2001) ‘Mesomechanical Constitutive Modelling’, World Scientific,
Singapore.
81. Qi H. J., Nguyen T. D., Castro F., Yakacki C. M., & Shandas R. (2008). Finite
deformation thermo-mechanical behaviour of thermally induced shape memory
polymers. Journal of the Mechanics and Physics of Solids, 56(5), 1730-1751.
82. Crawford R.J., 1998, Plastics Engineering, 3rd edn, Elsevier Butterworth-Heinmann,
Oxford, p 112.
83. Hsiao B. S., Zuo F., Mao Y., Schick C., (2013) Handbook of polymer crystallization:
Experimental Techniques, John Wiley & Sons Inc, p. 2.

144
84. Charlesby A., Hancock N. H., ‘The effect of cross-linking on the elastic modulus of
polythene’, Proceedings of the Royal Society of London A: Mathematical, Physical
and Engineering Sciences 218 245-255, (1953).
85. [Campus Database]. (n.d). Retrieved from http://www.campusplastics.com/
86. I-Kuan Yang, Ping-Hung Tsai, ‘Rheology and structure change of poly(ether-block-
amide) segmented block copolymer’, J. Cent. South Univ. Technol. (2007).
87. Park S. W., and R. A. Schapery. "Methods of interconversion between linear
viscoelastic material functions. Part I—A numerical method based on Prony
series." International Journal of Solids and Structures 36.11 (1999): 1653-1675.
88. Crawford R.J., 1998, Plastics Engineering, 3rd edn, Elsevier Butterworth-Heinmann,
Oxford, p 116.
89. [Dynamic mechanical analysis]. (n.d). Retrieved from
http://www.pebax.com/en/properties/mechanical-properties/dynamic-
mechanical-analysis-dma/index.html
90. [Snell’s Law]. (2015). In Encyclopaedia Britannica. Retrieved from
http://www.britannica.com/EBchecked/topic/550450/Snells-law
91. G. Woan, 2000, The Cambridge Handbook of Physics Formulas, Cambridge University
Press, Cambridge, New York.
92. Bothe L., Rehage G., Autohesion of Elastomers. Rubber Chemistry and Technology:
November 1982, Vol. 55, No. 5, pp. 1308-1327.
93. Schauer T., ‘Development of a Computational Laser Bond Process Model’. Boston
Scientific Technical Report (confidential),(n.d.).
94. Schauer T., McGowan R., ‘Source Term Determination for the Computational Laser
Bond Process Model’. Boston Scientific Technical Report (confidential),(n.d.).
95. Holman J.P., Heat transfer, 9th ed. McGraw Hill, New York, 2002.
96. [Plastic Materials Selection Guide for Micro Machining and Micro Molding]. (2011).
Retrieved from http://www.mikrotech.com/micro-molding-materials/material-
selection-guide.html
97. [Pebax MED]. (n.d). Retrieved from http://www.pebax.com/en/pebax-
range/product-viewer/Pebax-MED/
98. [Finite deformation of thick-walled cylinders under internal stress and linear
perturbation from a state of finite deformation]. (2013). Retrieved from
http://caigroup.ucsd.edu/pdf/

145
99. Mattson S, ‘Laser welding RNF 100 (or MT-LWA) shrink tubing’. Boston Scientific
Technical Report, Document number 30729-00, (confidential),(2001).

146