MULTISCALE MULTIPHYSICS THERMO-MECHANICAL

MODELING OF AN MGB2 BASED CONDUCTION COOLED

MRI MAGNET SYSTEM

by

ABDULLAH AL AMIN

Submitted in partial fulfillment of the requirements

For the degree of Doctor of Philosophy

Department of Mechanical and Aerospace Engineering

CASE WESTERN RESERVE UNIVERSITY

January, 2018
 CASE WESTERN RESERVE UNIVERSITY

SCHOOL OF GRADUATE STUDIES

We hereby approve the dissertation of

ABDULLAH AL AMIN

candidate for the degree of Doctor of Philosophy*.

Committee Chair

Dr. Michael Martens and Dr. Ozan Akkus

Committee Member

Dr. Bo Li

Committee Member

Dr. Ya-Ting Tseng Liao

Committee Member

Dr. Robert Brown

Date of Defense

November 14, 2017

*We also certify that written approval has been obtained

for any proprietary material contained therein

 Dedication

To my father Mohammed Abdul Muttalib Sarker and the women of my life

- Gulshan Ara, my mother who raised me as a person I am today.

- Nishat Sultana, my beautiful wife who promised to share the joy and sorrow

of my life.
 Table of Contents

Table of Contents ................................................................................................................. i

List of Tables .......................................................................................................................vi

List of Figures ..................................................................................................................... vii

Preface ............................................................................................................................... xii

Acknowledgement ............................................................................................................ xiv

List of Abbreviation ......................................................................................................... xviii

List of Symbols ................................................................................................................... xx

Abstract ........................................................................................................................... xxvi

Chapter 1. Introduction to Magnetic Resonance Imaging ............................................. 1

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

1.2 Use of MRI Units................................................................................................... 3

1.3 MRI Market Prices ................................................................................................ 4

1.4 Alternative to LHe ................................................................................................ 6

1.5 Conduction Cooled MgB2 MRI System ................................................................. 8

1.6 Recent Efforts on MgB2 Systems .......................................................................... 9

1.7 Challenges of MgB2 Systems. ............................................................................. 10

1.8 Thesis Overview ................................................................................................. 10

i
 1.9 Related Publications ........................................................................................... 12

1.10 Conclusion .......................................................................................................... 14

Chapter 2. Literature Review of MRI Magnet Modeling .............................................. 16

2.1 Literature Review ............................................................................................... 16

2.2 Modeling of Wire Material Properties ............................................................... 18

2.2.1 Modeling of Elastic Properties .................................................................... 19

2.2.2 Modeling of Thermal Expansion Coefficient............................................... 22

2.2.3 Modeling of Thermal Conductivity ............................................................. 24

2.2.4 Modeling of Specific Heat ........................................................................... 27

2.3 Modeling The Magnet Bundles .......................................................................... 30

2.3.1 Modeling of Winding .................................................................................. 30

2.3.2 Modeling of Thermal Cool-down ................................................................ 34

2.3.3 Modeling of Electromagnetic Charging ...................................................... 36

2.4 Modeling of Quench........................................................................................... 38

2.4.1 MgB2 Quench .............................................................................................. 40

2.4.2 Analytical Formulation of Quench .............................................................. 42

2.5 Conclusion .......................................................................................................... 44

Chapter 3. Material Properties for Magnet Modeling ................................................. 46

3.1 Material Properties ............................................................................................ 46

ii
 3.2 Elastic Properties ................................................................................................ 47

3.3 Thermal Expansion Coefficient (α) ..................................................................... 53

3.3.1 Temperature Dependent Secant Coefficient (αse(Τ)) ................................. 54

3.3.2 Computed Average Thermal Expansion Coefficient ................................... 55

3.3.3 Calculated Thermal Strain (εth) ................................................................... 56

3.4 Thermal Conductivity (k) .................................................................................... 57

3.5 Specific Heat (C) ................................................................................................. 60

3.6 Resistivity (ρ) ...................................................................................................... 60

3.7 Conclusion .......................................................................................................... 61

Chapter 4. Numerical Homogenization of Superconducting Composite Wire for

Material Properties ........................................................................................................... 62

4.1 Composite Wire.................................................................................................. 62

4.2 MgB2 Composite Superconducting Wire............................................................ 64

4.3 Representative Volume Element (RVE) .............................................................. 65

4.4 Modeling of Composite wire .............................................................................. 67

4.4.1 Computational Homogenization of Elastic Properties ............................... 67

4.4.2 Reducing the Full RVE ................................................................................. 74

4.4.3 Computational Homogenization of Thermal Expansion Coefficient .......... 78

4.4.4 Computational Homogenization of Thermal Conductivity ......................... 79

iii
 4.4.5 Computational Homogenization of Specific Heat....................................... 82

4.5 Finite Element Discretization of RVE .................................................................. 83

4.6 Summary of the Homogenized Properties of the Wires .................................... 85

4.7 Conclusion .......................................................................................................... 93

Chapter 5. Multiphysics Modeling of Magnet Manufacturing ..................................... 96

5.1 Full body MRI Background Magnet .................................................................... 96

5.2 Details of the full Body MRI Background Magnet .............................................. 98

5.3 Overview of the Magnet Manufacturing and Energization ............................... 99

5.4 Modeling of Magnet Manufacturing and Operation ....................................... 103

5.4.1 Winding Stress .......................................................................................... 104

5.4.2 Thermal Cool-down Stress ........................................................................ 106

5.4.3 Electro-Magnetic Stress ............................................................................ 107

5.4.4 Strain Development in the Magnet Bundles ............................................ 110

5.4.5 Effect of Support Conditions ..................................................................... 119

5.5 Global-Local Analysis of Composite Wire ........................................................ 126

5.6 Effect of Creep .................................................................................................. 127

5.7 Conclusion ........................................................................................................ 128

Chapter 6. Multiphysics Modeling of Magnet Quench .............................................. 130

6.1 Modeling of Magnet Quench ........................................................................... 130

iv
 6.2 Modeling of Transition ..................................................................................... 130

6.3 Simplification of Composite Wire .................................................................... 131

6.4 ANSYS Simulation of Stress and Strain Development ...................................... 132

6.5 Results .............................................................................................................. 134

6.6 Conclusion ........................................................................................................ 135

Chapter 7. Conclusion ................................................................................................. 136

7.1 Challenge in FEA Modeling............................................................................... 136

7.2 Experimental Comparison ................................................................................ 137

7.3 Future Work ..................................................................................................... 138

INDEX .............................................................................................................................. 140

APPENDIX A. Material Properties ............................................................................... 142

APPENDIX B. Sample APDL Codes............................................................................... 146

B.1. Sample APDL Script for Homogenization of a Symmetry Reduced RVE for Elastic

Properties .................................................................................................................... 146

B.2. Sample APDL Script for Homogenization of Specific Heat ............................... 160

B.3. Sample APDL Script for Homogenization of the Thermal Conductivity ........... 168

B.4. Sample APDL script for Localization ................................................................. 176

B.5. Sample APDL Script for Magnet Analysis ..................................................... 183

References ...................................................................................................................... 225

v
 List of Tables

Table 1.1 : Price of superconducting wire .......................................................................... 7

Table 3.1: Poisson’s ratios of the constituent materials of superconducting wires ........ 52

Table 4.1: Volume fraction of the composite superconducting MgB2 wires ................... 65

Table 4.2 Comparison of the elastic property values using symmetry and antisymmetry

boundary conditions ................................................................................................. 77

Table 4.3 Comparison of elastic modulus calculation ...................................................... 90

Table 5.1. Design specification of 1.5 T conduction cooled MgB2 superconducting

magnet ...................................................................................................................... 98

Table 5.2 Material property of wire constituents and mandrel ..................................... 101

Table 5.3 Summary of the homogenized material property of the composite wire...... 102

Table 5.4 Summary of numerically homogenized material property of the composite

Rec#1027 wire at 300 K .......................................................................................... 116

Table A.1 Poisson's ratio of materials ............................................................................. 142

Table A.2 Modulus of elasticity of materials(GPa) ......................................................... 142

Table A.3 Thermal strain of materials (mm/m) ............................................................. 143

Table A.4 Thermal conductivity of materials (W/m-K) ................................................... 144

Table A.5 Specific Heat of materials (J/Kg-K) .................................................................. 145

vi
 List of Figures

Figure 1.1 Number of MRI scanners per million of population around the world. Data are

from source [2]............................................................................................................ 3

Figure 2.1. Typical configuration of composite superconducting wires. .......................... 19

Figure 2.2. Cylindrical coordinate system used in this work. ........................................... 31

Figure 3.1. Stress state in 3D. ........................................................................................... 46

Figure 3.2. Elastic modulus variation with temperature. [222]–[227] ............................. 51

Figure 3.3. Thermal strain of a bar of length ‘L’. .............................................................. 53

Figure 3.4. Thermal strain of a bar of length ‘L’. .............................................................. 54

Figure 3.5. Thermal strain of constituent materials. [225], [227] .................................... 57

Figure 3.6. Thermal conductivity of constituent materials. [224], [234]–[236] ............... 58

Figure 3.7. Heat Capacity of constituent materials. [224],[237-239] ............................... 59

Figure 3.8. Electrical resistivity of constituent materials. [224], [240]–[243] .................. 60

Figure 4.1. Typical configuration of composite superconducting wires. The wire

constitutes the top volume of length scale ‘L’ .......................................................... 62

Figure 4.2. Microscopic images and CAD models of the superconducting wires for MRI

background magnets (a) Cir#518, (b) Cir#632, (c) Cir#600, (d) Rec#1027. .............. 64

Figure 4.3. Typical configuration of composite superconducting wires. .......................... 66

Figure 4.4. Dimensions of the RVE used for numerical homogenization. ........................ 68

Figure 4.5. Determining the RVE of a composite wire...................................................... 73

Figure 4.6 Reduction of RVE into 1/8th of the volume. ..................................................... 76

vii
Figure 4.7. Steps in determining the representative volume element for a

superconducting magnet bundle made of superconducting metal matrix

composites. (a) Cir 518 RVE is homogenized and 1st principal strain is shown for

applied strain along direction 1. (b) Cir 632 RVE is homogenized and 1 st principal

strain is shown for applied temperature at 300 K. (c) Cir 600 RVE is homogenized

and temperature plot is shown when temperature gradient is applied at the top

and bottom surface of RVE. (d) Rec 1027 is homogenized and temperature plot is

shown for calculation of specific heat. The plot represents the temperature

increase from 300 K. ................................................................................................. 79

Figure 4.8. Applied boundary condition for homogenizing the RVE. From the top left to

the bottom right, the first six boundary conditions are for calculating the stiffness

matrix. Last row in the picture represents the boundary conditions for thermal

expansion coefficient, thermal conductivity (k11), and specific heat respectively. .. 80

Figure 4.9. Homogenized material properties for Cir#518 superconducting wire. ......... 84

Figure 4.10. Homogenized material properties for Cir#600 superconducting wire. ....... 85

Figure 4.11. Homogenized material properties for Cir#632 superconducting wire......... 86

Figure 4.12. Homogenized material properties for Rec#1027 superconducting wire. .... 87

Figure 4.13. Comparison of analytical and numerical homogenization schemes of elastic

stiffness for all four wires. ........................................................................................ 89

Figure 4.14. Comparison of analytical and numerical homogenization schemes for

thermal strains. ......................................................................................................... 91

viii
Figure 4.15 Comparison of thermal conductivity between analytical and numerical

homogenization schemes for different wires. .......................................................... 92

Figure 4.16. Comparison of specific heat between analytical and numerical

homogenization schemes for an MgB2 superconducting wire. ................................ 93

Figure 4.17. Moving across multiple length scales using the numerically homogenized

material properties of the composite superconducting wires. ................................ 94

Figure 5.1. 2D Axisymmetric (left) view of the coil half and isometric view (right) of the

coil bundles assembly. .............................................................................................. 97

Figure 5.2. Manufacturing process of a superconducting solenoid. ................................ 99

Figure 5.3. Wire cross section of 36 filament MgB2 based superconducting wire.

Microscopic picture on left and computer aided designed (CAD) model on right. The

orthogonal directions 1, 2, and 3 are shown in the figure. .................................... 100

Figure 5.4. Comparison of finite element and analytical approach of solving the winding

stress development in coil bundle 5 of the system. ............................................... 103

Figure 5.5. Comparison of finite element vs analytical solution of cool-down (298 K to 10

K) of a solenoid considered by Arp. ........................................................................ 106

Figure 5.6. Comparison of hoop stress variation along the radial direction of a solenoid.

The analytical approach as provided by Caldwell is in good agreement with the two

different FEA approach using ANSYS. ..................................................................... 108

Figure 5.7. Radial Stress along the radial direction of the mid-plane of all five coil

bundles. (Mandrel regions are shaded in gray) ...................................................... 109

ix
Figure 5.8. Hoop stress and strain plot along the radial direction of five bundles of the

system. .................................................................................................................... 110

Figure 5.9. Four different types of stresses and strains on the coil bundle 5 along the

radial direction after the winding of layers around the mandrel is complete. ...... 111

Figure 5.10. 1st and 3rd principal stress ((a), (c)) and strains ((b), (d)) along the radial

distance of the all five coil bundles......................................................................... 112

Figure 5.11. 1st and 3rd Principal mechanical stress ((a) and (c)) and strain ((b) and (d))

after the bundles are cooled down to operating temperature of 10K. .................. 113

Figure 5.12. 1st and 3rd principal mechanical stresses ((a) and (c)) and strains ((b) and

(d)) after the coil bundles are charged with the operating current. ...................... 115

Figure 5.13. 1st principal mechanical strain in all five bundles of the coil at the time of

electromagnetic charging after wire winding and cooling down to operating

temperature. ........................................................................................................... 117

Figure 5.14. Locations of all five coil bundles of the symmetric system and four different

support conditions for winding............................................................................... 119

Figure 5.15. 1st principal strain on all five bundles of the 1.5 T magnet system. Mandrel

regions are shaded in gray. The plot is at the mid-plane from the mandrel inner

surface to the outermost radial location of the bundles........................................ 120

Figure 5.16. Tangential strain development on bundle 5 at the end of each step. ....... 121

Figure 5.17. Maximum shear stress in bundle 5 at different steps of manufacturing and

operations. .............................................................................................................. 122

x
Figure 5.18. 1st principals strain due to different winding support cases after the

magnets are charged to 1.5 T field ......................................................................... 123

Figure 5.19. Hoop strain on bundle 5 with case II .......................................................... 124

Figure 5.20. Localization of the 1st Principal strain and Maximum Shear stress. ........... 126

Figure 6.1. Unit cell for ANSYS simulations to calculate stresses and strains. ............... 131

Figure 6.2. Material properties used in ANSYS simulations: a) Thermal strain; b) Elastic

modulus. ................................................................................................................. 132

Figure 6.3. Strains and stresses calculated in ANSYS. a) Tensile strain in the MgB2

superconductor; b) Shear stress (r-θ component) in the epoxy insulation............ 134

Figure 7.1. Test coil setup. a) Test coil setup with instrumentations b) Cryogenic cooling

system. .................................................................................................................... 137

xi
 Preface

Milk production in a dairy firm has been low recently, and the farmers are

scratching their heads for a viable solution to the problem. Hence, they reach out to the

university scholars to resolve the issues. An engineer, psychologist, and a physicist are

hired as a consultant to find a solution. The group of researchers spent some time to

understand the problem, worked hard for another couple months and came out with their

unique solution separately. The engineer explained the low production is caused by the

smaller pumps that are used to milk the cow. If a larger pump is brought, that will solve

the problem, and milk production will resume to the previous rate. The psychologist

suggested the gloomy mood of the cow is responsible for the low milk production. She

suggested the farm needs painting with natural green color to make the cows happy as

coloring will resemble the fresh green grassland. Now comes the turn of the physicist.

With a very confident voice, the physicist declared, she has the solution. However, it only

works for a cow with a spherical shape and in a vacuum.

The humorous anecdote above is an old cliché, used in different forms for different

perspectives but perfectly sums up the limitations of computational modeling. However,

computational modeling and analysis provide a sense of a system’s performance before

it is built. Computational modeling is precisely where the Finite Element Analysis (FEA)

pitched in and helped in predicting the performance of a conduction cooled MRI magnet.

A perceived helium crisis accelerated the development of conduction cooled MRI systems

starting in 2010. Superconductivity of magnesium diboride (MgB2) was discovered in

2001, and its use in superconducting composite wires was slowly developing. Although
xii
MgB2 was thought of as the future replacement of superconducting MRI magnets,

experimental development was limited due to the concern over significant manufacturing

costs. In this case, computational modeling to predict the performance and analyze the

feasibility has been useful. Therefore, in this dissertation, the steps to model the MRI

magnet system using FEA is discussed from the perspective of strain development, since

a primary concern about MgB2 magnets is the strain sensitiveness. I hope that the work

presented in this dissertation will help other researchers around the world to further

analyze the MRI magnet systems and reach conclusions much quicker than before.

xiii
 Acknowledgement

As a God believer, I praise the lord that He has given me time, energy and opportunity

to pursue and complete the work.

My undergraduate institution Bangladesh University of Engineering and Technology

played an important role in preparing me for the graduate school. As a citizen of the

country— Bangladesh with her developing economy, I am one of the privileged to have

access to the country’s best engineering school for a paltry sum of $100 a year. I

understand that fellow citizens’ tax paying moneys have provided us a quality education.

I am beholden to Bangladesh and her citizens to have put their trust in me.

Before starting my Ph.D., I received my Master of Science in Mechanical Engineering

from the University of Akron. I am grateful to the University of Akron for the opportunities

they had provided.

Case Western Reserve University (CWRU) has provided me access to a lot of scientific

journals, dissertations, many valuable software programs, training resources, and many

research facilities. Also, the intellectual and diverse environment on campus has guided

me to shape my research. I never felt away from home during the time of my Ph.D.

research. I am invaluably indebted to CWRU for providing me opportunities and believing

in my capabilities.

I would like to extend my heartiest gratitude to my magnificent advisers Dr. Michael

Martens and Dr. Ozan Akkus. They have guided me through difficult times to solve

seemingly impossible research problems, eased my frustrations and always helped me

xiv
out with novel ideas while working on those problems. Without their help and guidance,

it would be impossible to finish the work in a timely fashion.

My dissertation committee member, respected and distinguished professor Robert

Brown has advised me a lot regarding my research, future goal and career. I will always

admire his knowledge, his dominance over language and vocabulary. I will strive to be a

person similar to him one day, and I am grateful for what he has enlightened in me.

Dr. Bo Li and Dr. Ya-Ting Tseng Liao have enthusiastically spent their valuable time

discussing my research work. They advised me to a lot to improve my work and make it

stand out among others. I am thankful to them for their efforts.

Dr. Tanvir Baig advised me a lot and guided me through my journey at CWRU. He

speaks the same language, and it had been a boon to discuss novel ideas for four years in

a foreign university in my native language. I am thankful to him for guiding me in difficult

times and showing me the path to success.

I am also grateful to Dr. Robert J. Deissler, who has helped me to understand many

concepts with magnet design and magnet quench. His perception of mechanical and

thermal strain has helped me understand the failure criterion of MgB2 better.

Dr. Laith Sabri advised me with the basics of FEA, ANSYS and anti-symmetry boundary

conditions. I will remember his advice of ‘divide and conquer’ approach for a complicated

FEA problem in my future career. I am grateful to him for teaching me valuable

techniques, and also for the time we shared together.

xv
 Graduate school is a journey where we share our everyday experiences with other

fellow graduate students. Bhumi Bhusal, Mingdong Fan, Charles R. Poole, and already

graduated Zhen Yao were such friend and fellow students who cooperated in a lot of non-

research related activities. Over the years, I have enjoyed talking, working, traveling, and

playing with them. I wish I could always have them beside me. Worthy friends like them

will always make a rough day enjoyable, and I am grateful to them for the time they

shared with me.

I am thankful to the Ohio Third Frontier and National Science Foundation for their

financial support to complete my Ph.D. research work. I believe they will continue

supporting other fantastic research projects for years to come.

Much of my research work was in collaboration with Hyper Tech Research and Ohio

State University’s ‘Center for Superconducting and Magnetic Materials.' David Doll,

Michael Tomsic, and Matthew Rindfleisch have helped me a lot with their industrial

insights of wire design as well as magnet modeling. Dr. Michael Sumption and his

graduate student Danlu Zhang are helping our group with the experimental tests. Their

valuable inputs have made our magnet design easier. I would like to extend my gratitude

to these amazing people.

I would not become an engineer at first hand without the friendly companionship of

my college friends Dr. Md Mahbubul Islam, Mahmudur Rahman Faisal, Naim Hossain, Dr.

Hamim Ahmed, Mahmud Bin Abdus Salam, Sarat Das, Rajib Kumar Saha, and many others.

My friends keep me on track whenever I needed their help, whenever I had to share my

xvi
thoughts about my failures. The journey through the graduate school would not have

been successful without their conscious support.

Dr. Mohammad Akram Hossain and Muhammad Noman Hasan are those of my

college friends who continue to be my graduate school fellow at the University of Akron

and Case Western Reserve University. I appreciate their assistance and amicability

throughout my journey in the graduate school.

Valuable advice from Dr. Mohammed Ziaur Rahman and Dr. Munawar Sultana has

helped me improve my dissertation. I appreciate their effort on making my dissertation

better and flawless.

I would like to acknowledge my grandmother for her continued support and

inspiration as I grew up and made my journey from the early childhood towards the end

of the graduate school. I believe, she will continue to be the encouragement of my life.

Lastly, I would like to thank my younger brother Abdullah Al Momin for his continued

support. He took the responsibility of my parents so that I can live in a foreign country

and finish my work with little interruption.

xvii
 List of Abbreviation

CT – Computational Tomography

CTE – Coefficient of thermal expansion

DSV – Diameter Spherical Volume

FEA – Finite Element Analysis

HTS – High Temperature Superconductor

IROM – Inverse Rule of Mixture

LHe – Liquid Helium

LTS – Low temperature superconductor

MD – molecular dynamics

MMC – Metal matrix composite

MRI – Magnetic Resonance Imaging

Nb – Niobium

Nb3Sn – Niobium and Tin superconducting compound

NMR – Nuclear Magnetic Resonance

ROM – Rule of Mixtures

RUC – Representative Unit Cell

xviii
RVE – Representative Volume Element

SN2 – Solid nitrogen

TEC – Thermal expansion coefficient

xix
 List of Symbols

𝑆 (𝑣) , Compliance Tensor for Voigt Approximation

𝑉 (𝑐) , Volume fraction of constituent component of the composite

𝐶 (𝑐) , Elastic stiffness Tensor of the components of the composite

𝑆 𝑉𝑅𝐻 , Voigt-Reuss-Hill compliance tensor

𝐶𝛼𝛽 , Numerically homogenized stiffness tensor

𝜎𝛼 , Stress along direction 𝛼

𝜀𝛽0 , Applied strain along direction 𝛽

𝛼𝑖𝑗 average thermal expansion coefficient tensor at meso-scale

̅̅̅̅,

𝛣𝑖𝑗𝑘𝑙 , stress influence function tensor

𝛼𝑖𝑗 , average thermal expansion coefficient tensor at micro-scale

𝜎𝑖𝑗𝑟 , local stress tensor

0
𝜎𝑘𝑙 , uniform overall stress tensor

𝑘𝑒𝑓𝑓 , effective thermal conductivity of the composite

𝑓𝑖 , volume fraction of the components of the composite

𝑘𝑖 , thermal conductivity of the components of the composite

𝑓𝑓 , volume fraction of the filament

xx
𝑘𝑓 , thermal conductivity of the filament

𝑘𝑚 , thermal conductivity of the matrix

𝑞𝑖𝑖 , heat flux along direction ii

𝑘𝑖𝑗 , thermal conductivity tensor

𝑇𝑖,𝑗 , temperature gradient in spatial direction

𝐶𝜀 , 𝐶𝑣 Specific heat at constant strain/volume

𝐶𝜎 , 𝐶𝑝 Specific heat at constant stress/pressure

𝑞⃛, change in volumetric heat

𝛥𝑇, change in temperature

𝐶𝑖 (𝑇), specific heat of constituent materials of composite at temperature T

𝐶𝑒𝑓𝑓 (𝑇), effective specific heat of the composite

𝛾𝑒𝑓𝑓 , bulk density of the composite

𝛾𝑖 , specific heat of the constituent of the composite

𝜎𝑟 (𝑎), radial stress at radial location ‘a’ from the inner radial surface of magnet bundle

𝑢𝑟 (𝑎), radial displacement at radial location ‘a’

𝐸𝑚 , modified effective modulus of magnet bundle

𝜎𝜃 , tangential or hoop stress in magnet bundle

xxi
𝜎𝜃𝑖 , tangential stress at i-th layer of magnet bundle

𝜎𝑤 , applied winding stress on the wire during winding process

𝜎𝑤𝑘 , applied winding stress at k-th layer of the magnet bundle

𝜈𝑟𝜃 , 𝜈𝜃𝑧 , 𝜈𝑟𝑧 , Poisson’s ration in the 𝑟𝜃, 𝜃𝑧 and 𝑟𝑧 plane respectively

𝑝𝑘 , pressure on the k-th layer of the magnet bundle

𝑡𝑗 , thickness of the j-th layer of the magnet bundle

𝜀𝑟 , radial strain

𝛼𝑟 , thermal expansion coefficient in radial direction

𝜀𝜃 , tangential or hoop strain

𝛼, thermal expansion coefficient

𝛼𝑟 , thermal expansion coefficient in radial direction

𝛼𝜃 , thermal expansion coefficient in tangential direction

𝑇0 , initial temperature

𝑇𝑓 , final temperature

𝑟, radial distance from the inner surface of the bundle

𝐸𝜃 , Elastic modulus along tangential/hoop direction

𝐸𝑟 , Elastic modulus along radial direction of the bundles

xxii
𝐽, current density in the bundles

𝐵, Magnetic field strength

𝐵𝑧 , axial component of the magnetic field

𝐵𝑟 , radial component of the magnetic field

𝑋𝑟 , body force on the magnet bundles

𝑘(𝑇), thermal conductivity of the material at temperature T

𝑘𝜃 , thermal conductivity of the magnet bundle along tangential or hoop direction

𝑘𝑧 , thermal conductivity of the magnet bundle along axial direction

𝑘𝑟 , thermal conductivity of the magnet bundle along radial direction

𝐸𝑐 , Electric field criterion for quench

𝐼𝑐 (𝐵, 𝑇), critical current for the conductor under applied magnetic field B and

temperature T.

𝑇𝑡 , superconducting to resistive mode transition temperature

𝑇𝑜𝑝 , operating temperature

𝛿𝑤𝑖𝑟𝑒 , thickness of the wire

𝛿𝑖𝑛𝑠𝑢𝑙𝑎𝑡𝑜𝑟 , thickness of the insulator

𝜀𝑥𝑥 , 𝜀𝑦𝑦 , 𝜀𝑧𝑧 strain along direction 𝑥𝑥, 𝑦𝑦, 𝑧𝑧

𝜀𝑥𝑦 , 𝜀𝑥𝑧 , 𝜀𝑦𝑧 , strain in plane 𝑥𝑦, 𝑥𝑧, 𝑦𝑧

xxiii
𝜎𝑥𝑥 , 𝜎𝑦𝑦 , 𝜎𝑧𝑧 stress along direction 𝑥𝑥, 𝑦𝑦, 𝑧𝑧

𝜏𝑥𝑦 , 𝜏𝑥𝑧 , 𝜏𝑦𝑧 , shear stress in plane 𝑥𝑦, 𝑥𝑧, 𝑦𝑧

𝐸𝑥 , 𝐸𝑦 , 𝐸𝑧 , Elastic modulus along direction 𝑥, 𝑦, 𝑧

𝜈𝑥𝑦 , 𝜈𝑥𝑧 , 𝜈𝑦𝑧 Poisson’s ration in the plane 𝑥 − 𝑦, 𝑥 − 𝑧, 𝑦 − 𝑧

𝐺𝑥𝑦 , 𝐺𝑥𝑧 , 𝐺𝑦𝑧 Shear modulus in plane 𝑥 − 𝑦, 𝑥 − 𝑧, 𝑦 − 𝑧

𝜎𝑖𝑗 , stress tensor

𝜀𝑖𝑗 , strain tensor

𝐶𝑖𝑗𝑘𝑙 , stiffness tensor

𝜀𝑡ℎ , thermal strain

𝛥𝐿, change in length due to applied temperature

𝛼𝑠𝑒 (𝑇), instantaneous thermal expansion coefficient at temperature 𝑇

̅̅̅̅̅̅̅̅̅
𝛼𝑠𝑒 (𝑇), average thermal expansion coefficient over a temperature range.

𝑅, overall resistance

𝐴, cross sectional area of the material

𝐿, length of the material

𝜎̅̅,
̅̅𝑖𝑗 average stress tensor

𝜀𝑖𝑗 average strain tensor

̅̅̅,

xxiv
𝑉, volume of the element

𝑢𝑖 , displacement along direction 𝑖

𝑡𝑖 , traction along direction 𝑖

0
𝜀𝑖𝑗 , applied strain tensor

𝜎𝑖𝑗0 , applied stress tensor

𝑆𝑖𝑗𝑝𝑞 (𝑥), material compliance tensor at point 𝑥

xxv
 Multiscale Multiphysics Thermo-Mechanical Modeling of an MgB2 Based

Conduction Cooled MRI Magnet System

Abstract

by

ABDULLAH AL AMIN

The past decade has experienced a surge in the price of liquid helium (LHe) affecting the

cryogenic industry severely. As MRI machines are the most abundant consumer of LHe,

disruption in the market growth is unavoidable if alternative technologies are unavailable.

Replacement of current low-temperature superconductors such as Niobium-Titanium

(NbTi) and Niobium-Tin (Nb3Sn) with an affordable high-temperature superconductor

such as Magnesium diboride (MgB2) is a promising solution. However, strain sensitivity of

the recently discovered superconductor MgB2 places a strict design limit on an MRI

magnet. Since the manufacturing of the MRI magnet costs millions of dollars, a

computational model is sought to analyze the feasibility of the novel conduction cooled

MgB2 based full body MRI magnet. MRI magnet bundles (solenoids) consist of commercial

grade superconductors which are a composite metal matrix. Thus computational

modeling of MRI magnet systems spans across multiple scales and different physics fields.

In this work, a complete multiphysics-multiscale approach to analyzing the feasibility of a

next-generation conduction-cooled MRI magnet has demonstrated acceptable strain

(<0.2%) and stress (<100 MPa) development, suggesting the practicality of such system.

xxvi
A complete non-linear finite element analysis (FEA) model for 1.5 T full body MRI magnet

has been developed using ANSYS, and a prototype magnet bundle is under investigation

for experimental tests at Ohio State University’s ‘Center for Superconducting and

Magnetic Materials.' The experimental results are promising and assure this novel

technology as a replacement to the current LHe cooled MRI magnets.

xxvii
 Chapter 1. Introduction to Magnetic Resonance Imaging

“Read, and your Lord is the most generous, who taught by the pen, taught people that

which he did not know” — Quran 96: 3-5

1.1 Introduction

MRI, an acronym for Magnetic Resonance Imaging, is a non-invasive imaging

technique based on the nuclear magnetic resonance (NMR) of particle nuclei,

predominantly the proton in a hydrogen atom. The word ‘magnetic’ refers to the

interaction between an external magnetic field, and the magnetic moments of the nuclei.

The word ‘resonance’ refers to the excitation of the magnet moments via an applied radio

frequency (RF) magnetic field that is resonant with the natural precessional frequency of

the nuclear spin when placed in a magnetic field. Larmor frequency is the frequency at

which the magnetic moment precesses about the direction of the externally applied

magnetic field. The measurement of the magnetization created by the nuclei is used to

contrast an image of an object, such as a human for instance.

A significant milestone in the development of MRI was 1973 work by Paul

Lauterbur and Peter Mansfield in which they independently introduced a linear magnetic

field gradient used to correlate the Larmor frequency with a spatial location. With a

magnetic field gradient, portions of the sample located in regions of the higher magnetic

field will have a higher Larmor frequency. As an interesting side note, the acronym MRI

(magnetic resonance imaging) evolved from NMR; the word ‘nuclear’ fell out of use to

1
Chapter 1

avoid the incorrect association with nuclear radiation and its damaging effects on human

tissue [1].

The quality of an MRI image improves as the strength of the primary magnetic

field increases. Therefore, higher magnetic field systems, typically 1.5 or 3 T MRI units

have come out as the most popular machines. MRI systems with magnetic field strengths

of 0.5, 7 and 9.4 T also exists, but lower field systems generate lower-quality images, and

the higher field systems incur an increased cost of manufacturing. Hence, the cost-

magnetic field optimization settled for 1.5 T or 3.0 T as most common imaging units.

Magnetic field generation at such high level requires large current (~200 Amps) within the

magnet wire. The ability of superconductors to conduct current at high density with

minimal Joule heat loss1 is the underlying reason behind their primary choice in MRI

machines. A superconducting wire is the prerequisite to minimize the heat generated due

to the ohmic losses of a normal conducting wire. Without the availability of such

superconducting wire, it would be impractical to operate a 1.5 T or 3 T magnet due to the

heat generated from the ohmic losses of a normal conducting material.

1
Energy loss occurring due to the resistance of the wire.

2
 Introduction to Magnetic Resonance Imaging

Figure 1.1 Number of MRI scanners per million of population around the world. Data

are from source [2]

1.2 Use of MRI Units

MRI is a noninvasive high-quality imaging technique that does not produce

harmful radiation. With its ability to image non-ossified biological tissues and organs, and

provide a high contrast between types of soft tissue, MRI has the edge over other

conventional imaging techniques such as X-ray, Computational Tomography (CT), and

Ultrasound Imaging. In comparison to ultrasound imaging, MRI can distinguish between

different tissues, such as fat, and water, for instance. Thus, due to its advantages over

other imaging techniques, MRI scanning is a favorite imaging technique that is adopted

worldwide. Unfortunately, the high cost of manufacturing and operation has limited the

access of patients to MRI, especially in the developing world. Figure 1 shows the number

of MRI units worldwide per million population [2]. From the figure, it is clear that access

3
Chapter 1

to MRI is restricted and significant improvement in the technology to cut down the

associated cost is desirable.

1.3 MRI Market Prices

Typical MRI systems cost vary significantly depending upon their field strength and

capabilities. Usually, low field or open MRI machines cost less than the high field closed

MRI counterparts. Low field MRI, with field strength 0.2 ~ 0.5 T, can cost as little as

$150,000. However, prices may be as high a $3 million for a state of the art 3 T system

[3], [4]. In addition to the system, MRI units usually need a suite in the hospital that meets

safety requirements. Construction of these suites can add additional hundred thousands

of dollars. On top of that, staff clinicians and doctors charge for an MRI exam, and

radiologists are needed to interpret the images. These accumulated costs contribute to

price range of $511 to $2,815 for a single MRI scan [5]. Thus, cost is the reason why the

use of MRI is limited even though it has widespread popularity as an imaging modality.

A typical MRI magnet is constructed with superconducting wire and uses liquid

helium (LHe) to cool the superconductor down to cryogenic temperature (~ 4 K). MRI

magnets use predominantly NbTi or Nb3Sn superconducting wires to construct the coil

bundles, with critical temperatures of 9 K [6] and 18.3 K [7] respectively. However, MRI

magnets operate at a temperature of 4.2 K to improve the operating efficiency. At this

cryogenic temperature, Helium is in liquid form since the boiling point is 4.2 K and the

melting temperature is 0.95 K [8]. No other material or element can remain liquid at such

low temperature. Therefore, use of LHe is indispensable to convectively cooled NbTi or

4
 Introduction to Magnetic Resonance Imaging

Nb3Sn based MRI magnets. A typical 1.5 T MRI system requires about 2000 L of LHe to

cool the system to the operating temperature initially. Also, an uncontrolled temperature

rise due to mechanical, electrical or magnetic disturbances may initiate a “quench2”. A

quench is a phenomenon where the temperature in a portion of the magnet suddenly

rises beyond the critical temperature making that part of the magnet resistive which

generates ohmic heating. Over the lifespan of an MRI machine, a magnet may experience

several quenches whether intentional or unwanted. Due to the heat generated during the

quench, the LHe becomes a gas and must be vented to the atmosphere to avoid explosive

pressure buildup. As a result, the lost LHe must be replenished, and maintenance cost

goes up. In total, in an MRI manufacturing and operation part of the budget $60,000 ~

$100,000 is included solely for LHe [9].

In recent years, LHe has experienced a price hike attributed to supply shortage.

Since the reserve of natural minerals are limited and only a handful of processing plants

refine LHe, the prices have increased fourfold compared to 1999. In 2016, a total of 47

million cubic meters of Grade-A helium was consumed out of which, 14.1 million cubic

meters (30%) were used in MRI industry alone [10]. Thus, a significant dependence on

LHe with an increased number of MRI units will contribute to the price increases and will

make MRI scans more expensive if no alternative for the use of LHe becomes available.

2
In superconducting magnet research area, quench is the sudden rise in the temperature contrary to the
widespread notion of sudden temperature drop.

5
Chapter 1

1.4 Alternative to LHe

The core of an MRI machine, the background3 magnet, is designed and built with

superconducting wire due to the demand for high magnetic field strengths of 1.5 T or

more. With approximately 35,000 machines installed [11], [12] worldwide, the MRI

industry is the leading market for superconducting wires. Currently, MRI magnet wires

are mostly made of niobium-titanium (NbTi) superconducting wire, which has a critical

temperature of 9K [13]. Most popular clinical MRI systems operate at a magnetic field

strength of 1.5 Tesla to 3.0 Tesla over the center of the imaging region, the “diametrical

spherical volume” (DSV)4. With the burgeoning market of the MRI industry expected to

reach $7.4 billion [14], the rising cost of LHe as discussed in the previous section will

contribute to the overall price and on-going maintenance of an MRI system. This

expanding market for MRI coupled with decreasing helium reserves [15], has motivated

the MRI scientific community to look for a desirable alternative to LHe in the design and

operation of the background magnet.

One of the alternatives to a “wet” magnet filled with LHe is the design of a “dry”

magnet using solid nitrogen (SN2) cooled MRI magnet [9], [16], [17]. The enthalpy of

melting SN2 provides a thermal stability and longer warm-up period under power outage

[9], [10]. Since these systems must still be cooled to 4.2 K, a cryocooler is necessary even

for SN2 systems. Another alternative is the design of a conduction cooled MRI magnet. In

3
Background magnets are responsible for generating uniform high strength magnetic field in an MRI.
4
DSV is the region where certain magnetic field homogeneity is maintained. For a 3 T full body MRI, typical
DSV is 40 cm with a field homogeneity of <1 ppm.

6
 Introduction to Magnetic Resonance Imaging

these designs, the MRI magnet is conductively cooled using two stages of cryocooling.

Such an MRI system would require only about 1-3 liters of LHe (as compared to 1000-

2000 for a wet magnet) to set up the refrigeration system and cool down the cold head.

The magnets are cooled [18], [19] via heat conduction to the cold head of the cryo-cooler.

High temperature superconductors (HTS) such as MgB2 (Tc = 39K) [20], YBCO (Tc = 93K)

[19], [21], BSCCO (Tc = 108K) [22] and Bi-2223 (Tc = 90K to 95K) [23] are all good candidates

for such conduction cooled magnet designs since their critical temperatures are much

higher compared to NbTi. One advantage of using HTS is the increased temperature

margin

Table 1.1 : Price5 of superconducting wire

Wire Technology (dia

Price ($/m) Price (kA/m at 4T and 4.2 K)
0.8 mm)

MgB2 2.04 ~ 3.59 [16], [24] 5.19 ~ 12 [25], [26]

YBCO 42.87 [24] 140 [27] (at 2 T & 30 K)

BSCCO 18.55 [24] 62.23 [26] ( at 2 T & 30 K)

Nb3Sn 7.78 [24] 11 ~ 58.76 [26] ( at 5 T)

NbTi 1.42 [24] 2.7 ~ 25.52 [26]

5
Prices are adjusted for inflation for the year 2017 using the calculation provided by
http://money.cnn.com/calculator/pf/inflation-adjustment/

7
Chapter 1

between the critical temperature and the operating temperature. Due to the higher

critical temperature, it is possible to operate the magnets safely at relatively higher

temperature compared to the conventional NbTi or Nb3Sn systems. The added

temperature margin allows for higher temperature gradients within the magnet coils

which help facilitate conduction cooling system. The higher operating temperature also

cuts down the power required to cool the system and saves a significant part of the cool-

down cost. Also, the cost is an issue when considering HTS where MgB2 is ahead with a

price of approximately $2/m as compared to $20/m for Bi2223 and $40/m for YBCO [16].

The prices of different types of superconductors are summarized in Table 1.1.

1.5 Conduction Cooled MgB2 MRI System

A conduction cooled MRI system removes the heat in the magnet entirely through

the conduction process between the magnet coils and the cryocooler. Therefore, a

conductively cooled MRI magnet system eliminates the need for LHe except for the use

as refrigerant within the cryocooler. Several works have demonstrated the feasibility of

such superconducting wires for MRI magnet designs, and development of conduction-

cooled designs for a full body MRI magnets is an active area of interest [18], [21], [23],

[28]. Among these available options, the MgB2 superconducting wire is of particular

interest due to its potential application in full body MRI industry, small mechanical and

thermal anisotropy and high critical temperature of 39 K [29]. Since its discovery in 2001,

several works have been published showing the viability of MgB2 for MRI magnet designs

operating at a temperature of 10 K or higher [20], [30]–[33]. Development of MgB2 wires

to date has been experimental in context, assessing different types of monofilament [34]

8
 Introduction to Magnetic Resonance Imaging

and multifilament wire configurations [35], [36]. Development of MgB2 wire continued,

and recent experiments have demonstrated increased ultimate failure strain limit, and

critical current carrying capability [37]–[43].

1.6 Recent Efforts on MgB2 Systems

The high critical temperature, low manufacturing cost, and excellent performance

under the higher applied magnetic field, combined with a high thermal mass density, have

made MgB2 an attractive material for a superconductor. Hence, a significant effort has

been put into making high-performance superconducting wires out of MgB2. The leading

methods for manufacturing the wire are the powder in tube [44] technique for

manufacturing monofilament superconducting wire and MgB2 tapes [45]. Since both wire

and tape have become commercially available, researchers from both academia and

industry has been interested in developing MgB2 based magnet systems.

Using a 46 m commercial superconducting tape from Columbus Superconductor,

Stanvall et al. built a react-and-wind solenoid [46]. Zhang et al., tested thermal stability of

MgB2 based system. with a relatively longer 60 m wire [47], [48]. The field strength of 1.5

MgB2 magnets was tested for quench [49] analysis and persistent mode operation [50] by

Li et al. Larger 3 T field strength magnet with a bore diameter of 250 mm magnet

manufactured by 3 km wire and wind-and-react process experienced large remnant

voltage [51], [52]. As a result, persistent mode operation was not possible. ASG

superconductors manufactured double pancake 0.5 T open MRI systems and

demonstrated scan images using the MgB2 tapes from Columbus superconductor [53],

9
Chapter 1

[54]. Francis Bitter Magnet Laboratory at MIT built a prototype of 700 mm dia 10 coil

MgB2 system using SN2 [32], [55]. However, the superconducting persistent joint was the

challenge that Yao et al. [55] solved, but later it was found to work only for monofilament

wires [34], [56].

1.7 Challenges of MgB2 Systems.

While there is an increasing focus on MgB2 wires, the low failure strain of the MgB2

wires [38], [57]–[59] increase the demand for knowledge of the strain distribution during

coil design and magnet operation. A magnet experiences strain development caused by

the applied pretension at the time of the winding process. Also, anisotropic deformation

during cooling to cryogenic temperature (~10 K) caused by thermal contraction attributed

to the mismatched thermal strain of wire constituents and mandrel material induces extra

strain on the bundles. At the time of operation, the magnet is electrically energized, and

the Lorentz force generates further strain in the wire [60]. Also when the magnet is in

operation, the unwanted phenomenon may trigger a quench. It is important to take

protective measures against quench. Otherwise, the uncontrolled temperature may

result in significant damage, sometimes resulting in complete destruction of the magnet

system.

1.8 Thesis Overview

In Chapter 1, an introduction to MRI and the basic working principles are

discussed. Additionally, the prices of MRI units and the patient accessibility are presented.

10
 Introduction to Magnetic Resonance Imaging

Alternatives to existing MRI technologies are briefly mentioned before efforts with MgB2

are discussed.

Chapter 2 summarizes the work of other researchers. The literature review is in

no way comprehensive since the research area is continuously developing. The entire MRI

magnet system is a massive topic to be covered in a single dissertation, and it would be

even difficult to fit it in a book. Although, Chapter 2 is expected to serve as an introduction

to the several aspects of MRI background magnet system manufacturing, energization

and quench.

Material properties are an important input to any computational modeling. An

MRI background magnet is made up of composite superconducting wires. Chapter 3

discusses the various materials properties required to model the system computationally.

Often, it is challenging to obtain the material properties at cryogenic temperature. In this

chapter, material properties are compiled and presented for a different temperature

ranges from room temperature to cryogenic temperature 10 K, a temperature at which

the magnet is expected to operate.

Chapter 4 explains the modeling of composite superconducting wire for different

material properties required for magnet analysis. In a sense, this chapter is the first stage

of the multiscale magnet analysis. The chapter discusses techniques to numerically

homogenize the composite wires at a different temperature. The homogenized

properties are then used to analyze the magnet in the later chapters.

11
Chapter 1

MRI magnet manufacturing and operation consists of different steps. The steps to

build and electromagnetically energize the magnet are discussed in Chapter 5. FEA

modeling results for different stages of the magnet manufacturing and operation are

presented in the chapter. Failure criteria is defined from published articles, and the

magnet model is verified against the strain and stress limits. Analysis of the magnet results

presented in this chapter explains the feasibility of conduction cooled MgB 2 MRI magnet

system.

Chapter 6 explains the quench phenomenon of the magnet. Strain development

due to the quench is analyzed, and the results are presented. The developed strain

superposed with the accumulated strain from the manufacturing and energization steps

determines the feasibility of the entire magnet system as a whole.

Chapter 7 concludes the work that is presented in the previous six chapters. This

chapter also discusses the scope of future work and summarizes the experimental efforts

that is ongoing at Ohio State University’s ‘Center for Superconducting and Magnetic

Materials.'

1.9 Related Publications

Journal Publications

1. Deissler R J, Baig T, Charles P, Amin A A, Doll D, Tomsic M, and Martens, "A

Computational Study to Find an Optimal RRR Value for a 1.5 T Persistent-Mode

Conduction-Cooled MgB2 MRI Magnet from a Quench Protection Point of View.",

IEEE Trans. Appl. Supercond, Volume 99, Issue 4, June, 2017

12
 Introduction to Magnetic Resonance Imaging

2. Baig T, Amin A A, Deissler R J, Sabri L, Poole C, Brown R W, Tomsic M, Doll D,

Rindfleisch M, Peng X, Robert Mendris, Ozan Akkus, Michael Sumption, and

Michael Martens, "Conceptual designs of conduction cooled MgB2 magnets for 1.5

and 3.0 T full body MRI systems.", Superconductor Science and Technology,

Volume 30, Issue 4, March, 2017

3. Amin A A, Baig T N, Deissler R J, Sabri L A, Doll D, Tomsic M, Akkus O, and Martens

M A, "Mechanical Analysis of MgB2 Based Full Body MRI Coils Under Different

Winding Conditions.", IEEE Trans. Appl. Supercond, Volume 27, Issue 4, June, 2017

4. Deissler R J, Baig T, Poole C, Amin A A, Doll D, Tomsic M, and Martens M,

"Numerical simulation of quench protection for a 1.5 T persistent mode MgB2

conduction-cooled MRI magnet.", Superconductor Science and Technology,

Volume 30, Issue 2, December 2016

5. Amin A A, Baig T, Deissler R J, Yao Z, Tomsic M, Doll D, Akkus O, and Michael

Martens, "A multiscale and multiphysics model of strain development in a 1.5 T

MRI magnet designed with 36 filament composite MgB2 superconducting wire.",

Superconductor Science and Technology, Volume 29, Issue 5, March, 2016 .

Conference Proceedings

1. Amin A A, B Bhusal , TN Baig , RJ Deissler , L Sabri , O Akkus , and MA Martens, "A

comparative study of coil winding techniques of a full body 1.5 T MgB2 based MRI

magnets.", ISMRM 25th annual meeting & exhibition, Hawaii, USA April, 2017

2. Amin A A, TN Baig, RJ Deissler, L Sabri, D Doll, M Tomsic, O Akkus, and MA Marten,

"Effect of Mechanical Support Conditions of Winding on the Strain Development

13
Chapter 1

of a Composite MgB2 Based Full Body MRI Coil.", Applied Superconductivity

Conference, Denver, Colorado, USA October, 2016

3. RJ Deissler, TN Baig, CR Poole, Amin A A, D Doll, M Tomsic, and M Martens, "A

Computational Study to Find an Optimal RRR Value for a 1.5 T Persistent-Mode

Conduction-Cooled MgB2 MRI Magnet from a Quench Protection Point of View.",

Applied Superconductivity Conference, Denver, Colorado, USA October, 2016

4. Amin A A, B Bhusal, TN Baig, RJ Deissler, L Sabri, O Akkus, and MA Martens,

"Variation in strain characteristics for multiscale multiphysics models of a 1.5T

conduction cooled MRI system based on a 36 filament MgB2 composite wire.",

ISMRM 24th annual meeting & exhibition, Singapore City, Singapore May, 2016

5. Amin A A, TN Baig, Z. Yao, and MA Martens, "Stress and Strain Sensitivity Study of

1.5T Conduction Cooled MgB2 Magnet Design.", ISMRM 23rd annual meeting &

exhibition, Toronto, Canada May, 2016

1.10 Conclusion

Extensive researches to establish MgB2 systems as an alternative to conventional

full-body MRI system is ongoing. However, it is critical to understand the strain

development in the magnetic coils of MgB2 because of the low failure strain of the

material. A full body MRI system is significantly larger compared to the smaller lab

developed research magnets. Since critical current6 carrying capability is related to the

applied strain for MgB2 wires, it is necessary to understand the strain development of a

6
Maximum amount of current a superconductor can carry as a superconductor. Beyond the critical current
limit, the wires turn resistive again.

14
 Introduction to Magnetic Resonance Imaging

magnet manufacturing and operation process. While Finite Element Analysis (FEA) of NbTi

superconducting magnets has been carried out by researchers [61]–[67], mechanical

strain analysis of a conduction cooled MgB2 based superconducting solenoid magnet is

yet to be explored. Furthermore, existing magnet modeling techniques rely on the

simplified material modeling of the composite wire. Thus, a composite modeling of the

superconducting wire is required for more extensive analysis. The purpose of this study

is to develop a multiscale and multiphysics model to predict the final stress state in 1.5 T

react-and-wind MgB2 MRI system following the winding, cooling, electromagnetic

energization and finally the quench process.

15
 Chapter 2. Literature Review of MRI Magnet Modeling

“If I have seen further it is by standing on the shoulders of Giants” — Isaac Newton.

2.1 Literature Review

A superconducting magnet technology for MRI consists of designing,

manufacturing, and operation. In each stage, the magnet induces a significant stress

development. Typically, a stress generation of ~400 MPa has been predicted for 11.7 T

Iseult MRI when operating at a temperature of 4 K [68]. Before the discovery of

superconductivity in 1911 by Kamerlingh Onnes [69], most solenoids were of low

magnetic field strength, and the concerns with electromagnetic stress1 were negligible.

In the early years of the superconductive solenoid, scientists were on the lookout for new

and novel ways to generate higher magnetic fields [70], [71]. Those early days of high

magnetic field research, use of electric pulse was prevalent to avoid the high heat

generation in coils attributed to Joule heating [72]. This was a countermeasure against

failure due to thermal strains of the coil. However, failure due to sizeable electromagnetic

force was not of concern until high magnetic field solenoids are of interest. Cockroft et al.

[72] was the first to consider a simple solenoid and derived equations to calculate the

stress in radial and tangential direction by considering the force derived using magnetic

1
Stresses in the solenoid due to electromagnetic force.

16
 Literature Review of MRI Magnet Modeling

field components found from solving Maxwell’s equations. Stress from other sources such

as winding, cool-down and quench development is also significant, and in the early days

were of much simpler origin. The wires used for magnetic field generation in the early

days were homogenous materials made up of copper or other low resistance conductors.

As the research on high magnetic field strength intensified, superconductors were

brought into the scenario around 1955 [73], much later than the discovery of

superconductivity in 1911. The very first demonstration of superconductivity was with

Niobium (Nb) wire at 4.2 K. Six years later in 1961, Niobium-Tin (Nb3Sn) was demonstrated

to carry critical current densities much higher than Nb wires and could produce a

magnetic field of 8.8 T [7]. Due to the brittleness of Nb3Sn, it was difficult to manufacture

a long strand of wire. Thus Nb3Sn wire was built inside a Nb tube, and composite

superconducting wires came into existence [7]. Later on, multifilament composite

superconducting wires demonstrated higher performance in comparison and application

become widespread [74]. Thus, the composite nature of the superconducting wires

meant there was sufficient ductility to carry the critical current carrying capability. This

composite structure is also beneficial for the production of the strain sensitive MgB2

based superconductors. However, it became increasingly difficult to computationally

model and predict the strain scenario of a superconducting solenoid made of

multifilament composite superconducting wires. Fortunately, there are now existing

composite mechanics theories that can be successfully implemented in the modeling of

superconducting composite wire material properties as a homogenized structure. The

characteristics behavior of these homogenized wires can then be fed into the magnet

17
Chapter 2

manufacturing and operation process to estimate the strain development and check the

feasibility of the system regarding mechanical integrity. In this chapter, the details of

modeling the composite superconducting wires, magnet manufacturing and quench

modeling will be detailed.

2.2 Modeling of Wire Material Properties

Superconducting wires are composites made up of superconducting filament

embedded within a matrix made up of regular conductors, i.e., copper. The importance

of a copper matrix is realized during quench, when the superconductor in the wire

transitions from the superconducting to the normal state, and the large current then

passes through the material in the matrix. As the current travels through the matrix, it

generates ‘Joule heating,' which raises the temperature of the magnet causing thermal

stress which might destroy the magnet.

The matrix also helps protect superconducting wire damage since the matrix

materials are capable of sustaining considerable stresses, more than the superconductor

alone. Also, a comparatively larger thermal conductivity assists in quicker heat dissipation

which is beneficial for quench detection and protection. There is a sheath layer at the

outer periphery of the wires. Usually, the sheath layer is electrically insulating, preventing

unwanted current conduction in the lateral direction. In most cases, there is another extra

layer of Monel2 sheath surrounding the copper matrix that takes most of the mechanical

loads, allowing more room for strain development for strain-sensitive superconductors.

2
Nickel and copper alloy with ~67% of nickel.

18
 Literature Review of MRI Magnet Modeling

Figure 2.1 shows typical superconducting composite wire configurations. Due to their

composite structure, and absence of straight-forward periodic boundaries, it is difficult to

estimate the material properties (elastic properties, thermal expansion coefficients,

thermal conductivities) using simple techniques. Moreover, high strain sensitivity of MgB2

wires requires accurate estimation of material properties for MRI background magnet

modeling. However, considering every delicate detail of each filament will introduce

enormous amount of unknown, making it challenging to solve, even for the latest

supercomputing clusters. Thus, numerical homogenization of composite wire is a

Figure 2.1. Typical configuration of composite superconducting wires.

reasonable approach, and a means to obtain material properties of the wire as a whole,

with the only expense being acceptable error in estimation.

2.2.1 Modeling of Elastic Properties

The composite superconductors for conduction cooled MgB2 MRI background

magnet consist of repetitive microstructures. Numerical homogenization of this repetitive

microstructure may provide effective elastic properties if they are considered as a unit

cell with appropriate boundary conditions for magnet modeling. Asymptotic

homogenization [75] is an appropriate method; which is based on the superposition of

19
Chapter 2

global and periodic local perturbations. Explanation of the concepts of asymptotic

homogenization is available elsewhere [76]–[78]. The method is effective in modeling

elastic properties of the superconducting wire at multiple scales [79]–[81]. However, if

the microstructures are not repetitive; regions of possible effective behavior regarding

upper and lower bounds for elasticity are introduced depending on some microstructure

characterizing parameters. These methods date back to simple work of Voigt [82] and

Reuss [83]. The difference between these two methods varies under the assumption that

either the strain or stress tensor remains constant throughout the scale of the

microstructural constituent component of the composite respectively. The Voigt method

to approximate the elastic properties is written as

𝐶 (𝑣) = ∑ 𝑉 (𝑐) 𝐶 (2.1)

𝑐=1

In the equation, volume fraction, elastic stiffness and elastic compliance tensors

of the constituents are denoted by 𝑉 (𝑐) , 𝐶 (𝑣) and 𝑆 (𝑣) respectively. Similarly, Reuss

method to estimate the elastic properties is written as

𝑆 (𝑣) = ∑ 𝑉 (𝑐) 𝑆 (𝑐) (2.2)

𝑐=1

The Voigt method is also known as the “Rule of Mixture” or “ROM, ” and the Reuss

method is known as the “inverse rule of mixture (IROM)” in the superconducting research

community. Later, Hill [84] demonstrated that Voigt’s approximation and Reuss's

approximation are the upper and lower bounds of the true effective stiffness tensor

20
 Literature Review of MRI Magnet Modeling

irrespective of the geometry of the composite. Improvement of these two methods was

proposed by Hill [84] and Paul [85]; their techniques introduced bounds depending on

polycrystal orientation distribution functions. In simple terms, Hill proposed the stiffness

of a composite is the average of the Voigt and Reuss bounds.

1
𝑆 𝑉𝑅𝐻 = (𝑆 𝑉 + 𝑆 𝑅 ) (2.3)
2

In the equation 𝑆 𝑉 , 𝑆 𝑅 , 𝑆 𝑉𝑅𝐻 are the elastic stiffness values of Voigt, Reuss and

Voigt-Reuss-Hill approximation respectively. More refined and narrower bounds,

introduced by Hashin and Shtrikman [86]–[88] improved the material model for property

estimation.

Modeling of the nonlinear constitutive behavior of composite also requires

recognition of representative volume element (RVE) or representative unit cell (RUC) with

appropriate boundary conditions. Modeling of the rigid perfectly plastic composite was

introduced by Bishop and Hill [89], [90] extending the work of Taylor [91], [92]. The

technique constitutes a generalization for non-linear composite in the Voigt-Reuss-Hill

bounds. Extension of self-consistent [93]–[96] methods by incremental procedure [97],

simplified secant methods [98], [99] and extension of Hashin-Shtrikman [100]–[102] are

also suitable to model the non-linearity of composite material behavior. Other variational

principles [103]–[105] considers power-law [106] and elastic-perfectly-plastic response

[107] to model non-linearity in composites. However, estimating the linear elastic

21
Chapter 2

response of a composite structure containing fiber3 and matrix4 is much more

straightforward and requires applying a unit strain in each of the six-directions

sequentially; details of the techniques have been elaborated in references [108]–[110].

The components of the stiffness tensor C are calculated from each of the equations as

1
𝐶𝛼𝛽 = ∫ 𝜎 (𝑥 , 𝑥 , 𝑥 )𝑑𝑉 𝑤𝑖𝑡ℎ 𝜀𝛽0 = 1 (2.4)
𝑉 𝑉 𝛼 1 2 3

In the equation, 𝜎𝛼 is the calculated stress in the macroscopic model for any

applied unit strain 𝜀𝛽0 . Making use of finite element and Gauss-Legendre quadrature, it is

possible to estimate the components of the C matrix. However, the commercially

available FEA program ANSYS makes it easier to calculate the value of integral as stress

values for each element can be calculated individually and then find a summation.

2.2.2 Modeling of Thermal Expansion Coefficient

Modeling of thermal expansion coefficients for composites started later than the

effective mechanical property estimation. A comprehensive review of the fundamental

composite mechanical property estimation by Hashin [111] excluded the thermal

expansion coefficient approximation. Before Schapery [112] discussed the energy

principle method to approximate thermal expansion coefficient, limited work in literature

for isotropic, particulate-filled composites could be found [113]. A first important work

on thermal expansion coefficient was by Levin [114] in a Russian journal which was an

3
The relatively smaller volume fraction inclusion in a composite
4
Matrix has comparatively higher volume fraction in a composite

22
 Literature Review of MRI Magnet Modeling

extension of Hill’s method [115]. Levin formula provides the overall thermal strain vector

using principal of virtual work as

1 𝑇
𝛼𝑖𝑗 =
̅̅̅̅ ∫ 𝛣𝑖𝑗𝑘𝑙 𝛼𝑖𝑗 𝑑𝑉 (2.5)
𝑉
𝑉

In the equation, ̅̅̅̅

𝛼𝑖𝑗 and 𝛼𝑖𝑗 are the average thermal expansion coefficient tensors

at meso-scale and at micro-scale. 𝛣𝑖𝑗𝑘𝑙 is the stress influence function tensor. 𝛣𝑖𝑗𝑘𝑙 relates

0
uniform overall stress field 𝜎𝑘𝑙 and local stress 𝜎𝑖𝑗𝑟 using the following equation [116],

𝜎𝑖𝑗𝑟 = 𝛣𝑖𝑗𝑘𝑙 𝜎𝑘𝑙

0
(2.6)

Over the years, these techniques improved and facilitated modeling of

heterogeneous multiphase composites in different ways. Takao [117] formulated

effective thermal expansion coefficient for anisotropic short fiber-reinforced composite

based on Eshelby’s [118] equivalent inclusion method. The first FEA approach to estimate

effective thermal expansion coefficient was by Adams [119] for unidirectional

composites. Later efforts to utilize FEA intensified and were applied to model metal

matrix composite [120] as well as superconductors [121]. Besides, FEA methods based on

RVE dilute approximation [122], self-consistent [123] and modified Mori-Tanaka method

[124], [125] could be used for thermal strain tensor estimation. However, FEA methods

are extensively used in this work since ANSYS is the chosen approach for analyzing the

magnet and the wires.

23
Chapter 2

2.2.3 Modeling of Thermal Conductivity

The first approach to estimate the thermal conductivity of a composite was by

Bruggeman [123]. The Bruggeman model is expected to have better accuracy for high

filler composites [126] as a mean field approach is used to estimate the interactions of

randomly dispersed inclusions. Before Bruggeman’s self-consistent method, simple linear

mixing rule (parallel model)5 and series model have been used to estimate the thermal

conductivity of the composites. These two methods are used for thermal conductivity

calculation for composites with limited accuracy for some definite microstructures. The

linear mixing rule is based on the assumption that the temperature gradient is uniform

for the composite and that the heat flux is the weighted sum of the fiber and matrix. The

overall thermal conductivity of the composite as calculated by linear mixing rule is given

in equation form as:

𝑘𝑒𝑓𝑓 = 𝛴𝑓𝑖 𝑘𝑖 (2.7)

In the equation, 𝑘𝑒𝑓𝑓 is the effective thermal conductivity of the composite, 𝑓𝑖 is

the volume fraction of the composite constituents, and 𝑘𝑖 is the thermal conductivity of

those components. The series model considers the heat flux as constant throughout the

composite whereas the total thermal gradient of the composite is the weighted average

of each of the components. In equation form the series thermal conductivity

approximation of the composite is represented as:

5
Parallel model of composite thermal conductivity calculation

24
 Literature Review of MRI Magnet Modeling

−1
𝑓𝑖
𝑘𝑒𝑓𝑓 = (𝛴 ( ) ) (2.8)
𝑘𝑖

Maxwell-Garnett (MG) [127] is also a widely used technique to estimate the

effective thermal conductivity. The technique is developed for spherical inclusion that is

non-interacting with the matrix of the composite. Fricke [128] extended the method to

model the composite that has spheroidal inclusions. Further developments [129], [130]

of such thermal conductivity estimation focused on including the interfacial interaction

between fiber and matrix. However, the inability of these methods to predict the thermal

conductivity of a composite with the higher volume fraction of the filaments popularized

the use of Bruggeman [123] model. For a binary composite, effective thermal conductivity

is calculated using the equation:

𝑘𝑓 − 𝑘𝑒𝑓𝑓 𝑘𝑚 − 𝑘𝑒𝑓𝑓
𝑓𝑓 ( ) + (1 − 𝑓𝑓 ) ( )=0 (2.9)
𝑘𝑓 − 2𝑘𝑒𝑓𝑓 𝑘𝑚 + 2𝑘𝑒𝑓𝑓

In the equation, 𝑓𝑓 is the volume fraction of the filament, 𝑘𝑓 and 𝑘𝑚 are the

thermal conductivity of the filament and matrix respectively. There had been other

theoretical approaches [87], [93], [131]–[137] to improve the prediction of thermal

conductivity of the composite with different filler size, shape and volume fraction.

However, most of the analytical approach is limited to considering the interfacial

interaction with the matrix. Moreover, when the structure of the composite is relatively

well defined, it is more appropriate to use the FEA technique to model the composite

rather than assuming a random dispersion of the filaments.

25
Chapter 2

Numerical methods such as Finite Difference (FD) and FEA provides an edge to

calculate effective thermal conductivity when interfacial interaction and organized

filament structured composites are considered. FEA based approaches have the

advantage to consider the composite morphologies and rely on solving the heat equation.

𝑞𝑖𝑖 = 𝑘𝑖𝑗 𝑇𝑗,𝑗 (2.10)

In the equation, 𝑞𝑖𝑖 , 𝑘𝑖𝑗 and 𝑇𝑗,𝑗 are the heat flux, thermal conductivity tensor, and

temperature gradient in spatial direction jj respectively. One of the earliest approaches

to estimate the composite thermal conductivity by FEA was demonstrated by Kumlutus

et al. [138]. The approach did not consider anisotropy and was based on FD methods.

Later works followed by other researchers excluded the interfacial resistance, except

Singh et al. [139]. The RVE approach to estimate thermal conductivity was first introduced

by Banarjee et al. [140] for a 2D geometry. An element-free Galerkin method to calculate

the effective thermal conductivity including interfacial conductance was firs

demonstrated by Singh et al. [139]. Recently, a 3D RVE approach using FEA for composite

with disordered nanotubes was be found [141]. A multiscale approach involving

molecular dynamics (MD) modeling combined with FEA of composite is novel, and it will

allow researchers to model composite thermal conductivities for varying anisotropy of

multiple fillers and complexities. The superconducting wires in consideration is metal

matrix composite (MMC), and the orientation of the filaments are organized as well as

well defined. Thus FEA can be used effectively for multiscale analysis of heterogeneous

solids [142]–[150], high temperature superconducting wires [151], and plain-weave

fabrics [152].
26
 Literature Review of MRI Magnet Modeling

2.2.4 Modeling of Specific Heat

The specific heat of the material is a scalar quantity and is much simpler to

estimate for a composite than other material properties in consideration. However, the

effective specific heat for the composite is not merely the weighted average of the

constituent materials, i.e., filaments and matrix. This is because, even a uniform

temperature rise in a composite with displacement boundary conditions6 may generate

different local strain values, although the average strains are constant. This method to

calculate the specific heat is termed as specific heat at constant strain (Cε). The specific

heat can also be calculated by applying traction boundary7 condition or specific heat at

constant stress (cσ). Since the specific heat at constant strain is difficult to measure

experimentally, its use is more common in numerical or theoretical investigations,

whereas the specific heat at constant stress/pressure is experimentally measured.

However, the difference between the two types of specific heat is insignificant for solids

[153]. A composite usually contains materials of varying thermal expansion coefficients

so a local variation of strains and stresses of the constituent materials may be present

when temperature loading is applied. Therefore, even though the specific heat values are

a scalar quantity, a composite specific heat differs from the simple weighted average of

the fundamental values [154]. The methods to calculate the specific heat of composite

was first adopted from Levin’s [114] work. The theories were modified and extended to

6
The displacement at the boundaries are specified.
7
Stress is specified as the boundary conditions

27
Chapter 2

calculate the thermal conductivity and specific heat by Rosen [154] using the basic

equation as

𝑞⃛ = 𝑐𝛥𝑇 (2.11)

In the equation, 𝑞⃛, 𝛥𝑇, and 𝑐 are the change in volumetric heat, change in

temperature, and specific heat, respectively. The specific heat at constant deformation

and constant pressure for a reference temperature is defined as [155]:

𝜕𝑄
𝐶𝑣 (𝑇) = [ ] (2.12 a)
𝜕𝑇 𝜀𝑖𝑗

𝜕𝑄
𝐶𝑝 (𝑇) = [ ] (2.12 b)
𝜕𝑇 𝜎𝑖𝑗

In the equation 𝜕𝑄, is the amount of heat necessary for a uniform temperature

change of 𝜕𝑇. Analytically, the temperature dependent specific heat for superconducting

wires is calculated using the rule of mixtures [156], [157]:

𝑓𝑖 𝛾𝑖 𝐶𝑖 (𝑇)
𝐶𝑒𝑓𝑓 (𝑇) = ∑ (2.13)
𝛾𝑒𝑓𝑓
𝑖

where, 𝐶𝑖 is the individual specific heat, 𝑓𝑖 is the volume fraction, and 𝛾𝑖 is the

mass density of the constituents of the wire composite. The bulk density 𝛾𝑒𝑓𝑓 of the

composite is calculated using the simple weighted average as:

𝛾 = ∑ 𝑓𝑖 𝛾𝑖 (2.14)
𝑖

28
 Literature Review of MRI Magnet Modeling

In FEA, the conservation of the transient thermal energy equation is considered

and solved to calculate the specific heat of the homogenized RVE. Ignoring the effect of

heat of convection and radiation, the energy balance equation turns into the following

form

𝜕𝑇
∇ ∙ (𝑘(𝑇) ∙ 𝑇) + 𝑞⃛(𝑥, 𝑦, 𝑧) = 𝛾𝐶𝑝 (𝑇) (2.15)
𝜕𝑡

Equation (2.15) is solved in ANSYS with all the assigned material properties to the

components of the RVE and applied boundary conditions. Since, specific heat at constant

pressure is the major concern, traction boundary conditions are applied on the surface of

the RVE. The surfaces of the RVE is kept insulated, thus the temperature rise comes only

from the applied internal heat generation and is uniform over the volume. Thus, there is

a negligible temperature gradient inside the microscopic structure of the RVE. Therefore,

the conduction part of equation (2.15) is ignored and it takes the refined form of

𝜕𝑇
𝑞⃛(𝑥, 𝑦, 𝑧) = 𝛾𝐶𝑝 (𝑇) (2.16)
𝜕𝑡

Initially, a reference temperature ′𝑇′ is set to solve the equation to estimate the

specific heat at that temperature. A specified internal heat generation of 𝜕𝑞 is applied to

the RVE for a specified time of ‘𝑡′. The amount of applied heat generation is also chosen

in such a way that the temperature rise is significant enough to be recorded without any

numerical error, but insignificant enough to avoid any non-linearity in the estimation at

the temperature ′𝑇′. After time ‘𝑡′, heat generation is stopped and the steady state heat

29
Chapter 2

equation is solved to record the final average temperature of the RVE. The total amount

of heat input 𝜕𝑄 per unit time 𝑡 is calculated from the equation

𝜕𝑞
𝜕𝑄 = 𝑞⃛(𝑥, 𝑦, 𝑧) = (2.17)
𝑡

The difference between the initial and average final temperature 𝜕𝑇 is calculated

from the FEA results and input into the equation (2.17) to estimate the specific heat at

the temperature ′𝑇′.

2.3 Modeling The Magnet Bundles

2.3.1 Modeling of Winding

Winding stresses are applied on the superconducting wire at the time of magnet

winding manufacturing. Applying tension on the wire while winding it around the mandrel

helps to generate tensile load in the radial outward direction of the cylindrical coil bundle

[67], [158], [159]. Gray et al. [160] considered the bundle as transversely isotropic

solenoid materials and carried out the analysis of a superconducting solenoid for the first

time. The analysis was later extended for orthotropic problems by introducing finite

element analysis techniques instead of analytical equations [161]. Lontai and Marston

[162] were the first to utilize classical elasticity theories to derive equations for stress and

strains in solenoids. Burkhard [163] derived similar expressions independently, but the

analysis was isotropic and did not require computer analysis. Later, a computer code

“STANSOL” developed by Johnson et al. [164] was capable of calculating stress-strain due

to winding. However, the detailed procedure to include the effect of winding was unclear

since most of the research remained unpublished as they were mostly part of fusion
30
 Literature Review of MRI Magnet Modeling

energy research. Work by Arp [165] is the earliest published work that is available through

internet scholarly article search. Arp assumed no variation in tensile stress on the wire

along the axial direction. Thus, stress on a specific layer during winding process remains

Figure 2.2. Cylindrical coordinate system used in this work.

numerically similar. For convenience, Figure 2.2 shows the directional convention used in

this thesis. According to Arp’s [165] assumption, the negligible variation of winding

tension along the axial direction facilitates the assumption of plane strain approximation

even though numerical results have little deviation if plane stress is used instead. Also,

any sliding between the winding layers is disregarded.

2.3.1.1 Transverse isotropic Solution

During the winding process, the wires are wound around a mandrel. In most cases,

the mandrels are made of strong metals that are non-magnetic, such as austenitic

stainless steel. Sometimes aluminum mandrels are used. However, it is necessary that the

mandrel materials are non-magnetic so that there is little interaction between the

mandrel and the generated magnetic field. Once each layer of winding is complete, the

average elastic modulus of the winding and the mandrel changes, and this change is taken

into account by elementary mechanics as

31
Chapter 2

𝜎𝑟 (𝑎)
𝐸𝑚 = (2.18)
𝑢𝑟 (𝑎)
𝑎

In the equation, 𝐸𝑚 is the modified effective modulus at a location ‘a’, the

outermost radial location of the wire winding in a cylindrical form. This equation is

analogous to the simple stress strain relationship (𝜎 = 𝛦𝜀).

V. Arp [165] derived the winding stress with winding pretension of σw on a

solenoid for orthotropic material and presented in the form as

𝑟𝜎𝑟 = 𝐶 − 𝐷

𝑟𝜎𝜃 = 𝑟𝜎𝑤 + 𝑘𝐶 + 𝑘𝐷 (2.19)

𝑢 = (𝑘𝐿 − 𝑉)𝐶 + (𝑘𝐿 + 𝑉)𝐷

In the equation the constants are

𝑟 𝑘
𝐶 ≡ −[1 + 𝐸𝑚 (𝑘𝐿 − 𝑉)] ( ) 𝐼
𝑎

𝑎 𝑘
𝐷 ≡ −[1 − 𝐸𝑚 (𝑘𝐿 − 𝑉)] ( ) 𝐼
𝑟
(2.20)
1
1 − νθ𝑧 ν𝑧θ ν𝑟θ + νθ𝑧 ν𝑧𝑟 Eθ 2
L= ,V = ,k = ( )
Εθ Er Er

𝑏/𝑎
𝑎𝜎𝑤 𝑑𝑥
I≡∫
𝑟/𝑎 [1 + 𝐸𝑚 (𝑘𝐿 + 𝑉)]𝑥 𝑘 − [1 − 𝐸𝑚 (𝑘𝐿 − 𝑉)]𝑥 −𝑘

In the equation νθ𝑧 , νθr and ν𝑧r are the Poisson’s ratio in 𝜃 − 𝑧, 𝑟 − 𝜃, 𝑟 − 𝑧 plane

and 𝛦𝜃 and 𝐸𝑟 are the elastic modulus in 𝜃 and 𝑟 direction, respectively.

32
 Literature Review of MRI Magnet Modeling

2.3.1.2 Combined Homogeneous Cylinder Method (CHCM)

Another approach addressing the winding stress of the wire is to consider the

pressure on each layer and approximate them as thin cylinders. According to this method,

each layer of a coil bundle is approximated as concentric thin cylinders [62], [67]. Starting

from the basic elasticity equation, which takes the form of Hooke’s law in simplified form.

The equation in contracted notation becomes,

3
𝜎𝑖 𝜎𝑗
𝜀𝑖 = − ∑ 𝜈𝑗𝑖 , 𝑖 = 1,2,3 𝑤ℎ𝑒𝑟𝑒 𝜃 = 1, 𝑟 = 2, 𝑧 = 3 (2.21)
𝐸𝑖 𝐸𝑗
𝑗=1
𝑗≠𝑖

where 𝜎𝑖 , 𝜀𝑖 and 𝐸𝑖 are the stress, strain and elastic modulus in three orthogonal

direction (𝑟, 𝜃 and 𝑧), respectively, and 𝜈𝑗𝑖 is Poisson’s ratio. Since the stress in the axial

direction of the bundle (direction 2 of wire) is negligible compared to the applied winding

tension, the plane-stress approximation is more appropriate than the plane-strain [165].

When a layer of the bundle is wound with pretension, the applied radial pressure on the

lower layers can be calculated by considering each layer of a coil as a concentric

homogeneous cylinder using the following equation [62], [67].

𝜎𝑤0 ,𝑘 𝑡𝑘 𝜎𝑤0 ,𝑘 𝑡𝑘
𝑝𝑘 = = (2.22)
𝑟1 + ∑𝑖𝑗=1 𝑡𝑗 𝑟𝑤,𝑘

Here, σw0 ,k is the winding prestress on the kth layer, 𝑡𝑘 is the thickness of the kth

layer, 𝑟1 is the inner radial distance of the first layer of the coil, and 𝑡𝑘 is the thickness of

33
Chapter 2

each of the wound layers. The hoop stress developed on the lower layers due to this

applied pressure becomes

𝑝𝑘 (𝑟1 + ∑𝑖−1
𝑗=1 𝑡𝑗 )
𝜎𝜃𝑖 = (2.23)
∑𝑖−1
𝑗=0 𝑡𝑗

If the layer itself is wound with a prestress, 𝜎𝑘 , this amount of pretension will

superpose with the calculated circumferential stress given by equation (2.23). Hence, the

total hoop stress developed due to winding of the kth layer (σwk ) would be

𝜎𝑤𝑘 = 𝜎𝑤𝑜 ,𝑘 + ∑ 𝜎𝜃𝑖 (2.24)

𝑖=𝑘+1

The calculation of circumferential stress using these equations are known as the

Combined Homogeneous Cylinder Method (CHCM) [67]. In this work, the CHCM analytical

approximation is used to help validate the calculations of finite element analysis.

However, CHCM is limited by its inability to predict the stress development if there are

multiple materials, each with a different elastic modulus. Also, for complex geometry and

boundary conditions, CHCM does not offer a straightforward approach.

2.3.2 Modeling of Thermal Cool-down

Magnets for MRI systems are cooled to a cryogenic temperature of 4.2 K for NbTi

and Nb3Sn based systems. The proposed LHe free magnet is targeted to operate at 10 K

temperature. Thus the entire system is required to cool-down to 10 K from room

temperature of 298 K. This cooling down of the entire system creates stresses because

the thermal strain behavior of the mandrel and coil bundles is different and is highly non-

34
 Literature Review of MRI Magnet Modeling

linear in temperature. The analytical expression has been developed by researchers for

simple geometric structures considering transverse isotropic materials and under the

assumption of no heat loss, no contact resistance and linear material property responses.

However, they are a means of good approximation of a magnet bundle with complex

geometry and boundary conditions. Thus an analytical solution sometimes provides

useful comparison with FEA. It offers fidelity, and ability to handle a complex geometry.

2.3.2.1 Thermal Cool-down Stress Prediction by Analytical Approach

As the temperature drops from room temperature (T0) to a cryogenic temperature

(Tf), the added effect on the strain in cylindrical coordinates due to the temperature

change is expressed as,

𝑇𝑓 (𝑟) 𝑇𝑓 (𝑟)
𝜕𝑢 𝑢
𝜀𝑟 = − ∫ 𝛼𝑟 𝑑𝑇 , 𝜀𝜃 = − ∫ 𝛼𝜃 𝑑𝑇 (2.25)
𝜕𝑟 𝑟
𝑇0 𝑇0

The linear coefficients of thermal expansion (CTE) in the radial and circumferential

directions are given as αr and αθ, respectively; these are functions of temperature. By

implementing a simple power series expansion, Arp et al. [165] presented the expression

for radial and hoop stress as

𝑛
𝑐𝑟 𝑘−1 𝑑𝑟 −𝑘−1 𝑙𝑟𝑖 − (1 + 𝑖)𝑙𝜃𝑖 𝑖
𝜎𝑟 = − +∑ 𝑟
𝑘𝐿 − 𝑉 𝑘𝐿 + 𝑉 𝐿(1 + 𝑖)2 − 𝑘 2
𝑖=0
(2.26)
𝑛
𝑘𝑐𝑟 𝑘−1 𝑘𝑑𝑟 −𝑘−1 1 + 𝑖 𝑙𝑟𝑖 − (1 + 𝑖)𝑙𝜃𝑖 𝑖
𝜎𝜃 = + +∑ 𝑟
𝑘𝐿 − 𝑉 𝑘𝐿 + 𝑉 𝐿 (1 + 𝑖)2 − 𝑘 2
𝑖=0

The terms L, V, k, c, and d are defined as:

35
Chapter 2

1
1 − 𝜈𝜃𝑧 𝜈𝑧𝜃 𝑣𝜃𝑟 + 𝜈𝜃𝑧 𝜈𝑧𝑟 𝐸𝜃 2
𝐿= ,𝑉 = ,𝑘 = ( )
𝛦𝜃 𝐸𝑟 𝐸𝑟

𝑛 𝑎1+𝑖 𝑏1+𝑖
𝑘𝐿 − 𝑉 [𝑙𝑟𝑖 − (1 + 𝑖)𝑙𝜃𝑖 ] ( − 𝑘 )
𝑏𝑘 𝑎
𝑐= ∑
𝑏 𝑘 𝑎 𝑘 𝑖=0 (1 + 𝑖)2 − 𝑘 2 (2.27)
𝐿 ((𝑎) − ( ) )
𝑏

𝑛
𝑘𝐿 + 𝑉 [𝑙𝑟𝑖 − (1 + 𝑖)𝑙𝜃𝑖 ](𝑎1+𝑖 𝑏 𝑘 − 𝑏1+𝑖 𝑎𝑘 )
𝑑= ∑
𝑏 𝑘 𝑎 𝑘 (1 + 𝑖)2 − 𝑘 2
𝐿 (( ) − ( ) ) 𝑖=0
𝑎 𝑏

where νθz and νzθ are Poisson’s ratio and 𝛦𝜃 and 𝐸𝑟 are the modulus of elasticity

of the material in the specified directions and ‘a’ and ‘b’ are the inner and outer radii of

the cylinder.

2.3.3 Modeling of Electromagnetic Charging

The magnetic stresses in superconducting magnet become a concern when large

field strength magnets are built. In a 1924 article Kapitza [70], [71] demonstrated the

ability to produce a large field (30 T pulsed magnet) using a solenoid. In 1936 Francis Bitter

laboratory of MIT produced a water-cooled 10 T magnet that generates pressure in the

range of 300 MPa [159], [166]. Over the years, high field magnets experienced mechanical

failures [167], and complete understanding of the stress development in a magnet

became necessary. In the early years, magnetic forces were calculated directly from

magnetic field as explained by Kapitza [71], and Cockcroft [72] and are then substituted

on the linear momentum balance equation8. In the 1960s, extensive studies with

8
Continuum mechanics representation of Newton’s second law of motion

36
 Literature Review of MRI Magnet Modeling

nonlinear continuum mechanics peaked, and the constitutive relationship between

stress, strain and the magnetic field has been rigorously studied and formed [168], [169].

In a broad range, stresses in solenoids are classified into two categories: ferromagnetic

structures and non-ferromagnetic structures [166]. Non-ferromagnetic structures are the

superconductors carrying large currents through them. Since the focus of this work is

mainly MRI magnets, discussions will be limited to non-ferromagnetic structures.

When the superconducting solenoid is energized, the electromagnetic forces are

determined from the radial and axial components of J×B where J and B are the vector

representation of the current density and magnetic field, respectively. These radial and

axial directional forces, when considered as a body force in the stress equation, result in

stresses in the radial and hoop directions. The hoop stress appears to be dominating

compared to the radial stress [166] and hence is of major concern when designing the

superconducting solenoid.

2.3.3.1 Analytical Approach for Predicting Electromagnetic Stress

Starting from the linear momentum equation in cylindrical form, the

electromagnetic force due to electric current passing through the bundle is considered as

a body force. So the equation becomes:

𝜕𝜎𝑟
𝑟 + (𝜎𝑟 − 𝜎𝜃 ) + 𝑋𝑟 = 0 (2.28)
𝜕𝑟

where Xr is the body force which relates to the magnetic field Bz as

37
Chapter 2

𝑋𝑟 = 𝐽𝐵𝑧 (2.29)

and Bz is the axial component of the magnetic field. Caldwell [170], [171], has

derived the equation for radial and hoop stress for a uniform solenoid in a final form as

𝜕𝜎𝑟
𝑟 + (𝜎𝑟 − 𝜎𝜃 ) + 𝑋𝑟 = 0 (2.30)
𝜕𝑟

𝜎𝜃 1 𝑟12 𝑝 1 𝑝
𝜎𝑟 = (1 − 2 ) − 𝐼0 − 2 (1 − ) 𝐼2
2 𝑟 2 𝑟 2
(2.31)
𝐽{𝐵1 (𝑟22 + 𝑟1 𝑟2 − 2𝑟12 ) + 𝐵2 (2𝑟22 − 𝑟1 𝑟2 − 𝑟12 )}
𝜎𝜃 = 𝑟
6𝑟 𝑙𝑛 (𝑟2 )
1

where B1 and B2 are the radial component of the magnetic field at the inner radius

(𝑟1 ) and outer radius (𝑟2 ), 𝑝 = 1 − 𝜈

2.4 Modeling of Quench

The earliest effort to model a superconducting magnet quench dates back to 1963

by Stekly [172] after a theoretical understanding was discussed by Cherry et al. [173] even

though the transition between superconducting to normal mode was a common problem

for all superconducting systems [173]. Earlier investigations confirmed the reduction in

the magnetic field at the superconducting to normal (s-n) transition region for a thin plate.

Underlying reasons were unknown initially, and later scientists attributed the

phenomenon to joule heating [174] that contributed to the reduced critical current rating

38
 Literature Review of MRI Magnet Modeling

of the superconductors. Initial studies focused on thin film superconductors, which

provide advantages of a 1-D reduction of the complicated heat equation. The analytical

solution to the 1-D heat equation then expresses the temperature rise with time under

the assumption that the ordinary thermal conduction equation holds true for an infinite

plate length [175]. Superconducting wires could also be thought of as a long strand of

wire, and, subsequently, voltage and temperature could be modeled in a similar fashion

by assuming a solution in the form of a traveling wave [173]. Approaches to match

analytical prediction with experiments were demonstrated. However, the model is very

simplified and fails to address the non-linear behavior [176]. Significant deviation of the

analytical equation from the experimental results due to the unavailability of the

appropriate model for MRI coil has motivated some researchers to fit experimental data

in an equation form relating time, magnetic field transition current [177]. Later, advances

in the understanding of the quench presented more refined, accurate and appropriate

model of the magnet quench and Wilson [178, Ch. 9] developed computer code that has

been widely used. Several different techniques to address the 3-D model of magnet

quench was approached. Bottura et al. [179] developed a comprehensive model for

quench modeling with necessary equations, utilizing the method of coupling of 1-D heat

equation to formulate the basis for quench analysis. They introduced a parallel solution

to the coupled 1-D equation, estimating the composite superconducting wire as a volume

average. Presence of anisotropy was well recognized for superconducting composite

wires. Estimation of the material property for the wire was calculated considering

directional variation in material properties. A solution to the formulated coupled 1-D

39
Chapter 2

equations was presented in the follow-up article by making use of Taylor-Galerkin method

[180]. This technique has been successfully implemented to model quench for the ITER

magnet [181]. Besides, global collocation procedures9 [182], [183], a method of lines

[184] and FD methods have also been implemented as an alternative to FEA methods.

Although FEA methods are more accurate and flexible [179] compared to FD, they require

significant computational approach. Use of implicit schemes to avoid instability in the

solution further increases the computational effort, making the transient problem

challenging and computationally intensive in the early days of computational

development. Over time, extensive research with quench modeling has addressed AC loss

and simultaneous consideration of heat propagation [185] instead of coupled three 1-D

equations sets.

2.4.1 MgB2 Quench

The MgB2 critical temperature is ~ 39 K [29], making it a high-temperature

superconductor (HTS). Stability analysis of low-temperature superconductor (LTS) is

already well established [178], [151], and based on the concept of normal zone

propagation10 (NZP) [186]. In that perspective, the scenario is different for MgB2 which is

slightly different from the usual HTS, since the critical temperature is larger than the

operating temperature but relatively lower than that of HTS. In LTS magnets, the quench

is modeled by solving a heat conduction equation under the assumption that the normal

zone propagates with a moving boundary penetrating the superconducting region.

9
methods to solve ODE that comes from reduction of PDE
10
resistive zone of superconducting magnet

40
 Literature Review of MRI Magnet Modeling

However, in an HTS magnet, the initiation of a quench requires a significantly large

amount of energy compared to an LTS magnet since the heat capacity is relatively larger.

Also, an increased temperature margin makes it difficult for the normal zone to propagate

(two to three orders of magnitude slower) [187], thus resulting in the local hot spot [188],

[189]. Thus, a quench in an HTS magnet is a global phenomenon rather than a local zone

propagation, in contrast to LTS magnets [190]. LTS magnets rely on NZP whereas, the

temperature profile of the magnet is a more important indicator of stability for HTS

magnets [189], [191]–[193]. Furthermore, a quench is related to the electromagnetic

phenomenon and requires the simultaneous solution of the magnetic and thermal field

[151], [188], [194]–[200].

The MgB2 quench has been of interest because of its role in MRI magnets. As the

critical temperature of MgB2 is between the LTS and HTS, it behaves differently than those

LTS or HTS magnets [201]. Experimental studies have demonstrated a higher minimum

quench energy (MQE) requirement for an MgB2 quench to occur [202]. In the early stage

of MgB2 wire development, most of the quench studies were with monofilament tape

superconductor [201]–[204] as well as multifilament tape [204], [205] by adiabatic

cooling. Round MgB2 wires with both adiabatic cooling [206] and LHe cooling were

studied and the NZP velocity (NZPV) was found to vary between 0.015 m/s to 1 m/s. The

corresponding MQE values were 10 mJ to 1000 mJ [202]–[204], [207]–[209]. The ultimate

goal of MgB2 based MRI magnet is to eliminate LHe, thus a quench study of short samples

under conduction cooled setup was performed, and MQE and NZPV values are reported

[205]. Simulation of a solenoid coil is different from a simple wire as heat transfer takes

41
Chapter 2

place in the radial (r) direction of the coil. One approach to the issue of transverse heat

transfer is to take the average of the material properties of the wires and insulation layers.

The method eliminates the use of extensive modeling of each wiring layer, requiring less

computational time. This technique has been applied to MgB2 tapes, and considerable

agreement between numerical and experimental results has been reported [210], [211].

Efforts have been made to model quench progress consecutively from 2D multiple wire

stack simulation [212] to 3D coil bundles [213], [214]. Improving upon the wire quench

simulation, Stenvall [211] developed a Matlab code in combination with FEMLAB to

estimate the voltage, current, and temperature of a solenoid taking into consideration

the external circuitry. In their later work, Stenvall introduced a power law behavior [215],

[216] considering MgB2 as comparable to LTS to estimate the MQE due to finite ‘n’

values11 [217] because conventional methods [218], [219] introduced some level of

inaccuracy. Depending on the comparable resistivity there are two different methods to

model the transition region12 as explained by Ristic et al. [220].

2.4.2 Analytical Formulation of Quench

The scope of this work is limited to conduction cooled magnets. Thus there is a

slight difference between our modeling equation as compared to the other LTS LHe

cooled MRI background magnets. However, the underlying equation to estimate the

temperature profile is the same quasi-linear heat conduction equation [211],

11
the exponent of the current voltage characteristics curve
12
Superconducting to resistive mode transition

42
 Literature Review of MRI Magnet Modeling

𝜕𝑇
∇ ∙ (𝑘(𝑇) ∙ 𝑇) + 𝑞̇ (𝑥, 𝑦, 𝑧) = 𝛾𝐶𝑝 (𝑇) (2.32)
𝜕𝑡

where, 𝑘, 𝑇, 𝑞̇ and 𝐶𝑝 are the thermal conductivity, temperature, heat generation

and heat capacity respectively. This equation in cylindrical coordinate transforms into

[156], [157]

1 𝜕 𝜕𝛵 𝜕 𝜕𝛵 1𝜕 𝜕𝛵
2
(𝑘𝜃 (𝛵) ) + (𝑘𝑧 (𝛵) ) + (𝑟𝑘𝑟 (𝛵) )
r 𝜕𝜃 𝜕𝜃 𝜕𝑧 𝜕𝑧 𝑟 𝜕𝑟 𝜕𝑟
(2.33)
𝜕𝑇
̇ (𝑥, 𝑦, 𝑧) + 𝑞𝑒𝑥𝑡
+ 𝑞𝑖𝑛𝑡 ̇ (𝑥, 𝑦, 𝑧) = 𝛾𝐶𝑝 (𝑇)
𝜕𝑡

In the equation, the heat generation is a combination of external heat generation

̇ (𝑥, 𝑦, 𝑧)) and internal heat generation (𝑞𝑖𝑛𝑡

(𝑞𝑒𝑥𝑡 ̇ (𝑥, 𝑦, 𝑧)). External heat generation is

attributed to energy from an external quench initiation heater or protection heaters.

Internal heat generation is attributed to internal heating which is a choice between the

power law and joule heating [216]

𝑛
𝐼 𝐼 𝐼
̇ (𝑥, 𝑦, 𝑧) =
𝑞𝑖𝑛𝑡 min {𝐸𝑐 ( ) , 𝜌(𝐵, 𝑇) } (2.34)
𝐴 𝐼𝑐 (𝐵, 𝑇) 𝐴

In the equation, 𝐼 is the engineering current flowing through the superconducting

wire, A is the cross sectional area of the wire and is calculated from ∆𝑟 ∙ ∆z where ∆𝑟 and

∆z are the radial and axial dimension of the wire respectively. 𝐸𝑐 and 𝐼𝑐 (𝐵, 𝑇) is the

electric field criterion and critical current for superconductivity, n is the ‘n’ value of the

wire from the voltage-current power relationship.

The NZPV in the longitudinal direction is calculated by an analytical Wheatstone-

Rose formula [159]

43
Chapter 2

𝐽 𝜌(𝑇𝑡 )𝑘𝜃 (𝑇𝑡 )

𝑣𝑙 = ×
𝛾 √ 1 𝑑𝑘𝜃 𝑇𝑡 𝑇 (2.35)
(𝐶𝑝 (𝑇𝑡 ) − | 𝐶 (𝑇)𝑑𝑇) ∫𝑇 𝑡 𝐶𝑝 (𝑇)𝑑𝑇
𝑘𝜃 𝑑𝑇 𝑇 ∫𝑇𝑜𝑝 𝑝 𝑜𝑝
𝑡

In the equation above, ρ and Cp are resistivity and heat capacity respectively, J is

the current density, γ is the mass density, 𝑇𝑡 is the superconducting to normal mode

transition temperature, and 𝑇𝑜𝑝 is the operating temperature. When the material

property is constant with varying temperature, equation (2.35) simply reduces in,

𝐽 𝜌𝑘𝜃
𝑣𝑙 = ×√ (2.36)
𝛾𝐶𝑝 (𝑇𝑡 − 𝑇𝑜𝑝 )

This simple equation can be used for the composite wire to estimate the analytical

NZPV for superconducting magnets. After the longitudinal NZPV is calculated, the

transverse NZPV is estimated from the following relationship

1 𝛿𝑤𝑖𝑟𝑒 𝑘𝑟 (𝑇)
𝑣𝑡 = 𝑣𝑙 √ (2.37)
2 𝛿𝑖𝑛𝑠𝑢𝑙𝑎𝑡𝑜𝑟 𝑘𝜃 (𝛵)

Here, 𝑘𝑟 (𝑇) and 𝑘𝜃 (𝛵) is the transverse (radial, r) and longitudinal (tangential, θ)

thermal conductivity, 𝛿𝑤𝑖𝑟𝑒 and 𝛿𝑖𝑛𝑠𝑢𝑙𝑎𝑡𝑜𝑟 is the thickness of the composite wire and

thickness of the insulation layer, respectively.

2.5 Conclusion

Modeling of a full body MRI magnet is a complex task since the magnets are made

of bundles, and each bundle is made up of composite wires. The composite wires itself

has definite micro-structures. Thus modeling of the entire magnet requires extensive

44
 Literature Review of MRI Magnet Modeling

analysis of the wires, magnet bundles and magnet systems. If every detail from the wire

length scale to the magnet bundle length scale is considered, the FEA approach

constitutes an enormous number of unknowns and equation, making it difficult to solve

in a reasonable time frame. Thus, the multiscale approach provided the opportunity to

model the magnet system. Also, magnet manufacturing, cool-down, and operation

consist of interactions among different physics. Thus it is necessary to investigate the

existing methods and techniques to solve the problem. This chapter is an effort to

summarize the past efforts to model the superconducting wire and magnet. The

discussion is in no way comprehensive and only serves to provide an idea of the

techniques to model a superconducting magnet.

45
 Chapter 3. Material Properties for Magnet Modeling

“The nitrogen in our DNA, the calcium in our teeth, the iron in our blood, the carbon in

our apple pies were made in the interiors of collapsing stars. We are made of star stuff.”

— Carl Sagan, Cosmos (1980)

Figure 3.1. Stress state in 3D.

3.1 Material Properties

The material properties of the wire constituent are necessary to develop the

multiscale multiphysics model of the full body MRI magnet system. The wire under

46
 Material Properties for Magnet Modeling

consideration consists of MgB2, Niobium, Copper, Monel, and Epoxy (CTD 101k). The

material properties required for magnet winding, cool-down, electromagnetic

energization, and quench are the elastic modulus, Poisson’s ratio, thermal expansion

coefficient, thermal conductivity, heat capacity and resistivity.

3.2 Elastic Properties

Elastic properties define the mechanical stress and strain response of the material.

In simplest terms, assuming the material is isotropic, and Hooke’s law holds true, stress is

directly proportional to strain and is related by material’s elastic modulus, E.

𝜎 = 𝐸𝜀 (3.1)

In the equation, 𝜎 is the applied stress and 𝜀 is the strain. In most cases, the stress-

strain relationship is not so simplified and depends on position as well as orientation of

the surface containing the stresses. Therefore, stress is usually represented in a tensor

form as

𝜎𝑥𝑥 𝜏𝑥𝑦 𝜏𝑥𝑧

[𝜎] = [𝜏𝑦𝑥 𝜎𝑥𝑥 𝜏𝑦𝑧 ] (3.2)
𝜏𝑧𝑥 𝜏𝑧𝑦 𝜎𝑥𝑥

In the equation, the first subscripts denote the vector direction of the area on

which the stress is acting and the second subscript denotes the direction along which the

stress is acting according to Figure 3.1. For example, 𝜎𝑥𝑥 is stress in the surface vector in

direction x and the force is also acting along the direction x. Similarly, 𝜏𝑥𝑦 is the stress on

the plane that has vector direction along x but it is acting along direction y. Strain tensor

can also be represented in a similar fashions as stress. However, coordinate

47
Chapter 3

transformation does not apply if strains are written exactly in a form stress are

represented. The issue is solved by writing the strain in a special form as

2𝜀𝑥𝑥 𝛾𝑥𝑦 𝛾𝑥𝑧 𝜀11 𝜀12 𝜀13

1
[𝜀] = [ 𝛾𝑦𝑥 2𝜀𝑥𝑥 𝛾𝑦𝑧 ] = [𝜀12 𝜀22 𝜀23 ] (3.3)
2 𝜀13 𝜀23 𝜀33
𝛾𝑧𝑥 𝛾𝑧𝑦 2𝜀𝑥𝑥

In the equation, 𝜀𝑥𝑥 , 𝜀𝑦𝑦 , 𝜀𝑧𝑧 denotes normal strains and 𝛾𝑥𝑦 , 𝛾𝑦𝑧 , 𝛾𝑥𝑧 denotes

shear strains. Under the assumption that a body is constrained and rigid body motion is

absent, the strains are related to applied displacement as

𝜕𝑢𝑥 𝜕𝑢𝑥 𝜕𝑢𝑦

𝜀𝑥𝑥 = ; 𝛾𝑥𝑦 = 𝛾𝑦𝑥 = ( + )
𝜕𝑥 𝜕𝑦 𝜕𝑥
𝜕𝑢𝑦 𝜕𝑢𝑥 𝜕𝑢𝑧
𝜀𝑦𝑦 = ; 𝛾𝑥𝑧 = 𝛾𝑧𝑥 = ( + ) (3.4)
𝜕𝑦 𝜕𝑧 𝜕𝑥
𝜕𝑢𝑧 𝜕𝑢𝑦 𝜕𝑢𝑧
𝜀𝑧𝑧 = ; 𝛾𝑦𝑧 = 𝛾𝑧𝑦 = ( + )
𝜕𝑧 𝜕𝑧 𝜕𝑦

If a material is isotropic and homogeneous, the strain is directly proportional to

the applied stress. Considering a 3D hexahedron as shown in Figure 3.1, if the strain is

applied in x-direction, then the material will experience strain in the other two directions.

In isotropic material, when stress is applied in one direction, the strain in other two

orthogonal directions are directly linked through the Poisson’s ratio. When the material

response is not symmetric about three orthogonal planes, they are said to be anisotropic

material. For an anisotropic material, strain in a specific direction is connected to all nine

different applied stress. In equation form, this relationship is represented as

48
 Material Properties for Magnet Modeling

𝜀𝑥 = 𝑆11 𝜎𝑥𝑥 + 𝑆12 𝜎𝑦𝑦 + 𝑆13 𝜎𝑧𝑧 + 𝑆14 𝜏𝑥𝑦 + 𝑆15 𝜏𝑦𝑧 + 𝑆16 𝜏𝑧𝑥 + 𝑆17 𝜏𝑥𝑧
(3.5)
+ 𝑆18 𝜏𝑧𝑦 + 𝑆19 𝜏𝑦𝑥

It is easier to write the relations in matrix form, and thus it takes the form

𝜀𝑥𝑥 𝑆11 𝑆12 𝑆13 . . . . . 𝑆19 𝜎𝑥𝑥

𝜀𝑦𝑦 . . 𝜎𝑦𝑦
𝜀𝑧𝑧 . . 𝜎𝑧𝑧
𝛾𝑥𝑦 . . 𝜏𝑥𝑦
𝛾𝑦𝑧 = . . 𝜏𝑦𝑧 (3.6)
𝛾𝑧𝑥 . . 𝜏𝑧𝑥
𝛾𝑥𝑧 . . 𝜏𝑥𝑧
𝛾𝑧𝑦 . . 𝜏𝑧𝑦
{𝛾𝑦𝑥 } [𝑆 𝑆92 𝑆93 . . . . . 𝜏
𝑆99 ] { 𝑦𝑥 }
91

When the material response is symmetric about three orthogonal planes, the

material is said to be orthotropic. Also, the shear stresses are symmetric meaning 𝛾𝑥𝑦 =

𝛾𝑦𝑥 , and the above equation takes the form,

𝜀𝑥𝑥 𝑆11 𝑆12 𝑆13 0 0 0 𝜎𝑥𝑥

𝜀𝑦𝑦 𝑆21 𝑆22 𝑆23 0 0 0 𝜎𝑦𝑦
𝜀𝑧𝑧 𝑆31 𝑆32 𝑆33 0 0 0 𝜎𝑧𝑧
𝛾𝑥𝑦 = 0 𝜏𝑥𝑦 (3.7)
0 0 𝑆44 0 0
𝛾𝑦𝑧 0 0 0 0 𝑆55 0 𝜏𝑦𝑧
{ 𝛾𝑧𝑥 } [ 0 0 0 0 0 𝑆66 ] { 𝜏𝑧𝑥 }

In the equation, the tensor [S] is compliance tensor, and elastic properties are

input as matrix form for stress-strain calculation. Equation (3.7) is related to the elastic

modulus and Poisson’s ratio in the following terms:

49
Chapter 3

1 𝜈𝑦𝑥 𝜈𝑧𝑥
− − 0 0 0
𝐸𝑥 𝐸𝑦 𝐸𝑧
𝜀𝑥𝑥 𝜈𝑥𝑦 1 𝜈𝑧𝑦 𝜎𝑥𝑥
𝜀𝑦𝑦 − − 0 0 0 𝜎𝑦𝑦
𝐸𝑥 𝐸𝑦 𝐸𝑧
𝜀𝑧𝑧 𝜎𝑧𝑧
𝛾𝑥𝑦 = − 𝑥𝑧𝜈 𝜈𝑦𝑧 1 𝜏𝑥𝑦 (3.8)
− 0 0 0
𝛾𝑦𝑧 𝐸𝑥 𝐸𝑦 𝐸𝑧 𝜏𝑦𝑧
{ 𝛾𝑧𝑥 } 0 0 0 𝐺𝑥𝑦 0 0 { 𝜏𝑧𝑥 }
0 0 0 0 𝐺𝑦𝑧 0
[ 0 0 0 0 0 𝐺𝑧𝑥 ]

In equation (3.8), 𝐸𝑥 , 𝐸𝑦 , 𝐸𝑧 , are elastic modulus along direction 𝑥, 𝑦, 𝑧. 𝜈𝑥𝑦 , 𝜈𝑥𝑧 ,

𝜈𝑦𝑧 are Poisson’s ratio in the plane 𝑥 − 𝑦, 𝑥 − 𝑧, 𝑦 − 𝑧. Finally, 𝐺𝑥𝑦 , 𝐺𝑥𝑧 , 𝐺𝑦𝑧 are shear

modulus in plane 𝑥 − 𝑦, 𝑥 − 𝑧, 𝑦 − 𝑧. If the material’s response is similar in all three

directions, for an applied stress the strain response is similar in any direction of the plane.

In this case, 𝜀𝑥𝑥 = 𝜀𝑦𝑦 = 𝜀𝑧𝑧 and 𝛾𝑥𝑦 = 𝛾𝑦𝑧 = 𝛾𝑥𝑧 . Equation (3.7) then takes the simplified

form as

𝜀𝑥𝑥 𝑆 𝑆12 𝜎𝑥𝑥

{𝛾𝑥𝑦 } = [ 11 ]{ } (3.9)
𝑆21 𝑆22 𝜏𝑥𝑦

In an effort, the find analogy with Hooke’s law, the equation (3.9) can be re-

written in the form

𝜎𝑥𝑥 𝐶 𝐶12 𝜀𝑥𝑥

{𝜏𝑥𝑦 } = [ 11 ]{ } (3.10)
𝐶21 𝐶22 𝛾𝑥𝑦

In this equation, [C] is the stiffness tensor. It is related to the compliance tensor

[S] as [𝐶] = [𝑆]−1 . The equation is easier and more concise if written in contracted

notation as

50
 Material Properties for Magnet Modeling

𝜀𝑘𝑙 = 𝐶𝑖𝑗𝑘𝑙 𝜎𝑖𝑗 (3.11)

It is difficult to write the compliance tensor 𝐶𝑖𝑗𝑘𝑙 in its four dimensional form, but

we see that the values of 𝑖, 𝑗, 𝑘, 𝑙 varies from 1 to 3 and the first component in the matrix

is 𝐶1111 and the last component is 𝐶3333 .

300
MgB2 6
Niobium
250 Copper
Monel 5

Thermal Strain (mm/m)

CTD 101k
Elastic Modulus (GPa)

200
4

150
3

100 2

50 1

0 0
0 50 100 150 200 250 300 0
Temperature (K)
Figure 3.2. Elastic modulus variation with temperature. [222]–[227] 1000

1000
The material properties are experimentally investigated, and temperature
800
rmal Conductivity (W/m - K)

dependence is recorded by the different research group and material manufacturers.

Specific Heat (J/Kg-K)

100
Since for the MgB2 magnet design, all of the phenomena happen in the temperature range
600
10 ~ 300 K; elastic property values are thus compiled in this temperature range. Usually,
10

51 400
1

MgB2
200
Chapter 3

most of the constituent materials of superconducting composite wire exhibit isotropic

behavior. The elastic modulus values for MgB2 [221], [222], Niobium [223], Copper [224],

[225], Monel [224] and CTD 101k [226] is shown in the graph along with varying

temperature in Figure 3.2.

Table 3.1: Poisson’s ratios of the constituent materials of superconducting wires

Material Poisson’s ratio (ν) at 298 K

MgB2 0.181 [221]

Copper 0.355 [225]

Niobium 0.4 [227]

Monel 0.315 [228]

Glidcop® AL-60 0.32 [229]

Epoxy 0.355 [224]

Stainless steel 316L 0.305 [224]

The material property of MgB2 is extracted from an investigation by Prikhna [221].

The elastic modulus is reported 273 GPa with a Poisson’s ratio of 0.181 and bulk modulus

of 142.5 GPa. Although Poisson’s ratio decreases with an increase in temperature for

solids [230], the change is insignificant for most of the materials. Also, there is not enough

data available for the components under investigation. Thus the value is assumed

constant over the temperature range. The elastic property of the material is also possible
52
 Material Properties for Magnet Modeling

to calculate using Density Functional Theory (DFT) [231] or Molecular Dynamics [232], but

the extensive effort is required to estimate the properties of MgB2 with varying

temperature. Thus, due to the unavailability of the material property data, Poisson’s ratio

values listed in Table 3.1 are used for all the materials in consideration.

3.3 Thermal Expansion Coefficient (α)

When the temperature of a bar resting at a temperature 𝑇0 as shown in Figure 3.3

is elevated to a temperature of 𝑇𝑓 , the length of the bar changes. If 𝑇𝑓 > 𝑇0 , then the

length will increase and the length will decrease if 𝑇𝑓 < 𝑇0 . Assuming, the length of the

bar changes linearly with the change in temperature, the relationship is expressed as

[233],

Figure 3.3. Thermal strain of a bar of length ‘L’.

𝛥𝐿
𝜀𝑡ℎ = = 𝛼(𝑇𝑓 − 𝑇0 ) (3.12)
𝐿

The thermal strain is related to temperature change with a coefficient that is

termed as the thermal expansion coefficient (TEC), 𝛼. When the ends of the bar are

restrained, the bar does not experience change in length but develops internal stress. The

stress developed is calculated using the equation

53
Chapter 3

𝜀𝑡𝑜𝑡𝑎𝑙 = 𝜀𝑡ℎ + 𝜀𝑚𝑒𝑐ℎ = 0 (3.13)

𝜎
Following the Hooke’s law, the mechanical strain is substituted by 𝐸 and equation

(3.13) takes the form as

𝜎𝑚
= 𝛼(𝑇𝑓 − 𝑇0 ) (3.14)
𝐸

Figure 3.4. Thermal strain of a bar of length ‘L’.

In the equation, 𝜎𝑚 is the compressive stress developed due to the increase in

temperature and is mathematically opposite to the stress as described by Hooke’s law

𝜎
initially: (𝐸 ). Thus, using this equation it is possible to estimate the stress due to

temperature change. However, the thermal expansion coefficient also changes with

temperature change. The general trend is that the value of the thermal expansion

coefficient decreases as the temperature drops. In the analysis, the temperature

dependent thermal expansion coefficient is an input in FEA using three different form,

and they will be discussed in the following sections.

3.3.1 Temperature Dependent Secant Coefficient (αse(Τ))

Temperature-dependent secant coefficient is calculated by equation (3.12). At

any instant of the temperature, the thermal strain is calculated according to equation

(3.12) and takes the final form of

54
 Material Properties for Magnet Modeling

𝜀𝑡ℎ = 𝛼𝑠𝑒 (𝑇)(𝑇𝑓 − 𝑇0 ) (3.15)

In the equation, 𝛼𝑠𝑒 (𝑇) is the ‘TEC’ as a function of temperature. Thermal strains

in each orthogonal direction is calculated using equation (3.15) and is added as a

functional term in equation (3.7). In FEA, the strain tensors are calculated from the

applied boundary condition and traction loads. Thus, the modified form of equation (3.7)

takes the form

𝜎𝑥𝑥 𝐶11 𝐶12 𝐶13 0 0 0 𝜀𝑥𝑥

𝜎𝑦𝑦 𝐶21 𝐶22 𝐶23 0 0 0 𝜀𝑦𝑦
𝜎𝑧𝑧 𝐶31 𝐶32 𝐶33 0 0 0 𝜀𝑧𝑧
𝜏𝑥𝑦 = 0 𝛾 + [𝜀𝑡ℎ ] (3.16)
0 0 𝐶44 0 0 𝑥𝑦
𝜏𝑦𝑧 0 0 0 0 𝐶55 0 𝛾𝑦𝑧
{ 𝜏𝑧𝑥 } [ 0 0 0 0 0 𝐶66 ] ( { 𝛾𝑧𝑥 } )

The thermal strain in equation (3.16) is estimated using equation (3.15) where the

shear strains are assumed to be zero. In tensor form, the estimated thermal strain takes

the form

𝑥 (𝑇)
𝛼𝑠𝑒
𝑦
𝛼𝑠𝑒 (𝑇)
𝑧
[𝜀𝑡ℎ ] = (𝑇𝑓 − 𝑇0 ) 𝛼𝑠𝑒 (𝑇) (3.17)
0
0
[ 0 ]

3.3.2 Computed Average Thermal Expansion Coefficient

If the temperature change is uniform such that the assumption of a steady state

condition is valid, the use of average TEC values are more appropriate. Average TEC values

are the average of TEC values over a specific temperature range. Rather than considering

individual TEC values, average values are used over the temperature range. The FEA is

55
Chapter 3

performed over a single step using the average TEC values; the technique saves

computational effort yielding quicker results. Average TEC values are calculated using the

equation

𝑇𝑓
∫𝑇 𝛼𝑖𝑛 (𝛵)
̅̅̅̅̅̅̅̅̅
𝛼𝑠𝑒 (𝑇) = 0 (3.18)
(𝑇𝑓 − 𝑇0 )

3.3.3 Calculated Thermal Strain (εth)

Thermal strains are straightforward to estimate from experiment through a strain

gauge. Also in FEA, the input of thermal strain reduces a layer of computation, reducing

computation time for an analysis. Thermal strains are directly calculated using equation

(3.12). However, the specification of the reference temperature is crucial. If the reference

temperature is not set appropriately, the calculation of strains yields incorrect results.

The modeling of thermal cool-down process and modeling of quench requires carefully

selecting the reference temperatures. When the magnet system is cooled down from

room temperature (298 K), the reference temperature is 298 K. Thermal strain of all the

estimated material and wire constituents are ‘zero’ at this reference temperature. On the

other hand, for modeling of the quench, the reference temperature is 10 K, the magnet

operation temperature in this work. Thus, the thermal strain at this temperature is zero

for all the constituent components of the superconducting wire. Thermal strain of all the

wire constituents are presented in Figure 3.5, are provided by Hyper Tech Research or

extracted from other available resources [224], [226].

56
 Material Properties for Magnet Modeling

0.006 MgB2
Niobium
Copper
Thermal Strain (absolute)
0.005
Monel
CTD 101k (through thickness)
0.004 CTD 101k (Warp and Fill)
Reference Temp (10 K)

0.003

0.002

0.001

0.000
0 50 100 150 200 250 300
Temperature (K)
Figure 3.5. Thermal strain of constituent materials. [225], [227]

3.4 Thermal Conductivity (k)

Thermal conductivity is a material’s ability to conduct heat from the higher

temperature point to the lower temperature. Thermal conductivity is evaluated from the

Fourier law of heat conduction. In anisotropic material, the amount of heat flux passing

through a material is governed by the Fourier law of heat conduction as

𝑞𝑖𝑖 = −𝑘𝑖𝑗 𝑇𝑖,𝑗 (3.19)

57
Chapter 3

1000

Thermal Conductivity (W/m-K)

100

10

MgB2
0.1 Niobium
Copper
Monel
CTD 101k
0.01
0 50 100 150 200 250 300
Temperature (K)
Figure 3.6. Thermal conductivity of constituent materials. [224], [234]–[236]
In this equation, 𝑘𝑖𝑗 is the thermal conductivity tensor for a temperature gradient

in spatial direction of 𝑇𝑖,𝑗 . The Fourier law of heat conduction holds true under the

assumption that the characteristic length scale of the temperature gradient is significantly

larger than the ‘microscopic’ length scale of the medium. Therefore, in expanded tensor

form the heat conduction equation is rewritten in the form [143]

𝑞𝑥𝑥 𝑘𝑥𝑥 𝑘𝑥𝑦 𝑘𝑥𝑧

[𝑞𝑦𝑦 ] = −[𝑘] ∙ ⃗∇𝑇 = [𝑘𝑦𝑥 𝑘𝑦𝑦 𝑘𝑦𝑧 ] ∙ ⃗∇𝑇 (3.20)
𝑞𝑧𝑧 𝑘𝑧𝑥 𝑘𝑧𝑦 𝑘𝑧𝑧

58
 Material Properties for Magnet Modeling

If the material is orthotropic, i.e., the material is symmetric about the three

mutually orthogonal planes equation (30) reduces to

𝑞𝑥 𝑘𝑥𝑥 0 0
𝑞
[ 𝑦] = [ 0 𝑘𝑦𝑦 0 ] ∙ ⃗∇𝑇 (3.21)
𝑞𝑧 0 0 𝑘𝑧𝑧

1000
MgB2
Niobium
800 Copper
Monel
Specific Heat (J/Kg-K)

CTD 101k
600

400

200

0
0 50 100 150 200 250 300
Temperature (K)
Figure 3.7. Heat Capacity of constituent materials. [224],[237-239]
The thermal conductivity of the constituent materials of the composite wire is

assumed isotropic and a function of temperature. The thermal conductivities are

extracted from references [224], [234]–[236] and is presented in Figure 3.6.

59
 Chapter 3

3.5 Specific Heat (C)

Heat capacity, C, by definition indicates the amount of heat required to raise the

temperature of a material to a certain degree within a specified time. Considering the

Fourier heat equation for transient cases,

𝜕𝑇
𝑐𝜌 = [k]∇2 𝑇 + 𝑄 (3.22)
𝜕𝑡

In the equation, Q is the internal heat generation, 𝑐 is the heat capacity and 𝜌 is

the density. Units of heat capacity in SI is Joule per kelvin and experimentally calculated

using the equation (2.12 a). The values of specific heat with varying temperature is given

in Figure 3.7 as complied from different references for MgB2 [237], [238], niobium [224],

copper [239], monel [224], and CTD 101k as epoxy [224].

1E-3
RRR 50 MgB2
RRR 100 1E-3
Niobium
RRR 150 Copper 100
Resistivity of Copper (-m)

Resistivity of Copper (-m)

1E-4 RRR 300 1E-5 Monel

RRR 500

1E-7

1E-5
1E-9

1E-11
1E-6

1E-13

1E-7 1E-15
1 10 100 1 10 100
Temperature (K) Temperature (K)

Figure 3.8. Electrical resistivity of constituent materials. [224], [240]–[243]

3.6 Resistivity (ρ)

The resistivity of a material defines how strongly a material opposes the flow of

electric current. Depending on the purity of copper, the resistivity can vary a little with

temperature. Residual resistivity ratio (RRR) values are a good indication of

60
 Material Properties for Magnet Modeling

crystallographic impurities. The RRR values are defined as a ratio of resistivity at 300 K to

resistivity at 0 K. Practically, it is impossible to achieve a temperature of 0 K. However,

some approximation is made to estimate the resistivity at that temperature. Resistivity is

defined using the equation as follows

𝑅𝐴
𝜌= (3.23)
𝐿

In the equation, ‘R’ is the measured resistance, ‘A’ is the cross-sectional area, and

‘L’ is the length of the specimen. The resistivities for different RRR values of copper and

different constituent material are presented in figure 8. The data are obtained from

references [224], [240]–[243].

3.7 Conclusion

In this chapter, the basic material properties of the constituents of the composite

is discussed. Their temperature dependent material properties are extracted and

compiled from different resources ranging from 10 K to 300 K. These material properties

are necessary for the modeling of the MRI magnet system. The material properties listed

in this work are from different research groups. Depending on testing condition and

materials used, it is possible that the values vary slightly. However, it is expected that the

material properties will to good approximation represent the real world behavior when

they are used in superconducting composite wires

61
 Chapter 4. Numerical Homogenization of Superconducting

Composite Wire for Material Properties

“Model building is the art of selecting those aspects of a process that are relevant to the

question being asked” — John Holland, Hidden Order: How Adaptation Builds Complexity,

(1995)

4.1 Composite Wire

Superconducting wires are metal matrix composite (MMC)1 made of different

components (Figure 4.1). Typical wires consist of superconducting fibers inside a copper

matrix. MgB2 is a ceramic compound and is highly strain sensitive. The MgB2

Figure 4.1. Typical configuration of composite superconducting wires. The wire

constitutes the top volume of length scale ‘L’

1
A specific type of composite that contains at least two different types of material of which one is metal
and the other is ceramic or polymer.

62
 Numerical Homogenization of Superconducting Composite Wire for Material Properties

superconducting wire consists of a complex composite. The superconducting MgB2

filaments are surrounded by a layer of metal such as niobium, titanium or vanadium in

order to prevent the diffusion of copper into the MgB2 powder filaments during the

necessary thermal heat treatment at 650° C. The MgB2-Niobium/Titanium/Vanadium

filaments are contained in a matrix of stabilization metal, e.g. copper or aluminum.

Because the thermal and electrical conductivities of such metals are higher compared to

the wire composite’s other materials, most of the electrical current and heat is carried

away by the stabilization matrix during a magnet quench, in which the wire loses its

superconducting properties. A quench [156], [157], [159] of an MRI superconducting

magnet causes a substantial temperature rise within a concise time (~2s) due to

deposition of its high stored energy (on the order of megajoules) at the location of the

quench. MRI magnets are always equipped with quench protection systems to limit the

temperature rise and prevent potential safety hazards. Although the magnet is quenched

during an emergency shutdown, the magnet will experience irreversible deformation

leading to destruction, whether the quench is uncontrolled or the wires are ill-designed

[244]. A copper alloy sheath, e.g., Monel or Glidcop, required for drawing purposes,

surrounds the stabilization matrix and filaments and carries most of the mechanical loads

attributed to the elastic modulus. At the outer-most location, a layer of epoxy acts as an

electrical insulator for transient voltage in a wire bundle during the charging and

quenching of the magnet. A typical superconducting MRI magnet consists of a set of

individual bundles containing multiple layers, each consisting of multiple turns. The

63
Chapter 4

microscopic image and their CAD model of the wires considered for the magnet design

are shown in Figure 4.2.

Figure 4.2. Microscopic images and CAD models of the superconducting wires for MRI

background magnets (a) Cir#518, (b) Cir#632, (c) Cir#600, (d) Rec#1027.

4.2 MgB2 Composite Superconducting Wire

Hyper Tech Research Inc manufactures different types of MgB2 superconducting

wire for manufacturing the full body MRI magnet system. Hyper Tech Research has

provided the microscopic images of the wires (Figure 4.2). The microscopic images are

then imported into the 3D modeling software Creo Parametric. 3D models of the wires

are imported into ANSYS, meshed accordingly with 3D brick element SOLID 186. Suitable

material properties are then assigned to the specified volume. A summary of the wire

composition is listed in Table 4.1. Methods to numerically homogenize the composite

wire are elaborated by Barbero [108], [109] and have been previously employed by Boso

[81], [121] and Amin [158] for composite superconducting wires. A similar approach as

discussed initially is taken to homogenize the composite wires numerically. These

64
 Numerical Homogenization of Superconducting Composite Wire for Material Properties

properties are used in the FEA magnet model to compute the stress and strain

development during winding, cool-down and electromagnetic charging.

4.3 Representative Volume Element (RVE)

Computational modeling of MMC superconducting wire is cumbersome and

computationally intensive if every finest detail is considered. Such level of detail requires

extensive computational resources as a solution to an enormous number of unknowns

are needed. Thus, homogenization is a viable approach that connects materials at

Table 4.1: Volume fraction of the composite superconducting MgB2 wires

Volume Fraction (%)

Diameter or
Wire ID Epoxy (CTD
Dimensions (mm) MgB2 Niobium Copper Monel
101k)

CIR#518 1 8.78 8.65 19.37 26.20 37.00

CIR#632 0.84 9.57 9.93 15.93 23.74 40.83

CIR#600 0.834 8.31 11.23 16.30 23.16 41.00

REC#1027 1.03×1.554 6.91 16.56 18.02 27.27 31.24

different length scale, i.e., microscale to the macro scale. Shifting between these two

length scales facilitates the modeling of an entire background magnet of meter length

scale considering a composite superconducting wire of millimeter-scale even with a

regular moderate performance workstation. In the most straightforward approach,

homogenization of composite wire for elastic properties (Figure 4.3) has two bounds:

Reuss model and Voigt model. In the Voigt model, the strain tensor is assumed constant

65
Chapter 4

over the micro and macro scale geometry. In other words, the effective modulus is

calculated using the formula

𝐸 = 𝛦0 𝜑0 + 𝛦1 𝜑1 + 𝛦2 𝜑2 (4.1)

In the equation, 𝜑0 , 𝜑1 and 𝜑2 are the volume fraction of the fiber, matrix and

epoxy respectively. 𝛦0 , 𝛦1 and 𝛦2 indicates the elastic modulus of the materials

Figure 4.3. Typical configuration of composite superconducting wires.

respectively.

Similarly, for the Reuss model, the stress tensor or stiffness tensor is assumed

constant over the length scale and the effective elastic modulus is estimated using the

equation

1 𝜑0 𝜑1 𝜑2
= + + (4.2)
𝐸 𝛦0 𝛦1 𝛦2

These two approximations hold true when the filament densities are relatively low

as there then exists a linearity. However, when the density of the fibers increases,

nonlinearity is introduced, and computational modeling is required as compared to the

straightforward use of the equations (4.1) and (4.2) above. Mathematically, the

dimensions of the filaments should be infinitesimally smaller than the dimensions of the

66
 Numerical Homogenization of Superconducting Composite Wire for Material Properties

𝐷
matrix for the homogenization to be accurate i.e. → ∞ (Figure 4.1). However,
𝑑

considering the accuracy of the result and the computational cost involved, typical values

𝐷
of 𝑑 are chosen as 10 [245]. If an ensemble average is used, the value could be as low as

2 [246]. Another important consideration is the length scale of the macroscopic problem

𝐿
for the homogenized structure. Ideally, → ∞, but then again computational cost is
𝐷

sometime a big concern over the system fidelity, limiting the value of the ratio with a

finite number. The basic idea is that the dimension of the homogenized structure should

be small enough such that it is a valid material point at the macro scale to represent the

composite structure, and it is large enough at the micro length scales at which it

represents the behavior of microstructures with acceptable fidelity and efficiency.

4.4 Modeling of Composite wire

Several techniques to estimate the properties of the wire is reviewed in Chapter

2. In this section, techniques used in this study to model the material properties of the

superconducting background magnet system will be elaborated.

4.4.1 Computational Homogenization of Elastic Properties

Available numerical methods that preserve reasonable fidelity with acceptable

efficiency includes but are not limited to finite element analysis (FEA), boundary element,

and finite difference methods. Among several other available techniques to numerically

homogenize a composite superconducting wire, the simple variational approach

introduced by Luciano [110] combined with FEA, and presented in Barbero’s [108], [109]

book is discussed and implemented in this work. A similar approach to modeling material

67
Chapter 4

properties of Nb3Sn wire strands has also been implemented by Collins [247] and Boso

[81], [121]. A full wire model of the representative volume element (RVE) is designed and

meshed in which unit normal and shear strains are applied one at a time [110].

The geometry of the composite wire introduced by Barbero [108], [109] considers

cylindrical filaments in a matrix. Such microstructure behaves as an orthotropic manner

Figure 4.4. Dimensions of the RVE used for numerical homogenization.

at mesoscale, and the periodic microstructure facilitates the use of an RVE for finite

element homogenization. The compliance matrix of a material that behaves in an

orthotropic manner relates the stress and strain according to equation (3.7) as described

in chapter 3. Homogenized material property in its simplest form is obtained through the

basics of volume average methods. Under a macroscopic applied average displacement

or traction boundary condition, average stress or strain developed in the RVE is given by

the equation,

1
𝜎̅̅
̅̅𝑖𝑗 = ∫𝜎 𝑑𝑉 (4.3)
𝑉 𝑣 𝑖𝑗

68
 Numerical Homogenization of Superconducting Composite Wire for Material Properties

1
𝜀𝑖𝑗 =
̅̅̅ ∫𝜀 𝑑𝑉 (4.4)
𝑉 𝑣 𝑖𝑗

Homogeneous boundary conditions of constant stress or strain under equation

(4.3) and (4.4) are obtained either by applying a displacement or traction boundary

condition.

0
𝑢𝑖 (𝑆) = 𝜀𝑖𝑗 𝑥𝑗 (4.5)

𝑡𝑖 = 𝜎𝑖𝑗0 𝑛𝑗 (4.6)

In the equation, 𝑢𝑖 and 𝑡𝑖 are the displacement and traction boundary conditions

with 𝑥𝑗 and 𝑛𝑗 the direction parallel and perpendicular to the applied displacement and

traction. In this work, average stress methods are chosen to estimate the elastic

properties as demonstrated by Barbero [108], [109]. The underlying equation to estimate

the effective elastic constant of the stiffness matrix for a material point x is

∗
1 𝑘𝑙 𝑘𝑙
𝑆𝑖𝑗𝑘𝑙 = ∫ 𝑆 (𝑥)(𝑢𝑝,𝑞 + 𝑢𝑞,𝑝 )𝑑𝑉 (4.7)
2𝑉 𝑉 𝑖𝑗𝑝𝑞

where, 𝑆𝑖𝑗𝑝𝑞 (𝑥) is the material compliance tensor at point x and developed strain

1 0 𝑘𝑙 𝑘𝑙 0
𝜀𝑝𝑞 (𝑥) = 2 𝜀𝑘𝑙 (𝑢𝑝,𝑞 + 𝑢𝑞,𝑝 ) under the applied unit strain of 𝜀𝑘𝑙 = 1. This applied unit

strain generates an average unit strain over the RVE element [248]. The constants of each

of the stiffness matrix, S* is calculated using equation (45). Considering the origin of the

coordinate at the center of the rectangular RVE shown in figure 14, the sides are x1, x2

and x3 respectively in the specified directions. Using equation (24), the applied strain

equation transforms into (Figure 4.4):

69
Chapter 4

𝑋 𝑋
𝑢𝑥 ( 22 , 𝑦, 𝑧) − 𝑢𝑥 (− 22 , 𝑦, 𝑧)
0
𝜀̅̅̅̅
𝑥𝑥 = 𝜀𝑥2 =
𝑋2
𝑋3 𝑋
𝑢𝑦 (𝑥, 2 , 𝑧) − 𝑢𝑦 (𝑥, − 23 , 𝑧)
𝜀̅̅̅̅ 0 (4.8)
𝑦𝑦 = 𝜀𝑥3 =
𝑋3
𝑋1 𝑋
𝑢𝑧 (𝑥, 𝑦, 2 ) − 𝑢𝑧 (𝑥, 𝑦, − 21 )
0
𝜀𝑧𝑧 = 𝜀𝑥1
̅̅̅̅ =
𝑋1

For ease of use, setting X1 = 2a1, X2 = 2a2, and X3 = 2a3 and changing into X1-X2-X3

coordinate equation (4.8) changes into

0
𝑢𝑖 (𝑎1 , 𝑥2 , 𝑥3 ) − 𝑢𝑖 (−𝑎1 , 𝑥2 , 𝑥3 ) = 2𝑎1 𝜀𝑖1
0 (4.9)
𝑢𝑖 (𝑥1 , 𝑎2 , 𝑥3 ) − 𝑢𝑖 (𝑥1 , −𝛼2 , 𝑥3 ) = 2𝑎2 𝜀𝑖2
0
𝑢𝑖 (𝑥1 , 𝑥2 , 𝑎3 ) − 𝑢𝑖 (𝑥1 , 𝑥2 , −𝛼3 ) = 2𝑎3 𝜀𝑖3

Using the equation (4.9) each component of the stiffness tensor 𝑺 is calculated by

setting each component of the strain tensor in equation (3.7) to a unit values. For

example, the first column of the compliance matrix is calculated by setting,

𝜀10 = 1, 𝑎𝑛𝑑 𝜀20 = 𝜀30 = 𝛾40 = 𝛾50 = 𝛾60 = 0 (4.10)

whereas the stress and strains are related using the same equation (3.11) as

𝜎1 𝐶11 𝐶12 𝐶13 0 0 0 𝜀1

𝜎2 𝐶21 𝐶22 𝐶23 0 0 0 𝜀2
𝜎3 𝐶31 𝐶32 𝐶33 0 0 0 𝜀3
𝜏4 = 0 𝛾4 (4.11)
0 0 𝐶44 0 0
𝜏5 0 0 0 0 𝐶55 0 𝛾5
{ 𝜏6 } [ 0 0 0 0 0 𝐶66 ] 6 }
{ 𝛾

70
 Numerical Homogenization of Superconducting Composite Wire for Material Properties

Application of boundary condition from equation 4.10 converts to the boundary

conditions of equation (4.9) into following forms

𝑢1 (𝑎1 , 𝑥2 , 𝑥3 ) − 𝑢1 (−𝑎1 , 𝑥2 , 𝑥3 ) = 2𝑎1

𝑢2 (𝑎1 , 𝑥2 , 𝑥3 ) − 𝑢2 (−𝑎1 , 𝑥2 , 𝑥3 ) = 0
𝑢3 (𝑎1 , 𝑥2 , 𝑥3 ) − 𝑢3 (−𝑎1 , 𝑥2 , 𝑥3 ) = 0
.
.
. (4.12)
𝑢𝑖 (𝑥1 , 𝑎2 , 𝑥3 ) − 𝑢𝑖 (𝑥1 , 𝑎2 , 𝑥3 ) = 0
.
.
.
𝑢𝑖 (𝑥1 , 𝑥2 , 𝑎3 ) − 𝑢𝑖 (𝑥1 , 𝑥2 , −𝑎3 ) = 0

The first term of equation (4.12) indicates the applied displacement along

direction 1 on plane 𝑛
̂.
1 The next two terms indicate zero traction boundary conditions

along the other orthogonal direction on the plane 𝑛

̂.
1 The last two terms indicates the

displacement along the planes 𝑛

̂2 and 𝑛
̂3 as zero and the traction boundary conditions

also as zero. The applied strain generates an average stress that is possible to calculate

using the FEA program ANSYS. The average values of the calculated stresses then

represent the constants of the stiffness matrix S as,

1
𝑆𝑖1 = 𝜎̅𝑖 = ∫ 𝜎 (𝑥 , 𝑥 , 𝑥 ) (4.13)
𝑉 𝑉 𝑖 1 2 3

The compliance tensor S is related to the elastic property constants. Replacing the

desired constants in Equation (4.11) with the elastic modulus, Poisson’s ratio, and shear

modulus provides the following form of the stress-strain equation:

71
Chapter 4

1 𝜈12 𝜈13
− − 0 0 0
𝐸1 𝐸2 𝐸3
𝜀11 𝜈21 1 𝜈23 𝜎11
𝜀22 − − 0 0 0 𝜎22
𝜀33 𝐸1 𝐸2 𝐸3 𝜎33
𝜀12 = − 𝜈31 𝜈32 1 𝜎12 (4.14)
− 0 0 0
𝜀13 𝐸1 𝐸2 𝐸3 𝜎13
{𝜀23 } 0 0 0 𝐺12 0 0 {𝜎 23 }
0 0 0 0 𝐺13 0
[ 0 0 0 0 0 𝐺23 ]

Each of the constants in the stiffness matrix in Equation (4.11) is calculated by

applying an average strain as described by Luciano [110], detailed by Barbero [249] and

implemented for superconducting wires by Boso [121]. To both obtain and solve for each

of the individual constants in the stiffness matrix, the unit strain is applied to six different

cases. In tensor form, the six individual strain loading terms have the following form:

1 1 0 0 0
1 0 0 0 0 0 0 0 0 0 0 0 0 1
2 2 0 0
𝜀 0 = [0 0 0] , [0 1 0] , [0 0 0] , 1 , 0 0 0 , 2 (4.15)
0 0 0 0 0 0 0 0 1 2 0 0 1 1
[0 0 0] [0 0]
0 0] [2 2

The last three cases in equation (4.15) represent the unit shear strain, which is

divided into equal parts to constitute the unit strain for each shear loading. The strain

tensor in equation (4.11) is usually taken as the Cauchy strain tensor because it is

symmetric compared to the deformation gradient tensor. Thus, the strain tensor of

equation (4.11) is expressed regarding the displacement gradient as follows:

72
 Numerical Homogenization of Superconducting Composite Wire for Material Properties

1 𝜕𝑢1 𝜕𝑢1 1 𝜕𝑢1 𝜕𝑢2 1 𝜕𝑢1 𝜕𝑢3

𝜀11 = ( + ) 𝜀12 = ( + ) 𝜀13 = ( + )
2 𝜕𝑥1 𝜕𝑥1 2 𝜕𝑥2 𝜕𝑥1 2 𝜕𝑥3 𝜕𝑥1
1 𝜕𝑢2 𝜕𝑢1 1 𝜕𝑢2 𝜕𝑢2 1 𝜕𝑢2 𝜕𝑢3
𝜀21 = ( + ) 𝜀22 = ( + ) 𝜀23 = ( + ) (4.16)
2 𝜕𝑥1 𝜕𝑥2 2 𝜕𝑥2 𝜕𝑥2 2 𝜕𝑥3 𝜕𝑥2
1 𝜕𝑢3 𝜕𝑢1 1 𝜕𝑢3 𝜕𝑢2 1 𝜕𝑢3 𝜕𝑢3
𝜀31 = ( + ) 𝜀32 = ( + ) 𝜀33 = ( + )
2 𝜕𝑥1 𝜕𝑥3 2 𝜕𝑥2 𝜕𝑥3 2 𝜕𝑥3 𝜕𝑥3

The strain on the RVE is applied by using Equation (4.16) in terms the of

displacement boundary conditions in ANSYS. A hexahedral RVE as shown in Figure 4.4

with dimensions a1, a2, and a3.

Figure 4.5. Determining the RVE of a composite wire.

The applied unit strain creates a complex strain state inside the RVE. Equation

(4.11) for a unit applied strain (ε11 = 1) then becomes:

𝜎11 𝐶1111 𝐶1122 𝐶1133 0 0 0 1

𝜎22 𝐶2211 𝐶2222 𝐶2233 0 0 0 0
𝜎33 𝐶3311 𝐶3322 𝐶3333 0 0 0 0
𝜎12 = 0 (4.17)
0 0 𝐶1212 0 0 0
𝜎13 0 0 0 0 𝐶1313 0 0
{𝜎23 } [ 0 0 0 0 0 𝐶2323 ] 0}
{

73
Chapter 4

The values of the elements of the stiffness tensor C in Equation (4.17) can be

estimated from the average values of the stress, which is calculated using equation (4.13).

4.4.2 Reducing the Full RVE

Determining the RVE for fiber in the composite is thoroughly discussed in the book

by Barbero [108]. The periodic microstructure of a fiber in the matrix is identified as

shown in Figure 4.5. Periodic boundary conditions are applied on the six surfaces, twelve

edges and eight nodes of the RVE. Modifying the equation (4.12) according to the RVE

0 1
detected, the constraint equation for edges for applied shear strain of 𝜀12 = 2 for 𝑥1 =

±𝑎1 and 𝑥2 = ±𝑎2 becomes [108]

𝑢1 (𝑎1 , 𝑎2 , 𝑥3 ) − 𝑢1 (−𝑎1 , −𝑎2 , 𝑥3 ) = 𝑎2

𝑢2 (𝑎1 , 𝑎2 , 𝑥3 ) − 𝑢2 (−𝑎1 , −𝑎2 , 𝑥3 ) = 𝑎1 −𝑎3 < 𝑥3 < 𝑎3
𝑢3 (𝑎1 , 𝑎2 , 𝑥3 ) − 𝑢3 (−𝑎1 , −𝑎2 , 𝑥3 ) = 0
(4.18)
𝑢1 (𝑎1 , −𝑎2 , 𝑥3 ) − 𝑢1 (−𝑎1 , 𝑎2 , 𝑥3 ) = −𝑎2
𝑢2 (𝑎1 , −𝑎2 , 𝑥3 ) − 𝑢2 (−𝑎1 , 𝑎2 , 𝑥3 ) = 𝑎1 −𝑎3 < 𝑥3 < 𝑎3
𝑢3 (𝑎1 , −𝑎2 , 𝑥3 ) − 𝑢3 (−𝑎1 , 𝑎2 , 𝑥3 ) = 0

In a similar fashion for 𝑥1 = ±𝑎1 and 𝑥3 = ±𝑎3 , the edge constraint equations

becomes

𝑢1 (𝑎1 , 𝑥2 , 𝑎3 ) − 𝑢1 (−𝑎1 , 𝑥2 , −𝑎3 ) = 0

𝑢2 (𝑎1 , 𝑥2 , 𝑎3 ) − 𝑢2 (−𝑎1 , 𝑥2 , −𝑎3 ) = 𝑎1 −𝑎2 < 𝑥2 < 𝑎2
𝑢3 (𝑎1 , 𝑥2 , 𝑎3 ) − 𝑢3 (−𝑎1 , 𝑥2 , −𝑎3 ) = 0
(4.19)
𝑢1 (𝑎1 , 𝑥2 , −𝑎3 ) − 𝑢1 (−𝑎1 , 𝑥2 , 𝑎3 ) = 0
𝑢2 (𝑎1 , 𝑥2 , 𝑎3 ) − 𝑢2 (−𝑎1 , 𝑥2 , 𝑎3 ) = 𝑎1 −𝑎2 < 𝑥2 < 𝑎2
𝑢3 (𝑎1 , 𝑥2 , −𝑎3 ) − 𝑢3 (−𝑎1 , 𝑥2 , 𝑎3 ) = 0

And for 𝑥2 = ±𝑎2 and 𝑥3 = ±𝑎3 the equation becomes

74
 Numerical Homogenization of Superconducting Composite Wire for Material Properties

𝑢1 (𝑥1 , 𝑎2 , 𝑎3 ) − 𝑢1 (𝑥1 , −𝑎2 , −𝑎3 ) = 𝑎2

𝑢2 (𝑥1 , 𝑎2 , 𝑎3 ) − 𝑢2 (𝑥1 , −𝑎2 , −𝑎3 ) = 0 −𝑎1 < 𝑥1 < 𝑎1
𝑢3 (𝑥1 , 𝑎2 , 𝑎3 ) − 𝑢3 (𝑥1 , −𝑎2 , −𝑎3 ) = 0
(4.20)
𝑢1 (𝑥1 , 𝑎2 , −𝑎3 ) − 𝑢1 (𝑥1 , −𝑎2 , 𝑎3 ) = 𝑎2
𝑢2 (𝑥1 , 𝑎2 , −𝑎3 ) − 𝑢2 (𝑥1 , −𝑎2 , 𝑎3 ) = 0 −𝑎1 < 𝑥1 < 𝑎1
𝑢3 (𝑥1 , 𝑎2 , −𝑎3 ) − 𝑢3 (𝑥1 , −𝑎2 , 𝑎3 ) = 0

Similarly, the constraint equations for the edges are applied, and they are [108]

𝑢1 (𝑎1 , 𝑎2 , 𝑎3 ) − 𝑢1 (−𝑎1 , −𝑎2 , −𝑎3 ) = 𝑎2

𝑢2 (𝑎1 , 𝑎2 , 𝑎3 ) − 𝑢2 (−𝑎1 , −𝑎2 , −𝑎3 ) = 𝑎1
𝑢3 (𝑎1 , 𝑎2 , 𝑎3 ) − 𝑢3 (−𝑎1 , −𝑎2 , −𝑎3 ) = 0

𝑢1 (𝑎1 , 𝑎2 , −𝑎3 ) − 𝑢1 (−𝑎1 , −𝑎2 , 𝑎3 ) = 𝑎2

𝑢2 (𝑎1 , 𝑎2 , −𝑎3 ) − 𝑢2 (−𝑎1 , −𝑎2 , 𝑎3 ) = 𝑎1
𝑢3 (𝑎1 , 𝑎2 , −𝑎3 ) − 𝑢3 (−𝑎1 , −𝑎2 , 𝑎3 ) = 0
(4.21)
𝑢1 (−𝑎1 , 𝑎2 , 𝑎3 ) − 𝑢1 (𝑎1 , −𝑎2 , −𝑎3 ) = 𝑎2
𝑢2 (−𝑎1 , 𝑎2 , 𝑎3 ) − 𝑢2 (𝑎1 , −𝑎2 , −𝑎3 ) = −𝑎1
𝑢3 (−𝑎1 , 𝑎2 , 𝑎3 ) − 𝑢3 (𝑎1 , −𝑎2 , −𝑎3 ) = 0

𝑢1 (𝑎1 , −𝑎2 , 𝑎3 ) − 𝑢1 (−𝑎1 , 𝑎2 , −𝑎3 ) = −𝑎2

𝑢2 (𝑎1 , −𝑎2 , 𝑎3 ) − 𝑢2 (−𝑎1 , 𝑎2 , −𝑎3 ) = 𝑎1
𝑢3 (𝑎1 , −𝑎2 , 𝑎3 ) − 𝑢3 (−𝑎1 , 𝑎2 , −𝑎3 ) = 0

Applying these constraint equations are necessary to generate required shear

strain. It is also important to reduce the constraint equations as shown in the equations

above because when a constraint equation is applied to a specified node, it is not possible

to apply another constraint equation on the same node. Therefore, the necessary

equations are reduced to eighteen sets of edge constraint equation and 12 sets of vertex

equations as shown in (4.18)-(4.21). Applying all these constraint equations is time-

consuming and increases the solution time in FEA program. There is an alternate way,

where it is possible to make use of symmetry and asymmetry boundary conditions and

75
Chapter 4

reduce the problem into 1/8th of the RVE (Figure 4.6). Reduction of the RVE into a smaller

symmetric part not only reduces the number of elements, but it also eliminates the time

Figure 4.6 Reduction of RVE into 1/8th of the volume.

required for applying the constraint equations. Thus the solution time is faster compared

to the full RVE solution and there is only insignificant variation in the property values

calculated.

Antisymmetric boundary conditions are displacement boundary conditions with

the rotation applied about an axis. Making use of antisymmetry, the composite wire is

homogenized, and the results are compared for deviation when estimating the shear

modulus of the RVE. The applied boundary conditions are detailed in Figure 4.8.

Comparison of the homogenized elastic property values is summarized in Table 4.2. The

RVE considered has similar dimensions and material properties as elaborated in [108, p.

76
 Numerical Homogenization of Superconducting Composite Wire for Material Properties

170]. It is clear that the introduction of symmetry and antisymmetry reduces the problem

without a significant change in the homogenization results. Therefore, reduced

symmetric geometry is used to homogenize the composite wire properties.

Table 4.2 Comparison of the elastic property values using symmetry and

antisymmetry boundary conditions

Reduced
Elastic Property Full Geometry (%) Error
Geometry

𝑬𝟏𝟏 98.2993 98.3023 0.00313

𝑬𝟐𝟐 7.47651 7.47695 0.00582

𝑬𝟑𝟑 7.47818 7.47431 0.05166

𝝂𝟐𝟏 22.6959 22.7329 0.16281

𝝂𝟑𝟏 22.7169 22.7183 0.00591

𝝂𝟏𝟐 298.394 298.802 0.13687

𝝂𝟑𝟐 540.105 540.483 0.06994

𝝂𝟏𝟑 298.637 298.863 0.07578

𝝂𝟐𝟑 539.988 540.76 0.14307

𝑮𝟐𝟑 2.42369 2.42341 0.0114

𝑮𝟑𝟏 2.58252 2.42534 6.08638

𝑮𝟏𝟐 2.58186 2.44192 5.42018

77
Chapter 4

4.4.3 Computational Homogenization of Thermal Expansion Coefficient

The thermal expansion coefficients can be estimated from the calculated average

stress with appropriate boundary conditions. The constitutive equation relating the

thermal expansion and material elasticity is:

𝐶𝑖𝑗𝑘𝑙 : 𝜎𝑖𝑗 + 𝛥𝛵𝛼𝑘𝑙 = 0 (4.22)

where 𝛥𝛵 is the temperature difference, and 𝛼𝑘𝑙 is the thermal expansion

coefficient. Initially, no external strain is applied when the unit temperature difference is

applied on the RVE. The change in thermal loading generates strain inside the RVE due to

the different thermal expansion coefficients of the constituent materials. All of the

materials except CTD 101k have an isotropic thermal expansion coefficient. CTD 101k has

different thermal expansion coefficients in different plane orientations due to the

temperature gradients. Thus, the homogenized RVE exhibits significant orthotropic

thermal expansion behavior. In order to calculate the thermal expansion coefficient from

the average strain change, the following equation is used,

1 1
𝛼𝑘𝑙 = 𝜀𝑘𝑙 = ∫ 𝜀𝑘𝑙 𝑑𝑉
̅̅̅̅ (4.23)
𝛥𝑇 𝑉 𝑉

78
 Numerical Homogenization of Superconducting Composite Wire for Material Properties

Figure 4.7. Steps in determining the representative volume element for a superconducting

magnet bundle made of superconducting metal matrix composites. (a) Cir 518 RVE is

homogenized and 1st principal strain is shown for applied strain along direction 1. (b) Cir 632 RVE

is homogenized and 1st principal strain is shown for applied temperature at 300 K. (c) Cir 600 RVE

is homogenized and temperature plot is shown when temperature gradient is applied at the top

and bottom surface of RVE. (d) Rec 1027 is homogenized and temperature plot is shown for

calculation of specific heat. The plot represents the temperature increase from 300 K.

4.4.4 Computational Homogenization of Thermal Conductivity

The homogenization of the thermal conductivity involves Fourier’s law of heat

conduction, which relates the heat flux with the temperature gradient:
79
 Chapter 4

Figure 4.8. Applied boundary condition for homogenizing the RVE. From the top left to the

bottom right, the first six boundary conditions are for calculating the stiffness matrix. Last row in

the picture represents the boundary conditions for thermal expansion coefficient, thermal

conductivity (k11), and specific heat respectively.

𝑞𝑖𝑖 = 𝑘𝑖𝑗 𝑇𝑗,𝑗 (4.24)

where 𝑇𝑗,𝑗 is the temperature gradient along direction 𝑗, 𝑘 is the thermal

conductivity tensor, and 𝑞𝑖𝑖 is the heat flux along direction 𝑖. The thermal conductivity of

80
 Numerical Homogenization of Superconducting Composite Wire for Material Properties

the RVE can be estimated by applying either a uniform temperature gradient or uniform

heat flux at the boundaries [250]. Since the goal of this article is to estimate the thermal

conductivity at different temperatures, the applied unit temperature gradient must be

small enough to exclude any nonlinearities but large enough to avoid numerical

truncation errors [153]. Thus, once a unit temperature is applied across the cell, equation

(4.24) takes the following tensor form:

𝑞11 𝑘11 𝑘12 𝑘13 𝑇1,1

𝑞
[ 22 ] = [𝑘21 𝑘22 𝑘23 ] [ 0 ] (4.25)
𝑞33 𝑘31 𝑘32 𝑘33 0

The thermal conductivity of a heterogeneous metal matrix composite can be

estimated from equation (4.25). Similar techniques have been used for multiscale analysis

of heterogeneous solids [142]–[150], high-temperature superconducting wires [251], and

plain-weave fabrics [152]. The volume average of the heat flux for the applied

temperature difference is then calculated in ANSYS using an equation similar to equations

(4.13) and (4.23):

1 1
𝑘𝑖𝑗 = [ ∫ 𝑞 𝑑𝑉] (4.26)
𝑇𝑗,𝑗 𝑉 𝑉 𝑖𝑖

Equation (4.26) is used to estimate the thermal conductivity with FEA methods.

The analytical rule of mixtures method can be used to calculate the thermal conductivity

along the longitudinal direction of the composite wire because all the constituent

components are in a parallel orientation along the longitudinal (azimuthal) direction

(direction 1). The method for estimating the thermal conductivity by linear mixing rule of

81
Chapter 4

a composite is discussed in section 2.2.3 and modified form from equation (2.7) turns

into:

𝑘𝑙 = ∑ 𝑓𝑖 𝑘𝑖 (4.27)

The thermal conductivity along the radial (direction 2) and axial directions

(direction 3) are mainly driven by the conductivity of the epoxy since its thermal

conductivity is at least an order of magnitude smaller compared to other components

(Figure 4.7). Thus, the thermal conductivity of the RVE of the composite in the direction

2 and direction 3 is calculated following the series model using a simplified version of the

equation (2.8) as presented in [156]:

𝑏
𝑘𝑖𝑗 = 𝑘 (𝑇) (4.28)
𝛿 𝑖𝑛𝑠𝑢𝑙𝑎𝑡𝑜𝑟

In Equation (4.28), 𝑏 and 𝛿 are the thickness of epoxy and thickness of the wire

along the direction of heat flow and, 𝑘𝑖𝑛𝑠𝑢𝑙𝑎𝑡𝑜𝑟 (𝑇) is the thermal conductivity of the

epoxy at a specified temperature.

4.4.5 Computational Homogenization of Specific Heat

The methods to calculate the specific heat of the composite wire is elaborated in

section 2.2.4. In FEA, the transient heat equation is solved to estimate the specific heat.

Zero strain boundary conditions are applied for calculating the specific heat of constant

pressure, and zero traction boundary condition is applied to the specific heat of constant

volume. The underlying equations are (2.12 a).

82
 Numerical Homogenization of Superconducting Composite Wire for Material Properties

The boundary conditions for the different material property estimations are

presented in Figure 4.8, which explains how the applied boundary conditions are

implemented for each case. However, the traction-free boundary condition 𝜎𝑖𝑗 = 0 is not

shown to avoid complexity.

4.5 Finite Element Discretization of RVE

The RVE is determined from the microstructural repetition on a bundle of the

magnet system as shown in Figure 4.7. In the superconducting magnet bundles, a layer of

wire is wound around a mandrel or former. Typically, the mandrel consists of non-

magnetic steel alloys. The wires are wound around the mandrel in terms of loops. A layer

consists of a certain number of loops, and the number of the layers builds the magnet

bundles. The composite wire itself contains a layer of epoxy. After the winding is

complete, degassed liquid epoxy is passed through the S-glass braiding around each wire

under vacuum to remove any unwanted air pockets (VPI). Thus, the entire magnet bundle

could be thought of as a matrix of epoxy where the wires are present as fibers. As

explained in Figure 4.7, one magnet bundle cross-section schematic is presented for

different types of wire cross-sections. When circular wires are used, the wires constitute

a layer by a certain number of loops and are considered as the fibers inside the epoxy

matrix. The RVE is estimated for a bundle containing circular wires as shown in Figure 4.7.

The RVE is then modeled in a CAD software Creo Parametric and imported into ANSYS to

generate a mesh for analysis. The material properties of the heterogeneous MMC

microstructures are approximated by using standard finite element formulation and

applying either displacement or traction boundary conditions on the RVE [108], [121],

83
Chapter 4

[250]. However, symmetric boundary conditions [252], [253] reduces the number of

elements, which results in a smaller set of equations, and thus, the computational effort

is reduced without much loss in accuracy. Thus, the RVE is determined so that the

symmetric boundary could be implemented. Calculating the off-diagonal elements of the

compliance matrix in equation (4.17) is significantly simplified if anti-symmetry boundary

conditions are used. Numerical homogenization is much simpler with the symmetric

boundary condition for thermal analysis (thermal conductivity and heat capacity) as

4.0

3.5 Direction 2
100 0.40
Elastic & Shear Modulus (GPa)

Direction 3
3.0 Direction 1

Thermal Strain (mm/m)

E 1-1 E 2-2 Poisson's ratio
E 3-3 G 2-3
0.35 2.5
G 1-3 G 1-2
2-3 -2 2.0
-3
50 0.30 1.5

1.0

0.25 0.5

0.0
0 50 100 150 200 250 300 0 50 100 150 200 250 300
Temperature (K) Temperature (K)

500
Thermal Conductivity (W/m-K)

400
100
Specific Heat (J/kg-K)

Direction 1
Direction 2 300
Direction 3
10
200

100
1 FEA
Analytical
0
0 50 100 150 200 250 300 0 50 100 150 200 250 300
Temperature (K) Temperature (K)

Figure 4.9. Homogenized material properties for Cir#518 superconducting wire.

compared to the elasticity analysis.

84
 Numerical Homogenization of Superconducting Composite Wire for Material Properties

4.6 Summary of the Homogenized Properties of the Wires

The four different composite wires are numerically homogenized for the elastic

modulus, thermal strain, thermal conductivity, and specific heat as the boundary

conditions defined in Section 3 are applied. The elastic properties of the composite wire

are assumed to be orthotropic as suggested by other composite material analyses [108],

[109], [155]. Similar trends are observed in the homogenized material properties of the

wire (Cir#600) according to Figure 4.9. The elastic modulus and shear modulus each show

a decreasing trend with the temperature increase. This trend is natural for a metal matrix

composite and agrees with another metal composite study [254]. The change in the shear

modulus of the wire composites is relatively insignificant compared to its elastic

0.45 4.0

100 Direction 1
3.5
Direction 2
Elastic & Shear Modulus (GPa)

0.40
3.0
Direction 3
E 1-1 E 2-2
Thermal Strain (mm/m)

80
E 3-3 G 2-3
Poisson's Ratio

0.35 2.5
G 1-3 G 1-2
60  2-3  1-2
2.0
 1-3
0.30
1.5
40

0.25 1.0

20
0.5
0.20
0.0
0
0 50 100 150 200 250 300 0 50 100 150 200 250 300
Temperature (K) Temperature (K)

500

Direction 1
Direction 2
Thermal Conductivity (W/m-K)

Direction 3 400
100
Specific Heat (J/kg-K)

300

10
200

1 100
FEA
Analytical
0
0 50 100 150 200 250 300 0 50 100 150 200 250 300
Temperature (K) Temperature (K)

Figure 4.10. Homogenized material properties for Cir#600 superconducting wire.

85
Chapter 4

properties as shown in Figure 4.9. The Poisson’s ratio in the plane 2-3 of the composite

demonstrates significantly higher values compared to other planes (plane 1-2 and plane

1-3) and could be explained by considering the stress-strain relationship according to

Equation (4.14). Because of the higher elastic modulus in the direction 1, the Poisson’s

ratio on the other orthogonal planes is comparatively lower. A similar trend continues for

other wires of circular cross-section as shown in Figure 4.10, Figure 4.11, and Figure 4.12.

The dimension of round wire facilitates almost uniform elastic properties in the wire

cross-sectional plane. Figure 4.9, Figure 4.10, and Figure 4.11 indicate that the elastic

modulus of the circular wires is similar to the elastic modulus in the direction 2 and

direction 3 of the wire. The numerical differences in their properties are nearly negligible.

4.0

100 Direction 1
3.5
Direction 2
Elastic & Shear Modulus (GPa)

0.40
Direction 3
E 1-1 E 2-2 3.0
Thermal Strain (mm/m)

80 E 3-3 G 2-3
Poisson's Ratio

G 1-3 G 1-2 2.5

0.35
 2-3  1-2
2.0
60  1-3
1.5
0.30
40
1.0

0.25 0.5
20
0.0
0 50 100 150 200 250 300 0 50 100 150 200 250 300
Temperature (K) Temperature (K)

500

Direction 1
Direction 2
Thermal Conductivity (W/m-K)

Direction 3 400
100
Specific Heat (J/kg-K)

300

10
200

1 100
FEA
Analytical
0
0 50 100 150 200 250 300 0 50 100 150 200 250 300
Temperature (K) Temperature (K)

Figure 4.11. Homogenized material properties for Cir#632 superconducting wire.

86
 Numerical Homogenization of Superconducting Composite Wire for Material Properties

The same trends are also observed for the Poisson’s ratio for circular wires. However, due

to the slight variation in the dimension in the two different directions of the wire cross-

0.30
3.5 Direction 1
Direction 2
E 1-1 E 2-2 3.0 Direction 3
E 3-3 G 2-3

Thermal Strain (mm/m)

100
Elastic & Shear Modulus (GPa)

G 1-3 G 1-2 2.5

Poisson's ratio
2-3 -2
-3 2.0

1.5
50

1.0

0.5

0.0
0 0.25
0 50 100 150 200 250 300 0 50 100 150 200 250 300
Temperature (K) Temperature (K)

Direction 1
Direction 2
Thermal Conductivity (W/m-k)

400
100
Direction 3

Specific Heat (J/kg-K)

Experiment

10
200
Experiment

FEA
1
Analytical
0
0 50 100 150 200 250 300 0 50 100 150 200 250 300
Temperature (K) Temperature (K)

Figure 4.12. Homogenized material properties for Rec#1027 superconducting wire.

section, the elastic modulus along direction 2 and direction 3 are slightly different for the

rectangular cross-section wire. As shown Figure 4.12, a difference of 60 GPa is recorded

between the two different primary directions (direction 1 as compared to direction 2 &

direction 3). Again, the difference in elastic moduli between direction 2 and direction 3 is

small but noticeable with an average value of 3 GPa.

The thermal strains are calculated according to the methods described in section

2.2.2. Since the thermal strain generation for any given temperature is significantly

different for the two major directions (through the thickness and warp & fill) for epoxy

87
Chapter 4

(CTD 101k) the resultant thermal strain for the composite wires is also orthotropic. From

Figure 4.9, Figure 4.10, and Figure 4.11 the change in dimensions due to the temperature

change is significantly smaller along direction 1 compared to the other two directions (2

& 3), which originated from the thermal strain properties of the epoxy. Since this change

is small at cryogenic temperatures, the stress-strain development at the early stage of a

quench is delayed [157]. Nonetheless, at higher temperatures, the thermal strains are

relatively large with ~1.1 mm/m at 300 K for rec#1027. For the circular wires, the variation

increases with a maximum of ~1.4 mm/m for cir#600.

The thermal conductivities of the composite wires are estimated from the

resulting heat flux after a uniform temperature difference is applied across the two planes

of the composite wires as elaborated in section 0. Along with direction 1, the heat flow is

facilitated by the metals inside the wired composite. However, the heat flow is

significantly disrupted by the presence of insulating epoxy material along directions 2 and

3, and thermal insulation like behavior is observed along direction 2 and direction 3. Since

the Rec#1027 wire has different spatial dimensions along direction 2 and direction 3, the

thermal conductivity values are slightly varied in those dimensions, unlike the circular

wires, which have the same spatial dimensions in both direction 2 and direction 3. These

types of directional variations are completely absent for the specific heat, since it is a

scalar quantity and depends on the composite’s ability to retain heat. Thus, the specific

heat calculated using numerical homogenization agrees well with the analytical

equations. The experimental values of the thermal conductivity and specific heat are

obtained from experimental studies on an equivalent test cube for Rec#1027 wire

88
 Numerical Homogenization of Superconducting Composite Wire for Material Properties

supplied by Hyper Tech Research Inc. The tests were performed by C-Therm technologies

with TCi instruments at room temperature (300 K), and the results are plotted in Figure

4.12, which show a relatively good agreement between the numerical and experimental

values. The slight overestimate for the experimental thermal conductivity and

underestimate for the experimental specific heat values can be attributed to the smaller

average epoxy (CTD 101k) thickness of the test block. Due to the limitation of the test

facility for C-Therm TCi, the test specimen must be larger than the device’s 17 mm heat

sensor. In the process of building the test block out of the wire#1027, the epoxy thickness

could also be smaller than the actual wire. Since, the epoxy used in the wire has a lower

120 120

100 100
Elastic Modulus (GPa)

Elastic Modulus (GPa)

Voigt Voigt
Reuss Reuss
80 80 Voigt-Reuss-Hill
Voigt-Reuss-Hill
Direction 1 Direction 1
Direction 2 Direction 2
60 Direction 3 60 Direction 3

40 40

(a) Cir#518 (b) Cir#600

20 20
0 50 100 150 200 250 300 0 50 100 150 200 250 300
Temperature (K) Temperature (K)
120
140

100 Voigt
120
Elastic Modulus (GPa)

Elastic Modulus (GPa)

Voigt Reuss
Reuss Voigt-Reuss-Hill
80 100 Direction 1
Voigt-Reuss-Hill
Direction 1 Direction 2
Direction 2 Direction 3
60 80
Direction 3

60
40

(c) Cir#632 40 (d) Rec#1027

20
0 50 100 150 200 250 300 0 50 100 150 200 250 300
Temperature (K) Temperature (K)

Figure 4.13. Comparison of analytical and numerical homogenization schemes of

elastic stiffness for all four wires.

89
Chapter 4

thermal conductivity and higher specific heat, a reduction in the epoxy volume explains

the higher estimated thermal conductivity, and lower estimated specific heat.

The elastic properties are compared with the analytical approximations of Voigt,

Reuss and Voigt-Reuss-Hill described in Section 2.2.1. The Voigt approximation for a

composite usually provides the upper bound of the elastic properties whereas the Reuss

approximation provides the lower bound. Hill proposed the average of the two bounds

Table 4.3 Comparison of elastic modulus calculation

Computational
Rule of
Composite wires Experiment (GPa) Homogenization
Mixture (GPa)
(GPa)

CS36M wire [255].

Columbus 154 113 165

superconductors

HYT30S wire [57].

Hyper Tech Research, 119 140 141

Inc.

as an approximate of the composite. However, all three of these approximations have

their limitations. According to Figure 4.13, the Voigt approximation is very close to the

elastic properties along direction 1, but fails to predict the material behavior in the other

two orthogonal directions. The Reuss approximation is based on the assumption that the

strain tensor remains constant, but differs significantly from the approximation in

90
 Numerical Homogenization of Superconducting Composite Wire for Material Properties

direction 2 and direction 3 of the composite. The average of the Voigt and Reuss

approximations is proposed by Hill, which is close to the elastic property values but still

fails to predict its elastic properties. The elastic modulus of a 36 filament composite MgB2

wire [255] is compared with experimental values obtained by an indentation test and

presented in table 2.

4.0 4.0

Direction 1 Direction 1
3.5 3.5
Direction 2 Direction 2
3.0 Direction 3 3.0
Direction 3
Thermal Strain (mm/m)

Thermal Strain (mm/m)

Direction 1 (analytical) Direction 1 (analytical)
2.5 Direction 2 & 3 (analytical) 2.5 Direction 2 & 3 (analytical)

2.0 2.0

1.5 1.5

1.0 1.0

0.5 0.5

(a) Cir#518 (b) Cir#600

0.0 0.0
0 50 100 150 200 250 300 0 50 100 150 200 250 300
Temperature (K) Temperature (K)
4.0

Direction 1 3.5 Direction 1

3.5
Direction 2 Direction 2
Direction 3 3.0 Direction 3
3.0
Thermal Strain (mm/m)

Thermal Strain (mm/m)

Direction 1 (analytical) Direction 1 (analytical)

2.5 Direction 2 & 3 (analytical) 2.5 Direction 2 & 3 (analytical)

2.0 2.0

1.5 1.5

1.0 1.0

0.5 0.5

0.0 (c) Cir#632 (d) Rec#1027

0.0
0 50 100 150 200 250 300 0 50 100 150 200 250 300
Temperature (K) Temperature (K)

Figure 4.14. Comparison of analytical and numerical homogenization schemes for

thermal strains.
The analytical and numerical thermal strain values (Figure 4.14) show agreement

along direction 1 for all the different wire types. However, the difference between the

analytical and numerical homogenization of the thermal strain values along directions 2

and 3 for the circular wires significantly increases as the number of MgB2 filaments

91
 Chapter 4

decrease along directions 2 and 3. The numerically homogenized thermal conductivities

of the composites are also compared with analytical calculations. Although the analytical

approximation successfully predicts the thermal conductivity along direction 1, the

Thermal Conductivity (W/m-K)

Thermal Conductivity (W/m-K)

100 100

Direction 1 Direction 1
Direction 2 Direction 2
Direction 3 Direction 3
10 Direction 1 (analytical) 10 Direction 1 (analytical)
Direction 2 (analytical) Direction 2 (analytical)
Direction 3 (analytical) Direction 3 (analytical)

1
1

(a) Cir#518 (b) Cir#600

0 50 100 150 200 250 300 0 50 100 150 200 250 300
Temperature (K) Temperature (K)
Thermal Conductivity (W/m-K)

Thermal Conductivity (W/m-k)

100 100

Direction 1 Direction 1
Direction 2 Direction 2
Direction 3 Direction 3
10 Direction 1 (analytical) Direction 1 (analytical)
10
Direction 2 (analytical) Direction 2 (analytical)
Direction 3 (analytical) Direction 3 (analytical)

1
1
(c) Cir#632 (d) Rec#1027

0 50 100 150 200 250 300 0 50 100 150 200 250 300
Temperature (K) Temperature (K)

Figure 4.15 Comparison of thermal conductivity between analytical and numerical

homogenization schemes for different wires.

thermal conductivity along directions 2 and 3 are slightly different due to the simplified

thermal conductivity equation. The thermal conductivity of the components other than

epoxy has been disregarded in this simplified equation because of the very low thermal

conductivity of epoxy.

The experimental specific heat of a composite superconducting MgB2 wire [256]

is also compared with the numerically homogenized values. The composite wire has 18
92
 Numerical Homogenization of Superconducting Composite Wire for Material Properties

filaments of MgB2, and its specific heat was measured at different temperatures ranging

from 2-250 K. The measured results are compared with the numerically homogenized

results with the acceptable agreement and are presented Figure 4.12. This agreement
400

350

300
Specific Heat (J/Kg-K)

250

200

150

100

50 FEA estimation
Measured
0
0 50 100 150 200 250
Temperature (K)

Figure 4.16. Comparison of specific heat between analytical and numerical

homogenization schemes for an MgB2 superconducting wire.

demonstrates that numerical homogenization techniques could be used to estimate the

specific heat of composite MgB2 wires.

4.7 Conclusion

The properties of the superconducting wire composites must be properly

understood before validating the viability of an optimized superconducting magnet

design. Since the material properties are temperature dependent, and the wire is a

complex composite structure, a high fidelity computational model of the superconducting

93
Chapter 4

wire and the magnet should be constructed. This concept has been applied, for the first

time, to four different composite superconducting MgB2 wires by constructing 3D FEA

Figure 4.17. Moving across multiple length scales using the numerically

homogenized material properties of the composite superconducting wires.

models of the microstructure and analyzing the results to extract the homogenized

material properties. The computational homogenization using FEA shows more accurate

estimates of the material properties as compared to the analytical techniques discussed

in the paper because FEA precisely incorporates the geometric complexities of the wire

composite. The analysis also enables the consideration of the orthotropic material

behavior of the composite wire and the need for an FEA approach to accurately estimate

orthotropic elasticity. Also, the cross-sectional shape of the wire could impact the final

design. The shear modulus for rectangular wires differs significantly compared to the

circular wires. Homogenization of elastic property values obtained from FEA analysis is

94
 Numerical Homogenization of Superconducting Composite Wire for Material Properties

compared to experimental MgB2 composite wire and show good agreement.

Computational homogenization results for the rectangular wire (wire#1027) have been

compared with the experimental results of the thermal conductivity and specific heat

values and agree within a reasonable limit. This agreement proves the feasibility of

numerical homogenization as a substitute for the expensive and time-consuming

experimentation necessary to model computationally the composite wire for numerical

studies of superconducting MRI magnet design as shown in Figure 4.17. Furthermore, the

computational effort could be significantly reduced without substantial sacrifice in the

accuracy if homogenization techniques are implemented. Once the homogenization of

the composite wire provides the necessary material properties, they can be used to model

the multiscale magnet analysis as shown in Figure 4.17.

95
 Chapter 5. Multiphysics Modeling of Magnet Manufacturing

“Experimental confirmation of a prediction is merely a measurement. An experiment

disproving a prediction is a discovery.” — Enrico Fermi

5.1 Full body MRI Background Magnet

The full body MRI background magnet consists a set of magnet bundles. The

position of the magnet bundles is optimized for highly homogeneous magnetic field down

to ~6 ppm. Figure 5.1 shows the optimized location of the bundles for a 1.5 T full body

MRI magnet system. However, for magnet bundles built out of strain sensitive materials

such as NbTi and MgB2, it is crucial to monitor the strain development. While multiscale

and multiphysics FE analysis of NbTi superconducting magnets has been carried out by

researchers [61]–[67], mechanical strain analysis of a conduction cooled MgB2 based

superconducting solenoid magnet is yet to be explored. The purpose of the work in this

chapter is to develop a multiscale and multiphysics model to predict the final stress state

in 1.5 T MgB2 MRI coils following the production, cooling and magnetization stages. The

study interacts with different physics field and thus is a multiphysics problem as it

accounts for mechanically, thermally and magnetically induced strains. It is also a

multiscale problem because it executes a homogenization scheme to connect the wire

length scale studies to the coil length scale studies.

96
 Multiphysics Modeling of Magnet Manufacturing

Figure 5.1. 2D Axisymmetric (left) view of the coil half and isometric view (right) of

the coil bundles assembly.

In this chapter, modeling of the strain distribution in MgB2 coils by using (FEA) is

sought. FEA gives an advantage over analytical solutions for studying strain in coils

because analytical studies are limited regarding handling complex geometric shapes,

imposed boundary conditions and in addressing the variation of elastic moduli between

mandrel, epoxy, and wire. Moreover, analytical solutions are limited to isotropic material

properties whereas an orthotropic material property approximation for composite wire

would be more appropriate. As a result, current magnet designers rely mostly on

computer programs based on finite element analysis (FEA) compared to analytical

approaches such as filament winding of pressure vessels, [67], [165] thermal cooling [165]

and electromagnetic charging [170]. Therefore, in this study we have calculated the total

(accumulated) mechanical strain (T) resulting from winding (w), cool-down (th), and

electromagnetic charging (e) experienced by the MgB2 coil system.

97
Chapter 5

5.2 Details of the full Body MRI Background Magnet

A 1.5 T magnetically shielded, full-body system consisting of ten coils and an

engineering current density of 118.5A/mm2 (Figure 5.1). The magnet system has four

excitation coil bundles that generate the desired magnetic field. The largest coil bundle

(bundle 5) is the shield bundle. This bundle generates an opposing magnetic field that

Table 5.1. Design specification of 1.5 T conduction cooled MgB2 superconducting magnet

Coil Inner Radius Outer Radius Starting Axial Ending Axial Location

(m) (m) position (m) (m)

Bundle 1 0.5001 0.5321 0.0496 0.1167

Bundle 2 0.5005 0.5327 0.2273 0.3181

Bundle 3 0.5000 0.5321 0.4639 0.5764

Bundle 4 0.5000 0.5321 0.6595 0.8699

Bundle 5 0.9059 0.9309 0.6122 0.8063

Current Density, J = 118.5A/mm2

prevents the magnetic field from coming out of the 5 gauss line1. Since the full body MRI

magnet system is symmetric about the mid-plane, the total geometry reduces to quarter

symmetry, which saves computational effort. The dimensions of the coil windings for the

1
The line is the boundary within which any ferromagnetic materials or objects are strictly prohibited.

98
 Multiphysics Modeling of Magnet Manufacturing

entire magnet system are provided in Table 5.1. The location of the coil bundles is

optimized based on the methods described elsewhere [20].

5.3 Overview of the Magnet Manufacturing and Energization

The manufacturing and operation of an MRI magnet system undergo three

different stages — the winding of superconducting wire around a mandrel, the thermal

cool-down of the entire system to the operating temperature of 10 K and the energization

of the system electromagnetically by passing the design current through the wire.

 t e w+th+e = T

Figure 5.2. Manufacturing process of a superconducting solenoid.

In this work, a 1.5 T MgB2 based 36 filaments superconducting wire is used to

manufacture each coil bundle by winding the wires around a non-ferromagnetic stainless

steel mandrel. During the winding process, typically a 7 to 15 lb pretension is applied on

the wire to ensure adequate packing. 31.14 MPa (15 lb) of pretension is applied on the

wire at the time of the winding process to model maximum strain development in the

MRI system.

After the winding process, the magnet bundles are assembled into the vacuum

vessel, and the copper straps connected to the bundle for conduction cooling are linked

to the second stage of the cryo-cooler. The entire assembly is thermally shielded inside

99
Chapter 5

an aluminum thermal shield. Next, the evacuated space containing the mechanically

integrated coils and insulation are subjected via cryocoolers to a uniform thermal

gradient. A uniform temperature difference of 288 K is applied to model the cool-down

of the entire system from room temperature of 298 K to the operating temperature of

the coils of 10 K. Once, the system is cooled down to the operating temperature, the

magnets are energized with an operating current of 255.87 A, a current value that

corresponds to an engineering current density of 118.5 A/mm2 to generate a magnetic

field of 1.5 T in the DSV. The process affecting the MgB2 wire performance starting with

winding and ending with electromagnetic charging is summarized in Figure 5.2. As

mentioned before, a 36 filament MgB2 based superconducting wire is considered for the

MRI system. As the modeling of an entire bundle considering each wire constituent

individually would be computationally intensive, the wire is first analyzed and numerically

Figure 5.3. Wire cross section of 36 filament MgB2 based superconducting wire.

Microscopic picture on left and computer aided designed (CAD) model on right. The

orthogonal directions 1, 2, and 3 are shown in the figure.

homogenized as an orthotropic material. This wire length scale considers the wire

constituents, MgB2, Niobium, Copper, Monel and Epoxy as isotropic. The homogenized

wire material properties then substitute for the material properties of the coil bundle

100
 Multiphysics Modeling of Magnet Manufacturing

Table 5.2 Material property of wire constituents and mandrel

Average Thermal

Modulus of Expansion
Poisson’s ratio (ν) at
Material Elasticity (GPa) at Coefficient
298 K
298 K (10K~298K)

μm/m.K

MgB2 273 [221] 0.181 [221] 4.23 [259]

Copper 129.5 [225] 0.355 [225] 10.9 [225]

Niobium 103 [227] 0.4 [227] 9.28 [260]

Monel 179 [228] 0.315 [228] 12.5 [228]

19.83(Through

Thickness), 6.23
Epoxy 19.7 [226] 0.355 [224]
(Warp or Fill)

[226]

302 Stainless steel 190 [224] 0.305 [224] 12 [224]

considering complete orthotropy, which is the coil bundle length scale. The coil bundles

are solved for all three processes— winding, cool-down and electromagnetic excitation.

There are well verified existing analytical approaches [62], [63], [65]–[67] to model

winding, thermal cool-down [165] and electromagnetic charging [60], [64], [166], [170],

[257], [258], but the computational approach based on FEA is well favored considering

101
Chapter 5

Table 5.3 Summary of the homogenized material property of the composite wire

Homogenized 36
Material Property (Directions are shown in Figure 2)
Filament Wire

Modulus of Elasticity (Direction 1) 153 GPa

Modulus of Elasticity (Direction 2) 123 GPa

Modulus of Elasticity (Direction 3) 119 GPa

Shear Modulus (Direction 1-2) 43.9 GPa

Shear Modulus (Direction 2-3) 36.1 GPa

Shear Modulus (Direction 3-1) 40.9 GPa

Poisson’s ratio (Direction 1-2) 0.284

Poisson’s ratio (Direction 2-3) 0.332

Poisson’s ratio (Direction 3-1) 0.291

Average thermal expansion coefficient (10K~298K) (α23) 10.71 μm/m-K

Average thermal expansion coefficient (α1) 8.62 μm/m-K

orthotropic material approximation, boundary conditions, geometric complexity, and

multivariate elastic moduli of wire and mandrel. Hence, while employing FEA analysis;

shifting from the wire length scale to the coil bundle length scale makes the problem

multiscale and modeling the coil bundle for thermal cool-down, and electromagnetic

energization for stress-strain turns the model into multiphysics modeling. Hence, a
102
 Multiphysics Modeling of Magnet Manufacturing

multiscale multiphysics model is required for the entire 1.5 T MRI magnet bundle.

5.4 Modeling of Magnet Manufacturing and Operation

The wire considered for the design has dimensions of 1.19 mm by 1.82 mm

including insulation (Figure 5.3). The composite wire has a total of 36 filaments of MgB2

(volume fraction 18%) enclaved by a layer of Niobium (25%) inside a Copper (16%) matrix.

The entire matrix is surrounded by Monel (29%). The entire wire is enveloped by a layer

of epoxy of commercial grade (12%).

Figure 5.4. Comparison of finite element and analytical approach of solving the winding

stress development in coil bundle 5 of the system.

Since the modeling of the individual wire with great detail in a bundle would be

cumbersome and computationally intensive, a numerical homogenization approach as

detailed in section 4.4.1 is employed. The constituent and calculated material properties

103
Chapter 5

are summarized in Table 5.2 and Table 5.3. The homogenized material property values

are used in the analysis of all five bundles in the 1.5 T system.

5.4.1 Winding Stress

The amount of pretension on the wire depends on the magnetic field design, wire

cross-section and strain limitation of wire materials. For a 1.5 T MRI system, a maximum

31.14 MPa pretension is relatively typical for this wire size and will be assumed for

modeling of the bundle winding.

5.4.1.1 Winding Stress Prediction by Finite Element Analysis

ANSYS, a commercial Finite Element Analysis (FEA) software, is used to model the

winding process of a 1.5 T conduction cooled MgB2 based superconducting magnet

system. The basic equation for the finite element analysis is summarized with the most

straightforward equation as,

{σ} = [C]{𝜀 𝑒𝑙 } (5.1)

In the equation, [C] is the elastic stiffness matrix, {εel} represents the elastic strain

vector, and {σ} indicates the stress developed on the modeled geometry. To imitate the

winding process, the element birth and death technique in ANSYS is employed. Quadratic

2D element Plane 183 is chosen for the geometry since the problem is considered 2D

axisymmetric for simplification. Quadratic elements are chosen over linear elements

(Plane 182) because they better approximate the stress at the mandrel-bundle interface

with their shape functions. The initial state condition is used to apply the prestrain that

corresponds to the pretension of 31.14 MPa on the designated wire. After modeling the

104
 Multiphysics Modeling of Magnet Manufacturing

entire geometry, a uniform prestrain is applied to the coil bundle, and all corresponding

layer elements are deactivated by setting the stiffness to an ANSYS defined low value of

10-6. Then, to imitate the winding process, each layer is activated by returning the stiffness

values to the assigned material property values. To compare the agreement between the

analytical and FEA approaches, the outermost coil bundle (Bundle 5 in Figure 5.1) is

modeled and compared for both the analytical and FEA results. It is clear from Figure 5.4

that the analytical CHCM model (section 5.4.1) is incapable of precisely calculating the

stress as the value of the elastic modulus changes due to the change from stainless steel

mandrel to coil bundle. This is because the CHCM equations do not consider the modulus

of elasticity of different materials. On the other hand, the hoop stress values are within

~3% in the coil bundles if the mandrel and coil are made of the same material. Hence, the

analytical CHCM method can be used to validate the finite element analysis result, but

105
Chapter 5

cannot address the change in material stiffness values, which limits the use of CHCM in a

mandrel-bundle winding modeling system.

Figure 5.5. Comparison of finite element vs analytical solution of cool-down (298

K to 10 K) of a solenoid considered by Arp.

5.4.2 Thermal Cool-down Stress

Once the construction and assembly of the magnet system is complete, the next

step is to cool the magnet from room temperature to the cryogenic temperature of 10 K

that is required for MgB2 to act as a superconductor with optimum performance. This

cooling down of the entire system creates stresses because the thermal strain behavior

of the mandrel and coil bundles is different and each is highly non-linear.

106
 Multiphysics Modeling of Magnet Manufacturing

5.4.2.1 Thermal Stress Prediction by Finite Element Analysis

To validate the finite element analysis method with the analytical equations, the

same superconducting solenoid considered by Arp [165] was solved and compared. In

ANSYS, quadratic plane element Plane 183 is used, and the linear thermal expansion

coefficient is assigned in both the radial and circumferential (hoop) directions with the

reference temperature set to T0 (298 K). The temperature of the magnet is assumed to be

spatially uniform during the entire cool down process from T0 to Tf. The FEA equation

now has the extra thermal strain term due to temperature change as shown in following

equation

{σ} = [C]({𝜀} − {εth }) (5.2)

The thermal strain {εth } in the equation is calculated according to equation (3.12)

𝑇 (𝑟)
in integral from as 𝜀 𝑡ℎ = ∫𝑇 𝑓 𝛼𝑑𝑇 and {𝜀} is the total strain vector.
0

5.4.3 Electro-Magnetic Stress

When the superconducting solenoid is energized, the electromagnetic forces are

determined from the radial and axial components of J×B where J and B are the vector

representation of the current density and magnetic field, respectively. These radial and

axial directional forces, when considered as a body force in the stress equation, result in

stresses in the radial and hoop directions. The hoop stress appears to be dominating

compared to the radial stress [166] and hence is of major concern when designing the

superconducting solenoid.

107
 Chapter 5

Figure 5.6. Comparison of hoop stress variation along the radial direction of a solenoid.

The analytical approach as provided by Caldwell is in good agreement with the two

different FEA approach using ANSYS.

5.4.3.1 Finite Element Analysis of Electromagnetic Energization

ANSYS coupled field element plane 13 is used to model the stress due to

electromagnetic charging. ANSYS provides two different methods: sequential load

transfer and direct coupled field analysis. For the former, Maxwell’s equations are solved

first, and then forces at each nodal location are calculated. At the next step, the nodal

forces are transferred to a new analysis, where the stress-strain equations are solved. In

contrast, in the direct coupled analysis, the stress and strains are a direct output of the

coupled electromagnetic and solid mechanics equation. The two different FEA analyses,

direct coupled and sequential coupled field, are employed to model Caldwell’s [170],

108
 Multiphysics Modeling of Magnet Manufacturing

[171] superconducting solenoid and then compared with the analytical solution.

However, ANSYS’s limitation of handling load transfer from previous analyses (winding

and cool-down) for a direct couple field analysis restricts the solution method to

sequential coupling. Figure 5.6 shows the agreement between the direct coupled, the

sequential coupled and the analytical solution. It appears that the implementation of the

sequentially coupled solution introduces a slight deviation from the direct coupled

solution, but the variation is within 5% of the analytical range.

Figure 5.7. Radial Stress along the radial direction of the mid-plane of all five coil

bundles. (Mandrel regions are shaded in gray)

109
Chapter 5

5.4.4 Strain Development in the Magnet Bundles

The radial stress and strain for each bundle are plotted in Figure 5.7 along the

radial direction from the inner surface of the mandrel to the outer surface of the

outermost layer of each coil bundle. Coil bundles 1 through 4 have the same number of

winding layers with almost the same radial positions. Hence, the radial stress and strain

developments in the coil bundle after winding are almost the same. From Figure 5.7 it is

noted that the maximum radial stress occurs within the coil near the interface between

the wires and the mandrel. The stresses are zero at the inner and outer boundary surfaces

as these two boundaries are free to move in the radial direction. This application of

Figure 5.8. Hoop stress and strain plot along the radial direction of five bundles of

the system.

winding stress helps compensate for the tensile stress developed by the magnetic forces

during operation [165] and thus assists in minimizing the developed strains while keeping

them within the failure limit. The pretension also helps to compensate for the transverse

shear stress created by the Lorentz force [261] at the time of magnet operation and

prevents the sliding of the layers relative to one another. This can be confirmed by
110
 Multiphysics Modeling of Magnet Manufacturing

comparing the strain development on the bundle at two stages: after winding and after

electromagnetic charging. Moreover, pretension also helps in reducing the strain in the

epoxy and reduces the possibility of the epoxy cracking [262]. Furthermore, Arp et al.

[165] showed that it is possible to control more precisely the total stress development at

the time of electromagnet charging if the applied pretension is varied on the layers of the

coil. These findings underscore the importance of controlling the pretension on the wire

during the winding process of a superconducting solenoid.

When considering the stress and strain results from the structural analysis of a

superconducting magnet bundle, it is found that the radial stresses are negligible

Figure 5.9. Four different types of stresses and strains on the coil bundle 5 along the

radial direction after the winding of layers around the mandrel is complete.

compared to the hoop (circumferential) stresses. It is clear from Figure 5.8 that the hoop

stress varies in the range of -50 MPa to 30 MPa, whereas the radial stress reaches only a

maximum value of -1 MPa. Hence, previous works have looked into hoop stress

(circumferential stress) [62], [64], [67] and von Mises stress [65], [66] to determine the

111
Chapter 5

failure criteria. However, in the numerical modeling, the wire inside the bundle is a

composite wire made of specific proportions of MgB2, niobium, copper, monel and epoxy.

Hence, the superconducting filaments of MgB2 act as a fiber-matrix composite, which

undergoes tensile or compressive loading as the magnet is built, cooled down and

charged. As summarized by Orifici et al. [263] for composite fibers under tensile loading,

Figure 5.10. 1st and 3rd principal stress ((a), (c)) and strains ((b), (d)) along the radial

distance of the all five coil bundles.

it is usually the maximum strain that predicts the fiber failure. Besides, Vaghar et al. [264]

has included von Mises strains as one of the failure criteria for their Nb 3Sn

superconducting magnets design. So it is essential to consider the principal and von Mises

112
 Multiphysics Modeling of Magnet Manufacturing

stress and strains along with hoop (circumferential) stress and strains of the coil bundle.

Moreover, uniaxial tensile testing of the superconducting wire at cryogenic temperatures

reveals that the wire becomes resistive if the mechanical strain goes below -0.6% or above

0.4% [38], [58], [265], [266]. Therefore, it is also essential to plot the third principal stress

and strain as well to check if the magnet wire experiences mechanical strain below -0.6%

at any point during the magnet operation. Hence, as a means of visualizing the stress and

strain scenario, different types of stresses and strains are plotted for bundle 5 after the

Figure 5.11. 1st and 3rd Principal mechanical stress ((a) and (c)) and strain ((b) and (d))

after the bundles are cooled down to operating temperature of 10K.

winding is complete. From the figure, it is seen that both the tensile and compressive

stress and strain scenarios are better picked up by the 1 st and 3rd principal stresses and

strains in the bundle coordinate system. This is because σ1 > σ2 > σ3 according to the

113
Chapter 5

definition of principal stress; where σ1, σ2, σ3 are the stresses in the three principal

directions respectively. Hence, if the bundles are checked for tensile stresses and strains,

σ1 and ε1 would provide the scenario where the bundle materials are in tensile loading.

Correspondingly, for a compressive loading, the 3rd principal stress and strain would

provide the bundle scenarios. On the other hand, the Von Mises stress and strain are

limited in predicting the failure of a composite fiber-matrix configuration, because

composite structures will individually have different compressive and tensile strengths,

which do not meet the requirement of von Mises equivalent stresses and strains to be

used as failure criteria [267]. Moreover, the von Mises stress and strain never go below

zero and fail to predict the compressive stress and strains. As a result, the first and third

principal stresses and strains are considered as primary failure criteria for the design and

thus plotted to give an overview of the stress-strain scenario of the magnet bundles. The

stress-strain state after the winding is complete is summarized in Figure 5.10. From this

figure, it is noted that the first four bundles having the same inner and outer radii exhibit

similar stress and strain development as to be expected and hence they are superposed.

However, since the fifth bundle has a different radial thickness and different radial

location, it has slightly different stress and strain development. The bundles experience

the maximum tensile stress of 30 MPa at the outermost layers. In contrast, the 3rd

principal stresses are peaked near the inner surface of the mandrel (-50 MPa). The peak

values for 1st and 3rd principal strains are 0.019% and - 0.025%, respectively.

The stress-strain scenario changes as the system is cooled down to an operating

temperature of 10 K. Since, the mandrel is made of stainless steel, with a higher thermal

114
 Multiphysics Modeling of Magnet Manufacturing

expansion coefficient than the composite wire material, the mechanical strains determine

the failure criteria [268]. Therefore, total mechanical strains (w+th) are to be checked for

the failure criteria. Hence, total mechanical strains are plotted in Figure 5.11 to visualize

the strain development in the entire system. The thermal expansion coefficient of

stainless steel is comparatively higher than the composite wire; thus, the mandrel tends

to shrink more relative to the wires as the temperature drops. As the temperature is

dropping, the layers adjacent to the mandrel are pulled by the mandrel on one side, but

restrained by the upper layers on the other side, which would generate a tensile loading

Figure 5.12. 1st and 3rd principal mechanical stresses ((a) and (c)) and strains ((b) and

(d)) after the coil bundles are charged with the operating current.

115
Chapter 5

along the radial direction. This occurrence of the tensile load is responsible for the sudden

change in the 1st and 3rd principal strains in that region. The 1st and 3rd principal

mechanical stress and strains are also plotted in Figure 5.11.

Table 5.4 Summary of numerically homogenized material property of the

composite Rec#1027 wire at 300 K

Material Property (Directions are shown in Figure 4.4) Homogenized wire

Modulus of Elasticity (Direction 1) 112 GPa
Modulus of Elasticity (Direction 2) 57.9 GPa
Modulus of Elasticity (Direction 3) 59.6 GPa
Shear Modulus (Plane 1-2) 17.5 GPa
Shear Modulus (Plane 2-3) 13.4 GPa
Shear Modulus (Plane 3-1) 18 GPa
Poisson’s ratio (Plane 1-2) 0.26
Poisson’s ratio (Plane 2-3) 0.288
Poisson’s ratio (Plane 3-1) 0.255
Avg. thermal expansion coefficient (direction. α1) 10.1 μm/m-K
Avg. thermal expansion coefficient (direction α2) 12.9 μm/m-K
Avg. thermal expansion coefficient (direction α3) 12.6 μm/m-K

From the figure, the maximum tensile stress of ~87 MPa is observed near the

mandrel outer surface, and maximum compressive stress of ~52 MPa is observed near

the innermost layers. The strains also peak near ~0.039% at the mandrel inner surface,

whereas the compressive strain peaks at 0.026% near the first layer of the bundle. It is

interesting to note that the strain curves for the first four bundle now have slight

116
 Multiphysics Modeling of Magnet Manufacturing

deviations from each other caused by their different axial lengths due to the effect of

added strain from thermal contraction.

At the time of operation, the Lorentz forces act on the coil bundle. This force is

directed primarily radially outward on the bundles and generates tensile stress in the

bundle layers. From the winding and thermal cool down process, there are residual

compressive stresses already present in the mandrel and coil layers. The tensile stress

generated counteracts the residual compressive stress and explains the importance of

Figure 5.13. 1st principal mechanical strain in all five bundles of the coil at the time of

electromagnetic charging after wire winding and cooling down to operating

temperature.

applying pretension in the wire at the time of the winding process. A careful design of

applied pretension allows more room for electromagnetic stress development before the

117
Chapter 5

wire bundle crosses the design limit of strains as previously described. From the principal

stress and strain plot (Figure 5.12), it is observed that the maximum tensile stress

develops in the mandrel outer radius, whereas the stress in the coil bundle peaks at ~25

MPa. As far as the strain development is concerned, the maximum tensile strains of

0.016% occur in the innermost layer of the bundle. Some other researchers [38], [58],

[59], [265], [266] have shown that the critical current carrying capacity of composite wire

reduces as the strain is being developed. So the strain development in the wire bundle as

design criteria must stay below the limit of 0.2% (half of the failure criteria) in the tensile

direction and -0.3% in the compressive direction at the time of electromagnetic operation

[38], [57], [58], [266]. From the analysis of all five coil bundles, as principal strains are

considered according to maximum strain criteria [263], the entire system develops a

maximum mechanical strain of 0.016%, which is well below the design criterion of 0.2%.

Also, the 3rd principal mechanical strain development is in the range of -0.035%,

which is well below the -0.3% failure criterion. The entire coil bundle system is shown in

Figure 5.13, a ¾th expansion of the axisymmetric bundle and total 1st principal mechanical

strain is plotted. From the figure, the maximum mechanical strain of 0.06% is observed at

the outermost axial location (direction 2 of wire) in coil bundle 4 which is also within the

design criteria (-0.3% to 0.2%). Thus the analysis indicates the safe operation of the

magnet for a design field of 1.5 T.

118
 Multiphysics Modeling of Magnet Manufacturing

5.4.5 Effect of Support Conditions

Arp [165] has shown with analytical equations that the support condition at the

time of winding process may affect the final stress-strain state of the magnet bundle

system. The composite wire used for the study is Rec#1027 as analyzed in section 4.4. The

homogenization of the composite wire is done using the methods detailed and discussed

in section 4.4. The summary of the wire material properties at 300 K is presented in the

table.

Figure 5.14. Locations of all five coil bundles of the symmetric system and four different

support conditions for winding.

At the time of winding— four different types of support on the mandrel can be

used. Different types of support conditions are shown with schematics in Figure 5.14. No

mandrel support during winding allows the mandrel to deform inwardly due to the

applied pretension on the wire (Case I). Only the radial support at the inner radial location

of the mandrel restricts all inward deformation of the mandrel (Case II: radial

119
Chapter 5

displacement, ur = 0). The combination of radial support and axial support restricts the

deformation of the mandrel in both the radial and axial directions (Case III: radial

Figure 5.15. 1st principal strain on all five bundles of the 1.5 T magnet system. Mandrel

regions are shaded in gray. The plot is at the mid-plane from the mandrel inner surface

to the outermost radial location of the bundles.

displacement, ur = axial displacement, uz = 0). Only the axial support at the two extreme

axial location of the mandrel restricts any axial deformation of the mandrel but allows

any deformation inward in the radial location (Case IV: axial displacement, uz = 0). After

the winding is complete, all the supports are removed, and the coil is thermally cooled

down to 10 K and electromagnetically charged to a 1.5 T magnetic field.

The entire magnet system is solved utilizing a multiscale multiphysics FEA method

as detailed in a prior article [158] and elaborated in section 5.4. The principal strains

represent the composite failure criteria more accurately [263] rather than the von Mises

120
 Multiphysics Modeling of Magnet Manufacturing

criteria and should be the primary concern. Also, the composite MgB2 superconducting

wire fails at 0.4% tensile strain under uniaxial tensile loading as demonstrated

experimentally [38], [57], [58], [266]. Hence, Figure 5.15 shows the 1st principal strain

development in all of the five bundles after the winding is complete. There was no

mandrel support in the radial direction for this case. From this figure, maximum strain

Figure 5.16. Tangential strain development on bundle 5 at the end of each step.

development is observed in bundle 5 (shield coil). Therefore, considering only bundle 5

would be enough to picture the overall scenario of maximum strain development. The

coil bundles undergo different processes— winding, cool-down and electromagnetic

charging (excitation) and accumulate strain throughout these processes. Figure 5.16

shows tangential (hoop) strain development on bundle 5. The tangential strain is plotted

in this case because this strain helps to understand the strain change with ease across

121
Chapter 5

different processes: winding, cool-down and electromagnetic charging. Form the figure;

it is observed that the bundles are under tensile strain. As the system is cooled down to

10 K operating temperature, the strain drops and part of the bundle is now under

compressive strain. After excitation, the strain in the bundle increases and becomes

tensile but stays well below engineering design strain limit of 0.2% (half of the 0.4%

irreversible strain).

Figure 5.17. Maximum shear stress in bundle 5 at different steps of manufacturing and

operations.
During winding, if the floating coil technique is employed, the mandrel and coil

bundle stay free of each other. Implementation of this winding technology facilitates the

free movement of the bundle when the magnet is excited. However, the conduction-

cooled magnet would experience degradation in heat transfer and affect the cooling

122
 Multiphysics Modeling of Magnet Manufacturing

system attributed to decreased contact conductance if floating coil technique is

Figure 5.18. 1st principals strain due to different winding support cases after the

magnets are charged to 1.5 T field

employed. Hence, the mandrel and bundles are attached to each other in this case.

Maximum shear stress along the mid-plane from inner to outer radial locations is plotted

in Figure 5.17 to understand the failure in the epoxy. A maximum shear stress of 28 MPa

is observed at the outer surface of the bundle. The shear strength of epoxy at 4 K is 232

MPa [269]. The developed shear stress is well within the range of the failure strength of

an epoxy and hence indicates a safe operation of the magnet system.

5.4.5.1 Stress and strains for different support conditions

The effect of using different support condition at the time winding is better

imaged at the final electromagnetic energization stage. Four different boundary

123
Chapter 5

conditions as discussed in the previous section 5.4.5 has been employed during the

winding process and subsequently removed before the cool-down and electromagnetic

charging. The boundary conditions are described previously in the section ‘Support

conditions, ’ and the 1st principal strain development after electromagnetic charging is

plotted in Figure 5.18. In case of no radial support on the mandrel, the mandrel and

Figure 5.19. Hoop strain on bundle 5 with case II

bundle both are allowed to deform inwardly due to the applied compressive stress by the

pretension (31.14 MPa) [158] on the wire. This inward deformation assists in stress

relaxation in the tangential direction and helps reduce the tensile stress in the lower

layers at the time of winding as shown in Figure 5.19. The tangential stresses are the most

dominant stress components and constitutes the majority of the 1st principal strain.

Therefore, the 1st principal strain closely follows the trend of tangential strain. For case

IV, the mandrel is supported in the axial direction. Hence, the 1st principal strain is similar

124
 Multiphysics Modeling of Magnet Manufacturing

to case I.

However, when case II and case III are employed, the deformation of the mandrel

is restricted in the inward radial direction. This restriction allows the tangential stress to

stay almost constant through the bundle along the radial direction. During cool-down, the

stresses reduce as the mandrel shrinks more in comparison to the bundle. This shrinkage

allows extra space for the bundle to relax as the strain developed. As a result, from Figure

5.19, it is observed that the strain values drop after the cool-down process. When the

bundles are charged with the excitation current, they expand radially outward due to the

Lorentz force. Again, as the inner surface of the mandrel is supported with constraints,

most stress development occurs in the mandrel. As a result, higher strain development is

observed in the mandrel while strain in the bundle is comparatively smaller. Similarly,

when case III is used, strain development is maximum in the mandrel while strains in the

bundle stay low with a decreasing trend.

From Figure 5.18 when different boundary conditions are compared for the strain

development, it appears that supporting the mandrel with radial support with axial

support (Case III) is the most beneficial in terms of keeping the strains low. It is observed

that not only the strain development is low but also the variation of the strain is narrow

which is around 0.001% with a maximum strain occurring at the bundle-mandrel interface

of 0.0207% to the outer most radial location of 0.0197%. This uniform strain development

introduces a stabilization factor to the design that would help during quench induced

strain due to the hot spot.

125
Chapter 5

Along with the strain development in the bundle, epoxy cracking is another major

concern to design the magnet [270]. To check the integrity of the magnet bundle,

maximum shear stress is also plotted for different support conditions. Again from Figure

5.18, it is clear that the shear stress is smaller in the bundle (~13 MPa) when support

condition II or III is employed. This is about 12 MPa lower than support condition I and IV.

Hence, it is suggested that the use of radial support along with axial support during wire

winding minimizes the strain and shear stress development.

5.5 Global-Local Analysis of Composite Wire

The calculated directional strain results when transferred to the RVE, estimates

the stress-strain scenario of the composite. When utilized the technique, it is possible to

accurately predict the stress and strain states in the components of the composite [108].

Figure 5.20. Localization of the 1st Principal strain and Maximum Shear stress.

The methodologies used are detailed in section 4.4. Making use of the same set of

equation, it is possible to apply six independent strain states simultaneously on the RVE.

126
 Multiphysics Modeling of Magnet Manufacturing

As the numerically homogenized RVE helped in reducing the computational efforts, the

model is solved for the directional strain states for a specific location. After the solution

yields specific values of 𝜀𝑖𝑗 , the values are input to the same numerical homogenization

script to calculate the 1st principal strain in the MgB2 and the maximum shear stress in the

epoxy. Figure 5.20 shows the localization strain and stress states in the Cir#518 wire. From

the figure, it is noted that the strain values are still smaller thant he 0.2% strain limit and

the stress values are also smaller than the design limit of 100 MPa.

5.6 Effect of Creep

The manufactured full body MRI magnet system experiences shear stress about

50 MPa [271], at the time of electromagnetic operation. This amount of constant loading

can be a concern of creep failure in the epoxy used. However, there has been limited

study regarding the creep failure of the S-glass epoxy [272]. The reported study was at

elevated temperature of 750˚ C. Also, materials usually demonstrate higher resilience to

creep failure at lower temperature [273]. Moreover, fiber-reinforced polymer epoxy are

stronger in terms of creep failure even at room temperature. They can sustain creep test

load of up to 77% of the ultimate tensile strength even at room temperature [274]. The

possibility of failure due to creep is much lower at cryogenic temperature. The CTD 101k

considered for the computational study of the full body magnet has compression strength

in the range of 1300 MPa [226]. This indicates that the possibility of epoxy failure due to

creep loading is very slim and reasonable to ignore for an MRI magnet study.

127
Chapter 5

5.7 Conclusion

This chapter considers a full body conduction cooled MgB2 superconducting

magnet system designed specifically for a field strength of 1.5 T. The system is modeled

at three stages of manufacturing and operation: the winding of coil bundles in layers

around a mandrel, the cool-down of the magnets, and ultimately, the electromagnetic

charging. The constituent wire and support matrix is considered as a composite of MgB2

made up of niobium, monel, copper, and epoxy. A combination of analytical

approximations and Finite Element Analysis are used to verify the calculation of the stress

and strain in the coil and mandrel, and the analysis is simplified by reducing the geometry

to 2D axisymmetric at the mid-plane of the entire system.

After winding, cool-down and electromagnetic charging, the strain state of the coil

bundles provides a clear image of the strain development of the entire system. It is

determined that the failure criteria is the principal mechanical strains developed in the

magnet bundle. The failure is initiated when an irreversible degradation occurs in the

wire, and the wire becomes resistive due to the strain exceeding -0.6% or 0.4% at the time

of electromagnetic operation. For design purposes, considering a safety margin, a limit of

-0.3% to 0.2% is imposed. From the Finite Element Analysis using ANSYS APDL, the

maximum principal mechanical strain is estimated at ~0.06% at the outermost axial

location in coil bundle 4. All calculated strains are within the safety limit of -0.3% to 0.2%

strain for MgB2. Hence, it is expected that the entire magnet system is safe to operate

concerning strain development at 1.5 T. Finally, further studies will investigate the

mechanical integrity of the magnet during a quench.

128
 Multiphysics Modeling of Magnet Manufacturing

Additionally, the effect of support conditions for the winding process on the final stress-

strain state is studied. Four different types of support conditions are investigated using

the multiscale multiphysics FEA method to study the effect of strain and shear stress

development on the coil bundles. Studies have suggested the use of the mandrel radial

support along with axial support on the axial locations of the mandrel is beneficial to

maintain a low strain and shear stress on the magnet bundles. These results are validated

for a magnet system with a stainless steel mandrel and 18 filaments MgB2 composite

superconducting wire.

129
 Chapter 6. Multiphysics Modeling of Magnet Quench

“Everything should be made as simple as possible, but not simpler” — Albert Einstein

6.1 Modeling of Magnet Quench

Quench is the sudden rise in the magnet bundle due to the transition from

superconducting to the resistive mode of magnet operation. It is essential to model the

quench of a superconducting magnet as means to predict the temperature rise. The

modeling of temperature rise also helps to design the protection electronics circuit

system for the magnet. The protection system prevents the sudden temperature rise and

averts localized heating to save the magnet from destruction due to the strain generation.

6.2 Modeling of Transition

There are two available models for superconducting coil quench modeling at the

transition region. As explained by Ristic et al. [220], they are

i. No current sharing between superconducting fiber and matrix

ii. Current sharing between superconducting fiber and matrix

In the no current sharing model, the resistance of the superconducting fiber at the

transition region is at least an order of magnitude larger than the matrix of the composite

wire. The matrix of the composite wire is usually copper. In that case, since the resistivity

is lower for copper, the current transition from superconducting fiber to the copper

130
 Multiphysics Modeling of Magnet Quench

matrix occurs sharply, and virtually no heat is generated in the superconducting material.

In this case, the ‘n’ value of the superconducting material is usually large (n>30).

In the current sharing model, the resistivity of the superconducting fibers is

comparable to the copper matrix. Thus, at the transition state, current is shared between

the superconducting material and the matrix material. Hence, there is heat generation in

both the superconducting fiber and matrix. Typically, the HTS have a comparable

resistivity in the transition range, and current sharing model is more appropriate for these

types of superconducting coils. However, the transition region is very short and has an

insignificant effect on the heating of the magnet bundle. Therefore, the “no current

sharing” between superconducting fiber and the copper matrix is a valid assumption for

the modeling of a quench.

Figure 6.1. Unit cell for ANSYS simulations to calculate stresses and strains.

6.3 Simplification of Composite Wire

The numerically homogenized model considers the composite wire as a new RVE,

and the strain in the MgB2 is challenging to estimate. Thus to address this issue, the

131
Chapter 6

homogenized composite wire is replaced with a new representative unit cell (RUC) where

all multiple filaments of MgB2 is replaced with a single filament surrounded by the copper

matrix as shown in the figure. Finally, at the outer most location, the epoxy encloses the

RVE. In this RUC, the volume fraction of the epoxy and MgB2 is kept constant, whereas,

nobium, monel, and copper constitute the homogenized region.

Figure 6.2. Material properties used in ANSYS simulations: a) Thermal strain; b)

Elastic modulus.

6.4 ANSYS Simulation of Stress and Strain Development

The temperature in the coils is calculated using a lab-grown finite difference

Matlab code [156], [157]. The temperature of the geometry is then imported into ANSYS

to calculate the strain in the MgB2 and the stress in the epoxy. In the ANSYS simulations,

the geometry of the wire was simplified as shown in Figure 6.1. Although simplified, this

geometry can be used to calculate both the strain developed on the MgB 2 filaments and

the stress in the epoxy. In the FD code, it was assumed that the temperature in the

Monel/Cu/Nb/MgB2 conductor was uniform in the radial and axial directions: All

132
 Multiphysics Modeling of Magnet Quench

temperature gradients occur in the insulation or along the wire in the azimuthal direction.

For input to the ANSYS calculations, the temperature in the epoxy was determined by a

linear interpolation of the conductors on either side of the insulation.

The thermal strains and modulus of elasticity for the materials in the ANSYS

calculations are shown in Figure 6.2. Figure 6.2 (a) shows the thermal strains (referenced

to 10 K) for the Monel/Cu/Nb composite, the MgB2, and the epoxy insulation, where the

thermal strain is given by (3.12) In the simulations the materials are assumed isotropic

except for the thermal expansion coefficient of the epoxy. For the epoxy there are two

thermal expansion coefficients values: one in the warp or fill direction, which is in a

direction along the wire (𝜃 direction); and the other in the through thickness direction,

which is perpendicular to the wire (𝑟 and 𝑧 directions). The MgB2, is assumed isotropic

and the Monel/Cu/Nb composite was homogenized using the Reuss method [108], [275],

where the stress is assumed uniform in the composite. Figure 6.2 (b) shows the modulus

of elasticity for the Monel/Cu/Nb composite, the MgB2, and the epoxy insulation. The

Poisson’s ratios used are 0.33 for the Monel/Cu/Nb composite, 0.3 for the epoxy, and

0.18 for the MgB2 [158].

133
Chapter 6

Figure 6.3. Strains and stresses calculated in ANSYS. a) Tensile strain in the MgB2

superconductor; b) Shear stress (r-θ component) in the epoxy insulation.

6.5 Results

The maximum temperatures in all the simulations were below 200 K. Therefore,

as noted previously, we would expect that the mechanical strains are within safe limits

during the quench [159]. The temperature data from the quench simulations was used as

an input into an ANSYS model of the magnet to verify the safe mechanical strain

development. The heating of the wire during a quench creates tensile strain in the MgB2

because MgB2 has a thermal expansion coefficient that is smaller than the other materials

in the wire. The maximum strain from the thermal expansion occurs at the location of the

hot spot. Also, there is a strain on the wire from prestressing during winding, a strain from

cooldown, and a strain from electromagnetic (EM) energization [158]. These strains are

accounted for in the accumulated strain which is the relevant quantity for comparison to

the strain limit.

134
 Multiphysics Modeling of Magnet Quench

The maximum tensile strain acting on the MgB2 superconductor resulting from prestress

and cooldown, EM charging, and the quench are shown in Figure 6.3. The smallest time

corresponds to the instant the heaters are initiated, and the largest time corresponds to

the temperature reaching a maximum. It is seen that the maximum accumulated tensile

strain acting on the MgB2 is about 0.06%, which is well below the failure criteria of 0.4%

[38], [57], [58], [266] (design criteria of 0.2%). Figure 6.3 (b) shows the maximum shear

stress acting on the epoxy insulation. The r-θ component is shown, since this is the shear

stress component that reaches the largest value. The shear stress is seen to reach about

27 MPa, which is well below the epoxy shear strength of 100 MPa [226] (design criteria

of 50 MPa).

6.6 Conclusion

The calculation of 1st principal strain in the MgB2 and maximum shear stress in

epoxy demonstrates that the 1.5 T magnet system is safe to operate with the protection

system that is discussed by Poole [156] and Deissler [157]. The MgB2 demonstrate

irreversible strain damage beyond the value of 0.4% and the epoxy fails under applied

shear stress values of 100 MPa. The FEA calculation has demonstrated that for a 1.5 T

MgB2, MRI magnet system, a maximum of 0.06% of 1st principal strain is developed at the

mid-plane of the bundle 1. The value is well within the design limit of 0.2%. Additionally,

the maximum shear stress in the epoxy is estimated as 27 MPa. The maximum shear stress

value is also within the safety limit and ensures the temperature rise of 200 K without any

irreversible damage to the wire.

135
 Chapter 7. Conclusion

“The White Rabbit put on his spectacles. ‘Where shall I begin, please your Majesty?’ he

asked. ‘Begin at the beginning,’ the King said gravely, ‘and go on till you come to the end:

then stop.’” - Alice’s Adventures in Wonderland

7.1 Challenge in FEA Modeling

Complete modeling of an MRI magnet system consists of a lot of different aspects.

Stress-strain modeling of the entire magnet system is only a part of that. Even the stress-

strain study of the magnet system is based on successfully modeling material properties,

wire modeling, winding modeling, cool-down modeling, electromagnetic energization

modeling and at the very last, quench modeling. Accurate modeling of the system

requires appropriate compilation of the constituent material properties of the composite

wires at a different temperature. Due to the challenges faced in experimentally testing

and standardizing all the materials at cryogenic temperature, data is limited. Thus in this

instance, material properties have to be interpolated to fill in the blanks. This assumption

provides a way to model the wire and eventually the superconducting magnet system

numerically. However, this does imply some computational error. Nonetheless, the

unavailability of the data has left the research community with no alternate choice when

modeling a superconducting magnet. Therefore, a complete FEA method has been

developed to predict the stress-strain of a conduction cooled 1.5 T MRI magnet keeping

136
 Conclusion

in mind the assumptions. The multiscale multiphysics FEA model of the full body

conduction cooled 1.5 T MRI magnet interacts between the composite wire length scale

and the magnet bundle length scale. This multiscale approach facilitates the modeling of

the entire magnet system using a workstation in a reasonable time frame. The interaction

between different physics phenomenon (elastic-thermal-electro-magnetic) makes it

possible to model each of the steps and stages of magnet manufacturing and operation.

Thus, the model presented in this thesis is capable of predicting the stress and strain of a

full body MgB2 based MRI magnet system. The model is also possible to be used in other

superconducting MRI magnet system when stress and strains are a concern.

(a) (b)

Figure 7.1. Test coil setup. a) Test coil setup with instrumentations b) Cryogenic

cooling system.

7.2 Experimental Comparison

It is always exciting to come into an agreement with experimental values with

computational prediction. However, manufacturing of MRI magnet system is expensive

137
Chapter 7

and requires great care. Therefore, a prototype MgB2 coil bundle has just been

manufactured by Hyper Tech Research and is under test at Ohio State ‘Center for

Superconducting and Magnetic Materials’. The tests are ongoing and experimental data

are yet to be compiled. The test coil and cryo-cooling system is shown in Figure 7.1.

Preliminary test data show slight deviation on the final cryogenic temperature. At the

time of writing this dissertation, the troubleshooting is still under investigation. There

have been several attempts to cool the system to the desired temperature, but the

successful attempt to reach the design temperature has yet to be made. However, once

all the thermal leakage is fixed and vacuum pressure is maintained, the prototype coil will

be ready to provide test data and comparison of results will be possible. It is expected

that the experimental results will deviate to a certain extent as there have been numerous

assumption during coil modeling. Whether or not the assumptions will be replaced with

more accurate modeling depends on the approach, fidelity, and accuracy required.

7.3 Future Work

Apart from the assumptions in the model of the magnet, it is possible to improve

the model itself in several aspects. Firstly, the material properties of the components used

are limited at cryogenic temperature. It is possible to estimate the material properties

such as elastic modulus, thermal expansion coefficient, thermal conductivity, and specific

heat using molecular dynamics modeling. However, the appropriate force field is still not

fully developed for MgB2 specifically. Density functional theory (DFT) is another approach

to predict the material property. However, DFT is limited regarding a number of atoms

considered in a computational model. The limited number of the atoms under study has

138
 Conclusion

some limitation predicting the material behavior in the continuum scale. Future studies

may focus on developing the theories and force fields to provide the molecular simulation

tools in order to estimate the material properties at cryogenic temperature. Additionally,

the temperature-magnetic field-current relationship of the composite superconducting

wire has been experimentally obtained. Development of a model to predict the behavior

of the superconducting composite will allow a more accurate model of the magnet

quench. The work presented in this dissertation is developed in collaboration with Hyper

Tech Research and Ohio State University Center for Superconducting and Magnetic

Materials. The recently manufactured single bundle conduction cooled MgB2 coil bundle

prototype that is under investigation at Ohio State University will provide experimental

tests of the stress, strain and temperature data in the near future. They will provide a

comparison and suggest further improvements for a more complete FEA model of the

magnet system. Manufacturing of a full body MRI magnet system is expensive and may

cost a million dollars; however, it is necessary to manufacture and test one before the

next generation system is approved for the market. If experimental and computational

results agree within reasonable error, it will help understand the behavior of the system

much better than past investigations. This model will then serve its intended purpose to

make an optimized design for high-performance conduction cooled MgB2 MRI magnet

systems.

139
 INDEX

B
F
background magnet, 6
finite element analysis, 30, 34, 67, 97, 104, 105, 107

C
G
ceramic compound, 62
Global-Local Analysis, 126
complex composite, 63, 93

Composite Wire, 62, 126, 131 H

cool-down, 8, 17, 34, 45, 47, 56, 65, 97, 99, 100, 109,
Hill approximation, 20, 21, 23, 90, 91, 230, 231
121, 124, 125, 128, 136
Homogenization, 62, 67, 78, 79, 82, 90, 94, 146, 160,
Cool-down Stress, 35, 106
168, 176, 231, 233, 234
Creep, 127, 231

D M

Magnet Bundles, 30, 110

Discretization, 83
magnetic resonance

E nuclear, 1

mandrel, 10, 30, 31, 34, 83, 97, 99, 101, 102, 104,
Elastic Properties, 19, 47, 67, 146
106, 110, 114, 116, 117, 119, 121, 122, 124, 125,
electromagnetic charging, 65, 97, 100, 108, 111, 121,
128, 129, 210, 211
124, 128
metal matrix composite, 23, 26, 62, 81, 85
Electro-Magnetic Stress, 107
MRI, 1
energization, 11, 12, 15, 47, 99, 102, 123, 134, 136

Energization, 99, 108 P

Experiment, 90
principal mechanical strain

140
 1st, 118
S
3rd, 118

Specific Heat, 27, 60, 82, 160

Q Support Conditions, 13, 119

quench
T
magnet quench, 5, 9, 10, 11, 12, 13, 15, 17, 18, 38,

39, 40, 41, 42, 43, 47, 56, 63, 88, 125, 128, 130, Thermal Conductivity, 24, 57, 79, 168, 232, 234

131, 134, 135, 136, 139, 234, 236, 237 Thermal Expansion Coefficient, 22, 53, 55, 78, 101

Quench, 40 Transition, 130, 225, 229

R V

representative unit cell Voigt, 20, 21, 65, 90, 230

RUC, 21, 132

W
Resistivity, 60, 61, 239

Reuss (Inverse Rule of Mixture), 20, 21, 65, 66, 90, 91, winding, 10, 13, 15, 17, 30, 31, 32, 33, 34, 47, 65, 83,
133, 230 97, 99, 104, 109, 110, 113, 117, 119, 121, 122,
Rule of Mixture, 20, 90 123, 126, 128, 129, 134, 136, 229
RVE Winding Stress, 104
Representative Volume Element, 21, 23, 26, 29,

30, 65, 68, 69, 73, 74, 76, 78, 81, 82, 83, 84,

126, 127, 131, 132, 146, 151, 172

141
 APPENDIX A. Material Properties

Table A.1 Poisson's ratio of materials

Temperature (K) MgB2 Niobium Copper Monel CTD 101k

10~300 0.181 0.4 0.355 0.32 0.355

Table A.2 Modulus of elasticity of materials(GPa)

Temperature (K) MgB2 Niobium Copper Monel CTD 101k

10 76.00 252.76 138.59 181.46 19.42
20 76.00 252.90 138.55 181.41 18.96
30 76.00 252.98 138.53 181.36 18.52
40 76.00 252.90 138.41 181.22 18.09
50 76.00 252.59 138.26 181.10 17.68
60 76.00 252.10 138.07 181.00 17.29
70 76.00 251.53 137.85 180.85 16.92
80 76.00 251.00 137.59 180.77 16.56
90 76.00 250.61 137.27 180.62 16.22
100 76.00 250.30 136.93 180.51 15.89
110 76.00 249.99 136.58 180.23 15.58
120 76.00 249.70 136.25 179.92 15.29
130 76.00 249.44 135.92 179.52 15.02
140 76.00 249.20 135.58 178.92 14.76
150 76.00 248.96 135.23 178.47 14.52
160 76.00 248.70 134.86 177.97 14.30
170 76.00 248.40 134.46 177.48 14.09
180 76.00 248.10 134.04 176.77 13.90
190 76.00 247.84 133.61 176.20 13.72
200 76.00 247.62 133.18 175.68 13.56
210 76.00 247.42 132.76 175.31 13.42
220 76.00 247.20 132.34 174.88 13.30
230 76.00 246.96 131.91 174.55 13.19
240 76.00 246.70 131.47 174.25 13.10
250 76.00 246.43 130.99 173.95 13.03
260 76.00 246.20 130.49 173.67 12.97
270 76.00 246.03 129.94 173.80 12.93
280 76.00 245.90 129.35 173.89 12.91
290 76.00 245.77 128.73 174.08 12.90

142
 300 76.00 245.60 128.07 174.71 12.91

Table A.3 Thermal strain of materials (mm/m)

CTD 101k
Temperature MgB2 Niobium Copper Monel
Direction 1 Direction 2 & 3
10 0 0 0 0 0 0
20 0.000243 0.00145 0.00124 0.00378 0.0284 0.0977
30 -9.9E-05 0.00737 0.00698 0.00756 0.0613 0.207
40 -0.00019 0.0206 0.0222 0.0299 0.0985 0.326
50 0.00112 0.0417 0.0502 0.0597 0.14 0.456
60 0.00446 0.0693 0.0929 0.0946 0.185 0.596
70 0.0082 0.103 0.151 0.136 0.235 0.746
80 0.0212 0.141 0.222 0.185 0.287 0.905
90 0.0341 0.183 0.307 0.238 0.342 1.0737
100 0.0521 0.229 0.402 0.303 0.4 1.2505
110 0.0756 0.277 0.507 0.38 0.461 1.4352
120 0.102 0.329 0.622 0.46 0.524 1.6273
130 0.134 0.382 0.744 0.549 0.59 1.8266
140 0.168 0.437 0.873 0.644 0.658 2.0327
150 0.213 0.494 1.0079 0.746 0.728 2.2451
160 0.255 0.552 1.149 0.85 0.8 2.463
170 0.301 0.612 1.2952 0.965 0.873 2.6859
180 0.355 0.673 1.4459 1.0818 0.947 2.9137
190 0.408 0.735 1.6006 1.2075 1.0226 3.1466
200 0.462 0.798 1.7586 1.3315 1.0984 3.384
210 0.52 0.862 1.9194 1.4584 1.1747 3.6248
220 0.58 0.928 2.0826 1.5876 1.2513 3.868
230 0.646 0.994 2.2479 1.7168 1.3282 4.1134
240 0.719 1.061 2.4151 1.839 1.4051 4.3616
250 0.787 1.1287 2.5842 1.966 1.4818 4.6123
260 0.861 1.1969 2.7552 2.0924 1.5582 4.8647
270 0.943 1.2654 2.9282 2.2191 1.6339 5.118
280 1.0211 1.3343 3.1036 2.3465 1.7088 5.3715
290 1.099 1.4037 3.2817 2.4685 1.7824 5.6251
300 1.1832 1.4734 3.4631 2.6248 1.8544 5.8791

143
 Table A.4 Thermal conductivity of materials (W/m-K)

Temperature (K) MgB2 Niobium Copper Monel CTD 101k

10 2.00E+00 5.50E+01 1.54E+03 1.74E+00 9.05E-02
20 4.79E+00 8.40E+01 2.42E+03 4.30E+00 1.33E-01
30 6.99E+00 8.70E+01 2.14E+03 6.90E+00 1.63E-01
40 8.68E+00 7.40E+01 1.49E+03 9.00E+00 1.87E-01
50 9.62E+00 6.50E+01 1.01E+03 1.10E+01 2.09E-01
60 1.06E+01 5.80E+01 7.41E+02 1.20E+01 2.25E-01
70 1.10E+01 5.40E+01 6.04E+02 1.30E+01 2.45E-01
80 1.13E+01 5.20E+01 5.29E+02 1.40E+01 2.58E-01
90 1.19E+01 5.10E+01 4.87E+02 1.47E+01 2.72E-01
100 1.20E+01 5.10E+01 4.62E+02 1.52E+01 2.84E-01
110 1.22E+01 5.05E+01 4.49E+02 1.58E+01 2.97E-01
120 1.23E+01 5.00E+01 4.35E+02 1.63E+01 3.09E-01
130 1.25E+01 5.00E+01 4.29E+02 1.68E+01 3.19E-01
140 1.26E+01 5.00E+01 4.22E+02 1.73E+01 3.29E-01
150 1.27E+01 5.00E+01 4.19E+02 1.78E+01 3.38E-01
160 1.28E+01 5.00E+01 4.15E+02 1.82E+01 3.46E-01
170 1.30E+01 5.00E+01 4.13E+02 1.86E+01 3.55E-01
180 1.33E+01 5.00E+01 4.10E+02 1.90E+01 3.63E-01
190 1.35E+01 5.00E+01 4.09E+02 1.94E+01 3.70E-01
200 1.37E+01 5.00E+01 4.07E+02 1.98E+01 3.76E-01
210 1.40E+01 5.00E+01 4.06E+02 2.00E+01 3.82E-01
220 1.43E+01 5.00E+01 4.05E+02 2.03E+01 3.87E-01
230 1.46E+01 5.00E+01 4.03E+02 2.05E+01 3.93E-01
240 1.48E+01 5.00E+01 4.02E+02 2.08E+01 3.99E-01
250 1.51E+01 5.00E+01 4.01E+02 2.10E+01 4.04E-01
260 1.52E+01 5.00E+01 4.00E+02 2.12E+01 4.09E-01
270 1.52E+01 5.00E+01 3.99E+02 2.14E+01 4.15E-01
280 1.53E+01 5.00E+01 3.98E+02 2.16E+01 4.20E-01
290 1.54E+01 5.00E+01 3.97E+02 2.18E+01 4.25E-01
300 1.55E+01 5.00E+01 3.96E+02 2.20E+01 4.31E-01

144
 Table A.5 Specific Heat of materials (J/Kg-K)

Temperature (K) MgB2 Nb Copper Monel CTD 101k

10 0.34 2.20 0.87 1.70 29.41
20 0.67 11.30 7.27 7.10 95.59
30 3.95 35.00 26.64 21.00 167.66
40 10.82 68.00 58.86 45.00 228.07
50 25.17 99.00 96.78 78.00 293.97
60 44.89 127.00 135.26 110.00 345.08
70 77.81 152.00 170.90 150.00 390.68
80 114.65 173.00 202.22 180.00 431.15
90 151.36 189.00 229.13 210.00 475.50
100 187.74 202.00 251.94 240.00 518.61
110 228.86 211.50 269.57 262.50 553.07
120 271.08 221.00 287.19 285.00 587.52
130 308.21 227.50 299.94 299.50 618.39
140 359.53 234.00 312.69 314.00 649.25
150 400.08 238.50 321.97 325.00 680.95
160 438.56 243.00 331.26 336.00 712.65
170 475.10 246.00 338.26 345.50 737.72
180 512.11 249.00 345.26 355.00 762.80
190 549.37 251.50 350.69 362.50 786.50
200 584.18 254.00 356.12 370.00 810.20
210 616.78 256.00 360.37 377.50 834.57
220 649.25 258.00 364.62 385.00 858.93
230 680.11 259.50 368.00 392.00 877.88
240 711.10 261.00 371.38 399.00 896.83
250 735.31 262.50 374.06 404.50 915.99
260 762.86 264.00 376.73 410.00 935.15
270 784.10 265.00 378.94 415.00 952.14
280 801.42 266.00 381.14 420.00 969.13
290 813.37 267.00 390.82 425.00 974.58
300 817.00 268.00 400.50 430.00 980.04

145
 APPENDIX B. Sample APDL Codes

B.1. Sample APDL Script for Homogenization of a Symmetry Reduced RVE for Elastic

Properties

finish
/clear
*GET, time1, Active,,TIME, WALL
/com, Thermal properties
/title, Homogenization Of Composite
/filename, Homogenization_Composite_Thermal, 1
*dim, directory, string, 160,3
!!!!!!!!!!!!analysis directory
directory(1,1) = 'F:\ANSYS'
!!!!!!!!!!!!geometry directory
directory(1,2) = 'F:\ wire_518'
!!!!!!!!!!!!materials directory
directory(1,3) = 'F:\ material_properties'
/CWD, directory(1,1)
CDOPT,IGES
/CWD, directory(1,2)
CDREAD,SOLID,'','',,'bundle_analysis_asm','igs'

/prep7
ASEL, ALL
ASEL, U, LOC,Z,0
ADELE,ALL,,,1
ASEL, ALL
AOVLAP, ALL

146
/pnum, area,1
ASEL, S, AREA,,6
Asel, a, area,,8
AADD, ALL
ASEL, S, AREA,,5
ASEL, A, AREA,,7
AADD, ALL
allsel, all
numcmp, all
!selecting epoxy
ASEL, S, AREA,,1,4
AADD, ALL
asel, all
aglue, all
numcmp, all

asel, s, area,,45
CM, epoxy, area
!selecting monel
ASEL, S, AREA,,15
asel, a, area,,35
CM, monel, area
!selecting copper
asel, s, area,,2
asel, a, area,,3
CM, cu, area
!selecting mgb2
asel, s, area,,5,14
asel, a, area,,25,34
cm, mgb2, area

147
!selecting niobium
asel, all
cmsel, u, epoxy
cmsel, u, monel
cmsel, u, cu
cmsel, u, mgb2
cm, nb, area
/CWD, directory(1,1)

a2 = 0.6e-3 !direction 2
a3 = 1.03923e-3 !direction 3
a1 = 0.5e-3 !direction 1
tol = 1e-3
max_temp_steps = 30
*DIM, EZ, ARRAY,max_temp_steps,1
*DIM, EX, ARRAY,max_temp_steps,1
*DIM, EY, ARRAY,max_temp_steps,1
*DIM, NUXY, ARRAY,max_temp_steps,1
*DIM, NUXZ, ARRAY,max_temp_steps,1
*DIM, NUYZ, ARRAY,max_temp_steps,1
*DIM, GXY, ARRAY,max_temp_steps,1
*DIM, GYZ, ARRAY,max_temp_steps,1
*DIM, GXZ, ARRAY,max_temp_steps,1
*DIM, ALP_theta, ARRAY, max_temp_steps,1
*DIM, ALP_radial, ARRAY, max_temp_steps,1
*DIM, ALP_axial, ARRAY, max_temp_steps,1
*DIM,C,ARRAY,6,6
*DIM,A,ARRAY,6,1
*DIM,C11,ARRAY,max_temp_steps,1

148
*DIM,C12,ARRAY,max_temp_steps,1
*DIM,C13,ARRAY,max_temp_steps,1
*DIM,C22,ARRAY,max_temp_steps,1
*DIM,C23,ARRAY,max_temp_steps,1
*DIM,C33,ARRAY,max_temp_steps,1
/PREP7 ! Pre-processor module
ASEL, ALL
APLOT
ARSCALE,ALL, , ,1/1000,1/1000,1/1000, ,1,1
ASEL, ALL
APLOT
/CWD, directory(1,3)
/INPUT,'MgB2_E','txt',,, 0
/INPUT,'niobium_E','txt',,, 0
/INPUT,'copper_E','txt',,, 0
/INPUT,'monel_E','txt',,, 0
/INPUT,'CTD_E','txt',,, 0
/INPUT,'MgB2_thsx','txt',,, 0
/INPUT,'niobium_thsx','txt',,, 0
/INPUT,'copper_thsx','txt',,, 0
/INPUT,'monel_thsx','txt',,, 0
/INPUT,'CTD_thsx','txt',,, 0
/CWD, directory(1)
ET,1,183
ET,2,SOLID186 ! Choose quad 186 element type
cmsel, s, mgb2
AATT, 1
cmsel, s, nb
AATT, 2
cmsel, s, cu

149
AATT, 3
cmsel, s, monel
AATT, 4
cmsel, s, epoxy
AATT, 5
ALLSEL, ALL
ALLSEL, ALL

!ARSYM,X,ALL
!ARSYM,Y,ALL
!NUMMRG,ALL, , , ,LOW
esize,0.01/1000
amesh,all
*GET,max_k,KP,0,num,max
KSEL, S, Loc, X, 0
KSEL, R, LOC, Y, 0
KSEL, R, LOC, Z, 0
*GET,min_k,KP,0,num,max
K,max_k+1,,,a1
L,max_k+1, min_k

*GET,max_ln,LINE,0,num,max
LESIZE,max_ln,,,5
TYPE, 2
EXTOPT,ATTR,1
EXTOPT,ACLEAR,1
VDRAG,ALL, , , , , ,max_ln ! Number of divisions on the
matrix
*create,srecover !,mac ! Create macro to calculate average
stress
/nopr

150
ETABLE, ,VOLU, ! Get element volume
ETABLE, ,S,X ! Get element stress
ETABLE, ,S,Y
ETABLE, ,S,Z
ETABLE, ,S,XY
ETABLE, ,S,XZ
ETABLE, ,S,YZ
ETABLE, ,S,YZ
SMULT,SXV,VOLU,SX,1,1, ! Stress by element volume
SMULT,SYV,VOLU,SY,1,1,
SMULT,SZV,VOLU,SZ,1,1,
SMULT,SXYV,VOLU,SXY,1,1,
SMULT,SXZV,VOLU,SXZ,1,1,
SMULT,SYZV,VOLU,SYZ,1,1,
SSUM
*get,totvol,ssum,,item,volu ! Integer stress along total
volume
*get,totsx ,ssum,,item,sxv
*get,totsy ,ssum,,item,syv
*get,totsz ,ssum,,item,szv
*get,totsxy ,ssum,,item,sxyv
*get,totsxz ,ssum,,item,sxzv
*get,totsyz ,ssum,,item,syzv

Sxx0 = totsx/totvol ! Compute average RVE stress

Syy0 = totsy/totvol
Szz0 = totsz/totvol
Sxy0 = totsxy/totvol
Sxz0 = totsxz/totvol
Syz0 = totsyz/totvol

151
/gopr
*end !srecover
FINISH

*DO, i, 1, max_temp_steps
/prep7
temp_applied = i*10
/SOLU ! Solution module
ANTYPE,STATIC ! Set static analysis

lsclear,all ! Boundary conditions Column 1

TREF, temp_applied
asel,s,loc,x,0 ! Model X direction = 2 material
direction
asel,a,loc,x,a2
da,all,ux,0

asel,s,loc,y,0 ! Model Y direction = 3 material

direction
asel,a,loc,y,a3
da,all,uy,0
asel,s,loc,z,0 ! Model Z direction = 1 material
direction
da,all,uz,0
asel,s,loc,z,a1
da,all,uz,a1
allsel,all
lswrite,1
lsclear,all ! Boundary conditions Column 2
TREF, temp_applied

152
asel,s,loc,x,0 ! Model X direction = 2 material
direction
da,all,ux,0
asel,s,loc,x,a2
da,all,ux,a2
asel,s,loc,y,0 ! Model Y direction = 3 material
direction
asel,a,loc,y,a3
da,all,uy,0
asel,s,loc,z,0 ! Model Z direction = 1 material
direction
asel,a,loc,z,a1
da,all,uz,0
allsel,all
lswrite,2
lsclear,all ! Boundary conditions Column 3
TREF, temp_applied
asel,s,loc,x,0 ! Model X direction = 2 material
direction
asel,a,loc,x,a2
da,all,ux,0
asel,s,loc,y,0 ! Model Y direction = 3 material
direction
DA,all,uy,0
asel,s,loc,y,a3
da,all,uy,a3
asel,s,loc,z,0 ! Model Z direction = 1 material
direction
asel,a,loc,z,a1
da,all,uz,0
asel,all
lswrite,3

153
lsclear,all ! Boundary conditions Column 1
TREF, temp_applied
asel,s,loc,x,0 ! Model X direction = 2 material
direction
da,all,asym
asel,s,loc,x,a2
da,all,uy,0.5*a2
asel,s,loc,y,0 ! Model Y direction = 3 material
direction
DA,ALL,ASYM
asel,s,loc,y,a3
da,all,ux,0.5*a3
asel,s,loc,z,0 ! Model Z direction = 1 material
direction
asel,a,loc,z,a1
da,all,uz,0
asel,all
lswrite,4
lsclear,all ! Boundary conditions Column 1
TREF, temp_applied
asel,s,loc,x,0 ! Model X direction = 2 material
direction
asel,a,loc,x,a2
da,all,ux,0
asel,s,loc,y,0 ! Model Y direction = 3 material
direction
da, all, asym
asel,s,loc,y,a3
da,all,uz,0.5*a3
asel,s,loc,z,0 ! Model Z direction = 1 material
direction
da, all, asym

154
asel,s,loc,z,a1
da,all,uy,0.5*a1
allsel,all
lswrite,5
lsclear,all ! Boundary conditions Column 1
TREF, temp_applied
asel,s,loc,x,0 ! Model X direction = 2 material
direction
da, all, asym
asel,s,loc,x,a2
da,all,uz,0.5*a2
asel,s,loc,y,0 ! Model Y direction = 3 material
direction
asel,a,loc,y,a3
da,all,uy,0
asel,s,loc,z,0 ! Model Z direction = 1 material
direction
da, all, asym
asel,s,loc,z,a1
da,all,ux,0.5*a1
asel,all
lswrite,6
lsclear,all ! Boundary conditions Column 1
allsel,all
TREF, 10
BF, ALL, TEMP, temp_applied
asel,s,loc,x,0 ! Model X direction = 2 material
direction
asel,a,loc,y,0
asel,a,loc,z,0 ! Model Z direction = 1 material
direction
da, all, symm

155
asel,s,loc,x,a2
da, all, ux, 0
asel,s,loc,y,a3
da, all, uy, 0
asel,s,loc,z,a1
da,all,uz,0
allsel,all
lswrite,7

LSSOLVE,1,7 ! Solve all load sets

FINISH ! Exit solution module
/post1
SET,1 ! First column coefficients
*use,srecover
C(1,1) = Szz0
C(2,1) = Sxx0
C(3,1) = Syy0
SET,2 ! Second column coefficients
*use,srecover
C(1,2) = Szz0
C(2,2) = Sxx0
C(3,2) = Syy0
SET,3 ! Third column coefficients
*use,srecover
C(1,3) = Szz0
C(2,3) = Sxx0
C(3,3) = Syy0
SET,4 ! Fith column coefficients
*use,srecover
C(4,4) = Sxy0

156
SET,5 ! Fith column coefficients
*use,srecover
C(5,5) = Syz0
SET,6 ! Sixth column coefficients
*use,srecover
C(6,6) = Sxz0
SET, 7
*use,srecover
A(1,1) = -Szz0
A(2,1) = -Sxx0
A(3,1) = -Syy0
A(4,1) = -Sxy0
A(5,1) = -Sxz0
A(6,1) = -Syz0
*MOPER, S_Mat, C, INVERT
*MOPER, ALPHA, S_Mat, MULT, A
ALP_theta(i,1) = alpha(1,1)/temp_applied
ALP_radial(i,1) = alpha(3,1)/temp_applied
ALP_axial(i,1) = alpha(2,1)/temp_applied
!ALP_theta(i,1) = alpha(1,1)
!ALP_radial(i,1) = alpha(3,1)
!ALP_axial(i,1) = alpha(2,1)
E1 = 1/S_Mat(1,1)
E2 = 1/S_Mat(2,2)
E3 = 1/S_Mat(3,3)
nu21 = -E2*S_Mat(1,2)
nu31 = -E3*S_Mat(1,3)
nu12 = -E1*S_Mat(2,1)
nu32 = -E3*S_Mat(2,3)
nu13 = -E1*S_Mat(3,1)

157
nu23 = -E2*S_Mat(3,2)
G23 = 1/S_Mat(4,4)
G31 = 1/S_Mat(5,5)
G12 = 1/S_Mat(6,6)

Ez(i,1) = E1
Ex(i,1) = E2
EY(i,1) = E3
NUXY(i,1) = (NU32+NU23)/2
NUXZ(i,1) = (NU21+NU12)/2
NUYZ(i,1) = (NU31+NU13)/2
GXY(i,1) = G23
GYZ(i,1) = G31
GXZ(i,1) = G12
*ENDDO
*create, ansuitmp
*cfopen, 'ANSYS_ELASTIC_Values','txt'

*VWRITE, EZ(1,1), EX(1,1), EY(1,1), NUXY(1,1), NUXZ(1,1),

NUYZ(1,1), GXY(1,1), GYZ(1,1), GXZ(1,1)
(12g16.8)
*cfclose
*END
/INPUT, ansuitmp
*create, ansuitmp
*cfopen, 'TEC_Values','txt'
*VWRITE, ALP_radial(1,1), ALP_axial(1,1), ALP_theta(1,1)
(3g16.8)
*cfclose
*END

158
/INPUT, ansuitmp
*GET, time2, Active,,TIME, WALL
total = time2-time1

159
B.2. Sample APDL Script for Homogenization of Specific Heat

finish
/clear
keyw,PR_SGVOF,1
/com, Thermal properties
/title, Homogenization Of Composite
/filename, Homogenization_Composite_Thermal, 1
*dim, directory, string, 160,3
!!!!!!!!!!!!analysis directory
directory(1,1) =
'F:\Abdullah\Writing\Articles\Journal004\wire_modeling\wire
_518\heat_capacity\ansys'
!!!!!!!!!!!!geometry directory
directory(1,2) =
'F:\Abdullah\Writing\Articles\Journal004\Wire_geom\wire_518
'
!!!!!!!!!!!materials directory
directory(1,3) =
'F:\Abdullah\Writing\Articles\Journal004\material_propertie
s'

*GET,TT1,ACTIVE,0,TIME,WALL
/CWD, directory(1,1)
CDOPT,IGES
/CWD, directory(1,2)

CDREAD,SOLID,'','',,'bundle_analysis_asm','igs'

/prep7
ASEL, ALL
ASEL, U, LOC,Z,0
ADELE,ALL,,,1
160
ASEL, ALL
AOVLAP, ALL

/pnum, area,1
ASEL, S, AREA,,6
Asel, a, area,,8
AADD, ALL
ASEL, S, AREA,,5
ASEL, A, AREA,,7
AADD, ALL
allsel, all
numcmp, all
!selecting epoxy
ASEL, S, AREA,,1,4
AADD, ALL
asel, all
aglue, all
numcmp, all
asel, s, area,,45
CM, epoxy, area
!selecting monel
ASEL, S, AREA,,15
asel, a, area,,35
CM, monel, area
!selecting copper
asel, s, area,,2
asel, a, area,,3
CM, cu, area
!selecting mgb2
asel, s, area,,5,14

161
asel, a, area,,25,34
cm, mgb2, area
!selecting niobium
asel, all
cmsel, u, epoxy
cmsel, u, monel
cmsel, u, cu
cmsel, u, mgb2
cm, nb, area

/CWD, directory(1,1)
!rf=3.5 ! Radius fiber in microns
a2 = 0.905/1000 ! x2 length in microns
a3=0.595/1000 ! x3 length in microns
a1 = 0.05/1000 ! x1 length in microns
tol = 1e-3
final_temp_steps = 31 !31 ! 30 data points until 300 K
time_steps = 0.1 ! values are from convergence test
were 0.1
final_time=1 ! time in seconds
q_gen = 1e3
rho_mgb2 = 1540
rho_nb = 8570
rho_cu = 8960
rho_monel = 8830
rho_epoxy = 1100
/PREP7 ! Pre-processor module
ASEL, ALL
APLOT
ARSCALE,ALL, , ,1/1000,1/1000,1/1000, ,1,1

162
ASEL, ALL
APLOT
/CWD, directory(1,3)
/INPUT,'MgB2_Cp','txt',,, 0
/INPUT,'niobium_Cp','txt',,, 0
/INPUT,'copper_Cp','txt',,, 0
/INPUT,'monel_Cp','txt',,, 0
/INPUT,'CTD_Cp','txt',,, 0
/INPUT,'MgB2_k','txt',,, 0
/INPUT,'niobium_k','txt',,, 0
/INPUT,'copper_k','txt',,, 0
/INPUT,'monel_k','txt',,, 0
/INPUT,'CTD_k','txt',,, 0
/CWD, directory(1,1)
MP, DENS, 1, rho_mgb2
MP, DENS, 2, rho_nb
MP, DENS, 3, rho_cu
MP, DENS, 4, rho_monel
MP, DENS, 5, rho_epoxy
ET,1,77
ET,2,SOLID90 ! Choose SOLID90 element type
numcmp, all
CMsel, s, mgb2
AATT, 1
CMsel, s, nb
AATT, 2
CMsel, s, cu
AATT, 3
CMsel, s, monel
AATT, 4

163
CMsel, s, epoxy
AATT, 5
CMSEL, S, mgb2, area
ASUM, FINE
*get, a_mgb2, area, all, area
CMSEL, S, nb, area
ASUM, FINE
*get, a_nb, area, all, area
CMSEL, S, cu, area
ASUM, FINE
*get, a_cu, area, all, area
CMSEL, S, monel, area
ASUM, FINE
*get, a_monel, area, all, area
CMSEL, S, epoxy, area
ASUM, FINE
*get, a_epoxy, area, all, area
ASEL, ALL
ASUM, FINE
*get, a_tot, area, all, area
f_mgb2 = a_mgb2/a_tot
f_nb = a_nb/a_tot
f_cu = a_cu/a_tot
f_monel = a_monel/a_tot
f_epoxy = a_epoxy/a_tot

density =
f_mgb2*rho_mgb2+f_nb*rho_nb+f_cu*rho_cu+f_monel*rho_monel+f
_epoxy*rho_epoxy
!ARSYM,X,ALL

164
!ARSYM,Y,ALL
!NUMMRG,ALL, , , ,LOW
esize,0.01/1000
amesh,all
!asel, all
!arefine, all,,,1
*GET,max_k,KP,0,num,max
KSEL, S, Loc, X, 0
KSEL, R, LOC, Y, 0
KSEL, R, LOC, Z, 0
*GET,min_k,KP,0,num,max
K,max_k+1,,,a1
L,max_k+1, min_k
LESIZE,1,1
*GET,max_ln,LINE,0,num,max
TYPE, 2
EXTOPT,ATTR,1
EXTOPT,ACLEAR,1
VDRAG,ALL, , , , , ,max_ln ! Number of divisions on the
matrix
*DIM, Cp, ARRAY, final_temp_steps, 2
*DO, i, 1, 2!final_temp_steps
/SOLU
TREF, (i-1)*10
TUNIF, (i-1)*10
cp(i,1) = (i-1)*10
ANTYPE, TRANS
!TIMINT, OFF, STRUC
!CONVTOL, HEAT
!CONVTOL, F

165
AUTOTS, ON
OUTRES, ,ALL
KBC, 1
DELTIM,time_steps,time_steps,time_steps
TIME, final_time
ALLSEL, ALL
BFV, ALL, HGEN, q_gen*i

ALLSEL, ALL
SAVE
SOLVE
/post1
SET, LAST
ETABLE, ,VOLU, ! Get element volume
ETABLE, ,TEMP
SMULT, T_V,VOLU,TEMP,1,1
SSUM
*get,totvol,ssum,,item,volu ! Integer stress along total
volume
*get,tot_temp,ssum,,item,T_V
avg_temp = tot_temp/totvol
del_Temp = avg_temp-cp(i,1)
cp(i,2) = (i*q_gen)/(density*del_Temp)
*ENDDO
*create, ansuitmp
*cfopen, 'Cp_values','txt'
*VWRITE, cp(1,1), cp(1,2)
(2g16.8)
*cfclose
*END

166
/INPUT, ansuitmp
*GET,TT2,ACTIVE,0,TIME,WALL
total_time = tt2-tt1
save

167
B.3. Sample APDL Script for Homogenization of the Thermal Conductivity

finish
/clear
/com, Thermal properties
/title, Homogenization Of Composite
/filename, Homogenization_Composite_Thermal, 1
*GET,TT1,ACTIVE,0,TIME,WALL
/CWD, 'F:\ansys'
CDOPT,IGES
/CWD, F:\wire_518
CDREAD,SOLID,'','',,'bundle_analysis_asm','igs'
/prep7
ASEL, ALL
ASEL, U, LOC,Z,0
ADELE,ALL,,,1
ASEL, ALL
AOVLAP, ALL
/pnum, area,1
ASEL, S, AREA,,6
Asel, a, area,,8
AADD, ALL
ASEL, S, AREA,,5
ASEL, A, AREA,,7
AADD, ALL

allsel, all
numcmp, all
!selecting epoxy
ASEL, S, AREA,,1,4

168
AADD, ALL
asel, all
aglue, all
numcmp, all
asel, s, area,,45
CM, epoxy, area
!selecting monel
ASEL, S, AREA,,15
asel, a, area,,35
CM, monel, area
!selecting copper
asel, s, area,,2
asel, a, area,,3
CM, cu, area
!selecting mgb2
asel, s, area,,5,14
asel, a, area,,25,34
cm, mgb2, area

!selecting niobium
asel, all
cmsel, u, epoxy
cmsel, u, monel
cmsel, u, cu
cmsel, u, mgb2
cm, nb, area

/CWD,
'F:\Abdullah\Writing\Articles\Journal004\wire_modeling\wire
_518\thermal_conductivity\ansys'
a2 = 0.6e-3 !direction 2

169
a3 = 1.03923e-3 !direction 3
a1 = 0.05e-3 !direction 1
tol = 1e-3
temp_points = 30
/PREP7 ! Pre-processor module
ASEL, ALL
APLOT
ARSCALE,ALL, , ,1/1000,1/1000,1/1000, ,1,1
ASEL, ALL
APLOT
/CWD,
F:\Abdullah\Writing\Articles\Journal004\material_properties

/prep7
/INPUT,'MgB2_k','txt','F:\Abdullah\Writing\Articles\Journal
004\material_properties',, 0
/INPUT,'niobium_k','txt','F:\Abdullah\Writing\Articles\Jour
nal004\material_properties',, 0
/INPUT,'copper_k','txt','F:\Abdullah\Writing\Articles\Journ
al004\material_properties',, 0
/INPUT,'monel_k','txt','F:\Abdullah\Writing\Articles\Journa
l004\material_properties',, 0
/INPUT,'CTD_k','txt','F:\Abdullah\Writing\Articles\Journal0
04\material_properties',, 0
/CWD,
'F:\Abdullah\Writing\Articles\Journal004\wire_modeling\wire
_518\thermal_conductivity\ansys'
ET,1,77
ET,2,SOLID90 ! Choose SOLID90 element type

numcmp, all
cmsel, s, mgb2

170
AATT, 1
cmsel, s, nb
AATT, 2
cmsel, s, cu
AATT, 3
cmsel, s, monel
AATT, 4
cmsel, s, epoxy
AATT, 5
ALLSEL, ALL
ALLSEL, ALL
!ARSYM,X,ALL
!ARSYM,Y,ALL
!NUMMRG,ALL, , , ,LOW
esize,0.01/1000
amesh,all

!asel, all
!arefine, all,,,1
*GET,max_k,KP,0,num,max
KSEL, S, Loc, X, 0
KSEL, R, LOC, Y, 0
KSEL, R, LOC, Z, 0
*GET,min_k,KP,0,num,max
K,max_k+1,,,a1
L,max_k+1, min_k
LESIZE,1,1
*GET,max_ln,LINE,0,num,max
TYPE, 2
EXTOPT,ATTR,1

171
EXTOPT,ACLEAR,1
VDRAG,ALL, , , , , ,max_ln ! Number of divisions on the
matrix
*create,srecover ! create macro to calculate average
heat flux
/POST1 ! Post-processor module
/nopr
ETABLE, ,VOLU, ! Get element volume
ETABLE, ,TF,X ! Get element stress
ETABLE, ,TF,Y
ETABLE, ,TF,Z
SMULT,TFXV,VOLU,TFX,1,1,
SMULT,TFYV,VOLU,TFY,1,1, ! Stress by element volume
SMULT,TFZV,VOLU,TFZ,1,1,
SSUM
*get,totvol,ssum,,item,volu ! Integer stress along total
volume
*get,totsx ,ssum,,item,TFXV
*get,totsy ,ssum,,item,TFYV
*get,totsz ,ssum,,item,TFZV

Sxx0 = (totsx*a2)/totvol ! Compute average RVE stress

Syy0 = (totsy*a3)/totvol
Szz0 = (totsz*a1)/totvol
/gopr
*END
*dim, k, ARRAY, temp_points,4
*Do, i, 1, temp_points
/SOLU ! Solution module
TREF, 10*i
T1=10*i

172
T2=T1+1
lsclear,all ! Boundary conditions Column 1
Asel,s,loc,x,0 ! Model X direction = 2 material
direction
Asel,a,loc,x,a2+tol
SFA,all,1,HFLUX,0,
Asel,s,loc,y,0 ! Model Y direction = 3 material
direction
Asel,a,loc,y,a3
SFA,all,1,HFLUX,0,
Asel,s,loc,z,0 ! Model Z direction = 1 material
direction
DA,all,TEMP,T1
Asel,s,loc,z,a1
da,all,TEMP,T2
allsel,all
lswrite,1

lsclear,all ! Boundary conditions Column 2

asel,s,loc,x,0 ! Model X direction = 2 material
direction
DA,all,TEMP,T1
asel,s,loc,x,a2
da,all,TEMP,T2
asel,s,loc,y,0 ! Model Y direction = 3 material
direction
asel,a,loc,y,a3
SFA,all,1,HFLUX,0,
asel,s,loc,z,0 ! Model Z direction = 1 material
direction
asel,a,loc,z,a1
SFA,all,1,HFLUX,0,

173
asel,all
lswrite,2

lsclear,all ! Boundary conditions Column 3

asel,s,loc,x,0 ! Model X direction = 2 material
direction
asel,a,loc,x,a2
SFA,all,1,HFLUX,0,
asel,s,loc,y,0 ! Model Y direction = 3 material
direction
DA,all,TEMP,T1
asel,s,loc,y,a3
da,all,TEMP,T2
asel,s,loc,z,0 ! Model Z direction = 1 material
direction
asel,a,loc,z,a1
SFA,all,1,HFLUX,0,
asel,all
lswrite,3
LSSOLVE,1,3 ! Solve all load sets
FINISH ! Exit solution module
/POST1 ! Post-processor module
SET,1 ! First column coefficients
*use, srecover
K(i,1) = -Szz0
SET,2 ! Second column coefficients
*use,srecover
K(i,2) = -Sxx0

SET,3 ! Third column coefficients

*use,srecover

174
K(i,3) = -Syy0
k(i,4) = i*10
*ENDDO
*GET,TT2,ACTIVE,0,TIME,WALL
Run_time=(TT2-TT1)*60
*CREATE,ansuitmp
*CFOPEN,'Kth_ISo','txt',' '
*VWRITE,k(1,4),k(1,1), k(1,2), k(1,3)
(4e16.8)
*CFCLOSE
*END
/INPUT,ansuitmp

175
B.4. Sample APDL script for Localization

finish
/clear
keyw,PR_SGVOF,1
*GET, time1, Active,,TIME, WALL
/com, Thermal properties
/title, Homogenization Of Composite
/filename, Homogenization_Composite_Thermal, 1
*dim, directory, string, 160,4
!!!!!!!!!!!!analysis directory
directory(1,1) = 'F:\Ansys'
!!!!!!!!!!!!geometry directory
directory(1,2) = 'F:\wire_518'
!!!!!!!!!!!!materials directory
directory(1,3) = 'F:\material_properties'
!!!!!!!!!!!!macro directory
directory(1,4) = 'F:\ANSYS_macros'
/PSEARCH, directory(1,4)
/CWD, directory(1,1)
CDOPT,IGES
/CWD, directory(1,2)
CDREAD,SOLID,'','',,'bundle_analysis_asm','igs'
/prep7
ASEL, ALL
ASEL, U, LOC,Z,0
ADELE,ALL,,,1
ASEL, ALL
AOVLAP, ALL
/pnum, area,1

176
ASEL, S, AREA,,6
Asel, a, area,,8
AADD, ALL
ASEL, S, AREA,,5
ASEL, A, AREA,,7
AADD, ALL
allsel, all
numcmp, all
!selecting epoxy
ASEL, S, AREA,,1,4
AADD, ALL
asel, all
aglue, all
numcmp, all
asel, s, area,,45
CM, epoxy, area
!selecting monel
ASEL, S, AREA,,15
asel, a, area,,35
CM, monel, area
!selecting copper
asel, s, area,,2
asel, a, area,,3
CM, cu, area
!selecting mgb2
asel, s, area,,5,14
asel, a, area,,25,34
cm, mgb2, area
!selecting niobium
asel, all

177
cmsel, u, epoxy
cmsel, u, monel
cmsel, u, cu
cmsel, u, mgb2
cm, nb, area
/CWD, directory(1,1)
a2 = 0.6e-3 !direction 2
a3 = 1.03923e-3 !direction 3
a1 = 0.5e-3 !0.05e-3 !direction 1
tol = 1e-5
max_temp_steps = 30
*DIM, EZ, ARRAY,max_temp_steps,1
*DIM, EX, ARRAY,max_temp_steps,1
*DIM, EY, ARRAY,max_temp_steps,1
*DIM, NUXY, ARRAY,max_temp_steps,1
*DIM, NUXZ, ARRAY,max_temp_steps,1
*DIM, NUYZ, ARRAY,max_temp_steps,1
*DIM, GXY, ARRAY,max_temp_steps,1
*DIM, GYZ, ARRAY,max_temp_steps,1
*DIM, GXZ, ARRAY,max_temp_steps,1
*DIM, ALP_theta, ARRAY, max_temp_steps,1
*DIM, ALP_radial, ARRAY, max_temp_steps,1
*DIM, ALP_axial, ARRAY, max_temp_steps,1
*DIM,C,ARRAY,6,6
*DIM,A_strain,ARRAY,6,1
*DIM,C11,ARRAY,max_temp_steps,1
*DIM,C12,ARRAY,max_temp_steps,1
*DIM,C13,ARRAY,max_temp_steps,1
*DIM,C22,ARRAY,max_temp_steps,1
*DIM,C23,ARRAY,max_temp_steps,1

178
*DIM,C33,ARRAY,max_temp_steps,1

/PREP7 ! Pre-processor module

ASEL, ALL
APLOT
ARSCALE,ALL, , ,1/1000,1/1000,1/1000, ,1,1
ASEL, ALL
APLOT
/CWD, directory(1,3)
/INPUT,'MgB2_E','txt',,, 0
/INPUT,'niobium_E','txt',,, 0
/INPUT,'copper_E','txt',,, 0
/INPUT,'monel_E','txt',,, 0
/INPUT,'CTD_E','txt',,, 0
/INPUT,'MgB2_thsx','txt',,, 0
/INPUT,'niobium_thsx','txt',,, 0
/INPUT,'copper_thsx','txt',,, 0
/INPUT,'monel_thsx','txt',,, 0
/INPUT,'CTD_thsx','txt',,, 0
/DEVICE,VECTOR,1
/CWD, directory(1)
ET,1,183
ET,2,SOLID186 ! Choose quad 186 element type
cmsel, s, mgb2
AATT, 1
cmsel, s, nb
AATT, 2
cmsel, s, cu
AATT, 3
cmsel, s, monel

179
AATT, 4
cmsel, s, epoxy
AATT, 5
ALLSEL, ALL
ALLSEL, ALL
!ARSYM,X,ALL
!ARSYM,Y,ALL
!NUMMRG,ALL, , , ,LOW
esize,0.02/1000
amesh,all
arsym, x, all
arsym, y, all
nummrg, node, 1e-10
nummrg, kp, 1e-6
numcmp, all
*GET,max_k,KP,0,num,max
KSEL, S, Loc, X, 0
KSEL, R, LOC, Y, 0
KSEL, R, LOC, Z, 0
*GET,min_k,KP,0,num,max
K,max_k+1,,,a1
L,max_k+1, min_k
*GET,max_ln,LINE,0,num,max
LSEL, S, line,,max_ln-1,max_ln
LESIZE,all,,,5
TYPE, 2
EXTOPT,ATTR,1
EXTOPT,ACLEAR,1
VDRAG,ALL, , , , , ,max_ln ! Number of divisions on the
matrix

180
nummrg, node, 1e-10
nummrg, kp, 1e-6
numcmp, all
allsel, all
*GET, max_node,node,0,num,max
NGEN,2,max_node,ALL, , , , , -a1,1,
EGEN,2,max_node,ALL,
nummrg, nodel, 1e-8
FINISH
ALLSEL, ALL
save
/SOLU ! Solution module
temp_applied = 300
TREF, temp_applied
*GET, time1, Active,,TIME, WALL
!*use,ceRVE,a1,a2,a3,dir_11,dir_22,dir_33,dir_23,dir_13,dir
_12
ceRVE,a1,a2,a3,0.32e-3,-0.377e-4,-0.176e-3,0.001332,0,0
*GET, time2, Active,,TIME, WALL
CE_time = time2-time1
*GET, time3, Active,,TIME, WALL
SOLVE ! Solve analysis
*GET, time4, Active,,TIME, WALL
CE_time = time2-time1
solu_time = time4-time3
/device, vector, 0
!!!!!!!!Maximum Shear Stress
/post1
ALLSEL, ALL
i = 1

181
SET,,,,,,,i
AVPRIN,0,,
ETABLE,S1_%i%,S,1
AVPRIN,0,,
ETABLE,S3_%i%,S,3
SADD,S_Max_%i%,S1_%i%,S3_%i%,0.5,-0.5,0,
total_time = (time4-time1)*60

182
B.5. Sample APDL Script for Magnet Analysis

FINISH
/clear, nostart
!CHANGE DIRECTORY
/MKDIR, Test1
/CWD, D:\ANSYS
/DIRECTORY,,EMSTRESS,,
/filename, EMStress, 1
/com, STRESS MODELING of Electromagnetic Charging
/prep7
/title, STRESS MODELING of Electromagnetic Charging
ET,1,PLANE233,,,1 ! Plane13, AZ DOF, AXISYMMETRIC, FOR
AIR
ET,2,PLANE233,,,1 ! Plane13, AZ,UX,UY,TEMP DOF,
AXISYMMETRIC, FOR Steel and Coil material

MP, EX, 1, 1.19E+11 !MATERIAL 1 IS WIRE

MP, EY, 1, 1.23E+11
MP, EZ, 1, 1.53E+11
MP, MURX, 1, 1
MP, NUXY, 1, 0.332
MP, NUYZ, 1, 0.284
MP, NUXZ, 1, 0.291
MP, GXY, 1, 3.61E+10
MP, GYZ, 1, 4.39E+10
MP, GXZ, 1, 4.09E+10
MP, ALPX, 1, 1.07899E-05
MP, ALPY, 1, 1.07899E-05
MP, ALPZ, 1, 9.25988E-06

183
MP, EX, 2, 190E9 !MATERIAL 2 IS Steel
MP, MURX, 2, 1
MP, NUXY, 2, 0.305
MP, ALPX, 2, 1.2E-5
MP, MURX,2,1
MP, MURX,3,1 ! Material is air Relative permiability
Mthick = 0.01
EMUNIT,MKS ! MKS UNITS
!COIL Specification
R1_i = 0.500058015
R1_o = 0.532188015
R2_i = 0.500572511
R2_o = 0.532702511
R3_i = 0.500001488
R3_o = 0.53213
R4_i = 0.5
R4_o = 0.53213
R5_i = 0.905863086
R5_o = 0.930853086

Z1_l = 0.0496339006784
Z1_h = 0.116770401
Z2_l = 0.227363742
Z2_h = 0.318088742
Z3_l = 0.463904874
Z3_h = 0.576403874
Z4_l = 0.659515695
Z4_h = 0.869997695
Z5_l = 0.61216759
Z5_h = 0.80631909

184
!Layer Number
Bundle_1 = 27
Bundle_2 = 27
Bundle_3 = 27
Bundle_4 = 27
Bundle_5 = 21
tol = 1E-6
tagcp = 1
!FORMATION OF AIR
RECTNG, 1.1, 2, 0, 1.1
RECTNG, 0, 0.3, 0, 1.1
RECTNG, 2, 3, 0, 3
RECTNG, 0, 2, 1.1, 3
RECTNG, 0.3, 1.1, 0, 1.1
ASEL, ALL
AGLUE, ALL
NUMCMP, AREA
RECTNG, R1_i-MThick, R1_o, Z1_l, Z1_h
ASBA,3,6
ASEL, ALL
AGLUE, ALL
NUMCMP, ALL
RECTNG, R2_i-MThick, R2_o, Z2_l, Z2_h
ASBA,5,6
AGLUE, ALL
NUMCMP, ALL
RECTNG, R3_i-MThick, R3_o, Z3_l, Z3_h
ASBA,5,6
AGLUE, ALL
NUMCMP, ALL

185
RECTNG, R4_i-MThick, R4_o, Z4_l, Z4_h
ASBA,5,6
NUMCMP, ALL
RECTNG, R5_i-MThick, R5_o, Z5_l, Z5_h
ASBA,5,6
NUMCMP, ALL
AGLUE, ALL
NUMCMP, ALL
!formation of COIL
NUMSTR, AREA, 100
RECTNG, R1_i-MThick,R1_i,Z1_l,Z1_h
*DO, j,1,Bundle_1
RECTNG, R1_i+(j-1)*((R1_o-
R1_i)/Bundle_1),R1_i+(j)*((R1_o-R1_i)/Bundle_1),Z1_l,(Z1_h-
z1_L)/2+Z1_l
RECTNG, R1_i+(j-1)*((R1_o-
R1_i)/Bundle_1),R1_i+(j)*((R1_o-R1_i)/Bundle_1),(Z1_h-
z1_L)/2+Z1_l,z1_h
!ASEL, S, AREA,,5+j,
*ENDDO
AGLUE, ALL
NUMCMP, ALL
NUMSTR, AREA, 200
RECTNG, R2_i-MThick,R2_i,Z2_l,Z2_h
*DO, j,1,Bundle_2

RECTNG, R2_i+(j-1)*((R2_o-
R2_i)/Bundle_2),R2_i+(j)*((R2_o-R2_i)/Bundle_2),Z2_l,(Z2_h-
z2_L)/2+Z2_l
RECTNG, R2_i+(j-1)*((R2_o-
R2_i)/Bundle_2),R2_i+(j)*((R2_o-R2_i)/Bundle_2),(Z2_h-
z2_L)/2+Z2_l,z2_h
*ENDDO

186
NUMSTR, AREA, 300
RECTNG, R3_i-MThick,R3_i,Z3_l,Z3_h
*DO, j,1,Bundle_3
RECTNG, R3_i+(j-1)*((R3_o-
R3_i)/Bundle_3),R3_i+(j)*((R3_o-R3_i)/Bundle_3),Z3_l,(Z3_h-
z3_L)/2+Z3_l
RECTNG, R3_i+(j-1)*((R3_o-
R3_i)/Bundle_3),R3_i+(j)*((R3_o-R3_i)/Bundle_3),(Z3_h-
z3_L)/2+Z3_l,z3_h
*ENDDO
NUMSTR, AREA, 400
RECTNG, R4_i-MThick,R4_i,Z4_l,Z4_h
*DO, j,1,Bundle_4
RECTNG, R4_i+(j-1)*((R4_o-
R4_i)/Bundle_4),R4_i+(j)*((R4_o-R4_i)/Bundle_4),Z4_l,(Z4_h-
z4_L)/2+Z4_l
RECTNG, R4_i+(j-1)*((R4_o-
R4_i)/Bundle_4),R4_i+(j)*((R4_o-R4_i)/Bundle_4),(Z4_h-
z4_L)/2+Z4_l,z4_h
*ENDDO
NUMSTR, AREA, 500
RECTNG, R5_i-MThick,R5_i,Z5_l,Z5_h
*DO, j,1,Bundle_5
RECTNG, R5_i+(j-1)*((R5_o-
R5_i)/Bundle_5),R5_i+(j)*((R5_o-R5_i)/Bundle_5),Z5_l,(Z5_h-
z5_L)/2+Z5_l
RECTNG, R5_i+(j-1)*((R5_o-
R5_i)/Bundle_5),R5_i+(j)*((R5_o-R5_i)/Bundle_5),(Z5_h-
z5_L)/2+Z5_l,z5_h
*ENDDO
ASEL, S, AREA,,100,599
APLOT
ASEL, ALL
AGLUE, ALL

187
NUMCMP, ALL
!Selecting bundles macro
*CREATE, BUNDLE, MAC

ASEL, S, AREA,,5,6
ASEL, A, AREA,,8,32 !Bundle 1 lower layers
ASEL, A, AREA,,7,7 !Bundle 1 upper layers
ASEL, A, AREA,,34,58
ASEL, A, AREA,,33
APLOT
ASEL, A, AREA,,60,61 !Bundle 2 lower layers
ASEL, A, AREA,,63,87
ASEL, A, AREA,,62,62 !Bundle 2 upper layers
ASEL, A, AREA,,170,194
ASEL, A, AREA,,166
ASEL, A, AREA,,88,89 !Bundle 3 lower layers
ASEL, A, AREA,,91,115
ASEL, A, AREA,,90,90 !Bundle 3 upper layers
ASEL, A, AREA,,195,219
ASEL, A, AREA,,167,167
ASEL, A, AREA,,116,117 !Bundle 4 lower layers
ASEL, A, AREA,,119,143
ASEL, A, AREA,,118,118 !Bundle 4 upper layers
ASEL, A, AREA,,220,244
ASEL, A, AREA,,168,168
ASEL, A, AREA,,144,145 !Bundle 5 lower layers
ASEL, A, AREA,,147,165
ASEL, A, AREA,,146,146 !Bundle 5 upper layers
ASEL, A, AREA,,245,263
ASEL, A, AREA,,169,169

188
APLOT
*END
!Select layers and define as macros
*CREATE, Bundle1layer1, MAC
ASEL, S, AREA,,5
ASEL, A, AREA,,7
ESLA
*END
*CREATE, Bundle1layer2, MAC
ASEL, S, AREA,,6
ASEL, A, AREA,,34
ESLA
*END
*CREATE, Bundle1layer3, MAC
ASEL, S, AREA,,8
ASEL, A, AREA,,35
ESLA
*END
*CREATE, Bundle1layer4, MAC
ASEL, S, AREA,,9
ASEL, A, AREA,,36
ESLA
*END
*CREATE, Bundle1layer5, MAC
ASEL, S, AREA,,10
ASEL, A, AREA,,37
ESLA
*END
*CREATE, Bundle1layer6, MAC
ASEL, S, AREA,,11

189
ASEL, A, AREA,,38
ESLA
*END
*CREATE, Bundle1layer7, MAC
ASEL, S, AREA,,12
ASEL, A, AREA,,39
ESLA
*END
*CREATE, Bundle1layer8, MAC
ASEL, S, AREA,,13
ASEL, A, AREA,,40
ESLA
*END
*CREATE, Bundle1layer9, MAC
ASEL, S, AREA,,14
ASEL, A, AREA,,41
ESLA
*END
*CREATE, Bundle1layer10, MAC
ASEL, S, AREA,,15
ASEL, A, AREA,,42
ESLA
*END
*CREATE, Bundle1layer11, MAC
ASEL, S, AREA,,16
ASEL, A, AREA,,43
ESLA
*END
*CREATE, Bundle1layer12, MAC
ASEL, S, AREA,,17

190
ASEL, A, AREA,,44
ESLA
*END
*CREATE, Bundle1layer13, MAC
ASEL, S, AREA,,18
ASEL, A, AREA,,45
ESLA
*END
*CREATE, Bundle1layer14, MAC
ASEL, S, AREA,,19
ASEL, A, AREA,,46
ESLA
*END
*CREATE, Bundle1layer15, MAC
ASEL, S, AREA,,20
ASEL, A, AREA,,47
ESLA
*END
*CREATE, Bundle1layer16, MAC
ASEL, S, AREA,,21
ASEL, A, AREA,,48
ESLA
*END
*CREATE, Bundle1layer17, MAC
ASEL, S, AREA,,22
ASEL, A, AREA,,49
ESLA
*END
*CREATE, Bundle1layer18, MAC
ASEL, S, AREA,,23

191
ASEL, A, AREA,,50
ESLA
*END
*CREATE, Bundle1layer19, MAC
ASEL, S, AREA,,24
ASEL, A, AREA,,51
ESLA
*END
*CREATE, Bundle1layer20, MAC
ASEL, S, AREA,,25
ASEL, A, AREA,,52
ESLA
*END
*CREATE, Bundle1layer21, MAC
ASEL, S, AREA,,26
ASEL, A, AREA,,53
ESLA

*END
*CREATE, Bundle1layer22, MAC
ASEL, S, AREA,,27
ASEL, A, AREA,,54
ESLA
*END
*CREATE, Bundle1layer23, MAC
ASEL, S, AREA,,28
ASEL, A, AREA,,55
ESLA
*END
*CREATE, Bundle1layer24, MAC

192
ASEL, S, AREA,,29
ASEL, A, AREA,,56
ESLA
*END
*CREATE, Bundle1layer25, MAC
ASEL, S, AREA,,30
ASEL, A, AREA,,57
ESLA
*END
*CREATE, Bundle1layer26, MAC
ASEL, S, AREA,,31
ASEL, A, AREA,,58
ESLA
*END
*CREATE, Bundle1layer27, MAC
ASEL, S, AREA,,32
ASEL, A, AREA,,33
ESLA
*END
!Bundle 2
*CREATE, Bundle2layer1, MAC
ASEL, S, AREA,,60
ASEL, A, AREA,,62
ESLA
*END
*CREATE, Bundle2layer2, MAC
ASEL, S, AREA,,61
ASEL, A, AREA,,170
ESLA
*END

193
*CREATE, Bundle2layer3, MAC
ASEL, S, AREA,,63
ASEL, A, AREA,,171
ESLA
*END
*CREATE, Bundle2layer4, MAC
ASEL, S, AREA,,64
ASEL, A, AREA,,172
ESLA
*END
*CREATE, Bundle2layer5, MAC
ASEL, S, AREA,,65
ASEL, A, AREA,,173
ESLA
*END
*CREATE, Bundle2layer6, MAC
ASEL, S, AREA,,66
ASEL, A, AREA,,174
ESLA
*END
*CREATE, Bundle2layer7, MAC
ASEL, S, AREA,,67
ASEL, A, AREA,,175
ESLA
*END
*CREATE, Bundle2layer8, MAC
ASEL, S, AREA,,68
ASEL, A, AREA,,176
ESLA
*END

194
*CREATE, Bundle2layer9, MAC
ASEL, S, AREA,,69
ASEL, A, AREA,,177
ESLA
*END
*CREATE, Bundle2layer10, MAC
ASEL, S, AREA,,70
ASEL, A, AREA,,178
ESLA
*END
*CREATE, Bundle2layer11, MAC
ASEL, S, AREA,,71
ASEL, A, AREA,,179
ESLA
*END
*CREATE, Bundle2layer12, MAC
ASEL, S, AREA,,72
ASEL, A, AREA,,180
ESLA

*END
*CREATE, Bundle2layer13, MAC
ASEL, S, AREA,,73
ASEL, A, AREA,,181
ESLA
*END
*CREATE, Bundle2layer14, MAC
ASEL, S, AREA,,74
ASEL, A, AREA,,182
ESLA

195
*END
*CREATE, Bundle2layer15, MAC
ASEL, S, AREA,,75
ASEL, A, AREA,,183
ESLA
*END
*CREATE, Bundle2layer16, MAC
ASEL, S, AREA,,76
ASEL, A, AREA,,184
ESLA
*END
*CREATE, Bundle2layer17, MAC
ASEL, S, AREA,,77
ASEL, A, AREA,,185
ESLA
*END
*CREATE, Bundle2layer18, MAC
ASEL, S, AREA,,78
ASEL, A, AREA,,186
ESLA
*END
*CREATE, Bundle2layer19, MAC
ASEL, S, AREA,,79
ASEL, A, AREA,,187
ESLA
*END
*CREATE, Bundle2layer20, MAC
ASEL, S, AREA,,80
ASEL, A, AREA,,188
ESLA

196
*END
*CREATE, Bundle2layer21, MAC
ASEL, S, AREA,,81
ASEL, A, AREA,,189
ESLA
*END
*CREATE, Bundle2layer22, MAC
ASEL, S, AREA,,82
ASEL, A, AREA,,190
ESLA
*END
*CREATE, Bundle2layer23, MAC
ASEL, S, AREA,,83
ASEL, A, AREA,,191
ESLA
*END
*CREATE, Bundle2layer24, MAC
ASEL, S, AREA,,84
ASEL, A, AREA,,192
ESLA
*END
*CREATE, Bundle2layer25, MAC
ASEL, S, AREA,,85
ASEL, A, AREA,,193
ESLA
*END
*CREATE, Bundle2layer26, MAC
ASEL, S, AREA,,86
ASEL, A, AREA,,194
ESLA

197
*END
*CREATE, Bundle2layer27, MAC
ASEL, S, AREA,,87
ASEL, A, AREA,,166
ESLA

*END
!Bundle 3
*CREATE, Bundle3layer1, MAC
ASEL, S, AREA,,88
ASEL, A, AREA,,90
ESLA
*END
*CREATE, Bundle3layer2, MAC
ASEL, S, AREA,,89
ASEL, A, AREA,,195
ESLA
*END
*CREATE, Bundle3layer3, MAC
ASEL, S, AREA,,91
ASEL, A, AREA,,196
ESLA
*END
*CREATE, Bundle3layer4, MAC
ASEL, S, AREA,,92
ASEL, A, AREA,,197
ESLA
*END
*CREATE, Bundle3layer5, MAC
ASEL, S, AREA,,93

198
ASEL, A, AREA,,198
ESLA
*END
*CREATE, Bundle3layer6, MAC
ASEL, S, AREA,,94
ASEL, A, AREA,,199
ESLA
*END
*CREATE, Bundle3layer7, MAC
ASEL, S, AREA,,95
ASEL, A, AREA,,200
ESLA
*END
*CREATE, Bundle3layer8, MAC
ASEL, S, AREA,,96
ASEL, A, AREA,,201
ESLA
*END
*CREATE, Bundle3layer9, MAC
ASEL, S, AREA,,97
ASEL, A, AREA,,202
ESLA
*END
*CREATE, Bundle3layer10, MAC
ASEL, S, AREA,,98
ASEL, A, AREA,,203
ESLA
*END
*CREATE, Bundle3layer11, MAC
ASEL, S, AREA,,99

199
ASEL, A, AREA,,204
ESLA
*END
*CREATE, Bundle3layer12, MAC
ASEL, S, AREA,,100
ASEL, A, AREA,,205
ESLA
*END
*CREATE, Bundle3layer13, MAC
ASEL, S, AREA,,101
ASEL, A, AREA,,206
ESLA
*END
*CREATE, Bundle3layer14, MAC
ASEL, S, AREA,,102
ASEL, A, AREA,,207
ESLA
*END
*CREATE, Bundle3layer15, MAC
ASEL, S, AREA,,103
ASEL, A, AREA,,208
ESLA
*END
*CREATE, Bundle3layer16, MAC
ASEL, S, AREA,,104
ASEL, A, AREA,,209
ESLA
*END
*CREATE, Bundle3layer17, MAC
ASEL, S, AREA,,105

200
ASEL, A, AREA,,210
ESLA
*END
*CREATE, Bundle3layer18, MAC
ASEL, S, AREA,,106
ASEL, A, AREA,,211
ESLA

*END
*CREATE, Bundle3layer19, MAC
ASEL, S, AREA,,107
ASEL, A, AREA,,212
ESLA
*END
*CREATE, Bundle3layer20, MAC
ASEL, S, AREA,,108
ASEL, A, AREA,,213
ESLA
*END
*CREATE, Bundle3layer21, MAC
ASEL, S, AREA,,109
ASEL, A, AREA,,214
ESLA
*END
*CREATE, Bundle3layer22, MAC
ASEL, S, AREA,,110
ASEL, A, AREA,,215
ESLA
*END
*CREATE, Bundle3layer23, MAC

201
ASEL, S, AREA,,111
ASEL, A, AREA,,216
ESLA
*END
*CREATE, Bundle3layer24, MAC
ASEL, S, AREA,,112
ASEL, A, AREA,,217
ESLA
*END
*CREATE, Bundle3layer25, MAC
ASEL, S, AREA,,113
ASEL, A, AREA,,218
ESLA
*END
*CREATE, Bundle3layer26, MAC
ASEL, S, AREA,,114
ASEL, A, AREA,,219
ESLA
*END
*CREATE, Bundle3layer27, MAC
ASEL, S, AREA,,115
ASEL, A, AREA,,167
ESLA
*END
!Bundle 4
*CREATE, Bundle4layer1, MAC
ASEL, S, AREA,,116
ASEL, A, AREA,,118
ESLA
*END

202
*CREATE, Bundle4layer2, MAC
ASEL, S, AREA,,117
ASEL, A, AREA,,220
ESLA
*END
*CREATE, Bundle4layer3, MAC
ASEL, S, AREA,,119
ASEL, A, AREA,,221
ESLA
*END
*CREATE, Bundle4layer4, MAC
ASEL, S, AREA,,120
ASEL, A, AREA,,222
ESLA
*END
*CREATE, Bundle4layer5, MAC
ASEL, S, AREA,,121
ASEL, A, AREA,,223
ESLA
*END
*CREATE, Bundle4layer6, MAC
ASEL, S, AREA,,122
ASEL, A, AREA,,224
ESLA
*END
*CREATE, Bundle4layer7, MAC
ASEL, S, AREA,,123
ASEL, A, AREA,,225
ESLA
*END

203
*CREATE, Bundle4layer8, MAC
ASEL, S, AREA,,124
ASEL, A, AREA,,226
ESLA
*END
*CREATE, Bundle4layer9, MAC
ASEL, S, AREA,,125
ASEL, A, AREA,,227
ESLA
*END
*CREATE, Bundle4layer10, MAC
ASEL, S, AREA,,126
ASEL, A, AREA,,228
ESLA
*END
*CREATE, Bundle4layer11, MAC
ASEL, S, AREA,,127
ASEL, A, AREA,,229
ESLA
*END
*CREATE, Bundle4layer12, MAC
ASEL, S, AREA,,128
ASEL, A, AREA,,230
ESLA
*END
*CREATE, Bundle4layer13, MAC
ASEL, S, AREA,,129
ASEL, A, AREA,,231
ESLA
*END

204
*CREATE, Bundle4layer14, MAC
ASEL, S, AREA,,130
ASEL, A, AREA,,232
ESLA
*END
*CREATE, Bundle4layer15, MAC
ASEL, S, AREA,,131
ASEL, A, AREA,,233
ESLA
*END
*CREATE, Bundle4layer16, MAC
ASEL, S, AREA,,132
ASEL, A, AREA,,234
ESLA
*END
*CREATE, Bundle4layer17, MAC
ASEL, S, AREA,,133
ASEL, A, AREA,,235
ESLA
*END
*CREATE, Bundle4layer18, MAC
ASEL, S, AREA,,134
ASEL, A, AREA,,236
ESLA
*END
*CREATE, Bundle4layer19, MAC
ASEL, S, AREA,,135
ASEL, A, AREA,,237
ESLA
*END

205
*CREATE, Bundle4layer20, MAC
ASEL, S, AREA,,136
ASEL, A, AREA,,238
ESLA
*END
*CREATE, Bundle4layer21, MAC
ASEL, S, AREA,,137
ASEL, A, AREA,,239
ESLA
*END
*CREATE, Bundle4layer22, MAC
ASEL, S, AREA,,138
ASEL, A, AREA,,240
ESLA
*END
*CREATE, Bundle4layer23, MAC
ASEL, S, AREA,,139
ASEL, A, AREA,,241
ESLA
*END
*CREATE, Bundle4layer24, MAC
ASEL, S, AREA,,140
ASEL, A, AREA,,242
ESLA
*END
*CREATE, Bundle4layer25, MAC
ASEL, S, AREA,,141
ASEL, A, AREA,,243
ESLA
*END

206
*CREATE, Bundle4layer26, MAC
ASEL, S, AREA,,142
ASEL, A, AREA,,244
ESLA
*END
*CREATE, Bundle4layer27, MAC
ASEL, S, AREA,,143
ASEL, A, AREA,,168
ESLA
*END
!Bundle 5
*CREATE, Bundle5layer1, MAC
ASEL, S, AREA,,144
ASEL, A, AREA,,146
ESLA
*END
*CREATE, Bundle5layer2, MAC
ASEL, S, AREA,,145
ASEL, A, AREA,,245
ESLA
*END
*CREATE, Bundle5layer3, MAC
ASEL, S, AREA,,147
ASEL, A, AREA,,246
ESLA
*END
*CREATE, Bundle5layer4, MAC
ASEL, S, AREA,,148
ASEL, A, AREA,,247
ESLA

207
*END
*CREATE, Bundle5layer5, MAC
ASEL, S, AREA,,149
ASEL, A, AREA,,248
ESLA
*END
*CREATE, Bundle5layer6, MAC
ASEL, S, AREA,,150
ASEL, A, AREA,,249
ESLA
*END
*CREATE, Bundle5layer7, MAC
ASEL, S, AREA,,151
ASEL, A, AREA,,250
ESLA
*END
*CREATE, Bundle5layer8, MAC
ASEL, S, AREA,,152
ASEL, A, AREA,,251
ESLA
*END

*CREATE, Bundle5layer9, MAC

ASEL, S, AREA,,153
ASEL, A, AREA,,252
ESLA
*END
*CREATE, Bundle5layer10, MAC
ASEL, S, AREA,,154
ASEL, A, AREA,,253

208
ESLA
*END
*CREATE, Bundle5layer11, MAC
ASEL, S, AREA,,155
ASEL, A, AREA,,254
ESLA
*END
*CREATE, Bundle5layer12, MAC
ASEL, S, AREA,,156
ASEL, A, AREA,,255
ESLA
*END
*CREATE, Bundle5layer13, MAC
ASEL, S, AREA,,157
ASEL, A, AREA,,256
ESLA
*END
*CREATE, Bundle5layer14, MAC
ASEL, S, AREA,,158
ASEL, A, AREA,,257
ESLA
*END
*CREATE, Bundle5layer15, MAC
ASEL, S, AREA,,159
ASEL, A, AREA,,258
ESLA
*END
*CREATE, Bundle5layer16, MAC
ASEL, S, AREA,,160
ASEL, A, AREA,,259

209
ESLA
*END
*CREATE, Bundle5layer17, MAC
ASEL, S, AREA,,161
ASEL, A, AREA,,260
ESLA
*END
*CREATE, Bundle5layer18, MAC
ASEL, S, AREA,,162
ASEL, A, AREA,,261
ESLA
*END
*CREATE, Bundle5layer19, MAC
ASEL, S, AREA,,163
ASEL, A, AREA,,262
ESLA
*END
*CREATE, Bundle5layer20, MAC
ASEL, S, AREA,,164
ASEL, A, AREA,,263
ESLA
*END
*CREATE, Bundle5layer21, MAC
ASEL, S, AREA,,165
ASEL, A, AREA,,169
ESLA
*END
!Selecting and adding all mandrel element
*CREATE, MANDREL, MAC
ASEL, S, AREA,,59,59 !Bundle 1 mandrel

210
 ASEL, A, AREA,,264,264 !Bundle 2 mandrel
ASEL, A, AREA,,265,265 !Bundle 3 mandrel
ASEL, A, AREA,,266,266 !Bundle 4 mandrel
ASEL, A, AREA,,267,267 !Bundle 5 mandrel
ESLA
*END
*CREATE, MANDREL_BUNDLE, MAC
BUNDLE
ASEL, A, AREA,,59,59 !Bundle 1 mandrel
ASEL, A, AREA,,264,264 !Bundle 2 mandrel
ASEL, A, AREA,,265,265 !Bundle 3 mandrel
ASEL, A, AREA,,266,266 !Bundle 4 mandrel
ASEL, A, AREA,,267,267 !Bundle 5 mandrel
*END
BUNDLE
AATT,1,,2 !Material assignment to coil and mandrel
MANDREL
AATT,2,,2 !Overwrite mandrel material assignment to
steel or mandrel material
ASEL, S, AREA,,1,4
ASEL, A, AREA,,268,268
APLOT
AATT,3,,1 !Assigning material air and using element
type 1 (AZ DOF)

!Default Element Size

ASEL, ALL
AESIZE, ALL, 0.01
!Coil bundle element size
MANDREL_BUNDLE
APLOT

211
!AESIZE, ALL, 0.001
AESIZE, ALL, 0.005
! Element size near coils
ASEL, S, AREA,,268,268
APLOT
!AESIZE, ALL, 0.009
AESIZE, ALL, 0.02
! Element size second near the coils
ASEL, S, AREA,,1,2
ASEL, S, AREA,,4
APLOT
!AESIZE, ALL, 0.01
AESIZE, ALL, 0.03
ASEL, ALL
MSHK, 2
MSHA, 0,2D
AMESH, ALL
!ASEL, S, AREA,,268
!APLOT
!AREFINE,ALL,,,2
!Mandrel and bundle refinement
!Coil bundle element size
MANDREL_BUNDLE
APLOT
AREFINE,ALL,,,2
MANDREL
AREFINE, ALL,,,2
!BUNDLE 1
NSEL, S, LOC, Y, Z1_l-tol,Z1_l+tol
NSEL, R, LOC, X, R1_i,R1_o

212
NPLOT
!CP, tagcp, UY, ALL
!tagcp = tagcp+1
NSEL, S, LOC, Y, Z1_h-tol,Z1_h+tol
NSEL, R, LOC, X, R1_i,R1_o
NPLOT
!CP, tagcp, UY, ALL
!tagcp = tagcp+1
!BUNDLE 2
NSEL, S, LOC, Y, Z2_l-tol,Z2_l+tol
NSEL, R, LOC, X, R2_i,R2_o
NPLOT
!CP, tagcp, UY, ALL
!tagcp = tagcp+1
NSEL, S, LOC, Y, Z2_h-tol,Z2_h+tol
NSEL, R, LOC, X, R2_i,R2_o
NPLOT
!CP, tagcp, UY, ALL
!tagcp = tagcp+1
!BUNDLE 3
NSEL, S, LOC, Y, Z3_l-tol,Z3_l+tol
NSEL, R, LOC, X, R3_i,R3_o
NPLOT
!CP, tagcp, UY, ALL
!tagcp = tagcp+1
NSEL, S, LOC, Y, Z3_h-tol,Z3_h+tol
NSEL, R, LOC, X, R3_i,R3_o
NPLOT
!CP, tagcp, UY, ALL
!tagcp = tagcp+1

213
!BUNDLE 4
NSEL, S, LOC, Y, Z4_l-tol,Z4_l+tol
NSEL, R, LOC, X, R4_i,R4_o
NPLOT
!CP, tagcp, UY, ALL
!tagcp = tagcp+1
NSEL, S, LOC, Y, Z4_h-tol,Z4_h+tol
NSEL, R, LOC, X, R4_i,R4_o
NPLOT
!CP, tagcp, UY, ALL
!tagcp = tagcp+1
!BUNDLE 5
NSEL, S, LOC, Y, Z5_l-tol,Z5_l+tol
NSEL, R, LOC, X, R5_i,R5_o
NPLOT
!CP, tagcp, UY, ALL
!tagcp = tagcp+1

NSEL, S, LOC, Y, Z5_h-tol,Z5_h+tol

NSEL, R, LOC, X, R5_i,R5_o
NPLOT
!CP, tagcp, UY, ALL
!tagcp = tagcp+1
NSEL, S, LOC, X, 3-tol,3+tol
NPLOT
!CP, tagcp, AZ, ALL
!tagcp = tagcp+1
allsel, all
cdwrite, comb, EMSTRESS, cdb
!cdread,comb,EMSTRESS,cdb

214
FINISH
/prep7
ET,1,PLANE13,,,1 ! Plane13, AZ DOF, AXISYMMETRIC, FOR
AIR
ET,2,PLANE13,,,1
FINISH
/SOLU
/NERR,,99999999,,0,0
ANTYPE, 0
!CNVTOL, F, ,1e-7
CNVTOL, CSG, ,1e-7
!Flux Parallel BC
LSEL, S, LOC, X,3,3
LPLOT
DL, ALL,,AX,0
!Assigning the current density
BUNDLE
APLOT
ASEL, U, AREA,,144,145
ASEL, U, AREA,,147,165
ASEL, U, AREA,,146,146 !Bundle 5 upper layers
ASEL, U, AREA,,245,263
ASEL, U, AREA,,169,169
APLOT
ESLA
EPLOT
BFE, ALL, JS,,,,118500000

ASEL, S, AREA,,144,145 !Bundle 5 lower layers

215
ASEL, A, AREA,,147,165
ASEL, A, AREA,,146,146 !Bundle 5 upper layers
ASEL, A, AREA,,245,263
ASEL, A, AREA,,169,169
APLOT
ESLA
EPLOT
BFE, ALL, JS,,,,-118500000
ALLSEL, ALL
SAVE
SOLVE
FINISH
/filename, EMSTRESS_MAG,1
save
/filename, EMSTRESS,1
/PREP7
ET,2,PLANE182,,,1
/SOLU

/NERR,,99999999,,0,0
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
!Define Boundary Condition!!!!!!!!!!!!!!!!!!!
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
!fixing the midplane of each bundle Y-directional
displacement to zero
!BUNDLE 1
NSEL, S, LOC, Y, ((Z1_l+Z1_h)/2)-tol,((Z1_l+Z1_h)/2)+tol
NSEL, R, LOC, X, R1_i-Mthick,R1_o
!NPLOT
D, ALL, UY, 0

216
!BUNDLE 2
NSEL, S, LOC, Y, ((Z2_l+Z2_h)/2)-tol,((Z2_l+Z2_h)/2)+tol
NSEL, R, LOC, X, R2_i-Mthick,R2_o
!NPLOT
D, ALL, UY, 0
!BUNDLE 3
NSEL, S, LOC, Y, ((Z3_l+Z3_h)/2)-tol,((Z3_l+Z3_h)/2)+tol
NSEL, R, LOC, X, R3_i-Mthick,R3_o
!NPLOT
D, ALL, UY, 0

!BUNDLE 4
NSEL, S, LOC, Y, ((Z4_l+Z4_h)/2)-tol,((Z4_l+Z4_h)/2)+tol
NSEL, R, LOC, X, R4_i-Mthick,R4_o
!NPLOT
D, ALL, UY, 0
!BUNDLE 5
NSEL, S, LOC, Y, ((Z5_l+Z5_h)/2)-tol,((Z5_l+Z5_h)/2)+tol
NSEL, R, LOC, X, R5_i-Mthick,R5_o
!NPLOT
D, ALL, UY, 0
ANTYPE, 0
TIME2
NLGEOM, ON
NROPT, FULL
MANDREL_BUNDLE
ESLA
inis,write,1,,,,,epel
save
DELTIM, 0.5

217
NSUBST, 2
solve
Mandrel_BUNDLE
APLOT
ESLA
EKILL, ALL
e_x = -6.66396000000000e-05
e_y = -6.42872844000000e-05
e_z = 0.000203655600000000
/NERR,,99999999,,0,0
MANDREL
ESLA
EALIVE, ALL
FINISH
*DO, j,1, Bundle_1
/SOLU
/DELET, EMSTRESS,rst
Time, j+3
DELTIM, 0.5
NSUBST, 2

inis,read,EMSTRESS,ist
*DO, i,1,4 !4
layer = 'Bundle%i%layer%j%'
layer
ESLA
EALIVE, ALL
inistate,set,dtyp,epel
inistate,defi,,,,,e_x,e_y,e_z
*ENDDO

218
 *IF, j, LE,21, THEN
layer = 'Bundle5layer%j%'
layer
ESLA
EALIVE, ALL
inistate,set,dtyp,epel
inistate,defi,,,,,e_x,e_y,e_z
*ENDIF
MANDREL_BUNDLE
ESLA
inis,write,1,,,,,epel
SAVE
SOLVE

FINISH

*ENDDO
FINISH
/filename, EMSTRESS_WINDING,1
save
/filename, EMSTRESS,1
/SOLU
/NERR,,99999999,,0,0

!Thermal Cooling
TIME, j+4
MANDREL_BUNDLE
ESLA
TUNIF, -288
SAVE

219
SOLVE
FINISH
/filename, EMSTRESS_cooling,1
save
/filename, EMSTRESS,1
/solu
/NERR,,99999999,,0,0

!!!!!!Electromag
TIME, j+5
ALLSEL, ALL
LDREAD,FORC,,,,,,rmg,,
MANDREL_BUNDLE
ESLA
EPLOT
SAVE
SOLVE
FINISH

/filename, EMSTRESS,1
save

/post1
PATH, COIL1, 2,50,50
PPATH,1,,(R1_i-MTHICK),(Z1_h+Z1_l)/2,0
PPATH,2,,R1_o,(Z1_h+Z1_l)/2,0
PDEF,S_X,S,X
PDEF,S_y,S,Y
PDEF,S_z,S,Z
PDEF,S_XY,S,XY

220
PDEF,S_YZ,S,YZ
PDEF,S_XZ,S,XZ
PDEF,S_1,S,1
PDEF,S_3,S,3
PDEF,S_vm,S,EQV
PDEF, e_1st_th,EPTH,1
PDEF, e_z_th,EPTH,z
PDEF,e_vm_th,EPTH,EQV
PDEF,e_r_tot,EPTO,x
PDEF,e_y_tot,EPTO,y
PDEF,e_z_tot,EPTO,z
PDEF,e_xy_tot,EPTO,xy
PDEF,e_yz_tot,EPTO,yz
PDEF,e_xz_tot,EPTO,xz
PDEF,e_1st_tot,EPTO,1
PDEF,e_2nd_tot,EPTO,2
PDEF,e_3rd_tot,EPTO,3
PDEF,e_vm_tot,EPTO,EQV
PDEF,e_in_tot,EPTO,INT
PRPATH, S_X, S_Z, S_1, S_3, S_VM
PRPATH, e_r_tot, e_1st_tot, e_3rd_to, e_vm_tot, e_z_tot

PATH, COIL2, 2,50,50

PPATH,1,,(R2_i-MTHICK),(Z2_h+Z2_l)/2,0
PPATH,2,,R2_o,(Z2_h+Z2_l)/2,0
PDEF,S_X,S,X
PDEF,S_z,S,Z
PDEF,S_1,S,1
PDEF,S_3,S,3
PDEF,S_vm,S,EQV

221
PDEF, e_1st_th,EPTH,1
PDEF, e_z_th,EPTH,z
PDEF,e_vm_th,EPTH,EQV
PDEF,e_r_tot,EPTO,x
PDEF,e_1st_tot,EPTO,1
PDEF,e_3rd_tot,EPTO,3
PDEF,e_vm_tot,EPTO,EQV
PDEF,e_z_tot,EPTO,z
PLPATH, e_1st_tot, e_1st_th
PRPATH, S_X, S_Z, S_1, S_3, S_VM
PRPATH, e_r_tot, e_1st_tot, e_3rd_to, e_vm_tot, e_z_tot

PATH, COIL3, 2,50,50

PPATH,1,,R3_i-MTHICK,(Z3_h+Z3_l)/2,0
PPATH,2,,R3_o,(Z3_h+Z3_l)/2,0
PDEF,S_X,S,X
PDEF,S_z,S,Z
PDEF,S_1,S,1
PDEF,S_3,S,3
PDEF,S_vm,S,EQV
PDEF, e_1st_th,EPTH,1
PDEF, e_z_th,EPTH,z
PDEF,e_vm_th,EPTH,EQV
PDEF,e_r_tot,EPTO,x
PDEF,e_1st_tot,EPTO,1
PDEF,e_3rd_tot,EPTO,3
PDEF,e_vm_tot,EPTO,EQV
PDEF,e_z_tot,EPTO,z
PLPATH, e_1st_tot, e_1st_th
PRPATH, S_X, S_Z, S_1, S_3, S_VM

222
PRPATH, e_r_tot, e_1st_tot, e_3rd_to, e_vm_tot, e_z_tot

PATH, COIL4, 2,50,50

PPATH,1,,R4_i-MTHICK,(Z4_h+Z4_l)/2,0
PPATH,2,,R4_o,(Z4_h+Z4_l)/2,0
PDEF,S_X,S,X
PDEF,S_z,S,Z
PDEF,S_1,S,1
PDEF,S_3,S,3
PDEF,S_vm,S,EQV
PDEF, e_1st_th,EPTH,1
PDEF, e_z_th,EPTH,z
PDEF,e_vm_th,EPTH,EQV
PDEF,e_r_tot,EPTO,x
PDEF,e_1st_tot,EPTO,1
PDEF,e_3rd_tot,EPTO,3
PDEF,e_vm_tot,EPTO,EQV
PDEF,e_z_tot,EPTO,z
PLPATH, e_1st_tot, e_1st_th
PRPATH, S_X, S_Z, S_1, S_3, S_VM
PRPATH, e_r_tot, e_1st_tot, e_3rd_to, e_vm_tot, e_z_tot

PATH, COIL5, 2,50,50

PPATH,1,,R5_i-MTHICK,(Z5_h+Z5_l)/2,0
PPATH,2,,R5_o,(Z5_h+Z5_l)/2,0
PDEF,S_X,S,X
PDEF,S_z,S,Z
PDEF,S_1,S,1
PDEF,S_3,S,3
PDEF,S_vm,S,EQV

223
PDEF, e_1st_th,EPTH,1
PDEF, e_z_th,EPTH,z
PDEF,e_vm_th,EPTH,EQV
PDEF,e_r_tot,EPTO,x
PDEF,e_1st_tot,EPTO,1
PDEF,e_3rd_tot,EPTO,3
PDEF,e_vm_tot,EPTO,EQV
PDEF,e_z_tot,EPTO,z
PLPATH, e_1st_tot, e_1st_th
PRPATH, S_X, S_Z, S_1, S_3, S_VM
PRPATH, e_r_tot, e_1st_tot, e_3rd_to, e_vm_tot, e_z_tot

224
 References

[1] R. W. Brown, Y.-C. N. Cheng, E. M. Haacke, M. R. Thompson, and R. Venkatesan,

Magnetic resonance imaging: physical principles and sequence design, Second
edition. Hoboken, New Jersey: John Wiley & Sons, Inc, 2014.
[2] “MRI units density by country 2015 | Statistic,” Statista, 02-May-2017. [Online].
Available: https://www.statista.com/statistics/282401/density-of-magnetic-
resonance-imaging-units-by-country/. [Accessed: 02-May-2017].
[3] B. Hüsing, L. Jäncke, and B. Tag, Impact Assessment of Neuroimaging: Final Report.
vdf Hochschulverlag AG, 2006.
[4] M. Parizh, Y. Lvovsky, and M. Sumption, “Conductors for commercial MRI magnets
beyond NbTi: requirements and challenges,” Supercond. Sci. Technol., vol. 30, no.
1, p. 014007, 2017.
[5] A. Gengler, “How to Get the Same Health Care at a Quarter of the Cost,” Money,
02-May-2017. [Online]. Available: http://time.com/money/2993644/why-health-
care-costs-vary-widely/. [Accessed: 02-May-2017].
[6] T. G. Berlincourt, “Pulsed-Magnetic-Field Studies of Superconducting Transition
Metal Alloys at High and Low Current Densities,” Atomics International. Div. of
North American Aviation, Inc., Canoga Park, Calif., 1962.
[7] J. E. Kunzler, E. Buehler, F. S. L. Hsu, and J. H. Wernick, “Superconductivity in
${\mathrm{Nb}}_{3}$Sn at High Current Density in a Magnetic Field of 88 kgauss,”
Phys. Rev. Lett., vol. 6, no. 3, pp. 89–91, Feb. 1961.
[8] J. S. Dugdale and F. E. Simon, “Thermodynamic Properties and Melting of Solid
Helium,” Proc. R. Soc. Lond. Math. Phys. Eng. Sci., vol. 218, no. 1134, pp. 291–310,
Jul. 1953.
[9] J. Ling, “A persistent-mode MgB₂ 0.5-T/240-mm solid-nitrogen-cooled magnet for
MRI,” Massachusetts Institute of Technology, 2016.
[10] J. E. Hamak, “U.S. Geological Survey, Mineral Commodity Summaries,” USGS
Mineral Information, Jan-2017. [Online]. Available:
https://minerals.usgs.gov/minerals/pubs/commodity/helium/mcs-2017-heliu.pdf.
[Accessed: 02-May-2017].
[11] Y. Lvovsky and P. Jarvis, “Superconducting systems for MRI-present solutions and
new trends,” IEEE Trans. Appl. Supercond., vol. 15, no. 2, pp. 1317–1325, Jun. 2005.
[12] Z. M. Wang, “Superconducting magnet of magnetic resonance imaging system,” in
2013 IEEE International Conference on Applied Superconductivity and
Electromagnetic Devices (ASEMD), 2013, pp. 530–533.
[13] C. Spencer, P. A. Sanger, and M. Young, “The temperature and magnetic field
dependence of superconducting critical current densities of multifilamentary
Nb3Sn and NbTi composite wires,” IEEE Trans. Magn., vol. 15, no. 1, pp. 76–79, Jan.
1979.
[14] Y. Lvovsky, E. W. Stautner, and T. Zhang, “Novel technologies and configurations of
superconducting magnets for MRI,” Supercond. Sci. Technol., vol. 26, no. 9, p.
093001, Sep. 2013.

225
[15] Z. Cai, R. H. Clarke, B. A. Glowacki, W. J. Nuttall, and N. Ward, “Ongoing ascent to
the helium production plateau—Insights from system dynamics,” Resour. Policy,
vol. 35, no. 2, pp. 77–89, 2010.
[16] J. Ling, “Monofilament MgB 2 Wires for MRI Magnets,” Massachusetts Institute of
Technology, 2012.
[17] D. Patel et al., “Solid cryogen: a cooling system for future MgB2 MRI magnet,” Sci.
Rep., vol. 7, p. 43444, Mar. 2017.
[18] H. Morita et al., “10 T conduction cooled Bi-2212/Ag HTS solenoid magnet system,”
IEEE Trans. Appl. Supercond., vol. 11, no. 1, pp. 2523–2526, Mar. 2001.
[19] B. J. Parkinson, R. Slade, M. J. D. Mallett, and V. Chamritski, “Development of a
Cryogen Free 1.5 T YBCO HTS Magnet for MRI,” IEEE Trans. Appl. Supercond., vol.
23, no. 3, pp. 4400405–4400405, Jun. 2013.
[20] T. Baig, Z. Yao, D. Doll, M. Tomsic, and M. Martens, “Conduction cooled magnet
design for 1.5 T, 3.0 T and 7.0 T MRI systems,” Supercond. Sci. Technol., vol. 27, no.
12, p. 125012, Dec. 2014.
[21] H. Miyazaki et al., “Development of a 5.1 T conduction-cooled YBCO coil composed
of a stack of 12 single pancakes,” Phys. C Supercond., vol. 484, pp. 287–291, Jan.
2013.
[22] M. A. Young, J. A. Demko, M. J. Gouge, M. O. Pace, J. W. Lue, and R. Grabovickic,
“Measurements of the performance of BSCCO HTS tape under magnetic fields with
a cryocooled test rig,” IEEE Trans. Appl. Supercond., vol. 13, no. 2, pp. 2964–2967,
Jun. 2003.
[23] Q. Wang et al., “Design and Fabrication of a Conduction-Cooled High Temperature
Superconducting Magnet for 10 kJ Superconducting Magnetic Energy Storage
System,” IEEE Trans. Appl. Supercond., vol. 16, no. 2, pp. 570–573, Jun. 2006.
[24] Z. Melhem, High Temperature Superconductors (HTS) for Energy Applications.
Elsevier, 2011.
[25] T. C. Cosmus and M. Parizh, “Advances in Whole-Body MRI Magnets,” IEEE Trans.
Appl. Supercond., vol. 21, no. 3, pp. 2104–2109, Jun. 2011.
[26] M. A. Green and B. P. Strauss, “Things to Think About When Estimating the Cost of
Magnets Made With Conductors Other Than Nb-Ti,” IEEE Trans. Appl. Supercond.,
vol. 27, no. 4, pp. 1–5, Jun. 2017.
[27] M. A. Green and S. S. Chouhan, “A Cyclotron Magnet Case Study: Would Replacing
the LTS Coils With HTS Coils Make Sense?,” IEEE Trans. Appl. Supercond., vol. 26,
no. 3, pp. 1–5, Apr. 2016.
[28] Y. S. Choi, D. L. Kim, B. S. Lee, H. S. Yang, and T. A. Painter, “Conduction-Cooled
Superconducting Magnet for Material Control Application,” IEEE Trans. Appl.
Supercond., vol. 19, no. 3, pp. 2190–2193, Jun. 2009.
[29] S. L. Bud’ko, G. Lapertot, C. Petrovic, C. E. Cunningham, N. Anderson, and P. C.
Canfield, “Boron Isotope Effect in Superconducting MgB2,” Phys. Rev. Lett., vol. 86,
no. 9, pp. 1877–1880, Feb. 2001.
[30] K. Vinod, R. G. A. Kumar, and U. Syamaprasad, “Prospects for MgB2
superconductors for magnet application,” Supercond. Sci. Technol., vol. 20, no. 1,
p. R1, Jan. 2007.

226
[31] M. Modica et al., “Design, construction and tests of MgB2 coils for the development
of a cryogen free magnet,” IEEE Trans. Appl. Supercond., vol. 17, no. 2, pp. 2196–
2199, 2007.
[32] W. Yao, J. Bascuñán, W.-S. Kim, S. Hahn, H. Lee, and Y. Iwasa, “A Solid Nitrogen
Cooled $${$$\backslash$ hbox ${$MgB$}$$}$ _ ${$2$}$ $‘Demonstration’ Coil for
MRI Applications,” IEEE Trans. Appl. Supercond., vol. 18, no. 2, pp. 912–915, 2008.
[33] S. Mine et al., “Second Test Coil for the Development of a Compact 3 T Magnet,”
IEEE Trans. Appl. Supercond., vol. 23, no. 3, pp. 4601404–4601404, Jun. 2013.
[34] J. Ling et al., “Monofilament $$\backslash$ hbox ${$MgB$}$ _ ${$2$}$ $ Wire for a
Whole-Body MRI Magnet: Superconducting Joints and Test Coils,” IEEE Trans. Appl.
Supercond., vol. 23, no. 3, pp. 6200304–6200304, 2013.
[35] R. Penco and G. Grasso, “Recent Development of MgB2-Based Large Scale
Applications,” IEEE Trans. Appl. Supercond., vol. 17, no. 2, pp. 2291–2294, Jun.
2007.
[36] M. Tomsic et al., “Overview of MgB2 Superconductor Applications,” Int. J. Appl.
Ceram. Technol., vol. 4, no. 3, pp. 250–259, Jul. 2007.
[37] P. Kováč, L. Kopera, T. Melišek, M. Rindfleisch, W. Haessler, and I. Hušek,
“Behaviour of filamentary MgB 2 wires subjected to tensile stress at 4.2 K,”
Supercond. Sci. Technol., vol. 26, no. 10, p. 105028, Oct. 2013.
[38] M. Dhallé et al., “Scaling the reversible strain response of MgB2 conductors,”
Supercond. Sci. Technol., vol. 18, no. 12, p. S253, Dec. 2005.
[39] G. Z. Li, Y. Yang, M. A. Susner, M. D. Sumption, and E. W. Collings, “Critical current
densities and n-values of MgB2 strands over a wide range of temperatures and
fields,” Supercond. Sci. Technol., vol. 25, no. 2, p. 025001, Feb. 2012.
[40] G. Z. Li et al., “The critical current density of advanced internal-Mg-diffusion-
processed MgB2 wires,” Supercond. Sci. Technol., vol. 25, no. 11, p. 115023, Nov.
2012.
[41] A. Yamamoto, J. Shimoyama, S. Ueda, Y. Katsura, S. Horii, and K. Kishio, “Improved
critical current properties observed in MgB2 bulks synthesized by low-temperature
solid-state reaction,” Supercond. Sci. Technol., vol. 18, no. 1, p. 116, Jan. 2005.
[42] Y. Ma, X. Zhang, G. Nishijima, K. Watanabe, and X. Bai, “Significantly enhanced
critical current densities in MgB2 tapes made by a scaleable, nano-carbon addition
route,” Appl. Phys. Lett., vol. 88, no. 7, p. 072502, 2006.
[43] S. X. Dou et al., “Enhancement of the critical current density and flux pinning of
MgB2 superconductor by nanoparticle SiC doping,” Appl. Phys. Lett., vol. 81, no. 18,
p. 3419, 2002.
[44] B. A. Glowacki, M. Majoros, M. Vickers, J. E. Evetts, Y. Shi, and I. McDougall,
“Superconductivity of powder-in-tube MgB 2 wires,” Supercond. Sci. Technol., vol.
14, no. 4, p. 193, 2001.
[45] “Large transport critical currents in unsintered MgB2 superconducting tapes,” Appl.
Phys. Lett., vol. 79, no. 2, pp. 230–232, Jul. 2001.
[46] A. Stenvall et al., “A checklist for designers of cryogen-free MgB2 coils,” Supercond.
Sci. Technol., vol. 20, no. 4, p. 386, 2007.

227
[47] D. Zhang et al., “Research on Stability of $\rm MgB_2$ Superconducting Magnet
for MRI,” IEEE Trans. Appl. Supercond., vol. 21, no. 3, pp. 2100–2103, Jun. 2011.
[48] D. Zhang et al., “Research on $${$$\backslash$ rm MgB$}$ _ ${$2$}$ $
Superconducting Magnet With Iron Core for MRI,” IEEE Trans. Appl. Supercond.,
vol. 20, no. 3, pp. 764–768, 2010.
[49] X. H. Li, X. J. Du, Q. Bao, L. Q. Kong, L. Y. Ye, and L. Y. Xiao, “Design, development
and experiment of a 1.5 T MgB 2 superconducting test magnet,” Cryogenics, vol.
49, no. 6, pp. 286–290, 2009.
[50] X. Li et al., “A Small 1.5 T Persistent Current Operating Test Magnet Using
$${$$\backslash$ rm MgB$}$ _ ${$2$}$ $ Wire With High $ j_ ${$$\backslash$ rm
c$}$ $ Joints,” IEEE Trans. Appl. Supercond., vol. 21, no. 3, pp. 1616–1619, 2011.
[51] S. Mine et al., “Development of a Compact 3 T $$\backslash$ hbox ${$MgB$}$ _
${$2$}$ $ Magnet,” IEEE Trans. Appl. Supercond., vol. 24, no. 3, pp. 1–4, 2014.
[52] S. Mine et al., “Development of a 3 T–250 mm Bore $$\backslash$ hbox ${$MgB$}$
_ ${$2$}$ $ Magnet System,” IEEE Trans. Appl. Supercond., vol. 25, no. 3, pp. 1–4,
2015.
[53] M. Razeti et al., “Construction and Operation of Cryogen Free Magnets for Open
MRI Systems,” IEEE Trans. Appl. Supercond., vol. 18, no. 2, pp. 882–886, Jun. 2008.
[54] M. Modica et al., “Design, Construction and Tests of MgB2 Coils for the
Development of a Cryogen Free Magnet,” IEEE Trans. Appl. Supercond., vol. 17, no.
2, pp. 2196–2199, Jun. 2007.
[55] W. Yao, J. Bascuñán, S. Hahn, and Y. Iwasa, “A Superconducting Joint Technique for
$${$$\backslash$ rm MgB$}$ _ ${$2$}$ $ Round Wires,” IEEE Trans. Appl.
Supercond., vol. 19, no. 3, pp. 2261–2264, 2009.
[56] D. K. Park et al., “$${$$\backslash$ rm MgB$}$ _ ${$2$}$ $ for MRI Magnets: Test
Coils and Superconducting Joints Results,” IEEE Trans. Appl. Supercond., vol. 22, no.
3, pp. 4400305–4400305, 2012.
[57] P. Kováč, L. Kopera, T. Melišek, M. Rindfleisch, W. Haessler, and I. Hušek,
“Behaviour of filamentary MgB 2 wires subjected to tensile stress at 4.2 K,”
Supercond. Sci. Technol., vol. 26, no. 10, p. 105028, Oct. 2013.
[58] H. Kitaguchi and H. Kumakura, “Superconducting and mechanical performance and
the strain effects of a multifilamentary MgB2/Ni tape,” Supercond. Sci. Technol., vol.
18, no. 12, p. S284, Dec. 2005.
[59] K. Katagiri et al., “Stress–strain effects on powder-in-tube MgB2 tapes and wires,”
Supercond. Sci. Technol., vol. 18, no. 12, p. S351, Dec. 2005.
[60] M. Guan, X. Wang, L. Ma, Y. Zhou, and C. Xin, “Magneto-Mechanical Coupling
Analysis of a Superconducting Solenoid Magnet in Self-Magnetic Field,” IEEE Trans.
Appl. Supercond., vol. 24, no. 3, pp. 1–4, Jun. 2014.
[61] T. Schild et al., “The Iseult/Inumac Whole Body 11.7 T MRI Magnet Design,” IEEE
Trans. Appl. Supercond., vol. 18, no. 2, pp. 904–907, Jun. 2008.
[62] L. Li, J. Cheng, Z. Ni, H. Wang, Y. Dai, and Q. Wang, “Preliminary Mechanical Analysis
of a 9.4-T Whole-Body MRI Magnet,” IEEE Trans. Appl. Supercond., vol. 25, no. 2,
pp. 1–7, Apr. 2015.

228
[63] X. Jiang et al., “A 7-T Actively Shielded Superconducting Magnet for Animal MRI,”
IEEE Trans. Appl. Supercond., vol. 23, no. 3, pp. 4400604–4400604, Jun. 2013.
[64] Q. Wang et al., “A superconducting magnet system for whole-body metabolism
imaging,” Appl. Supercond. IEEE Trans. On, vol. 22, no. 3, pp. 4400905–4400905,
2012.
[65] Y. Dai et al., “Structural Design of a 9.4 T Whole-Body MRI Superconducting
Magnet,” IEEE Trans. Appl. Supercond., vol. 22, no. 3, pp. 4900404–4900404, Jun.
2012.
[66] J. Chen and X. Jiang, “Stress Analysis of a 7 T Actively Shielded Superconducting
Magnet for Animal MRI,” IEEE Trans. Appl. Supercond., vol. 22, no. 3, pp. 4903104–
4903104, Jun. 2012.
[67] L. Li, Z. Ni, J. Cheng, H. Wang, Q. Wang, and B. Zhao, “Effect of Pretension, Support
Condition, and Cool Down on Mechanical Disturbance of Superconducting Coils,”
IEEE Trans. Appl. Supercond., vol. 22, no. 2, pp. 3800104–3800104, Apr. 2012.
[68] F. Nunio et al., “Mechanical Design of the Iseult 11.7 T Whole Body MRI Magnet,”
IEEE Trans. Appl. Supercond., vol. 20, no. 3, pp. 760–763, Jun. 2010.
[69] H. K. Onnes, “Further experiments with liquid helium,” in Proceedings of the KNAW,
1911, vol. 13, pp. 1910–1911.
[70] P. L. Kapitza, “A Method of Producing Strong Magnetic Fields,” Proc. R. Soc. Lond.
Math. Phys. Eng. Sci., vol. 105, no. 734, pp. 691–710, Jun. 1924.
[71] P. Kapitza, “Further Developments of the Method of Obtaining Strong Magnetic
Fields,” Proc. R. Soc. Lond. Math. Phys. Eng. Sci., vol. 115, no. 772, pp. 658–683,
Aug. 1927.
[72] J. D. Cockcroft, “The design of coils for the production of strong magnetic fields,”
Philos. Trans. R. Soc. Lond. Ser. Contain. Pap. Math. Phys. Character, vol. 227, pp.
317–343, 1928.
[73] G. B. Yntema, “Superconducting winding for electromagnet,” in Physical Review,
1955, vol. 98, pp. 1197–1197.
[74] E. Buehler and H. J. Levinstein, “Effect of Tensile Stress on the Transition
Temperature and Current‐Carrying Capacity of Nb3Sn,” J. Appl. Phys., vol. 36, no.
12, pp. 3856–3860, Dec. 1965.
[75] E. Sanchez-Palencia, Non-Homogeneous Media and Vibration Theory. Springer,
1730.
[76] M. Lefik and B. Schrefler, “FE modelling of a boundary layer corrector for
composites using the homogenization theory,” Eng. Comput., vol. 13, no. 6, pp. 31–
42, 1996.
[77] M. Kamiński and B. A. Schrefler, “Probabilistic effective characteristics of cables for
superconducting coils,” Comput. Methods Appl. Mech. Eng., vol. 188, no. 1–3, pp.
1–16, Jul. 2000.
[78] M. Lefik and B. A. Schrefler, “Application of the homogenisation method to the
analysis of superconducting coils,” Fusion Eng. Des., vol. 24, no. 3, pp. 231–255,
Jun. 1994.

229
[79] D. P. Boso, M. Lefik, and B. A. Schrefler, “Multiscale analysis of the influence of the
triplet helicoidal geometry on the strain state of a Nb3Sn based strand for ITER
coils,” Cryogenics, vol. 45, no. 9, pp. 589–605, Sep. 2005.
[80] D. P. Boso, M. Lefik, and B. A. Schrefler, “A multilevel homogenised model for
superconducting strand thermomechanics,” Cryogenics, vol. 45, no. 4, pp. 259–
271, Apr. 2005.
[81] D. P. Boso, M. Lefik, and B. A. Schrefler, “Homogenisation methods for the thermo-
mechanical analysis of Nb3Sn strand,” Cryogenics, vol. 46, no. 7–8, pp. 569–580,
Jul. 2006.
[82] W. Voigt, “Uberdie Beziehung Zwischen den Beiden Elastizitatskonstanten
Isotroper Korper,” Wied Ann, vol. 38, no. 2, p. 573, 1889.
[83] A. Reuss, “Berechnung der Fließgrenze von Mischkristallen auf Grund der
Plastizitätsbedingung für Einkristalle .,” ZAMM - J. Appl. Math. Mech. Z. Für Angew.
Math. Mech., vol. 9, no. 1, pp. 49–58, Jan. 1929.
[84] R. Hill, “The Elastic Behaviour of a Crystalline Aggregate,” Proc. Phys. Soc. Sect. A,
vol. 65, no. 5, p. 349, 1952.
[85] B. Paul, “Prediction of elastic constants of multi-phase materials,” DTIC Document,
1959.
[86] Z. Hashin and S. Shtrikman, “On some variational principles in anisotropic and
nonhomogeneous elasticity,” J. Mech. Phys. Solids, vol. 10, no. 4, pp. 335–342, Oct.
1962.
[87] Z. Hashin and S. Shtrikman, “A variational approach to the theory of the elastic
behaviour of polycrystals,” J. Mech. Phys. Solids, vol. 10, no. 4, pp. 343–352, Oct.
1962.
[88] Z. Hashin and S. Shtrikman, “A variational approach to the theory of the elastic
behaviour of multiphase materials,” J. Mech. Phys. Solids, vol. 11, no. 2, pp. 127–
140, Mar. 1963.
[89] J. F. W. Bishop and R. Hill, “XLVI. A theory of the plastic distortion of a
polycrystalline aggregate under combined stresses.,” Lond. Edinb. Dublin Philos.
Mag. J. Sci., vol. 42, no. 327, pp. 414–427, 1951.
[90] J. F. W. Bishop and R. Hill, “CXXVIII. A theoretical derivation of the plastic properties
of a polycrystalline face-centred metal,” Lond. Edinb. Dublin Philos. Mag. J. Sci., vol.
42, no. 334, pp. 1298–1307, Nov. 1951.
[91] G. I. Taylor and H. Quinney, “The plastic distortion of metals,” Philos. Trans. R. Soc.
Lond. Ser. Contain. Pap. Math. Phys. Character, vol. 230, pp. 323–362, 1932.
[92] S. G. I. Taylor, Plastic Strain in metals. 1938.
[93] Z. Hashin, “Assessment of the Self Consistent Scheme Approximation: Conductivity
of Particulate Composites,” J. Compos. Mater., vol. 2, no. 3, pp. 284–300, Jul. 1968.
[94] A. V. Hershey, “The elasticity of an isotropic aggregate of anisotropic cubic
crystals,” J. Appl. Mech.-Trans. ASME, vol. 21, no. 3, pp. 236–240, 1954.
[95] R. Hill, “A self-consistent mechanics of composite materials,” J. Mech. Phys. Solids,
vol. 13, no. 4, pp. 213–222, Aug. 1965.
[96] B. Budiansky, “On the elastic moduli of some heterogeneous materials,” J. Mech.
Phys. Solids, vol. 13, no. 4, pp. 223–227, Aug. 1965.

230
[97] R. Hill, “Continuum micro-mechanics of elastoplastic polycrystals,” J. Mech. Phys.
Solids, vol. 13, no. 2, pp. 89–101, Apr. 1965.
[98] J. W. Hutchinson, “Bounds and Self-Consistent Estimates for Creep of
Polycrystalline Materials,” Proc. R. Soc. Lond. Math. Phys. Eng. Sci., vol. 348, no.
1652, pp. 101–127, Feb. 1976.
[99] M. Berveiller and A. Zaoui, “An extension of the self-consistent scheme to
plastically-flowing polycrystals,” J. Mech. Phys. Solids, vol. 26, no. 5, pp. 325–344,
Oct. 1978.
[100] J. R. Willis, “The Overall Elastic Response of Composite Materials,” J. Appl. Mech.,
vol. 50, no. 4b, pp. 1202–1209, Dec. 1983.
[101] J. R. Willis, “Variational Estimates for the Overall Response of an Inhomogeneous
Nonlinear Dielectric,” in Homogenization and Effective Moduli of Materials and
Media, Springer, New York, NY, 1986, pp. 247–263.
[102] D. R. S. Talbot and J. R. Willis, “Variational Principles for Inhomogeneous Non-linear
Media,” IMA J. Appl. Math., vol. 35, no. 1, pp. 39–54, Jul. 1985.
[103] P. P. Castañeda, “New variational principles in plasticity and their application to
composite materials,” J. Mech. Phys. Solids, vol. 40, no. 8, pp. 1757–1788, Nov.
1992.
[104] P. P. Castañeda, “The effective mechanical properties of nonlinear isotropic
composites,” J. Mech. Phys. Solids, vol. 39, no. 1, pp. 45–71, Jan. 1991.
[105] D. R. S. Talbot and J. R. Willis, “Some simple explicit bounds for the overall
behaviour of nonlinear composites,” Int. J. Solids Struct., vol. 29, no. 14, pp. 1981–
1987, Jan. 1992.
[106] P. M. Suquet, “Overall potentials and extremal surfaces of power law or ideally
plastic composites,” J. Mech. Phys. Solids, vol. 41, no. 6, pp. 981–1002, Jun. 1993.
[107] T. Olson, “Improvements on Taylor’s upper bound for rigid-plastic composites,”
Mater. Sci. Eng. A, vol. 175, no. 1, pp. 15–20, Feb. 1994.
[108] E. J. Barbero, Finite Element Analysis of Composite Materials Using ANSYS®, Second
Edition. CRC Press, 2013.
[109] E. J. Barbero, Introduction to composite materials design. CRC press, 2010.
[110] R. Luciano and E. Sacco, “Variational methods for the homogenization of periodic
heterogeneous media,” Eur. J. Mech. - ASolids, vol. 17, no. 4, pp. 599–617, Jul. 1998.
[111] Z. Hashin, “Theory of mechanical behavior of heterogeneous media,”
PENNSYLVANIA UNIV PHILADELPHIA TOWNE SCHOOL OF CIVIL AND MECHANICAL
ENGINEERING, 1963.
[112] R. A. Schapery, “Thermal Expansion Coefficients of Composite Materials Based on
Energy Principles,” J. Compos. Mater., vol. 2, no. 3, pp. 380–404, Jul. 1968.
[113] L. E. Nielsen, “Mechanical Properties of Particulate-Filled Systems,” J. Compos.
Mater., vol. 1, no. 1, pp. 100–119, Jan. 1967.
[114] V. M. Levin, “On the Coefficients of Thermal Expansion of Heterogeneous
Materials,” Mekhanika Tverd. Tela, p. 88, 1967.
[115] R. Hill, “Theory of mechanical properties of fibre-strengthened materials: I. Elastic
behaviour,” J. Mech. Phys. Solids, vol. 12, no. 4, pp. 199–212, Sep. 1964.

231
[116] G. Dvorak, Micromechanics of Composite Materials, vol. 186. Dordrecht: Springer
Netherlands, 2013.
[117] Y. Takao and M. Taya, “Thermal expansion coefficients and thermal stresses in an
aligned short fiber composite with application to a short carbon fiber/aluminum,”
J. Appl. Mech., vol. 52, no. 4, pp. 806–810, 1985.
[118] J. D. Eshelby, “The determination of the elastic field of an ellipsoidal inclusion, and
related problems,” in Proceedings of the Royal Society of London A: Mathematical,
Physical and Engineering Sciences, 1957, vol. 241, pp. 376–396.
[119] D. F. Adams, “Inelastic analysis of a unidirectional composite subjected to
transverse normal loading,” J. Compos. Mater., vol. 4, no. 3, pp. 310–328, 1970.
[120] O. Pierard, C. Friebel, and I. Doghri, “Mean-field homogenization of multi-phase
thermo-elastic composites: a general framework and its validation,” Compos. Sci.
Technol., vol. 64, no. 10, pp. 1587–1603, Aug. 2004.
[121] D. P. Boso, “A simple and effective approach for thermo-mechanical modelling of
composite superconducting wires,” Supercond. Sci. Technol., vol. 26, no. 4, p.
045006, 2013.
[122] Y. Benveniste, G. J. Dvorak, and T. Chen, “On diagonal and elastic symmetry of the
approximate effective stiffness tensor of heterogeneous media,” J. Mech. Phys.
Solids, vol. 39, no. 7, pp. 927–946, Jan. 1991.
[123] D. A. G. Bruggeman, “Dielectric constant and conductivity of mixtures of isotropic
materials,” Ann Phys Leipz., vol. 24, pp. 636–679, 1935.
[124] Y. Benveniste and G. J. Dvorak, “On a correspondence between mechanical and
thermal effects in two-phase composites,” in Micromechanics and inhomogeneity,
Springer, 1990, pp. 65–81.
[125] Y. Benveniste, G. J. Dvorak, and T. Chen, “Stress fields in composites with coated
inclusions,” Mech. Mater., vol. 7, no. 4, pp. 305–317, 1989.
[126] H. Chen et al., “Thermal conductivity of polymer-based composites: Fundamentals
and applications,” Prog. Polym. Sci., vol. 59, no. Supplement C, pp. 41–85, Aug.
2016.
[127] J. C. M. Garnett, “Colours in Metal Glasses, in Metallic Films, and in Metallic
Solutions. II,” Philos. Trans. R. Soc. Lond. Ser. Contain. Pap. Math. Phys. Character,
vol. 205, pp. 237–288, 1906.
[128] H. Fricke, “A Mathematical Treatment of the Electric Conductivity and Capacity of
Disperse Systems I. The Electric Conductivity of a Suspension of Homogeneous
Spheroids,” Phys. Rev., vol. 24, no. 5, pp. 575–587, Nov. 1924.
[129] P. L. Kapitza, “Heat Transfer and Superfluidity of Helium II,” Phys. Rev., vol. 60, no.
4, pp. 354–355, Aug. 1941.
[130] C.-W. Nan, R. Birringer, D. R. Clarke, and H. Gleiter, “Effective thermal conductivity
of particulate composites with interfacial thermal resistance,” J. Appl. Phys., vol.
81, no. 10, pp. 6692–6699, May 1997.
[131] R. L. Hamilton and O. K. Crosser, “Thermal Conductivity of Heterogeneous Two-
Component Systems,” Ind. Eng. Chem. Fundam., vol. 1, no. 3, pp. 187–191, Aug.
1962.

232
[132] H. Hatta and M. Taya, “Effective thermal conductivity of a misoriented short fiber
composite,” J. Appl. Phys., vol. 58, no. 7, pp. 2478–2486, Oct. 1985.
[133] Y. Agari, A. Ueda, and S. Nagai, “Thermal conductivity of a polymer composite,” J.
Appl. Polym. Sci., vol. 49, no. 9, pp. 1625–1634, Sep. 1993.
[134] S. C. Cheng and R. I. Vachon, “A technique for predicting the thermal conductivity
of suspensions, emulsions and porous materials,” Int. J. Heat Mass Transf., vol. 13,
no. 3, pp. 537–546, Mar. 1970.
[135] L. E. Nielsen, “Generalized Equation for the Elastic Moduli of Composite Materials,”
J. Appl. Phys., vol. 41, no. 11, pp. 4626–4627, Oct. 1970.
[136] L. E. Nielsen, “Thermal conductivity of particulate-filled polymers,” J. Appl. Polym.
Sci., vol. 17, no. 12, pp. 3819–3820, Dec. 1973.
[137] T. B. Lewis and L. E. Nielsen, “Dynamic mechanical properties of particulate-filled
composites,” J. Appl. Polym. Sci., vol. 14, no. 6, pp. 1449–1471, Jun. 1970.
[138] D. Kumlutaş, İ. H. Tavman, and M. Turhan Çoban, “Thermal conductivity of particle
filled polyethylene composite materials,” Compos. Sci. Technol., vol. 63, no. 1, pp.
113–117, Jan. 2003.
[139] I. V. Singh, M. Tanaka, and M. Endo, “Effect of interface on the thermal conductivity
of carbon nanotube composites,” Int. J. Therm. Sci., vol. 46, no. 9, pp. 842–847, Sep.
2007.
[140] B. Banerjee and D. O. Adams, “Micromechanics-based determination of effective
elastic properties of polymer bonded explosives,” Phys. B Condens. Matter, vol.
338, no. 1, pp. 8–15, Oct. 2003.
[141] X. Li et al., “Computational modeling and evaluation of the thermal behavior of
randomly distributed single-walled carbon nanotube/polymer composites,”
Comput. Mater. Sci., vol. 63, no. Supplement C, pp. 207–213, Oct. 2012.
[142] I. Özdemir, W. A. M. Brekelmans, and M. G. D. Geers, “Computational
homogenization for heat conduction in heterogeneous solids,” Int. J. Numer.
Methods Eng., vol. 73, no. 2, pp. 185–204, Jan. 2008.
[143] F. Gori, J. F. Ciparisse, and S. Corasaniti, “Thermal anisotropic properties of
composite materials,” in 2012 IEEE Aerospace Conference, 2012, pp. 1–8.
[144] M. Kamiński, “Homogenization technique for transient heat transfer in
unidirectional composites,” Commun. Numer. Methods Eng., vol. 19, no. 7, pp.
503–512, Jul. 2003.
[145] A. El Moumen, T. Kanit, A. Imad, and H. El Minor, “Computational thermal
conductivity in porous materials using homogenization techniques: Numerical and
statistical approaches,” Comput. Mater. Sci., vol. 97, pp. 148–158, Feb. 2015.
[146] E. Andreassen and C. S. Andreasen, “How to determine composite material
properties using numerical homogenization,” Comput. Mater. Sci., vol. 83, pp. 488–
495, Feb. 2014.
[147] X. Liu, K. Rouf, B. Peng, and W. Yu, “Two-step homogenization of textile composites
using mechanics of structure genome,” Compos. Struct., vol. 171, pp. 252–262, Jul.
2017.
[148] A. Yoshimura, A. M. Waas, and Y. Hirano, “Multiscale homogenization for nearly
periodic structures,” Compos. Struct., vol. 153, pp. 345–355, Oct. 2016.

233
[149] C. C. Mei and B. Vernescu, Homogenization methods for multiscale mechanics.
Singapore ; Hackensack, NJ: World Scientific, 2010.
[150] F. Otero, S. Oller, X. Martinez, and O. Salomón, “Numerical homogenization for
composite materials analysis. Comparison with other micro mechanical
formulations,” Compos. Struct., vol. 122, pp. 405–416, Apr. 2015.
[151] J. Lehtonen, R. Mikkonen, and J. Paasi, “A numerical model for stability
considerations in HTS magnets,” Supercond. Sci. Technol., vol. 13, no. 3, p. 251,
2000.
[152] A. Dasgupta and R. K. Agarwal, “Orthotropic Thermal Conductivity of Plain-Weave
Fabric Composites Using a Homogenization Technique,” J. Compos. Mater., vol. 26,
no. 18, pp. 2736–2758, Dec. 1992.
[153] S. Schindler, J. Mergheim, M. Zimmermann, J. C. Aurich, and P. Steinmann,
“Numerical homogenization of elastic and thermal material properties for metal
matrix composites (MMC),” Contin. Mech. Thermodyn., vol. 29, no. 1, pp. 51–75,
Jan. 2017.
[154] B. W. Rosen and Z. Hashin, “Effective thermal expansion coefficients and specific
heats of composite materials,” Int. J. Eng. Sci., vol. 8, no. 2, pp. 157–173, Feb. 1970.
[155] J. Aboudi, S. M. Arnold, and B. A. Bednarcyk, Micromechanics of Composite
Materials: A Generalized Multiscale Analysis Approach, 1 edition. Amsterdam ;
Boston: Butterworth-Heinemann, 2012.
[156] C. Poole, T. Baig, R. J. Deissler, D. Doll, M. Tomsic, and M. Martens, “Numerical
study on the quench propagation in a 1.5 T MgB 2 MRI magnet design with varied
wire compositions,” Supercond. Sci. Technol., vol. 29, no. 4, p. 044003, Apr. 2016.
[157] R. J. Deissler et al., “Numerical simulation of quench protection for a 1.5 T
persistent mode MgB 2 conduction-cooled MRI magnet,” Supercond. Sci. Technol.,
vol. 30, no. 2, p. 025021, 2017.
[158] A. A. Amin et al., “A multiscale and multiphysics model of strain development in a
1.5 T MRI magnet designed with 36 filament composite MgB 2 superconducting
wire,” Supercond. Sci. Technol., vol. 29, no. 5, p. 055008, May 2016.
[159] Y. Iwasa, Case Studies in Superconducting Magnets. Boston, MA: Springer US, 2009.
[160] W. H. Gray and J. K. Ballou, “Electromechanical stress analysis of transversely
isotropic solenoids,” J. Appl. Phys., vol. 48, no. 7, pp. 3100–3109, Jul. 1977.
[161] W. H. Gray and J. E. Akin, “Finite Element Stress Analysis of Orthotropic Solenoids,”
Oak Ridge National Lab., Tenn. (USA); Tennessee Univ., Knoxville (USA), CONF-
771029-61, Jan. 1977.
[162] L. M. Lontai and P. G. Marston, “A 100 kilogauss quasi-continuous cryogenic
solenoid, part I(Stress analysis of 100 kilogauss quasi-continuous cryogenic
solenoid),” 1965., pp. 723–732, 1965.
[163] H. Burkhard, “Magnetic- and stress-fields of cylindrical non-uniform current density
superconducting coils,” Appl. Phys., vol. 6, no. 3, pp. 357–362, Apr. 1975.
[164] N. E. Johnson, W. H. Gray, and R. A. Weed, “Stress analysis of non-homogeneous,
transversely isotropic superconducting solenoids,” 1976.
[165] V. Arp, “Stresses in superconducting solenoids,” J. Appl. Phys., vol. 48, no. 5, pp.
2026–2036, May 1977.

234
[166] F. C. Moon, Magneto-solid Mechanics. University Microfilms, 2005.
[167] J. M. Daniels, “High power solenoids; stresses and stability,” Br. J. Appl. Phys., vol.
4, no. 2, p. 50, Feb. 1953.
[168] Y.-H. Pao, “Electromagnetic forces in deformable continua,” in In: Mechanics today.
Volume 4.(A78-35706 14-70) New York, Pergamon Press, Inc., 1978, p. 209-305.
NSF-supported research., 1978, vol. 4, pp. 209–305.
[169] H. A. Haus and P. Penfield, “Electrodynamics of moving media,” 1967.
[170] J. Caldwell, “Electromagnetic forces in high field magnet coils,” Appl. Math. Model.,
vol. 6, no. 3, pp. 157–160, Jun. 1982.
[171] J. Caldwell, “The stress in the windings of a coil carrying an electric current,” J. Phys.
Appl. Phys., vol. 13, no. 8, p. 1379, 1980.
[172] Z. J. Stekly, “Theoretical and experimental study of an unprotected
superconducting coil going normal,” Advan Cryog Eng, vol. 8, 1963.
[173] W. H. Cherry and J. I. Gittleman, “Thermal and electrodynamic aspects of the
superconductive transition process,” Solid-State Electron., vol. 1, no. 4, pp. 287–
305, Sep. 1960.
[174] J. W. Bremer and V. L. Newhouse, “Thermal Propagation Effect in Thin
Superconducting Films,” Phys. Rev. Lett., vol. 1, no. 8, pp. 282–284, Oct. 1958.
[175] R. F. Broom and E. H. Rhoderick, “Thermal propagation of a normal region in a thin
superconducting film and its application to a new type of bistable element,” Br. J.
Appl. Phys., vol. 11, no. 7, p. 292, 1960.
[176] D. Hagedorn and P. Dullenkopf, “The propagation of the resistive region in high
current density superconducting coils,” Cryogenics, vol. 14, no. 5, pp. 264–270, May
1974.
[177] J. Williams, “Quenching in coupled adiabatic coils,” IEEE Trans. Magn., vol. 21, no.
2, pp. 396–399, Mar. 1985.
[178] Superconducting Magnets. Oxford, New York: Oxford University Press, 1987.
[179] L. Bottura and O. C. Zienkiewicz, “Quench analysis of large superconducting
magnets. Part I: model description,” Cryogenics, vol. 32, no. 7, pp. 659–667, Jan.
1992.
[180] L. Bottura and O. C. Zienkiewicz, “Quench analysis of large superconducting
magnets. Part II: model validation and application,” Cryogenics, vol. 32, no. 8, pp.
719–728, Jan. 1992.
[181] L. Bottura, “A Numerical Model for the Simulation of Quench in the ITER Magnets,”
J. Comput. Phys., vol. 125, no. 1, pp. 26–41, Apr. 1996.
[182] A. Shajii and J. P. Freidberg, “Quench in superconducting magnets. I. Model and
numerical implementation,” J. Appl. Phys., vol. 76, no. 5, pp. 3149–3158, 1994.
[183] U. Ascher, J. Christiansen, and R. D. Russell, “Algorithm 569: COLSYS: Collocation
Software for Boundary-Value ODEs [D2],” ACM Trans Math Softw, vol. 7, no. 2, pp.
223–229, Jun. 1981.
[184] C. A. Luongo, R. J. Loyd, F. K. Chen, and S. D. Peck, “Thermal-hydraulic simulation
of helium expulsion from a cable-in-conduit conductor,” IEEE Trans. Magn., vol. 25,
no. 2, pp. 1589–1595, 1989.

235
[185] Y. M. Eyssa and W. D. Markiewicz, “Quench simulation and thermal diffusion in
epoxy-impregnated magnet system,” IEEE Trans. Appl. Supercond., vol. 5, no. 2, pp.
487–490, Jun. 1995.
[186] C. N. Whetstone and C. E. Roos, “Thermal Phase Transitions in Superconducting Nb-
Zr Alloys,” J. Appl. Phys., vol. 36, no. 3, pp. 783–791, 1965.
[187] J. Paasi, J. Lehtonen, T. Kalliohaka, and R. Mikkonen, “Stability and quench of a HTS
magnet with a hot spot,” Supercond. Sci. Technol., vol. 13, no. 7, p. 949, 2000.
[188] H. Lim and Y. Iwasa, “Two-dimensional normal zone propagation in BSCCO-2223
pancake coils,” Cryogenics, vol. 37, no. 12, pp. 789–799, Dec. 1997.
[189] S. B. Kim, A. Ishiyama, H. Okada, and S. Nomura, “Normal-zone propagation
properties in Bi-2223/Ag superconducting multifilament tapes1,” Cryogenics, vol.
38, no. 8, pp. 823–831, Aug. 1998.
[190] R. Mikkonen, J. Lehtonen, and J. Paasi, “Philosophy of a quench; two case studies-
LTS wiggler magnet and HTS/spl mu/-SMES,” IEEE Trans. Appl. Supercond., vol. 9,
no. 2, pp. 596–599, 1999.
[191] H. Kumakura et al., “Performance tests of Bi-2223 pancake magnet,” Cryogenics,
vol. 38, no. 6, pp. 639–643, Jun. 1998.
[192] T. Kiss et al., “Heat propagation and stability in a small high Tc superconductor coil,”
Phys. C Supercond., vol. 310, no. 1–4, pp. 372–376, Dec. 1998.
[193] J. W. Lue, M. S. Lubell, D. Aized, J. M. Campbell, and R. E. Schwall, “Quenches in a
high-temperature superconducting tape and pancake coil,” Cryogenics, vol. 36, no.
5, pp. 379–389, May 1996.
[194] J. Leveque, D. Netter, J. P. Caron, C. Martinez, and A. Rezzoug, “Influence of the
dispersion of physical parameters on the thermal stability of superconducting
coils,” Supercond. Sci. Technol., vol. 11, no. 1, p. 149, 1998.
[195] M. Lindmayer and H. Mosebach, “Quenching of high-T/sub c/-superconductors and
current limitation numerical simulations and experiments,” IEEE Trans. Appl.
Supercond., vol. 7, no. 2, pp. 1029–1032, 1997.
[196] J. Lehtonen, R. Mikkonen, and J. Paasi, “Stability considerations of a high-
temperature superconductor tape at different operating temperatures,” Phys. C
Supercond., vol. 310, no. 1, pp. 340–344, 1998.
[197] E. E. Burkhardt and J. Schwartsz, “Three-dimensional numerical analysis of the
stability of Ag/Bi/sub 2/Sr/sub 2/Ca/sub 2/Cu/sub 3/O/sub x/tape conductors,”
IEEE Trans. Appl. Supercond., vol. 7, no. 2, pp. 199–202, 1997.
[198] S. B. Kim and A. Ishiyama, “Transient stability analysis in Bi-2223/Ag
superconducting tapes,” IEEE Trans. Appl. Supercond., vol. 7, no. 2, pp. 203–206,
1997.
[199] M. Wetzko, M. Zahn, and H. Reiss, “Current sharing and stability in a Bi-2223/Ag
high temperature superconductor,” Cryogenics, vol. 35, no. 6, pp. 375–386, 1995.
[200] H. Hashizume and A. Toda, “Electromagnetic thermal analysis of low/high Tc
superconducting wire,” IEEE Trans. Magn., vol. 34, no. 5, pp. 3016–3019, Sep. 1998.
[201] A. Stenvall, A. Korpela, R. Mikkonen, and G. Grasso, “Stability considerations of
multifilamentary MgB 2 tape,” Supercond. Sci. Technol., vol. 19, no. 2, p. 184, 2006.

236
[202] M. Fu et al., “Quench characteristics and normal zone propagation of an MgB 2
superconducting coil,” Supercond. Sci. Technol., vol. 17, no. 1, p. 160, 2004.
[203] U. Gambardella et al., “Stability Measurements in Multifilamentary MgB2 Tapes,”
IEEE Trans. Appl. Supercond., vol. 17, no. 2, pp. 2937–2940, Jun. 2007.
[204] J. Pelegrín et al., “Experimental and numerical analysis of quench propagation on
MgB 2 tapes and pancake coils,” Supercond. Sci. Technol., vol. 26, no. 4, p. 045002,
2013.
[205] L. Ye, D. Cruciani, M. Xu, S. Mine, K. Amm, and J. Schwartz, “Magnetic field
dependent stability and quench behavior and degradation limits in conduction-
cooled MgB 2 wires and coils,” Supercond. Sci. Technol., vol. 28, no. 3, p. 035015,
2015.
[206] E. Martínez, E. A. Young, M. Bianchetti, O. Muñoz, S. I. Schlachter, and Y. Yang,
“Quench onset and propagation in Cu-stabilized multifilament MgB 2 conductors,”
Supercond. Sci. Technol., vol. 21, no. 2, p. 025009, 2008.
[207] X. Wang, S. V. P. S. S. Pamidi, U. P. Trociewitz, and J. Schwartz, “Self-field quench
behavior of multifilamentary wires in liquid helium,” Cryogenics, vol. 48, no. 11–12,
pp. 469–477, Nov. 2008.
[208] M. Fu et al., “Minimum quench energy and normal zone propagation velocity in
MgB 2 superconducting tape,” Phys. C Supercond., vol. 402, no. 3, pp. 234–238,
2004.
[209] H. Van Weeren et al., “Adiabatic normal zone development in MgB/sub
2/Superconductors,” IEEE Trans. Appl. Supercond., vol. 15, no. 2, pp. 1667–1670,
2005.
[210] V. Cavaliere, M. Cioffi, A. Formisano, R. Martone, and G. Masullo, “Three-
dimensional quench propagation in HTS coils,” IEEE Trans. Magn., vol. 42, no. 4, pp.
975–978, Apr. 2006.
[211] A. Stenvall, A. Korpela, R. Mikkonen, and G. Grasso, “Quench analysis of MgB 2 coils
with a ferromagnetic matrix,” Supercond. Sci. Technol., vol. 19, no. 6, p. 581, 2006.
[212] M. Alessandrini, G. Majkic, E. T. Laskaris, and K. Salama, “Modeling of Longitudinal
and Transverse Quench Propagation in Stacks of Superconducting $hbox MgB_2$
Wire,” IEEE Trans. Appl. Supercond., vol. 19, no. 3, pp. 2437–2441, Jun. 2009.
[213] X. Du et al., “Numerical Analysis on the Quench Process and Protection of
Conduction Cooled $${$$\backslash$ rm MgB$}$ _ ${$2$}$ $ Magnet,” IEEE Trans.
Appl. Supercond., vol. 20, no. 3, pp. 2102–2106, 2010.
[214] D. Zhang et al., “Research on Stability of $${$$\backslash$ rm MgB$}$ _ ${$2$}$ $
Superconducting Magnet for MRI,” IEEE Trans. Appl. Supercond., vol. 21, no. 3, pp.
2100–2103, 2011.
[215] A. Stenvall, A. Korpela, J. Lehtonen, and R. Mikkonen, “Formulation for solving 1D
minimum propagation zones in superconductors,” Phys. C Supercond. Its Appl., vol.
468, no. 13, pp. 968–973, Jul. 2008.
[216] A. Stenvall, R. Mikkonen, and P. Kovac, “Relation Between Transverse and
Longitudinal Normal Zone Propagation Velocities in Impregnated $${$$\backslash$
rm MgB$}$ _ ${$2$}$ $ Windings,” IEEE Trans. Appl. Supercond., vol. 19, no. 3, pp.
2403–2406, 2009.

237
[217] R. G. Jones, E. H. Rhoderick, and A. C. Rose-Innes, “Non-linearity in the voltage-
current characteristic of a type-2 superconductor,” Phys. Lett. A, vol. 24, no. 6, pp.
318–319, Mar. 1967.
[218] H. van Weeren, “Magnesium diboride superconductors for magnet applications,”
info:eu-repo/semantics/doctoralThesis, University of Twente, Enschede, 2007.
[219] E. Martínez et al., “Quench development and propagation in metal/MgB2
conductors,” Supercond. Sci. Technol., vol. 19, no. 1, p. 143, 2005.
[220] M. Ristic, J. V. M. McGinley, and F. Lorenzoni, “Numerical Study of Quench
Protection Schemes for a Superconducting Magnet,” IEEE Trans. Appl. Supercond.,
vol. 21, no. 5, pp. 3501–3508, Oct. 2011.
[221] T. A. Prikhna, “Properties of MgB2 bulk,” ArXiv Prepr. ArXiv09124906, 2009.
[222] M. Sugano, A. Ballarino, B. Bartova, R. Bjoerstad, A. Gerardin, and Christian
Scheuerlein, “Evaluation of Young’s modulus of MgB 2 filaments in composite wires
for the superconducting links for the high-luminosity LHC upgrade,” Supercond. Sci.
Technol., vol. 29, no. 2, p. 025009, 2016.
[223] K. J. Carroll, “Elastic Constants of Niobium from 4.2° to 300°K,” J. Appl. Phys., vol.
36, no. 11, pp. 3689–3690, Nov. 1965.
[224] J. E. Jensen, R. G. Stewart, W. A. Tuttle, and H. Brechna, Brookhaven National
Laboratory Selected Cryogenic Data Notebook: Sections X-XVIII, vol. 2. Brookhaven
National Laboratory, 1980.
[225] http://www.copper.org/resources/properties/cryogenic/, .
[226] Composite Technology Developement, “CTD 101 k Material properties.” [Online].
Available: http://www.ctd-materials.com/wordpress/wp-
content/uploads/2014/05/CTD-101G-DS-2006.pdf.
[227] https://en.wikipedia.org/wiki/Niobium, .
[228] http://www.specialmetals.com/assets/documents/alloys/monel/monel-alloy-
400.pdf, .
[229] “Glidcop copper dispersion strengthened with aluminum oxize.” Argonne National
laboratory.
[230] W. KÖSTER and H. FRANZ, “Poisson’s Ratio for Metals and Alloys,” Metall. Rev., vol.
6, no. 1, pp. 1–56, Jan. 1961.
[231] M. de Jong et al., “Charting the complete elastic properties of inorganic crystalline
compounds,” Sci. Data, vol. 2, p. 150009, Mar. 2015.
[232] K. M. Liew, X. Q. He, and C. H. Wong, “On the study of elastic and plastic properties
of multi-walled carbon nanotubes under axial tension using molecular dynamics
simulation,” Acta Mater., vol. 52, no. 9, pp. 2521–2527, May 2004.
[233] M. R. Eslami, R. B. Hetnarski, J. Ignaczak, N. Noda, N. Sumi, and Y. Tanigawa, Theory
of Elasticity and Thermal Stresses, vol. 197. Dordrecht: Springer Netherlands, 2013.
[234] N. J. Simon, E. S. Drexler, and R. P. Reed, “NIST Monograph 177 Properties of copper
and copper alloys at cryogenic temperature,” US Gov. Print. Off., pp. 2–27, 1992.
[235] P. E. Fabian, T. S. Bauer-McDaniel, and R. P. Reed, “Low temperature thermal
properties of composite insulation systems,” Cryogenics, vol. 35, no. 11, pp. 719–
722, Nov. 1995.

238
[236] M. Schneider et al., “Heat and charge transport properties of MgB2,” Phys. C
Supercond., vol. 363, no. 1, pp. 6–12, Oct. 2001.
[237] R. M. Swift and D. White, “Low temperature heat capacities of magnesium diboride
(MgB2) and magnesium tetraboride (MgB4),” J. Am. Chem. Soc., vol. 79, no. 14, pp.
3641–3644, 1957.
[238] Y. Wang, T. Plackowski, and A. Junod, “Specific heat in the superconducting and
normal state (2–300 K, 0–16 T), and magnetic susceptibility of the 38 K
superconductor MgB2: evidence for a multicomponent gap,” Phys. C Supercond.,
vol. 355, no. 3–4, pp. 179–193, Jun. 2001.
[239] G. K. White and S. J. Collocott, “Heat Capacity of Reference Materials: Cu and W,”
J. Phys. Chem. Ref. Data, vol. 13, no. 4, pp. 1251–1257, Oct. 1984.
[240] R. A. Matula, “Electrical resistivity of copper, gold, palladium, and silver,” J. Phys.
Chem. Ref. Data, vol. 8, no. 4, pp. 1147–1298, Oct. 1979.
[241] G. W. Webb, “Low-Temperature Electrical Resistivity of Pure Niobium,” Phys. Rev.,
vol. 181, no. 3, pp. 1127–1135, May 1969.
[242] J. M. Rowell, “The widely variable resistivity of MgB 2 samples,” Supercond. Sci.
Technol., vol. 16, no. 6, p. R17, 2003.
[243] http://www.specialmetals.com/assets/documents/alloys/monel/monel-alloy-
400.pdf, .
[244] M. J. White et al., “Design, Operation, and Safety of Single-Room Interventional
MRI Suites: Practical Experience From Two Centers,” J. Magn. Reson. Imaging, vol.
41, no. 1, pp. 34–43, Jan. 2015.
[245] H. Yin and Y. Zhao, Introduction to the Micromechanics of Composite Materials. CRC
Press, 2016.
[246] W. J. Drugan and J. R. Willis, “A micromechanics-based nonlocal constitutive
equation and estimates of representative volume element size for elastic
composites,” J. Mech. Phys. Solids, vol. 44, no. 4, pp. 497–524, 1996.
[247] B. Collins, “Investigation of Nb3Sn Based Superconductors Through Hierarchical
Models,” 2013.
[248] R. Luciano and E. Sacco, “Variational methods for the homogenization of periodic
heterogeneous media,” Eur. J. Mech. - ASolids, vol. 17, no. 4, pp. 599–617, Jul. 1998.
[249] E. J. Barbero, Finite Element Analysis of Composite Materials Using ANSYS®, Second
Edition. CRC Press, 2013.
[250] T. Kanit, S. Forest, I. Galliet, V. Mounoury, and D. Jeulin, “Determination of the size
of the representative volume element for random composites: statistical and
numerical approach,” Int. J. Solids Struct., vol. 40, no. 13–14, pp. 3647–3679, Jun.
2003.
[251] J. Lehtonen, R. Mikkonen, and J. Paasi, “Effective thermal conductivity in HTS coils,”
Cryogenics, vol. 40, no. 4–5, pp. 245–249, Apr. 2000.
[252] W. Song, V. Krishnaswamy, and R. V. Pucha, “Computational homogenization in
RVE models with material periodic conditions for CNT polymer composites,”
Compos. Struct., vol. 137, pp. 9–17, Mar. 2016.
[253] J. J. Espadas-Escalante, N. P. van Dijk, and P. Isaksson, “A study on the influence of
boundary conditions in computational homogenization of periodic structures with

239
 application to woven composites,” Compos. Struct., vol. 160, pp. 529–537, Jan.
2017.
[254] D. J. Lloyd, “Particle reinforced aluminium and magnesium matrix composites,” Int.
Mater. Rev., vol. 39, no. 1, pp. 1–23, 1994.
[255] G. Lenoir and V. Aubin, “Mechanical Characterization and Modeling of a Powder-
In-Tube MgB 2 Strand,” IEEE Trans. Appl. Supercond., vol. 27, no. 4, pp. 1–5, Jun.
2017.
[256] C. S. Myers, M. A. Susner, L. Motowidlo, J. Distin, M. D. Sumption, and E. W.
Collings, “Specific Heats of Composite Bi2212, $\hbox{Nb}_{3} \hbox{Sn}$, and
$\hbox{MgB}_{2}$ Wire Conductors,” IEEE Trans. Appl. Supercond., vol. 23, no. 3,
pp. 8800204–8800204, Jun. 2013.
[257] W. Wu et al., “Design of a 7 T Superconducting Magnet for Lanzhou Penning Trap,”
IEEE Trans. Appl. Supercond., vol. 20, no. 3, pp. 989–992, Jun. 2010.
[258] M. Guan, X. Wang, C. Xin, W. Wu, and L. Ma, “Structural Mechanics Exploration for
Multicomponent Superconducting Solenoids by Hoop Strain Tests During Cooling
and Excitation,” J. Supercond. Nov. Magn., vol. 27, no. 5, pp. 1179–1185, Nov. 2013.
[259] J. J. Neumeier et al., “Negative thermal expansion of Mg B 2 in the superconducting
state and anomalous behavior of the bulk Grüneisen function,” Phys. Rev. B, vol.
72, no. 22, p. 220505, 2005.
[260] G. K. White, “Thermal expansion of vanadium, niobium, and tantalum at low
temperatures,” Cryogenics, vol. 2, no. 5, pp. 292–296, Sep. 1962.
[261] E. S. Bobrov, J. E. C. Williams, and Y. Iwasa, “Experimental and theoretical
investigation of mechanical disturbances in epoxy-impregnated superconducting
coils. 2. Shear-stress-induced epoxy fracture as the principal source of premature
quenches and training theoretical analysis,” Cryogenics, vol. 25, no. 6, pp. 307–316,
1985.
[262] G. Pasztor and C. Schmidt, “Dynamic stress effects in technical superconductors
and the ’’training’’ problem of superconducting magnets,” J. Appl. Phys., vol. 49,
no. 2, pp. 886–899, Feb. 1978.
[263] A. C. Orifici, I. Herszberg, and R. S. Thomson, “Review of methodologies for
composite material modelling incorporating failure,” Compos. Struct., vol. 86, no.
1–3, pp. 194–210, Nov. 2008.
[264] M. R. Vaghar, H. Garmestani, and W. D. Markiewicz, “Elastoplastic stress analysis
of Nb3Sn superconducting magnet,” J. Appl. Phys., vol. 80, no. 4, p. 2490, 1996.
[265] P. Kovac et al., “Tensile and Bending Strain Tolerance of Ex Situ MgB2/Ni/Cu
Superconductor Tape,” IEEE Trans. Appl. Supercond., vol. 25, no. 2, pp. 1–7, Apr.
2015.
[266] C. Zhou et al., “Intrawire resistance, AC loss and strain dependence of critical
current in MgB2 wires with and without cold high-pressure densification,”
Supercond. Sci. Technol., vol. 27, no. 7, p. 075002, Jul. 2014.
[267] R. De Groot, M. C. Peters, Y. M. De Haan, G. J. Dop, and A. J. Plasschaert, “Failure
stress criteria for composite resin,” J. Dent. Res., vol. 66, no. 12, pp. 1748–1752,
Dec. 1987.

240
[268] P. Kovác et al., “Dependence of the critical current in ex situ multi- and mono-
filamentary MgB2 /Fe wires on axial tension and compression,” Supercond. Sci.
Technol., vol. 16, no. 5, p. 600, 2003.
[269] N. A. Munshi and H. W. Weber, “Reactor Neutron and Gamma Irradiation of
Various Composite Materials,” in Materials, F. R. Fickett and R. P. Reed, Eds.
Springer US, 1992, pp. 233–239.
[270] Y. Iwasa, “Mechanical disturbances in superconducting magnets-a review,” IEEE
Trans. Magn., vol. 28, no. 1, pp. 113–120, Jan. 1992.
[271] A. A. Amin et al., “Mechanical Analysis of MgB2 Based Full Body MRI Coils Under
Different Winding Conditions,” IEEE Trans. Appl. Supercond., vol. 27, no. 4, pp. 1–
5, 2017.
[272] A. R. Boccaccini, G. West, J. Janczak, M. H. Lewis, and H. Kern, “Tensile behavior
and cyclic creep of continuous fiber-reinforced glass matrix composites at room
and elevated temperatures,” J. Mater. Eng. Perform., vol. 6, no. 3, pp. 344–348,
1997.
[273] N. E. Dowling, Mechanical Behavior of Materials, 4 edition. Boston: Pearson, 2012.
[274] W. K. Goertzen and M. R. Kessler, “Creep behavior of carbon fiber/epoxy matrix
composites,” Mater. Sci. Eng. A, vol. 421, no. 1, pp. 217–225, Apr. 2006.
[275] A. Reuss, “Berechnung der flie\s sgrenze von mischkristallen auf grund der
plastizitätsbedingung für einkristalle.,” ZAMM-J. Appl. Math. Mech. Für Angew.
Math. Mech., vol. 9, no. 1, pp. 49–58, 1929.

241